Underground

Ventilation
EDITED BY Purushotham Tukkaraja
UNDERGROUND VENTILATION

Underground Ventilation contains the proceedings of the 19th North American Mine Ventilation
Symposium held at the South Dakota School of Mines & Technology (South Dakota Mines) in Rapid
City, South Dakota, June 17-22, 2023. South Dakota Mines organized this symposium in collaboration
with the Underground Ventilation Committee (UVC) of the Society for Mining, Metallurgy & Exploration
(SME).

The Mine Ventilation Symposium series has always been a premier forum for ventilation experts,
practitioners, educators, students, regulators, and suppliers from around the world to exchange
knowledge, ideas, and opinions. Underground Ventilation features sixty-seven selected technical
papers in a wide range of ventilation topics including: auxiliary and primary systems, mine fans,
case studies, computational fluid dynamics applications, diesel particulate control, electric machinery,
mine cooling and refrigeration, mine dust monitoring and control, mine fires and explosion prevention,
mine gases, mine heat, mine ventilation and automation, occupational health and safety, renewable/
alternative energy, monitoring and measurement, network analysis and optimization, and planning and
design.
PROCEEDINGS OF THE 19TH NORTH AMERICAN MINE VENTILATION SYMPOSIUM
(NAMVS 2023), 17-22 JUNE 2023, RAPID CITY, SOUTH DAKOTA, USA

Underground Ventilation

Edited by
Purushotham Tukkaraja, Ph.D., QP
Mining Engineering & Management, South Dakota Mines, Rapid City, SD, USA
Front Cover Image: © Zitrón - Ventilation solutions for underground mines

First published 2023

by CRC Press/Balkema
4 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

and by CRC Press/Balkema

2385 NW Executive Center Drive, Suite 320, Boca Raton FL 33431

CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business

© 2023 selection and editorial matter, Purushotham Tukkaraja; individual chapters, the
contributors

Typeset by Integra Software Services Pvt. Ltd., Pondicherry, India

The right of Purushotham Tukkaraja to be identified as the author of the editorial material,
and of the authors for their individual chapters, has been asserted in accordance with sections
77 and 78 of the Copyright, Designs and Patents Act 1988.

All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form
or by any electronic, mechanical, or other means, now known or hereafter invented, including
photocopying and recording, or in any information storage or retrieval system, without
permission in writing from the publishers.

Although all care is taken to ensure integrity and the quality of this publication and the
information herein, no responsibility is assumed by the publishers nor the author for any
damage to the property or persons as a result of operation or use of this publication and/or
the information contained herein.

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

Library of Congress Cataloging-in-Publication Data

A catalog record has been requested for this book

ISBN: 978-1-032-55146-3 (hbk)

ISBN: 978-1-032-55147-0 (pbk)
ISBN: 978-1-003-42924-1 (ebk)
DOI: 10.1201/9781003429241
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Editor(s), ISBN 978-1-032-55146-3

Table of contents

Preface xi
Organizing committees xiii
Underground Ventilation Committee (UVC) xiii
Review committee xiii
Sponsors and Exhibitors xv

Auxiliary ventilation
An investigation of booster fan placements in a large opening underground stone mine
utilizing CFD 3
N. Gendrue, S. Liu & S. Bhattacharya
Quantifying assemblage losses in auxiliary ventilation systems 15
E. De Souza

Case studies of mine ventilation

Kucing-Liar (KL) mine LOM preliminary ventilation design – PT Freeport Indonesia 25
A.A. Habibi, I. Setiawan, R. Prasojo & C. Stewart
A case study in successful use of spot cooling for underground shaft sinking 34
M. Brown, D.W. Durieux, C. McGuire & D. Witow

Computational fluid dynamics applications in mine ventilation

CFD study of cavern ventilation 45
M.A. Carvajal-Meza & J.P. Hurtado-Cruz
Model order reduction of high-fidelity underground mine model 54
M. Vaze, J. Nyqvist & S. Dasgupta
Minimizing the aerodynamic impact of a new cooling plant installed upstream of an existing
surface fan and heater arrangement through CFD analyses 60
J.K. Shaw, L.K. Falk, C. Allen & M. Kaufman
CFD modeling of a large-opening stone mine using COMSOL multiphysics 69
K.V. Raj & V. Gangrade
An investigation of shock loss factors at ventilation raise junctions in underground hard rock
mines using computational fluid dynamics 83
H. Zhang, L.K. Falk & C. Allen

v
Diesel particulate control
Comparing diesel and GDiesel® exhaust exposures in an underground mining laboratory 95
R.J. Reed, J.L. Burgess & E.A. Lutz
Improvement of size-selective sampling of diesel aerosols in underground mines 104
A.D. Bugarski, T.L. Barone, J.A. Hummer, T. Lee, S. Vanderslice & S. Friend
Results of diesel exhaust nanoparticle experimental sampling in a cabin of LHD loader
operating in an active ore heading area 115
S. Sabanov, N. Magauiya, A. Zeinulla, A. Abil, A. Qureshi, M. Torkmahalleh,
G. Nurshaiykova & D. Rakhimov
Importance of using real-time and microscopic analysis techniques to characterize DPM in
underground mines 120
A.A. Habibi, K.O. Homan & A.D. Bugarski
Development and evaluation of innovative diesel particulate filter technology 129
J. Stachulak, B. Rubeli, D. Young, K. Watson & B. McLean
DPM reduction through emission assisted maintenance - PT Freeport Indonesia program
update 135
A.A. Habibi, E. Pinto, M. Mardon, K. Wijayanto & C. Rose

Electric machinery in mine ventilation

Underground mine ventilation design: Diesel vs. electrical equipment 149
C.A. Rawlins

Mine cooling and refrigeration

Selection, design challenges and construction of Vale’s Coleman mine 10 MW surface
refrigeration plant 163
S.G. Hardcastle, J.K. Shaw & C. Allen
Mine refrigeration using geothermal energy – is this a viable decarbonization strategy? 174
D.W. Durieux
Onaping Depth Project underground refrigeration plant update 183
D.W. Durieux, C. McGuire, T. Mehedi, D. Witow & E. Pilkington
A direct equation for sigma heat and wet-bulb temperature for underground ventilation
applications 192
A. Pandey, S. Jayaraman Sridharan & B.S. Sastry
Case study: Refrigeration requirements during mineshaft excavation as a function of heat
stress index 200
K. Tom
Development of energy efficient and sustainable cooling strategies for hot underground mines 208
J.E. Fox

Mine dust monitoring and control

Effect of auxiliary scrubbers on respirable coal mine dust particle size and composition 221
F. Animah, A. Greth, C. Keles & E. Sarver
Recovery of respirable dust from fibrous filters for particle analysis by scanning electron
microscopy 230
A. Greth, S. Afrouz, F. Animah, C. Keles & E. Sarver

vi
Effects of vertical air-blocking ring of drill shroud on dust control for surface mine drilling
operation using CFD 239
Y. Zheng, J.D. Potts & W.R. Reed
Investigation on the effect of water pressure on spray performance for removal of respirable dust 247
H. Jiang, S. Klima, T. Beck & Y. Zheng
Real-time measurements of respirable crystalline silica, kaolinite, coal, and calcite 256
W. Arnott, C. Kocsis, X. Wang, B. Osho, P. Nascimento, S. Taylor, B. Bingham, C. Murphy &
M. Sandink
Comparing respirable dust characteristics from full-scale cutting tests of three rock samples
with conical picks at three stages of wear 264
S. Slouka, E. Sidrow, C. Tsai & J. Brune
Accuracy of low-cost particulate matter sensor in measuring coal mine dust- a wind tunnel
evaluation 274
M.M. Zaid, G. Xu & N.A. Amoah
Parametric studies to maximize the dust protection performance of the two-level manifold
canopy air curtain and computational fluid dynamics modeling 285
N.A. Amoah, G. Xu & A.R. Kumar
Respirable coal mine dust research: Characterization and toxicity analysis based on dust
sources 296
V.P. Salinas, M.C. Das, G. Rubasinghege, P. Roghanchi & K. Zychowski
Comparison of respirable coal and silica dust monitoring systems for underground mining
applications 305
A. Medina, A. Vanegas, E. Madureira, P. Roghanchi, R. Rajapaksha, L. Uecker, T. Rawson &
C. Harb
Evaluation of different surfactants’ performance in varying coal dust concentration through
logistic regression analysis 313
Z. Zhao, A. Ghosh, P. Chang & Y. Liu
Process ventilation solutions for mitigation of combustible and non-combustible dust
hazards at mining operations 322
J. Finn
Development of VR-CFD-based training tool for dust control in gateroad development 330
M. Qiao, T. Ren, J. Roberts, J. Hines, C. Chow & A. Clayton

Mine fans
Stall impact on axial fans and testing of anti-stall rings 341
J. Fernandez
Practical values for the evaluation of fan system efficiencies 353
J. Bowling, G. Schult & J. Van Diest

Mine fires and explosion prevention

Longwall maingate and tailgate proactive sponcom and gas management strategy-
An operational safety share on risk management 363
B. Belle & R. Balusu
Real-time methane prediction with small dataset in underground longwall coal mining using AI 377
D.C. Demirkan, H.S. Duzgun, A. Juganda, J. Brune & G. Bogin

vii
Evaluation of different suppression techniques for lithium-ion battery fires 384
L. Yuan, W. Tang, R.A. Thomas & J. Soles
Characterization and preliminary assessment of diesel fire prior to setting up large size
battery fire experiment 393
R.I. Pushparaj, G. Xu, A. Iqbal & O.B. Salami
Fire-induced temperature attenuation under the influence of a single ceiling smoke extraction
point in a bifurcated drift 399
O.B. Salami, G. Xu, A.R. Kumar, R.I. Pushparaj & A. Iqbal
Application of pressure balancing techniques at the West Elk coal mine 411
C. Kocsis, F. Calizaya, J. Johnson, T. Dias, N. Nunes, E. Lindgren, G. Atchley & J. Poulos

Mine gases
Underground coal methane gas forecasting using multivariate time series with one and two
auxiliary variables 423
J.C. Diaz, Z. Agioutantis, S. Schafrik & D.T. Hristopulos
Complete degasification of longwall panels in U.S. coal mines 431
P.C. Thakur
Study of in-situ coal seam gas content for Australian coal and gas outburst management:
A field data analysis and laboratory experiment 439
Z.B. Li, T. Ren, M. Qiao, D. Black & J. Juric
Airflow patterns and blast fume dispersion in different mining methods 451
S. Jayaraman Sridharan, A. Adhikari, P. Tukkaraja & J. Connot
The oxiperator for Ventilation Air Methane (VAM) 462
E. Prabhu, M. Prabhu & J.E. Fox

Mine heat
Findings and learnings from thermal parameter studies at four LKAB sites 469
F.K.R. Klose, A.L. Martikainen & T.H. Jones
Scenario-driven evaluation of heat sources in underground production scheduling 484
J.A. Buaba, E. Udofia, A. Newman & A.J. Brickey

Mine ventilation and automation

Permeability determination for potential interaction between shale gas wells and
the coal mine environment due to longwall-induced deformations under deep
cover 499
M.L. Harris, S.J. Schatzel, K.M. Ajayi, M. Van Dyke, P. Zhang, V. Gangrade, J.D. Addis,
H. Dougherty & E. Watkins
Analysis of variation in longwall-induced permeability under different mining depths 507
K.M. Ajayi, Z. Khademian, S.J. Schatzel, M.L. Harris & J.D. Addis
New applications of jet fans in underground mines for haulage ramps and block cave
ventilation control 515
C. Stewart

viii
Occupational health and safety in mine ventilation
Ventilation research findings for enhanced worker safety when mining near unconventional
gas wells in longwall abutment pillars 529
S.J. Schatzel, K.M. Ajayi, Z. Khademian, R. Kimutis, M.L. Harris, M. Van Dyke, J.D. Addis,
H. Dougherty & E. Watkins
Optimization for fire evacuation applying maximum flow cost algorithm to improve the
time-response in underground coal mines 541
S. Lotero, H. Khaniani, V. Androulakis, M. Hassanalian, S. Shao & P. Roghanchi
Heat mitigation for underground coal mine refuge alternatives 550
D.S. Yantek & L. Yan
Occupancy derating for underground coal mine refuge alternatives 559
D.S. Yantek & L. Yan

Renewable / alternative energy in mine ventilation

Incorporating droplet dynamics to improve the reduced-order model of spray freezing for
mine heating applications 569
M. Mohit, S. Akhtar, M. Xu & A.P. Sasmito
Analysis of small-scale lithium-ion batteries under thermal abuse 578
A. Iqbal, G. Xu, R.I. Pushparaj & O.B. Salami

Ventilation monitoring and measurement

Gauge and tube surveys: What is their future and that of underground measurements generally
as mines transition towards greater use of Big Data and Artificial Intelligence systems? 589
D.J. Brake
Improving the accuracy of field airflow measurements for tunnel ventilation fans 595
R.E. Ray & E. Fuster

Ventilation network analysis and optimization

Case study on the abnormal airflow diagnosis method using atmospheric monitoring data 607
L. Zhou, D. Bahrami & R.A. Thomas
Ventilation system upgrades at the Waste Isolation Pilot Plant 615
K.G. Wallace & I. Peña
Ventilation network optimization: Realizing energy savings while promoting worker health
and safety 624
A.K. Ngcibi, M. Mochubele & F.S. Bergh

Ventilation planning and design

Design highlights for Agnico Eagle’s Macassa 4 Shaft primary ventilation systems 635
K. Boyd, T. Mehedi, D. Witow & M. Pinheiro-Harvey
Upper Keel mine ventilation strategy at Eagle mine 642
K. Boyd, D. Witow, C. McGuire & C. Gobbs
Empirical and numerical investigation on the optimal length of eddy airflow in dead-end tunnel 649
R. Morla, J. Chen, S. Karekal, A. Godbole, P. Tukkaraja & P. Chang
Author index 559

ix
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Editor(s), ISBN 978-1-032-55146-3

Preface

This volume contains the proceedings of the 19th North American Mine Ventilation Sympo­
sium held at the South Dakota School of Mines & Technology (South Dakota Mines) in
Rapid City, South Dakota, June 17-22, 2023. South Dakota Mines organized this symposium
in collaboration with the Underground Ventilation Committee (UVC) of the Society for
Mining, Metallurgy & Exploration (SME).

The North American Mine Ventilation Symposium series was initiated by the UVC in 1982. The
UVC is a Joint Technical Committee of the Coal and Energy and the Mining and Exploration
Divisions of the SME. The purpose of the UVC is to promote engineering interest and techno­
logical progress in the ventilation of mines, tunnels, and other subsurface openings. The UVC
accomplishes its purpose by conducting technical sessions at SME-AIME meetings, sponsoring
the North American Mine Ventilation Symposium with host universities and other organizations,
and soliciting papers for publication in Mining, Metallurgy & Exploration Journal and SME con­
ference proceedings. The UVC offers an affiliation home for SME members, and others engaged
in the practice of underground ventilation. In these ways, the UVC seeks to encourage research,
education, publications, and technology transfer in the field of underground ventilation.

The North American Mine Ventilation Symposium, held every two to three years since 1982,
provides a forum for practitioners, educators, and researchers to exchange the latest informa­
tion on the ventilation of mines, tunnels, and other underground facilities.

Past Symposia were held and organized by…

1982 – University of Alabama – Howard L. Hartman

1985 – University of Nevada, Reno – Pierre Mousset-Jones

1987 – Pennsylvania State University – Jan Mutmansky

1989 – University of California, Berkeley – Malcolm J. McPherson

1991 – West Virginia University – W. J. Wang

1993 – University of Utah – Ragula Bhaskar

1995 – University of Kentucky – Andrzej Wala

1999 – University of Missouri-Rolla – Jerry Tien

2002 – Queen’s University, Canada – Euler De Souza

xi
2004 – University of Alaska Fairbanks – Sukumar Bandopadhyay and Rajive Ganguli

2006 – Pennsylvania State University – Jan Mutmansky

2008 – University of Nevada, Reno – Pierre Mousset-Jones

2010 – Laurentian University – Stephen Hardcastle and Dale McKinnon

2012 – University of Utah – Felipe Calizaya

2015 – Virginia Tech University – Kray Luxbacher and Emily Sarver

2017 – Colorado School of Mines – Jurgen F. Brune

2019 – McGill University and University of British Columbia, Canada

– Ali Madiseh, Agus Sasmito, Ferri Hassani, and Jozef Stachulak

2021 – South Dakota School of Mines & Technology – Purushotham Tukkaraja

With the help of the organizing committee, a solid 3-day program was assembled, with tech­
nical papers, panel discussions, and keynote presentations organized in 21 sessions. A total
of 82 abstracts and 67 final papers were received. Session ventilation themes include case
studies, computational fluid dynamics applications, diesel particulate control, electric
machinery, mine cooling and refrigeration, mine dust monitoring and control, primary and
auxiliary systems, mine fans, mine fires and explosion prevention, mine gases, mine heat,
ventilation management, automation, occupational health and safety, renewable/alternative
energy, monitoring and measurement, network analysis and optimization, and planning and
design.

I would like to thank the UVC and Review Committee members for their help with peer-
reviewing papers, chairing technical sessions, and advice to make this symposium a success.

Finally, I would like to thank the Center for Alumni Relations and Advancement (CARA)
and the Office of Marketing and Communications at South Dakota Mines for assisting with
the symposium registration, advertisement, and website services.

Purushotham Tukkaraja, Ph.D., QP

Symposium Chair

xii
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Editor(s), ISBN 978-1-032-55146-3

Organizing committees

Underground Ventilation Committee (UVC)

Arash Habibi, Freeport-McMoRan, USA David Brokering, Freeport-McMoRan, USA

Arun Rai, Compass Minerals, USA Jack Trackemas, CDC-NIOSH, USA
Bharath Belle, Anglo American, Australia John Bowling, SRK Consulting, USA
Craig Stewart, Minware, Australia Purushotham Tukkaraja, South Dakota Mines, USA

Review committee

Adrianus Halim Euler De Souza Pramod Thakur

Agus Sasmito Felipe Calizaya Ramakrishna Morla
Aleksandar Bugarski Frank von Glehn Rao Balusu
Alex Hatt Gerrit Goodman Richard Ray
Ali Haghighat Guang Xu Rick Brake
Alex Rawlins Heather Dougherty Rohit Pandey
Andrea Brickey Hua Jiang Rudrajit Mitra
Ankit Jha Ian Loomis Sekhar Bhattacharyya
Arash Habibi Jack Trackemas Shimin Liu
Ashish Kumar John Bowling Srivatsan Jayaraman
Bharat Belle Jon Fox Sridharan
Brian Prosser Joseph Finn Stephen Hardcastle
B S Sastry Jozef Stachulak Steven Schafrik
Calen Beaune Jurgen Brune Vaibhav Raj
Charles Kocsis Kayode Ajayi Vasu Gangrade
Cheryl Allen Keith Wallace William Reed
Craig Stewart Lihong Zhou Yi Zheng
Daniel Stinnette Liming Yuan Zach Agioutantis
Darryl Witow Marcia Harris
David Brokering Matthew Gray
D P Mishra Moe Momayez
Duncan Chalmers Myriam Francoeur
Emanuele Cauda Pedram Roganchi
Emily Sarver Pierre Mousset-Jones

xiii
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Editor(s), ISBN 978-1-032-55146-3

Sponsors

Platinum
Freeport-McMoRan
G+ Plastics
Hatch
Howden
Minetek
SRK Consulting, Inc.

Gold
Accutron Instruments
Turnstone – ABC VS & JENNMAR

Silver
BBE Consulting
Maestro Digital Mine
Mecanicad
Spendrup Fan Co.

Exhibitors
Accutron Instruments
BBE Consulting
CDC Dust Control & Air Blow Fans
CFT Compact Filter Technic
G+Plastics
Howden
Hyperflo
Maestro Digital Mine
Mecanicad
Minetek
Pinssar
Quick Supply Co
Spendrup Fan Co
SRK Consulting
TLT-Turbo
Turnstone – ABC VS & JENNMAR
Zitron

xv
Auxiliary ventilation
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

An investigation of booster fan placements in a large opening

underground stone mine utilizing CFD

N. Gendrue, S. Liu & S. Bhattacharya

Department of Energy and Mineral Engineering, G3 Center and Energy Institute, The Pennsylvania State
University, University Park, PA, USA

ABSTRACT: The optimization and planning of the mine ventilation system is a key compo­
nent of mine operation given that ventilation related costs can range between 20% and 50% of
the total operating cost of the mine. In large opening mines (LOMs) utilizing perimeter venti­
lation schemes minimal guidance is available for determining optimal booster fan (BF) place­
ments. A ventilation survey was conducted and published previously by the author which
established a CFD model for a section of an underground room and pillar large opening lime­
stone mine. In this work the previously created CFD model was utilized to investigate a total
of 15 BF positions with a focus on recirculation patterns, overall airflow within the BF entry,
and airflow around face area. It was found that the maximum airflow around the face areas
can be achieved with fan positioned on the same side of the entry as face area; with the max­
imum airflow through the BF entry being achieved when the fan is placed in the center of the
entry on the upstream side of the pillar line. The recirculation percentages were similar in all
cases reaching a maximum between 35%-40% of the total air movement. However, the highest
recirculation percentages also facilitated, via air entrainment, the highest airflow magnitudes
through the BF entry. The booster fan’s ability to stimulate airflow through adjacent entries
was found to be reduced by approximately 30% for each adjacent entry. Therefore, the recom­
mendation was given to position the BF within 3 entries of the face to achieve adequate
airflow.

1 INTRODUCTION

Mine ventilation is an essential part of any underground mining system and accounts for
between 20% and 50% of the total operating cost of the mine (Carter, 2018, Leonida, 2019,
Babu et al., 2015). In large opening mines (LOMs, mines with cross-sections >~1000 ft2), par­
ticularly in stone mines utilizing perimeter ventilation schemes, the placement of booster fans
is typically empirically determined by rule of thumb or trial and error until satisfactory condi­
tions are achieved. Many LOMs do not utilize typical auxiliary ventilation systems with an
auxiliary fan and bag. Therefore, proper BF placement is imperative considering the lack of
other ventilation controls.
In place of typical auxiliary systems many LOMs utilize large diameter fans that produce
high-volume with low-pressure; and commonly referred to as box fans or propellor fans. Com­
pared to jet fans these fans provide the increased airflow needed to maintain regulatory com­
pliance; however, the low-pressure tradeoff creates other concerns over the increased role of
natural ventilation plays in the system due to the low system pressure, ventilation efficiency,
recirculation, and short circuiting of mining sections. Specifically for large opening stone
mines there is little quantitative guidance on how to combat these issues. This work will inves­
tigate the effect of BF placement within the entry and its influence on recirculation and face
ventilation efficiencies.

DOI: 10.1201/9781003429241-1

3
 In Gendrue et al. (Gendrue et al., 2023), a ventilation survey was conducted and used to
create a computational fluid dynamics (CFD) model of a section of a large opening stone
mine. The simulations resulted in a discussion of the mines selected BF placement and its
effectiveness as well as discussions on measurement locations for ventilation surveys. In this
work, the same model will be utilized to simulate different BF placements and discussions on
the BF effectiveness around the face area and airflow directions and quantities will be given.
This work aims to provide a foundation for future CFD BF placement studies on perimeter
ventilation systems and discuss important considerations when selecting a position for a BF.

2 BACKGROUND

Previous mine ventilation studies related to face ventilation focus on auxiliary ventilation of
drifts or dead-end headings with an emphasis on contaminant distributions and removal. Vari­
ables typically considered when evaluating face ventilation are fan selection, curtain or auxiliary
tubing placement, airflow penetration depth, entrainment ratios, recirculation, and jet fan
angles being common among the studies listed below. However, large opening stone mine venti­
lation schemes are unique and must be evaluated separately. Studies specifically in LOM envir­
onments are limited however, the same quantitative variables are used to evaluate the systems.
Dunn et al. (Dunn et al., 1983) studied how BF inclination, fan elevation and ultimately air­
flow around open entries and the face area in large and medium sized airways. However, their
results and discussions were derived from field data which inherently limits the analysis due to
the smaller number of sampling locations when compared to numerical simulations. Nonethe­
less, their recommendations for optimizing face ventilation were to: (i) utilize a pushing system
rather than a pulling system, however, no pulling systems were investigated, (ii) place fans to
direct flow diagonally across the face, (iii) place fans one pillar inby the general flow to minimize
recirculation. They also noted that larger capacity fans ventilate more effectively and inclination
and elevations of the fans have little effect on face ventilation. Krog and Grau (Krog and Grau,
2006) discussed the use of vane axial or propellor fans in terms of their recirculation, overall
airflow, and entrainment ability for use in LOMs. Ultimately giving recommendations that pro­
pellor fans be placed in the fresh air stream and work best for regional ventilation applications
while vane axial fans be placed 1 entry behind the airstream and work best for dead end ventila­
tion i.e., headings. They also tested pushing and pulling systems for a new mine utilizing
a textbook example of a split mine ventilation scheme concluding that a BF positioned just
outby a portal is effective for mine airflow. Later, Grau and Krog (Grau III and Krog, 2009)
discussed BF placement and mine design criteria such as long pillars in terms of ventilation effi­
ciency for LOMs. Recommendations were given to place the booster fans outby the last open
crosscut. Furthermore, it was noted that fans positioned in the middle of the entry will increase
recirculation and that “not all recirculation is detrimental.”
These three studies, however, do not cover all the common ventilation schemes used today in
LOMs. The study by Krog and Grau (Krog and Grau, 2006) studied only one adjacent entry
when three to five entries are common in perimeter schemes. They also investigated a split mine
ventilation scheme for a new mine. Grau and Krog (Grau and Krog, 2009) utilized a 3-entry
heading which operated under a pseudo split ventilation scheme with fresh air on one side and
return air on the other. The BF as directed perpendicular to the face opposite what is done in
perimeter ventilation where the fan is directed parallel with the face. Furthermore, the study by
Dunn et al. (Dunn et al., 1983) may initially seem like a perimeter ventilation scheme but all the
tested fan positions were at a 90° angle from the general airflow direction which is more analo­
gous to a unit ventilation scheme. These studies also utilized smaller fans than are typically seen
today with the largest studies being an 8 ft. in Krog and Grau, 2006. Therefore, these studies are
fundamentally different not only in terms of fan technology but also in terms of their fundamen­
tal ventilation system design as defined by Krog et al. (Krog et al., 2004) as split, unit, and per­
imeter schemes.
BF placement and ventilation system designs have been noted in many studies as the most
or one of the most important factors in LOM ventilation systems, however, there are minimal

4
studies that discuss optimal BF placement (Thimons et al., 1985, Dunn et al., 1983, Grau and
Krog, 2009, Grau III et al., 2006). Thus, BF placement must be reexamined with current fan
capacities and in a perimeter ventilation context with the fan directing airflow parallel to the
face area rather than at a 90° angle to it.

3 METHOD

3.1 Partner mine description and CFD model creation

A full-scale geometry was developed based on the mine map from a limestone mine located in
Pennsylvania, USA. The entry widths and heights average 12.2 m (40 ft.) and 9.1 m (30 ft.)
respectively, with 24.4 m (80 ft.) pillar centers. The mine utilizes curtains, stoppings, and booster
fans to deliver fresh air across the mine with a perimeter ventilation scheme, with curtains estab­
lished approximately 5-6 entries back from the mining face. Figure 1 shows the plan view for
the mine with the curtains, stoppings, and booster fans highlighted in black. The general airflow
pattern is also shown with the red and green arrows with the different colors on the mine map
indicate the different levels of benching. The yellow highlighted section in Figure 1 indicates the
section of the mine that was utilized for regional CFD modeling. This section of the mine was
chosen for simulations because the shape is consistent with a typical in a perimeter ventilation
scheme. Furthermore, no benching was done throughout the region allowing for a consistent
mining height which is common around the face areas of perimeter schemed mine. These two
factors allow the results to be broadened to other LOMs utilizing perimeter schemes.
To summarize, the model creation that was reported in Gendrue et al. (Gendrue et al., 2023),
the boundary conditions for the model were measured directly from the mine during a ventilation
survey utilizing a velocity (mass) inlet condition with a pressure outlet. Curtain lines on the east
and north side of the geometry as well as the mine boundary on the south and west were utilized
as the edges of the geometry. Two booster fans are located within the model domain, one near the
center of the geometry labeled inner BF (IBF) and the second near the exhaust portal labeled
exhaust BF (EBF). The EBF was held consistent throughout all simulations with only the location
of the IBF changing. The specific BF location that was utilized by the mine at the time of the
survey can be seen in Figure 2A as location 1 in the original fan locations callout. Table 1 shows
the measured airflow velocity from the mine survey as well as the BF settings and descriptions.
The measurement locations and labeling system used in Table 1 can be seen in Figure 2B

Figure 1. Plan view of the partner limestone mine with curtains, stoppings, and booster fans highlighted
in black, with the selected CFD simulation region highlighted in yellow.

5
Table 1. Measured airflow velocities and fan settings observed during the ventilation survey that were
used for the boundary conditions of the CFD simulation, measurement locations and names are shown
in Figure 2B.

3.2 Booster fan placement scenarios

To evaluate multiple BF locations two scenarios were created totaling 15 BF locations as
shown by the blue fan symbols in Figure 2A. Scenario one hereby called original fan location
consisted of the original mine selected BF location, on the floor in the center of the entry the
6N and pillar line 7W in Figure 2B. Eight other fan positions within the same entry and pillar
line were investigated in a 3x3 grid pattern around the mines original position for a total of 9
positions in the original fan location scenario. The 3x3 grid pattern was set up within the
entry so there was 5 ft. of air space between the edge of the IBF and the pillars edge. Given
the fans are 12 ft. diameters the center of the fan was aligned 11 ft. from the pillar and crosscut
edge. The center of the entry and center of the pillar line was then used for the positions in the
middle of the grid. All nine positions are marked as ‘original fan location’ can be seen in the
callout of Figure 2A. This 3x3 grid pattern enables fan positions at the front, back, and center
of the pillar line to be examined while also investigating positions nearer to each pillar edge
and the center of the entry. The second scenario was created because two different Groups
Dunn et. al., and Krog and Grau (Dunn et al, 1983, and Krog and Grau, 2006) came to simi­
lar conclusions about placing the fan in a forcing fashion one pillar inby the general airflow
without providing any data in their works for pulling fans. Therefore, two new locations were
investigated one that was three pillars upstream to examine a more forcing fan location and
another location three pillars downstream to study the effect of pulling airflow through the
region. These two locations are hereby called inby fan location and outby fan location respect­
ively and are labeled as such in Figure 2A. Only 3 fan positions were investigated for each of
the inby and outby locations positioned in a similar fashion to the original scenario: in the
center of the pillar line 11 ft. from the pillar edge in the outer positions and in the center of the
entry for the middle position. The same boundary conditions were used for all simulations
with only the location of the BF changing within the domain.

4 RESULTS AND DISCUSSION

4.1 Calculation of airflow from simulation files

The airflow through north to south entries labeled 1N-8N and east to west entries labeled
1W-12W in Figure 2B was calculated from the simulation file at the centerline of pillars.
Figure 2B shows the XZ and YZ planes that were used to calculate the airflow in each simula­
tion along with the corresponding labels. The airflow values were calculated directly from the
simulation file by velocity vector integration multiplied by the calculated cross-sectional area
of the entry. Since the cells in the mesh are not guaranteed to be uniform in size and the fluid
was treated as incompressible, an area weighted average was used for vector integration. The
volume of airflow entering the model excluding inlet 9 due to its location in the domain based
on the ventilation survey was ~280 m3/s (~487,000 CFM) including the airflow measured from
the S-CC and N-CC crosscuts which are 3 entries outby and 5 entries outby the original BF
pillar line. Thus, any airflow values calculated from the simulations above 280 m3/s indicates

6
some level of recirculation. The air flowing South-North (1N-8N) will be referenced as ‘entry’
airflow while the air flowing West-East (1W-12W) will be referenced as ‘crosscut’ airflow. Air­
flow that is opposite the BF direction (flowing from north to south) in the entries and from
east to west in the crosscuts is labeled as negative airflow and is used to calculate
a recirculation ratio. The recirculation ratio is defined as the sum of negative airflow across
a pillar line divided by the total airflow across the pillar line.

Figure 2. A) Booster fan placements inside the model geometry, 15 different booster fan locations were
simulated with 9 arranged in a grid pattern between the same pillar marked original fan location, another
3 locations are 2 pillars inby the original location and the final group of 3 locations being 2 pillars outby
the original location B) Location of airflow measurements from the simulation files, entries parallel to the
booster fan direction are labeled 1N-8N while the crosscut entries are labeled 1W-12W with the original
booster fan placement shown as the blue fan symbol.

4.2 Scenario one: Original fan location

The calculated airflow values from the entries and crosscuts of the nine original BF positions
are shown in Figures 3 and 5. The positive and negative airflow values are shown for each
pillar line segregated by the entry, 1N through 8N in Figure 3. The BF entry label on the y-axis
is the pillar line on the 7W plane shown in Figure 2B. Location 7 shows the highest positive
airflow in entry 6N of ~680 m3/s two pillars outby the BF with location 3 showing the lowest
airflow of 407 m3/s in the BF pillar line. Locations 3 and 5 show the highest absolute airflow
across the pillar line with the BF entry while all other locations reach a maximum one to 3
pillars downstream the BF. Location 2, 3, 5, and 9 all show similar patterns with high airflow
at the BF pillar line followed by a decrease in airflow one pillar downstream. This is due to
the placement of the BF and the pillar geometry. As noted in Krog and Grau (Krog and
Grau, 2006), when higher velocity is observed through the main entry the airflow catches the
corners of the pillars and is directed perpendicularly through the crosscuts. This is what was
observed at the first crosscut in location 3,4,5,6 and 9 causing a lower airflow value to be
observed along the first pillar line downstream. The velocity contour around the breathing
height (Z=2m) for original locations 6 and 8 are shown for comparison in Figure 4 with the
airflow around the BF circled in black. In Figure 4A the airflow hits the first pillar corner
downstream in entry 6N leading to immediate and intense recirculation around the left adja­
cent BF pillar. comparatively, the airflow in Figure 4B fully engulfs entry 6N leading to higher

7
airflows and the traditional entrainment pattern at the first crosscut. Furthermore, the airflow
entering the model through inlets 1-8 in Figure 4A is mainly pulled though entry 8N while the
inlet airflow in Figure 4B is more uniformly distributed throughout all entries. This shows the
importance of BF placement within the context of pillar geometry. The pillar geometry was
created from the mine map, and when pillars are offset, or not perfectly aligned due to blast­
ing, then BF airflow and effectiveness can be minimized due to the airstream disruption as
described above. In typical mine ventilation surveys that an operator would perform daily it is
common to measure airflow downstream of the BF and in an adjacent entry when time per­
mits. With these minimal measurement locations, the region cannot be fully quantified. In
Figure 4A and 4B the red circled areas on the left upstream side of the region have minimal
airflow of under 0.5 m/s and are insignificantly affected by all fan positions in scenario one.

Figure 3. Airflow through entries 1N-8N for scenario one, fan locations are marked 1 through 9 in
a counterclockwise direction, positive airflow indicates flow in the booster fan direction (downstream)
and negative airflow indicates flow opposite the booster fan direction (upstream).

The right fan positions in the entry (locations 2, 3, and 9) show the lowest total air move­
ment across the pillar lines and through entry 6N in Figure 3; these locations also have the
highest downstream positive airflow through entries 7N and 8N which is likely hindering air
entrainment as noted in Krog and Grau (Krog and Grau, 2006) that a more uniform distribu­
tion of airflow will not induce entrainment. Since this uniform airflow does not occur when
the BF is on the right side of the entry it is likely due to the fan positioned on the same side of
the entry as the adjacent curtain line. Indicating that a fan positioned on the same side of the
entry as a curtain/stopping line in adjacent entries will minimize BF airflow due to the loss of
the booster fans entrainment abilities.
Table 2 shows the recirculation percentage across the pillar lines 1W-12W as defined in
Figure 2B in section 4.1. Recirculation percentages above 30% are highlighted in orange while
percentages below 10% are highlighted in blue. When no recirculation occurs, the cell is
marked with a dash (-). With the BF positioned in the center of the entry (positions 1, 4,
and 8) there are similar airflow magnitudes in Figure 3 as to when the BF is positioned on the

8
left side of the entry but with less recirculation as seen in Table 2. An increase in airflow and
recirculation can be seen in the first two pillars upstream from the BF when the fan is posi­
tioned on the left-side of the entry (positions 5, 6, and 7) which is not seen with the center
positioning (Org 1). The left-side locations (Org 5, 6, and 7) also account for the highest recir­
culation percentages in Table 2. With the fan positioned on the left side the cone effect of the
BF is maximized allowing for the airflow to be distributed across the entry without pillar/rib
interaction; however, with these fan positions the recirculation around the left adjacent pillars
are the highest compared to all other positions in Figure 3. Due to the large airflow (velocity)
gradient between the BF entry and the adjacent entries in the left positions it is unsurprising
these locations also have the highest airflow due to recirculation and entrainment properties.
Interestingly, in Figure 3 the position 8 graph is more similar to the left-side graphs in terms
of total airflow, recirculation amounts, and face airflow. Indicating that more total air move­
ment can be achieved when the BF is positioned in either position 8, in the middle of the entry
in line with the upstream side of the pillar edge, or in the left positions, on the same side of the
entry as the open/face area.

Table 2. Recirculation percentages across the Pillar lines (1W-12W) from

scenario one showing all 9 booster fan positions.

Figure 5 shows the recirculation percentage through the entries as in Section 4.1. When the
BF is positioned in the center of the entry (positions 1, 4, and 8) there are similar airflow mag­
nitudes in Figure 3 as to when the BF is positioned on the left side of the entry but with less
recirculation as seen in Figure 5. An increase in airflow and recirculation can be seen in the
first two pillars upstream from the BF when the fan is positioned on the left-side of the entry
(positions 5, 6, and 7) which is not seen with the center positioning. The left-side locations
also account for the highest recirculation percentages for all pillar lines (1W-12W) in Figure 5.
With the fan positioned on the left side the cone effect of the BF is maximized allowing for the
airflow to be distributed across the entry without pillar/rib interaction; however, with these
fan positions the recirculation around the left adjacent pillars is the highest compared to all
other positions in Figure 3. Due to the large airflow (velocity) gradient between the BF entry
and the adjacent entries in the left positions it is unsurprising these locations also have the
highest airflow due to recirculation and entrainment properties. Interestingly, in Figure 3 the
position 8 graph is more similar to the left-side graphs in terms of total airflow, recirculation
amounts, and face airflow. Indicating that more total air movement can be achieved when the
BF is positioned in either position 8, in the middle of the entry in line with the upstream side
of the pillar edge, or in the left positions, on the same side of the entry as the open/face area.
Figure 5 shows the airflow through the crosscuts labeled1W-12W with each bar representing
a north to south pillar lines labeled as right or left of the BF entry. Again, positive airflow in
Figure 5 represents air moving from west to east in the geometry in Figure 2B. The total bar
height represents the magnitude of airflow through the crosscuts or the exchange between the

9
Figure 4. Velocity contour of positions 6 and 8 of scenario one near the breathing plane (Z=2m), A)
location 6, B) location 8.

entries. The BF location in scenario one is between crosscuts 6W and 8W. Given the large
amounts of air entrainment seen in positions 1, 4, 6, 7, 8, in Figure 3 it is expected that the air­
flow in the first and second crosscuts (7W and 8W) are high in magnitude and positive on the left
and negative on the right side of the BF entry indicating air flowing inwards towards the BF
entry i.e., entrainment. Furthermore, the maximum airflow at these locations in Figure 3 is
reached a maximum either one pillar (80 ft.) or two pillars (160 ft.) downstream the BF which is
consistent with the finding of Krog and Grau (Krog and Grau, 2006) as described by Figure 6 in
their work that maximum airflow for propeller fans being reached ~ 52 m (170 ft.) downstream.
Moving to ventilation effectiveness across the face area the ability to dilute contaminates
generated near the face is also related to crosscut flow due to the geometry, the only way to
transport containments out of the mine is towards the general airstream in the BF entry which
must be achieved through crosscut airflow. Therefore, it is important to maximize the crosscut
airflow near the face (5 and 4 pillars to the left of the BF entry). Locations 5, 6, and 8 in
Figure 5 shows similarly high values for airflow 5 entries to the left of the BF entry (between
entries 1N and 2N) compared to all other fan positions. This may be due to the higher overall
airflow generated by the BF across the region however, location 7 also has high absolute air­
flow of ~804 m3/s which does not translate to higher face airflow. Interestingly at position 7 in
Figure 5, marked by the red box, the first left adjacent entry (1 left) has the highest positive
and overall airflow through it which does not translate to high face airflow. Thus, leading to
the conclusion that not only is high regional airflow necessary to ventilate further adjacent
entries but some amount of airstream disruption via air diversion from pillar corners is
needed to maximize crosscut exchange between entries.
To quantify the influence of the BF to adjacent entries the percent reduction in airflow
exchange (crosscut airflow) between the BF entry and adjacent entries can be calculated.
When face airflow is higher, as seen in positions 5, 6, and 8 a reduction in crosscut airflow can
be seen starting from 1 pillar to the left of the BF entry until approximately 4 pillars to the left
as shown by the red arrows in position 5 on Figure 5. When averaged across these positions
a total decrease in airflow of 91% on the positive side and 84% on the negative side is seen.
Indicating that crosscut airflow is reduced by approximately 46% to 55% for every consecutive
adjacent entry from the BF entry. Therefore, the BF effectiveness to facilitate airflow past
four adjacent entries is seen to be less than 15% of the BF entry and may not be sufficient to
be relied upon for mine ventilation planning.

4.3 Scenario two: Three pillars inby and outby the original fan location
Six more simulations were conducted three pillars inby and three pillars outby the original BF
location. Booster fan positions 1, 2, and 6 were selected for these inby and outby locations.
The calculated airflow values were obtained in the same way as the previous section.

10
Figure 5. Airflow through the crosscuts (1W-12W) from scenario one, fan locations marked 1 through 9
in a counterclockwise direction, positive airflow indicates flow from west to east in the crosscuts (1W to
12W) while negative airflow values indicate airflow in the opposite direction.

Figure 6 shows the calculated airflow values through entries 1N through 8N with the ori­
ginal inby and outby BF locations marked on the y-axis. Inby location 1 shows the highest
airflow through entry 6N with a value of ~ 575 m3/s with inby position 6 showing the lowest
airflow through entry 6N of ~330 m3/s. Inby location 1 showed significant negative airflow
values in entries 3N to 5N throughout the middle portion of the geometry, which facilitates
the high airflow in entry 6N through air entrainment as discussed in Section 4.2. The outby
fan locations in Figure 6 shows minimal airflow through the majority of the geometry due
to the BF being in the outby position. However, similar maximum airflow values are still
reached downstream from the BF. All three of the outby fan positions did not have negative
airflow until 3 pillars upstream from the BF location which allows the majority of the sec­
tion to have positive airflow. This leads to all three outby simulations (similar to original
location 2) having positive airflow throughout the face area (entries 1N-3N). However, the
magnitude of the positive airflow seen around the face area was not significantly larger than
any of the other simulations that saw negative airflow. Indicating that if avoiding negative
airflow is a necessity then a more outby BF location would be desired with a preference
towards a fan placement in the middle of the entry or on the same side as the adjacent cur­
tain/stopping lines.
The recirculation percentages in Table 3 of the outby and inby fan locations are similar in
magnitude to that of the original locations. The recirculation at the original, and outby loca­
tion 6 have the highest magnitude and occurrence while inby location 1 has the highest mag­
nitude and occurrence. Recirculation across the BF pillar line can be thought of as the
‘overall sections recirculation’ given that the air is directly recirculating from downstream to
upstream of the BF. Location 6 has the largest magnitude in all positionings, 36% for outby,
30% for original, and 33% for inby positions. This is consistent with the results from Table 2
showing the highest recirculation percentages in original locations 5, 6, and 7 of scen­
ario one.
Lastly, the crosscut airflow from the inby, outby, and original locations are shown
in Figure 7. A similar reduction in airflow in adjacent entries can be seen in both the
inby and outby locations. Inby location 6 shows a significantly different pattern than

11
the rest of the locations due the high airflow through entry 4W. Similar to Figure 4A,
the airflow in entry 6N in the inby location 6 simulation caught the corner of a pillar
and split the air stream between entry 6N and crosscut 4W. This air stream continued
towards the face and split resulting in positive airflow downstream and negative air­
flow upstream through entry 1N as shown by the red circle in the inby location 6
graph in Figure 6. Even with this extreme example of airflow being directed towards
the face the average reduction in crosscut airflow still remains and is 79% and 74% for
positive and negative airflow respectably in the inby locations. The average reduction
for outby locations is 91% and 77% for positive and negative airflow respectably.
These reductions correspond to a 25-30% reduction per pillar line in all cases, indicat­
ing that positioning the BF at an inby or outby locations does not have a significant
effect on crosscut airflow.

Table 3. Recirculation percentage of the inby, original, and outby fan locations,
fan locations 1, 2, and 6.

Figure 6. Airflow through entries 1N-8N for booster fan locations 1, 2, and 6 at the inby, original, and
outby fan locations, positive airflow indicates flow in the booster fan direction (downstream) and nega­
tive airflow indicates flow opposite the booster fan direction (upstream).

12
Figure 7. Airflow through the crosscuts (1W-12W) from the inby, original, and outby fan locations, posi­
tive airflow indicates flow from entries west to east in the crosscuts (1W to 12W) while negative airflow
values indicate airflow in the opposite direction.

5 SUMMARY AND CONCLUSION

A CFD model was created based on a mine ventilation survey of a typical perimeter ventilation
schemed large opening room and pillar underground mine. 15 BF positions were investigated in
two scenarios: the first scenario consisted of nine fan positions around the original BF placement
selected by the mine operator with the second scenario consisting of three fan positions three pil­
lars inby and outby the original location. The airflow around the BF was investigated through
recirculation patterns and overall airflow within the BF entry and the room and pillar region with
a focus around the face velocity. The following conclusions were made based on the simulations:
(1) The highest total airflow in the section may not always be obtained when the BF is posi­
tioned at the center of the entry but rather when the BF is positioned on the same side of
the entry as the open room and pillar areas.
(2) As noted in Krog and Grau (Krog and Grau, 2006) and confirmed in these simulations
air entrainment is maximized when high velocity gradients or non-uniform distribution
of airflow is seen through the adjacent BF entries. Furthermore, the maximum airflow
through the BF entry was confirmed through the simulations to be ~ 160 ft downstream
the BF which is consistent with their findings of ~170 ft.
(3) Not only is high regional airflow necessary to ventilate further adjacent entries from the
BF but some amount of airstream disruption via air diversion from pillar corners is
needed to maximize crosscut exchange between entries.
(4) The airflow through the crosscuts observes a 46%-55% reduction per entry indicating
that booster fans in this geometry have minimal effect past 4 adjacent entries. Leading
to a recommendation to keep the BF within 3-4 entries of the face.
(5) In situations where recirculation must be avoided a BF located more outby from the
center of the section is preferable. This outby fan location was shown to create ‘positive’
airflow around the face areas i.e., non-recirculating airflow.
While all mine geometries are unique and BF positioning should be investigated on a case-
by-case basis with the purpose of the BF in mind, these simulations and recommendations can
act as a starting point for LOM operators who utilize a perimeter ventilation scheme.

13
ACKNOWLEDGEMENT

This work was financially supported by The National Institute of Occupational Safety and
Health (NIOSH) under contract No. 75D30119C05743.

REFERENCES

Babu, V.R., Maity, T., Prasad, H., 2015. Energy saving techniques for ventilation fans used in under­
ground coal mines—A survey. J. Min. Sci. 51, 1001–1008. https://doi.org/10.1134/S1062739115050198.
Carter, R., 2018. Focusing the Flow. Eng. Min. J. 219, 50–56.
Dunn, M., Kendorski, F., Rahim, M., Mukherjee, A., 1983. Testing Jet Fans in Metal/nonmetal Mines
With Large Cross-sectional Airways.pdf.
Gendrue, N., Liu, S., Bhattacharyya, S., Clister, R., 2023. An investigation of airflow distributions with
booster fan for a large opening mine through field study and CFD modeling. Tunn. Undergr. Sp.
Technol. 132.
Goodman, G., Taylor, C., Thimons, E., 1992. Jet Fan Ventilation in Very Deep Cuts-A Preliminary
Analysis.
Grau III, R.H., Krog, R.B., Robertson, S.B., 2006. Maximizing the ventilation of large-opening mines.
Proc. 11th U.S./North Am. Mine Vent. Symp. - 11th U.S./North Am. Mine Vent. Symp. 2006 53–59.
https://doi.org/10.1201/9781439833391.ch8.
Grau III, H., Krog, R., 2009. Using mine planning and other techniques to improve ventilation in
large-opening mines. Min. Eng. 61, 46–50.
Hargreaves, D.M., Lowndes, I.S., 2007. The computational modeling of the ventilation flows within
a rapid development drivage. Tunn. Undergr. Sp. Technol. 22, 150–160. https://doi.org/10.1016/j.
tust.2006.06.002.
Krog, R.B., Grau III, R.H., Mucho, T.P., Robertson, S.B., 2004. Ventilation planning layouts for large
opening mines. Soc. Mining, Metall. Explor. 1–9.
Krog, R.B., Grau, R.H., 2006. Fan selection for large-opening mines: Vane-axial or propeller fans -
Which to choose? Proc. 11th U.S./North Am. Mine Vent. Symp. - 11th U.S./North Am. Mine Vent.
Symp. 2006 535–542.
Leonida, C., 2019. Changing the Face of Mine Ventilation.
Nguyen, V.D., Heo, W.H., Kubuya, R., Lee, C.W., 2019. Pressurization ventilation technique for con­
trolling gas leakage and dispersion at backfilled working faces in large-opening underground mines:
CFD analysis and experimental tests. Sustain. 11. https://doi.org/10.3390/SU11123313.
Parra, M.T., Villafruela, J.M., Castro, F., Méndez, C., 2006. Numerical and experimental analysis of
different ventilation systems in deep mines. Build. Environ. 41, 87–93. https://doi.org/10.1016/j.
buildenv.2005.01.002.
Thimons, E., Kohler, J., 1985. Measurement of air velocity in mines.

14
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Quantifying assemblage losses in auxiliary ventilation systems

E. De Souza
AirFinders Inc., Kingston, Canada

ABSTRACT: While mine ventilation systems may account for 40% to 50% of the energy con­
sumption of a mine operation, auxiliary ventilation alone may be accountable for half of this con­
sumption. In effect, auxiliary ventilation systems comprise a significant portion of a mine
operation’s base energy demand and is consequently responsible for a large percentage of the total
mine operating costs. This paper presents how engineering design principles can be applied to
design efficient and reliable auxiliary ventilation systems, especially focusing on assemblage losses.
Case studies are presented to demonstrate the effect of design, installation and maintenance prac­
tices on system reliability and operating costs. In particular, the effect of assemblage losses is
quantified in terms of operating efficiencies, energy consumption and costs.

1 INTRODUCTION

Increasing costs of electricity have resulted in emphasis on energy-efficient designs and oper­
ation for all energy-consuming systems in mining. Since ventilation systems normally account
for 25-40% of the total energy costs and 40-50% of the electrical consumption of a mine oper­
ation (De Souza, 2018, 2013), the optimization of ventilation systems is today a ventilation
engineer’s top priority.
While modern duct-fan systems require precise engineering design, meticulous attention to
installation and regular maintenance practices, many installations are often designed based on
outdated rules of thumb and with disregard to best installation practices. In the course of
many years of investigations of duct-fan systems, the author has found them to be, in general,
fairly energy inefficient, with many systems operating at efficiencies below 65% and with air
leakages ranging between 25% and 75% (De Souza, 2004).
Auxiliary ventilation systems comprise a significant portion of a mine operation’s base energy
demand and may be accountable for half of a mine ventilation system energy consumption.
Typical auxiliary ventilation systems are fairly energy inefficient; factors affecting mechanical
and ventilation efficiencies loss include design flaws, installation practices and air leakage.
General solutions and tactics for improving auxiliary ventilation systems as presented in
this paper come from multiple ventilation audits performed by the author. They target sub-
system components which affect shock losses and mechanical and ventilation efficiencies. By
increasing the efficiency of auxiliary ventilation system components and by correcting
inappropriate designs or system degradation caused by poor maintenance, the overall capacity
of the system in delivering air to the required active faces can often be improved.

2 CASE APPLICATION

A case application associated with extensive engineering work conducted by the author is pre­
sented in this section to demonstrate how, by conducting detailed ventilation efficiency audits,
simple low-cost solutions can be devised to increase efficiency, reduce power consumption,
and lower operating costs.

DOI: 10.1201/9781003429241-2

15
 The case study is based on the auxiliary ventilation system illustrated in Figure 1. A series
of analysis are performed to quantify the contribution of each component of the complete
system to energy consumption and costs: inlet bell, screen, silencers, system friction, couplings,
bends, duct exit. The effect of air leakage and of installation practices is also quantified. The
analysis was based on analytical procedures and scientific guidelines developed by the author
(De Souza & Dirige, 2022) and use of specialized software (AirFinders, 2022).
The auxiliary ventilation system, to be installed in a development drift to supply 18.88 m3/s
air for a diesel production fleet rated at 298.3 kW, has the following design characteristics:
– layflat duct - new, 1.2192 m diameter, provided in 15.24 m long sections, 10 sections of
ducting. Multi clip joints.
– spiral duct, new, 1.2192 m diameter, 1 section 15.24 m for bend. Multi clip joints.
– total column length - 167.64 m.
– fan: 1.219-0.80-1780 (fan diameter-hub diameter-rpm) with a 149.14 kW motor. 600 V.
Motor efficiency 95%. Power factor 0.84.
– inlet bell - 1.651 m diameter, 0.2159 m long.
– screen - wire mesh screen of 95% net free area.
– silencers - two podless flow through silencers.
– bend - right angle normal bend of 1.524 m radius.
– cost of power: $0.08/ kW.hr.
– operation - 24 hours/day, 365 days/year.
– standard density conditions.
Three cases, with operational features presented in Table 1, are considered. Case 1 represents
the system design to meet the flow requirements of at the face of 18.88 m3/s and considers
a design air leakage of 20%. The fan supply flow is 23.6 m3/s. An ‘installation quality factor’ of
10%, representing a ‘good’ installation, is used to adjust the system static resistance pressure.
Case 2 represents the system as installed. Attained flows at the face of 15.56 m3/s do not meet
requirements, and is based on an air leakage of 30%. The fan supply flow is 22.23 m3/s. An
‘installation quality factor’ of 30%, representing a ‘poor’ installation, is used. Case 3 represents
the system with the fan blade setting adjusted to achieve the required face flow of 18.88 m3/s.
Air leakage is not constrained, remaining at 30%. The fan supply flow is 26.97m3/s. The system
installation quality is not improved; an ‘installation quality factor’ of 30% is used.

Table 1. Operational features of three duct system cases.

Case 1 2 3
Face Flow 18.88 m3/s 15.56 m3/s 18.88 m3/s
Leakage 20% 30% 30%
Fan Flow 23.6 m3/s 22.23 m3/s 26.97 m3/s
Installation Quality Good Poor Poor
Inlet Bell Installed No Bell No Bell
Screen 95% Net Free Area 30% Net Free Area 30% Net Free Area
Silencers Proper Connection Non-Aerodynamic Non-Aerodynamic
Friction Head Loss Same Resistance Same Resistance Same Resistance
Coupling Losses Same Resistance Same Resistance Same Resistance
Bend Proper Installation ‘Kinked’ ‘Kinked’
Duct Exit Full Section Reduced Section Reduced Section

The contribution of each duct system component to power consumption and operating
costs is presented in the following sections.

2.1 Inlet bell

The fan inlet bell ensures smooth air flow through the fan intake and serves to minimize
entrance losses. Case 1 has a proper inlet bell installed and in Cases 2 and 3 the fan is installed

16
without an inlet bell. Table 2 presents a summary of results for the 3 cases. For cases 2 and 3,
substantial increases in operating cost of 723% and of 1,370% are noted when an inlet bell is
not used. Table 2 clearly shows that, when an inlet bell is installed, significant energy and cost
savings can be achieved.

Table 2. Head losses, power, and operating costs for inlet bell.
Case Head Loss (Pa) Power (kW) Cost/Year ($/y) % Change in Cost
1 14.73 0.59 412.28 –
2 108.90 4.84 3,393.06 723
3 160.27 8.64 6,058.39 1,370

Figure 1. Auxiliary ventilation system configuration.

2.2 Screen
The fan screen prevents debris from entering the fan. Case 1 has a screen of 95% net free area
installed and in Cases 2 and 3 the fan screen is partially blocked by the deposition of debris.
Table 3 presents a summary of results for the 3 cases. For cases 2 and 3, substantial increases
in operating cost of 340% and of 686% are noted when the screen is partly blocked with debris
and not well maintained.

Table 3. Head losses, power, and operating costs for screen.

Case Head Loss (Pa) Power (kW) Cost/Year ($/y) % Change in Cost
1 11.29 0.45 315.99 –
2 44.65 1.99 1,391.19 340.26
3 65.71 3.54 2,483.91 686.06

2.3 Silencers
Silencers provide a level of noise reduction to meet specific needs and for compliancy with
regulations. Case 1 has silencers properly connected to the fan and in Cases 2 and 3 the silen­
cers have a non-aerodynamic connection to fan. Table 4 presents a summary of results for the
3 cases. For cases 2 and 3, relatively large increases in operating cost of 19.7% and of 45% are
noted when the silencers are not properly installed.

17
Table 4. Head losses, power, and operating costs for silencers.
Case Head Loss (Pa) Power (kW) Cost/Year ($/y) % Change in Cost
1 49.77 1.99 1,392.95 –
2 53.50 2.39 1,666.93 19.67
3 53.50 2.89 2,022.36 45.19

2.4 Friction head losses for layflat duct

For the three cases, the layflat duct has the same frictional resistance, however the resistance
pressures vary as a function of the airflow volumes passing through the duct column due to
leakage effects. Table 5 presents a summary of results for the 3 cases. Because of the reduced
flows, case 2 has a reduced operating cost of 19.1% and, for case 3, because of the increased
fan flow, an increase in operating cost of 44.4% is noted.

Table 5. Head losses, power, and operating costs for layflat duct friction losses.
Case Head Loss (Pa) Power (kW) Cost/Year ($/y) % Change in Cost
1 604.66 21.60 15,136.12 –
2 469.54 17.47 12,240.01 –19.13
3 691.04 31.19 21,854.54 44.39

2.5 Friction head losses for spiral duct

For the three cases, the spiral duct has the frictional resistance, however the resistance pres­
sures vary as a function of the fan flow. Table 6 presents a summary of results for the 3 cases.
Because of the reduced fan flow, case 2 has a reduced operating cost of 1.2% and, for case 3,
because of the increased fan flow, an increase in operating cost of 76.4% is noted.

Table 6. Head losses, power, and operating costs for spiral duct friction losses.
Case Head Loss (Pa) Power (kW) Cost/Year ($/y) % Change in Cost
1 226.75 9.06 6,346.37 –
2 201.23 8.95 6,269.83 –1.21
3 296.16 15.97 11,195.12 76.40

2.6 Coupling losses for layflat duct

For the three cases, the layflat duct has the same resistance due to couplings, however the
resistance pressures vary as a function of the airflow volumes passing through the duct
column due to leakage effects. Table 7 presents a summary of results for the 3 cases. Because
of the reduced flows, case 2 has a reduced operating cost of 19.1% and, for case 3, because of
the increased fan flow, an increase in operating cost of 44.4% is noted.

Table 7. Head losses, power, and operating costs for layflat duct coupling losses.
Case Head Loss (Pa) Power (kW) Cost/Year ($/y) % Change in Cost
1 64.27 2.30 1,608.96 –
2 49.91 1.86 1,301.10 –19.13
3 73.46 3.31 2,323.12 44.39

2.7 Coupling losses for spiral duct

For the three cases, the spiral duct has the same resistance due to couplings, however the resist­
ance pressures vary as a function of the fan flow. Table 8 presents a summary of results for the

18
3 cases. Because of the reduced fan flows, case 2 has a reduced operating cost of 1.2% and, for
case 3, because of the increased fan flow, an increase in operating cost of 76.4% is noted.

Table 8. Head losses, power, and operating costs for spiral duct coupling losses.
Case Head Loss (Pa) Power (kW) Cost/Year ($/y) % Change in Cost
1 71.42 2.85 1,998.86 –
2 63.38 2.82 1,974.75 –1.21
3 93.28 5.03 3,526.02 76.40

2.8 Bend
Case 1 has a properly designed bend and, in Cases 2 and 3, the installed bend is ‘kinked’,
resulting in a higher resistance pressure. Table 9 presents a summary of results for the 3 cases.
For cases 2 and 3, relatively large increases in operating cost of 97.6% and of 252.8% are
noted when the bend is not properly installed.

Table 9. Head losses, power, and operating costs for bend losses.
Case Head Loss (Pa) Power (kW) Cost/Year ($/y) % Change in Cost
1 45.40 1.81 1,270.75 –
2 80.59 3.58 2,510.85 97.59
3 118.60 6.40 4,483.27 252.80

2.9 Exit losses

Case 1 has a properly installed duct end, with its full cross-section open, and in Cases 2 and 3
the installed duct end has its exit reduced in section, resulting in a higher resistance pressure.
Table 10 presents a summary of results for the 3 cases. For cases 2 and 3, relatively large
increases in operating cost of 109.2% and of 273.5% are noted when the duct end is not prop­
erly installed.

Table 10. Head losses, power, and operating costs for exit losses.
Case Head Loss (Pa) Power (kW) Cost/Year ($/y) % Change in Cost
1 157.07 5.02 3,516.58 –
2 337.29 10.50 7,356.40 109.19
3 496.41 18.74 13,134.63 273.51

2.10 System component contribution summary

Table 11 presents the contribution of each individual component, relative to Case 1, to annual
operating costs. For case 2 the overall system component contribution to costs is 19.1% and
for case 3 it reaches 109.6%. This clearly indicates the potential for cost savings when an auxil­
iary ventilation system is properly designed, commissioned, and maintained.

2.11 Fan operation

Figure 2 presents the fan curve and operating points for the 3 cases and Table 12 presents
details of the fan operation. For cases 2 and 3, relatively large percent increases in fan input
power and operating cost of 10% and of 94.5%, relative to case 1, are noted. It is pointed that,
for case 2, the system is not in compliance since the face supplied flow does not meet regula­
tory requirements. Also, for case 3, the fan operates very close to stall. To reduce risk in both
cases, the installed system components should be improved including, installing an inlet bell,
maintaining the fan screen, correcting the silencer connections, installing a proper bend, and

19
 Table 11. Component percent contribution to annual operating costs.
Component Percent $/year change relative to case 1
Case 2 Case 3
Inlet bell 723.0 1,369.5
Screen 340.3 686.1
Silencers 19.7 45.2
Friction layflat –19.1 44.4
Friction spiral –1.2 76.4
Joints layflat –19.1 44.4
Joints spiral –1.2 76.4
Bend 97.6 252.8
Exit 101.2 273.5
Overall 19.1 109.6

the correcting duct outlet. Also, the duct column installation quality should be improved, and
air leakage controlled. These actions will bring the system operation close to the design
specifications.

Figure 2. Fan characteristics and operating points for the three cases.

Table 12. Fan operation for the three cases.

Input
Flow Blade Angle Efficiency Brake Power Power Cost/Year % Increase
Case (m3/s) TP (Pa) (degrees) (%) (kW) (kW) ($/y) in Cost
1 23.6 1,615.30 19.5 64.5 59.1 62.2 43,773 –
2 22.23 2,049.48 19.5 70 65.09 68.51 48,160 10
3 26.97 2,983.49 28 70 114.94 120.99 84,797 94.5

20
2.12 Operational costs
Table 13 presents annual fan operating costs as a function of the number of duct installations.
Independent on the number of installations, fan operating costs increase by 10% and 93.7%
for cases 2 and 3, relative to case 1. Typical hard rock mines may have well over 40 auxiliary
ventilation systems installed to support development and production activities. For the case
study presented, annual cost savings reaching some $1.65M can be realized when a system is
installed and maintained according to design.
It is noted that the comparative analysis was based on direct costs only. It is recognized that
ventilation training, asset management and preventive maintenance produce significant reduc­
tions in ventilation operating costs. While indirect costs are site dependent, the author has
observed that when workers acquire practical ventilation training prior to working under­
ground, which normally comes at a very nominal cost, considerable improvements in ventila­
tion system performance are readily realized.

Table 13. Annual operating costs for multiple duct system installations.
Number of Duct System Installations
Case 1 5 10 20 30 40
1 43,773 218,865 437,730 875,460 1,313,190 1,750,920
2 48,160 240,800 481,600 963,200 1,444,800 1,926,400
3 84,797 422,983 847,965 1,695,930 2,543,895 3,391,860

3 CONCLUSIONS

This paper has demonstrated how the use of proper engineering design for optimal duct
system installations would dramatically result in reduced energy consumption and in reduced
operating costs. When properly designed, installed, and maintained, an auxiliary ventilation
system can operate efficiently with substantial power and operating cost savings.
In a case study presented, it has been illustrated that savings in fan energy consumption
exceeding 94% can be achieved by correctly commissioning and maintaining a properly
designed auxiliary ventilation system.

REFERENCES

AirFinders, 2022. AirFinders Force Auxiliary Ventilation System Design. Engineering Design software.
De Souza, E. 2004. Auxiliary ventilation operation practices. Proceedings of 10th U.S./ North American
Mine Ventilation Symposium. Anchorage:Balkema. 341–348.
De Souza, E. 2013. Improving the energy efficiency of mine fan assemblages. Proceedings of the 23rd
World Mining Congress, Montreal, QC. pp 9.
De Souza, E. 2018. Cost saving strategies in mine ventilation. CIM Journal. Volume 9. Issue 2.
De Souza, E. & Dirige, P. 2022. Auxiliary mine ventilation manual. Version 2.0. Workplace Safety
North. Pp 261.

21
Case studies of mine ventilation
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Kucing-Liar (KL) mine LOM preliminary ventilation design –

PT Freeport Indonesia

A.A. Habibi
Technical Expert, PT Freeport Indonesia

I. Setiawan
Chief Ventilation Engineer, PT Freeport Indonesia

R. Prasojo
Mine Engineer, PT Freeport Indonesia

C. Stewart
Principal Engineer, Minware Consulting

ABSTRACT: A ventilation study was conducted on Kucing Liar (KL) mine. KL is the add­
ition to PT Freeport Indonesia (PTFI) panel cave operations and is anticipated to operate in
parallel to Grasberg Block Cave (GBC). The study consisted of model construction, budget
preparation and investigating multiple scenarios to meet the minimum budget requirements.
Utilization of new engine technologies controlled partial air reuse, push-pull and utilization of
Battery Electric Vehicles (BEVs) were considered during this feasibility study. A main intake
and exhaust drift size optimization study was also conducted to determine the number of add­
itional main fans and drift sizes required to support the KL mine. The results show the
requirement for two additional 5MW main fans, two intake drifts and two exhaust drifts
(7 m x7 m). The results also show the requirement for eight 6m ventilation raises to support
primary air delivery to KL mine. Proof of concept trials are deemed necessary to determine
the feasibility of applying some of the proposed concepts in underground operations (such as
BEVs). The trial results will be used to complete a detailed ventilation network design.

1 INTRODUCTION

GBC mine development started in 2004 with initial drifting and the first drawbell blast in 2018.
GBC mine is expected to achieve full production rates of 130 kt/d to 160 kt/d of ore by 2025.
Currently, the three active underground mining operations in the Grasberg Mining District are
the GBC block cave mine (target 130-160 kt/d ore), the Deep Mill Level Zone (DMLZ) panel
cave mine (target 80 kt/d), and the Big Gossan open stoping operation (7 kt/d). The Deep Ore
Zone (DOZ) panel cave mine (80 kt/d) has been closed and is no longer in operation.
The KL orebody is a new large caving operation where development started in 2021 and pro­
duction is due to begin in 2027. The Grasberg Open Pit commenced production in 1990 with
active pit production ending in late 2019. Figure 1 is a schematic view of the Grasberg District
orebodies. The concentrating plant has a peak capacity of about 240 kt/d. Peak production targets
from the underground district will remain at 240 kt/d and will be entirely from underground. The
GBC will be the primary source of mill feed targeted at 130-160 kt/d (Brannon et al. 2020).

1.1 Existing ventilation infrastructure and network layout

KL mine is located at a lower elevation from GBC. The main exhaust and intake drifts (labled
the Grasberg Vent Drift (GVD) system) will be shared between the existing GBC mine and future

DOI: 10.1201/9781003429241-3

25
KL mine. Five Howden mixed flow rated at 5 MW each are currently installed in a parallel in an
exhaust configuration to support the GBC mine. Fresh air from the mill valley is downcast to the
mine through four parallel primary intakes (GVD 1-4) mined at 6.8 mW × 9.0 mH, at 2.6 km
from the portals to the footprint. Air is distributed to the working areas of the mine using ventila­
tion drifts and raises mined at various dimensions. The return air reports back to the under­
ground main exhaust fans and out into the valley. Figure 2 displays the main components of the
ventilation system. The system capacity is 3500 m3/s when all five fans are running.
Based on the recent LOM baseline estimate, the current GBC mine (which will be operating
in parallel with KL block cave) consumes approximately 25 m3/s/ktpd. Other block cave
mines typically have ratios from 17 to 40 m3/s/ktpd (Brannon et al. 2020). while the ratio cur­
rently sits at the lower end of the scale, a Ventilation On Demand (VOD) system has been put
in place to improve the efficiency of air distribution.

Figure 1. PTFI mining district overview.

Air velocity and diesel exhaust dilution requirements form the main design criteria. The air­
flow velocity requirements air primarily for dust control in the ore flow system, and diesel
dilution for development and production activities.

Figure 2. GBC and KL mines network layouts.

26
1.2 Ventilation budget
The KL mine shares primary intake and exhaust drifts with the GBC mine the ventilation cap­
acity of this system is insufficient for the combined GBC and KL development and produc­
tion, and will need to be expanded by the year 2026 with additional intake and exhaust drift(s)
and additional primary ventilation fan(s).
Based on current activity forecasts, GVD ventilation capacity is required to increase from
3500 m3/s to 5100 m3/s, proportional to planned production increases from the combined
mines. KL mine requirements peak briefly at 2400 m3/s, although as this peak is short-term,
2200 m3/s is used as an economic design target for long-term requirements.
Freeport PT mine airflow requirements are based on diesel equipment and minimum air vel­
ocity criteria, depending on location and activity.
• All working areas must maintain a minimum 0.3 m/s velocity standard.
• Fixed facility infrastructure with working personnel adopts a higher 0.5 m/s velocity
standard.
• Remotely operating diesel equipment (MineGEM) requires a minimum air velocity of
0.8 m/s to ensure equipment cooling.
• Diesel activities with personnel present, until recently were required to meet an 0.08 m3/s/kW
standard at a nominal machine utilization factor (typically 80%). This has since been simpli­
fied to 0.067 m3/s/kW at 100% machine installed engine power.
• For BEV ventilation budget calculations, minimum velocity of 0.6 to determine the budget,
which exceeds the minimum velocity requirement. This approach resulted in higher airflow
requirement compared to kW-based approach.
The above criteria represent minimum flows that should be met or exceeded at the work
location. To allow for operational loss and leakage, a leakage factor is applied to some activ­
ities, and a ‘balancing’ factor to maintain minimum airflow delivery to multiple locations is
also applied. In addition, an air density factor at the primary fans is applied to account for
density differences due to elevation.

2 DESIGN SCENARIOS

The GBC mine production is scheduled to increase to 136 ktpd by 2024. KL mine develop­
ment has commenced, with steady-state full production planned for 90 ktpd by 2033. The
mine will supplement GBC mine production which reduces to 103 ktpd by 2033, giving
a combined total tonnage from both mines from 2033 of 193 ktpd. Five scenarios were con­
sidered as case studies for this design.

2.1 Scenario 1. Base case - diesel mine using existing strategies

The base case represents a diesel equipment mine, operated to existing strategies and airflow
guidelines. To meet planned production requirements, the proposed baseline ventilation
design increases total airflow through the combined GBC + KL primary system from
3500 m3/s to 5100 m3/s, an increase of 42%, in line with the production increase. Economic
analysis using mining and fan capital costs and discounted life of asset operating costs
revealed there was no prospect of increasing airflow through existing mining infrastructure
due to high system resistance, existing fan performance limitations, and other practical consid­
erations (such as mine and ventilation disruption during construction).
The primary surface system design requires an additional two (2) exhaust drifts, with pri­
mary fans (two 5MW mixed flow fans) totaling 5.2 km development drifts. In addition, two
intake drifts totaling 4 km DEQ are required. Air to KL mine is delivered from the GVD
exhaust and intake system via four (4) intake raises and four (4) exhaust raises totalling
2.9 km at 6m in diameter.

27
2.2 Scenario 2 - KL mine air resuse
An air reuse option redistributes 600 m3/s of cleaned infrastructure exhaust to the extraction
level, effectively reducing the airflow drawn to and from the surface by 600 m3/s. This permits
a reduction in primary main fan infrastructure requirements, and a resultant reduction in cap­
ital and operating costs.
Infrastructure level exhaust is proposed for reuse as it is lower (per unit volume) in diesel
activity (DPM is more difficult to scrub and reduce). Reintroduction of the scrubbed air to
the extraction horizon is a form of ‘series’ ventilation meaning gases will not continue to build
as may be the case in a recirculated design.
This proposal is speculative, in that no effective validated air-reuse scrubber arrangement
has been proven for this mine. To do this, the characterization of dust and contaminants
would need to be performed, and available technologies evaluated for cost and efficiency.
A proposed arrangement is to parallel 80 m3/s capacity scrubbing units together in at least
two locations on the infrastructure air return level, linked by raises into the extraction level.
Several proven wet and dry technologies exist for removing fine silica and nuisance dust
from the air. Removal of DPM from the air requires dry filter scrubbing which should be
avoided from a likely cost and maintenance perspective. The most likely robust fine dust
removal technology is ultra-fine wet spray chambers, using water atomized with the assistance
of air compressors. For design purposes, power totaling 400 kW is assumed to overcome 2
kPa of filter media or sprays with collectors, and 1.5 kPa air reinjection into the mine extrac­
tion level.

2.3 Scenario 3 - BEV utilization at extraction level combined by air reuse

The use of BEVs for loader extraction activities is expected to reduce airflow requirements by
275 m3/s. To achieve a step change to 600 m3/s airflow reduction (required to eliminate an
intake and exhaust drift), a further 325 m3/s is assumed to be re-used – a limited version of the
previous option.
The option reduces infrastructure requirements similar to the previous air reuse case (one
less intake and exhaust drift/fan) but is slightly cheaper and lower in power cost due to the
reduction of air scrubbers. The capital and productivity assumptions of BEV loaders are
excluded from this cost.

2.3.1 BEV design details

Consideration of BEVs for production mucking and loading activities was undertaken to
determine potential ventilation improvements to the KL mining zone. BEVs are considered
zero-emission vehicles, emitting no discernible gases or diesel particulates. Oxygen levels
remain constant and heat emissions are estimated at less than 20% compared to diesel equip­
ment (McGuire, et. Al. 2022).
The minimum airflow requirement for BEVs is therefore limited by only:
• Minimum air velocity requirements
• Build-up and removal of dust that would otherwise cause health & safety issues, or visibility
problems.
• Other noxious gases that enter the atmosphere – e.g., blasting fumes or oxidation of ores.
• Heat build-up from all sources, but primarily from external sources such as rock strata or
hot water, instead of BEV machinery.
Fire risk and resultant toxic gas emissions from BEV fires have been flagged as serious haz­
ards. Despite several mining BEV fires reported worldwide, no evidence to date suggests BEV
fires in underground mining are more common or more hazardous than diesel machine fires
(Stewart, 2022). Nonetheless, hydrogen fluoride emissions during lithium battery fires have
been flagged as a potentially lethal atmospheric hazard, and when coupled with burning
hydrocarbon products (such as tires) and potentially reduced airflows for BEV use, the con­
centrations of fire combustion products may be more hazardous than an equivalent well-

28
ventilated diesel machine fire. A BEV LHD trial is underway (in 2023) as a follow-up to
hazard and risk assessment strategy identifying BEV-specific safety issues, controls and mine
design changes.

2.3.2 BEV design strategy

The prime potential for ventilation reductions in BEVs comes from the removal of the diesel
airflow allocation rule. For PTFI, 0.067m3/s/kW would no longer be the minimum airflow
standard for operating machines.
BEV heat emissions are unlikely to cause heat build-up even with reduced airflow. Moisture
and humidity, partial products of diesel exhaust, are also reduced contributing to dryer atmos­
pheres and potentially reduced wet bulb temperatures.
Currently, PTFI uses a diesel airflow requirement of 0.067 m3/s/kW to satisfy Government
regulations and maintain suitable working temperature of the mine. This reduction in sug­
gested airflow requirements is further supported by BEV field test work in Canada which
revealed equivalent heat output of electric loaders in production activities is reduced by
a factor of 5.3 or to 18.7% of an equivalent diesel loader (McGuire, et. Al. 2022), which could
arguably support reductions in flow rates to as low as 0.013 m3/s/kW to provide the same
WBGT increase.
Alternatively, a minimum airflow velocity approach can be taken, ensuring there is suffi­
cient capacity to remove BEV and strata heat, dust and gas build-up. Applying a minimum
airflow velocity of 0.5 m/s to KL Mining schedules and airway sizes for example, in most
cases would result in an equivalent airflow ranging from 0.025 to 0.040 m3/s/kW of equivalent
electric power. Being zero-emission vehicles, the type of BEV is largely irrelevant for ventila­
tion considerations.
Apart from diesel activity areas, other active areas at PTFI have a budget airflow allocation
of 0.5 m/s while passive areas can be reduced to 0.3 m/s. Assuming dust and heat can be man­
aged at reduced airflow, a conservative estimate of BEV airflow can be made as follows.
An active airflow allocation of 0.5 m/s is assumed for BEV-operated areas, including Mine­
GEM remote loaders. This allows for reasonable dust and blast fume removal and provides
adequate cooling for exposed working personnel
The proposed guide reduces extraction panel ventilation to 7 m3/s while development with
electric trucks and loaders could be reduced to 17-25 m3/s (depending on drift size).
Figure 3 shows budgeted estimates for airflow requirements for different BEV options. An
extraction level loader option is considered, reducing extraction airflow requirements from
730 m3/s to 460 m3/s (a reduction of 270 m3/s or 12% of total KL Mine flow). To create a step
reduction in ventilation infrastructure (i.e., one less intake and exhaust drift), an additional
330 m3/s reduction in demand is still required, most likely through some air reuse.

2.4 Scenario 4. full BEV development and extraction

This scenario engages the use of BEVs for all major mining activities. This permit airflow
reductions in both development and production areas, although haulage is maintained at simi­
lar airflow levels due to the 0.5 m/s requirement to ventilate active chutes (meaning diesel
trucking for haulage transport could continue).
While reduced air usage for BEVs does have risks (greater potential concentration of air­
borne dust, slower removal of gases and blast fumes), the reduction in total airflow for KL
mine (29%) is seen as realistic and achievable and avoids the need to rely on speculative scrub­
ber technology for air reuse.
A full BEV option (trucks and loaders) considering both development and extraction activ­
ities, results in a 650 m3/s airflow reduction (29% of total KL mine flow). The lower airflow
requirements (if combined with mine air reuse and scrubbing for the partial BEV case) will
reduce capital mining infrastructure by one GVD intake and exhaust drift, and one less pri­
mary fan.
The comparison highlighting airflow versus BEV options is shown below in Figure 3.

29
Figure 3. KL mine ventilation budget reduction.

2.5 Primary ventilation drift optimization (GVDs)

The GVD primary ventilation drifts will need to support both the existing GBC block and the
new KL block. An optimization study was performed to determine the additional number and
drift dimensions of primary ventilation drifts to support the additional activity in the KL
mine for each ventilation scenario. The option with the lowest combined capital cost of
mining ventilation drifts and new primary fans, and discounted power costs over the life of
the mine is the preferred velocity. Base case scenario for the analysis is assuming diesel
powered mobile equipment. The Net Present Value (NPV) analysis has been discounted at
10% for this study and is based on budget PTFI power and mining costs. A 95% total cost
variance limit was chosen to reflect variable confidence in the assumptions used and ensure
the recommended outcomes are close to the maximum economic value.
While larger drift sizes show a slight increase for higher economic velocities, as values for
most sizes are close, it is suggested an optimum velocity of around 10.5 to 11 m/s for any drift
size should be adopted for planning purposes, with a peak upper limit of 14.0 m/s. Internal
shaft velocities can vary between 13.1 m/s and 20.5 m/s. Figure 4 shows an analysis example
for the 6.8 m X 9 m drift.

Table 1. Calculated optimum velocity for different drift profiles.

Drift Size Optimum Velocity Upper Limit 95% Lower Limit 95% Max Upper Airflow
W (m) H (m) m/s m/s m/s m3/s

4.4 4.0 10.8 13.3 8.6 234

6.8 6.0 11.0 13.6 8.6 555
7 7.0 11.0 13.6 8.6 666
6.8 9.0 11.8 14.6 9.4 894
6.0 m Raise 16.5 20.5 13.1 580

30
Figure 4. Velocity optimization for primary ventilation drifts.

Where personnel are required to travel in high-velocity drifts, dust and coarse particle
entrainment in the air can create safety hazards, particularly with eye injuries. The KL
design for PTFI uses an economic velocity range that may exceed 10 m/s for infre­
quently travelled airways while maintaining 10 m/s or less for common travel ways. Spe­
cial safety precautions and travel restrictions will apply for travelling in airways
exceeding 10 m/s. Nonetheless, every effort is made to reduce air velocity to below the
maximum economic limit. For example, for Scenario 1 (fully diesel mine, no air re-use)
two GVD intakes have been added to the system as optimum velocity analysis indicates
that this can be done at a cost-neutral basis, the impact being higher capital costs being
offset by longer-term lower operating costs.

2.6 KL Block profile size optimization

A service level drift optimization process was conducted for KL footprint to reduce the
mining cost in intake and exhaust level (service level). It was determined that at the rear of the
KL mining block, it is permissible to decrease the size of the intake and exhaust drifts, where
cumulative air requirements decrease.
Based on the distribution of open regulators, the rearmost 300m sections of exhaust
and intake drifts in the KL footprint could be reduced from 7.0m x 7.0m to 5.5m
x 5.5m or smaller, reducing overall mining capital costs without any significant detri­
mental effect on ventilation cost or delivery. The concept is shown in Figure 5. show­
ing the region of reduced airflow and potentially smaller exhaust and fresh air drift
sizes (the red shading indicates higher concentration of airflows as block delivery and
exhaust system accumulate airflow).
The reduction of rear KL Block drifts sizes would only be recommended for the
KL block if the mining footprint was not expanded further west, or if the mine
was prepared to strip or mine additional drifts if the footprint did extend west in
future.

31
Figure 5. KL service level profile size optimization.

3 SIMULATION AND ANALYSIS RESULTS

Figure 6 shows the results for selected scenarios. The base case diesel mine, utilizing existing
airflow concepts and designs, is unsurprisingly the highest cost case, requiring at least two full
intake and exhaust drifts and fans and higher power for the increased ventilation flow.
The full BEV option delivers the lowest overall capital and operating cost, reducing primary
infrastructure requirements and reducing power costs. A partial BEV and air reuse option
deliver the next lowest capital and operating costs, while an air reuse option with diesel equip­
ment is the third lowest cost option.

Figure 6. KL service level profile size optimization.

32
4 CONCLUSION

No commitment to any specific mine ventilation design scenario could be made until further
research is completed on BEV equipment options, and air reuse strategies. An intermediate
KL Mine Development plan required for all scenarios is recommended for one additional
7.0m x 7.0m GVD intake, exhaust drift, primary fan and four 6.0m internal shafts each for
the intake and exhaust connections. In summary, the intermediate mining and infrastructure
plan will increase GVD exhaust capacity from 3750 m3/s to nearly 4500 m3/s.
Mining an additional GVD exhaust and intake access of 7.0m x 7.0m linear development
meters was recommended. While the optimum development size should technically be 6.8m
x 9.0m high, a more rapid single-pass mining option of 7.0m x 7.0m is acceptable which gives
near equivalent financial optimization at the loss of 100 m3/s of airflow.
The addition of a 5MW primary mixed flow fan was recommended and added to the plan
as well as four KL exhaust raises (360m each at 6.0m diameter) and four equivalent KL intake
raises.
The intermediate plan is required at a minimum for every conceivable scenario until the via­
bility of options of BEV or air reuse is resolved. Further work will be undertaken to validate
BEV viability and air reuse viability. An additional intake and exhaust drift with a primary
fan and internal raise will only be required if the reuse of infrastructure air or BEV use cannot
be made viable.
The full BEV option provides the lowest-cost ventilation option and, subject to other viabil­
ity analyses of BEVs, is recommended. Partial air re-use is viewed with some risk and contro­
versy, particularly if diesel equipment is continued to be used, as fine DPM is difficult to
remove.
Electric vehicles (BEVs) are seen as an inevitable implementation in underground mines
given the demonstrable benefits in lower heat, reduced airflow requirement and the elimin­
ation of gas and DPM emissions. As the mining ventilation plan for scenarios 1, 3 and 4 are
similar, the implementation of BEVs does not influence the intermediate mine ventilation
infrastructure plan, only the later size and type an air-re-use facility.

REFERENCES

Brannon, C. Brard, D. Pascoe, N. Priantna, A. 2020, “Development and production update for the Gras­
berg Block Cave mine – PT Freeport Indonesia”, 8th International conference and exhibition on Mass
Mining virtual conference Proc, pp 747 to 760, 9-11 December 2020, Santiago Chile. ISBN 978-956-
19-1196-3
McGuire, C, Witow D, Mayhew M, Bowness, K. 2022, “Comparison of heat, noise and ore handling
capacity of battery-electric versus diesel LHD”, Australian Mine Ventilation Conference, Gold Coast,
QLD, Australia, 10-12 October 2022.
Stewart, C.M., 2022. “Ventilation considerations and modelling of lithium battery fires in underground
mining.” Australian Mine Ventilation Conference, Gold Coast, QLD, Australia, 10-12 October 2022.

33
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

A case study in successful use of spot cooling for underground

shaft sinking

M. Brown, D.W. Durieux, C. McGuire & D. Witow

Hatch

ABSTRACT: This paper presents engineering design and equipment selection for a successful
temporary spot cooling installation to support underground shaft sinking that has been in oper­
ation for the last 14 months: September 2021 to November 2023. The need for cooling was
driven by the depth of shaft sink (starting from >1,900m below surface), however this system
was subject to many of the common challenges preventing the widespread use of spot cooling,
including limited process water & dewatering capability, heat rejection equipment placement in
the path of blasting fumes, limited airflow quantity for heat rejection, and layout constraints due
to the existing and upcoming mine services installations and construction. Components were
selected for mine-duty with consideration for the dusty environment. Use of hybrid cooling
towers allow for increased heat rejection capacity from evaporative cooling while maintaining
a fully closed-loop condenser water circuit. Skid-mounting of all components allowed for easy
placement and relocation. Use of HDPE piping lashed to existing ground support allowed for
maximum layout flexibility and minimized installation time. Performance, operational features,
and additional lessons learned, including feedback from operations personnel, are shared.

1 INTRODUCTION

This paper concerns a case study project which is currently in execution phase. This project
required a small-scale underground “spot cooling” system deployment to support shaft sinking
operations in the construction of a new winze. Mechanical refrigeration at this relatively small
scale is widely understood and deployed in industrial surface applications; however, the ability to
effectively reject heat within the confines of the underground workplace requires careful design
due to a variety of factors including limited footprint and headroom availability, high air temper­
atures, low available air flow volumes, and dirty/dusty air conditions. Additionally, where these
spot cooling installations are proposed for temporary applications, there is often a constraint on
available electrical services and water supply/removal capabilities, and little desire to invest in
reticulation systems to support the temporary facility.
None of these challenges are, on their own, sufficient to disqualify mechanical refrigeration
from spot cooling applications. Operational challenges can be attributed to multiple small issues
concurrently as opposed to a single, broad failure in design. Instead, careful attention needs to be
paid in the design and specification of equipment for underground service, to obtain fit-for-
purpose equipment which maximizes the chance of success given the known constraints. This
paper presents one such case study in which the selection of spot cooling equipment specifically
for the challenging underground service has resulted in a robust and reliable operation.

2 APPLICATION DESCRIPTION

The case study considers a project in the process of an expansion to a deep orebody which will be
accessed by an internal winze shaft from approximately 1,200m to 2,600m below surface. The

DOI: 10.1201/9781003429241-4

34
lower leg (1,900m to 2,600m) of the winze is being developed using conventional “blind” shaft
sinking methodology with a Galloway stage and equipping stage progressing in parallel. Prior to
starting development of the winze, there was no refrigeration capacity present at the mine.
Due to the significant depth of the sinking activity, refrigeration is required to manage
workplace temperature conditions and maintain safe, productive advance rates for the largely
manual mining workplace. The mine will operate a large refrigeration plant for future mech­
anized development and production activities, but this plant will not be available until the
completion of the sinking process. Therefore, a temporary interim solution was required,
which was to be fit-for-purpose for the reduced ventilation demands of a shaft sinking applica­
tion as opposed to mechanized drill & blast with multiple crews and diesel vehicles. The appli­
cation is considered to be in-line with typical “spot cooling” applications in underground
mining due to the small air flow rate and dedicated cooling of a single workplace, as compared
to “bulk air cooling” of a mine’s entire intake air stream.
A simplified schematic of the mine’s ventilation system is presented in Figure 1. Custom
ductwork was installed to allow intake air from the downcast portion of the shaft, bypassing
the heat from diesel mobile equipment operating in the service decline to support development
and construction activities throughout the mine.

Figure 1. Ventilation Schematic (Flows in m3/s).

The cooling system is located at 1,900m below surface at a mid-shaft station. During the
plant’s period of operation, none of the mine’s permanent systems and infrastructure (e.g.,
process water supply, dewatering, and electrical distribution) are available at the plant loca­
tion. Therefore, minimizing the duty with which these services were impacted by the cooling
plant was a key design consideration.

3 PROCESS CONDITIONS

Ventilation design was undertaken by creating a detailed thermodynamic model in Ventsim

software. Surface ambient conditions for the refrigeration design were selected based on the
ASHRAE 2% humidification case from a nearby weather station. These surface conditions
were used as an input for the Ventsim model, which determines the corresponding air tem­
perature at the intake of the spot cooling system.

35
 The outlet temperature from the air-cooling plant was selected by the design engineers
based on heat modelling, reflecting conditions to mitigate workplace heat stress risk. The heat
stress index of choice at the site is wet bulb globe temperature (WBGT), and the maximum
workplace temperature target is 28°C WBGT, based on the ACGIH limit for 100% work with
a moderate work intensity.
Process conditions for the system are summarized in Table 1.

Table 1. Summary of Spot Cooling Process Conditions.

Parameter Units Value

Air Flow (Volumetric) m3/s 12.0

Mass Flow (Dry Air) kg/s 15.9
Inlet Air Temperature °C wb/db 28.4/36.0
Chilled Air Temperature °C wb/db 12.0/12.0
Barometric Pressure kPa 121.3
Nominal Air Cooling Duty kW(R) 740
Air Temperature Available for Heat Rejection °C wb/db 30.0/36.8
Maximum Air Flow Available for Heat Rejection m3/s 98.0

One critical and uncommon characteristic of the ventilation system is the use of fiberglass-
reinforced polymer (FRP) construction for the sinking ventilation duct. This material has better
thermal insulative properties than steel or flexible fabric duct, allowing for warmer chilled air tem­
perature at the cooling coil outlet for a given temperature at the discharge of the duct. This selection,
in addition to equipment selections described in Section 4 below, was made with the primary object­
ive of minimizing electrical loading required to achieve a safe workplace for the shaft sinking crews.

4 EQUIPMENT SELECTION

Equipment selection for the spot cooling system focused on ensuring robust performance at
the required process conditions while attempting to minimize impacts to other mine systems
(e.g., water and electrical) during the temporary installation period. A schematic of the final
system is shown in Figure 2.
The general configuration was largely dictated based on the overall process and mechanical con­
ditions. Specifically, the air-cooling requirements fell well within typical applications for indirect
water-cooling coils (i.e., tube and fin heat exchangers). This equipment type also worked well to
facilitate tie-ins to the fresh air ductwork supporting the mine development. Although the cooling
coil is located in the mine’s fresh air stream, there is still anticipated to be dust present due to the
nature of underground operations. Thus, the design was specified to include a wider-than-standard
fin spacing of 7 fins per inch (vs. common surface applications as high as 12 fins per inch), which
reduces risk of dust accumulation plugging the coils and reducing both airflow and heat transfer
efficiency. 7 fins per inch was selected in consultation with the equipment supplier as a compromise
between risk of fouling and the reduced heat transfer performance offered by the reduced fin quan­
tity. The increased spacing has the added benefit of reducing the coil air side pressure drop, thereby
minimizing the fan power required for the temporary ventilation system.
Water is transferred in a closed loop circuit between the air-cooling coil bank and a water-cooled
chiller. The large physical space and air-flow requirements associated with an air-cooled chiller were
deemed not feasible with an underground installation location. Additionally, presence of dust in
underground workplaces presents significant operational and maintenance challenges with air-cooled
chilling equipment. Frequent inspection and replacement of air side filters are required which can put
strain on owner or contractor maintenance resources. Thus, this equipment type was not considered
for the application. A screw compressor with dual compressor arrangement was selected to maximize
compressor lift available in the event of challenging heat rejection conditions (i.e., high condenser
temperature/pressure), and R-134a refrigerant is used due to its Class A1 (non-flammable and non-
toxic) properties.

36
Figure 2. Schematic of the Cooling System.

The last component of the spot cooling system is the heat rejection circuit. Design of this area
was of critical importance in ensuring consistent, robust performance of the overall system. Design
of heat rejection systems in underground applications requires careful consideration and can be
challenging due to a variety of factors including high air temperatures, low available air flow vol­
umes, and dirty/dusty air conditions. For the spot cooling system, hybrid cooling towers were ultim­
ately selected for heat rejection. The towers operate through indirect heat exchange between the air
and a closed-loop water coil connected to the chiller. They also utilize a small spray water stream
on the coil exterior to facilitate latent heat transfer, and notably this evaporative cooling water
source is completely independent of the chiller system (which prevents contamination of the conden­
ser cooling water circuit).
Overall, this heat rejection configuration provided many benefits to the overall installation.
First, it enabled maximum thermal performance without introducing potential fouling and
water quality concerns with the chiller. The project did consider indirect heat transfer through
the use of tube and fin heat exchangers similar to the air-cooling coil; however, this configur­
ation was eliminated. The lack of evaporative cooling and the high dry bulb temperature of the
mine air stream available for heat rejection (only approximately 5°C cooler than the condenser

37
water stream) resulted in prohibitive quantities of ventilation air required to achieve the design
heat rejection.
Secondly, the use of indirect heat exchange with only an external spray design simplified the
process water circuit by minimizing make-up and blow-down water requirements as compared
to direct evaporative cooling. The range of acceptable water chemistry is also greater in the
cooling towers, which are typical galvanized steel construction, in comparison to the special­
ized material present in the chiller heat exchangers which likely requires chemical dosing and
water treatment consideration to prolong the life. This was an important consideration as the
mine water systems are limited during temporary development phases, both for supply of
fresh process water for make-up and for dewatering capacity to accommodate blow-down.
Finally, the tower design attempts to minimize concerns with dirty/dusty air conditions and
potential heat exchanger fouling (which will worsen heat transfer performance). Although the
tower does contain a coil, it is designed to reduce air-side pressure drop while still maximizing
area and heat transfer performance. The external spray water system also helps promote coil
cleaning and removal of potential dust build-up on the heat exchange surface.
It is also noted that the system was designed from a process and mechanical perspective to
minimize electrical system impacts, again an important consideration due to electrical system
constraints during early mine development periods. For example, reasonable airflow rates and
velocities were targeted to minimize pressure drops and resulting electrical motor sizes. In add­
ition, the closed-loop water circuit temperatures were mindful of chiller lift considerations and
enabled a coefficient-of-performance of slightly below 4 at maximum design conditions
(impacting the compressor motor size and electrical consumption).
Because this spot cooling system is directly coupled to the shaft sinking ventilation system,
reliable operation is vital to prevent downtime on the project critical path. Improvements in
reliability were achieved by installing piped-in standby pumps on both closed-loop water cir­
cuits. Additionally, both the cooling coils and the cooling towers consist of two units installed
in parallel. While this does not achieve full redundancy, it removes the risk of a “single point
of failure” in the cooling system causing a complete outage. Consultation with equipment sup­
pliers indicated that the system could operate (at reduced capacity) in the event of failure of
any one compressor, cooling tower or cooling coil. Preventative maintenance and cleaning of
the system was identified as a key enabler of reliable performance and would be performed on
the system regularly via a service agreement with a local supplier.

5 INSTALLATION

Another key aspect of the equipment design and selection was the ease of installation in an
underground mining environment. The system was designed such that installation and connec­
tions could be achieved by tradespeople within the workforce of typical underground mining
contractors, and not requiring any specific qualifications. Given the relatively small size and cap­
acity of the chiller, it was shipped to site fully assembled and charged with refrigerant, eliminating
the need for any qualified refrigerant technicians to participate in the installation process.
All pipe connections were Victaulic grooved, consistent with the mine’s standard for air, water
and dewatering lines. Complex piping arrangements including hard-piped headers and tie points
were fabricated in the manufacturer’s shop, minimizing the number of field connections required
and ensuring that all field connections required identical couplings. An example is shown in
Figure 3, in which four parallel inlets and four parallel outlets for the cooling coils are piped
back to a single DN150 (6-inch) grooved inlet and single DN150 grooved discharge connection.
Due to the mine layout and ventilation system, the cooling towers were installed approximately
200 metres from the chiller and cooling coils, adjacent to the return air raise. The main condenser
water piping across this distance was installed using HDPE piping, which offered several
advantages:
• Pipes were sufficiently lightweight to be supported from existing ground support screen
using chains (see Figure 4), eliminating the need to drill any rock anchors for pipe supports
for a temporary application,

38
Figure 3. Photo of shop-installed pipe headers on air-cooling coils.

• Thermal expansion and contraction can be easily accommodated by allowing the pipe to
sag between subsequent supports,
• Flexibility to adjust layout to work around existing infrastructure and avoid interferences,
• Significant reduction in the number of couplings required, improving installation time.
Flexible hose was selected for all water lines except the 200m run from the chiller to the
cooling towers. This provided flexibility to field-locate equipment skids and adjust to condi­
tions in the underground workplace.
Finally, in addition to local installation considerations, the system was specifically designed
to be easily transported from surface to underground. The system is operating in a shaft

Figure 4. Photos of HDPE Pipe Supported by Chain from Mine Ground Support Screen.

39
access mine, and therefore all components were selected and manufactured to be shipped
within cage size constraints. Of all the equipment included in the system, only the cooling
towers needed to be broken down for transport due to their overall height of 3.9m being
unsuitable for ramp transport on flatbed trucks available at the mine. A photo of the cooling
tower split in preparation for transport is shown in Figure 5.

Figure 5. Photo of Cooling Towers in Preparation for Transport.

Custom skids were engineered for each piece of equipment, which allowed for forklift trans­
port and creating a stable base that allowed the equipment to be installed on compacted mine
roadway ballast. Skids with fork pockets are visible in Figure 3 and Figure 5. The level of
flexibility was so pronounced that the complete chiller, cooling coil and evaporator water
system was installed in a temporary location to provide interim cooling in the summer
months, and then relocated to the final location only three months later.

6 PERFORMANCE

The temporary refrigeration system has been in service for 14 months as of the writing of this
paper in October, 2022. The system has been operating reliably and has yet to experience any
unplanned downtime. The only unplanned maintenance activity to date resulted from a failed
seal on the operating condenser water pump; however, the standby pump was switched into
service and the repair was made without any appreciable downtime to the plant.
A ventilation fan upgrade in 2022 has increased the total airflow delivery to the sinking
crews from 12 m3/s to approximately 20 m3/s since the start of operation. Periodic measure­
ments of the air side differential pressure have seen a negligible change over time, indicating
little dust build-up on the coils. The housing was opened after 12 months of service to clean
the coils, and no notable dust was observed. Differential pressure remained unchanged after
cleaning. A photo of the air-cooling equipment is presented in Figure 6.
Cooling tower performance has been stable throughout the operation. Dust filters on the
air intakes are cleaned periodically by spraying down with process water - a cleaning process
which can be done without shutdown of the units. The self-cleaning characteristics of the

40
Figure 6. Photo of sinking ventilation fan and air-cooling coil in service.

internal coils with water sprays have maintained consistent operation despite the dusty envir­
onment. A photo of the cooling towers in service is shown in Figure 7. It can be seen that the
intakes on the right-hand unit have recently been cleaned at the time of the photo.
Operator feedback has been positive, noting that the complete system is operated in
a hands-off manner. The chiller automatically modulates duty by adjusting screw compressor
speed using its on-board variable speed drive (VSD), and all other components operate con­
tinuously without need for any modulation or intervention.
System performance has been measured to align with the engineering design values for the
conditions to date. As sinking progresses and ambient temperatures increase, the system per­
formance at the design maximum values will be measured and validated.

Figure 7. Photo of cooling towers in service with recently cleaned air intakes on the right hand unit.

41
Computational fluid dynamics applications in mine ventilation
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

CFD study of cavern ventilation

M.A. Carvajal-Meza & J.P. Hurtado-Cruz

Universidad de Santiago de Chile, Santiago, Chile

ABSTRACT: The ventilation of mines and underground works has been studied for more
than 100 years, with extensive and rigorous results that currently aid in safely performing
underground operations of different kinds and characteristics. However, the study of under­
ground ventilation in large volumes such as caverns has seen relatively little open study, being
part of engineering studies without further disclosure. In general terms, it is the final use of
the cavern that determines the ventilation system to be used. In this regard, the present study
is generated from the need to define elementary ventilation systems for underground caverns
linked to the construction system, which allows for maximum pollutant drag with minimum
flow rate and reducing recirculating secondary flows. For this, Computational Fluid Mechan­
ics is used as a simulation and analysis tool through the Ansys Fluent software, generating
various simulation scenarios of steady and transient flows. The results show different ways of
ventilating according to the established boundary conditions and the construction geometries
used, which could help to orientate future ventilation designs from an academic exercise.

1 INTRODUCTION

1.1 Cavern design

A cavern can be defined as a cavity excavated underground, usually of large dimensions, greater
than the dimensions of the tunnels from which it is accessed. When considering the depth at
which the caverns can be found, it is possible to detach 3 groups: the first includes caverns on the
surface and in depths no greater than 3 times their diameter (Hoek, et al., 2002); the intermediate
group covers depths less than 100 meters and greater than 3 times its diameter; and the third
group, the deep ones, considers caverns located at depths greater than 100 meters (Sharma &
Judd, 1991).
Caverns, like all underground works, have different types of shapes and are generated from
a series of factors, requirements and parameters, combining to determine the dimensions of
the section to be excavated. The functionality and the desired use for the work must be taken
into account as well. Chile provides a good example, where buildings tend to have a vaulted
and arched roof to better support stresses given existing seismic and tectonic conditions (the
shape used in this study). Caverns can be classified according to the type of use that is given to
them, such as hydroelectric caverns, oil, hydrocarbon and gas storage, mining industry, for
laboratories, and other civil and/or military uses.
Specifically, in mining, caverns are one of the most important underground construction
works. Their applications cover crushing, grinding, flotation rooms, maintenance workshops
for loading and transport equipment, and storage for explosives and/or dangerous chemical
elements, among others. These factors make it important that the infrastructure generated is
economical, efficiently ventilated, and, above all, safe.
Cavern construction design is generally based on two access tunnels that arrive at the upper
and lower ends (either on the same face or in the opposite faces –this last is more used-),
making a recess from the upper part, and taking advantage of gravity to remove muck (see

DOI: 10.1201/9781003429241-5

45
Figure 1). This also aids with fortification as the walls are lowered, leaving the cavern ceiling
and walls reinforced. Also, can be find three tunnels, one arrives at the upper and two at the
lower opposite ends.

Figure 1. Examples of mining caverns of Codelco: (a) Crusher cavern of Dacita, El Teniente (b)
Crusher cavern of Chuquicamata Underground; (c) Crusher cavern of Diablo Regimiento, El Teniente.

1.2 CFD and modelling studies

The focus of this work is based on the study of cavern ventilation, an area that has not been
openly studied in detail over the years; models that can be replicated or that are capable of estab­
lishing air flow behavior inside the cavern are limited. From this, to develop this work, analogous
cases from other areas have been analyzed, with models of similar physical behavior, studying
ventilation in closed or semi-closed spaces on a smaller scale. Thus, taking advantage of dimen­
sionlessness at different scales of turbulent phenomena, the behavior of air is similar. Suspended
dust particle size is also very similar in all environments, even comparable to the size of small vir­
uses (0.3 μm). Among the studies analyzed, some of the most pertinent considered surgical wards
in hospitals (Anuraghava et al., 2021), subway stations (Guan et al., 2007) and mining ventilation
systems (Chang et al., 2019) (Zhou et al., 2020). Based on these works, it is considered a common
factor that all study models require a forced air injection and/or extraction system in closed
spaces. The paper “CFD modeling of ventilation and dust flow behavior above an underground
bin and the design of an innovative dust mitigation system” (Ren et al., 2014) establishes a CFD
analysis of the behavior of an air-injection ventilation system in coal mining. The K-Epsilon tur­
bulence model is also used together with the transport of particles in a discrete phase model
(DPM) and with a geometry similar to that of an underground cavern, where the dimensions of
the tunnel under study were considerably greater than those seen in other studies and helped
establish the behavior of dust particles within this geometry.
Various medical studies on COVID-19 prevention were considered for this work as well, includ­
ing effective ventilation in supermarkets (Mohamadi and Fazeli, 2022), classrooms (Jeong et al.,
2022), isolation rooms (Bhattacharyya et al., 2020), and hospitals (Bayatian et al., 2021, Obeidat
et al., 2021). The paper “CFD modeling of airborne virus diffusion characteristics in a negative
pressure room with mixed mode ventilation” (Anuraghava et al., 2021) shows the modeling car­
ried out for the ventilation of COVID-19 patients’ isolation rooms, using mixed ventilation sys­
tems for this purpose. The importance of this study lies in the fact that it establishes a mixed cross
ventilation system using a K-Epsilon turbulence model, together with a discrete phase particle
transport model with geometry similar to that of an underground cavern. These elements help
provide the bases to generate the mixed models that are applied.
All these works have been consulted in order to observe ventilation systems with 1 or 2 air out­
lets with different dimensions. This combination helps establish a convergence between them all in
the use of a turbulence model for particle transport, but especially in the meshing and spatial dis­
tribution, thereby validating similarities with the case of mining caverns and allowing the replicat­
ing of said models.

46
2 GEOMETRY DESIGN

2.1 Geometry design

First, we must establish the layout of the geometry to be simulated and its dimensions. For this
study, as the most of the caverns, maintain a parallelogram shape, the length being the longest
side, with extensions close to 50 meters in the case of mining (Villablanca, 2014). However, these
can reach lengths of up to 900 meters, as is the case of the Jinzhou project oil storage cavern
(Zhang et al., 2018). On the other hand, the height and width are limited depending on the geome­
chanical characteristics of the rock in the project. For this particular case, the study of cavern
model is based on two important works carried out in Chile, the Diablo Regimiento crushing
cavern and the Phase I grinding cavern of Andina, where it is assumed that in the upper part the
cavern has arc shape in order to better support the geomechanical conditions (Villablanca, 2014).
A form factor (f) was also established that uses the orthogonal dimensions of the cavern, coming
from rock fragmentation studies. Dimensionless factor f obeys a criterion used in mining in order
to provide rock fragments with a size factor, and to be able to compare them with each other. The
applied formula is shown in Equation 1. The factor is in the range 0 to 1, where if f is near 1 all
dimensions are the same (a cube), while if f is near 0 at least one dimension is very stretch. In this
work is applied to visualize how stretch or similar a cube is the cavern geometry.

This provides the results of factors shown in Table 1, from which measures with a factor
equal to the average of both (0.31) were chosen for this study. The value of 0.31 converges for
other case studies carried out, mainly those in health and social areas, including operating
rooms (Anuraghava et al., 2021), supermarkets (Mohamadi and Fazeli, 2022) and schools
(Jeong et al., 2022). Therefore, the dimensions chosen to carry out the base design according to
the geometry of the cavern are 30 m high, 21 m wide, 45 m long, and a size factor of 0.31.

Table 1. Cavern dimensions.

Study Case Height (m) Width (m) Length (m) Shape Factor (f)

Grind Cavern Phase I Andina 37 28 58 0.31

Crushing Cavern Diablo Regimiento 22 13 31 0.30
Average Value 29.5 20.5 44.5 0.31
Applied case 30 21 45 0.31

2.2 Arrangement of air inlet and outlet

In many cases, construction accesses to the cavern are taken into account to generate ventila­
tion designs. As mentioned, the construction of the cavern generally begins by connecting an
upper tunnel with a lower tunnel by means of a raise in order to carry out the developments
by recess. Therefore, only one air inlet is considered, and 1 or 2 outlets depending on the con­
struction design or subsequent use. Inlet and outlet drifts have square section 4x4m. All
designs were devised to have a direct connection between the air inlet and outlet, in order to
avoid any significant air vortices, stagnant air, or dead air spaces (Figure 2).
– Design 1: The air inlet is through an injection tunnel that is located in the upper center of
the cavern, while the outlet is on the opposite wall, lower right sector, and consists of an
extraction tunnel of the stale air found inside the cavern.
– Design 2: The air inlet is through an injection tunnel located in the upper left sector of the
cavern, while the outlet is on the opposite wall, lower right sector and consists of a tunnel
that is used to extract stale air found inside the cavern.
– Design 3: The air inlet is through an injection tunnel situated in the upper center of the
cavern, while the outlets (2) are located in the lower right and left sectors of the opposite
wall, and consist of two galleries used to extract the stale air that is inside the cavern.

47
 Special design: a design that considers the air inlet and outlet from the same face or wall,
for cases where there is no possibility of creating ventilation paths on the opposite wall. This
is considered in a separate analysis because, constructively, the excavation would only be from
below, when access to the cavern construction is limited (for example, disassembling a TBM
in a dead end).

Figure 2. Isometric view of designs 1 (left), 2 (center) and 3 (right).

3 CFD SETUP

3.1 Mesh, initial and boundary condition

The mesh is a structured hexagonal domain in the shape of regular parallelepipeds. A mesh
size analysis is performed, establishing two critical sizes with lengths of 20 and 25 cm, for
which a final size of 20 cm is established, generating 1.8 million cells (Table 2).

Table 2. Cells number by design.

Case Design Cell Number

1 1,852,108
2 1,892,369
3 1,882,534

Boundary conditions established are velocity inlet (upper tunnel as inlet) and pressure outlet
(lower tunnel), as shown in Table 3. The same table includes initial conditions, velocities, and
time steps used in the unsteady cases. Time independence tests were made to validate conver­
gence, based on Courant number less than 1. Unsteady cases were made to test flow instabilities
and some DPM cases to evaluate dust residence time, but most cases were a steady simulation.

Table 3. Boundary and initial conditions.

Condition Value Unit

Air Density 1.23 kg/m3

Gravitational acceleration -9.8 m/s2
Inlet velocity 0.5 – 2.5 – 5 – 10 m/s
Inlet Temperature 277 K
Pressure Outlet (gauge) 0 Pa
Wall Temperature 303 K
Time step for 0.5 m/s 0.4 s
Time step for 2.5 m/s 0.08 s
Time step for 5.0 m/s 0.04 s
Time step for 10 m/s 0.02 s

48
3.2 Turbulence model
In order to choose the turbulence model to be used, the researchers’ experience was considered
and an exhaustive bibliographical analysis was carried out, searching in various areas such as
transportation, energy, mining, health, and others (17 studies overall). Additionally, for this
work it not needed heat transfer. So that, the chosen turbulence model is k-epsilon Realizable,
with second order upwind spatial discretization and Simplec pressure-velocity coupling scheme.

3.3 Order of design-cases for results

The designs used have the distribution presented in Table 4 in the inlet and outlet positions of
the airflow. Designs 1 and 2 take the sides specular response into account, for one side to
another or in an inverse way, as the same behavior. In turn, the cases are classified according
to flow speed (see Table 5). Some authors (Vutukuri & Lama, 1986) suggest a minimum air
velocity of 0.5 m/s is needed to manage dust and consider that a maximum of air velocity of
2.5 m/s is required to minimize the dust concentration in the air. Additionally, the Chilean law
establish 2.5 m/s as the maximum allowed in working zones in Chile. 5 m/s is the air speed
recommended to conveyor drifts (6 m/s for main haulage routes) (McPherson, 2009). Finally,
10 m/s is the speed limit fixed by authors because is a exceed velocity which exceeds more than
double human comfortable limits and is near to economical airflow limit.

Table 4. Designs and spatial distribution.

Design Inlet Outlet

Design 1 Center Left

Design 2 Left Right
Design 3 Center Left-Right
Special Design Center Center

Table 5. Cases and airflow.

Case Air Velocity (m/s) Airflow (m3/s)

1 0.5 8
2 2.5 40
3 5.0 80
4 10 160

4 RESULTS

4.1 Designs-case results

First, an analysis of the airflow inside the cavern was carried out considering a minimum vel­
ocity limit equal to 0.004 m/s, which is the turbulent flow limit (Re≥4000), in order to avoid
gas/dust stratifications and clear away dust. In this way, the effective circulation inside the
cavern can be classified according to the percentage of sweeping or dragging of contaminants
showed in the Table 6. Figure 3 shows the Streamlines and Isosurface, specifically cases 1 and
4 (due to the extension of the work no presented all analyzed cases). The ranges of color cor­
responds to zero as minimum velocity (blue color) and the maximum velocity (red color) of
each case 0.5 and 10 m/s (cases 1 and 4), which emphasizing the differences. Isosurface dia­
grams represents the swept spaces wit turbulent flow (Re≥4000), and the empty volume repre­
sents the unswept spaces (Re<4000).
It should be noted that the simulated cases have been injected at the top and exit at the bottom,
since extraction at the top causes the particles to rise. This is very difficult to achieve for all par­
ticles due to low of air velocity, leaving long residence time and even permanent fine particles,
which is why they were not considered in this study. As Figure 3 and Table 6 show, the maximum
speed of 0.5 m/s (case 1) is insufficient to carry out a good sweep of contaminants. On the other

49
hand, the maximum speed of 10 m/s (case 4) gives good sweeping of contaminants, with the excep­
tion of the special design, wherein even with high flow speed the sweeping is not entirely good.
The special design has poor performance, even with high speed, because the inlet and outlet air­
ways are closely each other.
Table 6. Design-case combinations.
Air volume with effective Air volume with effective
Design Case circulation Design Case circulation

Design 1 1 70% Design 3 1 72%

2 75% 2 88%
3 82% 3 96%
4 95% 4 100%
Design 2 1 69% Special 1 43%
Design
2 80 2 65%
3 90 3 76%
4 97% 4 82%

Figure 3. Designs and cases for lower (left side) and higher velocities (right side) tested.

50
4.2 Optimized designs-case
From the exposed cases, the ones with the best performance have been selected, i.e., with
very high or total effective circulation. These were considered with the potential to be opti­
mized by applying the DPM model, for a turbulent flow with the presence of mineral par­
ticles requiring a minimum fluidization speed of 0.04 m/s (Sakar, 2015), to ensure an
effective drag of larger particles. To optimize cases, an analysis of particle entrainment and
particle residence time is required to establish the optimal contaminant aeration model.
Dust particles with a minimum diameter of 1 μm and a maximum of 1 mm to be injected are
considered (Ormeño, 2014).
Additionally, a mixed circuit push-pull (push from the top and pull from the bottom) was
also applied to designs 2 case 4 and design 3 case 3, to analyze if the flow improves taking into
account an additional pressure output (exhaust pressure). Figures 4 and 5 show particle Iso­
surfaces with lower velocity than 0.04 m/s (Re<4000) and particle residence time (seconds),
respectively, for Design 3-Case 4 (a) and Design 2-Case 4 mixed (b). Table 7 shows particle
drag is bad for most of cases, which means some particles staying permanently into the
vortex, except Design 3 case 4 and Design 2 case 4 Mixed. Design 3 case 4 is the best, since it
only requires air injection and not a mixed air circuit (Push-Pull).
High speeds are directly related to better ventilation and removal of polluting particles.
Although the geometries have been obtained from crushing caverns, the simulations carried
out do not have a specific focus on this use of caverns. This is a general approach for entrain­
ing pollutants to keep indoor air clean. In particular, in the case of crushing caverns, both the
air speeds at the outlet must be adequate so as not to generate contaminant dispersion, so it is
recommended to carry out a specific CFD study for these cases.

Table 7. Optimus design-cases.

Air volume with
Design Case effective circulation Particle drag? Optimus?

2 4 78% Bad No
3 3 89% Bad No
3 4 95% High Yes
2 4-Mix 95% Medium No
3 3-Mix 89% Bad No
Special 4 62% Bad No

Figure 4. Particle Isosurfaces with lower velocity than 0.04 m/s for Design 3-Case 4 (a) and Design
2-Case 4 mixed (b).

51
Figure 5. Particle residence time (s) for Design 3-Case 4 (a) and Design 2-Case 4 mixed (b).

5 CONCLUSIONS

Cavern ventilation designs are established to take advantage of the constructive galleries used
for their development. The objective of these studies is to avoid dead spaces without air move­
ment. These designs are studied through computational fluid mechanics, a k-epsilon turbu­
lence model and a particle transport model (DPM), and a type of hexagonal structured mesh
being recommended for this type of simulation. Although the proposed models generally meet
the basic purpose of ventilation, optimized designs are determined. Optimized designs allow
for entraining all the air in the cavern, but need significant velocities of 5 to 10 m/s. Design 3
with a speed of 5 m/s shows particularly high study opportunities, as it can deliver satisfactory
results.
In general, cases with areas with less air renewal are identified, particularly in the upper sector
of the face containing the exits and the space corresponding to the lower part of the wall with
the entrance; these require the implementation of boosters or fans. Designs with cross ventila­
tion are considered better, with an entrance and exit on opposite faces and positions, generating
greater air mobility inside the cavern. A good entrance speed is established with this method,
generally 5 m/s. If it is not possible to access sufficient ventilation in flow or crossover design,
the use of boosters is recommended to avoid leaving dead spaces, for which CFD tools should
be used. The special design (both accesses at the same face) has poor performance even at high
speed, so this design should be avoided from a ventilation point of view if possible.

6 FUTURE WORKS

Cavern ventilation design is a challenge to face case by case. As part of the research develop­
ment, future works other variables will incorporated as local booster fans to improve stagna­
tion zones, heat sources and heat dissipation, different shapes of cavern, others in and out
locations of the air and ascending ventilation (bottom to up). For these types of modelling or
similar, authors suggest to use DPM model to dust modeling, because is a very useful tool to
analyze it behavior according to velocities and shapes.

REFERENCES

Anuraghava, C., Abhiram, K., Sai Reddy, V. N. & Rajan, H. 2021. CFD modelling of airborne virus
diffusion characteristics in a negative pressure room with mixed mode ventilation. International Jour­
nal for Simulation and Multidisciplinary Design Optimization 12: 8.
Bayatian, M., Ashra, K., Amiri, Z. & Jafari, E. 2021. Computational Fluid Dynamics Simulation of
Airow and Air Pattern in the Living Room for Reducing Coronavirus Exposure. Research Square.
Bhattacharyya, S., Dey, K., Paul, A. R. & Biswas, R. 2020. A novel CFD analysis to minimize the spread
of COVID-19 virus in hospital isolation room. Chaos, Solitons & Fractals 139.

52
Chang, P., Xu, G., Zhou, F., Mullins, B., Abishek, S. & Chalmers, D. 2019. Minimizing DPM pollution
in an underground mine by optimizing auxiliary ventilation systems using CFD. Tunnelling and Under­
ground Space Technology 87: 112–121.
Guan, X., Jiang, Y. & Yao, Y. CFD simulation of a novel ventilation system of subway station in
Harbin. Building Simulation, 2007 Beijing, China. 432–346.
Hoek, E., Carranza-Torres, C., & Corkum, B. (2002). Hoek-Brown Failure Criterion. En North American
Rock Mechanics Symposium and The 17th Tunnelling Association of Canada Conference. Toronto:
Tunnelling Association of Canada. Recuperado el 25 de Abril de 2022
Jeong, D., Yi, H., Park, J.-H., Park, H. W. & Park, K. 2022. A vertical laminar airflow system to prevent
aerosol transmission of SARS-CoV-2 in building space: Computational fluid dynamics (CFD) and
experimental approach. Indoor and Built Environment 31(5): 1319–1338.
Mohamadi, F. & Fazeli, A. 2022. A Review on Applications of CFD Modeling in COVID-19 Pandemic.
Archives of Computational Methods in Engineering 29: 3567–3586.
Obeidat, B., Fawwaz Alrebei, O., Atef Abdallah, I., Darwish, E. F. & Amhamed, A. 2021. CFD Ana­
lyses: The Effect of Pressure Suction and Airflow Velocity on Coronavirus Dispersal. Applied Sciences
11(6): 7450.
Ormeño, M. 2014. Estudio de un sistema de limpieza aire-polvo utilizando cámara de remanso y filtración
por agua. Universidad de Santiago de Chile.
Ren, T., Wang, Z. & Cooper, G. 2014. CFD modelling of ventilation and dust flow behaviour above an
underground bin and the design of an innovative dust mitigation system. Tunnelling and Underground
Space Technology 41: 241–254.
Sakar, D. K. 2015. Thermal power plant. Design and Operation, Amsterdam, Netherlands, Elsevier Inc.
Sharma, S. & Judd, W., 1991. Underground opening damage from earthquakes. Clemson, Engineering
Geology.
Villablanca, A. M. 2014. Guía de diseño de soporte de cavernas subterráneas en roca. Universidad de San­
tiago de Chile.
Zhang, Heng, et al. 2018. Optimization Scheme for Construction Ventilation in Large-Scale Underground
Oil Storage Caverns. 10, Basilea: MDPI, 2018, Applied Sciences, Vol. 8.
Zhou, Gang, et al. 2020. Dust removal effect of negatively-pressured spraying collector for advancing sup­
port in fully mechanized coal mining face: Numerical simulation and engineering application. Amster­
dam: Elsevier, January 2020, Tunneling and Underground Space Technology, Vol. 95.

53
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Model order reduction of high-fidelity underground mine model

M. Vaze, J. Nyqvist & S. Dasgupta

ABB

ABSTRACT: Multiphysics modeling is a globally sought-after methodology to simulate com­

plex flow systems with high fidelity, given the strides made by mathematical tools in capturing
complex flow patterns together with heat transfer and gas dispersion. An application that could
benefit from multiphysics modeling, relevant to ABB offerings is ventilation control of under­
ground mines. While ABB offers state-of-the-art ventilation control systems, with advanced
strategies like ventilation-on-demand or VOD, experts are ever ambitious to take the systems to
the next level. The objective of this paper is to explore the possibility of leveraging multiphysics
modeling, specifically computational fluid dynamics (CFD) modeling with species transport
analysis to simulate gas dynamics within mines. The end goal is to improvise ventilation control
strategies, ensuring reduction of pollution levels better than ever, obtaining guidelines from
a reduced order model or ROM, a user-configurable and implementable version of the model.

1 INTRODUCTION

ABB is a leading provider of mine ventilation control technology, in recognition of the need to
optimize mine ventilation given the high energy demands and costs associated with the technol­
ogy. This paper describes an exploration to investigate the possibility of leveraging multiphy­
sics-based models to take current mine ventilation control technology to the next level. The
modeling methodology specifically investigated here is computational fluid dynamics (CFD)
modeling which has the capability of evaluating time-varying complex fluid flow, gas mixing
and thermal events within three dimensional domains. Understanding such physical complex­
ities and application of the same in improvising mine control strategies can be a significant step
in improving ABB’s lead in the competition. In modern terminology such physics-based models
are also called “digital twins”, given their capability of replicating behavour of actual systems.
While ABB, with its rich legacy in control and optimization, has developed state-of-the-art tech­
nologies in mine ventilation control, it is ever desirable to continue upgrading offerings keeping in
mind the ever-growing competition and global trends in embracing modern technologies like digi­
tal twins, machine learning, artificial intelligence etc. Certain locations in the mine, like freshly
excavated cavities and dead ends, are critical from the safety point of view, given that these are
possible sites of ventilation air starvation and consequently toxic gas accumulation. In the interest
of improving mine safety standards with optimum usage of resources, it is important to accurately
estimate critical parameters like ventilation air velocity, toxic gas concentration and temperature
distribution in critical mine locations where sensors are inaccessible. Therefore, it is desirable to
develop a tool that can perform a thorough physics-based evaluation of these locations evaluating
critical thermo-fluidic parameters with high reliability. CFD modeling technique using finite elem­
ent or finite volume discretization methods, is by far the most popular physics-based technique in
resolving complex fluid dynamics, gas mixing and thermal events within enclosures. The technique
can generate three-dimensional holistic maps of the mentioned parameters and if required can per­
form time varying analysis. Direct or indirect incorporation of CFD models would enable devel­
opment of the next generation intelligent mine ventilation control strategies, ensuring a high
safety standard while catering to growing demands on saving ventilation costs and energy.

DOI: 10.1201/9781003429241-6

54
 Several researchers in the past have used CFD model to simulate and study gas dynamics
within mines and investigate technologies to control pollution levels. Kurnia et al. (2014) per­
formed CFD modeling studies to understand methane flow dispersion in underground mines and
evaluated a novel flow divertor to focus ventilation air specifically to zones with high methane
concentration for reduction in overall concentration below safety levels. Lihong et al. (2015) devel­
oped an experimental test set-up to replicate a mine like enclosure, simulated air flow across the
domain and measured air velocities at selected points. The measurements were used in the present
study to validate the modeling methodology. More recently, Liu et al. (2019) developed CFD
model to simulate mine gas dynamics and investigated the possibility of mitigating pollution
levels, by optimizing the mine ventilation strategy, specifically the ventilation air velocity profiles.
While CFD modeling is a high-fidelity tool in analyzing complex physical events, given the
computational time associated with the models, it is desirable to develop reduced order
models or ROMs, to facilitate integration with control system platforms.
ROMs retains the essential features of a full-scale model and enables simulation of any given
set of inputs, to yield the output in acceptable time. In short, ROM offers the following benefits:
• Optimizing product design using several iterations in less time.
• Offering the opportunity to develop on-line physics-based digital twins.
• Allowing low or mid-level experts to use simulations.
The subsequent sections describe the ROM development procedure starting with development
of the full scale CFD model.

2 METHODOLOGY

2.1 Initial CFD model

As a first and essential step, an initial full-scale CFD model of the system is developed. Figure 1
a below shows the computational domain, replicating a typical mine tunnel. As seen in the
figure, a mine vehicle is modeled as the source of carbon dioxide a major pollutant in the
exhaust gases. Not shown here, the flowrate of carbon dioxide and other gases is prescribed at
the exhaust of the vehicle provided as the bottom, as a boundary condition. In this analysis we
are mainly concerned with carbon dioxide distribution, given that almost 12% of the exhaust
gases comprise carbon dioxide. The other important boundary condition is the flowrate of the
ventilation air at the exit of the ventilation duct, modeled at the top. The vehicle exhaust gases
mix with the ventilation air and the resulting mixture is flushed out from the tunnel entry.
Figure 2 shows the discretization scheme or meshing scheme to enable the finite volume calcula­
tions. All critical zones were well resolved to account for the drastic change in flow patterns.

Figure 1a. Computational domain replicating a typical mine tunnel with ventilation duct and vehicle as
the source of pollutants.

55
Figure 1b. Discretization of computational domain.

As usual in CFD calculations the mass and momentum conservation equations are first
solved. The mass conservation equation is written as:

where i stands for 1, 2, 3 (all three coordinates). We see that the mass conservation equations
is in terms of the time average velocity. The momentum conservation equation is written as

We see�that the
� momentum equation is not independent of the velocity fluctuations. The
0 0
last term ρui uj , also known as Reynolds stress term (or stress tensor), imposes an additional
stress on the fluid and manifests as the spread of momentum due to turbulence. The k-omega
SST turbulent model (a variety of RANS or Reynold’s average navier stokes equation) was
used to account for turbulence. A y+ value of less than 1 was ensured at all walls during the
simulations, as per the requirements, indicating proper resolution of the boundary layer
physics.
To solve the dispersion and distribution of carbon dioxide and other gases the species trans­
port equation is solved.

Where C is the concentration at a given spatial point and time and D is the diffusivity of
carbon dioxide in air.
The CFD model described above, yielded comprehensive details of the velocity and gas con­
centration distribution over the three dimensional domain and as a function of time.
Figure 2a shows the distribution of carbon dioxide within the domain. In the current settings
it is encouraging to see that the concentration of carbon dioxide is lower towards the vehicle
end and higher towards the middle of the tunnel and the entry. This ensures minimal exposure
of the gas to personnel manning the vehicle. Figure 2b further shows gas concentration con­
tours across transverse planes at the middle of the tunnel, the tunnel entry and end.

2.2 Model validation

A crucial stage in a CFD modeling-based study is validation of the model using experimental
data. Obtaining correct and accurate experimental data on velocity and gas concentration in

56
Figure 2a. Distribution of carbon dioxide within mine tunnel.

Figure 2b. Carbon dioxide concentration distribution across key transverse planes.

Figure 3. Comparing experimentally measured and model computed air velocities within mine-like arti­
ficial enclosure. Trends in approximate agreement.

mines is difficult owing to challenges like inaccessibility, errors in measurement devices and
limitations in the number of measurement devices that need to be used. In this paper, experi­
mental velocity measurements from a previous publication (Lihong et al. 2015) was used for
validating the modeling methodology. Lihong et al. (2015) developed an artificial mine tunnel
set up or an enclosure, with air flowing through a convoluted tunnel. Air velocity was meas­
ured at selected points within the enclosure. In this study, the set up was replicated by devel­
oping a CFD model and the model was run using the parameters selected by Lihong et al.
(2015). Figure 3 below shows the comparison between the measured and computed velocities
at the selected points. It is seen that in terms of trends the experimental and computed values
are approximately in agreement. The mismatch between the absolute values is due to several

57
unknown parameters like the uncertainty in sensor location and details of additional features
in the experimental set up. However, it may be concluded that the modeling methodology
used in this study yields results almost close to reality, and is worthwhile to be used for the
subsequent development of the ROM.

2.3 Building the ROM

The CFD model described above was re-run across several parameters enlisted in the table
below:

Table 1. Input parameters.

Input Parameters Min Value Max Value Unit
1 Air flowrate 10 40 m3/s
Air Velocity 12.7 50.9 m/s
2 Exhaust gas flowrate 1.0 3 m3/s
Exhaust gas velocity 7.1 21.2 m/s

Design of experiments or DOE was performed using Optimal Space-Filling Design (OSF)
method. Initially 10 design points (DPs) or variations of input parameters were considered.
These DP’s were simulated using the CFD model. The input parameters were mapped to the
output resulting in generation of the ROM using the genetic algorithm, working behind the
scenes. The output of the ROM is the response surface (Figure 4) which graphically depicts
the relation between the primary output results, in this case mass averaged carbon dioxide
within the tunnel volume and the input parameters, or the ventilation air and vehicle exhaust
velocities.

Figure 4. ROM relating mass averaged CO2 concentration with air and gas velocities.

The accuracy and reliability of the ROM is checked using the goodness of fit procedure the
results of which are shown in Figure 5. The procedure involves comparing the predicted
values from the response surface for some selected points or test points and with the ones
obtained from the full-scale CFD model for the same test points. Square symbols show the
design points used to develop the ROM, while the circular symbols show the additional verifi­
cation or test points. It is required that all the verification points follow the regression line to
ensure proper goodness of fit. In this case the RMS error is 1 × 10-3, an acceptable error level.
This established credibility in the ROM as a reliable tool to analyze mine gas dynamics, a tool
that retains the accuracy of the full-scale model, while allowing a quick prediction of gas dis­
tribution under a given set of conditions.

58
Figure 5. Goodness of fit procedure results to check accuracy of ROM prediction.

3 CONCLUSION

CFD modeling to predict gas dynamics within mine tunnels, yields useful information that
could guide improvisation of current ventilation protocols. However, a gap in the modeling
method is the high computational times required by the model under any given set of condi­
tions, making it difficult to use the model as an online tool for ventilation optimization. How­
ever, the response surface generated by the ROM, a result of CFD calculations, relating
exhaust gas concentration to the input parameters can be an effective tool in the online control
of mine pollution if integrated with ABB control platforms. The background of the ROM
being three dimensional high-end CFD calculations, it was seen during our investigations,
that prediction accuracy could be retained within acceptable limits. The ROM approach inves­
tigated shows promise to add a new dimension to future ventilation control technology. This
study lays the foundation towards development of a new physics-based analytical technology
for mine ventilation control. As a future step, extensive experimental studies using state of the
art sensor systems in a controlled environment may be performed and the accuracy of the
CFD model can be improved beyond current standards.

REFERENCES

Kurnia, J.C. 2014. CFD simulation of methane dispersion and innovative methane management in
underground mining faces. Applied Mathematical Modeling. 3467–3484.
Lihong, Z. 2015. CFD modeling of methane distribution at a continuous miner face with various curtain
setback distances. Journal of Mining Science and Technology. 635–640.
Liu, H. 2019. A Case Study of an Optimized Intermittent Ventilation Strategy Based on CFD Modeling
and the Concept of FCT. 1–16.

59
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Minimizing the aerodynamic impact of a new cooling plant

installed upstream of an existing surface fan and heater
arrangement through CFD analyses

J.K. Shaw
BBE Consulting Canada

L.K. Falk & C. Allen

Vale Canada

M. Kaufman
FLOWCARE Engineering Inc.

ABSTRACT: Vale’s Coleman mine, operating at depths reaching 1850m, could experience
underground summer temperatures >40°Cdb/28°Cwb without cooling. Through 2019-20,
BBE provided EPCM for a 10.3MWR vapour-compression cooling plant that allowed 15%
ambient air mixing. The footprint was restricted on three sides: two rock walls and the existing
surface intake fans/heater facility. The designed proximity of the new system, immediately
upstream, could not impact current performance.
CFD analyses, by Flowcare, were used to ensure the 6m set-back design of eight bulk air
coolers (BACs), 4 wide by 2 high, did not impact flow or pressure of the existing fan/heater,
or cooled air delivery efficiency. Seasonal operation requirements, wind effects, and mitiga­
tions were considered. Configurations evaluated included: base case (no cooling plant), per­
mutations of the BAC design, and cross winds up to 30km/h.
The CFD analyses showed negligible aerodynamic impacts and minor influence on the
fan’s 2kPa total pressure and 566m3/s flow; but wind effects potentially caused up to 40%
cooling loss. The addition of side-shields and BAC fan deflectors were shown to reduce the
loss to a maximum of 6%. This paper discusses the CFD modeling, incremental design
improvements, and system performance data gathered to date.

1 INTRODUCTION

The shockwaves of the industrial revolution that resulted in the mechanization of under­
ground mining have resulted in most underground mines in the Sudbury basin reaching
depths exceeding 1800m. At this depth, auto-compression of air alone will result in work/rest
regimes being implemented during hot summer months, which are detrimental to productivity.
To minimize the impact of hot climatic conditions underground, refrigeration plants are
installed alongside heating plants, which are used to heat the air in the winter. Among these is
Vale’s Coleman mine, located in Levack Ontario.
Vale’s Coleman mine has been in operation for over 50 years and is critical to Vale’s Sud­
bury operations. Current mining activity in the 170 OB is at depths of up 1850 m below sur­
face which introduces challenges with respect to ventilation supply and acceptable climatic
conditions. Unlike Vale’s Creighton mine, Coleman does not have the benefit of natural cool­
ing capacity. Consequently, the mine experiences hot working conditions, especially during
the summer, that have impacted employees and ultimately reduced productive face time.

DOI: 10.1201/9781003429241-7

60
 Hot underground conditions result not only from auto-compression, which increase temper­
atures by approximately 10°Cdb (dry-bulb) and 5°Cwb (wet-bulb) per every 1000m depth
below surface, but also from fan, equipment, and mobile machinery heat. Strata heat is less of
an issue as the surrounding rock can even moderate temperatures in the summer months.
However, freshly broken rock can represent a significant heat input at the workface.
With summer surface ambient conditions that are often more than 30°Cdb/20°Cwb, under­
ground workplace temperatures will be in the range that require heat exposure management
through work-recovery regimens or stop work conditions above 40°Cdb or 32°Cwb. Thus, air
cooling is required to avoid production interruptions.
In the mine planning phase, Vale recognized the need for cooling and engaged BBE in 2016
to investigate novel cooling and traditional mechanical technologies assessing both under­
ground and surface options. It was identified that although the cooling aspect was feasible,
providing the required services, electrical power, clean water supply, and locating the heat
rejection without impacting mine activities would be problematic. While focused cooling
requires less refrigeration, capacity challenges with underground infrastructure exist. Alterna­
tively, surface cooling plants are more economical per unit of cooling but result in overall
higher cooling capacity as a larger volume of air is cooled. Ultimately, the findings of the
study recommended a surface cooling plant (Hardcastle, 2022).
BBE were contracted to design and construct the plant as a turnkey project in Decem­
ber 2018. The initial design included investigating alternative mechanical options and oppor­
tunistic coolth storage solutions, with the requirement of an operating turn-key system for the
summer of 2020. In light of the 18-month design-build timeline, the unproven coolth storage
options were discounted, and Vale opted for a 10.3 MWR conventional vapour compression
mechanical system. The project was fast-tracked to deliver as much cooling as possible for the
summer 2020 cooling season, which resulted in several innovative construction practices to
bring the plant online at half capacity in the middle of construction while work continued on
the remaining plant components.
Detailed design established a plant nominal duty of 10.3 MWR based on design input ambi­
ent air conditions of 26°Cdb/19°Cwb. Ambient conditions for the mine were defined using
several years of historical climate data available publicly from the “Sudbury A” climate sta­
tion (Government of Canada, 2022), as well as ASHRAE publicized data for the same climate
station. The design input ambient conditions align approximately with the yearly 2% cooling
design temperature in the ASHRAE climatic data (ASHRAE, 2017).
Surveying and groundbreaking began in October 2019, with the project adopting an
AGILE “just-in time” engineering approach to maintain project momentum. Primary engin­
eering for permitting was complete by January 2020. Half of the cooling duty was provided by
July 2020 while construction continued around the operating plant; the entire plant cooling
capacity became available in September 2022. Full details on the timeline of construction are
available in a recorded format (Shaw, 2022) presented at the MeMO 2022 conference.
The project criteria, timeline, and budget dictated a fit-for-purpose, modular, off-the-shelf
design where pre-engineered and pre-assembled modules were used wherever possible. The
only site-specific structural designs were the Bulk Air Cooler (BAC) module supports and
cooling tower stairs. One of the many challenges to this project, in addition to the short deliv­
ery timeline, included fitting the plant into the limited 20-metre by 30-metre space available
directly in front of the mine’s main intake fans without impacting fan performance or airflow
rate, and successful winter operation of the direct-fired natural gas mine air heaters positioned
upstream of the main fans.
The plant design resulted in the placement of the bulk air coolers shown on the right-hand
side of Figure 1. The BACs were situated with a 5-metre open-air gap between the outlet of
the coolers and the inlet wall of the fan house; this was intended to eliminate any airflow dis­
ruption or added resistance to the main ventilation fans. This configuration also allows the
fans and heaters to operate in the winter without any added resistance from the BACs and
permits access in front of the fans for maintenance. However, this configuration introduces
the potential that not all of the cooled air leaving the BACs is drawn into the main fans
intake.

61
 The BACs are film-packed crossflow modular cooling towers with fans mounted to one side
used to provide the required flow through the tower and to deliver it, across the open-air gap,
to the inside of the intake fan building envelope. The fans are shown on the left side of the
BACs in Figure 1.

Figure 1. Placement of Bulk Air Coolers (Elevation View) – Base Case.

There are a total of 8 BACs, arranged into 4 pairs of units, with 2 units installed on each
level as shown in Figure 2. The main fan intake system is designed to handle a wet basis mass
flow 662 kg/s. The BACs cool an engineered flow of 563 kg/s wet basis, resulting in a net
system mass underflow of 101 kg/s, or 15%.

Figure 2. BAC arrangement & Mass underflow.

2 COMPUTATIONAL FLUID DYNAMICS ANALYSIS

The direct proximity of the plant to the primary mine intake fans was considered through
computational fluid dynamic modelling with the help of FLOWCARE Engineering Inc. utiliz­
ing Ansys CFX-18.1 software. The solver high resolution advection scheme was used in con­
junction with a standard K-epsilon (k-ϵ) turbulence model to compute steady state solutions
for the model configurations of interest (Kaufman, 2021). The heat transfer models were pri­
marily isothermal, however thermal energy modelling was used to consider the effects of buoy­
ancy of the cooled air.
The results presented were obtained using a tetrahedral mesh, whose sensitivity was minim­
ized using nominal element sizes ranging from 2 to 20 inches, resulting in approximately
1 million nodes (including boundary inflation).

62
 The essence of the CFD work was to compare varying configurations and arrangements to
the accepted base case flow. The base case flow considers the previously discussed engineered
mass-underflow with the air cooler delivering 564 kg/s of cooled air while the total mass flow
through the fans is 665 kg/s. The pressures at the main mine fans were used to indicate the
pressure loss of the system. Alternative designs were iteratively evaluated relative to this base­
line performance.
Figure 3 confirms that without wind loading, most of the BAC outlet air is received at the
fan intake as indicated by the blue streamlines converging into the heater house. Excavation
for the pump pit, which recesses the pumps and pipework below the collection basins of the
BACs, revealed bedrock at an elevation higher than expected. This resulted in a 1.1m change
in the floor elevation, bringing the entire structure up with it. This introduced airflow impinge­
ment on the top lip of the heater house which is later quantified in this analysis.

Figure 3. Confirmation of base case.

The mass flow of cooled air present at the main fan outlets for the final design of the cool­
ing system was predicted to be 557 kg/s without wind present, and 537 kg/s with a 30 km/hr
crosswind detailed further in this paper. While these values are both lower than the 564 kg/s
design value, BBE accepted the losses considering the access gap required in the design.
The fan pressures predicted by the CFD for the final cooler design were marginally higher
in magnitude when compared with the absence of BACs. The reduction in flow, considered
negligible, was expected with the placement of the bulk air coolers. Fresh air fan operating

63
data over the last two years has validated the modelling, as there have been no perceptible
changes in flow rate, fan pressure, or motor power.

3 IMPINGEMENT ON FAN HOUSE CLADDING

Elevation of the BAC structure by 1.1m, in response to the pump pit geotechnical constraints,
caused secondary losses through impingement of air on the upper lip of the heater intake
building, as shown in Figure 4. To resolve this, various degrees of airflow directionality con­
trol were examined – from 18.5 to 10 degrees downwards from the primary flow axis.

Figure 4. Losses through impingement.

Figure 5 presents the CFD predicted flow profile, shown as streamlines, exiting the coolers
and entering the existing heater house for the final design model. Crosswind flows normal to
the BAC airflow were modelled and the air from the top set of coolers was directed downward
at 10 degrees. Figure 5 also includes side walls to control the influence of local wind disturb­
ances which are discussed in the next section of this paper.
With the addition of sidewalls and airflow deflection, the mass flow of cooled air at the fan
outlets was predicted to be 537 kg/s relative to the target of 564 kg/s which BBE indicated was
acceptable. Contrasted with the 518 kg/s with sidewalls only, directing the top row of air
cooler outlet flows downwards nets a 19 kg/s (or 3%) overall improvement.
The proposed airflow directionality control comes at a significant price and access penalty
since it would require repurposing and redesigning the existing BAC structure to support the
additional control infrastructure. Repurposing the structure in this way would result in access
restrictions to the front of the fans for maintenance and inspection and would necessitate
access via aerial work platform or the installation of new stair towers.
For these reasons, the addition of airflow deflectors has been postponed until existing
improvements are complete and the effect of enclosing the system on 5 sides can be evaluated.
Considering the engineered 101kg/s mass underflow through the BACs relative to the total
mass flow of 665 kg/s air through the fans, a trade-off study quantifying the losses can be
made empirically by measuring the in-situ system. This trade-off study, which will begin when
the construction of the sidewalls is complete, will evaluate the capital outlay required to justify
the forecast 3% gain in cooled air into the fan station.

64
Figure 5. Final design - Sidewalls and 10 degree decline.

4 WIND LOSSES

CFD analysis indicated that incident wind causes a flow disruption. Annual windspeed and
wind direction were reviewed from the “Sudbury A” climate station and the wind-rose shown
in Figure 6 was produced to weight the output of the CFD analysis relative to the hours
per year at various intensities and directions to focus on the summer months.
The worst-case scenario is a southeast wind which acts normal to the flow of air; this occurs
less than 10% of July. Air arriving from the south also negatively impacts the airflow but
rarely exceeds 30km/hr. The majority of July air arrives from the west and southwest which
benefits this installation as it is blocked by the fan and heater house.
CFD analysis of a south-easterly crossflow normal to the primary axis of airflow for the
system resulted in air losses shown in Figure 7 (155° from N), which assumed a worst-case
windspeed of 30km/hr. This figure shows a large divergence in the flow streamlines indicating
substantial losses to atmosphere. The analysis varied wind directions from 110° to 235° from
N to quantify losses. Table 1 shows a summary of the results from the variation in wind angle,
in terms of the proportion of cooled air arriving at the intake fans.
Based on the results of Table 1, the decision was made to start the engineering and construc­
tion of the proposed sidewalls to limit the atmospheric interference on the cooled air leaving
the BACs shown in Figure 8. This would improve the theoretical performance of the cooled
air delivery during heavy wind from 338 kg/s to 518 kg/s.

65
Figure 6. Wind Rose for Sudbury A – July.

Figure 7. CFD Streamlines for Base Case with 30km/hr Crosswind.

Table 1. Mass Flow of Cooled Air at Fan Inlet (Design = 564 kg/s).
Wind Angle No Shields North Shield South Shield Both Shields
kg/s kg/s kg/s kg/s

235° from N 451 478 478 531

200° from N 418 438 484 511
155° from N 338 378 273 518
110° from N 385 398 484 498

66
 The sidewalls will be in place for the 2023 cooling season, and their efficacy will be quanti­
fied by comparing the 2022 cooling performance without the side shields and published in an
upcoming paper. This paper will also attempt to correlate the high windspeeds with high
ambient wet bulb temperatures.

Figure 8. Sidewalls to limit losses.

5 WINTER OPERATION

When ambient temperatures fall below the setpoint of the cooling plant, approximately 15°C,
the plant is brought offline. In addition to the CFD performed to confirm the pressure drop
across the main fans, the burner velocity profile was also verified. The burners require
a specific and uniform air velocity across their profile, and analysis indicated this would be
relatively unaffected and uniform with the installation of the bulk air coolers.
To further minimize the pressure drop associated with the BAC infrastructure by allowing
air to bypass the coolers, two louvres were installed in the sidewalls and are designed to be
opened during the winter. The louvres are sliding barn-door type and are shown in light blue
in Figure 8 of the previous section and provide access for equipment to the front of the fans
and burners.

6 CONCLUSION

CFD analysis was used to perform an aerodynamic design review of a new cooling system at
Vale’s Coleman mine in Sudbury for an existing fan intake and burner system. The results of
the CFD were used throughout the design process. The primary objectives were to confirm no
negative impacts to the existing fan and burner system, as well as maximize the amount of
cooled air delivered to the mine.
Crosswinds were determined to significantly impact the volume of cooled air delivered, and
thus it was decided to install side walls to limit interaction with incident wind. The installation
of the sidewalls began in summer of 2022 and their performance/efficacy will be studied
during the 2023 cooling season.
Field conditions discovered during construction required the BAC centerline elevation to be
raised by approximately 1 metre, causing a portion of the cooled air to impinge on the heater
house and be lost. Capitally intensive steps would be required to direct the air from the top set of

67
coolers at an angle of 10° to 18.5° toward the opening of the heater house to achieve an approxi­
mate 3% gain in cooled air delivery. The current system with its engineered 101 kg/s mass under­
flow will be re-evaluated after the installation of sidewalls to quantify the cost and benefit to
direct airflow downwards towards in the inlet.
The overall design, with the installation of sidewalls, was satisfactory for its intended pur­
pose with acceptable losses to the environment, and with minimal impact to the fans and
burners. During winter, CFD-predicted pressure losses were found to be like those for the
summer operation.

REFERENCES

ASHRAE. 2017. ASHRAE. Climatic Design Conditions. Available online http://ashrae-meteo.info/v2.0/.

Government of Canada. 2022. Historical Climate Data. Weather, climate, and hazards. Available online
https://www.canada.ca/en/services/environment/weather.html.
Hardcastle, Stephen. 2022. Selection, Design Challenges and Construction of Vale’s Coleman Mine 10
MW Surface Refrigeration Plant. North American Mine Ventilation Symposium 2023.
Kaufman, Matt. 2021. Engineering Report. CFD Analysis of Cooling System at Vale McCreedy East
Mine Shaft.
Shaw, Jeff. 2022. Construction and Lessons Learned at Sudbury’s First Mine Air Cooling Plant. Main­
tenance, Engineering, and Reliability/Mine Operators Conference - MeMO 2022.

68
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

CFD modeling of a large-opening stone mine using COMSOL

multiphysics

K.V. Raj
National Intitute for Occupational Safety and Health, Spokane Mining Research Division, Spokane,
WA, USA

Vasu Gangrade
National Intitute for Occupational Safety and Health, Pittsburgh Mining Research Division,
Pittbsburgh, PA, USA

ABSTRACT: Large-opening stone mines are characterized by entry sizes twice or thrice the
size of typical underground coal or metal/nonmetal mines. Due to the large openings, the
volume of air needed to ventilate these mines is significantly higher than in coal or metal/non­
metal mines. This leads to low air velocities with low static pressure drop. Airflow in the mine
mostly relies on natural ventilation and auxiliary fans at the working face. The National Insti­
tute for Occupational Safety and Health (NIOSH) is conducting research on the ventilation of
large-opening stone mines to reduce worker exposure to dust and other contaminants such as
diesel particulate matter (DPM). To understand airflow in a large-opening mine, we con­
ducted a numerical modeling study using computational fluid dynamics (CFD). This paper
presents the results from the CFD modeling of airflow in a large-opening stone mine. The
CFD model is calibrated against the data previously collected at a large-opening stone mine.
The team ran different scenarios with fan placement along with the movement of truck to
simulate how effectively air is moving in the mine.

1 INTRODUCTION

Stone mines produce a wide range of raw material such as basalt, granite, limestone, marble, etc.
for the construction industry needed for infrastructure development. As of 2020, there are 4,248
stone mining operations with 110 underground operations in the United States (NMA, 2020). The
majority of the underground stone mines contain broken limestones according to the standard
industrial classification (SIC) system. These underground limestone mines are characterized by
openings which are larger than the openings compared to coal and metal/nonmetal mines. The
mining method for large-opening limestone mines is room and pillar which is similar to coal
mines. However, the dimensions of large-opening limestone mines are twice or thrice of the typical
coal mines. This larger dimension of the limestone mines creates a unique challenge of ventilation
in the mines. Due to these large dimensions, airflow in the mine is affected by natural ventilation
with little differential pressure change. There are limited studies done to address this unique venti­
lation issue related to large-opening limestone mines.
In the early 2000s, the National Institute for Occupational Safety and Health (NIOSH)
started conducting research to improve air quality by reducing contaminants such as diesel
particulate matter (DPM) in large-opening mines. These studies looked into ventilation, fan
selection, stoppings, and mine planning to reduce the contaminants level in large-opening
mines (Grau III et al., 2002). The research suggested that large-opening mines face three pri­
mary ventilation challenges: moving adequate volumes of ventilation air, controlling and

DOI: 10.1201/9781003429241-8

69
directing the airflow, and planning ventilation systems that work well with production require­
ments (Grau III et al., 2006). Multiple studies were also conducted that showed increased suit­
ability of propellor fans for low-resistance ventilation systems present in stone mines compared
to the vane axial fans commonly applied in coal mines (Grau III et al., 2004, Krog et al., 2004).
Studies on the effectiveness of different large-opening stone mine ventilation stoppings showed
steel stoppings were effective for permanent stopping locations, fabric stoppings were more
effective closer to blasting areas, and long stone pillars could perform the same role as a row of
stoppings (Grau III and Meighen, 2006, Grau III and Krog, 2009). Recently, researchers have
conducted a field study of large-opening underground mines looking at variations in pressure,
relative humidity, and temperature over time (Gendrue et al., 2021). Past research, however, has
not looked into numerical techniques to analyze the airflow patterns in large-opening mines.
A numerical technique such as computational fluid dynamics (CFD) has been used for solving
airflow problems in mine ventilation for quite some time. CFD modeling has been widely used
in many areas of airflow and particulate and gaseous contaminant modeling from underground
mine ventilation to open-pit ventilation (Raj et al., 2015, Raj and Bandopadhyay, 2017, Kumar
et al., 2021, Bhargava et al., 2021, Morla et al., 2021). Most recently, Watkins and Gangrade
conducted a study using ANSYS Fluent to optimize the auxiliary fan placement in a large-
opening stone mine (Watkins and Gangrade, 2022). There are many commercially off-the-shelf
and open-source CFD software programs, such as ANSYS Fluent, Cradle CFD, COMSOL
Multiphysics, and OpenFOAM, that are available and being used to solve airflow problems in
the mining industry. The majority of these CFD programs are based on Navier-Stokes equa­
tions, the energy equation, the mass conversion, and transport equations. CFD models have the
potential to provide a pattern of airflow in large-opening underground stone mines.
This study will present a two-dimensional modeling of a large-opening stone mine using the
COMSOL Multiphysics® CFD modeling program. This paper focuses on airflow patterns
with and without a fan in the model domain and how the airflow pattern changes with move­
ment of a truck in the model.

2 CFD AS A TOOL FOR UNDERSTANDING VENTILATION IN LARGE-OPENING

UNDERGROUND MINES

The geometry for this work was adopted from Grau III and Krog (2009) work to simulate the
airflow inside the mine with four scenarios. Figure 1 shows the model geometry. The entry of
the mine is 50 ft, and dimension of the pillars are 50 ft by 50 ft. The extent of the model in the
x-direction is 1,450 ft, and in the y-direction it is 1,000 ft. In the first scenario, we simulated
the airflow inside the mine without any mechanical ventilation such as a fan, then followed
the second scenario by adding a fan to the model. Grau III and Krog (2009) mentioned in
their work that truck movement inside the mine has an impact on the ventilation system. In
the third scenario, we simulated the movement of the truck along and against the airflow.
Finally, the fourth scenario presents the simulation of the airflow with movement of the truck
as well as presence of the fan in the model.

2.1 Scenario 1: Model with no fan

This model presents as a baseline model and shows the airflow pattern in the mine. Boundary
conditions used for the model are inlet, outlet, interior wall, and wall. The inlet, outlet, and
interior walls boundary conditions stands for the main entry, main return and stopping
respectively as shown in Figure 1. The rest of the geometry representing the wall boundary.
The interior wall boundary condition represents the stopping. COMSOL Multiphysics has
predefined mesh size settings from courser to finer mesh size, and for the first scenario which
is a simple model, we used the normal predefined mesh size setting where the mesh size ranged
from 0.202 ft to 17.6 ft. Figure 2 shows an expanded view of the element size distribution of
the triangular and quadrilateral elements. Steady-state simulation was performed using the
Realizable κ-ε turbulent model. The inlet boundary condition was set to a velocity of 2.4 m/s

70
Figure 1. Geometry of the model domain (Grau III and Krog (2009)).

Figure 2. An expanded view of the mesh size (in m) near interior walls in the model geometry.

based on work of Grau III and Krog (2009) and outlet boundary condition at zero pressure.
Velocity profile along with streamlines are shown in Figure 3. It can be seen from the figure
that most of the air comes from the inlet boundary to passing in a straight path through
Return 1 and goes out without much movement in the top right corner of the model. Even
though the velocity streamlines suggest some movement and recirculation of air near the
Return 2 side of the geometry, the velocity is relatively lower compared to velocity at
Return 1.

2.2 Scenario 2: Model with fan in working area near Return 2 and Return 1
In this scenario, a fan was placed in the working area of the model near Return 2 to move the
air coming from the inlet to the top right area of the model. Figure 4 shows the placement of
the fan in the model geometry. The fan in the model represents an 8-ft diameter propeller fan

71
Figure 3. Velocity profile along with streamlines of the base airflow model.

with capacity of 55 m3/s at 0.127 kPa. Being a 2D model, model input for the fan was the
speed instead of volumetric flow rate. Both side of the fan was treated as inlet boundary con­
dition with positive and negative fan speeds. Fan speed was 12 m/s based on the data collected
for the propeller fan. Figure 5 shows the mesh elements around the fan as very fine mesh is
needed around the fan geometry to resolve the flow. However, the inlet velocity was set to
4.8 m/s as per the work of (Grau III and Krog 2009). Figure 6 shows the velocity profile and
streamlines for the steady-state simulation. It can be seen from the figure that similar to the
model with no fan most of the air is following the path from the inlet to outlet, and there is air
movement in the top near Return 2 due to the fan. However, most of the air is being recircu­
lated rather than going out through Return 2. This means that fresh air from the inlet is not
being utilized by the fan, and due to recirculation contaminants like diesel particulate matter
(DPM), dust and other engine exhaust concentrations might increase over time. Grau III et al.

Figure 4. Model geometry with fan (in red circle) in the working area near Return 2.

72
Figure 5. Mesh elements around the fan in the model.

Figure 6. Velocity profile along with streamlines of the model with fan near Return 2.

(2002) suggested that fan placement affect ventilation in large-opening mines. To this end, the
fan position was changed from working area near Return 2 to Return 1. This was done to
push fresh air coming from inlet toward the working area. Grau III and Krog (2009) in their
study used a 36-in diameter axial-vane fan with a 23-in reducer mounted at the outlet.
In the model with the fan near Return 1, the inlet velocity was 2.4 m/s with the fan speed of
19.4 m/s based on the data (Figure 6). Figure 7 shows the geometry with a fan in the model
domain. Figure 8 presents the velocity profile and streamlines of the simulation. It can be
inferred from the figure that this fan placement worked as suggested by Grau III and Krog
(2009). There is a clear indication that fresh air from the inlet is being pushed towards the
working area at the top of the model and that there is no direct path of air from the inlet to
outlet through Return 1. This scenario will lead to the removal of contaminants from the
working area through Return 2.
Figure 9 shows the velocity profiles and average velocities along the cross section taken in
the middle of Return 2 for the above models. The model with no fan has an inverted

73
Figure 7. Model geometry with fan (in red circle) in the working area near Return 1.

U-shaped velocity profile with a maximum velocity of 1.8 m/s reaching in the middle. The
model with an 8-ft diameter fan has the maximum velocity of 3.6 m/s between the three
models but a decline over the length of the Return 2 cross section. The model with a 36-in
diameter fan has the maximum velocity of around 2 m/s. However, if we look at the average

Figure 8. Velocity profile and streamlines of the model with fan near Return 1.

velocities of the three models, it was found that the model with a 36-in diameter fan has an
average velocity of 1.4 m/s followed by a model with no fan. The average velocity for the
model with no fan was 1.3 m/s, whereas the model with an 8-ft diameter fan had an average
velocity of 1.1 m/s. This shows that with an 8-ft diameter fan in the model domain, most of
the air was recirculating rather than passing through the Return 2. Grau and others also sug­
gested that movement of a truck in the mine can affect the airflow in the mine. Modeling of
the movement of a truck along and against the airflow will be discussed in the following
sections.

74
Figure 9. Velocity profiles along the cross section of Return 2.

2.3 Scenario 3: Modeling with truck movement in the mine without fan
As mentioned in the previous section, movement of a truck has an impact on the airflow move­
ment in the mine, and in order to simulate that we are first modeling the movement of a truck
without any fan in the model domain. Trucks in the mine move from working area towards the
entry and from an entry towards the working area of the mine. When the truck moves towards
the entry, it moves against the airflow whereas the truck moves along the airflow when it moves
towards Return 1. Figure 10 presents the model geometry with a truck near the Return 1. For the
first model, the truck will move from Return 1 towards the inlet against the airflow. The truck is
represented by a rectangle with a dimension of 30 × 12 ft. An expanded view of the truck can be
seen in Figure 11. The truck speed for the modeling was assumed to be 9 m/s, which is 20 miles
per hour. The important part of modeling a moving body in an airflow simulation is the domain
in which the truck will move. For modeling truck movement in both directions, the moving
domain was defined as a rectangle of a dimension of 1,250 × 30 ft as shown in Figure 10.

Figure 10. Model geometry with truck near Return 1.

75
 Modeling a moving body in CFD is challenging as the model needs to solve for both airflow and
movement of the body in the fluid domain. Apart from that, the most critical aspect is the meshing
of the moving domain as well as the rest of the geometry. Mesh elements for the geometry other
than in the moving domain and around the truck were triangular, whereas the mesh elements in the
moving domain were mapped, forming rectangular mesh elements. Figure 12 shows expanded view
of the mesh size distribution along with mesh elements in the model domain. Mesh elements around
the truck and moving domain is also visible in Figure 12. The model was run for 38 seconds during
this time when the truck moved from the Return 1 to the other side of the moving domain.
Figure 13 shows the velocity profile and streamlines at time t = 0. It can be seen from the figure that
there is air moving towards the working area near Return 2. Figure 14 shows the average airflow
passing through Return 2 over the entire simulation run. As the truck passes through Return 1,
average airflow velocity in Return 2 is increased by around 50% and starts to lower over time. This
shows that there is positive impact of movement of a truck against the airflow.

Figure 11. An expanded view of the truck in the model.

Figure 12. An expanded view of the mesh elements of the moving domain.

76
 Similarly, the modeling of the truck moving from the inlet towards the Return 1 was con­
ducted. In this scenario, the model was run for 39 seconds, and the velocity profile along with
streamlines are presented in Figure 15. Figure 15 shows that air is moving towards Return 2
at a low velocity. The average air velocity passing through Return 2 over time is presented in
Figure 16. It can be seen from Figure 16 that there is no positive impact on the air velocity at
Return 2 as seen in the previous scenario; instead, we see the opposite—that the average air
velocity decreases over time.

Figure 13. Velocity profile and streamlines at initial time step with truck moving towards the inlet.

Figure 14. Average air velocity passing through Return 2 over time with truck moving towards the inlet.

2.4 Scenario 4: Modeling with truck movement in the mine with fan
In this scenario, a combination of the above two scenarios were modeled: the truck’s movement
and a fan running near Return 1 (Figure 7), which is a combination of the above two scenarios.
Again, modeling was conducted considering the truck movement towards and away from the
inlet. Figure 17 shows the velocity profile and streamlines at the initial time step. It can be seen

77
from the figure that the fan is pushing fresh air towards the working area near Return 2.
Figure 18 shows the average air velocity passing through Return 2 over time. The profile looks
similar to the model where no fan was present; however, due to fan, the average air velocity is
higher. The decrease in the air velocity was observed after the truck passed the fan. Figure 19 pre­
sents the velocity profile and streamlines when the truck is crossing paths with the fan flow. This
time is around 10 seconds. The truck movement disrupted the flow of the fan towards the working
area, restricting air coming from the inlet. This led to a decrease in the airflow at Return 2.

Figure 15. Velocity profile and streamlines at initial time step with truck moving away from the inlet.

Figure 16. Average air velocity passing through Return 2 over time with truck moving away from the inlet.

In the second model the truck was moving away from the inlet with a fan running near
Return 1. All the model settings were the same as in the scenario with truck moving and no fan
in the model domain. This model was run for 39 seconds. Figure 20 shows the velocity profile
and streamlines at initial time step. It can be seen from the figure that the flow profile near the
main entry and the truck looks similar to the velocity profile when there is no fan in the model
(Figure 15). The main change in this case can be seen as due to running of the fan near Return

78
1, which is pushing the fresh air towards the working area with the air passing through Return
2. This will help in the removal of pollutants generated at the working area. The model ran
without an issue until it crossed paths with the fan. The model crashed when the truck crossed
the path with a fan just after 27 seconds. Currently, this issue has not been resolved. In order to
look at the airflow with movement of the truck, the model was run by changing the position of

Figure 17. Velocity profile and streamlines at initial time step with fan and truck moving towards the inlet.

the truck just past the fan to see how the airflow changes. Figure 21 shows the velocity profile
and streamlines at t = 27 seconds. Figure 22 presents the average air velocity passing through
Return 2 over time. There is a brake in the line plot at 27 seconds to show the discontinuity
due to the model crash. However, it can be seen that after restarting the model with a truck
positioned just after the fan that it follows a similar trend as in Figure 15. The main difference
found out to be the average flow velocity due to the operating fan.

Figure 18. Average air velocity passing through Return 2 over time with fan near Return 1 and truck
moving towards the inlet.

79
Figure 19. Velocity profile and streamlines at t= 10 seconds with fan and truck moving towards the inlet.

Figure 20. Velocity profile and streamlines at initial time step with fan and truck moving towards the inlet.

3 SUMMARY

The ventilation of large-opening underground stone mines is challenging due to the larger
dimension of the mines. Ventilation issues can be resolved by use of the proper placement of
a fan in combination with a truck moving underground. We considered different scenarios
such as fan placement in the mine, movement of a truck, and a combination of fan and truck
movement in the mine to model. The CFD modeling results presented in this paper show that
placement of a fan can both positively and negatively affect the ventilation of fresh air near the
working area of the mine. Simulation results from movement of a truck has also some impact
on the airflow pattern in the mine affecting the ventilation; however, the movement of a truck
against the airflow affects the airflow more than the truck moving in the same direction of the
airflow. Modeling results from the combination of a fan and truck movement yielded similar
result as with movement of a truck alone, and the only difference was the magnitude of the
velocity due to the presence of a fan in the model domain. Finally, these are only the 2D

80
Figure 21. Velocity profile and streamlines at t = 27 seconds with fan and truck moving away from the
inlet.

Figure 22. Average air velocity passing through Return 2 over time with fan near Return 1 and truck
moving away from the inlet.

modeling results, which does not provide the airflow pattern in the third dimension, and a 3D
modeling is needed to fully understand the airflow pattern in large-opening stone mines.

DISCLAIMER

The findings and conclusions in this report are those of the author(s) and do not necessarily
represent the official position of the National Institute for Occupational Safety and Health,
Centers for Disease Control and Prevention. Mention of any company or product does not
constitute endorsement by NIOSH.

81
REFERENCES

Bhargava, R., Tukkaraja, P., Adhikari, A., Sridharan, S. J. & Vytla, V. V. S. (2021) Airflow characteristic
curves for a mature block cave mine. In Tukkaraja, P. (Ed. Proceedings of the 18th North American
Mine Ventilation Symposium. Rapid City, SD, CRC Press.
Gendrue, N., Liu, S., Bhattacharyya, S. & Spellman, C. (2021) Field survey of mine ventilation system
for large opening underground mines: Pressure, relative humidity, and temperature. In Tukkaraja, P.
(Ed. Proceedings of the 18th North American Mine Ventilation Symposium. Rapid City, SD, CRC
Press.
Grau III, R. H. & Krog, R. B. (2009) Ventilating large opening mines. Journal of the Mine Ventilation
Society of South Africa, 62, 8–14.
Grau III, R. H., Krog, R. B. & Robertson, S. B. (2006) Maximizing The Ventilation Of Large-Opening
Mines. In JM, M. & RV., R. (Eds.) Proceedings of the 11th U.S./North American Mine Ventilation
Symposium. University Park, PA, Taylor & Francis Group.
Grau III, R. H. & Meighen, G. M. (2006) Novel stopping designs for large-opening metal/nonmetal
mines. In JM, M. & RV., R. (Eds.) Proceedings of 11th U.S./North American Mine Ventilation Sympo­
sium. University Park, PA, Taylor & Francis Group.
Grau III, R. H., Robertson, S. B., Krog, R. B., Chekan, G. J. & Mucho, T. P. (2004) Raising the bar of
ventilation for large-opening stone mines. In Bandopadhyay, S. & Ganguli, R. (Eds.) Proceedings of
the 10th US/North American Mine Ventilation System. Anchorage, AK.
Grau III, R. H., Robertson, S. B., Mucho, T. P., Garcia, F. & Smith, A. C. (2002) NIOSH ventilation
research addressing diesel emissions and other air quality issues in nonmetal mines. SME Annual
Meeting. Phoenix, AZ.
Krog, R. B., Grau III, R. H., Mucho, T. P. & Robertson, S. B. (2004) Ventilation planning layouts for
large opening mines. SME Annual Meeting. Denver, CO.
Kumar, A. R., Henderson, K. M. & Schafrik, S. (2021) Scale modeling, PIV, and LES of blowing type
airflow in a deep cut continuous coal mining section. In Tukkaraja, P. (Ed. Proceedings of the 18th
North American Mine Ventilation Symposium. Rapid City, SD, CRC Press.
Morla, R., Karekal, S. & Godbole, A. (2021) Transient-flow modelling of DPM dispersion in unventil­
ated dead-end crosscuts and control strategy using curtain. In Tukkaraja, P. (Ed. Proceedings of the
18th North American Mine Ventilation Symposium. Rapid City, SD, CRC Press.
NMA (2020) Mine Safety & Health Administration’s Number of Coal and Non-fuel Mineral Operations
in the U.S.
Raj, K. V. & Bandopadhyay, S. (2017) CFD Modeling Of Cloud Cover for Pollutants Dispersion in
Deep Open-Pit Mines Under Arctic Air Inversion. SME Annual Meeting. Denver, CO, SME.
Raj, K. V., Bandopadhyay, S. & Ramani, R. V. (2015) Turbulent Models for Pollutant Transport in
Openpit Mines under Stable Boundary Layer. Transactions of Society for Mining, Metallurgy, and
Exploration, Inc, 338, 476–486.
Watkins, E. & Gangrade, V. (2022) Optimization of auxiliary fan placement for large-opening under­
ground stone mines. SME Annual Meeting. Salt Lake City, UT.

82
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

An investigation of shock loss factors at ventilation raise junctions in

underground hard rock mines using computational fluid dynamics

H. Zhang, L.K. Falk & C. Allen

Vale Canada Limited, Sudbury, ON, Canada

ABSTRACT: Shock loss occurs when there is a change in the area of a drift or airflow direc­
tion. These shock losses accumulate throughout the ventilation system and will increase the
mine resistance as underground development deepens or becomes more complex. Hence, shock
losses should be carefully considered when establishing the resistance through a series of
branches as they can be a significant contributor to the estimated mine resistance. Using com­
puter software to simulate a ventilation system is common but applying reasonable shock loss
factors in the software is not a simple or direct process. Determining shock loss values includes
the consideration of historical measurements, application of factors and formulas from litera­
ture based on the geometry and complexity of junctions, and the geometry variation at different
locations. For example, the shock loss at the bottom of a ventilation raise can be comprised of
a 90° bend and a sudden area change. This paper presents an approach for estimating shock
loss factors near raise junctions using computational fluid dynamics. Current and future work
will also be compared with field measurements. Results from this approach are compared with
software defaults as well as classical techniques in published literature and standards.

1 INTRODUCTION

The ventilation system is one of the most critical systems in underground mines. It provides
fresh air to personnel and clears contaminants out of mines. However, the efficacy and efficiency
of the ventilation system can be significantly impacted by the flow phenomena and energy losses
at junctions where there are changes in geometry or airflow direction. As these “shock losses”
accumulate through a ventilation system, they can impart a negative impact on fan pressure
requirements and energy consumption, as well reduced air quality and potential safety concerns.
Researchers (Jade & Sastry, 2008; Tien, 1987; Purushotham et al., 2010) have conducted labora­
tory studies to estimate shock loss factors and explore the factors (e.g., geometry, wall roughness)
that contribute to shock losses. Although there is plenty of field data (e.g., pressure data collected
using trailing hose and manometer) in published literature, it does not contain accompanying
information about the junction geometry. Therefore, one cannot easily establish a catalogue of
shock losses for different cases. The motivation to take these measurements comes from available
textbooks (e.g., (McPherson, 2009)) and publications that do not have practically relevant shock
loss factors considering the combined effect from turns and sudden expansion or sudden contrac­
tion. However, in operating mines, ventilation engineers and specialists encounter these questions
regularly. Hence, it is necessary to collect field data and calibrate models (e.g., computational fluid
dynamics (CFD) models, network models) to better predict shock loss factors and consequently
achieve higher fidelity ventilation designs and fan selections.
This paper presents two experiments conducted in two operating underground hard-rock
mines in Canada. This paper also presents a proposed method for isolating the shock loss
from overall pressure loss data in a CFD analysis. Three-dimensional (3D) CFD models were
created to represent the geometry of the experiment locations, and the CFD results are

DOI: 10.1201/9781003429241-9

83
compared with the field measurements. The findings of this paper will be of interest to mine
ventilation engineers and researchers who seek to conduct optimization and improve the effi­
ciency of underground mine ventilation systems.

2 EXPERIMENT DETAILS

Two sets of field measurements were conducted at two different mines. The data from each
mine is summarized below.

2.1 Experiment 1
Experiment 1 was conducted in a fresh air transfer network as shown in Figure 1(a) and 1(b). Two
horizontal drifts are connected by a 4.27 m (14 ft)-diameter, 288 m (945 ft)-long unsupported
round raise. The intent of the experiment was to record trailing hose pressure readings every
30.48 m (100 ft) down the raise to determine the raise inlet and outlet shock loss and the friction
loss in the raise. Data was collected using 365.76 m (1200 ft) of 0.0048 m (3/16 inch) inside diam­
eter vinyl tubing and a Model 5825 (TSI Incorporated, 2022) DP-Calc MicroManometer. Unfor­
tunately, the hose hung up within the first ~60.96 m (200 ft) rendering the efforts unsuccessful.
Additional data collected included: temperature, relative humidity, and barometric pressure (using
a Kestrel environmental meter (Nielsen-Kellerman Co., 2020)) in both the upper and lower trans­
fer drifts, refer to Table 1. The flow rate through the fresh air transfer was established with a pitot
tube survey of the two booster fans at the bottom level; the result was 216.2 m3/s (460 kcfm). The
pitot tube survey data are not presented in this paper due to the paper’s length limitation. Two
CFD models were created for Experiment 1 to investigate the shock loss that occurs entering and
exiting the raise. The classical calculation of the shock loss based on the field data is shown in
Table 2.

Figure 1. 3D view of Experiment 1.

Table 1. Data collected in Experiment 1.

Dry-bulb Wet-bulb Relative Barometric Air density
Location temperature (°C) temperature (°C) humidity (%) pressure (Pa) (kg/m3)

Upper level 31.4 22.7 48.4 111,500 1.27

Lower level 34.2 24.8 44.3 113,820 1.28

84
Table 2. The classical calculation to subtract raise friction loss and arrive at an estimated value of the
combined shock loss (i.e., 90° corner and sudden contraction) into the raise as shown in Figure 1(a).
Length of hose in the Measured pressure Friction loss1 Shock loss2 Velocity Shock
raise (m) loss (Pa) (Pa) (Pa) pressure3 (Pa) factor4

30.48 154.13 34.74 119.39 61.54 1.94

60.96 165.34 69.49 95.85 1.56
91.44 202.94 104.23 98.70 1.60

Notes:
1
The friction loss was calculated using a friction factor of 0.0053 kg/m3 at local air density (McPherson,
2009) based on Equation (1).
2
The shock loss is the difference between the measured pressure loss and the friction loss
3
Velocity pressure was calculated using velocity at the inlet (Figure 1(a)) based on Equation (2).
4
Shock factor is calculated using shock loss divided by velocity pressure (or using Equation (3)).

The friction pressure drop is calculated using Atkinson’s Equation (Atkinson, 1854).

Where P is the friction pressure drop, k1.2 is the Atkinson friction factor at standard air
density (1.2 kg/m3), L is the length of a given airway, per is the perimeter, A is the cross-
sectional area, Q is air volume, ρ is local air density.
The velocity pressure is calculated using Equation (2) (McPherson, 2009).

Where Pv is velocity pressure, and u is air velocity.

The pressure caused by shock loss can be calculated using Equation (3) (McPherson, 2009).

Where Pshock is the pressure loss caused by shock losses, X is the shock loss factor.
Based on Table 2, the average shock factor referred to the inlet velocity pressure is 1.70,
a combined factor considering sudden contraction and a sharp turn.

2.2 Experiment 2
Experiment 2 was conducted on two levels, as shown in Figure 2. There are two 1.57 m (5.17 ft)-
diameter, 41.16 m (135 ft) long, unsupported round raises between the levels. A 60.96 m (200 ft)-
long trailing hose was released from the upper level in both raises. Efforts to characterize internal
raise losses using trailing hose measurements were successful in this experiment because of the
shorter raise length and better planning based on the lessons learned from Experiment 1. Meas­
ured temperature, relative humidity, and barometric pressure data are presented in Table 3. The
traverse method with a Testo 417 vane anemometer (Testo SE & Co. KGaA, 2023) was used to
measure the average air velocity in the upper and lower drifts. Due to safety constraints associated
with working near open holes, individual raise flow rates were not measured; the measured total
flow has been assumed to split evenly into the two raises. Air velocity in the drifts and raises was
relatively low at ~1.5 m/s and ~6 m/s, respectively. Air volume measurements in Experiment 2 are
shown in Table 4. Pressure data collected in the field for raise 1 and raise 2 are shown in Tables 5
and 7, respectively. The calculated shock and friction losses for raise 1 and raise 2 are shown in
Tables 6 and 8, respectively. One CFD model was created for Experiment 2 to investigate the over­
all shock loss that occurred in the ventilation system.

85
Figure 2. 3D view of Experiment 2 (dimensions are in meters).

Table 3. Data collected in Experiment 2.

Dry-bulb Wet-bulb Relative Barometric Air
Location temperature (°C) temperature (°C) humidity (%) pressure (Pa) density (kg/m3)

Upper level 24.50 18.80 56.80 113,970 1.33

Lower level 25.40 19.60 56.50 114,350 1.33

Table 4. Air volumes measured in Experiment 2.

Location Inlet Outlet 1 Outlet 2

Air volume (m3/s) 32.74 23.23 9.52

Table 5. Pressure loss measurements taken and calculated for raise 1 in Experiment 2.
Differential static Calculated Calculated Calculated
pressure (measured in velocity pressure2 static pressure3 total pressure4
Sample location the field1) (Pa) (Pa) (Pa) (Pa)

Upper drift 0 0.18 33.37 33.55

6.1 m into raise -19.17 23.60 14.19 37.796
12.2 m into raise -19.92 23.60 13.45 37.046
18.3 m into raise -21.17 23.60 12.20 35.806
24.4 m into raise -28.88 23.60 4.48 28.08
30.5 m into raise -28.64 23.60 4.73 28.33
36.6 m into raise -29.13 23.60 4.23 27.83
42.7 m into raise -30.635 23.60 2.74 26.33
Lower drift -33.37 0.18 0 0.18

Notes:
1
Each sample was taken every 6.1 m (20 ft). Each sample data represents a 10-second average pressure
reading. A hose was dropped from the upper level to the lower level and retrieved from the lower level.
2
The velocity pressure is calculated based on 11.61 m3/s of air in the raise. It is assumed the raise has
a 3% overbreak (1.57 m in diameter as shown in Figure 2 while the design diameter is 1.52 m).
3
The static pressure is calculated based on the differential static pressure in column 2.
4
The total pressure is calculated as the sum of the velocity pressure and the static pressure.
5
This measurement was taken when the hose first entered into the raise from the lower level.
6
These total pressure data are outliers because they are greater than the upper drift total pressure.
Hence, they are not used to calculate the results shown in Table 6.

86
 Table 6. Pressure loss results calculated for raise 1 based on the field data shown in Table 5.
Inlet shock loss (Pa) Raise friction loss (Pa) Outlet shock loss (Pa)
1 2
5.47 4.19 26.153

Notes:
1
5.47 = 33.55-28.08. This result represents the inlet shock loss and a small friction loss (along
24.4 m in the raise) because of the field data outliers collected from 6.1 m to 18.3 m in the
raise.
2
4.19 = 43.95x(28.08-26.33)/(42.7-24.4)
3
26.15 = 26.33 – 0.18

Table 7. Pressure loss measurements taken and calculated for raise 2 in Experiment 2.
Differential static Calculated Calculated Calculated
pressure (measured in velocity pressure static pressure total pressure
Sample location the field) (Pa) (Pa) (Pa) (Pa)

Upper drift 0 0.18 33.62 33.80

6.1 m into raise -24.15 23.60 9.46 33.06
12.2 m into raise -25.40 23.60 8.22 31.81
18.3 m into raise -28.14 23.60 5.48 29.07
24.4 m into raise -29.881 23.60 3.74 27.33
30.5 m into raise -34.86 23.60 -1.25 22.35
36.6 m into raise -30.63 23.60 2.99 26.58
41.9 m into raise -29.13 23.60 4.48 28.08
Lower drift -33.62 0.18 0 0.18

Notes:
1
This measurement was taken when the hose first entered into the raise from the lower level.

Table 8. Pressure loss results calculated for raise 2 based on the field data shown in Table 7.
Inlet shock loss (Pa) Raise friction loss (Pa) Outlet shock loss (Pa)

0.741 5.732 27.903

Notes:
1
0.74 = 33.8-33.06
2
5.73 = 33.06-27.33
3
27.9 = 28.08 – 0.18

Based on classical shock factor equations for sudden contraction and sudden enlargement
(McPherson, 2009), the shock factors for the upper bend (from the upper drift to the raise) and
the lower bend (from the raise to the lower drift) are calculated to be 0.42 and 0.84, respectively.
The shock losses for the upper and lower bends are calculated to be 9.95 Pa and 19.89 Pa, respect­
ively. Compared to Tables 6 and 8, the lower bend shock losses (26.15 Pa and 27.9 Pa) calculated
based on the field pressure data are consistently higher than that (19.89 Pa) calculated using the
classical equations. In contrast, the upper bend shock losses (5.47 Pa and 0.74 Pa) calculated
based on the field pressure data are less than that (9.95 Pa) calculated using the classical equations.
It is apparent that the difference between the pressure losses calculated from the field data and the
classical equations is less at the lower bend than at the upper bend. The low air velocity in the
drift, actual raise diameter (e.g., overbreak), measurement uncertainty in collecting the field data
all contribute to the mentioned differences.

3 3D CFD MODEL SETUP

3D CFD models were created for each experiment matching the geometry and configuration
shown in Figures 1 and 2.

87
3.1 Assumptions
The geometry for the two experiments was simplified, including no arch on the back, no over­
break on walls and back, and idealized overbreak (i.e. smooth fillets) at raise intersections. In
addition, uniform flow at the inlet was applied.

3.2 Boundary conditions

The CFD modelling was conducted using Ansys Workbench 2022 R1 (ANSYS Inc., 2022b).
The model geometry was built in SpaceClaim (Ansys Inc., 2021), and the mesh was gener­
ated using Meshing (ANSYS Inc., 2010). Results were obtained using the CFX solver. The
solver high resolution advection scheme was used with the SST turbulence model (ANSYS
Inc., 2009) to compute steady-state isothermal analysis results for the flow domain(s) of
interest. An average of 150 iterations were required to demonstrate that iterative conver­
gence errors were not present and achieve convergence with root mean square (RMS)
residual values below 1.0e-4. The results presented herein were obtained using a mixed hexa­
hedral and tetrahedral mesh. Boundary layer inflation was applied following Ansys docu­
mentation for the SST turbulence model. Nominal element sizing of approximately 1/15th of
characteristic model geometry length scales (e.g. 0.1 m element sizing for a 1.525 m diameter
raise) was demonstrated to achieve results with minimal grid dependency. CFD models for
Experiment 1 shown in Figure 1(a) and 1(b) have 159,023 and 171,572 nodes, respectively.
CFD models for Experiment 2 ranged in mesh size from 94,892 (coarse mesh) to 210,903
(medium mesh) nodes. Mass flow inlet, pressure outlet, and no-slip wall boundary condi­
tions were applied for both experiments’ CFD models. A wall roughness height of
0.0175 m (White, 2015) is estimated for the drifts and raises. Results were analyzed in CFD-
Post (ANSYS Inc., 2022).

3.3 A CFD modelling approach for determining the shock loss and shock loss factor
One modelling approach to establish CFD estimated pressure losses associated with a shock
loss is to use three individual CFD models. For example, Figure 3 illustrates the CFD models
used to estimate a sudden expansion shock loss factor.

Figure 3. CFD models (plan view) required to estimate a sudden expansion shock loss factor.

In Figure 3, Model 1 and Model 2 are used to establish the pressure loss due to friction in
each geometry segment. Model 3 establishes the total pressure loss through the entire geom­
etry; the shock loss associated with the expansion is determined using Equation (4).

Where dPshock is the pressure drop caused by shock loss, dP3 is the total pressure difference
between the inlet and outlet in Model 3 in Figure 3, dP1 and dP2 are the total pressure differ­
ence between the inlet and outlet in Model 1 and Model 2, respectively.

88
4 RESULTS

CFD results for experiments 1 and 2 are shown in Tables 9 and 10, respectively. Pressure
result comparisons between the field data and the CFD model are shown in Table 11.

Table 9. CFD results for Experiment 1.

CFD Models Pti1 (Pa) Pto2 (Pa) dPt3 (Pa) Friction Loss4 (Pa) Shock Loss5 (Pa)

Figure 1(a) Both raise and drift 241.59 149.90 91.69 65.60
Raise only 176.36 150.28 26.09 26.09
Figure 1(b) Both raise and drift 206.56 11.01 195.54 167.67
Drift only 11.64 10.08 1.57 1.79
Raise only 176.36 150.28 26.09 26.09

Notes:
1
Area Average total pressure at the inlet based on the CFD model
2
Area Average total pressure at the outlet based on the CFD model
3
Pressure difference between the inlet and outlet
4
Friction loss is the total loss along the drift or raise
5
Shock loss is calculated using Equation (4)

Table 10. CFD results for Experiment 2.

CFD Models Pti (Pa) Pto (Pa) dPt (Pa) Friction Loss (Pa) Shock Loss (Pa)

All the raises and drifts 0 -33.67 33.67 18.88

1.57 m (diameter) Raise 31.82 29.72 2.10 14.801

Notes:
1
The friction loss in the raise is calculated based on the linear relationship between the fully-developed
airflow region (velocity contours apart from each other by 6.1 m shown in Figure 4) and the average
length of the two raises (42.94 m). So 14.80 is the result of 42.94x2.10/6.10.

Figure 4. Contour of velocity magnitude in the 1.57 m (5.17 ft)-diameter raise in Experiment 2.

The CFD results are all less than those measured in the field for Experiment 1. A reason for
the considerable disagreement between the results is the field data is point data, whereas the
CFD results are the area-average pressure at the monitoring planes. In addition, it is import­
ant to pay attention when comparing the CFD-calculated pressure loss with the field-
measured value. The CFD results are useful to offer a relative measure of the geometry
change impacts, but it worth noting that the geometry modelled in the CFD model is not

89
Table 11. Pressure result comparisons (between CFD and field data) for experiments 1 and 2.
Experiment 1 (Figure 1(a)) Experiment 2 Raise 2 (Table 7)
Field data collected at Difference Field Difference
Pressure CFD results 60.96 m below the collar (Pa) CFD results data (Pa)

Total 91.69 165.34 -73.65 33.67 33.62 0.05

Friction 26.09 69.49 -43.4 14.80 5.73 9.07
loss
Shock 65.60 95.85 -30.25 18.88 28.64 -9.76
loss

identical (e.g., arched drifts, uneven level) to the actual geometry and the numerical models
used in CFD representing the complex physics of turbulent airflow can also contribute to dis­
crepancies between the results in the CFD and field measurements. As well, the raise condition
(e.g., smooth wall, broken wall) is unknown, and the hose continuously moved because of the
high air velocity in the raise while the pressure was taken. These difficulties in taking accurate
pressure readings affected the field data accuracy.
From Table 11, the CFD total pressure results agree with that measured in the field for
Experiment 2. Similar to Experiment 1, the shock loss result from the CFD model is less than
that from the field. The field data indicates less friction loss in the raise compared to that in
the CFD model.

5 CONCLUSIONS AND FUTURE WORK

In this paper, two field experiments were conducted at two different mine sites. The two
experiments used the trailing hose method to measure the pressure loss. An average shock
factor referred to the inlet is 1.70 in Experiment 1. Results from the CFD model of Experi­
ment 1 do not agree well with the field data. Efforts to characterize internal raise losses using
trailing hose measurements were unsuccessful in Experiment 1 because the hose hung up in
the raise. However, it was successful in Experiment 2. There is a good agreement on the total
pressure loss between the results from the CFD model and the field data for Experiment 2
despite more complexity in the geometry of Experiment 2.
The authors identified the following factors that contribute to the total pressure losses:
• The overbreak of the raises during construction
• The connection type (e.g., sharp 90° turn, fillet) between the raise and horizontal drift
• Accuracy of the instrument (e.g., vane anemometers, manometers)
• Methods of air velocity measurement (e.g., fixed points, traverse)
• Wall roughness height
The authors also found that the Ansys CFX and Fluent solvers could generate differ­
ent results based on a preliminary study on the total pressure losses obtained with the
same mesh, near identical solver settings (understanding that some settings are fundamen­
tally different). A literature review (Acharya, 2016) has identified discrepancies between
Fluent and CFX results even when the model input parameters are as closely matched as
possible between the two solvers. Ongoing research into this topic and how it may
impact the use of CFD methods to estimate shock losses in mine ventilation systems will
be conducted by the authors.
In the future, a mesh independence study with various mesh elements or nodes will be con­
ducted to ensure the CFD results are independent of the mesh quantity and quality. In add­
ition, more field data will be collected at geometrically preferred (e.g., one raise between the
two levels instead of two raises) locations at different mines.

90
ACKNOWLEDGEMENTS

The authors want to thank Vale Canada Ltd. for the support of this work. The authors are
also grateful to their colleagues Fletcher Friesen, Paul Aho, Daniel Caton, Allen Jonas, Hugo
Ferrari, Stefano Girardo, Scott MacMillan, Peter McGirr, Brandan Randell, Deon Jenkins,
and Matthew Cheeseman for helping set up the field experiments and take measurements.

REFERENCES

Acharya, R. 2016. Investigation of differences in Ansys solvers CFX and fluent. Royal Institute of Technol­
ogy, KTH, Stockholm, Sweden.
Ansys Inc. 2021. Discovery SpaceClaim. Canonsburg, PA, USA.
ANSYS Inc. 2022. Ansys CFD-Post User’s Guide. Canonsburg, PA, USA.
ANSYS Inc. 2009. ANSYS CFX-Solver Theory Guide. Canonsburg, PA, USA.
ANSYS Inc. 2010. ANSYS Meshing User’s Guide. Canonsburg, PA, USA.
Atkinson, J.J. 1854. On the Theory of the Ventilation of Mines. North of England Institute of Mining
Engineers, 3: 118.
Jade, R.K. & Sastry, B.S. 2008. An experimental and numerical study of two-way splits and junctions in
mine airways. In 12th US/North American mine ventilation symposium. Reno, Nevada, USA.
McPherson, M.J. 2009. Subsurface Ventilation and Environmental Engineering. Fresno, CA, USA: Mine
Ventilation Services, Inc.
Purushotham, T., Sastry, B.S. & Samanta, B. 2010. Estimation of shock loss factors at shaft bottom junc­
tion using computational fluid dynamics and scale model studies. CIM Journal, 1(2): 130–139.
Testo SE & Co. KGaA. 2023. Instruction manual Testo 417 vane anemometer. West Chester, PA, USA.
Tien, J.C. 1987. Shock losses around underground overcasts. In J. Mutmansky, ed. 3rd Mine Ventilation
Symposium. State College, PA, USA.
TSI Incorporated. 2022. DP-CALC MicroManometer Model 5825 Operation and Service Manual. Shore­
view, Minnesota, USA.
White, F.M. 2015. Fluid Mechanics. 8th ed. New York, NY, USA: McGraw-Hill Education.

91
Diesel particulate control
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Comparing diesel and GDiesel® exhaust exposures

in an underground mining laboratory

R.J. Reed & J.L. Burgess

Division of Community, Environment, and Policy; Mel & Enid Zuckerman College of Public Health,
University of Arizona, Tucson, Arizona, USA

E.A. Lutz
School of Mining and Mineral Resources, College of Engineering, University of Arizona, Tucson, Arizona, USA

ABSTRACT: Diesel exhaust is a known carcinogen and has been linked to several negative
health outcomes, yet it is commonly used in industrial settings such as underground mining. Alter­
native fuels such as GDiesel® (GD), a natural gas/diesel blend, have the potential to reduce health
exposures and associated health effects. Mirroring a previous study in which use of GD in a 2005
Wagner load-haul-dump (LHD) vehicle with oxidation catalyst demonstrated significant reduc­
tions in diesel exhaust exposures, operator-location and area exposure samples were collected in
an underground mining laboratory with diesel (D) and then GD fuel while operating a JCI LHD.
Analytes of interest included total and respirable diesel particulate matter (tDPM and rDPM,
respectively), total and respirable elemental and organic carbon (tEC, rEC, tOC, rOC, respect­
ively), as well as formaldehyde (CH2O), nitric oxide (NO), and nitrogen dioxide (NO2). Use of
GD resulted in non-significant reductions in median rOC, tDPM, tOC, and NO2 concentrations,
non-significant increases in rDPM, rEC, tEC, and NO, and identical median CH2O concentra­
tions. A significant decrease in NO2 exposure concentrations (p=0.012) and increase tEC and
rEC exposure concentrations (p=0.023 and p=0.024, respectively) were observed. After controlling
for environmental confounders there was no difference observed in analyte concentrations
between the two fuels. Further research is needed to determine whether GD alone can significantly
reduce human exposures across vehicles and pollution configuration types.

1 INTRODUCTION

1.1 Background
Diesel fuel (D) is used in a variety of applications for several industries, including transportation,
agriculture, railroads, construction, and mining. In 2012 the International Agency for Research
on Cancer designated D exhaust as a Group 1 carcinogen in humans (WHO 2012). Large longitu­
dinal studies performed with underground miners have linked their exposure to D with increased
risk of lung cancer (Attfield et al 2012, Silverman et al 2012). Exposure to D engine emissions is
associated with adverse health effects including: chronic bronchitis, respiratory tract infections,
asthma exacerbation, and increased cardiovascular morbidity and mortality (Bonauto et al 2007,
Culp et al 2011, Kagawa 2002, Taylor 2006). Given the nature of D exposures in the mining
industry and the many associated health effects (Attfield et al 2012, Silverman et al 2012, Sydbom
et al 2001), reducing engine emissions has become an industrial hygiene priority. In the US, there
are nearly 14,000 mine sites that employ over 210,000 individuals in the workforce (BLS 2017,
MSHA 2014). Both underground and surface mines predominantly use diesel-powered vehicles at
their operations to perform tasks such as drilling, mucking, and hauling. Miners are regularly
exposed to D during their work, especially in areas where there is limited ventilation.

DOI: 10.1201/9781003429241-10

95
1.2 Regulatory compliance
In 2008 the Mine Safety and Health Administration (MSHA) published updated standards for
the allowable concentration of respirable (<1.0 µm with impactor) diesel particulate matter
(rDPM) in an effort to control D exhaust exposures. The administration has also recently pub­
lished a request for information (RFI) regarding diesel exhaust exposures and controls (MSHA
2016). The permissible exposure limit (PEL) of 160 µg/m3 as total (combined inorganic and
organic) carbon (CFR 2016) is monitored through air sampling and processing per the National
Institute for Occupational Safety and Health (NIOSH) analytical method 5040. Despite the
installation of controls such as mine ventilation, the underground rDPM exposure standard is
frequently exceeded. Since 2000, MSHA has made detailed information regarding the occupa­
tional health exposure assessments performed during inspections available to the public. Accord­
ing to data since the new diesel exhaust standard, diesel particulate matter (DPM) had a higher
median exposure concentration-to-PEL (ECP) ratio (0.47) than any other airborne contaminant
(for which 100 or more samples were taken) except silver fumes (0.88). This ratio indicates that
mine operators continue to struggle to control D exhaust exposures, and that governmental,
national, and industrial organizations should focus on developing better control options.

1.3 Biodiesel
Given the health hazards associated with and MSHA’s focus on D exposures, mine operators
seek effective control methods for rDPM. Some have explored the use of biodiesel (B) fuel
blends. Studies comparing D to B emissions and exposures report a decrease in total carbon
output but an increase in organic carbon, aldehydes, and nitrogen dioxide (Turrio-Baldassarri
et al 2004, Knothe, Sharp, & Ryan 2006, Li & McLaughlin 2005, Bugarski et al 2010, He et al
2009, Bugarski et al 2003). However, other research demonstrated an increase in aldehyde,
nitrogen dioxide (NO2), and organic carbon fraction concentrations with use of B mixtures
(Bugarski et al 2010, Liu et al 2009). Despite the increasing usage of B, there is a lack of infor­
mation on the human health effects of exposure to these emissions. Recent studies also suggest
that exposure to B exhaust may be as or more toxic than D (Donoghue & Bates 2000, Bhavar­
aju et al 2013, Brak & Bates 2003, Bunger et al 2000, Armstrong et al 1998, Mehus et al 2015).

1.4 GDiesel ®
GDiesel® (GD), an EPA-approved natural gas/diesel fuel mixture, became commercially avail­
able in 2011 and provides a second alternative to D fuel. GD is prepared by combining diesel
with natural gas, using a proprietary charged-catalytic reaction. The end product is an American
Society for Testing and Materials (ASTM)-designated D fuel with purportedly reduced tailpipe
emissions of diesel particulate matter (DPM) and oxides of nitrogen (NOx) compounds (NDEP
2010, ARC 2014). Our independent, simulated study using a 2005 Wagner load-haul dump
(LHD) with a diesel oxidative catalyst (DOC) but no diesel particulate filter (DPF) at an under­
ground mining laboratory showed a 66% reduction in rDPM and significant reductions for all
other analytes except carbon monoxide (CO) associated with use of GD (Lutz et al 2015). LHDs
are commonly used in underground mine settings and often associated with high exposures to
D (Bugarski et al 2004, Bugarski et al 2003). While promising, this GD study represents a single
vehicle, study design, and pollution configuration. Other studies have shown that D is influenced
by a variety of factors including type of equipment, load, and pollution control equipment
(Bugarski et al 2010, Liu et al 2009, Bugarski et al 2004, Durbin et al 2000, Tsai et al 2011).

1.5 Objective
The objective of this pilot study was to evaluate the exhaust exposures of GD, as compared to
D, using a JCI LHD with Deutz engine and DOC but no DPF. We hypothesized that use of
GD fuel in the JCI LHD would result in significant reductions for all analytes, comparable to
those observed in our previous study.

96
2 METHODS

2.1 Previous work

Lutz et al (2015) conducted a pilot study at the University of Arizona San Xavier Underground
Mining Laboratory (SX), a non-operational hard rock mine at which laboratory work is per­
formed. Exposure to vehicle emissions was evaluated in the ‘decline,’ an unventilated portion of
the SX with sloping underground opening for rubber-tired vehicle access. Study participants and
research staff alternatively performed mucking activities with a University-owned 2005 Wagner
B10-203 LHD vehicle with open cabin and DOC and observed from the decline rib. The D and
GD fuels were loaded into separate fuel tanks, and the LHD was operated for approximately
one hour after switching to the GD tank before beginning testing. Personal and area exposure
samples to rDPM, rEC, rOC, tDPM, tEC, tOC, formaldehyde (CH2O), acetaldehyde, naphtha­
lene, NO, NO2, and CO were collected. For a more detailed description, please reference the pub­
lication (Lutz et al 2015). The work by Lutz et al (2015) was limited to a single vehicle and
pollution control configuration – factors that have been shown to influence both exhaust emis­
sions and exposures (Bugarski et al 2003, Bugarski et al 2010, Durbin et al 2000, Tsai et al 2011).

Figure 1. Depiction of the decline at the University of Arizona San Xavier Underground Mining
Laboratory.
Note: Image not to scale.

2.2 Fuels, equipment and location

For the current study, ultra-low sulfur D #2 was obtained from a regional distributor (Ari­
zona Petroleum, Tucson, AZ). The GD fuel #2 was purchased directly from the producer
(Advance Refining Concepts, LLC., Reno, NV).

97
 A JCI LHD with inline, 4-cylinder, 55 horsepower at 2300 RPM Deutz engine (motor
number 7534-462) with an Engine Control Systems Purimuffler (part number A17-0313) DOC,
but no DPF, was borrowed from an underground mine in the region. It also had an open cabin.
This study was conducted at the SX decline (see Figure 1 below). Similar to the previous study
(Lutz et al 2015), the decline’s gate (position B) was covered with a vinyl tarp and the metal door
(position D) entering the ‘adit’ level was closed at the start of each exposure session to minimize
airflow into and out of the decline. The mine fan was not operated during the study sessions.

2.3 Procedure
Research staff operated the vehicle for five consecutive days using D. Two 200-minute exposure
sessions were completed each day, with one operator-location and one area sample taken during
each session. In order to change from D to GD fuel, the LHD fuel tank was emptied using
a fuel pump, filled with GD, and then driven for approximately one hour before beginning the
first GD exposure session. The LHD was then operated for a total of four days with GD fuel.
Due to water constraints in the mine and potential for significant dust dispersion, mucking was
not performed. The LHD bucket was filled once at the beginning of the study and the equip­
ment simply driven back and forth on the 30-meter decline mucking path by the same operator.
The bucket was not emptied until the end of the study. Each day the vehicle was not allowed to
idle before the first session and, before beginning the second, the tarp was removed for 30 min­
utes to allow natural ventilation to completely disperse carbon monoxide. CO dispersion was
confirmed with an MSA 4X Altair gas monitor (Cranberry Township, PA).

2.4 Exposure assessment

Universal PCXR 8 (SKC West, Inc., Fullerton, CA) and Escort ELF (Zefon International,
Inc., Ocala, FL) air sampling pumps, along with associated sampling tubing and holders or
cassettes, were placed in pockets of or clipped to safety vests. Because the operator’s cabin
was small it was not possible for the operator to wear the vest. One was therefore placed in
the cabin within approximately 36 inches of the operator’s breathing zone. A second ‘area’
vest was placed on the rib wall at a height of approximately six feet (position C).
Personal integrated sample collection and analysis was performed in accordance with the
NIOSH manual of analytical methods (NMAM). Specifically, GS-1 Respirable Cyclones with
37 mm jeweled impactor and flow rate of 1.7 L/min (NIOSH 2003a) were used for rDPM sam­
pling. As rDPM is simply the sum of rEC and rOC, these were also obtained from the sample
data. A 37 mm open face quartz fiber filter with flow rate of 2.0 L/min (NIOSH 2003a) was util­
ized to sample tDPM, tEC and tOC. A tandem triethanolamine/oxidizer with flow rate 0.025 L/
min (NIOSH 1994) was used to collect NO and NO2 samples. Sorbent tubes containing silica gel
and a flow rate of 0.1 L/min (NIOSH 2003b) were used in formaldehyde sample collection. All
sampling media were produced by the same manufacturer (SKC West, Inc., Fullerton, CA). Pre-
and post-sampling calibration was performed using a Bios Drycal® Defender 520 calibrator
(Mesa Labs, Inc., Butler, NJ) at the SX Training Center. An independent AIHA-accredited indus­
trial hygiene laboratory performed laboratory analysis (Galson Laboratories, East Syracuse, NY).

2.5 Data analysis

Laboratory results were copied, pasted to, and then formatted in a database table. Reported con­
centrations were time-weighted over an 8-hour period (see Figure 1 below) to obtain the time-
weighted average (TWA8). Environmental data were averaged over the time period corresponding
to each exposure session and paired using database relationships. All analysis was performed
using STATA 12.0 (StataCorp, College Station, TX). Unadjusted differences in distribution were
measured using the Wilcoxon rank-sum test. Multiple linear regression was used to identify which
of the four environmental measures had the least potential to be a confounder. In every case
except for formaldehyde, for which temperature had the least potential, the confidence interval for
wind speed was most likely to contain zero. To adjust for environmental confounders, binary
logistic regression was then performed using fuel type as the output, with the TWA8 and three

98
remaining environmental measures as inputs. An alpha error threshold of 0.05 was used
throughout.

Using fences at 150% of the interquartile range, seven outliers were identified, by fuel type,
and removed from the D (rOC, NO2) and GD (rDPM, rEC, rOC, NO2) datasets. Statistical
analysis was performed with and without outliers. In three of the four datasets there were no
changes in conclusions observed when outliers were included versus not included. However,
the rEC sample set changed when outliers were included. The rEC results with and without
outliers are provided in the results section.

3 RESULTS

3.1 Overview
There were a total of 20 D and 16 GD samples collected. Variations in samples collected were due
to LHD equipment malfunction during two of the GD sampling days. For the unadjusted rank-
sum difference, median rOC, tDPM, tOC, and NO2 TWA8 exposure concentrations were reduced
with use of GD (see Figure 2 below), but only NO2 was significantly so (p=0.012). Use of GD
resulted in increased median concentrations of rDPM, rEC, tEC, and NO, though only tEC and
rEC (after removing outliers) were significantly greater (p=0.023 and p=0.024, respectively).

Figure 2. Median time-weighted average exposure concentrations by analyte. Note that Diesel (D) has
been set as a reference and GDiesel (GD) marker placement represents a respective percentage compared
to D. Abbreviations: respirable diesel particulate matter (rDPM), respirable elemental carbon (rEC), res­
pirable organic carbon (rOC), total diesel particulate matter (tDPM), total elemental carbon (tEC), total
organic carbon (tOC), formaldehyde (CH2O), nitric oxide (NO), and nitrogen dioxide (NO2).

99
Median formaldehyde concentrations were identical. For more detailed results, see Tables 1, 2
and 3 below. After controlling for confounders, the likelihood of a given TWA8 being associated
with either fuel was not statistically significant for any analytes.

Table 1. Descriptive statistics for gaseous analytes formaldehyde (CH2O), nitric oxide (NO), and nitro­
gen dioxide (NO2), by fuel. Abbreviations: diesel fuel (D), and GDiesel fuel (GD). All units in ppm.
Analyte CH2O NO NO2

Fuel D GD D GD D GD
N 19 16 20 16 20 16
Mean 0.035 0.033 1.946 2.976 0.345 0.159
SD 0.013 0.010 0.783 2.322 0.589 0.098
Median 0.030 0.030 1.725 1.868 0.186 0.120*
Q1 0.027 0.025 1.469 1.470 0.156 0.102
Q3 0.041 0.040 2.619 4.100 0.293 0.171
Min 0.020 0.020 0.413 1.000 0.088 0.076
Max 0.058 0.052 3.559 7.877 2.813 0.417

* Significantly different from D at p<0.05.

Table 2. Descriptive statistics for respirable particulates: respirable diesel particulate matter (rDPM),
respirable elemental carbon (rEC), and respirable organic carbon (rOC), by fuel. Abbreviations: Diesel
fuel (D), and GDiesel fuel (GD). All units in µg/m3.
Analyte rDPM rEC rOC

Fuel D GD D GD GDa D GD
N 20 15 20 15 14 20 15
Mean 274.9 305.2 115.6 137.6 144.4 159.3 167.8
SD 112.2 84.3 36.4 32.6 19.5 85.4 100.8
Median 250.4 283.4 113.4 140.9 143.8* 150.9 138.2
Q1 203.5 245.0 90.5 126.9 130.7 92.7 114.3
Q3 347.8 336.7 146.6 160.7 160.7 200.8 172.5
Min 111.9 217.8 50.8 41.4 115.3 53.9 93.5
Max 533.8 549.8 177.6 183.6 183.6 377.4 507.5
a
With outliers removed,
* significantly different from D at p<0.05.

Table 3. Descriptive statistics for total particulates: total diesel particulate matter (tDPM), total elemen­
tal carbon (tEC), and total organic carbon (tOC), by fuel. Abbreviations: Diesel fuel (D), and GDiesel fuel
(GD). All units in µg/m3.
Analyte tDPM tEC tOC

Fuel D GD D GD D GD
N 20 16 20 16 20 16
Mean 551.7 526.4 136.2 161.6 413.3 361.4
SD 141.8 75.4 34.0 20.0 120.6 52.8
Median 551.1 500.0 134.3 155.8* 405.9 343.6
Q1 464.1 466.4 118.4 149.1 330.4 321.4
Q3 620.5 566.7 159.4 176.3 484.1 396.7
Min 283.3 460.6 66.7 123.3 212.5 298.8
Max 843.3 683.3 195.5 195.5 690.0 469.8

* Significantly different from D at p<0.05.

100
4 DISCUSSION

4.1 Findings
This study demonstrated little statistical difference between D and GD exhaust exposures in
a JCI LHD with DOC at the SX mine. The results of this study are inconsistent with results
observed by Lutz et. al. (2015), while using the Wagner LHD with DOC. Previously rDPM,
rEC, rOC, tDPM, tEC, tOC, CH2O, NO, and NO2 were significantly reduced, whereas this
study observed a significant decrease in NO2 alone, with significant increases in rEC and tEC.
These findings are also inconsistent with those described by the manufacturer of GD (NDEP
2010, ARC 2014). However, the Nevada Division of Environmental Protection (2010)
reported that use of GD may cause small increases in total hydrocarbon emissions. This
research is consistent with previous work that has demonstrated variability in exposure reduc­
tions based on the use of different equipment and pollution control configurations (Bugarski
et al 2004, Bugarski et al 2003, Bugarski et al 2010, Durbin et al 2000, Tsai et al 2011).
Compared to regular D fuel, this pilot study demonstrated a reduction in NO2 and increase in
tEC and rEC with use of GD, with no change in rDPM or other analytes. Given that the MSHA
standard is based on rDPM exposures, this work makes it unclear whether use of GD would
likely increase compliance with the federal regulation. Despite MSHA’s focus on rDPM, other
components of D contribute to adverse health effects. Insufficient data are available to indicate
whether reducing one or some of the contaminant exposures from D influences observed health
effects.
Use of compressed natural gas in D engines has been shown to reduce particulate emissions,
polycyclic aromatic hydrocarbons, formaldehyde, and genotoxic activity, compared to D fuel
Fritz & Egbuonu 1992, Greenwood et al 1996, Ayala et al 2002, Turrio-Baldassarri et al
2006). Despite this study’s findings, a reduction in exposures with GD fuel would seem reason­
able given the apparently cleaner burning qualities of natural gas.
Differences observed between fuel types are likely a function of the fuels’ chemical compos­
ition, vehicle engine conditions, such as temperature of combustion and concentration of
oxygen present, and emission source controls, such as DOCs and DPFs. For example, differ­
ences in DPM could be due to more complete combustion from higher engine temperatures,
or impurities and hydrocarbon chain lengths found in each fuel. In addition, EC is released
directly from the incomplete combustion of fossil fuels while OC, an aggregate of hundreds of
individual compounds spanning a wide range of chemical and thermodynamic properties, is
formed by a variety of processes, suggesting perhaps more complete combustion with D in the
JCI LHD. Variation in NOx concentrations could be due to higher combustion temperatures,
as well as interaction with a DOC. Higher engine temperatures and increased oxygen content
may lead to conditions that increase the formation of short hydrocarbons, such as formalde­
hyde and acetaldehyde. Increased combustion temperature may therefore explain the decrease
in DPM but increase in aldehydes associated with B in some studies.

4.2 Limitations
The limitations of our pilot study include use of a single vehicle and single pollution control
configuration (DOC, no DPF). In addition, this study occurred at a naturally ventilated
underground laboratory, rather than an operational mine where other pieces of equipment
and active, forced ventilation are typically present. During two of the scheduled GD exposure
sessions the LHD would not start due to heating of the starter solenoid. Heating of the engine
over time, as well as the chosen order of fuel use, may have biased these results. While efforts
were taken to minimize airflow into and out of the mine, outdoor environmental conditions
such as wind speed and barometric pressure may have influenced our observations.

4.3 Conclusions
The current lack of information regarding the alternative fuel’s emissions suggests that further
research is needed to determine whether GD can significantly reduce occupational and

101
environmental exposures across vehicle and pollution configuration types. We recommend
research that utilizes a crossover design, targets several of the most commonly used diesel-
fueled vehicles in the mining industry and monitors exposures and health effects over an
extended period. The current study, coupled with previous research [28], demonstrates how
changes to the types of equipment used, driving conditions and environmental conditions
make predicting exposure risks from diesel exhaust extremely difficult.

REFERENCES

Advanced Refining Concepts, LLC (ARC). 2014. “GDiesel® - The World’s Best Diesel Fuel.” Available
at http://www.clearrefining.com/newfactsheet1211.pdf
Armstrong, L.E., Soto, J.A.H. and Hacker, F.T., 1999. Urinary indices during dehydration, exercise, and
rehydration. Occupational Health and Industrial Medicine, 2(40), p. 97.
Attfield, M.D., Schleiff, P.L., Lubin, J.H., Blair, A., Stewart, P.A., Vermeulen, R., Coble, J.B. and
Silverman, D.T., 2012. The diesel exhaust in miners study: a cohort mortality study with emphasis on
lung cancer. Journal of the National Cancer Institute, 104(11), pp. 869–883.
Ayala, A., Kado, N.Y., Okamoto, R.A., Holmén, B.A., Kuzmicky, P.A., Kobayashi, R. and Stiglitz, K.
E., 2002. Diesel and CNG heavy-duty transit bus emissions over multiple driving schedules: Regulated
pollutants and project overview. SAE Transactions, pp. 735–747.
Bhavaraju, L., Shannahan, J., William, A., McCormick, R., McGee, J., Kodavanti, U. and Madden, M.,
2014. Diesel and biodiesel exhaust particle effects on rat alveolar macrophages with in vitro exposure.
Chemosphere, 104, pp. 126–133.
BLS: U.S. Department of Labor. 2017. “Industries at a Glance: Mining (except Oil and Gas): NCAIS
212. U.S. Bureau of Labor Statistics. Available at http://www.bls.gov/iag/tgs/iag212.htm.
Bonauto, D., Anderson, R., Rauser, E. and Burke, B., 2007. Occupational heat illness in Washington
State, 1995–2005. American journal of industrial medicine, 50(12), pp. 940–950.
Brake, D.J. and Bates, G.P., 2003. Fluid losses and hydration status of industrial workers under thermal
stress working extended shifts. Occupational and environmental medicine, 60(2), pp. 90–96.
Bugarski, A., Schnakenberg, G., Noll, J., Mischler, S., Patts, L., Hummer, J., Vanderslice, S., Crum, M.
and Anderson, R., 2003. ‘The Effectiveness of Selected Technologies in Controlling Diesel Emissions
in an Underground Mine—Isolated Zone Study at Stillwater Mining Company’s Nye Mine. Draft
Report,”NIOSH, Pittsburgh Research Laboratory.
Bugarski, A., Mischler, S., Noll, J., Schnakenberg, G., Crum, M. and Anderson, R., 2004. ‘An Evaluation
of the Effects of Diesel Particulate Filter Systems on Air Quality and Personal Exposure of Miners at
Stillwater Mine Case Study: Production Zone. Report to M/NM Diesel Partnership, April, 1.
Bugarski, A.D., Cauda, E.G., Janisko, S.J., Hummer, J.A. and Patts, L.D., 2010. Aerosols emitted in
underground mine air by diesel engine fueled with biodiesel. Journal of the Air & Waste Management
Association, 60(2), pp. 237–244.
Bünger, J., Müller, M.M., Krahl, J., Baum, K., Weigel, A., Hallier, E. and Schulz, T.G., 2000. Mutageni­
city of diesel exhaust particles from two fossil and two plant oil fuels. Mutagenesis, 15(5), pp. 391–397.
Code of Federal Regulations (CFR). 2016. “Limit on exposure to diesel particulate matter.” Title 30.
Part 57.5060.
Culp, K., Tonelli, S., Ramey, S.L., Donham, K. and Fuortes, L., 2011. Preventing heat-related illness
among Hispanic farmworkers. AAOHN journal, 59(1), pp. 23–32.
Donoghue, A.M., and Bates, G.P. 2000. The risk of heat exhaustion at a deep underground metalliferous
mine in relation to body-mass index and predicted Vo(2)max. Occupational Medicine-Oxford 50(4):
259–263.
Durbin, T.D., Collins, J.R., Norbeck, J.M. and Smith, M.R., 2000. Effects of biodiesel, biodiesel blends,
and a synthetic diesel on emissions from light heavy-duty diesel vehicles. Environmental science &
technology, 34(3), pp. 349–355.
Fritz, S.G., and Egbuonu, R.I. 1992. EMISSIONS FROM HEAVY-DUTY TRUCKS CONVERTED
TO CNG. Journal of Engineering for Gas Turbines and Power-Transactions of the Asme 114(3):
561–567.
Greenwood, S.J., Coxon, J.E., Biddulph, T. and Bennett, J., 1996. An investigation to determine the
exhaust particulate size distributions for diesel, petrol, and compressed natural gas fuelled vehicles.
SAE transactions, pp. 640–646.
He, C., Ge, Y., Tan, J., You, K., Han, X., Wang, J., You, Q. and Shah, A.N., 2009. Comparison of
carbonyl compounds emissions from diesel engine fueled with biodiesel and diesel. Atmospheric Envir­
onment, 43(24), pp. 3657–3661.

102
Kagawa, J., 2002. Health effects of diesel exhaust emissions—a mixture of air pollutants of worldwide
concern. Toxicology, 181, pp. 349–353.
Knothe, G., Sharp, C.A. and Ryan, T.W., 2006. Exhaust emissions of biodiesel, petrodiesel, neat methyl
esters, and alkanes in a new technology engine. Energy & fuels, 20(1), pp. 403–408.
Li, Y.X., McLaughlin, N.B., Patterson, B.S. and Burtt, S.D., 2006. Fuel efficiency and exhaust emissions
for biodiesel blends in an agricultural tractor. Canadian biosystems engineering, 48(2).
Liu, Y.Y., Lin, T.C., Wang, Y.J. and Ho, W.L., 2009. Carbonyl compounds and toxicity assessments of
emissions from a diesel engine running on biodiesels. Journal of the Air & Waste Management Associ­
ation, 59(2), pp. 163–171.
Lutz, E.A., Reed, R.J., Lee, V.S. and Burgess, J.L., 2015. Occupational exposures to emissions from
combustion of diesel and alternative fuels in underground mining—a simulated pilot study. Journal of
occupational and environmental hygiene, 12(3), pp .18–25.
Mehus, A.A., Reed, R.J., Lee, V.S., Littau, S.R., Hu, C., Lutz, E.A. and Burgess, J.L., 2015. Comparison
of acute health effects from exposures to diesel and biodiesel fuel emissions. Journal of occupational
and environmental medicine, 57(7), p. 705.
Mine Safety and Health Administration (MSHA). 2014. “U.S. Department of Labor: “Mining Safety
and Health At a Glance.” Available at http://www.msha.gov/MSHAINFO/FactSheets/MSHAFCT10.
asp (accessed May 01, 2014).
Mine Safety and Health Administration (MSHA). 2016. Exposure of Underground Miners to Diesel
Exhaust: A Proposed Rule by the Mine Safety and Health Administration.
Nevada Division of Environmental Protection (NDEP). 2010. “Notice of Proposed Action.”
National Institute of Occupational Safety and Health (NIOSH). 1994. U.S. Department of Health and
Human Services. NITRIC OXIDE and NITROGEN DIOXIDE, 6014. NIOSH Manual of Analytical
Methods (NMAM), Fourth Edition. National Institute of Occupational Safety and Health.
National Institute of Occupational Safety and Health (NIOSH). 2003a. U.S. Department of Health and
Human Services. DIESEL PARTICULATE MATTER, 5040. NIOSH Manual of Analytical Methods
(NMAM), Fourth Edition. National Institute of Occupational Safety and Health.
National Institute of Occupational Safety and Health (NIOSH). 2003b. U.S. Department of Health and
Human Services. FORMALDEHYDE, 2016. NIOSH Manual of Analytical Methods (NMAM),
Fourth Edition. National Institute of Occupational Safety and Health.
Silverman, D. T., Samanic, C. M., Lubin, J. H., Blair, A. E., Stewart, P. A., Vermeulen, R., Coble, J. B.,
Rothman, N., Schleiff, P. L., Travis, W. D., Ziegler, R. G., Wacholder, S., & Attfield, M. D. 2012.
The Diesel Exhaust in Miners study: a nested case-control study of lung cancer and diesel exhaust.
Journal of the National Cancer Institute, 104(11),855–868. https://doi.org/10.1093/jnci/djs034.
Sydbom, A., Blomberg, A., Parnia, S., Stenfors, N., Sandström, T. and Dahlen, S.E., 2001. Health effects
of diesel exhaust emissions. European Respiratory Journal, 17(4), pp. 733–746.
Tsai, J.H., Chen, S.J., Huang, K.L., Lee, W.J., Kuo, W.C. and Lin, W.Y., 2011. Characteristics of par­
ticulate emissions from a diesel generator fueled with varying blends of biodiesel and fossil diesel.
Journal of Environmental Science and Health, Part A, 46(2), pp. 204–213.
Turrio-Baldassarri, L., Battistelli, C.L., Conti, L., Crebelli, R., De Berardis, B., Iamiceli, A.L.,
Gambino, M. and Iannaccone, S., 2004. Emission comparison of urban bus engine fueled with diesel
oil and ‘biodiesel’blend. Science of the Total Environment, 327(1-3), pp. 147–162.
Turrio-Baldassarri, L., Battistelli, C.L., Conti, L., Crebelli, R., De Berardis, B., Iamiceli, A.L.,
Gambino, M. and Iannaccone, S., 2006. Evaluation of emission toxicity of urban bus engines: Com­
pressed natural gas and comparison with liquid fuels. Science of the Total Environment, 355(1-3), pp.
64–77.
World Health Organization. 2012. “IARC: Diesel Engine Exhaust Carcinogenic.” International Agency
for Research on Cancer.

103
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Improvement of size-selective sampling of diesel aerosols in

underground mines

A.D. Bugarski, T.L. Barone, J.A. Hummer, T. Lee & S. Vanderslice

National Institute for Occupational Safety and Health, Pittsburgh Mining Research Division,
Pittsburgh, Pennsylvania, USA

S. Friend
National Institute for Occupational Safety and Health, Health Effects Laboratory Division,
Morgantown, West Virginia, USA

ABSTRACT: The quantification of exposures to diesel particulate matter (DPM) could be

adversely affected by the presence of micron-sized aerosols in DPM samples containing organic
and elemental carbon from other than combustion sources. Therefore, size classification plays
an important role in the collection of representative DPM samples. The results of this evalu­
ation were used to assess the potential for an improvement in the size separation of diesel and
micron size aerosols by using lower cutoff-size impactors in DPM sampling cassettes. Three
impactors with the cut-off diameters of 258, 466, and 716 nm were characterized in the labora­
tory using polystyrene divinylbenzene beads, and consequently the impactors were evaluated in
the outby area of an underground coal mine incorporating diesel-powered equipment. Scan­
ning electron microscope analysis of the field samples showed that reduction of the cut-off
diameters of the impactors resulted in lower contamination of the DPM samples with micron-
sized particulates, in this case primarily mineral dust. The total mass concentrations of organic
and elemental carbon in the samples were found to be inversely affected by the cut-off size of
the orifices, indicating some potential losses of DPM in the impactors with smaller cut-offs.

1 INTRODUCTION

Due to widespread use of various types of diesel-powered vehicles in hard to cost-effectively ven­
tilate confined spaces, underground miners around the world are exposed to relatively high con­
centrations of submicron aerosols and gases (Cohen et al. 2002, Pronk et al. 2009, Peters et al.
2017, Susanto et al. 2018, MSHA 2022). Long-term exposure to aerosols and gases emitted by
“traditional” (Ruehl et al. 2015) diesel engines was found to have adverse effects on pulmonary
(Kim et al. 2018, Du et al. 2020), vascular (Mills et al. 2005, Nejad et al. 2013), and other systems
(Faherty et al. 2021). The International Agency on Research on Cancer (IARC) recognizes the
risks of lung cancer and categorized diesel engine exhaust as a carcinogen to humans (group 1)
(IARC 2012). Rapid development of engine, aftertreatment, fuel, and lubrication oil technologies
over the past couple of decades resulted in “clean” engines that produce substantially lower par­
ticulate and gaseous emissions (Fiebig et al. 2014, Ruehl et al. 2015, Dallman and Menon 2016).
The implementation of various control strategies and technologies such as replacement of older
with advanced engines (Ruehl et al. 2015, Bugarski et al. 2020a), retrofitting older engines with
exhaust aftertreatment technologies (Bugarski et al. 2009, Bugarski et al. 2020b), the use of alter­
native fuels (Bugarski et al. 2014, Bugarski et al. 2016), and replacement of diesel-powered
vehicles with battery-powered (Halim et al. 2022) electric vehicles have resulted in a gradual
reduction in the industry-wide average exposures of underground metal and nonmetal miners to
elemental carbon (EC) and total carbon (TC) concentrations (Bugarski and Potts 2018, Bugarski

DOI: 10.1201/9781003429241-11

104
et al. 2022, Gren et al. 2022). In addition, the exposure to exhaust from contemporary “clean”
diesel engines equipped with advanced exhaust aftertreatment systems were shown to have less
adverse health effects (McDonald et al. 2015). However, some occupations across the industry
are still exposed to relatively high concentrations of “traditional” diesel aerosols (Bugarski and
Potts 2018, Gren et al. 2022, MSHA 2022).
Monitoring ambient concentrations and personal exposures to diesel particulate matter
(DPM) is complicated by the fact that DPM describes very complex ever-changing products of
incomplete combustion of petroleum diesel and alternative fuels (Bugarski et al. 2014, Bugarski
et al. 2016). This is further complicated in many situations by the use of multiple diesel engines
of various generations and designs (Bugarski et al. 2020b) operated over a wide spectrum of
conditions at various locations in underground mines (Rubeli et al. 2004, Saarikoski et al.
2018, Saarikoski et al. 2019, Bugarski et al. 2020c, Bugarski et al. 2022). In addition, sampling,
analysis, and measurement of DPM is even further complicated by other factors including
small particle size, the dynamic nature of the formation and transformation of DPM, and the
interaction with other aerosols and environment (Bugarski et al. 2012, Giechaskiel et al. 2014).
The currently used methods for compliance monitoring of exposures of underground miners
to diesel aerosols at permissible exposure levels (PELs) of 160 µgTC/m³ (71 Fed. Reg. 33387 2006,
OHSA 2017) and 100 µgEC/m³ (AIOH 2013) are grounded in the use of the National Institute for
Occupational Safety and Health (NIOSH) Method 5040 (Birch 2002, NIOSH 2003, NIOSH
2016). These offline methods are founded on: (1) collecting representative samples of diesel aero­
sols on quartz fiber filters (QFFs) using size-selective sampling, and (2) subsequent carbon ana­
lysis of those filters using a thermo-optical transmittance (TOT) method (NIOSH 2003).
The sampling apparatus currently used for compliance sampling incorporates two consecutive
pre-selectors (MSHA 2020, SKC 2022). This apparatus was designed using parameters estab­
lished during development of the personal diesel aerosol sampler by the U.S. Bureau of Mines
(Cantrell and Rubow 1991, McCartney and Cantrell 1992): (1) a respirable cyclone operated at
1.7 lpm with a cut-off size of approximately 4.0 µm, which is used to eliminate the majority of
coarse, mostly mechanically generated dust from respirable dust, and (2) a four-orifice sharp-cut
impactor with a cut-off size of approximately 0.8 µm when operated at 1.7 lpm (McCartney and
Cantrell 1992, Noll et al. 2005) incorporated with the cassette is used to eliminate the remaining
dust from diesel aerosols. The design criteria for the impactors (Huang 2005) used with the per­
sonal diesel aerosol sampler (PDAS) (McCartney and Cantrell 1992) and consequently in the
SKC Inc. DPM cassette (Noll et al. 2005) were established using the results of the size-selective
aerosol measurements in four underground coal mines in the United States in the late 1980s
(Cantrell and Rubow 1991). Those measurements showed that contemporary diesel engines con­
tributed aerosols with a mass median diameter larger than 150 nm. The sharp-cut impactor with
cut-off size of 0.8 ± 0.1 µm was found to provide acceptable separation of coal dust and diesel
aerosols. The average loss of diesel aerosols was estimated to be 0.2% to 8.4%, and contamin­
ation with coarse aerosols was estimated to be 0.1% to 3.6% (Cantrell and Rubow 1991). It is
important to note that the concentrations of diesel aerosol in various areas in those mines where
diesel vehicles were used were relatively high (substantially above current PEL) and contributed
a substantial fraction of respirable aerosol mass (Cantrell et al. 1992).
The DPM collected on QFFs is analyzed for EC and organic carbon (OC) using TOT analysis
(NIOSH 2003). The results are used to determine total carbon (TC=EC+OC). The TOT analysis
was chosen over the traditionally used gravimetric analysis for its sensitivity and selectivity for
monitoring DPM at the Mine Safety and Health Administration (MSHA) targeted PEL levels
(Ramachandran and Watts 2003). The method was originally developed for EC analysis and later
adopted for TC analysis (MSHA 2016). EC was proposed as a more suitable surrogate than TC
for DPM monitoring (Birch and Noll 2004) because EC is: (1) selective of combustion byproducts
of diesel fuels, (2) not present in other submicron aerosols in underground mines such as cigarette
smoke, wood smoke, and oil mist, and is (3) present in small quantities in coal and other carbon­
aceous dust. However, since TC makes up the majority of “traditional” DPM and includes
important carcinogenic organic compounds, it was selected by the MSHA as a surrogate for
monitoring exposures of underground miners to DPM (71 Fed. Reg. 33387 2006). The appropri­
ateness of the methodology of using TC as a surrogate for assessing exposures of underground

105
miners to DPM have been challenged previously (Cohen et al. 2002, Noll et al. 2015). The defi­
ciencies in the precision and accuracy of this methodology were primarily attributed to a number
of issues including: (1) potential presence of OC interferences in an underground mining environ­
ment (Cohen et al. 2002, Birch and Noll 2004, Noll et al. 2015), (2) adsorption of gas phase (non-
DPM) OC on QFFs (Noll and Birch 2008, Bosch 2015), (3) contamination of QFFs in SKC
DPM cassettes (Noll et al. 2019), and (4) losses of DPM to coarse aerosols (Gaillard et al. 2019).
At present, MSHA does not require monitoring exposure to DPM in underground coal mines in
the United States. Monitoring DPM exposures of underground metal, nonmetal, and coal mines
in Australian states and Germany is performed using EC as a surrogate (AIOH 2013).
Since the presence of micron-sized particles in the DPM samples can interfere with TOT analysis,
the effective removal of micron-sized aerosols using size-selective sampling is critical to monitoring
DPM exposures in underground mining environments. The presence of coal dust that contains non­
trivial amounts of OC (Yang and Yu 2002) and to some extent EC (Birch and Noll, 2004) can cause
overestimation of OC and EC and errors in determination of the EC/OC split. If present in the
DPM samples, alkali, alkaline-earth, and transition metal salts could interfere with TOT analysis by
changing EC oxidation temperature and causing more charring (Wang et al. 2010). Carbonates pre­
sent in rock dusting material could also interfere with the DPM measurement (Cavalli et al. 2010).
The evolution of diesel technology resulted in lower emission rates, decrease in size, and changes
in chemical composition of emitted diesel aerosols. Measurement of size distributions of diesel aero­
sols in underground mines (Skubacz et al. 2017, Saarikoski et al. 2019, Salo et al. 2021, Bugarski
et al. 2022) shows that the size distributions of diesel aerosols in many contemporary underground
mines, particularly in the mines with a high penetration of advanced diesel technology, are character­
ized with smaller diameters from those used when designing the PDAS (McCartney and Cantrell
1992) and DPM cassette (SKC). In the late 1990s and early 2000s when EC and TC were selected as
surrogates for monitoring DPM exposures of underground metal and nonmetal mines in the United
States (71 Fed. Reg. 33387 2006, MSHA 2016), the contemporary diesel engines emitted DPM that
was primarily made of EC and OC (Lowenthal et al. 1994, McDonald et al. 2003). Although EC
and especially OC still contribute substantially to total particulate matter mass, the dominant con­
stituents of particulate mass emitted by advanced diesel engines equipped with diesel particulate filter
(DPF) systems are noncarbonaceous compounds such as nitrites, sulfates, and ammonium (Herner
et al. 2011, Khalek et al. 2011, Khalek et al. 2015, Ruehl et al. 2015).
The changes in the DPM concentrations and physical properties necessitates improvements
in the precision and accuracy of the methodology used for monitoring DPM using TC as
a surrogate, at levels substantially below current MSHA PEL of 160 µgTC/m³. Although previ­
ously proposed interventions (Noll et al. 2005, Noll et al. 2015, Noll et al. 2019) resulted in
measurable improvements, additional work is needed to minimize collection of micron-sized
particles that can interfere with TOT analysis. This evaluation was conducted to evaluate the
potential for an improvement in sampling methodology for the collection of diesel aerosols by
reducing and eventually optimizing cut-off size for the impactor in the DPM sampling cassette.

2 METHODOLOGY

Three impactors with different sized orifices were evaluated. The reference impactor (I-061) was
the one in the commercially available DPM cassettes (SKC 225-317). The additional two impact­
ors were constructed by modifying impactor assemblies in the existing DPM cassettes to develop
the sharp-cut impactors with lower cut-off sizes. In the modified assemblies, the original 0.61-
mm synthetic sapphire orifice inserts (SSOIs) were replaced with 0.52-mm SSOIs (O’Keefe Con­
trols Co., Monroe, CT, SAP-44-SA) to obtain I-052 and with 0.44-mm SSOIs (O’Keefe Controls
Co., Monroe, CT, SAP-44-SA) to obtain I-044. For the sake of simplicity, the distance between
nozzle and impaction plate was kept identical to the one in the existing DPM cassette.
The penetration efficiencies of the impactors were determined using polystyrene divinylben­
zene (PS/DVB) beads with nominal diameters of 200, 300, 400, 500, 600, 700, 800, and 900 nm
in aqueous suspension (Thermo Fisher Scientific, Fremont, California, Nanosphere™ Poly­
mer Microspheres). The PS/DVB beads were aerosolized using a 3-Jet Collison nebulizer (CH

106
Technologies, Westwood, New Jersey). The normalized cumulative size distributions of aero­
solized beads measured with the scanning mobility particle sizer (SMPS) with the electrostatic
classifier (TSI, Shoreview, MN, Models 3080L) and ultrafine condensation particle counter
(TSI, Shoreview, MN, Models 3776), and Aerodynamic Particle Sizer® (APS™) (TSI, Shore­
view, MN, Model 3321) are shown in Figure 1.

Figure 1. Normalized cumulative size distributions of aerosolized beads measured with SMPS and APS.

The impactors were field evaluated in the outby area of the underground coal mine during
two shifts (S2 and S3) to assess: (a) performance of the I-061, I-052, and I-044 impactors with
respect to penetration of micron-sized dust, and (b) differences in losses in diesel aerosols
between those three impactors. This evaluation was executed as a part of the study described
by Bugarski et al. (2022). The mine used “traditional” diesel engines in all categories of
vehicles, retrofit type DPFs on the nonpermissible equipment, and filtration systems with dis­
posable filter elements on permissible equipment. The aerosol sampling and measurements
were performed at the continuous miner section. The size distribution of aerosols measured
with the fast mobility particle sizer (FMPS, TSI Model 3091) during S2 and S3 were found to
be bimodal with the agglomeration mode electrical mobility CMDs between 52 and 86 nm
(Bugarski et al. 2022). By introducing the assumption that the density of diesel aerosols in the
mine correlate with equations introduced by Maricq and Xu (2004) and Liu et al. (2009), the
mass median diameters of those aerosols could be estimated to be between 123 and 172 nm.

Figure 2. Sampling apparatuses: (a) used to collect DPM for SEM analysis, and (b) used to collect
DPM for carbon analysis.

107
 The issue of penetration of micron-sized aerosols through the I-061, I-052, and I-044 impact­
ors was investigated qualitatively on the results of scanning electron microscopy (SEM) analysis
performed on the samples collected with the sampling apparatus shown in Figure 2a. Four differ­
ent kinds of samples were collected concurrently using: (a) open-face 5-piece cassette, (b) 5-piece
cassette preceded by cyclone and I-061, (c) 5-piece cassette preceded by cyclone and I-052, and
(d) 5-piece cassette preceded by I-044. The samples were collected on the 25-mm, 0.2-μm track-
etched polycarbonate membrane filters (Millipore Sigma, Whatman® Nuclepore™, Darstadt,
Germany). The samples collected at a 600-second sampling time were used for analysis.
Twenty-one sections of each of the filters, one in the center and four groups of five radially distrib­
uted around the center at 90° spacing were analyzed by scanning electron microscopy (SEM) (Model
S-4800 by Hitachi, Tokyo, Japan) using 10K magnification and energy dispersive x-ray spectroscopy
using 20 to 70K magnification (by Bruker Quantax, Madison, Wisconsin). The particle imaging was
performed at 20 kV, and elemental analysis was performed using 20-kV incident beam energy.
The issue of the differences in diesel aerosol (EC and OC) losses in the three evaluated impactors
was studied using the results of the carbon analysis on sets of at least four filter samples of the same
kind collected concurrently using the custom sampling apparatus shown in Figure 2b. The respir­
able cyclones (Zefon International, Ocala, FL, Zefon Nylon Dorr-Oliver Cyclone) were used ahead
of each of the impactors. The DPM samples were collected on the primary and secondary quartz
fiber filters (QFFs) (Millipore Sigma, Darmstadt, Germany, AQFA03700) housed inside of the
DPM cassettes. To minimize OC contamination, the QFFs were removed from the impactors and
preheated for two hours at a temperature of 800°C (Noll et al. 2019). The nominal sampling flow
rates of 2.0 lpm were maintained using critical flow orifices installed in two common manifolds.
A single vacuum pump (Oerlikon Leybold Vacuum, Cologne, Germany, Segovac SV25B) was used
to draw all samples. All DPM samples were analyzed at the NIOSH Pittsburgh Mining Research
Division using the NIOSH Method 5040 with modified temperature protocol. The analysis was
performed using an OC/EC Aerosol Analyzer (Sunset Laboratory Inc., Portland, OR).

3 RESULTS

The penetration curves for the assessed I-061, I-052, and I-044 impactors are shown in
Figure 3. The cut-off diameters (D50) for those impactors were estimated to be 716, 466, and
258 nm, respectively. The geometric standard deviations (GSDs) were calculated (Marple et al.
2001) as 1.359, 1.582, and 2.151. The D50 and GSD for I-061 are similar to those assessed for
the impactors in SKC DPM cassettes by Olson (2001) (746-770 nm and 1.391-1.420).

Figure 3. Penetration curves for the impactors.

Twenty-one images of each of the filters were examined for the presence of micron-sized
particles. A total of twenty, eleven, seven, and three micron-sized particles were identified on
the open face, I-061, I-052, and I-044, respectively. The SEM images of the filters collected as
open face and with I-061, I-052, and I-044 impactors preceded with respirable cyclones are
shown in Figure 4. Elemental analysis of the selected micron-sized particles found at the same

108
locations showed that those particles were made primarily of the elements found in explosion
suppression rock dust (Man and Teacoach 2009) that likely were aerosolized by movement of
the diesel-powered equipment.
The results of carbon analysis performed on the samples collected with I-061, I-052, and I-044
impactors are shown in Figure 5. The mean concentrations of OC, EC, and TC were calculated
using results for at least 4 samples of the same kind. Certain samples (Figure 5) were deemed out­
liers due to potential contamination during the manufacturing process and excluded from the cal­
culations. The error bars represent one standard deviation of means for the specific sets of
samples. The OC, EC, and TC concentrations between the impactors were compared using ana­
lysis of variance (ANOVA, OriginPro, OriginLab Corporation, Northampton, MA). All OC, EC,
and TC concentrations were significantly different (p<0.05) between the impactors except OC
concentrations in S2.

Figure 4. SEM images for the samples collected: (a) without pre-selectors, (b) with cyclone/I-061, (c)
with cyclone/I-052, and (d) with cyclone/I-044.

In the case of samples collected during S2, the OC, EC, and TC concentrations were 17.5%,
6.6%, and 11.0% lower for the samples collected with I-052 rather than I-061, and 34.3%, 14.1%,
and 22.2% lower for the samples collected with I-044 rather than I-061. In the case of samples col­
lected during S3, the OC, EC, and TC concentrations were 24.5%, 7.4%, and 15.6% lower for the
samples collected with I-052 rather than I-061 and 30.4%, 13.9%, and 21.7% for the samples col­
lected with I-044 rather than I-061.

Figure 5. OC, EC, and TC concentrations measured during: (a) S2, and (b) S3 (Number of samples: * -
four, ** - five, *** - six).

109
4 DISCUSSION

Reliable monitoring of exposures to TC, particularly the OC fraction of it, at the levels sub­
stantially below the current MSHA PEL of 160 µgTC/m³ is challenging due to earlier identified
sampling and analytical issues. One possible improvement involves minimizing penetration of
the micron-sized non-diesel aerosols that could potentially adversely affect the results of the
TOT analysis (Yang and Yu 2002, Birch and Noll, 2004, Wang et al., 2010, Cavalli et al.
2010). We examined the potential of the SKC DPM cassettes with the modified impaction
assemblies to eliminate micron-sized aerosols more effectively from DPM samples.
The results of SEM analysis performed on the limited set of samples collected in
a contemporary coal mine showed that a substantial number of micron-sized aerosols present in
the mine atmosphere penetrated the impactor currently used in the SKC DPM cassette (I-061,
D50 = 716 nm). The I-052 (D50 = 466 nm) and particularly the I-044 (D50 = 258 nm) impactors
were much more effective in preventing micron-sized aerosols from reaching the sampling filters.
The QFFs collected with I-061, I-052, and I-044 contained comparable amounts of EC and
OC. The observed differences in EC and OC were at the level of limit of quantification (LOQ)
for EC and OC of the NIOSH Method 5040. The losses in EC for I-052, 6.6 and 7.4% for S2
and S3, respectively, were found to be statistically significant, but still relatively minor. The
losses in EC for I-044, 14.1 and 13.9% for S2 and S3, respectively, were somewhat more pro­
nounced. The losses in OC for I-052, 17.5 and 24.5% for S2 and S3, respectively, and for
I-044, 34.3 and 30.4% for S2 and S3, respectively, were found to be more substantial than
those observed for EC. It is important to note that the uncertainty of OC measurements was
substantially higher for the majority of samples than that of EC measurements.
This limited set of data indicate that due to relatively large losses in TC, the evaluated I-044
might not be suitable for DPMs sampling in the mines with a large presence of equipment
powered by “traditional” diesel engines. Optimization of the preselector for specific applica­
tions would be necessary. Those efforts would require good understanding of the physical
properties, particularly size distribution of diesel aerosols at the specific site. Unfortunately,
this kind of data is currently available for a very few sites.
Additionally, the design of the impactor needs to be optimized. In the case of the I-044
impactor, the relatively large GSD of 2.151 indicates that the design, in particular sharpness
of the impactor, can be potentially further improved by reducing the distance between nozzle
and impaction plate.

5 CONCLUSION

The changes in concentrations and physical properties of diesel aerosols in contemporary under­
ground mines necessitates improvements in precision and accuracy of the methodology used for
sampling and analysis of DPM. This is particularly important if TC is used as a surrogate for moni­
toring exposure to DPM. The results of this evaluation showed the potential for improvements in
the currently used size-selective sampling of diesel aerosols in underground mines. The SKC DPM
cassettes with the modified impaction assemblies with lower cut-off sizes were shown to effectively
eliminate micron-sized aerosols from DPM samples. The cut-off size of the impactor could be fur­
ther optimized to minimize penetration of micro-sized aerosols and losses in EC or OC.
It is important to note that this evaluation was limited in scope and just proved the concept.
Measurements were done over relatively limited period(s) of time in a single section of an under­
ground coal mine that is using equipment powered by “traditional” diesel engines. Additional
studies are needed to develop, optimize, and evaluate impactors for the specific applications, par­
ticularly in the underground mines with the substantial use of “clean” diesel technologies.

6 DISCLAIMER

The findings and conclusions in this manuscript are those of the authors and do not necessar­
ily represent the official position of the National Institute for Occupational Safety and Health

110
(NIOSH), Centers for Disease Control and Prevention (CDC). Mention of company names or
products does not constitute endorsement by NIOSH or CDC.

REFERENCES

71 Fed. Reg. 33387 2006. 30 CFR 57: Diesel particulate matter exposure of underground metal and non­
metal miners. Limit on concentration of diesel particulate matter. Code of Federal Regulations. Wash­
ington, DC: U.S. Government Printing Office, Office of the Federal Register.
AIOH 2013. Diesel Particulate Matter & Occupational Health Issues. Position Paper. Australian Institute
of Occupational Hygiene. AIOH Exposure Standards Committee.
Birch, E.M. 2002. Occupational monitoring of particulate diesel exhaust by NIOSH method 5040.
Applied Occupational and Environmental Hygiene, Vol. 17(6): 400–405. https://doi.org/10.1080/
10473220290035390.
Birch, M.E., Noll, J.D. 2004. Submicrometer elemental carbon as a selective measure of diesel particulate
matter in coal mines. Journal of Environmental Monitoring 6:799–806. https://doi.org/10.1039/b407507b.
Bosch, R. 2015. Review of sampling and evaluation techniques for diesel exhaust particulate matter in the
mining industry and the application of a vapor phase organic carbon correction factor. Montana Tech of the
University of Montana. Graduate Theses Non-Theses. 61. http://digitalcommons.mtech.edu/grad_rsch/61.
Bugarski, A.D. Hummer, J.A., Vanderslice, S. 2016. Effects of hydrotreated vegetable oil on emissions of
aerosols and gases from light-duty and medium duty older technology engines, Journal of Occupa­
tional and Environmental Hygiene, 13:4, 293–302, http://dx.doi.org/10.1080/15459624.2015.1116695.
Bugarski, A.D., Cauda E., Janisko, S.J., Hummer, J.A., Patts, L.D. 2010. Aerosols emitted in under­
ground mine air by diesel engine fueled with biodiesel. Journal of Air and Waste Management Associ­
ation. 60, 237–244. https://doi.org/10.3155/1047-3289.60.2.237.
Bugarski, A.D., Hummer, J.A., Vanderslice, S., Barone, T. 2020a. Retrofitting and repowering as control
strategies for the curtailment of exposure of underground miners to diesel aerosols. Journal of Mining,
Metallurgy & Exploration, 37(2):791–802. https://doi.org/10.1007/s42461-019-00146-z.
Bugarski, A.D., Hummer, J.A., Vanderslice, S., Mischler, S. 2020b. Contribution of various types and
categories of diesel-powered vehicles to aerosols in an underground mine. Journal of Occupational
and Environmental Hygiene 17: 121–134. https://doi.org/10.1080/15459624.2020.1718157.
Bugarski, A.D., Hummer, J.A., Vanderslice, S., Shahan, M.R. 2020c. Characterization of aerosols in an
underground mine during a longwall move. Mining, Metallurgy & Exploration (MM&E)
37:1065–1078. https://doi.org/10.1007/s42461-020-00209-6.
Bugarski, A.D., Janisko S., Cauda E.G., Noll, J.D., Mischler, S.E. 2012. Controlling Exposure - Diesel
Emissions in Underground Mines. Society for Mining, Metallurgy, and Exploration. ISBN-13:
9780873353601.
Bugarski, A.D., Janisko, S.J., Cauda, E.G., Patts, L.D., Hummer, J.A., Westover, C., Terrillion, T. 2014.
Aerosols and criteria gases in an underground mine that uses FAME biodiesel blends. Ann. Occup.
Hyg. 58(8):971–982. http://dx.doi.org/10.1093/annhyg/meu049.
Bugarski, A.D., Potts, J.D. 2018. Exposures of Underground Miners to Diesel Particulate Matter in the
United States. 24th Annual Mining Diesel Emission Conference (MDEC) Conference. Toronto, Ontario,
Canada, October 2–4. Available from https://www.mdec.ca/2018/. (Accessed on January 14, 2021).
Bugarski, A.D., Schnakenberg, G.H., Jr., Hummer, J.A., Cauda E., Janisko, S.J., and Patts, L.D. 2009.
Effects of diesel exhaust aftertreatment devices on concentrations and size distribution of aerosols in
underground mine air. Environmental Science and Technology 43, 6737–6743. https://doi.org/10.3155/
1047-3289.60.2.237.
Bugarski, A.D., Vanderslice, S., Hummer, J.A., Barone, T.L., Mischler, S.E., Peters, S., Cochrane, S.,
Winkler, J. 2022. Diesel aerosols and gases in an underground coal mine. Mining, Metallurgy &
Exploration 39:937–945, https://doi.org/10.1007/s42461-022-00588-y.
Cantrell, B.K., Rubow K.L. 1991. Development of personal diesel aerosol sampler design and perform­
ance criteria. Mining Engineering 43(2):232–236.
Cantrell, B.K., Rubow, K.L., Watts, W.F. Jr., Carlson, D.H. 1992. Pollutant levels in underground coal
mines using diesel equipment. Society for Mining Metallurgy and Exploration Transaction 290:1901–1907.
Cavalli, F., Viana, M., Yttri, K.E., Genberg, J. and Putaud, J.P., 2010. Toward a standardised
thermal-optical protocol for measuring atmospheric organic and elemental carbon: the EUSAAR
protocol. Atmospheric Measurement Techniques, 3(1):79–89. www.atmos-meas-tech.net/3/79/2010/.
Cohen, H.J., Borak, J., Hall, T., Sirianni, G., Chemerynski S. 2002. Exposure of miners to diesel exhaust
particulates in underground nonmetal mines, AIHA Journal, 63:5, 651–658, https://doi.org/10.1080/
15428110208984753.

111
Dallmann, T., Menon, A. 2016. Technology pathways for diesel engines used in non-road vehicle and
equipment. The International Council on Clean Transportation. White paper. September.
Du, M., Hall, G.l., Franklin, P., Musk, A.W. (Bill), Mullins, B.J., de Klerk, N., Elliott, N.S.J., Sodhi-
Berry, N., Brims, F., Reid, A. 2020. Association between diesel engine exhaust exposure and lung
function in Australian gold miners. International Journal of Hygiene and Environmental Health 226
113507. https://doi.org/10.1016/j.ijheh.2020.113507.
Gaillard, S., Sarver, E., Cauda, E. 2019. A field study on the possible attachment of DPM and respirable dust in
mining environments. Journal of Sustainable Mining 18: 100–108. https://doi.org/10.1016/j.jsm.2019.02.008.
Faherty, T., MacKenzie, A.R., Shapiro, K., Delgado-Saborit, J.M., McFiggans, G., Alfarra, R.,
Raymond, J. 2021. The impact of acute diesel exhaust exposure on executive brain function. Journal
of Vision. Volume 21 (9): 2562. https://doi.org/10.1167/jov.21.9.2562.
Fiebig, M., Wiartalla, A., Holdenbaum, B., Kiesow, S. 2014. Particulate emissions from diesel engines:
Correlation between engine technology and emissions. J Occup Medic Toxic 9(1):6. https://doi.org/
10.1186/1745-6673-9-6.
Giechaskiel, B., Maricq, M., Ntziachristos, L., Dardiotis C., Wang, X., Axmann, H., Bergmann, A.,
Schindler W. 2014. Review of motor vehicle particulate emissions sampling and measurement: From
smoke and filter mass to particle number. Journal of Aerosol Science 67: 48–86. https://doi.org/
10.1016/j.jaerosci.2013.09.003.
Gren, L., Krais, A.M., Assarsson, E., Broberg, K., Engfeld, M., Lindh, C., Strandberg, B., Pagels, J.,
Hedmer, M 2022. Underground emissions and miners’ personal exposure to diesel and renewable
diesel exhaust in a Swedish iron ore mine. International Archives of Occupational and Environmental
Health 95:1369–1388,. https://doi.org/10.1007/s00420-022-01843-x.
Halim, A., Lööw, J., Johansson, J., Gustafsson, J., van Wageningen, A., Kocsis, K. [2022]. Improvement
of working conditions and opinions of mine workers when battery electric vehicles (BEVs) are used
instead of diesel machines — Results of field trial at the Kittilä Mine, Finland. Mining, Metallurgy &
Exploration (2022) 39:203–219. https://doi.org/10.1007/s42461-021-00506-8.
Herner, J.D., Hu, S., Robertson, W.H., Huai, T., Chang, M.-C.O., Rieger, P., Ayala, A. 2011. Effect of
advanced aftertreatment for PM and NOx reduction on heavy-duty diesel engine ultrafine particle
emissions. Environ Sci Technol 45:2413–2419.
IARC 2012. IARC Press Release No. 213, June 12. Diesel engine exhaust carcinogenic. International
Agency for Research on Cancer, World Health Organization International Agency for Research on
Cancer, Lyon, France
Khalek, I.A., Blanks, M.G., Merritt, P.M., Zielinska, B. 2015. Regulated and unregulated emissions
from modern 2010 emissions-compliant heavy-duty on highway diesel engines. J. Air Waste Manage.
Assoc. 65(8):987–1001. https://doi.org/10.1080/10962247.2015.1051606.
Khalek, I.A., Bougher, T.L., Merritt, P.M., Zielinska, B. 2011. Regulated and unregulated emissions
from highway heavy-duty diesel engines complying with U.S. Environmental Protection Agency 2007
emissions standards. J. Air Waste Manage. Assoc. 61(4):427–442. doi:10.3155/1047-3289.61.4.427.
Kim, J., Peters, C.E., Arrandale, V.H., Labreche, F., Ge, C.B., McLeod, C.B., Song, C., Lavoué, J., Davies, H.
W., Nicol, A.-M., Pahwa, M., Demers, P.A. 2018. Burden of lung cancer attributable to occupational diesel
engine exhaust exposure in Canada. Occupational and Environmental Medicine 75(9):617–22.
Liu, G.Z., Vasys, V.N., Dettmann, M.E., Schauer, J.J., Kittelson, D.B., Swanson, J. 2009. Comparison
of strategies for the measurement of mass emissions from diesel engines emitting ultra-low levels of
particulate matter. Aerosol Sci. Technol. 43, 1142–1152.
Lowenthal, D.H., Zielinska, B., Chow, J.C., Watson, J.G., Gautam, M., Ferguson, D.H., Neuroth, G.R.,
Stevens, K.D. 1994. Characterization of heavy-duty diesel vehicle emissions. Atmos Environ 28(4): 731–743.
Man, C.K., Teacoach, K.A. 2009. How does limestone rock dust prevent coal dust explosions in coal
mines?. Mining Engineering 61(9):69–69.
Maricq, M.M., Xu, N. 2004. The effective density and fractal dimension of soot particle from premixed
flames and motor vehicle exhaust. Aerosol Science 35, 1251–1274.
McCartney, T.C., Cantrell, B.K. 1992. A cost-effective personal diesel exhaust aerosol sampler. United
States Department of the Interior, Bureau of Mines, IC 9324:24–30.
McDonald, J.D., Doyle-Eisele, M., Seagrave, J.C., Gigliotti, A.P., Chow, J., Zielinska, B., Mauderly, J.
L., Seilkop, K.S., Miller, R.A., 2015. Part 1. Assessment of carcinogenicity and biologic responses in
rats after lifetime inhalation of new-technology diesel exhaust in the ACES bioassay. In: Advanced
Collaborative Emissions Study (ACES): Lifetime Cancer and Non-Cancer Assessment in Rats
Exposed to New-Technology Diesel Exhaust. Research Report 184. Health Effects Institute, Boston.
McDonald, J.D., Zielinska, B., Sagebiel J.C., McDaniel, M.R., Mousset-Jones, P. 2003. Source appor­
tionment of airborne fine particulate matter in an underground mine. Journal of the Air & Waste
Management Association, 53 (4): 386–395, https://doi.org/10.1080/10473289.2003.10466178.

112
Mills N.L., Törnqvist H., Robinson S.D., Gonzalez M., Darnley K., MacNee W., Boon, N.A.,
Donaldson, K., Blomberg, A., Sandstrom, T., Newby, D.E. 2005. Diesel exhaust inhalation causes
vascular dysfunction and impaired endogenous fibrinolysis. Circulation 112(25):3930–6. https://doi.
org/10.1161/CIRCULATIONAHA.105.588962.
MSHA 2016. Carbon (Diesel Particulate Matter) by Thermal/Optical Analysis. MSHA P-13. Revision
#003, U.S. Department of Labor. Mine Safety and Health Administration.
MSHA 2020. Particulate sampling methods. In: Health Inspection Procedures Handbook. PH20-V-4.
U.S. Department of Labor, Mine Safety and Health Administration. Available from: https://arlweb.
msha.gov/READROOM/HANDBOOK/PH20-V-4.pdf (Accessed on August 15, 2022.).
MSHA 2022. The DPM personal sampling compliance data. The United States Department of Labor.
Mine Safety and Health Administration (MSHA). https://arlweb.msha.gov/OpenGovernmentData/
OGIMSHA.asp. (Accessed on August 11, 2022).
Nejad S.H., Takechi, R., Mullins, B.G., Giles, C., Larcombe A.N., Bertolatti, D., Rumchev, K.,
Dhaliwal, S., Mamo J. 2013. The effect of diesel exhaust exposure on blood-brain barrier integrity and
function in a murine model. Journal of Applied Toxicology 35:41–47. https://doi.org/10.1002/jat.2985.
NIOSH 2003. Diesel Particulate Matter (as Elemental carbon). Method 5040. In: NIOSH Manual of
Analytical Methods. 4th ed. Cincinnati, OH: U.S. Department of Health and Human Services, Centers
for Disease Control and Prevention, National Institute for Occupational Safety and Health.
NIOSH 2016. Monitoring diesel exhaust in the workplace. In: NIOSH Manual of Analytical Methods
(NMAM), 5th Edition, Chapter DL, Cincinnati, OH: U.S. Department of Health and Human Services,
Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health. Avail­
able from: https://www.cdc.gov/niosh/nmam/pdfs/NMAM_5thEd_EBook.pdf. Accessed on July 2, 2022.
Noll, J., Birch, M.E. 2008. Effects of sampling artifacts on occupational samples of diesel particulate
matter. Environ. Sci. Technol. 42:5223–5228. https://doi.org/10.1021/es702883k.
Noll, J., Cauda, E., Vanderslice, S.; Barone, T. 2019. Quantification of the effects of carbon on filter
media in SKC DPM cassettes on measurements of diesel particulate matter in underground mines.
SME Annual Meeting.
Noll, J., Gilles. S., Wu, H.W., Rubinstein. E. 2015. The relationship between elemental carbon and diesel
particulate matter in underground metal/nonmetal mines in the United States and coal mines. In Aus­
tralia Journal of Occupational and Environmental Hygiene 12: 205–211. https://doi.org/10.1080/
15459624.2014.960577.
Noll, J.D., Timko, R.J., McWilliams, L., Hall, P., Haney, R. 2005. Sampling results of the improved
SKC diesel particulate matter cassette. Journal of Occupational and Environmental Hygiene, 2:1,
29–37, https://doi.org/10.1080/15459620590900320.
OHSA 2017. Occupational Health and Safety Act, Ontario Regulation 854. Mines and Mining Plants.
Ontario Ministry of Labor.
Olson, B. 2001. Particle calibration of the disposable personal diesel aerosol sampler (PDAS). Report
submitted to the National Institute for Occupational Safety and Health. Minneapolis, MN: University
of Minnesota, Particle Calibration Laboratory.
Peters S., de Klerk, N., Reid, A., Fritschi L., Musk, A., Vermeulen, R. 2017. Quantitative levels of diesel
exhaust exposure and the health impact in the contemporary Australian mining industry. Occup
Environ Med. 74(4):282–289. https://doi.org/10.1136/oemed-2016-103808.
Pronk, A., Coble, J., Stewart, P.A. 2009. Occupational exposure to diesel engine exhaust: A literature
review. J. Exp. Sci. Envir. Epid. 19:443–457 (2009).
Ramachandran, G., Watts, W. 2003. Statistical comparison of diesel particulate matter measurement
methods. AIHA Journal, 64 (3):329–337, https://doi.org/10.1080/1542811030898482.
Rubeli, B., Gangal, M., Butler, K., Alder, W. 2004. Evaluation of the contribution of light-duty vehicle to
the underground atmosphere diesel emissions burden. In: Ganguli R, Bandopadhyay S, eds. Proceed­
ings of the 10th U.S./North American Mine Ventilation Symposium. Leiden, Netherlands: Balkema.
Ruehl, C., Herner, J.D., Yoon, S., Collins, J.F., Misra, C., Na, K., Robertson, W.H., Biswas, S.,
Chang, M.-C.O., Ayala, A. 2015. Similarities and differences between traditional and clean diesel PM.
Emiss Control Sci Technol 1:17–23.
Saarikoski, S., Salo, L., Bloss, M., Alanen, J., Teinilä, K., Reyes, F., Vázquez, Y., Keskinen, J.,
Oyola, P., Rönkkö, T., Timonen, H. 2019. Sources and characteristics of particulate matter at five
locations in an underground mine. Aerosol and Air Quality Research, 19:2613–2624, https://doi.org/
10.4209/aaqr.2019.03.0118.
Saarikoski, S., Teinilä, K., Timonen, H., Aurela, M., Laaksovirta, T., Reyes, F., Váasques, Y.,
Oyola, P., Artaxo, P., Pennanen, A.S., Junttila, S., Linnainmaa, M., Salonen, R. O., Hillamo, R.
2018. Particulate matter characteristics, dynamics and sources in an underground mine. Aerosol. Sci.
Technol. 52(1):114–122. doi:10.1080/02786826.2017.1384788.

113
Salo, L., Rönkkö, T., Saarikoski, S., Teinilä, K., Kuula, J., Alanen, J., Arffman, A., Timonen, H., Keskinen, J.
2021. Concentrations and size distributions of particle lung-deposited surface area (LDSA) in an under­
ground mine. Aerosol and Air Quality Research 21, 200660. https://doi.org/10.4209/aaqr.200660.
SKC 2022. DPM Cassette with impactor Cat. No. 225–317 – Operating Instructions. Form 400081 Rev
1808. SKC Inc. Eighty Four, PA. https://www.skcltd.com/images/pdfs/40081.pdf. (Accessed on
August 10, 2022).
Skubacz, K., Wojtecki, Ł., Urban, P. 2017. Aerosol concentration and particle size distributions in
underground excavations of a hard coal mine, International Journal of Occupational Safety and Ergo­
nomics, 23:3, 318–327, https://doi.org/10.1080/10803548.2016.1198553.
Susanto, A., Purwanto, P., Sunoko, H.R., Setiani, O. 2018. Assessment of diesel particulate matter
exposure of underground miners in Indonesia. Journal of Ecological Engineering 19(4): 34–42, https://
doi.org/10.12911/22998993/89671.
Wang, Y., Chung, A. Paulson, S.E. 2010. The effect of metal salts on quantification of elemental and
organic carbon in diesel exhaust particles using thermal-optical evolved gas analysis. Atmospheric
Chemistry and Physics, 10(23): 11447–11457. https://doi.org/10.5194/acp-10-11447-2010.
Yang, H. and Yu, J.Z. 2002. Uncertainties in charring correction in the analysis of elemental and organic
carbon in atmospheric particles by thermal/optical methods. Environmental Science & Technology, 36
(23): 5199–5204. https://doi.org/10.1021/es025672z.

114
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Results of diesel exhaust nanoparticle experimental sampling in

a cabin of LHD loader operating in an active ore heading area

S. Sabanov, N. Magauiya, A. Zeinulla, A. Abil & A. Qureshi

School of Mining and Geosciences, Nazarbayev University, Astana, Kazakhstan

M. Torkmahalleh
University of Illinois, Chicago, IL, USA

G. Nurshaiykova & D. Rakhimov

East Kazakhstan State Technical University, Ust-Kamenogorsk, Kazakhstan

ABSTRACT: Experimental sampling of diesel exhaust nanoparticles using a Naneos Partector 2

instrument has been conducted in an underground polymetallic mine. This sampling was conducted
within 12 minutes in the operator open cabin of the mine face LHD loader R1700 (engine model
Cat@C11 ACERT, 241 kW, Tier 3/Stage IIIA Equivalent Engine) working in the active ore head­
ing area. The aim of this experimental sampling was to determine the exposures of the cabin oper­
ator to diesel exhaust nanoparticles. The LHD loader operated in a 70 meters length drift
(4.3x4.2m), where the auxiliary ventilation velocity was around 0.5 m/s. As a result, the averaged
data for Lung Deposited Surface Area (LDSA) and diameter were 7 470 µm²/cm³ and 90 nm,
respectively. The results of this study suggest that Naneos Partector 2 sampling instrument can be
employed in polymetallic mines for monitoring workplace surveillance. This study offers important
data on particle surface area. As the surface contact between the particles and human cells is cru­
cial, particle surface area might be a new approach to investigate particle toxicity.

1 INTRODUCTION

Diesel-powered equipment is widely used in underground mining because of its high efficiency,
flexibility during developpment stage, reliability, and durability (Bugarski et al., 2012). Combustion
of diesel is the main source of the toxic gases and particles in underground metal and non-metal
mines. When underground miners, particularly truck drivers, continuously breathe contaminated
air for a prolonged length of time, severe occupational disorders may develop. US Federal M/NM
Mine Regulations have set a limit of 160 μg/m3 of total carbon (TC) for an eight-hour time weight
average (TWA) (OSHA-MSHA). TC value is obtained by summing the EC and OC numbers
together, and it generally accounts for 80 percent of the DPM (Kimbal et al, 2012). In addition to
toxic gases, diesel engines also release ultrafine particles (UFP) with a size of less than 0.1 µm. Their
accumulation in respiratory system is hazardous and can result in severe respiratory illnesses. When
evaluated in various environmental circumstances, the Lung Deposited Surface Area Concentration
(LDSA) approach was successful in recognizing UFP (Braakhuis et al. 2014, Kuuluvainen et al.
2016, Kuula et al. 2020, Tran et al. 2020, Afshar-Mohajer et al. 2020, Huynh et al. 2018). Ultrafine
particles (UFPs) can carry high concentration of toxic substances, and therefore, they can cause
inflammation through oxidative stress responses, atherosclerosis, increase of blood pressure, and
even myocardial infraction (Delfino et al. 2005).
The Lung Deposited Surface Area Concentration (LDSA) is a different way of describing
particle toxicity. The challenge with current mass-based exposure limits is that they fail to
account for ultrafine particles (less than 100 nanometres), which has negligible contributions

DOI: 10.1201/9781003429241-12

115
to particle mass but can penetrate deep into human airways and lodge in lung alveoli and
have high toxicity contributions. This is especially concerning because most diesel engine par­
ticle emissions (measured by number) fall into the ultrafine range (Braakhuis et al., 2014). In
distinct metropolitan contexts, LDSA concentrations, size distributions, and height profiles
have been measured in ambient conditions (Kuuluvainen et al., 2016; Kuula et al., 2020; Tran
et al., 2020). LDSA concentrations have been connected to emissions from combustion
sources. Afshar-Mohajer et al. (2020) used a variety of measures, including LDSA, to investi­
gate the variability of aerosol concentrations in different processing zones of a taconite mine.
Huynh et al. (2018) used numerous particle concentration measures, including LDSA, to
investigate fine particle concentrations in six taconite mines. LDSA concentrations in their
investigation ranged from 50 to 300 μm2/cm3 depending on the processing region. Although
LDSA level varied from station to station, Afshar-Mohajer et al. (2020) also noticed the
remarkable repeatability of concentration each day. The aim of this experimental sampling
was to determine the exposures of the cabin operator to diesel exhaust nanoparticles.

2 SAMPLING METHODOLOGY

The study was carried out in a polymetallic mine in the Eastern Kazakhstan. Diesel exhaust
samples were collected at the level +335m, which has 450 tons/day of total material moved.
The mine is operating 365 days a year and working in two 11-hour shifts each day, with two
mid-shift blasting breaks. The mine employs a push-pull ventilation system with two main
fans and several auxiliary fans within mine working faces.
Measurements were conducted in an area of the active ore heading, requiring the use of the
LHD loader CAT R1700 (engine model Cat@C11 ACERT, 241 kW) operated in a 70 meters
length drift which is 4.3 meters in height and 4.2 meters in wide (Figure 1). Average air vel­
ocity in the drift was around 0.5 m/s. The auxiliary fans was on during the active ore haulage
process to pull fresh air through the two ventilation ducts. Figure 1 shows the positioning of
LHD loader during its operational process.
During measurements, the LHD diesel loader was operating in its operational cycle: load
(~18 seconds), haul (~41 seconds), and dump (~7 seconds). The main source of diesel fumes
was only an exhaust pipe of the LHD loader.

Figure 1. Schematical layout of the sampling site.

The sampling was conducted by the Naneos Partector 2 within 12 minutes in the operator
open cabin of the mine face LHD loader working in the active ore heading area. The loader
engine corresponds to the Tier 3 or Stage IIIA Equivalent Engine and exploits the high-
quality diesel fuel DT-L-K2 (standard GOST 305-82), which in the first approximation corres­
ponds to the Euro 3 standard and differs only in the cetane number.

116
 The Naneos Partector 2 multimeric nanoparticle detector was used to measure the Lung Depos­
ited Surface Area (LDSA) and particle diameter. Prior to sampling, the collecting tape was
replaced, and all required calibrations were completed. During the sampling time temperature was
24° C, atmospheric pressure 964 hPa and humidity 53%. During ore haulage, the Naneos Partec­
tor 2 was placed inside the cabin with the suction duct taped to the loader roof frame at the oper­
ators face level (approximately 2.3 m from the floor). This was done for better understanding the
accurate exposure level of the workers. The LHD loader was not equipped with a sealed cab
(environmental cab) to protect the workers from the toxic gases or aerosols. Therefore, the expos­
ure levels indicated by the Partector 2 assumed to be equal to the actual exposure of the driver.

3 RESULTS AND DISCUSSION

As a result of Naneos Partector 2 produced from 700 samplings (every second), the averaged
data for Lung Deposited Surface Area (LDSA) was 7 470 µm²/cm³ with minimum of 1 657
µm²/cm³ and maximum of 12 792 µm²/cm³ (Figure 2). The LHD loader diesel engine was idle
running for the first 150 seconds. This can be observed from the graphs (Figure 2) where
LDSA are around 2000-4000 µm²/cm³. Then the LHD started its regular operational cycle.
Indicated peaks are basically related when the LHD engine increased revolutions per minute
for its operational processes and also the loader position in the drift.

Figure 2. Naneos Partector 2 LDSA concentrations over elapsed time.

Average particle diameter was 90 nm with minimum of 40 nm and maximum of 150 nm

(Figure 3)
The reasonable type of control technology used to minimize miners’ exposures to diesel particu­
late emissions is to have an air-conditioned cab on the vehicle which filters the particulate. Such
cabs can reduce diesel exhaust nanoparticle exposure from 60% to 80% and reduce miners’ expos­
ure to noise and dust. However, E-cabs are useful only for workers who stay inside the cab; those
workers involved in mine activities that are executed from outside the cab are not protected by
environmental cabs (Khan 2018). The effectiveness of these enclosures is highly dependent on
operators closing doors and windows, the maintenance of filtration systems, door seals and other
components such as floor bungs and grommets around cabin penetrations (Black 2017).

117
Figure 3. Particles diameter over elapsed time.

4 CONCLUSIONS

Experimental sampling of diesel exhaust nanoparticles using a Naneos Partector 2 instrument was
conducted in the underground polymetallic mine. This sampling was conducted in the operator
open cabin of the mine face LHD loader R1700 (engine model Cat@C11 ACERT, 241 kW, Tier
3/Stage IIIA Equivalent Engine) working in the active ore heading area. The aim of this experi­
mental sampling was to determine the exposures of the cabin operator to diesel exhaust nanopar­
ticles. The LHD loader operated in a 70 meters length drift (4.3m x 4.2m), where the auxiliary
ventilation velocity was around 0.5 m/s. As a result, the averaged data for LDSA and particles
diameter were found 7 470 µm²/cm³ and 90 nm, re-spectively. The LDSA results were compared
to the sampling produced out of the cabin in the same drift for a longer time period, which were
almost twice lower than in the cabin. However average particle diameter left unchanged 90 nm.
This demonstrates that LHD loader open cabin operators are more vulnerable to diesel exhaust
nanoparticles. The results of this study suggest that the Naneos Partector 2 sampling instrument
can be employed in polymetallic mines for monitoring workplace surveillance. This study offers
important data on particle surface area. As the surface contact between the particles and human
cells is crucial, particle surface area might be a new approach to investigate particle toxicity.

ACKNOWLEDGMENT
This study was supported by Nazarbayev University Grant Program: FY2020-СRP-1 Collab­
orative Research Project # 091019CRP2104.

REFERENCES

Afshar-Mohajer, N., Foos, R., Volckens, J., Ramachandran, G. 2020. Variability of aerosol mass and
number concentrations during taconite mining operations. Applied Occupational and Environmental
Hygiene. 17: 1–14. https://doi.org/10.1080/15459624.2019.1688823.
Black J, Reid D (2017) Delivering a healthy atmosphere underground in a cost-conscious environment,
Mine Ventilation Journal, ISBN 978 1 925100 617: 159–167.

118
Braakhuis, HM, Park, MV, Gosens, I, De Jong, WH, Cassee, FR, 2014. Physicochemical characteristics
of nanomaterials that affect pulmonary inflammation, Particle and Fibre Toxicology, vol. 11, no. 1, p.
18.
Cauda, E.G., Ku, B.K., Miller, A.L., Barone, T.L. 2012. Toward developing a new occupational expos­
ure metric approach for characterization of diesel aerosols. Aerosol. Sci. Technol. 46 (12): 1370–1381.
Hennig, F., Quass, U., Hellack, B., Küpper, M., Kuhlbusch, T. A. J., Stafoggia, M., & Hoffmann, B.
2018. Ultrafine and Fine Particle Number and Surface Area Concentrations and Daily Cause-Specific
Mortality in the Ruhr Area, Germany, 2009–2014., Environmental Health Perspectives, 126 (2),
027008. doi:10.1289/ehp2054.
Hussain S, Boland S, Baeza-Squiban A, Hamel R, Thomassen LCJ, Martens JA, et al. 2009. Oxidative
stress and proinflammatory effects of carbon black and titanium dioxide nanoparticles: role of particle
surface area and internalized amount. Toxicology 260(1-3):142–149, PMID: 19464580, https://doi.org/
10.1016/j.tox.2009.04.001.
Huynh, T., Ramachandran, G., Quick, H., Hwang, J., Raynor, P. C., Alexander, B. H., & Mandel, J. H.
2018. Ambient fine aerosol concentrations in multiple metrics in taconite mining operations. Annals of
Work Exposures and Health, 63(1): 77–90. https://doi.org/10.1093/annweh/wxy086.
Khan, M. U., Gillies, A. D. 2019. Diesel particulate matter: Monitoring and control improves safety and
air quality. Advances in Productive, Safe, and Responsible Coal Mining: 199–213. https://doi.org/
10.1016/b978-0-08-101288-8.00009-2.
Kimbal, K C, Pahler, L, Larson, R and Vanderslice, J, 2012. Monitoring diesel particulate matter and
calculating diesel particulate densities using grimm model 1.109 real-time aerosol monitors in under­
ground mines, Journal of Occupational and Environmental Hygiene, 9(6):353–361.
Kuula, J., Kuuluvainen, H., Niemi, J.V., Saukko, E., Portin, H., Kousa, A., Aurela, M., Rönkkö, T.,
Timonen, H. 2020. Long-term sensor measurements of lung deposited surface area of particulate
matter emitted from local vehicular and residential wood combustion sources, Aerosol Sci. Technol.
54: 190–202. https://doi.org/10.1080/02786826.2019.1668909.
Kuuluvainen, H., Rönkkö, T., Järvinen, A., Saari, S., Karjalainen, P., Lähde, T., Pirjola, L.,
Niemi, J. V., Hillamo, R., Keskinen, J. 2016. Lung deposited surface area size distributions of particu­
late matter in different urban areas. Atmos. Environ. 136: 105–113. https://doi.org/10.1016/j.
atmosenv.2016.04.019.
OSHA-MSHA. Diesel Exhaust/Diesel Particulate Matter’, December 2013. https://www.osha.gov/Publi
cations/OSHA-3590.pdf.
Ramachandran, G., Paulsen, D., Watts, W., Kittelson, D., 2005 Mass, surface area and number metrics
in diesel occupational exposure assessment. Environmental Monitoring, 7 (7): 728–735. https://doi.org/
10.1039/b503854e.
Sabanov, S, Zeinulla, A, Brune, J & Amouei Torkmahalleh, M 2021, ‘The results of the evaluations of
the instrumentation used to monitor concentrations of DPM’, in Proceedings of 18th North American
Mine Ventilation Symposium: 133–138.
Sioutas, C., Delfino, R. J., & Singh, M. 2005. Exposure assessment for atmospheric ultrafine particles
(ufps) and implications in Epidemiologic Research. Environmental Health Perspectives, 113
(8),947–955. https://doi.org/10.1289/ehp.7939.
Tran, P.T.M., Ngoh, J.R., Balasubramanian, R. 2020. Assessment of the integrated personal exposure to
particulate emissions in urban micro-environments: A pilot study. Aerosol Air Qual. Res. 20: 341–357.
https://doi.org/10.4209/aaqr.2019.04.0201.

119
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Importance of using real-time and microscopic analysis techniques

to characterize DPM in underground mines

A.A. Habibi
Ventilation Expert, PT Freeport Indonesia

K.O. Homan
Associate Professor, Missouri University of Science and Technology, MO, USA

A.D. Bugarski
Pittsburgh Mining Research Division, National Institute for Occupational Safety and Health,
Pittsburgh, Pennsylvania, USA

ABSTRACT: Diesel equipment exhaust is a primary source of carbon-rich submicron particles

in the underground mine atmosphere. In this study, the characterization of the morphological and
physical properties of particles was used for identification of emission source and understanding
the effect of diesel particulate matter (DPM) controll strategies on DPM characteristics. The size
distributions and other physical properties of diesel aerosol were investigated on size-segregated
samples collected in an underground mining operation. Scanning transmission electron micros­
copy (STEM) and fast mobility particle sizer were used concurrently to investigate size distribu­
tion based on particle projected area and electrical mobility. Other morphological attributes, such
as fractal dimension, carbon net counts, and primary particle dimensions were examined for light
and heavy-duty equipment powered by for various engine sizes. The results of STEM analysis
showed presence of three types of chain-like agglomerates as well as volatile carbon particles. The
results showed that when used concurrently, real-time and microscopic techniques can provide
a wealth of information on characteristics of aerosols in underground atmosphere.

1 INTRODUCTION

Diesel engines are the main source of carbon-rich submicron aerosol in underground mines.
Other sources of carbon-rich aerosols are oil mist from drilling operations, cigarette smoke,
and suspended carbonate dust in ore (Zielinska et al., 2002; McDonald et al., 2003; Noll et al.,
2006). Diesel aerosols are classified in three modes: nucleation mode (3-30 nm), accumulation
mode (30-500 nm), and coarse mode (>500 nm) (Kittelson et al. 1998). Due to small size, the
nucleation mode aerosols contribute substantially to number concentrations, but marginally
to the particle total mass (Bugarski et al., 2012). In the U.S., the Mine Safety and Health
Administration (MSHA) limits exposure of underground metal and nonmetal miners to diesel
particulate matter (DPM) to 160 µg/m³ of total carbon (TC) (30 CFR 57.5060). The TC con­
centrations are determined using NIOSH Method 5040 (NIOSH 2016). The mining industry
implemented various engineering controls and technologies to reduce the exposure (Bugarski
et al. 2014, Bugarski et al. 2022). A detailed knowledge of the size, structure, morphology,
fractal dimension, and composition could be used to identify the sources and processes guid­
ing formation, transformation of diesel aerosols. The same knowledge can be used for identi­
fying effective control strategies and technologies, source apportionment, and to understand
the interaction between micron-size dust particles and DPM.
In order to achieve the objectives, size segregated sampling using a multi-stage impactor
with Transmission Electron Microscopy (TEM) grids was conducted. The sampling was

DOI: 10.1201/9781003429241-13

120
conducted during different mining operation phases to collect a range of diverse samples.
Additionally, interaction of dust and diesel agglomerates can be investigated using Energy
Dispersive Spectroscopy (EDS) and microscopic image analysis.

2 METHODOLOGY

2.1 Instrumentation
Real time instrumentation (Table 1) was used to measure the particles size distributions and
number concentrations based on aerodynamic mobility (ELPI). The Electrical Low Pressure
Impactor (ELPI+) was used to collect size-segregated samples for S/TEM analysis (Table 1).

Table 1. List of instrumentation used in the study.

Instrument or Device Measurement or Sampling Objective

Electrical Low Pressure Impactor (Dekati, ELPI),
(ELPI used for real time readings)
Electrical Low Pressure Impactor (Dekati, ELPI +)
(ELPI+ is a newer model and was used to collect
S/TEMgrid samples)
Continuous measurement of number concentra­
tions and size distributions of aerosols with aero­
dynamic diameters between 30 and 10,000 nm
Collection of size-segregated samples for S/TEM

2.2 S/TEM sample preparation

The TEM grids were prepared by taping the grids on the aluminum foil and/or on a filter of the
Dekati ELPI+ substrate. Samples were collected on carbon-coated TEM copper grids of 200,
300, and 400 mesh. These grids have a Formvar film support that prevented shaking during speci­
men preparation. Grid removal was challenging as there were multiple grids on the foils as shown
in Figure 1a. Preparing the samples involved multiple steps. In order to reduce the particle boun­
cing, the aluminum foils were greased. Once grease was applied to the foils in the stencil, the foils
were ready to be placed on the collection substrate. At this stage, the TEM grids already had
been taped to the foil. The collection plate holder held the foil on the substrate, which sat on the
jet plate. The cleaning process was a tedious task after each measurement. In order to facilitate
the sampling process, spare set of substrates, TEM grids were prepared on surface and replaced
underground during the test. Figure 1b shows this process. Sampling time was kept short (30
seconds to 60 seconds) to avoid overloading substrates and overlapping of particulate matter on
the TEM grids and the consequent adverse impact on microscopic analysis.

Figure 1. A) S/TEM sampling grids taped on the 25-mm aluminum foil. The soot traces are visible. b)
ELPI+ components disassembled for exchanging the substrates between the tests.

121
2.3 Sample collection
The measurements and sampling were conducted downwind of a 300m long active work zone
shown in Figure 2. The measurement and sampling station was located between of the point
where four ventilation drifts merged and the flow regulator.
The total flowrate through the zone was 26 – 33 m3/s (55,000 – 70,000 cfm). During the
measurement and sampling periods, the type of vehicles and corresponding activities were
logged. During that period, the relatively large number of light-duty (LD) vehicles was used to
deliver material and manpower . The heavy-duty (HD) equipment which was concurrently
used in the work zone is listed in Table 2.

Figure 2. Schematic of work zone and measurement/sampling station.

Table 2. List of HD diesel equipment.

Equipment Equipment Equipment EPA Engine Manufac­
# Manufacturer Type Type 2 Tier turer and Model Control Strategy

1 Wagner LHD HD Permissible Pre CATERPILLAR Disposable Filter Element

3306 PCNA (Donaldson P60451)
2 Fletcher Shield HD Non- 3 CUMMINS Active Regeneration
hauler Permissible QSB6.7
3 Fletcher Shield HD Non- 3 CUMMINS Active Regeneration
hauler Permissible QSB6.7
4 Sandvik LHD HD Permissible 2 CATERPILLAR Disposable Filter Element
3126B HEUI (Donaldson P60451)
5 Eimco LHD HD Non- 3 CUMMINS DOC
Permissible QSB4.5
6 Eimco LHD HD Non- Pre CATERPILLAR No Filteration
Permissible 3306 ATAAC

The ELPI was used for continuous real-time monitoring of number concentrations and size
distribution of aerosols caried by ventilation air to the measurement/sampling station. Con­
currently, the ELPI+ was used collect the size segregated samples for S/TEM and EDS ana­
lyses. The sampling station set-up are shown in Figure 3 a and 3b.

3 RESULTS AND DISCUSSION

The results from six tests were used to study the impact of equipment activities on the aerosols
number and size distributions. The vehicle log showed that the exhaust of the over 35 LD and
6 HD vehicles contributed to the aerosols concentrations. The filtration system consisted of

122
Figure 3. Sampling station set-up. a) ELPI and ELPI+, b) Sampling station.

DFEs, tier 3 engine with diesel oxidation catalyst convertor (DOC) and tier 3 engine with
active regeneration system (SMF-AR). Number and mass concentration measurements
The number and mass concentration of aerosols were measured by ELPI instrument. The
results are for all six tests are shown in Figure 4. The number and mass concentrations fluctuated
substantially during all periods. This appears to be the consequence of diversity in the tasks per­
formed by the HD and LD equipment. During Test 1, two HD pieces of equipment retrofitted
with SMF after treatment systems and several LD equipment were used most heavily. As shown
in Table 2, the sixe HD vehicles included those with DFP filtration systems and SMF-AR. The
high mass concentrations of predominantly micron size aerosols is attributed to high activity of
the SMF-equipped vehicles that exhausted relatively low number concentrations of submicron
aerosols but entrained substantial quantities of micron-sized dust. The results for test 2 – 6 show
the relatively higher number concentrations of submicron aerosols along with proportionally high
mass concentration of submicron aerosols primarily attributed to the diesel emissions of the HD
and LD vehicles operated upwind and in the zone.

Figure 4. ELPI number and mass concentrations of aerosols measured.

3.1 ELPI and S/TEM number and mass size distributions

The results of SEM (Scanning Electron Microscopy) analysis show that the aerosols were dis­
tributed in two or three log-normal modes. The modes consisted of aged agglomeration mode

123
aerosols with count median diameters (CMD) ranging from 160 nm to 210 nm and freshly gen­
erated aerosols mainly in nucleation modes and partially in agglomeration modes with CMD
ranging from 27 and 166 nm. The mechanically generated particulates (coarse particulate) con­
tributed substantially less to number concentration compared to combustion generated aerosols.
The number concentrations were noted to be higher when the diesel equipment was operated
closer to the sampling station. During all six sampling periods, presence of aged aerosols trans­
ported from the outby areas was noted. The number distributions of aerosols at the sampling
station represented a mixture of aged aerosols transported with ventilation air from outby areas
and fresh aerosols generated in the zone. These distributions were mostly bimodal.
ELPI+ was used to collect TEM grids to conduct microscopic morphological and image
analysis. Thirty-five samples were collected during six sampling periods. S/TEM analysis was
conducted to obtain Dprojected for the collected agglomerates. The results show that size distri­
bution obtained from S/TEM analysis are single modal and normally distributed. Three
counts of coarse agglomerates (larger than 1um) were sampled.
The mass size distributions were dominated by entrained dust generated by movement of
diesel-powered and battery powered equipment. The mass median diameter (MMD) ranging
from 2.6 µm to 30 µm was measured. The Mass concentrations were mostly normally distrib­
uted except in two cases. The diesel particulates contributed to the second modes. In the case
of coarse mode, the lognormal distribution was obtained by extrapolation and fitting the log-
normal curves (Chimera Technologies 2009). Figure 5 shows the normalized mass size distri­
bution during Test 6. The high spike at 2056 s was noted to be the result of entrained dust by
an equipment operating close to the sampling station. Figure 5 a and b show size and mass
distribution ELPI measurement and Dprojected STEM results.

Figure 5. A) ELPI size distribution and Dprojected from STEM analysis. b) mass size distributions
measured at the different instances during the Test 6.

3.2 Fractal dimension and shape factor

The fractal morphology of agglomerates collected during sampling was also analyzed. The
samples were collected during six sampling periods in which conditions varied based on the
engine operating conditions, accumulation and sampling conditions. In this study fractal
dimension was calculated based on the maximum length (Lmax). Figure 6 show the fractal
dimension. Fractal dimension is represented by the slope of power relationship between
number of primary particles in the agglomerates versus Lmax/d on the logarithmic scale. Sug­
gested values by Oh and Sorenson (1997) were used to account for the primary particles over­
lapping parameter (d), values of a=1.19 and kα = 1.81 used. Koylu et al. (1995) suggested
values of a=1.09 and kα = 1.21 to calculate the fractal dimension for the agglomerates. The
results show that the fractal dimension vary from 1.87 to 2.04 which is higher than what was
calculated in the sampling zone (except in the case where LHD was intentionally measured

124
without DFE). These results therefore show that larger entrained particles and aged agglomer­
ates have contributed to larger fractal dimensions.
Shape factor is used to describe the agglomerates elongation. If shape factor is close to 1.0,
it indicates that the agglomerates are spherical, where 0.1 shape factor is an indicative of long-
chained or elongated agglomerates (Nord et al., 2004; Mustafi and Raine, 2009). The results
show a majority of agglomerates with shape factor of higher than 0.5. Figure 6 (a) and (b)
show the results for fractal dimension and shape factor.

Figure 6. A) Fractal dimension of aerosols b) Results of shape factor analysis.

3.3 Carbon-rich primary particle types

S/TEM image analysis was conducted on 290 agglomerates collected using the ELPI+ muti-stage
impactor. The results showed that 78 percent of analyzed agglomerates (227 counts) had primary
particle with graphitic layered (onion shell) structure as in Figure 7 (a). Another 12 percent (35
counts) of agglomerates had chain-like agglomerates with amorphous carbon-rich primary par­
ticles as in Figure 7 (b). In addition, 10 percent (28 counts) of agglomerates were chain-like
agglomerates with small catalyst particles attached to primary particles. We observed that some of
the agglomerates had one carbon rich primary particles where others had multiple nuclei centers.
This could be due to the combustion process during which they were formed. Chen et al. (2005)
explains the generated primary particle nuclei centers during post-combustion processes could
increase in size by condensation of volatile compounds. They further explained that primary par­
ticles could collide with others during accumulation mode to form larger agglomerates.

Figure 7. TEM BF image of primary particles (a) with graphitic structure with multiple nuclei centers
and (b) amorphous structure.

125
 The diameter of carbon-rich primary particles dp are normally distributed for agglomerates
with graphitic structure with mean value of 23.6 nm and s = 6.3 nm and for the agglomerates
with catalyst with mean value of 24.2 nm and s = 6.2 nm. The dp agglomerates with amorph­
ous structure were bi-modal.

3.4 S/TEM analysis of volatile agglomerates and EDS results

Primary particles emitted from combustion engines can be categorized into two types: volatile,
and solid particles (Mathis et al. 2004). Morphological analysis at the interface of dust and
DPM particles show some volatile agglomerates attached to stable solid particles. The larger
agglomerates were observed at the higher stages of ELPI+. EDS analysis results depicts non-
diesel volatile interface where particles evaporated after exposure to electron beam for 20
seconds, as shown in Figure 8 c and d. The larger chain-like agglomerate (Figure 8 a, agglomer­
ate #2) seemed more stable than the smaller of the two (Figure 8 a, agglomerate #1).
The morphological and EDS analysis results were also used to identify the source of each
particle. The SE (Figure 8b) image showed that the larger particle is thicker and denser com­
pared to agglomerates #1 and #2. The EDS results indicate the large agglomerate contains
high counts of Sodium (Na) and low counts of Potassium (K), Magnesium (Mg), Aluminum
(Al), Silicon (Si). The source of Sulfur (S) and Fe is likely from lubricating oil and diesel fuel.

Figure 8. S/TEM image analysis a) STEM BF of particles at 400K magnification b) SE image of par­
ticles at 300K magnification c) STEM BF of volatile interface at time 0 second d) STEM BF of volatile
interface at time lag 20 seconds.

126
The EDS results show agglomerate #2, is carbon rich. Traces of Na and Si did not continue
to agglomerate #2 unlike S. The EDS result for agglomerate #1 showed the trace of S net
counts was insignificant, unlike agglomerate #2. Similarly, the Si, Al, Mg and Na net counts
were either similar or slightly lower in agglomerate #1. This suggests that agglomerate #1 was
carbon-rich, volatile, and from a different source than #2. The morphological S/TEM analysis
of the particles mentioned above showed a graphitic structure for the primary particle of
chain-like agglomerate #2. However, an amorphous structure was observed for primary par­
ticles in agglomerate #1.
The TEM and SE images of agglomerate #2 show small particles on the primary particles.
The EDS results indicate Na and S elements surrounding the primary particles. The diameter
of primary particles for agglomerate #2 ranged from 23 nm to 50 nm. For agglomerate #1,
the primary particles appear to be smaller than 30 nm. Agglomerate #2 with the Lmax of
642.81 nm, width of 379.65 nm and calculated SF of 0.59 seemed to have similar characteris­
tics to the samples collected in sampling zone. Agglomerate #2 therefore appears likely to be
a product of the diesel combustion process from equipment in the sampling zone. The
attached Na particles represent the dust in the uncontrolled zone whereas S is likely to be
from the engine. As for agglomerate #1, projected measurements were difficult to obtain as
some of the volatile primary particles evaporated while EDS mapping was underway. How­
ever, the combined results and observations of morphological and EDS mapping make it
likely that agglomerate #2 did not have a combustion source.

4 CONCLUSION

The present study was designed to evaluate importance of monitoring and sampling the
aerosols emitted by diesel-powered equipment performing multiple tasks in the section of
underground mine. The study showed that entrainment of the micron sized dust due to
movement of diesel-powered equipment is the primary contributor to mass concentra­
tions of particulates at the section. The mass median diameter of entrained dust was
found to range between10 µm to 30 µm. The dominant sources of the submicron aerosols
were various types of LD and HD diesel-powered equipment. Diesel aerosols were found
to be distributed between nucleation and agglomeration modes. We observed that fresh
aerosols were distributed in single mode and aged aerosols were distributed bimodaly.
The electron microscopy analysis of size segregated aerosols samples revealed that diesel aero­
sols agglomerates consist of three types of carbon-rich primary particles. with the majority of
those are made of carbon-rich primary particles with graphitic structure. The smaller part is made
of carbon-rich primary particles with amorphous structure or carbon-rich primary particles
infused with catalyst. The diameter of primary particle with amorphous structure is bi-modally
distributed.
The EDS results along with morphological attributes were used to investigate chemical compos­
ition of the fractal agglomerates attached to the micron size particles. EDS results show that the
diesel chain-like agglomerates with high carbon to oxygen ratio and significantly low net counts
of dust and mining environment elements such as Na, Si, Fe, K and S. Volatile chain-like agglom­
erates with amorphous primary particles seemed susceptible to beam damage. EDS results for
these agglomerates showed high C net counts and equal or slightly lower Na, Si and K.
The EDS results show that the typical elements (Na, S either from fuel or not, Si) were
observed in lower stages of ELPI+. The elements were observed at very low net counts during
control zone test and higher net counts during regular mining activity due to the presence of
dust. High C/O net counts ratio for chain-like agglomerates varied from 6.5 to 9.6. Most of the
chain-like agglomerates appeared to have chain-like primary particles with graphitic structure.
The interaction of dust and volatile chain-like agglomerate was investigated at higher stages of
ELPI+ multi-stage impactor. It was observed that amorphous carbon rich particles tend to be
present at the interface of dust and chain-like agglomerates. The primary particles at the interface
with large particles were sensitive to beam damage and evaporated during EDS mapping.

127
 The sampling during mining activity provided valuable data to compare the results with
controlled zone sampling and investigate: i. The interaction of carbon rich agglomerates at the
interface with large dust particles. ii. The morphological differences between the agglomerates.
iii. Investigate the volatile agglomerates observed during the sampling to determine the
sources of agglomerates.

5 DISCLAIMER

The findings and conclusions in this manuscript are those of the authors and do not necessar­
ily represent the official position of the National Institute for Occupational Safety and Health
(NIOSH), Centers for Disease Control and Prevention (CDC). Mention of company names or
products does not constitute endorsement by NIOSH or CDC.

REFERENCES

Bugarski, A.D., Janisko, S.J., Cauda, E.G., Patts, L.D., Hummer, J.A., Westover, C., Terrillion, T. 2014.
Aerosols and criteria gases in an underground mine that uses FAME biodiesel blends. Ann. Occup.
Hyg. 58(8):971–982. http://dx.doi.org/10.1093/annhyg/meu049.
Bugarski, A.D., Hummer, J.A., Vanderslice, S., Mischler, S. 2020. Contribution of various types and cat­
egories of diesel-powered vehicles to aerosols in an underground mine. Journal of Occupational and
Environmental Hygiene 17: 121–134. https://doi.org/10.1080/15459624.2020.1718157.
NIOSH 2016. Monitoring diesel exhaust in the workplace. In: NIOSH Manual of Analytical Methods
(NMAM), 5th Edition, Chapter DL, Cincinnati, OH: U.S. Department of Health and Human Services,
Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health. Avail­
able from: https://www.cdc.gov/niosh/nmam/pdfs/NMAM_5thEd_EBook.pdf. Accessed on July 2, 2022.
Zielinska B, Sagebiel J, McDonald J, Rogers CF, Fujita E, Mousset-Jones P (2002). Measuring diesel emis­
sions exposure in underground mines: a feasibility study. Research directions to improve estimates of
human exposure and risk from diesel exhaust, Special report. Boston: Health Effects Institute, pp. 181–232
Kittelson, 1998, D. B. Engines and nanoparticles: A review. J. Aerosol. Sci. 1998, 29 (5-6), 575–588.
Mathis, U., Kaegi, R., Mohr, M., Zenobi, R. (2004). TEM analysis of volatile nanoparticles from particle
trap equipped diesel and direct-injection spark-ignition vehicles. Atmospheric Environment 38 (2004)
4347–4355
Nord, K., Haupt, D., Ahlvik, P., and Egeback, K.-E. (2004). Particulate Emissions from an Ethanol
Fueled Heavy-Duty Diesel Engine Equipped with EGR, Catalyst and DPF, SAE Paper 2004-01-1987

128
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Development and evaluation of innovative diesel particulate filter

technology

J. Stachulak
MIRARCO Mining Innovation, Sudbury, Ontario, USA

B. Rubeli & D. Young

CanmetMINING, Ottawa, Ontario, USA

K. Watson & B. McLean

Vale, Copper Cliff, Ontario, USA

ABSTRACT: The introduction of diesel particulate filters (DPF) technology has been an
important step towards controlling diesel emission from heavy-duty diesel -powered under­
ground mining vehicles. The Johnson Matthey CRT® (Continuously Regenerating Trap)
technology that combines a diesel oxidation catalyst (DOC) with a DPF has be extensively
used to trap harmful diesel aerosols, CO and hydrocarbons from on-road vehicles. In order to
address concerns over adverse effects of CRT® system on NO2 emissions and allow for use of
these systems in underground mining environments, Johnson Matthey (JM) developed, the
Mining CRT® system which integrates CRT® technology with NO2 abatement components.
The effects of the system on particulate and gaseous emissions were evaluated in the engine
dynamometer laboratory. The evaluation was done using Caterpillar C11, Tier 3 engine oper­
ated over the ISO 8178, 8-mode test cycles. The engine was fueled with an ultra–low-sulfur
diesel fuel. The system was found to effectively reduce total particulate mass and number,
total hydrocarbon, and CO emissions. The novel fuel injection and NO2 decomposition cata­
lyst were found to be effective in preventing NO2 slip out of the system.

1 INTRODUCTION

This study evaluated the emissions from a Caterpillar C11, Tier 3 engine, obtained with
and without the application of JM Diesel-Particulate-Filter Low NO2-CRT® emission con­
trol device. The objective of this evaluation was to provide information on the emission
performance of the DPF for application in Underground mines where increase of NO2 in
the mine environment is undesirable. The testing was performed at the Diesel Emissions
Research Laboratory at CANMET- MMSL, Ottawa. The emission testing was performed
on an engine dynamometer under controlled conditions using the ISO 8178-C1, 8-mode test
cycle.
Measurements were made for the basic engine parameters and exhaust gas concentrations
were measured for carbon monoxide (CO), carbon dioxide (CO2), oxygen (O2), nitric oxide
(NO), oxides of nitrogen (NOX), total hydrocarbons (THC), and diesel particulate matter
(DPM). Measurements were also made for nano-particle distribution for each 8-mode cycle.
Several steady state and transient test cycles were performed to calibrate and fine tune the fuel
injection protocol for emission reduction device.
The device tested showed a significant reduction in exhaust CO (96%), THC (58%), NO2
(55%) and nanoparticle DPM (99%) while CO2 remained nearly unchanged.

DOI: 10.1201/9781003429241-14

129
 This study was a prelude to an extensive field evaluation in underground operation at
Vale’s Copper Cliff Mine in Sudbury. The JM’s Mining CRT was installed on Caterpillar
1700 LHD, 263 kW C11 Tier 3. The Mining CRT acquired 7380 hours without the need for
cleaning and replacement while maintaining NO2 emission at engine-out level (Stachulak
et al., 2022). The lessons learned from this successful application of Mining CRT technology
is not part of this paper and will be a topic of the separate publication.

2 EXPERIMENTAL APPROACH

The engine used for the testing was a rebuilt Caterpillar C11, Table 1 provides some engine
specification data

Table 1. Test engine specifications for model C11, Tier 3.

Make Caterpillar

Model C11, Tier 3

Serial number RSX05190
Displacement 11 Liter
Rated power, gross 353 hp @ 1800 rpm
Fuel rate at rated power 118.5 lb/hr
Peak torque 1089 lb.ft @1600 rpm
Peak torque speed 1600 rpm
Aspiration Turbocharged
Fuel system DI, Electronically controlled fuel injection
Max air intake restriction 15” H2O
Max exhaust backpressure 40” H2O
Low idle speed 700 rpm
High idle speed 2120 rpm

The diesel fuel used for this study was an ultra-low-sulfur diesel fuel with a sulfur value of
11 ppm (CAN/CGSB-3.517-2000). Some of the laboratory analyzed fuel properties are given
in Table 2.

Table 2. Some diesel fuel properties.

Properties ASTM test method Analysis

Specific gravity 60/60F D4052 0.8248

Density @ 15°C, kg/m3 D4052 824.1
Carbon, wt% D5291 86.32
Hydrogen, wt% D5291 13.94
Nitrogen, wt% D5291 <0.3
Flash point, °C D93 54.1
Sulphur, ppm D7039 10.8

2.1 Gases and DPM control device

JM supplied a Diesel-Particulate-Filter Low NO2-CRT® emission control device for this
study. The device was tested after it was tuned for optimum performance.
Prior to testing, the device was installed as shown in Figure 1.

130
Figure 1. Diesel-Particulate-Filter Low NO2-CRT® emission control device and Caterpillar C11 diesel
engine installed in the CANMET-MMSL test cell.

The device was calibrated by the manufacturer’s representative according to both steady
state and transients’ tests.

2.2 Test procedure

The engine exhaust emissions were measured at all modes of the ISO 8178-C1 test cycle. The
8-Mode test cycle (ISO 8178-1, 1996) for the Caterpillar C11 engine is defined in Table 3. The
integrated 8-Mode average values for all tests were calculated using the appropriate weighting
factors for each mode.

Table 3. ISO 8178-C1 8- Mode test cycle.

Mode # 1 2 3 4 5 6 7 8

Engine Speed, rpm 1800 1350 700

Torque, % 100 75 50 10 100 75 50 0
Weighting factor 0.15 0.15 0.15 0.1 0.1 0.1 0.1 0.15

2.3 Gaseous emission measurement

The raw exhaust gas concentrations were measured using a California Analytical Instruments
(CAI) gas cart. The gas cart consists of an exhaust gas sampling and conditioning system,
emission analyzers, 64-point gas divider for system calibration, and NOx efficiency tester. The
gas cart contains the following gas analyzers:
1. Low range carbon monoxide (CO), model CAI 100 IR, non-dispersive infrared (NDIR)
detection system, span ranges 0-100 ppm and 0-2000 ppm.

131
2. Carbon dioxide (CO2) and high carbon monoxide (HCO) model CAI 300 IR, non- disper­
sive infrared (NDIR) detection system, CO2 span ranges 0-5.0% and 0-20.0%, and CO
span range 0-1.0%.
3. Oxygen (O2), model CAI 300 Oxygen, paramagnetic (PMA) detection system, span ranges
0-5% and 0-25%.
4. Oxides of nitrogen (NO/NOx), model CAI 400 HCLD heated, chemiluminescence
(CLD) detection system, span ranges 0-1000 ppm and 0-3000 ppm. The gas cart has two
NO/NOx analyzers that provide simultaneous measurements of NO and NOx. The con­
centration of NO2 is determined by the difference of NOx and NO concentrations.
5. Total hydrocarbons (THC), model CAI 300 HFID, heated flame ionization detector
(HFID), span ranges 0-100 ppm and 0-1000 ppm.

2.4 Particulate measurement

The Diesel Particulate Matter (DPM) was measured gravimetrically using a Sierra BG-2 par­
ticulate partial flow sampling system. This is a fully automated test stand with a micro-
dilution chamber which draws a sample of exhaust gas directly from the engine exhaust stack
and dilutes the entire sample fraction. The system’s two mass flow controllers provide the
required control of flow rates and dilution ratios.
The particulate sample is collected on dual Pallflex T60A20 (Teflon coated glass fibre) 90-mm
filters. All filters are conditioned in an environmentally controlled balance room with air condi­
tions of 22 ± 3°C temperature and at 45 ± 8% relative humidity for at least one hour before
weighing. After conditioning, the filters were weighed and placed in stainless steel filter holders.
After particulate sample collection using the BG2 micro- dilution system, the filters are returned
to the controlled balance chamber for re- conditioning and weighed again to determine the DPM
mass. The DPM concentration is then calculated using the sample mass, exhaust gas sample flow
rate and sampling duration.

2.5 Nano-particle measurement

A Thermo Scientific Inc. scanning mobility particle sizer (SMPS) 3080 spectrometer was used
to measure the size distribution of diluted diesel exhaust in the size 14 nm to 710 nm distribu­
tion. Particles were classified with an 3081 SMPS electrostatic classifier and their concentra­
tion was measured with a TSI 3025A condensation particle counter (CPC). Diluted diesel
exhaust aerosol was delivered to the SMPS from a Nano-Met raw gas dilution system which
maintained sample temperature and dilution control during the test.

2.6 Diesel-particulate-filter low NO2-CRT®

The Diesel-Particulate-Filter Low NO2-CRT® was installed and operated by the manufac­
turer representative. The regeneration of the filter is by oxidation of carbon through NO2.
Figure 2 below shows the schematic layout of JM Mining CRT® system. After entering the
system, the exhaust gas comes into contact with an oxidation catalyst that reduces CO and
HC levels, and also oxidizes a part of the NO produced from the engine to NO2. The exhaust
then passes through a filter that removes diesel particulate at a high efficiency. The NO2 sub­
sequently reacts with the soot on the filter at normal exhaust operating temperatures to give
a continuous regeneration and low operating back pressure. After the filter a small amount of
diesel fuel is added that reacts with any excess NO2 on a special catalyst to convert it back to
NO. This gives a system that effectively removes particulate with a continuous passive regen­
eration mechanism while minimizing levels of NO2 that might cause a problem in the confined
spaces of a mine.

132
Figure 2. Diesel-particulate-filter low NO2-CRT®.

3 TEST RESULTS

Table 4 provides 8-Mode integrated average values of specific emission data for gases
and DPM.
The percent reduction in emissions compared to engine baseline data is calculated using the
Equation 1, and the results for all devices (Gangal et al., 2012) are shown in Table 4.

Note: a negative value indicates an increase in emission

The percent filter efficiency for emissions of nanoparticles compared to the base engine is
calculated using the Equation 2, and the results are also shown in Table 4.

Table 4. Specific exhaust emissions for the integrated 8-Mode test cycle, test #7 and #8.
Integrated Specific emissions Baseline Test 8 CRT Test 7 Reduction

CO2 g/hr 86817.3 87687.2 –1.0%

CO g/hr 279.7 11.5 95.9%
NO2 g/hr 33.5 15.1 55.0%
NO g/hr 242.3 250.1 –3.2%
NOX g/hr 275.8 265.3 3.8%
THC g/hr 23.2 9.7 58.0%
DPM g/hr 4.8 2.5 47.7%
SMPS dw/dlogDp 6.9.E+06 7.6.E+04 98.9%

The overall DPM conversion of 47.7% reported over the Low NO2 CRT® system is lower
than expected. DPM removal over a ceramic wall flow filter is usually >90%. This lower than
expected conversion over the filter system could be due to Diesel fuel droplets, derived from

133
the fuel injection that takes place between the filter and the NO2 decomposition system, slip­
ping from the system and then being measured as DPM. This theory is supported by the rela­
tively poor reductions (58%) of THC found over the Low NO2 CRT® system. A THC
conversion of >90% is usually expected over a CRT® system operated at these temperatures
and the higher tailpipe levels found over this “Low NO2” system probably result from the
Diesel that is added after the filter to reduce tailpipe NO2 levels. The high reduction of nano­
particles again suggests that the filter is working efficiently and that the measured mass of par­
ticulate is largely due to condensed Diesel fuel rather than soot from the engine. It is expected
that this level of Diesel fuel slip can be reduced in future tests by system optimization.

4 SUMMARY

The reduction of exhaust gas emissions by JM Diesel-Particulate-Filter Low NO2-CRT® was

evaluated on an engine dynamometer. Caterpillar C11, Tier 3 engine was emission tested over
the 8-Mode test cycles. An ultra–low-sulfur (11 ppm) diesel fuel was used for this testing. The
regulated gaseous emissions and particle counts were measured.
Table 4 shows that JM Mining CRT® system reduced CO by 96%, and nanoparticles by
99% over the 8-Mode test cycle. Moreover, Table 4 indicates that the CRT system reduced the
THC by 58% and NO2 by 55% over the 8-Mode test cycle which is in line with the system
design to keep THC and NO2 lower than engine baseline.

ACKNOWLEDGEMENTS

The authors would like to acknowledge the support and contributions of Vale, Toromont
CAT, Johnson Matthey Ltd. for their help and insight during this work.

REFERENCES

Gangal, M., Rubeli, B., Young, D., Leung, E. & Feres, V. 2012. Performance Evaluation of Diesel Par­
ticulate-Filter Low NO2 CRT®.
ISO 8178-1. 1996. 1996(E) PART 1 AND ISO 8178-4:1996(E) PART 4: RECIPROCATING INTERNAL
COMBUSTION ENGINES - EXHAUST EMISSION MEASUREMENT.
Stachulak, J., Gangle, M., Young, D., Watson, K. & McLean, B. 2022. Performance Evaluation of
Diesel Particulate-Filter Low NO2 CRT. In MDEC Annual Conference. Toronto, Ontario, Canada.

134
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

DPM reduction through emission assisted maintenance -

PT Freeport Indonesia program update

A.A. Habibi
PT Freeport Indonesia, Papua, Indonesia

E. Pinto
Superintendent-Fire and Emission Maintenance, Papua, Indonesia

M. Mardon
Technical Expert-Operations Maintenance, Papua, Indonesia

K. Wijayanto
Maintenance Planner-Operations Maintenance, Papua, Indonesia

C. Rose
Director-Occupational Health and PTFI Health & Safety, Phoenix, Arizona, USA

ABSTRACT: Reducing Diesel Particulate Matter has a direct impact on improving air qual­
ity and thus workforce health and safety. PT Freeport Indonesia has selected a multi-stage
reduction strategy to reduce DPM emissions in its underground operations. This paper pro­
vides an update on challenges, success and lessons learned during the first two stages of the
Emission Assisted Maintenance (EAM) program, including mobile equipment selection for
trials and instrumentation used for tailpipe measurements. The first stage consists of various
Diesel Particulate Filter trials to measure the effectiveness and feasibility of a large scale roll
out. Use of new engine technologies was trialed in the second stage. Additionally, the PT Free­
port Indonesia’s EAM program includes wet and traditional DPF cleaning stations and
a Digital Data Management dashboard to control, monitor and measure success. Lastly, the
paper summarizes the impact of high altitude in engine emissions and DPF life as well as pro­
viding insight regarding the opportunity to scale up the EAM program.

1 INTRODUCTION

Emission Assisted Maintenance programs have been integral to the underground mining
industry efforts to reduce exposure of miners to gases and aerosols emitted by diesel-powered
equipment. The long-term data generated through EAM programs have been used to enhance
preventive and corrective maintenance and maintain diesel engine emissions at Original
Equipment Manufacturers specifications and certification levels. These programs were focus­
ing not only on keeping in-use emissions under control, preferentially at the certification/
approval levels (30 CFR 75.1914, 30 CFR 57.5066) but also to assure effectiveness of OEM
and retrofit control technologies and strategies (Bugarski et al. 2012) over their useful service
time. The collected data from EAM can be used to help ensure engine health in each interim.
These programs are designed to identify significant increases in emissions due to inadequate
engine maintenance, engine and exhaust aftertreatment aging, system malfunctioning, poison­
ing components of an exhaust aftertreatment system, and potentially tampering with engine
and exhaust aftertreatment systems (BHP Billiton 2005, McGinn 2010). The EAM programs

DOI: 10.1201/9781003429241-15

135
reduced the DPM level in the underground mines and proved to be useful in reducing poten­
tial downtime needed to accommodate unscheduled repairs and improve availability of diesel-
powered equipment and viability of the underground mining operations.
The implementation of complex engine end exhaust aftertreatment systems in support of
low emissions requirements introduced with EPA Tier 4f/EU Stage IV and EU Stage
V standards resulted in an increased demand for advancements in maintenance practices and
embracing new technologies. Under circumstances where the underground mining diesel fleets
due to implementation of control strategies and technologies contribute less and less pollu­
tants to the underground environments, a small fraction of the poorly maintained fleet may
contribute a growing and disproportionate share of the overall pollutants. Therefore, the role
of EAM programs is as important as ever.

1.1 EAM background information

The Emission Assisted Maintenance programs developed in the 1980s and 1990s (Waytulonis
1987, Spears MW 1997) gradually have evolved to address new considerations that have emerged
with advancements in diesel engine and exhaust aftertreatment technologies. The programs that
were initially developed for mechanically controlled, naturally aspirated heavy-duty engines were
gradually enhanced to address maintenance of electronically controlled, turbocharged engine and
retrofit type exhaust aftertreatment devices (BHP Billiton 2005, McGinn 2010, Hines 2019]).
One of the major limitations of existing EAM programs is the small set of emissions data that
can be generated for the engines installed in the mobile machinery. A short list of engine operat­
ing conditions that are used in EAM programs include low idle, high idle, torque of hydraulic
converter stall for the engines equipped with automatic transmission and hydraulic pumps.
Obtaining pertinent data from the engine management system using SAE J1587 or J1939
data protocols (Cummins 2016) could provide underground operators with highly sought
information. This information reflects the engine condition while equipment is operating
under various duty cycles. The additional data from measurements with portable emissions
measurement sensors (Gallus et al. 2016, Giechaskie et al. 2018, Yang et al. 2018), on-board
sensors (e.g. NOx) (Demirgok 2018), or other instrumentation (Kadijk et al. 2016) could also
greatly enhance the quality of data used to support EAM programs.

1.2 PTFI EAM program goals and challenges

The PTFI mining district is located in Papua, Indonesia. It consists of three active mining
operations: Grasberg Block Cave (target 130-160 kt/d ore), Deep Mill Level Zone (target 80
kt/d), and the Big Gossan open stoping operation (7 kt/d). The Kucing Liar operation is
a new addition, a block cave, to the mining district. Figure 1 shows the mining district. DOZ
block cave mining has been completed and is currently closed.
Although the mining methodology is similar throughout the underground, the mobile
equipment (MEQ) type, elevation and operating conditions vary from one site to the other.
For example, DMLZ utilizes R1600H and AD55 as primary load and haul units, whereas
GBC utilizes R1700 LHDs and an electric train. BG mine primary load units are R2900 load­
ers. There are over 1,500 pieces of underground diesel equipment at PTFI. This large fleet can
be categorized as 15 different types of equipment. Although load and haul units are assigned
properties, the support and ancillary equipment may travel from one mine to the other.
The following are the main objectives of PTFI EAM program:
• Improve Health and Safety of working area by reducing DPM concentration in under­
ground working areas
• Improve awareness of the importance of EAM programs among maintenance and
operations
• Identify the challenges of scaling up EAM for a large number of fleet vehicles
• Identify and improve field appropriate tailpipe measurements for old and new engine
technologies

136
Figure 1. PTFI mining district overview.

2 EAM PROGRAM DETAILS

The EAM program is an entity under the PTFI Air Quality Improvement umbrella. The pro­
gram is designed to tackle DPM challenges. The committee consists of all engaged stakeholders
as shown in Figure 2. The committee was formed in 2016, with initial tender for the Stage 1 trial
in 4Q 2020 and the official trial starting in 4Q 2022. Figure 2 shows the PTFI air quality matrix.

Figure 2. PTFI air quality initiatives.

The DPF trial covered multiple stages. Stage 1 included trialing passive and active DPFs.
The approach allowed the researchers to understand the difficulties and challenges and
troubleshoot. Stage 2 consisted of trialing multiple Tier 4f.

2.1 DPM Committee

Formed in 2016, the DPM reduction at source initiatives is managed by the DPM Committee.
The committee consists of multiple stakeholders including Operations, Maintenance, EHS,

137
Supply Chain, ventilation and support advisors from corporate. In addition, OEMs and DPF
vendors are included in the committee via dotted line.

Figure 3. PTFI DPM committee hierarchy.

In addition to the members shown in Figure 3, the committee regularly updated the mem­
bers of PTFI up to Executive level on successes and failures. It is important to note that direct
support from senior management has been vital to overcoming major obstacles. The support
consists of Management of Change for field work, data processing and funding. Otherwise,
this project could have been viewed as a disruption.

2.2 Equipment selection

PTFI selected seven types of equipment. The step approach was chosen to be able to facilitate
the size of the trial. Due to the size of the PTFI fleet, a smaller group of DPFs were purchased
and installed on selected equipment. Figure 4 shows PTFI underground engines in operation.
The committee used simulation and available tailpipe measurements to determine the highest
polluters for Stage 1. Dynamic contaminant simulation analysis was used to take distance,
emission time and ventilation into account to determine the equipment which would deliver
more reduction to overall DPM concentrations at PTFI.
Heavy duty (HD) load and haul units were trialed in Stage 1. Table 1 shows the list of
trialed HD equipment, engine information and corresponding trialed DPFs from different
suppliers. Three Light Duty (LD) vehicles also were selected due to the importance of this
reducing their emissions. At the beginning and end of each shift, hundreds of LD Toyotas
commute in and out of the mine. The impact to the long stretch of the entrance portal was
projected as 15 percent of the overall contribution there. The simulation inputs were actual
fuel consumptions and field tailpipe emissions to ensure reliable output. DPF A, B and C are
passive, and DPF D was sintered metal DPF with an active regeneration system.

2.3 Trial methodology and DPF selection

2.3.1 Trial methodology

DPFs were installed on selected equipment along with an aftermarket datalogger. The dataloggers
collect information on temperature and backpressure during operation. The elevation and duty
cycle for some of the trial equipment varied due to the nature of the tasks. The load and haul unit
operated mainly with a high duty cycle and lower idling time; however, the support equipment

138
Figure 4. PTFI UG fleet engines.

Table 1. Heavy duty mobile equipment list.

MEQ Engine Model Certification Engine (kW) Capacity (L) Trialed DPF
LHD R1700G C11 VR/Tier2 270 11.1 DPF A
LHD R1600H C11 VR/Tier2 208 11.1 DPF A
LHD R2900G C15 VR/Tier2 305 15 DPF A & DPF B
TR AD30 C15 VR/Tier2 305 15 DPF B
TR AD55 C18 VR/Tier2 439 18 DPF A
Getman A64 QSB 6.7 Tier 3 168 6.7 DPF C & DPF D

(such as Getman emulsion, flat bed and mixer trucks) showed lower temperatures due to high
idling time. The trial planned to run the DPFs until they reached their first cleaning cycle. The
goal for LD vehicles was set for 4,000 km and for HD equipment at 3,000 engine hours.

2.4 DPF selection

The DPM committee selected DPFs from four vendors. Vendors A, B and C supplied trad­
itional passive DPFs equipped with Diesel Oxidation Catalyst (DOC), and only Vendor
D could supply the DPF with an active regeneration system. Both cordierite and Silicon Car­
bide substrate filters were used due to their high efficiency and capability of capturing more
than 95 percent of particulates. All trialed DPFs were Wall Flow Filters with a SiC core to
maximize DPM reduction. However, this decision added complexity to the project. A lack of
adequate exhaust temperature results caused high back pressure. Excessive thermal runaways
also could have created molten substrate and high DPM concentrations. Analyzing the tem­
perature profile and DPM concentrations could identify potential regeneration issues.

3 TAILPIPE MEASUREMENT AND INSTRUMENTATION

3.1.1 Gas analyzer

The first measurements created a baseline for each piece of equipment. The committee
adapted the use of ECOM J2KN Pro for trial units instead of Drager EM200-E because of its
ability to measure CO2, O2, and exhaust temperature in addition to CO, NO and NO2. CO2

139
and NOx are main indicators of diesel combustion. Because most legacy engines are not
equipped with On Board Diagnostic systems, a regular tailpipe measurement was conducted
on all trial units. This data was used to ensure the health of the engine and DPF/DOC as well
as indicators of any physical damage to the DPF substrate. The low and high idle measure­
ments were collected for LD vehicles, and low, high and stall measurements were taken for
HD equipment.

3.1.2 PM measurements
For DPM measurements, PTFI is using the Pacific Data System analyzer (DPM-RT) model 1
and 2. The instruments operates on light-scattering principles to estimate mass of particles.
DPM-RT 2 is equipped with a touch screen display and logging capabilities. PTFI EAM tail­
pipe measurements records the DPM concentrations before and after the DPF/DOC package.
The procedure is used based on instrument manufacturer recommendation. The 60 seconds
procedure consists of 20 seconds low idle, 20 seconds high idle (3200 rpm for LV and
2300 rpm for HD) followed by the last 20 seconds low idle. The device provides and average
results for the 60 seconds, as well as individual readings.
In addition, DPM committee has acquired Nanoparticle Emission Tester model 3795
(NPET) manufactured by TSI. This instrument is to be used on new engines (T4F/Stage V).
NPET uses the principle of a condensation particle counter (CPC) to measure particle concen­
tration in the tailpipe. Assuming a mean particle diameter, the particle number concentrations
can be converted to mass concentrations (TSI 2021). The decision was due to repeated zero
DPM-RT tailpipe measurements indicating low mass concentrations and high particle
number concentrations.

3.2 EAM DPM dashboard

The PTFI EAM program includes a Power Bi dashboard to process and analyze all the tail­
pipe measurement data. A Digital Form was used as the data entry platform to enter gas emis­
sions, engine hours and/or odometer and DPM concentration. Other studies have shown
a reasonable linear correlation between CO concentrations and DPM measurements for old
technology engines (Spears 1997). Limited literature is available for the newer engine tech­
nologies. However, Hallquist, et.al. 2012, showed that CO can still be used as a surrogate on
an Exhaust Gas Recirculation (EGR) equipped engine. The study showed a linear correlation
between CO and DPM measurements. Although it is expected to measure higher CO concen­
trations as a result of EGR, similar increases can be correlated to PM emissions.
The results show reasonable accuracy in field measurements. However, the Torque Conver­
tor Stall test results were susceptible for several reasons as the nature of the test could produce
different results based on the operator and the load. In order to bring some consistency to the
gas emission measurements, the DPM committee opted to utilize CO2 emissions for consistent
measurements.

3.2.1 Dashboard layout and KPI parameters

The PTFI EAM dashboard is intended to assist the maintenance workforce with proper prac­
tices. DPF units are exposed to a harsh environment due to dust particles. For legacy engines,
DPF, back pressure and temperature sensor components are all prone to getting damaged.
This is even more challenging for a modern engine after-treatment package.
Figure 5 shows the first revision of the DPM dashboard overview page. The page consists
of multiple report windows covering all parameters of the selected equipment. Windows 1 and
4 depict measured DPM per kW and concentration of DPM measurements, respectively. Win­
dows 2, 3 and 5 represent the trends for low, high and stall tests. All reports can be filtered
based on gas type (O2, CO2, CO, NO, NO2, NO, NOx), engine conditions (back pressure,
boost pressure) and environmental effects such as barometric pressure. Window 6 shows
engine hours/odometer (kilometer). This is helpful to quickly calculate the age of the DPF and
engine. Window 8 shows the tabulated snapshot report for detailed investigation and

140
troubleshooting. Lastly, Figure 9 shows NOx ratio (NO2/NOX) to ensure catalysts are not
adding to NO2 levels in underground workings. The results show NO2 levels are either
decreasing or staying at pre-DPF levels, and the DPF installation has not affected the level
adversely. An example will be discussed in the next section.

Figure 5. PTFI EAM dashboard – overview page, first revision.

3.2.2 Interactive improved dashboard

The DPM dashboard in its current state has proven to be an effective tool for troubleshooting
and Root Cause Analysis. It is also possible to use the BI dashboard to study the results in the
field. However, a delay in data upload could limit the access to the data which would lead to
the field mechanics to determine the information based on their instrumentation.
In other words, a need for the next revision of the dashboard is needed facilitate the deci­
sion-making process via machine learning and processed data. Since all the historical data for
all equipment is stored in the database, EAM scripts can be used to make EAM recommenda­
tions (failure prediction source) in the field at the time of tailpipe emission measurements.
Most scripts will need to be related to CO, NOx, DPM and CO2 concentration. It is noted
that modern engine technologies will need to have a separate process due to their complex
nature of after-treatment.

4 RESULTS AND DISCUSSION

4.1 Light duty vehicles

Results from LD vehicles were positive overall. The trial started with three LD Toyota V8
EURO 4 engines. After initial success, the trial was expanded to 15 vehicles. Following are the
lessons learned from the three-unit trial:
• Two units worked without any issues close to the target distance of 4,000 km
• One unit failed after 1,700 km. The operator shared complaints of low power while in 4th gear.
The DPF was taken out, and the RCA showed a 2 kPa restriction across the substrate.
Figure 6 shows the DPF in question. A micro camera was used to investigate the pores. Ash
plugs were found in several areas on the clean side indicating successful regeneration with an
estimated up to 70 percent plugged pores. The dirty side was entirely plugged. Further investi­
gation showed the engine had been out of tune pre-DPF.

141
• The deterioration of exhaust piping was found after two months on trial units. PTFI has
opted to use stainless steel for all DPF accessories and mufflers.
• It is vital for the engine to be tuned up prior to DPF installation.

Figure 6. Light Vehicle Toyota DPF photos, dirty and clean sides.

4.1.1 Decisions based on the trial outcome

PTFI has identified a combined approach to reduce the emissions from LD vehicles by the
end of 2025. A trial of new Toyota Land Cruisers with modern engines (T4F) showed signifi­
cant improvement on gas and DPM measurements as shown in Figure 7. Toyotas equipped
with T4F engines nearly have reduced CO and DPM by approximately 95 percent. The NOx
concentrations appear to be aligned in all cases and are slightly lower for the T4F Toyota.
PTFI proceeded by scaling the LD DPF installation for another 150 Toyotas. In addition,
140 new T4Fs are planned to be added in 2023. This will continue through 2025 until all LD
vehicles are equipped either with a DPF or T4F engine.

4.2 Heavy duty vehicles

The following are the lessons learned from the trial.
4.2.1 R1700G and R2900
Two DPFs were trialed (DPF A and B) on both LHDs. In the case of R1700G, only DPF
B successfully surpassed the target number of hours. Figure 8 shows the results for two of the
last several readings. The NOx, CO2 and CO readings have been consistent, and the DPM
reduction has stayed low and hardly measurable. The back pressure is noted as low, indicating
the DPF does not require cleaning yet.
In the case of the R2900, both DPF A and B performed to the expectations without causing
any downtime to operations. The reduction in CO and DPM was noted without having any
adverse effect on boost pressure.

142
Figure 7. Comparison of LD vehicles’ gas emissions.

Figure 8. R1700G tailpipe results.

4.2.2 AD55 and AD30 trucks

The haul units as part of this trial were successful. Both units performed without causing any
reportable issues. Although the duty cycle for trucks is different from one to another, it
appears the temperature profile and DPF size have been a good match.

143
• DPF A was installed on AD55
• DPF B was trialed on AD30.

4.2.3 Getman A64

The initial trial attempt was unsuccessful as the DPF C failed prematurely after 120 hours.
Upon completion of investigation, it was decided to redo the trial using a DPF with a larger
DOC. The 2nd attempt was successful, and the machine has been in operation since the instal­
lation. The machine has approached 1,500 hours at the time of writing this paper. DPF D also
will be installed on a Getman unit in 2Q 2023.

4.2.4 R1600H
DPF A failed prematurely repeatedly after only 50 to 100 hours. The investigation showed
high back pressure was recorded; however, temperature readings of over 350 degrees C were
reached more than 75 percent of the time.
Since no evident reason was found, the RCA investigated the impact of elevation on tailpipe
emissions. Tailpipe measurements were collected at sea level and site level (2,500 m to
2,900 m above sea level). The results show a significant increase of tailpipe emissions on CO,
NOx, and CO2 except AD60 and 1700k. Figure 9 shows the percent variance for each emis­
sion. This issue was reported to the OEM engine manufacturer. The DPM committee is
exploring other possible solutions for R1600H in 2023.

Figure 9. Impact of elevation in direct tailpipe emission measurements.

4.2.5 New engine technology

As part of the second stage, PTFI trialed several HD equipment with T4F/Stage V engines.
Four of these were LHDs with complete DPF+ Selective Catalytic Reduction (SCR) after­
treatment packages. The tailpipe measurements showed immediate reduction in CO and PM
and lower NOx levels. As mentioned, the aftertreatment package is rather complex with
higher failure rates which could increase the cost of ownership. All four LHDs prematurely
failed before reaching the acceptable target hours and required small and significant repairs
before going back in service. It is committee’s point of view that although the technology
meets emission standards, it appears the technology still requires improvement, and reliability
is still an issue. PTFI is working with OEM to ensure a successful trial of these units.
Additionally, engine repackage is being trialed on an R1700 loader by replacing the old
engine with a T4F engine. The upgrade is expected to be completed in 1Q 2023.

144
Figure 10. PTFI cleaning station.

5 CLEANING STATION

Due to the remoteness of the PTFI jobsite, PTFI has established an inhouse cleaning station.
The cleaning station is in operation and equipped with traditional and washing solutions. The
committee has opted to trial both approaches individually and in tandem to determine the
impact of each approach to the life of DPFs. Figure 10 shows the cleaning station. A digital
dashboard is required to be built to track the life of DPFs for this purpose. The committee
believes this approach is to be completed by the end of 2023.

6 CONCLUSION

A combination of DPF retrofit installation and modern technology engines were trialed at the
PTFI jobsite. The results from LD vehicles were found to be positive with the DPF lasting up
to 4,000 km. This successful trial resulted in installation of an additional 150 DPFs across the
LD fleet. Additionally, the trial of new T4F Toyota LVs showed significant reduction in CO
and PM concentrations. The PTFI LD fleet is expected to be replaced or be equipped with
DPF by the end of 2025.
The HD DPF trial was more challenging. The trial was successful on AD30 and AD55
trucks. LHD 2900G resulted in positive results from two different DPF vendors. DPF
B worked without any issues on R1700G, and DPF C appears to be the solution for Getman
MEQ. The R1600H trial was not successful, and the DPM committee is exploring other
options as part of 2023 plan.
The trial of T4F HD LHDs shows significant reduction in emissions; however, the reliabil­
ity of MEQ appears to be low, and maintaining the LHD in operational conditions has been
challenging. The T4F LHDs has not been able to achieve the target hours yet. PTFI has been
working with the OEM and is committed to resolve the reliability issues.
The PTFI EAM process is moving forward with a trial on five other types of MEQ. The
program is aiming to tackle more outby equipment, mainly used for service, transportation
and goods delivery. These efforts are aligned with the PTFI five-year DPM reduction forecast
to improve working conditions in the underground.
The PTFI DPM committee consists of members from all departments. It is crucially import­
ant to note that direct support from the highest levels management has been and will be vital
to the success of this project.

REFERENCES

30 CFR 57.5066. Safety and Health Standards. Underground metal and nonmetal mines. Maintenance
standards. Code of Federal regulations. Washington, DC: U.S. Government Printing Office, Office of
the Federal Register.

145
30 CFR 75.1914. Mandatory Safety Standards. Underground coal mines. Maintenance of diesel-powered
equipment. Mine Safety and Health Administration. Code of Federal regulations. Washington, DC:
U.S. Government Printing Office, Office of the Federal Register.
BHP Billiton [2005]. Diesel Emissions Management. Rev.1. BHP Billliton Limited.
Bugarski, A.D., Janisko S., Cauda E.G., Noll, J.D., Mischler, S.E. [2012]. Controlling Exposure - Diesel
Emissions in Underground Mines. Society for Mining, Metallurgy, and Exploration. ISBN-13:
9780873353601. Available from: http://smemi.personifycloud.com/PersonifyEbusiness/Store/Product
Details.aspx?productId=116967 (accessed on February 1, 2021).
Cummins [2016]. Ready for more. Underground. Underground Mining Engines. Cummins Inc.
Demirgok B [2018]. Development of an emissions monitoring methodology using on-board NOx sensors
and revision to current in-use emissions regulatory protocols. Graduate Theses, Dissertations, and
Problem Reports. 5473. Available from https://researchrepository.wvu.edu/etd/5473 (Accessed on Feb­
ruary 2, 2021).
Gallus, J., Kirchner, U., Vogt, R., Börensen, C., Benter, T. 2016. On-road particle number measure­
ments using a portable emission measurement system (PEMS). Atmospheric Environment 124:
37–45, doi: 10.1016/j.atmosenv.2015.11.012.
Giechaskiel, B., Schwelberger, M., Delacroix, C., Marchetti, M., Feijene, M., Prieger, K., Andersson, S.,
Karlssong, H.L. 2018. Experimental assessment of solid particle number Portable Emissions Measure­
ment Systems (PEMS) for heavy-duty vehicles applications. Journal of Aerosol Science 123: 161–170,
doi: 10.1016/j.jaerosci.2018.06.014.
Hallquist, A. M., M. Jersko, H. Fallgren, J. Westerlund and A. Sjodin (2012). “Particle and gaseous
emissions from individual diesel and CNG buses.” Atmos. Chem. Phys. Discuss. 12: 27737–27773.
Hines J [2019]. The role of emissions based maintenance to reduce diesel exhaust emissions, worker
exposure and fuel consumption. Doctor of Philosophy thesis, School of Health and Society, University
of Wollongong. Available from https://ro.uow.edu.au/theses1/854 (Accessed on February 2, 2021).
Kadijk G, Spreen LS, van der Mark PJ [2016]. Investigation into a periodic technical inspection test
method to check for presence and proper functioning of diesel particulate filters in light-duty diesel
vehicles. TNO 2016 R10735v2. Earth, Life & Social Sciences. The Netherlands.
McGinn S, Ellington R, Penney J, Graham C, Huzij B [2010]. Diesel emissions: mechanics’ maintenance
manual. DEEP: Diesel Emissions Evaluation Program.
Spears MW [1997]. An emissions-assisted maintenance procedure for diesel-powered equipment. Minne­
apolis, MN: University of Minnesota, Center for Diesel Research. National Institute for Occupational
Safety and Health (NIOSH). Contract No. USDI/1432 CO36900.
Spears, M. (1997). An Emissions-Assisted Maintenance Procedure (EAMP) for Diesel-Powere Equip­
ment. Minneapolis, MN, University of Minnesota, Center for Diesel Research.
TSI. (2021). “NANOPARTICLE EMISSION TESTER MODEL 3795.” Retrieved 10/ 21/2021,from
https://tsi.com/products/particle-counters-and-detectors/engine-exhaust-particle-counters/nanoparti
cle-emission-tester-model-3795/.
Waytulonis RW [1987]. An overview of the effects of diesel engine maintenance on emissions and
performance. In Proceedings: Bureau of Mines Technology Transfer Seminar, Louisville, KY, April 21,
1987, and Denver, CO, April 23, 1987. Information Circular 9141. Bureau of Mines. United States
Department of the Interior.
Yang, J., Durbin, T.D., Jiang, Y, Tange, T., Karavalakis, G., Cocker III, D.R., Johnson, K.C. 2018.
A comparison of a mini-PEMS and a 1065 compliant PEMS for on-road gaseous and particulate emis­
sions from a light duty diesel truck. Science of the Total Environment 640–641:364–376, doi: 10.1016/
j.scitotenv.2018.04.383.

146
Electric machinery in mine ventilation
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Underground mine ventilation design: Diesel vs. electrical

equipment

C.A. Rawlins
Director - RME Consulting Canada Inc

ABSTRACT: Mine ventilation planning using trackless diesel equipment is well known in
the industry. Battery Electrical Vehicle (BEV’s) type equipment is increasingly replacing diesel
vehicles underground and different ventilation design methodologies need to be evaluated for
safety and health of workers. Planning further evaluate the heat loads of equipment to counter
their negative impact on the ventilation system. It is therefore important to evaluate the over­
all ventilation system when comparing Diesel vs BEV’s thereby accounting for other aspects
such as blasting and re-entry periods for the mine including multi-blast development ends. In
general, is it anticipated that replacing diesel vehicles with BEV’s will imply reduced air quan­
tity in a mine, therefore contributing positively towards overall mine power and an improved
return on an investment. BEV’s and other types such as trolley assisted vehicles all fall within
the heat and airflow reduction application philosophy and the ventilation planning needs to
be addressed to ensure designs don’t under-or-over-estimate their positive contribution in
underground mines as we always need to determine the in-mine heat load for shallow and
deep mines to ascertain compliance with design parameters. This paper elaborates on ventila­
tion design parameters for comparison with diesel vs BEV application.

1 INTRODUCTION

1.1 Diesel fuel and equipment background

Since the start of mining and the implementation of mechanized equipment in the mining
industry, the underground mining environment has seen many different power applications to
assist man in the mining of an orebody and the removal of broken rock from the face area in
either development and/or production areas.
Time consuming manpower was used initially to remove the rock, then actual horsepower
or animal power was applied, and as technology improved in mechanical equipment with the
industrial revolution, diesel equipment was applied in surface vehicles applications and then in
the underground environment. Diesel equipment had to be regulated in underground applica­
tions due to the tailpipe exhaust contaminants such as Sulphur, Carbon Monoxide, diesel par­
ticles or soot (DPM – Diesel Particulate Matter)), etc., and the negative effects it has on
humans.
Diesel fuel (Act, 2003) is “fuel that can evaporate at atmospheric pressure, that boils within
the range of 130°C to 400°C and that is for use in diesel engines or any fuel that is sold or
represented as diesel fuel, as biomass-based diesel fuel or as a blend of biomass-based diesel
fuel and diesel fuel (carburant diesel).”
Diesel fuel in North America is regulated since 1993 to reduce the Sulphur content in fuel
from 5000 ppm. In Canada, since 1 June 2007 to 31 May 2010, Sulphur content shall not
exceed 50 ppm or 500 mg/kg of fuel and there after it shall not exceed 15 mg/kg. This is called
Ultra Low Sulphur Fuel (ULSF).

DOI: 10.1201/9781003429241-16

149
 Besides regulating the diesel fuel content with relation to the Sulphur content and other
contaminants such as DPM, etc., Ontario, Canada, and other countries further regulates the
air supply for diesel vehicles used underground. This is required to ensure the Threshold Limit
Value (TLV) of airborne pollutants are kept at or below their respective limit values. Diesel
engines are tested at CANMET in Canada and MSHA in the USA to determine the air
supply to the vehicle thus countering the tailpipe contaminants to acceptable levels. These
tested air supply values are available at the different internet sites as needed.
Furthermore, diesel engines are further known as Tier 1, 2, 3 up to the current Tier 4 type
engine. Table 1 (Metrovancouver, 2023) shows the diesel engine Tier and year of manufactur­
ing for reference.
• Tier 1-3 Standards. The first federal standards (Tier 1) for new nonroad diesel engines were
adopted in 1994 for engines over 50 hp and phased-in from 1996 to 2000. Tier 1-3 standards
are met through advanced engine design, with no or only limited use of exhaust gas after­
treatment (oxidation catalysts).
• Tier 4 Standards. On May 11, 2004, EPA signed the final rule introducing Tier 4 emission
standards, which were phased-in over the period of 2008-2015. The Tier 4 standards require
that emissions of PM and NOx be further reduced by about 90%. Such emission reductions
can be achieved using control technologies—including advanced exhaust gas aftertreatment.
• Tier 5 Standards. In November 2021, the California Air Resources Board held the first
public workshop development of Tier 5 emission standards that will seek to further reduce
NOx and PM emissions by 50-90%, depending on the engine power category, in the 2028-
2030 timeframe.

Table 1. Diesel engine tier class and time frame.

In review of the current diesel engine pollutant requirements for Tier 4 and later year
designs, the outlet tailpipe pollutants are reduced reasonably well. It is important to note that
in some provinces in Canada and in the USA, legislation is being adopted to apply the actual
tested air quantity value noted in the test results for the different engines and this is related to
the Tier type of the engine manufacturer.
Furthermore, using the legal (Ontario, Canada) value of 0.06 m3/s/kW (100 cfm/bhp) for
diesel engines underground, in conjunction with exhaust emission dilution methods and
advanced engine control technologies, this will maintain the tailpipe pollutants below the
respective TLV’s. This air quantity applied is further used to evaluate the heat load compo­
nent onto the surrounding air from the diesel equipment. It is generally known that a diesel
machine typically supplies between 1.5 and 3 times the rated engine power as a heat compo­
nent depending on the machines age and maintenance applied over time including the engine
technology. This is all related to the efficiency of an engine which is typically around
a maximum of about 40%.
Notwithstanding the fact that diesel engine manufacturers have reduced their engine tail­
pipe pollutants to very low values after tailpipe advanced filtering methods and engine control

150
measures, the industry is moving towards more and more electrical based machinery which
totally avoid inducing pollutants such as carbon monoxide, sulphur content and nitrous
oxides and other harmful contaminants (DPM’s, etc.) into the underground air.
Some research indicated that electrical motors (excluding mechanical resistance items such
as wheel bearings, drive shafts, etc.) operate at about 38°C while diesel motors operate around
98°C (excluding mechanical items and exhaust pipe temperatures reaching 400 - 550°C at
times), thus indicating a motor heat load reduction of about 61% when compared to diesel
engines. Therefore, diesel engines need to be evaluated from two specific aspects namely (a)
a pollutant aspect, and (b) a heat load aspect, whereas the electrical machinery needs to be
evaluated from a heat load perspective only.

2 VENTILATION DESIGN

For actual mine ventilation design from conceptual to implementation phase, there are steps
to consider during each phase of the project fuel and equipment background. When consider­
ing a mine doing ventilation design there are a couple of steps to follow as mentioned above
and depending on the project level of design, i.e., Conceptual, Pre-feasibility, Feasibility, or
finally the Implementation phase, certain aspects need to be investigated in detail (Rawlins,
2004). Aspects such as cooling requirements, heat loads, dust liberation, air quantity estimates
or actual detailed calculations.
In general, a project phase will be as follows.

Figure 1. Planning phases for typical mine ventilation project.

As indicated in Figure 1, the ventilation design starts at a conceptual phase whereby identi­
fying typing parameters such as production and development rock/ore production. This will

151
include the evaluation of the mining method to be applied for the project. The mining method
is important as this can influence the air quantity either more or less as needed. This further
will indicate either trackless or track-bound vehicle operation or both (hybrid type systems).
Using BEV’s, used as track-bound locomotives to move the broken rock from the produc­
tion area of the mine to the rock-passes at the shaft, in mining is not a new concept as it has
been applied to shallow and deep mining operations in mines from the early mining oper­
ations. Figure 2 illustrates a typical battery electrical locomotive in this type of application.

Figure 2. Typical battery electrical locomotive in an underground application.

2.1 Ventilation specifications

As for any mine design and depending on the design stage, some limited specifications are
needed to ensure the design is done according to specific parameters and criteria with assump­
tions included.
Some limited ventilation specifications would be the following as indicated in Table 2.
The key face air quantity determinants for a mechanised mine are:
1. The diesel unit exhaust outlet pipe dilution factor quoted as 0.06 m3/s/kW of diesel rated
power at the point of operation (General dilution value applied).
2. For BEV equipment the allowance is initially 0.5 m/s minimum per m2 of haulage area.
3. For BEV equipment the allowance is further made for wet bulb temperature not more than
27.5°C after the vehicle.
4. The amount of air required to satisfy a mining related re-entry period. In the case of multi
blast conditions the re-entry period is usually 30-minutes and around 2-hours for a mine
wide operation with central blasting operations. Gases will be tested for to ensure compli­
ance with TLV values.

152
Table 2. Limited example of ventilation specifications or design criteria.
Parameters Value Unit

Production rate 100,000 Tpa

Surface intake temperature (Mean summer) 20/30 °C (WB & DB)
Reject temperature 27.5 °C (WB)
Intake air velocity (Shaft; Men & Material transport) <12
Intake air velocity (Dedicated intake shaft) 0 – 12
Intake air velocity (Lateral haulage; men & material) 0.5 – 7
Intake air velocity (Conveyor belt ore transport) 1–3 m/s
Return air velocity (Upcast/downcast shaft/s) 14 – 25
Return air velocity (Return lateral haulage/s) 0 – 15
Air velocity (Fresh air intake with BEV’s) ≥ 0.5
Shaft resistance (K-value; men & material system; Lined) 0.02 – 0.025
Ns2/m4
Shaft resistance (K-value; lined/Raise bored) 0.005 to 0.01
Quartz TLV 0.1 mg/m3
DPM - TC 150 μg/m3
CO (STEL = 100 ppm) 30
NO 25
ppm
NO2 (STEL = 5 ppm) 3
SO2 (STEL = 5 ppm) 2
Heat stress limit (Wet bulb temperature) 27.5 °C
Noise continuous 82 dBA
Diesel/BEV Truck power 100 kW
Diesel/BEV Truck power 100 kW
Diesel equipment air requirement ≥0.06 m3/s/kW
Electrical equipment air requirement ≥0.03 m3/s/kW
Minimum air velocity in diesel LHD drifts/stopes 1.0 m/s
Minimum air velocity in drifts/stopes (No diesel LHD’s) 0.5 m/s
Re-entry periods (Multi-blasting or Secondary blasting) 30 Minutes

5. Sufficient air to dilute flammable gas encountered and to dilute dust that could be created
during normal mining operations, i.e., drilling, tramming, etc. To satisfy cooling of work­
ers, a minimum of 0.5 m/s is advocated, and where tramming is carried out from the face
area, a dust and fume clearing velocity of 1.0 m/s is required.
6. The heat load of the mine with reflection on the diesel machinery, VRT, fans, motors,
blasting, etc.

2.2 Ventilation air quantity determination(s)

The ventilation requirements for a mine in a phased design approach, i.e., conceptual to
implementation, will need to capture information such as:
a) All diesel vehicle air requirement(s), i.e., legal and/or tested dilution values applied, and
heat load based.
b) Depending on the project stage, determine type of diesel machinery with information such
as engine type and model, and engine specifications.
c) Mine blasting requirements such as multi-blast requirements (rapid development), second­
ary blasting in production areas, shaft sinking operations, and other blast requirements for
development ends, etc.
d) Contaminant removal such as for (1) dust, (2) heat, and (3) fumes.
e) Air velocity which can be country specific and regulatory requirements for example,
haulages, stopes, etc. (also refer to Table 2).
f) Minimum air velocity requirement for BEV air calculations as design criteria.

153
 As for any mine design and depending on the design stage, some limited specifications are
needed to ensure the design is done according to specific parameters and criteria with assumptions
included.
This information provides important parameters that influence the air requirements for an
underground mine. The amount of air required at the point of operation in any mine must
ensure that any gas/fumes caused by blasting and/or mechanical equipment is adequately
diluted to legally allowed Occupational Exposure Limits (OEL’s), and that any dust generated
by mining activities is directed away to return airways to ensure that workers are not exposed
to undue occupational risk. In the situation where a mine experience increased levels of heat
ingress, the ambient wet bulb temperature (i.e., WBT ≥ 27.5°C) and air velocities would dictate
the Air-Cooling Power (i.e., ACP- ~170 W/m2) of air (Rawlins, 2004 (3)).

2.2.1 Diesel engine ventilation standards

The generally applied diesel dilution factor and in some countries a legal requirement, is used
to determine the air quantity requirements of an underground diesel engine are 0.06 m3/s/kW
or 100 cfm/Bhp (cubic feet per minute per break horsepower).
When using this exhaust outlet pipe dilution factor at the point of operation the estimated
fresh air quantity requirement for a 100 kW engine is calculated as:
Fresh Air Quantity (Q) = 100 kW x 0.06 m3/s/kW = 6.0 m3/s

2.2.2 Air quantity calculations (Engine tests)

An air quality value or method for calculating (expression developed by French & Mildon,
1978, 1984, & 1990) the quantity required by CANMET for a diesel vehicle is the equation
used for diesel engine certification. This method calculates an Exhaust Quality Index (EQI)
also known as the MAPTEST. The calculations method is:

The denominators indicate the Canadian threshold limit values applied for each contamin­
ant at the time the EQI was developed (reference Canadian Standard - CSA-M424.2 (1990).
Once the EQI is determined the air density (kg/m3) and engine exhaust gas flow rate (kg/s) is
introduced to determine the air quantity certified for that machine usage underground.
The quantity is calculated in the following equation:

and symbols defined as:

Qa: Air quantity required to dilute pollutants (m3/s)
Mde: Mass flow of exhaust air (kg/s)
ρ: Dry air density (kg/m3)
The 100 kW diesel EQI, given its tailpipe measured values, would be:

Therefore, the calculated quantity, using an exhaust gas mass flow rate of 456 kg/hr, is,

say 6.1 m3/s air required for this engine. Another parameter to be calculated is the heat pro­
duced from diesel vehicles.

154
2.2.3 Diesel vehicle heat
An accurate means of calculating trackless diesel equipment heat load (McPherson, 1993) is
from a fuel consumption and fuel property point of view. As an example, the diesel consump­
tion for the LHD is 0.254 L/h per rated power (kW). The diesel usage is 25.4 L/h and the fuel
calorific value is 35400 kJ/L. Therefore,
Heat produced = Diesel usage x Calorific value of the diesel/3600 seconds per hour

The heat produced from diesel machinery comprises of both sensible and latent heat. The
latent heat component could be calculated from the water produced after combustion (taken
at 100% in this example) at a mean of five (5) litre per litre of fuel consumed (water pro­
duced = 5×25.4 = 126.9 L/h). Given that the latent heat of� evaporation of water is about
2450 kJ/kg then the latent heat produced would be 126:9�2450
3600 ¼ 86:4� kW. The sensible com­
ponent is �therefore (249.6
� – 86.4) = 163.2 kW. The trackless equipment rated output power
251:4�3600 �
would be ð0:254�35400 Þ ¼ 100 kW. The heat load factor imposed onto the surroundings would
�
be 249:6
100 ¼ 2:5 (100% utilisation at full power rating). Depending on the variable parameters
used such as the diesel consumption for this specific LHD type, so will the heat load change
(McPherson, 1993). Figure 3 shows the heat load imposed on the surrounding air in relation to
the 100 kW diesel engine operated in conjunction with its full power output requirement at dif­
ferent utilisation factors.

Figure 3. Heat load and heat load factor graph for a 100 kW diesel engine.

Furthermore, when a diesel machine operates underground, the heat load imposed on the
surroundings influences the ambient air temperature. Therefore, notwithstanding the air
quantity determined from a fume dilution point of view, the heat expelled from the machine
could mean that the quantity requirement can change to counter this effect. In mechanized

155
mines, trackless machine heat load could affect the ventilation system requirements and the
air quantity could be required to change.
The wet bulb reject temperature in most mines range between 26°C and 28.5°C. In our
example the reject wet bulb temperature WB is taken to be 27.5°C. It is therefore important to
know what the intake WB just before the LHD should be, in order not to allow the exit tem­
perature (after the LHD) to exceed the design reject temperature of say 27.5°C. A simplistic
psychrometric calculation (Rawlins and Phillips, 2005) is done using the following input
parameters after the LHD.
a) Barometric pressure (P) = 103.6 kPa
b) Exit air temperature (Te) after the LHD: 27.5°C wb
c) Air mass flow (M2) = 20 kg/s (17.3 m3/s)
Calculations indicated that the energy content on the LHD intake side should not exceed
(1670.9 – 249.7) 1421.2 kJ/s. The intake energy content corresponds to a sigma (S2) heat value
of 71.06 kJ/kg. The WB temperature on the intake side of the LHD should therefore not
exceed 24.4°C taken that the surrounding rock and other heat loads are excluded at this point.
A WB above 24.4°C before the LHD (LHD operating at 100% power) would increase the tem­
perature after the LHD to above the design WB temperature of 27.5°C.

2.2.4 Diesel and BEV vehicle comparison calculations

To do a comparison for diesel vs BEV type vehicles, it was thought to do a model simulation
for both situations with similar design input parameters. The only parameter that changed is
making use of diesel and BEV type equipment.
The following are the input parameters for the example calculation.
➢ Diesel & BEV engine motor size is 100 kW.
➢ Fresh air intake temperature is 20/30°C (WB & DB).
➢ Barometric pressure is 103.6 kPa underground (about 100 m from surface).
➢ Haulage size is 5 x 5 m (25 m2).
➢ Air mass flow rate is 20 kg/s (Scenario A).

Table 3. Various scenarios for both BEV and diesel vehicles comparisons for heat load.
Mass Heat
Temperature Quantity Velocity Quantity
flow Load
Heat Load
No. Description WB (C) m3/s m/s kg/s kW Factor m3/s/kW

1 BE Vehicle 21.6 17.0 0.68 20.0 106.1 1.1 0.17

Diesel
23.6 17.3 0.69 20.0 248.3 2.5 0.17
Vehicle
2 BE Vehicle 27.5 3.2 0.13 3.8 105.5 1.1 0.03
Diesel
27.5 7.9 0.31 8.9 249.7 2.5 0.08
Vehicle
3 BE Vehicle 22.2 12.3 0.50 14.5 107.0 1.1 0.12
Diesel
24.9 12.6 0.50 14.5 251.6 2.5 0.13
Vehicle
4 BE Vehicle 24.3 6.0 0.24 6.9 104.1 1.0 0.06
Diesel
29.7 6.0 0.24 6.6 249.2 2.5 0.06
Vehicle

As there is no specific or legal air quantity requirement for BEV operation in mines, the
only parameter to evaluate BEV’s is the imposed heat load on the environment. This is besides
other aspects such as versatility requirements, etc.
The different scenarios are as follows.

156
1. The airflow remained at 20 kg/s.
Comment:
BEV: Wet Bulb Temperature increased from 20°C to 21.6°C.
Diesel: Wet Bulb Temperature increased from 20°C to 23.6°C.
The BEV heat load (HL) was 106.1 kW and the Diesel HL was 248.3 kW.
2. The wet bulb temperature was increased to 27.5°C.
Comment:
Air quantity was reduced, i.e., BEV air quantity is 3.2 m3/s to reach 27.5°C, and Diesel
vehicle air quantity is 7.9 m3/s to reach 27.5°C.
Heat Loads were similar as before.
The haulage air velocity in both the BEV and the diesel vehicle haulage were below the
design air velocity of a minimum of 0.5 m/s, i.e., 0.13 m/s and 0.31 m/s respectively.
3. The haulage air velocity was 0.5 m/s as for the design requirements.
Comment:
Air quantity for both the BEV and the diesel vehicle situations are 12.3 m3/s and 12.6 m3/s
and therefore very similar.
Wet Bulb Temperature increased to 22.2°C and 24.9°C for the BEV and the diesel vehicles
respectively.
The BEV heat load (HL) was 107.0 kW, and the Diesel HL was 251.6 kW.
4. The haulage air quantity was set at 6.0 m3/s to coincide with the design value for diesel
equipment of 0.06 m3/s/kW.
Comment:
Wet Bulb Temperature increased to 24.3°C and 29.7°C for the BEV and the diesel vehicles
respectively. The diesel vehicle wet bulb temperature is above the design value requirement
of 27.5°C and this indicates that the minimum standard set for the diesel vehicle in this
situation is inadequate and the air quantity need to increase to at least 7.9 m3/s to maintain
the wet bulb design value as indicated in scenario No. 2. Furthermore, the minimum air
velocity is not attained with the 7.9 m3/s air quantity and needs further adjustment to at
least 12.6 m3/s to maintain the minimum air velocity in the haulage as a design require­
ment.
The BEV heat load (HL) was 104.1 kW, and the Diesel HL was 249.2 kW.
Summary comment:
The minimum air velocity range of 0.5 m/s would be the more appropriate measure in the
BEV and diesel vehicle cases as this parameter maintains the heat load and therefore the wet
bulb temperature in both the BEV and the diesel vehicle scenarios evaluated.
The re-entry period scenario is also evaluated and given below.

2.2.5 Re-entry period quantity requirement

To determine the quantity required for the re-entry period the development end scenario, the
following parameters were used for calculation:
• Volume: 5 m x 5 m x 113.4 m = 2835 m3
• Re-entry period required: 30 minutes.
• Air changes: 8

The haulage length was calculated to be about 113.4 m length to maintain 30-minutes after the
blast with the minimum air quantity of 12.6 m3/s to further maintain the minimum air velocity in
the haulage. Therefore, the haulage length should not exceed the length of 113.4 m as this is the
maximum length to maintain a 30-minute rapid re-entry period. From the evaluations done it is
evident that the air velocity minimum of 0.5 m/s is the determined factor for this system where the
100 kW BEV or diesel vehicle operates. This minimum air velocity will counter the heat load in

157
the diesel vehicle as well. This further indicates that the heat load factor is about 1.1 for the BEV
and 2.5 for the diesel vehicle. The air ratio evaluation indicated that the BEV will have a value of
about 0.03 m3/s/kW and the diesel vehicle air ratio will need to increase from 0.06 m3/s/kW to
about 0.08 m3/s/kW to maintain the heat load aspect for the system. The scenarios as described
are illustrated in Figure 4.

Figure 4. Exit temperature data related to scenario evaluated.

It is notable that the design value of 0.06 m3/s/kW is not adequate for the diesel vehicle,
however it is in order when the BEV is applied. The design air velocity of 0.5 m/s is the param­
eter to be maintained at this scenario evaluation for the 100 kW vehicle scenario for heat load.

3 CONCLUSIONS AND DISCUSSION

This paper describes ventilation design parameters when diesel vehicles and BEV’s are compared.
When considering this shallow mine scenario described, it was evident that given the param­
eters of the design such as wet bulb air temperature, minimum air velocity in a mine haulage,
etc., the heat load is mostly the important factor to evaluate. When normal project ventilation
design work is done, we determine the total mine air requirement from all diesel vehicles.
Once this value is calculated, then the air distribution is done preliminary and other factors
such as air leakage, auto-compression effects, etc., are made part of the design.
Evaluating a shallow or deep mine system, the heat load will be the most important factor
to be evaluated against the trackless fleet employed, i.e., either diesel or BEV type machines.
When doing this shallow mine comparison for BEV vs diesel vehicles, it was evident that
the general air to engine power ratio of 0.06 m3/s/kW is not adequate as a standalone factor to
evaluate. It is rarely the only factor employed when trackless diesel equipment is evaluated in
a ventilation design.
To maintain the given wet bulb design temperature of 27.5°C with the given 100 kW vehicle
engine, the BEV indicated a value of about 0.032 m3/s/kW and the diesel vehicle indicated

158
around 0.079 m3/s/kW. This indicates a 58.9% air requirement reduction when compared to
the diesel vehicle, although this air quantity determined indicated an air velocity below the
design value of 0.5 m/s. Therefore, the ai velocity design value is the prominent value to apply
in this design. Other evaluations and suggestions proposed by Halim and Kerai (2013) indi­
cated values such as 0.04 m3/s/kW for deep mines and 0.025 – 0.037 m3/s/kW for shallow
mines where this shallow mine evaluated in this paper seems to be in-line with their initial
findings.
Other aspects such as equipment versatility is important in certain circumstances and there­
fore, as with typical coal mines, there is always a diesel scoop running around in a mine to
assist in certain work to be done besides the primary cable electrical equipment applied in pro­
duction sections.
The heat load of a mine is important to be evaluated when the vehicles are included. This
parameter will indicate what the appropriate air quantity is to be applied to the project. The
thoughts of implementing BEV’s into a project and reducing the amount of air substantially is
not true and could lead to inadequate air supply from a heat load perspective, also from
a dust dilution perspective when the minimum air velocity range is applied. A mines overall
re-entry period after a main blast (including rapid haulage development as needed) is further
aspects that need to be part of the overall mine design to ensure the re-entry period in
adequate as per mine design.

REFERENCES

Act, C. E. P. (2003) Sulphur in Diesel Fuel Regulations. SOR/2002-254.

Halim, A. & Kerai, M. (2013) Ventilation Requirement for ‘Electric’Underground Hard Rock Mines:
A Conceptual Study. Australian Mine Ventilation Conference 2013: 01/ 07/2013-03/07/2013. The Aus­
tralian Institute of Mining and Metallurgy.
Mcpherson, M. J. (1993) Subsurface ventilation systems. Subsurface ventilation and environmental
engineering.
Metrovancouver (2023) Non-Road Diesel Engine Regulatory Program. Environmental Regulation &
Enforcement Services
Rawlins, C. (2004) Ventilation Design Strategy For Trackless Deep Underground Mines. MDEC,
Toronto, Canada.
Rawlins, C. & Phillips, H. (2005) Underground mine ventilation planning and design with regards to
heat load and cooling mechanisms. Proceedings of 8th International Mine Ventilation Congress. ADS
Gillies.

159
Mine cooling and refrigeration
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Selection, design challenges and construction of Vale’s Coleman

mine 10 MW surface refrigeration plant

S.G. Hardcastle & J.K. Shaw

BBE Consulting, Sudbury, Ontario, USA

C. Allen
Vale, Copper Cliff, Ontario, USA

ABSTRACT: Vale’s Coleman mine has extracted minerals from 200-1,700 m depth for over 50
years and is currently expanding 1,850 m. At depth, autocompression effects superimposed on
summer climatic conditions, generate increasingly adverse thermal working conditions with tem­
peratures exceeding 40°C DB and 28°C WB. These require heat exposure management through
work:recovery regimens and stop work conditions. Through mine planning, Vale recognized they
would need cooling, and in 2016 engaged BBE to investigate novel cooling and traditional tech­
nologies, assessing both underground and surface options to provide focused to the 170 Orebody.
It was identified that although the cooling aspect was feasible, providing the required electrical
power, clean water supply, and locating the heat rejection without impacting mine activities
would be problematic. In 2019, Vale again engaged BBE to design and supply a surface cooling
plant, this included investigating alternative mechanical options, opportunistic coolth storage solu­
tions, and the final provision of an operating turn-key system for the summer of 2020. Consider­
ing the 18-month timeline, the capital costs, unproven technologies, and production loss risks,
most of the options reviewed were discounted and Vale opted for a conventional vapor compres­
sion mechanical system. This paper discusses the challenges of surface and underground cooling,
novel options, the installation, and operational challenges of a surface air cooling plant in
a northern climate.

1 INTRODUCTION – COLEMAN MINE

Vale’s Coleman mine located 50 km NW of Sudbury, Ontario, Canada, has a humid continental
climate with warm/hot humid summers and cold winters. Ambient temperatures can range sea­
sonally from -32°C through to +32°C, and temperatures in excess of 25°C can be expected from
May through September. With mines reaching greater depths, such summer conditions will
required heat stress management protocols (from work:recovery regimes to stop work orders) to
be progressively introduced due to autocompression effects causing the temperatures to increase.
Coleman mine currently produces to depths of 1,700 m with development to 1,850 m, and due to
autocompression thermal working conditions could exceed 40°C DB and 32.0°C WB; these tem­
peratures are Vale’s respective stop work limits. Heat stress protocols are required to ensure the
health and safety of the work force but reduce a mine’s productivity and thereby its profitability.
Through its mine planning, Vale recognized they would increasingly need mine cooling, and
from 2016 onward have been exploring options prior to the final selection of a surface plant in
2019, and requiring it being operational for the summer of 2020.
As will be described in the following sections, the need for cooling gradually progressed
from providing focused localized cooling, through to recognizing the need for a surface plant.
At both stages, various options and alternative technologies were evaluated prior to deciding
upon a conventional mechanical vapor compression refrigeration system.

DOI: 10.1201/9781003429241-17

163
2 COOLING THE 170 OREBODY (OB)

2.1 Underground cooling

Vale’s first request was to perform a feasibility study for mechanical refrigeration spot/zone
cooling of the 170 OB, where mining occurs between 1,500m (5160L) and 1,850 m (6050L),
and to which 212 m3/s (450 kcfm) of air is supplied via an underground booster fan station on
the 5160L. This volumetric flow represents 37% of the mine’s total air supply, and cooling
could be applied from April through to October.
Localized cooling would typically be a mine’s first choice for cooling due to its greater pos­
itional efficiency, a smaller volume needs to be cooled, hence a lower cooling duty, and neces­
sitating less mine service requirements with respect to power, water supply and disposal.
However, underground construction comes at a premium cost relative to surface installations.
Mine service upgrades, available space, logistics, vehicle and personnel travel routes, and heat
rejection also need to be considered.
Within this area four specific locations were identified as potential areas for refrigeration
plant, namely the 3770L fresh air distribution area, 4215L, and the 5160L each handling the
full 212 m3/s air volume, and a more specific 5700L Fresh Air transfer drift location delivering
71 m3/s exclusively to the 170 Lower OB. The return air route from these locations was the
primary Kiruna truck haulage ramp. The respective cooling needs for these locations to main­
tain air temperatures below 40°C DB and 27.5°C WB were identified as being 5.6-6.0 MWR
for the full flow and 2.0 MWR for the smaller flow.

Figure 1. Slipstream fan and cooling coil schematic.

Figure 2. Underground bulk air conditioner spray chamber, laronde mine (courtesy of Agnico Eagle).

Two mechanical cooling options were considered, single and two stage direct contact cold
water open spray chambers, Bulk Air Coolers (BACs), and slipstream cooling coils circulating
cold water. Regardless of choice, they would be serviced from a 6 or 2 MWR refrigeration

164
plant room, and heat rejection would be through respectively a 8 or 3 MW direct contact
open hot water Condenser Spray Chamber (CSC). A slipstream coil solution with air also
being allowed to bypass the unit was chosen over a bulkheaded arrangement due to not want­
ing to restrict the airflow when cooling was not required. Only a CSC option for heat rejection
was considered at this stage as the recipient would be air returning from mine workings con­
taining dust, particulate, blast fumes etc. Figure 1 and Figure 2 respectively show a BAC that
was installed underground at Agnico Eagle’s Laronde mine, and a schematic drawing of
a coiling coil fan unit.
Some of the specific considerations to each cooling option and the heat rejection are:
• BAC have a larger footprint and cross-sectional area (a function of required air velocity),
need near level ground for a water dam, but has no specific air quality limitations.
• The cooling coils, have a small footprint, can accommodate a grade change but need a fan
and clean air.
• A horizontal CSC has similar requirements to the BAC, a vertical CSC requires an eleva­
tion change and additional pumping power.
The general dimensions of each element and the required services for the various options to
supply 6.0 MWR and 2.0 MWR are given in Table 1; where possible the mine’s standard road­
way dimensions were maintained with a view to repurpose existing openings.

Table 1. Conceptual Design details of 6.0 and 2.0 MWR cooling arrange­
ments for an underground installation.

Although all concept options were to a certain degree feasible at the respective locations,
each had its limitations. These ranged from minor groundwork, essentially repurposing exist­
ing spaces, through to major construction requiring new locations to accommodate the BAC,
Plant Room and CSC. The amount of construction could be limited by using cooling coils
over an open BAC spray chamber. However, the common issue to all options was the lack of
a suitable airstream, at a sufficient volume for heat rejection. The CSC process to reject heat
raises both the temperature and relative humidity of the host airstream, typical practice would
be to reject this air direct to a non-travelled return air route. For the 170OB, the primary
return air route was the main haulage ramp, and resultant conditions could be expected to
reach the order of 33°C DB with very high humidity nearing saturated conditions well in
excess of 27.5°C WB. These conditions would be prohibitive, even upon considering the
trucking is performed in AC operator cabs, the condensation on the ramp as the air ascends
would affect vehicle traction.
Considering this, alternate heat rejection options were explored. At some locations, the air
could be routed through a repurposed manway, but ultimately that air returned to the main
haulage ramp. In terms of air-based solutions, the most promising option was to use a drift
connecting through to an area of old mine workings which was already providing an inde­
pendent exhaust to surface. This air route was a small dimension 900 m long section of track-
drift, nominally at 3,500 ft, relatively close in elevation to the 3770L, however a pipe-raise to
accommodate two 355 mm NB pipes would be required to the higher elevation, and a fan

165
would be needed to induce the higher reject airflow requirement. The use of this airway as
a heat sink was attractive as it did not compromise other activities within the mine. However,
its use similarly came with consequences, it was anticipated that the hot humid air would cool
as it travelled through the track-drift and condensation would occur producing an estimated
2.5 l/s of water needing to be managed by the mine’s dewatering system. This would require
upgrading the local dewatering pump capacity at the far end of the drift, which would necessi­
tate the drift to be rehabilitated for its entire length at considerable cost.
At this point, other cooling options, the use of ice slurry delivered from surface, and heat
rejection, the use of mine water, were assessed. The ice option was discounted as a 25 kg/s
supply would add 25 l/s to the mine’s limited dewatering capacity. The use of mine water as
a heat sink was also considered impractical as it would require 150 l/s, and such volumes were
not available. Of all the underground options considered, the 6 MWR cooling of air using slip­
stream coils located on the 3770L with heat rejection to the adjacent track drift was the pre­
ferred solution. However, it would also require a major electrical supply upgrade within the
mine and on surface to accommodate peak demands of 2.5 MWE.

2.2 Surface/underground cooling systems

Due to the accumulating logistical limitations of a pure underground solution, various surface
options to cool only the air required for the 170OB and the mine’s total intake air were
explored. These included using a dedicated chilled air path from surface, a pre-existing duct
within the #1 Fresh Air Raise, and the option to circulate chilled water to and from under­
ground cooling coils from a plant on surface. This latter option would use a U-tube type of
system which has been employed in South African platinum mines and for the development of
Rio Tinto’s Resolution project to depths of 1,400 m with plans to extend it to 2,000 m.
The advantages of a surface plant over an underground install include:

• Ease of construction • Lower cost construction

• Less sophisticated plant can be used on surface • Higher efficiency - improved Coefficient of
Performance (CoP),
• Less shaft time to transport material underground • Limitless ambient air available for heat rejection
• Opportunity to prefabricate off-site • Opportunity to use Ammonia refrigerant.

The FAR contained a 1.8 m diameter duct that was previously used as an exhaust. At
a velocity of 35 m/s the duct could supply 93 m3/s, this was only deemed sufficient to delivery
in the order of 2.0-3.0 MWR to the upper 3770L location. This option was quickly eliminated
based upon its limited cooling capacity, and because the current integrity of the duct, and its
long-term viability could not be assured.
The U-tube option would require both high pressure piping and coiling heat exchangers, nom­
inally two 400 mm diameter 120 Schedule piping capable to withstand 2,500-4,000 psi. These
pipes would need to be installed within the existing #1 FAR or routed through an independent
pipe raise. The FAR concept was deemed impractical considering its was a decommissioned
stripped shaft and extensive work and mine downtime would be required for pipe installation.
The cost for directionally drilling a dedicated pipe raise was also considered prohibitive.
The remaining option was to cool the mines total 566 m3/s intake airflow with the first
order cost estimate being 2x that of selectively cooling only 212 m3/s. In addition, it would
have a larger power demand, greater than 3 MWE, and the mine already had power supply
limitations.

3 SURFACE COOLING OF COLEMAN MINE – ALTERNATE TECHNOLOGIES

In late 2018, after considering the overall logistics of a 170OB solution and the loss potential
for productive time, Vale issued a request for the design and turn-key installation of a surface

166
cooling plant to deliver nominally 10.5 MWR. The specification was to be able to cool 566 m3/
s of air to a maximum temperature of 21°C DB/ 15.4 °C WB at the collar of the fresh air raise
downstream of the fans; plus, the requirement for the system to be operational for the
Summer of 2020. As part of the contract award to BBE, Vale required a stage-gate analysis
evaluating the potential to use alternate technologies from the typical mechanical vapor com­
pression solution to leverage innovation funding for environmental initiatives. Within this
analysis specific attention was to be given to low water consumption and low energy use solu­
tions respecting potential limitations at the mine. At the outset of this work, the cooling strat­
egies to be reviewed are detailed in Figure 3.

Figure 3. Alternative cooling strategies considered in initial stage gate analysis.

3.1 Alternate, green, and opportunistic cooling options

Most of the options available to deep mines in Ontario and Quebec, with northern continental
climates, have been reviewed in respect to cost (Millar, 2014). The outcome of that analysis is
given in Table 2 where the methods are ranked by ascending cost $(CAD) per MWhr.

Table 2. Deep mine cooling options ranked by unit cost.

Rank Technology Unit Cost Rank Technology Unit Cost

1 Opportunistic Ice Stope 5 8 Double Effect Absorption 55

2 Opportunistic NHEA 6 9 Ejector 94
3 Very Opportunistic HAC 15 10 Opportunistic HAC 101
4 Lake Water Cooling 19 11 HAC 115
5 Ice Stope 27 12 Single Effect Absorption 129
6 Mechanical Chiller 37 13 Liquid Air 173
7 Ice Slurry 42 14 Engineered NHEA 481

Within the above ranking the wording “opportunistic” assumes the infrastructure required
for this technology is pre-existing. For example, Glencore’s Kidd Mine, Timmins, Ontario
(Counter, 2014), has near surface stopes within which ice forms during Winter, and in
Summer, it is then used to provide the equivalent of 8.5 MWR of cooling. Vale’s Stobie Mine
also has an ice-stope that was designed to offset winter heating needs (Stachulak, 1991); had it
been designed for the reverse process; it could provide in the order of 6.0 MWR of natural
cooling (Millar, 2016). The term NHEA refers to a seasonal Natural Heat Exchanger Area,
such as that employed at Vale’s Creighton mine, here air enters the mine through an extensive

167
broken rock mass in its former caved zone; it provides heating during the winter and cooling
during the summer (Stachulak, 1991). Consequently, Creighton mine does not have heating
costs and cooling is estimated to be equivalent to a 17.5 MWR plant (Millar, 2016).
A general description of each technology assessed follows:
1. Ice stopes are underground or surface-built modular thermal storage units that allow the pass-
through flow of cold winter air which contacts water directly, giving up energy to the water as
it freezes to form ice, in-turn heating the air to some temperature approaching the water being
supplied for freezing. In summer the process is reversed with warm air passing over the ice.
2. NHEAs are typically composed of previous excavations or pits. The rock surface and
broken rock mass maximize air-to-rock surface area and provides a degree of thermal iner­
tia for the air being drawn through the excavation. Seasonally they provide heating and
cooling.
3. HAC refers to a hydraulic air compressor that function on the principle of air entrainment
as turbulent waterflows descend some vertical distance. The entrained air has the added
benefit of direct contact with the water as it is compressed. This compressed air can be cap­
tured, transported, and decompressed to provide dry, sensible cooling.
4. Lake cooling takes advantage of large thermal masses and deep lakes to provide a source
of ~5-7°C water throughout the year. This can be of sufficient thermal quality to use dir­
ectly in cooling applications, reducing interface inefficiencies, and reducing capital costs.
5. Mechanical chillers are a commercial vapor compression based system.
6. Ice slurries are produced with vapor compression technology with the cooling fluid frozen into
a slurry. The benefit here is that ice shipped underground doesn’t suffer autocompression.
7. Absorption chillers work on refrigeration cycles like mechanical vapor-compression except
they use a heat input directly instead of a compressor. Double effect absorption uses steam,
whereas single effect absorption uses hot water.
8. Ejectors convert static pressure into velocity pressure. A sudden decrease in pressure evap­
orates a liquid-phase refrigerant (water) in a flash chamber, taking process heat with it,
and allowing it to be re-condensed elsewhere and cooled in contact with some heat rejec­
tion infrastructure.
9. Liquid air can be used to provide point-source, dry, sensible cooling provided the infrastruc­
ture exists to ship it. Air is liquefied in the same cycle as traditional vapor compression.
During the first stage-gate analysis ice stopes, an NHEA, a HAC, ice slurries, ejectors and
liquid air were eliminated due to either the lack of the “opportunistic” infrastructure, or the
inability to construct the delivery systems to depth.
Other options identified in Figure 3 were eliminated as they did not fit Vale’s requirement
for them to be a low energy, low water consumer, as per Table 3.
Considering the constraints discussed above and as listed specific to a technology, BBE and
Vale short-listed the analysis to the following alternative cooling technologies:
• Combined Heat and Power (CHP) Natural Gas Generator & Absorption Chiller
• Natural-Gas-Fired Absorption Chillers
• Glycol Thermal Storage & Distribution System
• Stratified Thermal Energy Storage
These were subsequently analyzed using a baseline requirement of 2 MWR. A natural gas
supply was available as it was used for winter heating.

3.2 Baseline vapor compression refrigeration system

The baseline conventional mine air surface refrigeration system consists of Bulk Air Cooler
(BAC), a vapor compression refrigeration plant and a Condenser Cooling Tower (CCT). The
systems are known for their reliability and being able to operate year-round in hot climates.
For Coleman, vapor compression systems also have the benefit of being able to perform with
a coefficient of performance (COP) of refrigeration that is relatively stable across the part-
load operational requirements throughout the cooling season.

168
Table 3. Other alternate cooling technologies eliminated.
Energy Analysis

Lake Water Lake water was removed due to the lack of a suitable deep lake close to the intake shaft.
Environmental permitting and related schedule implications were assumed to prohibitively
complicate the use of lake water in spray chambers or by heat pumps. A deep lake based
cooler sized to provide 2 MWR cooling at required design conditions would need a flow of
60 kg/s or more, with an estimated seasonal volume of 24,000 m3. The lake would need to
be significantly larger.
Process Process water flows of sufficient volume are not available near the intake fans.
Water
Solar Solar-thermal systems by their nature only function when sunlight is available. This tran­
Thermal sient source of input energy could be used as a heat source for an absorption chiller. Local
conditions are assumed unfavorable for solar thermal systems; solar radiation data was
not verified. A solar collector with an area of approximately 70 m x 70 m would be
required to serve a 2 MWR absorption chiller. This and the absorption infrastructure’s
large capital cost and relatively high maintenance requirements were barriers to adoption.
Exhaust Air Exhaust air energy recovery cannot be applied feasibly to this project as the exhaust shaft
is not situated in close enough proximity to consider a glycol energy transfer system or
a heat pipe.
Ground A ground thermal energy source coupled with a heat pump could be used to provide heat­
Source ing in winter and cooling in summer. However, in the absence of a natural geothermal
heat source, conventional ground source – heat pump systems are not suited to high
duties. A nominal 48 km of buried piping would be required to serve a 2 MWR heat
pump.

3.3 CHP natural gas generator & absorption chiller

As per its title, this option is a combined system to generate both power, electricity, and heat
as a by-product that can then be used by an absorption chiller. A natural-gas fired 4.75 MWE
electrical generator will satisfy a 10.3 MWR plant’s electrical demand and can be fitted with
an absorption cooler which derives its prime power from the 46% efficient natural gas gener­
ator providing a maximum cooling of 2 MWR. The challenge in this case is that to produce
this amount of cooling the generator needs to be under relatively high load, of which the
refrigeration plant itself can provide half. From a refrigeration point of view this indicates in
general terms why CHP systems are primarily power systems that have refrigeration as
a tangential benefit. These are feasible only in applications where there is justification for the
comparatively higher capital costs and operating expenses compared to on-grid power plants.
Because the natural-gas-powered electrical generator and absorption refrigeration machine
combination is only feasible if the gas-powered electrical generator demand was pre-existing,
this technology was not recommended for Coleman.

3.4 Natural gas fired absorption chillers

Absorption chillers alone were considered taking the approach of the base case vapor-
compression plant and upgrading the condenser infrastructure. The fundamentals of the
absorption refrigeration cycle coefficients of performance result in a significantly larger required
condenser infrastructure, roughly 2x the original planned cooling tower size increasing the cap­
ital cost. A lower COP also results in significantly higher operational expenditure compared to
the conventional electrical grid powered plant. Consequently, any absorption refrigeration
system is only a more environmentally acceptable solution if the heat source is both existing and
free. If combustion of natural gas is used to provide the necessary heat, an absorption refriger­
ation system has a higher carbon footprint and operating costs compared to a conventional
refrigeration system. For this reason, it was not given any further consideration.

169
3.5 Seasonal thermal (coolth) storage systems
Two seasonal thermal storage systems were reviewed for Coleman. The first relied on a phase
change from water to ice to provide heating during winter and cooling during summer driven by
a glycol circuit. In winter, glycol is cooled, and the air heated via an air heating radiator. The cold
glycol is pumped in closed circuit to a bladder located in water reservoirs. Through heat transfer,
the water freezes to form the cold storage. The now warmed glycol returns to the heat radiator to
complete the circuit. In summer, cold glycol is pumped from the thermal storage to a water heat
exchanger with the chilled water then available to provide the mine cooling. This process is similar
to district heating and cooling systems. For its intended duty 2 MWR and duration, it was esti­
mated a 0.55M gallon reservoir in excess of 10 m depth by 16 m diameter would be required.
The second approach uses water and an insulated ice thermal storage reservoir. In winter
the reservoir is filled with ice which could be produced by various methods. In summer the
melt water is circulated to cool air. Initial calculation of the sizing storage system shows the
reservoir to be in the order of 10 m depth by 45 m diameter.
Comparing these two systems the former would require a significant volume of glycol at an
outlay exceeding $0.5M, while the latter had a far more significant footprint.
A third novel system including water atomizing sprays (ultrasonic eductors) and MURL
foil dehumidification blades was also briefly reviewed but in the absence of detailed informa­
tion from the proponent the system could not be technically assessed.

3.6 Stratified thermal storage systems

Another option considered was the use of a stratified thermal storage system where during periods
of low ambient temperature and/or low cooling demand, the system stores surplus cold water. It is
gently added to the bottom of a deep cylindrical tank and warm water is removed from the top of
the tank. Likewise, during periods of high ambient temperature and/or high cooling demand, add­
itional cold water is removed from the bottom of the cylindrical tank and warm water is gently
added to the top of the tank. Cold water is denser than warm water. Provided the water is added
to the tank gently and uniformly, a thermocline develops, and this will prevent the cold and warm
water from mixing. To maintain the thermocline, the tank must be fully charged or discharged
once a day and preferably both. Although, the cost of running this system will be approximately
the same amount or slightly more than the base case design, it has been included here as it pre­
sents a relatively energy-neutral approach to peak-shift the electrical demand.

3.7 Cost comparison

Table 4 shows the cost comparison results baselined to providing 2 MWR for 506 hr of oper­
ation. Based upon this analysis, BBE advised the most promising alternative energy sources to
be considered for Coleman to be the combination of widely studied seasonal ice thermal stor­
age systems with established thermal peak shaving technologies such as the proposed stratified
thermal energy storage systems. However, the practical combination of these technologies
would need further detailed costing for a full-sized plant and added consideration of the rela­
tively large amounts of water and ice required for their operation. Another alternative would
be to combine these technologies into a design where the conventional vapor compression
system was only responsible for a part-duty of the overall demand i.e. 7 MWR with the
remainder peak demand supplied by alternate sources. This study was completed Q3, 2019.

Table 4. Costing comparison of short-listed alternate technologies versus vapor compression.

2 MWR Source Cooling Technology CAPEX ($/2 MWR) OPEX ($/506 hrs)

Vapour-Compression Plant $3.0 M $19,000

3 MWE Gas Generator c/w 2 MWR Absorber $5.0 M $107,000
Absorption Chillers Only $3.3M $240,000
Coolth Storage Reservoirs $10M (est.) N/A

170
4 SURFACE COOLING OF COLEMAN MINE – SITE LIMITATIONS

The planned location for the cooling plant placed constraints on the final cooling plant selec­
tion. The available space in front of the pre-existing fan/heater system was 29 m long by
20 m wide, see Figure 4, and had a pre-existing headframe infrastructure slab. The area was
also bounded by rock walls on the North and East sides, the mine air heater to the West and
an access road to the South. The water supply was constrained, being a shared resource from
a common pipeline also serving an adjacent mine property, and electrical supply was limited
with the mine being at the end of a transmission line. An additional constraint put on the
installation required it to not impact the existing intake fans which were already operating
near their maximum flow/pressure duty. After reviewing the costs, viability, logistics involved
in the various options and the 9-month remaining aggressive timeline for finalizing the design
and completion, Vale opted the selection of a full 10.3 MWR vapor compression system. The
prime driver was to guarantee the required amount of cooling be available throughout the
summer of 2020 to not impact production.

Figure 4. Location constraints for the cooling plant.

Figure 5. Conceptual and final design of BACs in relation to the mine heaters.

Table 5. Final design parameters for the Coleman surface cooling plant.
Design Parameter Value

Technology Selected: Vapour-Compression

Ambient air condition 19°Cwb/26°Cdb, 97.2 kPa
Bulk air cooler air flow 500 m³/s, 564 kg/s
Maximum total air flow down FAR 665 kg/s
Mixed air temperature down FAR 15.4°Cwb/18.2°Cdb
Refrigeration machine duty 10.3 MWR
Electrical Source Natural gas generators

171
4.1 Final design criteria
Figure 5 shows the initial conceptual and final arrangement of the BACs set back from the front
of the pre-existing heater/fan installation to minimize interference, and Table 5 provides the
design values. The design basis was to condition 88% of the required mine flow with the make-up
bypassing from the sides and above. Due to concerns of cooling loss under certain wind condi­
tions, side walls with an access door have subsequently been added and the top remains open.
The placement of the BACs in relation to the fan-heater house and wind effects were stud­
ied through CFD analyses, (Shaw and Falk, 2023). The mine opted to use rented natural gas
generators to meet the electrical needs of the cooling plant until power can be upgraded. To
address possible short-term water shortages, or interruptions, the design also included a water
storage capacity capable of providing a 3.5-hr operational buffer.

4.2 Selected equipment

The primary equipment components selected for the plant were as follows:
• Refrigeration Plant – 3 Carrier XR chillers, each providing ~3.5 MWR duty each
• BACs – 8 Baltimore Aircoil 1500E series custom horizontal flow cooling towers.
• CCTs - 3 SPX Marley NC84 series cooling towers, 1 tower dedicated to each machine.
• BAC Pumps –2 Flowserve 41K series pumps, 1 duty, 1 standby, providing 360 l/s flow.
• CSC Pumps – 2 Flowserve 41K series pumps, 1 duty, 1 standby, providing 320 l/s flow.
• Water treatment - Suez water treatment technologies providing biocide, anti-scalant, and
pH control on the condenser and evaporator water circuits.
• Power supply - 4 x 1 MWE Aggreko natural gas generators, 3 duty, 1 standby.

4.3 Construction timeline

The plant was constructed in 13 months from Vale handing the site over to BBE in Septem­
ber 2019 to the full plant commissioning in October 2020, Figure 6. The majority of the con­
struction work coincided with the start of COVID-19 and associated precautions thereafter.
This had major implications on the planned schedule limiting concurrent activities, impacting
the available trade workforce, and prolonging the delivery of major capital items plus the full
spectrum of materials required. To help offset some of these compounding delays, it was
necessary to accelerate the work using 2 shifts and weekends.

Figure 6. Final system installed at Coleman, from right to left: 8 BACs (4x2 arrangement), cooling
plant building containing 3 x 3.5 MWR chillers, and 3 CCTs, 1 per chiller machine.

This effort permitted the plant to be 50% operational for the summer of 2020; the final com­
pletion of the system was delayed from that point on due to having to work around a part
commissioned plant. The full capacity of the plant was available October 2020. Post commis­
sioning analyses have shown a potential loss of cooling due to certain wind conditions and

172
airflow characteristics. As shown by (Shaw and Falk, 2023) these have subsequently been
addressed with side walls being installed in 2022.

4.4 Northern climate challenges of surface cooling plants

As noted earlier, there are many advantages of installing mechanical refrigeration systems on
surface including ease of build, maintenance, available power, cost, and heat rejection. How­
ever, northern climates have unique challenges such as:
• 12 to 16 weeks of day and night-time temperatures suitable for optimal operation.
• Cooling needs for deep mines may begin 4 to 6 weeks before a cooling plant would nor­
mally be expected to operate optimally.
• Cooling demands may last 4-6 weeks beyond the typical timing of the plant being shut
down for the season.
• These shoulder periods result in BACs being both water generators and consumers through
diurnal temperature cycles.
• It is often the dry bulb temperatures that exceeds allowable work limits.
• With the introduction of electric mobile equipment, the problem of rising dry bulb temper­
atures will become more prevalent.
• It is necessary to have both heating and cooling systems at the intake of a fresh air raise.
• At the cooling system design phase, the mine environment parameters alongside the equip­
ment type and activity that will occur generating the heat load need to be identified.

5 CONCLUSIONS

Vale and BBE went to considerable lengths to identify a workable cooling solution for the
Coleman Mine. The major hurdle to using an underground solution related to the lack of heat
rejection capacity, and possible solutions each had their drawbacks. Delivering chilled water
or ice to the required cooling horizon were also found to be impractical. The final option was
to cool the mine’s full intake flow at surface. With that choice, alternative cooling methods
and concepts were evaluated to leverage environmental initiative innovation funds. However,
none could provide the required guaranteed coolth, and capable of being constructed and
operational within the specified time frame. This time limitation was driven by the need to
avoid reduced work schedules or stop work conditions that would impact production in the
coming and subsequent summer.
Once the approval was given for a 10.3 MWR plant, BBE were responsible for its design,
construction, and commissioning within a 13-month period. Despite the numerous delays and
the major impact of COVID-19, the plant was operational at 50% capacity for the summer of
2020 and full commissioning completed October 2020. Further optimization work and con­
struction started in the fall of 2022 to be ready for the 2023 cooling season.

REFERENCES

Counter, D.B., 2014. Kidd Mine – dealing with issues of deep and high stress mining – past, present, and
future, Proc. Seventh International Conference on Deep and High Stress Mining, Deep Mining 2014,
Australian Centre for Geomechanics.
Millar, D.L. 2014. A review of options for cooling ultra-deep mines in Northern Ontario and Quebec.
CIM MEMO Conference 2014.
Shaw, J. Falk, L. 2023. Minimizing the Aerodynamic Impact of a New Cooling Plant Installed Upstream
of an Existing Surface Fan and Heater Arrangement Through CFD Analyses. Proc. 19th North Ameri­
can Mine Ventilation Symposium.
Stachulak, J.S., 1991. Ventilation and unique air conditioning at Inco Limited, CIM Bulletin, Vol. 84, No. 950.
Trapani, K., Romero, A. and Millar, D., 2016. Deep mine cooling, a case for Northern Ontario: Part II.
International Journal of Mining Science and Technology, 26 (2016) 1033–1042.

173
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Mine refrigeration using geothermal energy – is this a viable

decarbonization strategy?

D.W. Durieux
Hatch, Ontario, Canada

ABSTRACT: With the ever-increasing drive towards decarbonization, the case for mine
refrigeration using geothermal means is gaining traction. Potential mines must however be
located in both favorable geo-political jurisdictions and areas with adequate geological condi­
tions. One such area is Sudbury, Ontario, Canada where a conceptual evaluation found that
geothermal additions system can be cheaper and reduce CO2 emissions when compared to
a conventional mine refrigeration system.

1 INTRODUCTION

The largest mining companies including BHP, Rio Tinto, Glencore, and Vale have all targeted
“net zero” operations by 2050. In addition, they have collectively set medium-term carbon
reduction goals of 30% by 2030 (BHP), 50% by 2030 (Rio Tinto), 40% by 2035 (Glencore),
and 33% by 2030 (Vale) respectively. Therefore, given the relatively high energy consumption
of conventional mine refrigeration systems, and the resulting CO2 emissions associated with
power supply, the focus on mine refrigeration system efficiency is at an all-time high. The less
energy the mine refrigeration system requires to produce a given cooling duty, the lower the
associated carbon footprint.
The design process used to cool mines is well understood and has been successfully imple­
mented for many decades. These cooling systems, and the associated equipment employed,
have typically been designed and fabricated for the harsh mining environment with a major
focus on robustness, often at the cost of refrigeration efficiency, or more accurately, coeffi­
cient of performance (COP). System efficiency, of course, has featured in design decisions for
mine refrigeration systems, but the additional CAPEX and complexity required for very high
system efficiency has historically not been justifiable. That is until now, where the addition of
a CO2 emissions criterion into upfront option evaluation strongly favors highly efficient sys­
tems. This paper attempts to compare a base case conventional mine refrigeration system,
located in the Greater Sudbury Area in Northern Ontario, Canada to several geothermal
modifications made to this base case system. This comparison can be seen as an option evalu­
ation exercise similar to what is done during conceptual design.

2 ELECTRICITY EMMISSIONS

Greenhouse gases (GHG) emitted as a result of electricity generation and consumption largely
depends on the source of energy used. Common energy sources include coal, natural gas,
nuclear, hydro, and renewables such as wind and solar.
In Canada, the consumption intensity is the metric used to quantify the GHG emissions
from electricity consumption and Environment and Climate Change Canada recommended
a consumption intensity of 28 g CO2 eq./kWh for Ontario in 2020.

DOI: 10.1201/9781003429241-18

174
3 GROUND SOURCE HEAT PUMPS

Kavanaugh & Rafferty (2014) describes the term ground-source heat pumps (GSHP) as an
all-inclusive term for a variety of systems that use ground, groundwater, or surface water as
a heat source and/or heat sink. This includes various types and sub-types which is summarized
in Table 1. The choice of GSHP type will be both system- and site-dependent and will likely
require trade-off studies to select the optimal geothermal system or to reject geothermal
altogether. This paper, however, will focus on the vertical GCHP system (item 1.1 in Table 1)
as it is the most common and most applicable to a wide geography.

Table 1. GSHP types.

No. Types Description Sub-Type

1 Ground- Closed-looped 1.1 Vertical GCHP using HDPE tubes in a vertical borehole.
coupled piping buried in 1.2 Horizontal GCHP including HDPE single pipe, multiple pipes,
heat pump the ground and coiled pipes.
(GCHP)
2 Ground- Opened-loop 2.1 Plate heat exchanger between heat pump and ground water loop.
water heat piping systems 2.2 Groundwater flowing through heat pump heat exchanger.
pump with water wells
(GWHP)
3 Surface- Systems con­ 3.1 Closed-loop similar to GCHP.
water heat nected to lakes, 3.2 Open-loop similar to GWHP.
pumps streams, or
(SWHP) other reservoirs.

The tried-and-tested standard configurations for vertical GCHP, according to Kavanaugh

& Rafferty (2014) are as follows, see Figure 1:
• High-density polyethylene (HDPE) tubes are placed in vertical boreholes.
• Tubes are thermally fused at the bottom of the bore to a closed return U-bend.
• Standard prefabricated vertical tube sizes range from DN 20 mm to DN 32 mm (NPS 3/4”
to 1 ¼”) located inside boreholes with diameters between 90 mm & 135 mm (3 ½” & 5 ¼”).
• Common bore depths range from 60 m to 90 m (200 ft to 300 ft) – and up to 150 m.
The annular space between the tubing and borehole wall is filled with a thermally enhanced
grout containing sodium bentonite and silica sand (or graphite). This ensure heat transfer is
maximized between the piping loop and ground, as well as ensuring the borehole is sealed
from outside contaminants. The rationale for the relatively small borehole diameters at shal­
low depth fitted with small pipe sizes are as follows:
• Allows for small rotary drilling rig that is highly mobile – a high powered drive not
required.
• Pre-fabricated HDPE pipes can be inserted into the borehole via custom-made reels.
The system characteristics described above are for relatively small systems (cooling duties
< 750 kW). Typical mine refrigeration system duties however are an order of magnitude larger
and use relatively high water flowrates both on the cooling and heat rejection circuits. Con­
ceptual analysis indicates that HDPE pipe material will not be feasible. This is primarily due
to its relatively poor thermal conductivity, which adversely impacts heat transfer, and
secondly, its high surface roughness, which results in higher pumping power. Steel pipes,
therefore, are preferred at this conceptual stage of analysis. Simply put, HDPE will be cheaper
and easier to install, relative to steel, but will require a larger number of boreholes resulting in
higher pumping power requirements which will reduces the overall system efficiency.
The application of vertical GCHP to conventional mine refrigeration systems are new to
the mining industry, and as with most new technologies, some level of innovation will be

175
required for successful implementation – both from the design engineers and installation
Contractors.
Hence, DN 65 mm (2 ½”) threaded carbon steel pipes (schedule 40) located inside
90 m (300 ft) deep, 160 mm (6 ¼”) diameter boreholes are used for the remaining analysis.

4 GROUND CONDITIONS

Typical Virgin Rock Temperature (VRT) equations used in the mining industry take the
form of Equation 1 which assumes VRT increases linearly with depth. This is generally not
the case as shown in Figure 2 for borehole temperatures in a hot region. It can be seen that
there is a relatively constant rock temperature until a depth of about 150 m. However, for
most mining applications, the main aim of the VRT equation is to provide reliable VRTs
at sufficient depth where heat from rock strata starts having a noticeable effect on the
total mine heat load. Therefore, the “surface temperature (indicated)” in Equation 1 does
not represent the actual surface rock temperature, but rather an intersection of the vertical
axis (temperature) at zero depth that results from optimizing the curve fit to the non-
surface regions of interest. In the case of the example shown in Figure 2, the actual surface
rock temperature is about 25.0°C while the VRT surface temperature (indicated) is 23.5°C.
The “surface temperatures (indicated)” documented for common mining regions are indi­
cative only and need to be validated with site-specific near-to-ground thermal property
tests.

Sudbury is the location for this comparison and analysis and thus a VRT equation is
required. Cleland’s (1933) measurements found that the Frood Mine in Sudbury had
a geothermal gradient of about 1.14°C per 100m (1°F/160ft) with an indicated surface tem­
perature of 7.6°C (45.7°F). Misner (1950) found that Creighton Mine in Sudbury had
a geothermal gradient of about 1.59°C per 100m (1°F/115ft) with an indicated surface tem­
perature of 7.2°C (45.0°F), while more recent borehole tests in the Sudbury area suggest
a geothermal gradient of about 1.9°C per 100m (1°F/95ft) with an indicated surface tempera­
ture of -2.5°C (27.5°F). The above variations for the Sudbury area suggest that there are great
variations in geothermal gradients as well as indicated surface temperatures; hence, site spe­
cific borehole testing is imperative for accurate geothermal system assessment. For this ana­
lysis, however, the VRT equation stated in Equation (2) will be used to represent the Sudbury
area as it represents an average of the data stated above.

5 REFRIGERATION SYSTEMS

Conventional large-scale mine refrigeration systems typically employ a direct-contact heat

rejection strategy, meaning that the warm ambient air comes into direct-contact with the hot
water from the refrigeration machine condenser. There are many advantages to this type of
heat rejection method, including thermal and practical, but the return water temperature from
the Condenser Cooling Tower (CCT) to the condenser heat exchanger is limited by the ambi­
ent air wet-bulb temperature. Indeed, the return water temperature is typically 3-4°C above
the ambient air wet-bulb temperature. This limits the system efficiency for conventional sys­
tems, or more accurately, limits the refrigeration machine Coefficient of Performance (COP)
for a given location.

176
 COP is a metric used to define the “efficiency” of a refrigeration machine or an overall
refrigeration system and is defined as the ratio of cooling capacity provided to the electrical
energy input required to achieve this cooling. The higher the COP the better. For example,
a COP of 5 means that 1 kWe of electrical power is required to produce 5 kW of cooling.
“Refrigeration machine COP” is specific to the refrigeration machine only (evaporator duty
to compressor power) while “overall system COP” includes the whole system (air cooling duty
to total input power inclusive of pumping and ancillary systems). For the vapor-compression
refrigeration machines used in conventional mine refrigeration systems, refrigeration machine
COP is a function of the refrigerant evaporating temperature and the condensing temperature.
The closer these temperatures are to each other, the lower the compressor power requirement
and the higher the COP, for a given compressor isentropic efficiency. This is shown in
Figure 3 for typical mine refrigeration machine temperatures.
The hypothesis therefore is that the addition of a geothermal heat sink(s) to conventional
mine refrigeration systems will result in a lower refrigerant condensing temperature and hence
higher refrigeration machine COP for a given evaporating temperature. Three geothermal
Cases will be compared to the conventional Base Case.

Figure 1. Typical GSHP U-tube layout (left-unconsolidated & right-consolidated) [Kavanaugh et al.].

The design-day condition is taken at the ASHRAE yearly 98th percentile. The refrigeration
system is assumed on surface (Z = 0 m) close to the mine fresh air intake. All kW values refer
to thermal duty while kWe is used for electrical power. Air mass flowrates are for dry air and
temperatures are written as wet-bulb/dry-bulb.
Base Case (Figure 4) system comprises water-cooled R-134a centrifugal vapor-compression
refrigeration machine. Air cooling is done via a Bulk Air Cooler (BAC) and heat rejection via
CCT, both direct-contact type. Cold water is circulated from the BAC to the refrigeration
machine evaporator via the BAC pump while hot water is circulated from the CCT to the
refrigeration machine condenser. Refrigerant is circulated between the evaporator and con­
denser via the compressor. Cold, fresh air is induced into the mine via existing primary fans
(not included in the analysis). Make-up water is required at the CCT to account for

177
Figure 2. Typical borehole temperature results: VRT (°C) = [0.01 (°C/m) x depth (m)] + 23.5 °C.

evaporation and water quality. The BAC condensate flows to the CCT make-up water circuit
to reduce the total make-up water requirement. The system COP is 5.6.
Case 1 (Figure 5) adds a closed-loop geothermal heat sink circuit to pre-cool the evaporator
water returning from the BAC. The vertical GCHP geothermal configuration described in Sec­
tion 2 is employed. The required evaporator duty decreases due to the addition of the geother­
mal heat sink. This results in a smaller overall refrigeration machine duty and compressor, as
well as a smaller CCT. But a larger BAC pump is required to overcome the borehole pipe
resistance. The system COP increases to 6.0.
Case 2 (Figure 6) replaces the CCT in the Base Case with a closed-loop geothermal heat
sink circuit for heat rejection. Again, the vertical GCHP geothermal configuration is used –
albeit for a much larger duty. The evaporator duty requirement remains the same as the Base
Case, but the condenser duty reduces because of the smaller compressor duty requirement.
The reduced compressor duty is driven by the superior heat transfer characteristics of the geo­
thermal heat sink compared to the base case CCT resulting in much lower condenser tempera­
ture. Thus, a lower compressor lift is required between the evaporator and condenser circuits
which improves compressor COP (Figure 3). But again, a larger CCT pump is required to
overcome the borehole pipe resistance. The system COP increases to 8.2.
Case 3 (Figure 7) combines both Case 1 and Case 2 with a resulting system COP of 7.3. The
decrease in system COP when compared to Case 2 is primarily owed to the large electrical
load of the two geothermal water circuit pumps.

6 COSTS AND CO2 EMISSIONS

The estimated savings for both cost and CO2 emissions, relative to the conventional base case,
are shown in Table 2. OPEX includes electrical power ($0.08 per kWh) and provision for an
Ontario carbon tax at $155 per tonne of CO2 equivalent (average forecast over the next 15
years). OPEX does not include items such as maintenance, stationary engineer staffing or
make-up water. CAPEX is an all-in cost and includes both the refrigeration systems and the
boreholes and pipes for the geothermal heat sinks. The all-in refrigeration rate used is
$1,800 per kW of evaporator duty which is based on historical project data for similar sys­
tems. The borehole and pipe rate used is $80 per meter which is largely based on costs

178
Figure 3. Refrigeration Machine COP as a function of evaporator & condenser temperatures.

Figure 4. Base Case: Conventional water-cooled surface refrigeration system.

179
Figure 5. Case 1: Pre-cooled evaporator water via geothermal heat sink.

Figure 6. Case 2: Heat rejection via geothermal heat sink.

180
Figure 7. Case 3: Pre-cooled evaporator water and heat rejection via geothermal heat sink.

presented in Kavanaugh & Rafferty (2014) and adjusted for inflation as well as steel pipes
versus HDPE. Total Owning Cost (TOC) is used to evaluate the net effect of both CAPEX
and OPEX over a 15-year period with a 2% discount rate. The analysis makes provision for
4-months of continuous operation at peak duty which somewhat accounts for part-load oper­
ation and the 2 excluded shoulder months typical in Ontario, Canada.

7 DISCUSSION

The results shown in Table 2 suggest that the most optimal configuration is that of Case 3,
since both TOC cost and CO2 emissions are lower than in the conventional Base Case system.
However, there are merits to the other configurations assessed as well. If the goal is maximum
CO2 reduction or minimal electrical consumption (e.g., remote site without access to grid
power), then Case 2 is the best configuration with a marginal TOC savings. Alternately, if the
project is CAPEX sensitive, Case 1 suggests the lowest CAPEX with the benefit of lower CO2
emissions and a lower overall TOC.
The total length of borehole pipes required for adequate heat transfer for the geothermal
systems is relatively high ranging from 2.2 km to 8.8 km. The boreholes must also be suffi­
ciently spaced to reduce thermal interaction and thus relatively large land areas are required.
But most mines are in rural areas and therefore available land is likely. The boreholes must
also be located fairly close to the mine fresh air intake where the cooling system will be
located.
The geothermal configurations require less, or no make-up water compared to the conven­
tional Base Case system. This may be a major factor where make-up water is not readily avail­
able and/or handling of the CCT blowdown water is problematic. It also reduces the need for
sophisticated water treatment systems.
The geothermal configurations require relatively small refrigeration machine compressor
motor which may negate the need for higher voltage electrical equipment specifically for the

181
Table 2. Cost and CO2 savings versus the conventional base case system.
Base Case Case 1 Case 2 Case 3

CAPEX $ 17.92 mil $ 15.60 mil $ 18.74 mil $ 16.69 mil

Refrigeration System $ 17.92 mil $ 13.85 mil $ 13.44 mil $ 10.39 mil
Boreholes & Pipes - $ 1.75 mil $ 5.30 mil $ 6.30 mil
OPEX (per year) $ 0.44 mil $ 0.41 mil $ 0.29 mil $ 0.33 mil
Total Owning Cost $ 23.52 mil $ 20.88 mil $ 22.53 mil $ 20.93 mil
CO2 eq. emissions 145 t 136 t 98 t 110 t
OPEX & CO2 savings Base Case 7% 34 % 25 %
TOC savings Base Case 11 % 4% 11 %

refrigeration machine. There is therefore opportunity for a common low-voltage strategy for
the geothermal systems. However, special compressors may be required for some of the very
low lifts.
Although not explored in this paper, the application of geothermal to mine air heating sys­
tems commonly employed in cold regions potentially offers greater savings than those
explored in this paper. For example, a system comprising one heat pump which provides both
cold air to the mine in the summer and warm air to the mine in the winter. This all-in-one
system may eliminate the need for conventional large air heaters which typically burn natural
gas, or at the very least, supplement the air heaters reducing the natural gas consumption and
GHG emissions.

8 CONCLUSION

It has been shown that geothermal additions, in the form of vertical GCHP heat sinks, to con­
ventional mine refrigeration systems potentially offer both CO2 emission reduction and cost
savings. Multiple configurations were explored, yielding varying improvements on either cap­
ital or operating cost, and offering flexibility for the mine to select the best technology to suit
their specific goals and constraints. Technological innovation is likely required to achieve the
ambitious net-zero goals set by some major mining companies, and geothermal is presented as
a relatively simple option to supplement conventional mine refrigeration systems.

REFERENCES

Cleland, R.H. 1933. Rock Temperatures and Some Ventilation Conditions in the Mines of Northern
Ontario, Canadian Institute of Mining, Metallurgy and Petroleum.
Environment and Climate Change Canada. 2022. Annex 13, Electricity in Canada: Summary and Inten­
sity Tables, National Inventory Report 1990-2020: Greenhouse Gas Sources and Sinks in Canada –
Part 3.
Kavanaugh, S. and Rafferty, K. 2014. Geothermal Heating and Cooling, ASHRAE.
Misener, A.D. and Thompson, L.G.D. 1950. Temperature Gradients in Ontario and Quebec, Canadian
Institute of Mining, Metallurgy and Petroleum, Transactions, Volume LIII, pp. 380–383.

182
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Onaping Depth Project underground refrigeration plant update

D.W. Durieux, C. McGuire, T. Mehedi & D. Witow

Hatch Ltd.

E. Pilkington
Sudbury Integrated Nickel Operations, a Glencore Company

ABSTRACT: The Onaping Depth Project is a nickel/copper underground mine project cur­
rently in execution by Glencore’s Sudbury Integrated Nickel Operations. The Onaping Depth
orebodies will be accessed via the existing Craig Mine workings after completion of sinking
a winze from 1,200 to 2,635m below surface. Due to the final depth of the workplace, a 16
MW(R) installed capacity underground refrigeration plant will be operated at a depth of
1,915m below surface to provide conditioned air to the workplace throughout the project
development and life-of-mine production phases. This paper will provide an update on the
design, construction and commissioning of the underground refrigeration plant, including les­
sons learned in the contracting and construction methodologies.

1 INTRODUCTION

Glencore’s Sudbury Integrated Nickel Operations (Sudbury INO) are in the progress of exe­
cuting the Onaping Depth Project, an ultra-deep nickel/copper mining complex in Sudbury
Ontario Canada. Refrigeration has been a key consideration for the project since the study
phases due to the autocompression heat associated with the orebody depth (in excess of 2,500
metres below surface), high in situ rock temperature (45°C in deep workplaces) and large ven­
tilation volume to support mechanized mining. Through a series of studies and trade-offs, an
underground refrigeration plant was identified as the preferred technology to ensure safe
working conditions in the deep zone. This selection was made feasible by the project’s decision
to mine with exclusively battery-electric equipment, allowing for the air volume circulated to
depth (and therefore required excavation size for bulk air cooling) to be reduced by approxi­
mately 40% when compared to a typical diesel mining method (Wisniewski 2017).
Since 2016, feasibility-level and detailed engineering have advanced on the Onaping Depth
underground refrigeration plant. Major mechanical equipment was purchased in 2017 and
delivered to site in 2019-2020. Construction commenced at the 1915L (nominal metres below
surface) complex in early 2021 and commissioning of the refrigeration plant was substantially
completed in Q1 2023. The plant has been designed with the challenges of underground
refrigeration in mind. A unique layout was developed with a centrally-located plant room
between the two open spray chambers, minimizing excavation requirements and reducing
large bore pipe lengths, all while respecting the geotechnical constraints of designing large
excavations at a depth of 1,915 metres below surface.

2 PROJECT BACKGROUND

The Onaping Depth deposit has been known since the 1990s, when exploration drilling from
the existing Craig-Onaping mine complex identified the main orebody. The 18.4Mt deposit

DOI: 10.1201/9781003429241-19

183
resides in the vicinity of 2.6km below surface and will be mined at a target production rate of
1.2 million tonnes per annum with a design life of 20 years. The refrigeration plant will oper­
ate for the complete life of mine in addition to two to three years of the Project period
required to access the orebody.
Multiple attempts to develop a mine design which would be economical with conventional
diesel mechanized mining methods were unsuccessful. Mechanical refrigeration is necessary due
to drill-indicated virgin rock temperatures in excess of 45°C and auto-compression across more
than 2,500m of vertical depth. These factors resulted in excessive workplace temperature condi­
tions in the absence of cooling intervention. In addition, capital and operating costs associated
with the volume of air required to support a diesel fleet were found to be prohibitive.
In 2016, the Onaping Depth Project carried out a feasibility study on the basis of an entirely
battery-electric fleet for the deep workings. This allowed for a significant reduction in the ven­
tilation volume (approximately 40%; Wisniewski 2017), and therefore both the ventilation and
refrigeration power consumption (approximately 45%; Wisniewski 2017). In addition, major
capital cost savings were realized from a reduction in excavation sizes and development quan­
tities in the primary airways, in addition to the reduction in ventilation and refrigeration infra­
structure cost. The project was given approval to commence execution in late 2016, with the
first development blasts fired in January 2017.
Lateral development for mine access construction within the first five years of the project
timeline has been executed with diesel equipment. All of this development is accessible from
the existing ramp system in Craig Mine, and is supported by existing diesel distribution and
fueling infrastructure. The lower workings of the mine, known as “The Depth”, will only be
accessible via a new internal winze excavated from 1150L to 2635L. All mobile equipment in
use in the Depth for development or production will be battery-electric vehicles (BEVs). Shaft
sinking is expected to be finished in 2023, at which time the first BEVs will be required on site.
This execution strategy allowed for five years of research and development from the
announcement of project execution until the first heavy vehicles (load-haul-dump (LHD)
machines and haul trucks) were required on site.

3 REFRIGERATION PLANT DESIGN

During early engineering phases, the Onaping Depth Project undertook detailed assessments
and trade-offs to determine the optimal cooling plant location. Overall, locating the cooling
plant underground at 1915L was deemed the most favourable from a technical and economic
standpoint when compared with a surface refrigeration installation. This was largely attrib­
uted to the improved positional efficiency of an underground plant location, as there is
a reduced temperature differential with the rock strata upstream of the cooling plant for an
underground plant location and therefore less driving force for heat transfer from the strata.
The previous work has identified that powerful synergies arise due to the combination of
underground cooling with the reduction in primary ventilation volume attainable when regu­
latory requirements for diesel exhaust dilution do not exist. This synergy results in reductions
in both capital and operating costs when compared to surface plants, and is explored in detail
by Brown et al.(2018).
Key process conditions forming the basis of the Onaping Depth underground cooling plant
design are summarized in Table 1.
Note that these parameters are specific to the life-of-mine (maximum) design condition for air
flow and refrigeration duty; the plant will be operated at a reduced air flow with a target chilled
air temperature of 10°C during the development and construction phases of the project.
The cooling plant consists of the following main components: bulk air cooler (BAC) for air-
cooling, refrigeration plant room (housing the refrigeration machines to provide chilled water to
the BAC), condenser spray chamber (CSC) for system heat rejection, water pumps and water
treatment for the recirculating water circuits between the refrigeration machines and spray cham­
bers. Both the BAC and CSC consist of two-stage, unpacked horizontal spray chambers with
sloped basin floors to minimize slimes build-up and facilitate cleaning. Nominal dimensions of

184
 Table 1. Summary of air side process parameters for refrigeration plant design.
Parameter Units Value
3
Air Flow (Volumetric) m /s 229
Mass Flow (Dry Air) kg/s 308
Inlet Air Temperature °C wb/db 27.6/30.5
Chilled Air Temperature °C wb/db 12.0/12.0
Barometric Pressure kPa 120.5
Nominal Air Cooling Duty kWT 14,600
Air Temperature Available for Heat Rejection °C wb/db 23.1/26.1
Maximum Air Flow Available for Heat Rejection m3/s 203

each spray chamber are 8.0m wide x 7.0m high inside shotcrete ground support, with the max­
imum height of 8.0m high at the low point in the sloped floor.
The Plant Room is 9.0m wide x 8.2m high and contains three R-134a single-stage centrifu­
gal compressor refrigeration machines (Johnson Controls Model YK), arranged in a series-
counterflow configuration. Water treatment systems consist of side-stream filters and chemical
dosing equipment including corrosion inhibitor, biocide, and alkali for pH control.
A summary of the water side process conditions is provided in Table 2.

Table 2. Summary of water side process parameters.

Parameter Units Value

Refrigeration Machine Duty (Installed Capacity) kW(R) 16,200

Refrigeration Machine Duty (Operating Duty) kW(R) 15,400
BAC Water Flow L/s 400
BAC Water Inlet Temperature °C 7.5
BAC Return Water Temperature °C 16.5
CSC Water Flow L/s 550
CSC Water Inlet Temperature °C 44.1
CSC Return Water Temperature °C 35.8
Heat Rejection Duty kW(R) 19,600

Figure 1. 3D model overview of onaping depth 1915L refrigeration plant.

A 3D rendering of the various system components on 1915L is shown in Figure 1. The

plant layout was designed with the following key design aspects in mind:

185
• Plant room centrally located to minimize pump and piping requirements between refriger­
ation machines and spray chambers,
• Chillers installed in series counterflow arrangement due to high lift and desire to avoid two-
stage centrifugal compressors,
• Sloped spray chamber basin floors to minimize excavation quantity,
• Equipment access and maintenance requirements including overhead crane,
• Geomechanical considerations for drift size and spacing.
The sloped floor of the BAC can be seen in Figure 2, which is a similar design for the CSC.
In addition to minimizing excavation volumes, the sloped concrete basins will facilitate wash-
down of collected dust residues during plant outages as needed. Stainless steel work platforms
are provided above the water level to allow for inspection and maintenance of the spray pipes
without having to drain the full system volume from the basin. The depressed centre section
visible is local to the pump suction (see section view in Figure 3). The spray chamber floor is
excavated approximately two metres above the adjacent pump room floor in order to provide
a suitable minimum suction head of 1.5 m at spray chamber basin low water level.

Figure 2. Long section view of BAC spray chamber.

Figure 3. Section view of sump and pump suction.

Design of the plant prioritized robust, reliable, and safe performance in an underground mining
environment. For example, the refrigeration machines were specified with high heat exchanger
fouling factors reflective of conditions in underground mining operations. Heat exchanger tubes

186
were specified to be thicker than standard and constructed of more robust copper-nickel alloy
material for long-term protection against both the corrosion and the erosion expected in an under­
ground environment with the expected water quality. The chiller units were also specified to use
a Class A1 refrigerant with low toxicity and no flame propagation to minimize risk to workers in
the event of a leak in the confined underground environment. In addition, the airflow in the plant
room is ventilated directly to exhaust, reducing personnel risk from refrigerant leaks.
Another design aspect critical to the successful operation of the plant was inclusion of water
treatment systems. An appreciable portion of the dust entrained in the mine ventilation air
stream is expected to be captured in the spray chambers as the water sprays act like a wet par­
ticulate scrubber. A side stream filter is included in the design of the CSC water circuit to
remove suspended solids and prevent clogging of the condenser tubes or CSC spray nozzles.
The BAC water circuit is expected to see less contamination as it is on the fresh air side of the
mine ventilation circuit; however, provision for installation of an identical side stream filter
has been made in the excavation and layout allowance in case of issues with suspended solids
on the evaporator water circuit.
In addition, chemical dosing to inhibit corrosion, scaling, and biofilm formation is included
in the design. This is critical to ensure proper water quality in the equipment, thus preventing
material degradation and ensuring proper heat transfer performance, as well as minimizing
personnel risk to biological contaminants such as legionella (Legionnaire’s Disease).
The refrigeration plant was standardized to operate at 4,160V, with motors for all chillers,
pumps, and the nearby primary ventilation fans and dewatering pumps all operating at this
specification. Power distribution to the level is accomplished by a pair of 25kV shaft feeders
from surface to the main substation near the entrance to the Plant Room, at which point it is
stepped down to 4,160V and 600V for ancillary loads.

4 CONTRACTING STRATEGY

Construction of underground refrigeration plants extends beyond the typical skill set of most
underground mining trade workforces. In particular, fabrication and construction of custom,
large bore pipes (DN450 and DN500) and work with refrigeration equipment are unusual in
the underground environment. However, some of the construction activities such as construc­
tion of concrete bulkheads, drilling of rock anchors for pipe supports and structural steel tie-
backs, etc. are well-suited to mining contractors with experience in drilling and underground
construction.
The project team sought to develop a contracting model which would allow for both the high
degree of specialization associated with the cooling systems and the expertise in underground
construction. A lump sum contract model was selected, with an approved bidders list consisting
of underground mining contractors and contract language which ensured that dedicated piping
fabricators and tradespeople would be engaged from more traditional “surface contractors”.
This decision also intended to put the onus on the successful contractor to receive personnel and
materials at the shaft station(s) underground and provide transport to the refrigeration plant
workplace – approximately 3.5 km down ramp from the lowest shaft station.
Comprehensive tender evaluations revealed that the contract arrangement resulted in
a detrimental outcome primarily due to subcontracting markups included in all bid prices. In
response, an alternate strategy was proposed in which the total scope was divided into mul­
tiple subcontracts, each consisting of a single specialized trade. Bidders lists were crafted such
that some contracts were only bid to mining contractors, while others were only bid to special­
ist surface contractors. The division of scope is listed in Table 3 below.
In addition to these four construction contracts, an Underground Logistics contractor was
brought onto site who took responsibility for delivery of all construction materials and per­
sonnel from the shaft station to the contract workplaces. Thus, the single lump-sum contract
became five specialized contracts which were all tendered competitively. Although an increase
in project management and contracts administration oversight is required, the reduction in
construction cost was approximately 27%.

187
Table 3. Construction contract summary.
Contract Scope Included Contractor Type

Messenger Wire Installation Installation of electrical messenger system includ­ Mining

Concrete
Structural/Mechanical/Piping
Electrical & Instrumentation
ing grounding & tensioning
Slab on grade, equipment foundations, formed Mining or Civil
basins, dam walls, shotcrete ventilation
bulkheads
Structural steel, mechanical equipment (chillers, Mechanical
pumps, auxiliary ventilation), and refrigeration
plant piping
Substation construction, cable pulls, instrumenta­ Electrical
tion installation, communications

A downside of the approach of using multiple contractors is the difficulty of managing

handover of completed workplaces to subsequent contractors. Delays in early contracts (e.g.
concrete) introduce the risk of delaying mobilization of subsequent contractors, which can
produce financial penalties and/or loss of key personnel. Careful construction management is
required which can respond to ongoing disputes, re-prioritize work to allow for handover of
isolated workplaces, and generally facilitate discussion and co-operation between multiple
contractors sharing the same workplaces.

5 CONSTRUCTION & COMMISSIONING PROGRESS

At the time of writing in November 2022, the refrigeration plant construction (aggregate of all
contracts) was substantially complete. Photos below show milestones at the completion of chiller
installation, with the overhead bridge crane used for installation and maintenance visible above
(Figure 4), and an end view of one chiller following piping tie-in and insulation (Figure 5).

Figure 4. Refrigeration plant room following completion of mechanical installations.

188
Figure 5. View of insulated evaporator and evaporator piping.

Figure 6. View of CSC pipes, pipe supports and access platform following construction.

189
 The BAC and CSC follow similar construction methodologies, with lower spray headers
supported from the platform structure and upper spray headers supported on beams spanning
the width of the spray chamber. This is visible in Figure 6.

Figure 7. Refrigeration plant schematic showing commissioning subsystems.

A commissioning strategy was developed which separates the refrigeration plant into five
subsystems which were commissioned independently as shown in Figure 7. This approach
improves flexibility in the execution of the field work – where work on any subsystem can
start as soon as construction completion allows.

6 CONCLUSIONS

For the Onaping Depth Project, replacement of diesel equipment with BEV’s in the mining
plan allowed reduced airflow and heat loads in the mine.
An underground cooling plant was selected to provide bulk air cooling in close proximity to
the deep mining operations. This was shown to be more economical than a surface cooling
plant for the specific circumstances of the Onaping Depth mine design.
Equipment selection included single-stage, water cooled, centrifugal mechanical chillers
with synthetic refrigerant and shell and tube heat exchangers.
Both bulk air cooling and condenser heat rejection were designed to use direct contact open
spray chambers adjacent to the refrigeration plant room.

190
 Contracting strategy for the underground construction works was found to be sub-economic
using a single lump sum contract, and cost savings were achieved by dividing work to specialist
contractors and intensively coordinating to minimize work clashes between these packages.

REFERENCES

Brown, M., Guse, T., Arsenault, S. & Rogers, B. Effects of an Electric Fleet on Mine Ventilation and
Refrigeration with a Comparison to an Equivalent Diesel Mine. Paper presented at the Roomvent &
Ventilation Conference; 2018 Jun 2–5; Epsoo:Finland.
Wisniewski, S. Onaping Depth Project. Presented at CIM Sudbury General Meeting; 2017 Mar 23; Sud­
bury:Canada.

191
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

A direct equation for sigma heat and wet-bulb temperature for

underground ventilation applications

A. Pandey
Department of Mining Engineering, BIT Sindri, India

S. Jayaraman Sridharan
South Dakota Mines, Rapid City, SD, USA

B.S. Sastry
Department of Mining Engineering, IIT Kharagpur, India

ABSTRACT: In an underground ventilation system where there is a change in the moisture

content of air, sigma heat is useful to identify and quantify the heat transfer processes. The
literature on psychrometry suggests that the sigma heat is a function of barometric pressure
and wet-bulb temperature. However, a direct expression of this relationship has been elusive.
This paper presents a direct and straightforward expression for sigma heat as a function of wet-
bulb temperature and barometric pressures within the range generally encountered in the under­
ground mine environment. Also, a direct equation for wet bulb temperature as a function of
sigma heat and barometric pressure is presented, which otherwise require lengthy calculations
involving iterations. The difference between calculated and predicted sigma heat is within +1.32
kJ/kgda and -1.39 kJ/kgda; and wet bulb temperature is within +0.28 °C and -0.64 °C, respect­
ively. The efficacy of these prediction equations is quantified with some examples involving the
performance parameters of cooling tower and spray chambers. Moreover, a psychrometric chart
with sigma heat as a parameter in place of enthalpy is produced with the help of MATLAB code
and presented in this paper.

Keywords: Sigma heat, Wet-bulb temperature, Mine ventilation, Underground mine envir­
onment, Prediction model, Psychometric chart

1 INTRODUCTION

Calculation of heat addition in a segment of an underground airway is essential to obtain

a broad picture of the heat transfer processes from different sources under different condi­
tions. Various researchers (Hemp, 1985; Maurya et al., 2015) presented a detailed description
of these heat sources in their research. As the unsaturated air travels through the airway of an
underground mine, a sensible heat transfer from the air to the water will take place as water
molecules leaves the partially wet walls, drains, roof, and floor of the airway. This increases
the latent heat of air and these exchanges will continue until saturation (McPherson, 1993). In
an adiabatic humidification of air, the sensible and latent heat of the air is balanced by one
another, and no net addition or heat loss is involved in the process.
When latent heat transfer from the mine airway is negligible compared with that due to convec­
tion, then the roadway is defined as dry roadway (Jordan, 1965). However, an underground mine
airway is always wet to some degree (Nie et al., 2018). Therefore, mass has also been added to the
air in the form of water molecules which already contained sensible heat before evaporation.

DOI: 10.1201/9781003429241-20

192
Hence, the enthalpy (H) of the air does not remain constant in an adiabatic humidification pro­
cess. However, enthalpy minus the sensible heat of evaporated liquid water would be a property
value that remained genuinely constant and known as sigma heat (S) (Watson, 1981). The energy
contained within the air can be described by S (Ryan & Euler, 2017).
Since the concept of sigma heat was initially been discussed by Carrier (1911) in 1911, it has
not found wide use by air conditioning engineers until it was recognized by Whillier (1969)
and Bluhm (1981). Later, Bluhm (1990) shown that the sigma heat differences are more accur­
ate than the traditional enthalpy differences in approximating total heat transfer from a wet
surface and, the use of S is more accurate than H in the typical approximate energy balance
relationships between water and air streams.
It should be noted that the sigma heat is a unique function of wet-bulb temperature (Tw) at
a given barometric pressure (P). Hence, thermal exchange in an airway can easily be quanti­
fied from psychrometric observations using S. Moreover, as a directly measurable parameter,
a change in Tw indicates heat transfer from the machines, strata, or any other source, immedi­
ately. Estimation of S is necessary for a variety of heat exchange problems in mines, such as
designing of spray chambers (Bluhm & Whillier, 1978) and in the calculation of the factor of
merit in cooling towers (McPherson, 1993).
Several researchers presented approximation equation for Tw as a function of dry-bulb tem­
perature (Td) and either dew point temperature (Tdp) or relative humidity (RH). The standard
method of determining Tw from Td and Tdp is “one of the most involved and time consuming
calculations in psychrometrics” because it require iterations involving several psychrometric
equations (Chau, 1980). Al-Ismaili et al. (2016) presented an iterative approach to find Tw
from Td, RH, and actual station pressure values. An empirical equation for Tw as a function
of Td and RH is given by Estrada-Jaramillo et al. (2014) using a meteorological database. To
determine Tw for a given S and P, McPherson (1993) adopted an iterative method by iterating
between several psychrometric equations. Stull (2011) presented an empirical equation for Tw
by using Gene-expression programming (Ferreira, 2006) regression method. He used wet-bulb
depression (Td-Tw) and RH as input to the model for P of 1 atm. Mine ventilation software
such as VentSim and Vuma use iterative methods for the psychrometric calculations.
Although many equations are given by earlier research, in many ventilation problems, due
to statutory compliance or design requirements, Tw needed to calculated from given heat
values, and in many cases Td and RH values might not be known beforehand. Therefore, an
expression for Tw for a given S and P is desirable in such circumstances.
The aim of this paper is to provide an equation for sigma heat only as a function of and Tw
and P, which has the advantage of reducing the complexity of the computation. Also, an
expression for Tw as a function of S and P is also presented. Moreover, example of cooling
tower and spray chambers calculation is given to show the efficacy of the prediction equa­
tions. Furthermore, a psychrometric chart based in-terms of S is also presented in this study
which will provide a more intuitive and visual understanding of the underground psychro­
metric processes.

2 METHODOLOGY

2.1 Psychrometric equations

Psychrometry is the science of air-water mixture. Over the years, many formulations of the psy­
chrometric properties of humid air have been proposed. They differ in the levels of complexity
and accuracy (McPherson, 1993; ASHRAE, 2001; Nelson & Sauer, 2002). In the present study
McPherson’s work (1993) is taken as the reference for modeling and evaluation purpose.
Psychrometric equations relevant to the present study are given in Eqs. 1 to 8. Sigma heat,
S, can be calculated either by using Eqs. 7 or 8. It should be noted that Eq. 7 requires H,
which along with other parameters, is a function of Td of air. However, by using Eq. 8 and
Eqs. 1, 2, and 5, S can be calculated for a given P and Tw. Both equations (Eqs. 7 and 8) are
considered on a dry air (da) basis (kJ/kgda).

193
 Where Lw is latent heat of evaporation of water, X is moisture content of air, P is barometric
pressure in kPa, e is actual vapor pressure, Cw is specific heat of liquid water (4.187 kJ/kgK), Td
is dry bulb temperature of air in °C, Cpa is specific heat of dry air at constant pressure (1.005
kJ/kgK), Cpv is specific heat of water vapor at constant pressure (1.884 kJ/kgK), esw is saturated
vapor pressure of air at Tw, and Xs is the saturated moisture content of air at Tw.

2.2 Prediction models

In order to understand the relation among S, P, and Tw, 5000 random values are generated for
P and Tw between the ranges 99 kPa to 120 kPa and 5 °C to 50 °C, respectively. The range of
values is chosen to cover the values encountered in Indian underground mines. The values of
S are calculated using Eq. 8 (and Eqs. 1, 2, and 5). The value of calculated S falls in the range of
16.6 kJ/kgda to 256.4 kJ/kgda, for the above range of P and Tw. Figure 1 shows the variation of
S with Tw and P. Figure 2 is plotted simply by switching the axes of Figure 1 for better visual
understanding of the variation of Tw with S and P.

Figure 1. Calculated sigma heat vs. wet-bulb temperature and barometric pressure.

194
Figure 2. Wet-bulb temperature vs. calculated sigma heat and barometric.

It can be seen from Figure 1 that S increases exponentially with an increase in Tw. Also,
Figure 3 shows that S decreases as P increases. Also, there is a higher rate of reduction in S as
P increases, at higher values of Tw.

Figure 3. Calculated sigma heat vs. barometric pressure at different wet-bulb temperature.

Based on the relationship observed in Figures 1, 2 and 3, prediction model for S (kJ/kgda)
and Tw (°C) are developed using regression analysis and expressed as:

195
 0:73
where x ¼ S:P
1000

3 PSYCHROMETRIC CHART

Using the psychrometric equations to plot the psychrometric curves, a numerical code had
been developed in MATLAB R2016a to plot a psychrometric chart in-terms of S, from 5°C to
50°C dry bulb temperature, and atmospheric pressure range of 99 kPa to 120 kPa. The rele­
vant thermodynamic psychrometric equations are used from published ASHARE handbook
(ASHARE, 2013). The psychrometric charts obtained from the current study at 1 atm or
101.325 kPa is shown Figure 4. Except constant Sigma heat lines, the chart presented in
Figure 4 resembles the presently available psychrometric charts, and is validated as described
later which establishes the credibility of the developed code and reliability of the present chart
for practical purposes.

Figure 4. Psychrometric chart for 1atm pressure and 5 to 50 °C in-terms of Sigma heat.

4 RESULTS AND DISCUSSION

As the depth of the mines increases, P increases irrespective of the cooling system used in mines.
Generally, in non-artificially cooled mines, P and Tw both increase simultaneously with an
increase in depth. Taking the above-mentioned factors and the general climatic condition in an
underground mine into consideration, the ranges of P (99 kPa to 120 kPa) and Tw (5 °C to 50 °
C) are selected. In order to know the performance of the proposed models, 5000 randomly gen­
erated data points of Tw and P are used to calculate S by Eq. 8 (calculated sigma heat) and
Eq. 9 (predicted sigma heat). These data points are plotted as a scatter plot and shown in
Figure 5. Also, same data set is used to calculate Tw as a function of P and S by Eq. 10.

196
Figure 5. Calculated vs. Predicted sigma heat for 5000 randomly generated values of sigma heat.

Figure 6. Calculated vs. Predicted sigma heat for 5000 randomly generated values of Wet-bulb
Temperature.

The relationship between the calculated and predicted sigma heat shown in Figure 5 & 6 is
easy to comprehend. An almost superimposed relationship between the data points and 1:1 or
reference line, suggests the good performance of the prediction model. Also, the Pearson’s cor­
relation coefficient (r) of 0.999 shows the strong positive correlation between calculated and
predicted S as well as for calculated and predicted Tw.

197
 Table 1. Maximum error, minimum error, and root mean square
error (RMSE) for the prediction models for S and Tw (Eq. 9 and 10
respectively).
Parameters Sigma heat (kJ/kgda) Tw (°C)

Maximum error 1.32 0.28

Minimum error -1.39 -0.64
RMSE 0.77 0.11

While calculating S for 5000 data points of P and Tw in MATLAB R2016a (MATLAB,
2016), the time taken for the calculation by the proposed equation (Eq. 9) is 25 % less than
that by using Eq. 8. The main reason for this is that the number of variables needed to be
addressed is less for the proposed equation. This saves the computation time and resources
required to calculate S, especially in the software tools applicable for modelling, designing,
and simulation of an underground mine environment.

4.1 Comparison among sigma heat values

In order to know the performance of presented psychrometric chart (Figure 4), the S values at
different Tw and 1 atm pressure obtained from chart and prediction equation (Eq. 9) are com­
pared with theoretical values of S (Eq. 8) and tabulated in Table 2.

Table 2. Sigma heat values from theoretical equation (Eq. 8), psychrometric chart (Figure 4) and predic­
tion equation (Eq. 9), with respective absolute percentage deviation from theoretical sigma heat values
at P = 1 atm.
Wet-bulb Theoretical Predicted Deviation from Sigma heat from psy­ Deviation from
temperature Sigma heat Sigma heat Theoretical value chrometric chart Theoretical value
(°C) (Eq. 8) (Eq. 9) (%) (Figure 6) (%)

5 18.47 18.60 0.71 18.48 0.02

10 28.96 29.63 2.32 28.96 0.01
15 41.33 42.20 2.09 41.34 0.02
20 56.16 56.86 1.26 56.17 0.03
25 74.14 74.38 0.32 74.18 0.04
30 96.21 95.79 0.44 96.27 0.06
35 123.54 122.48 0.86 123.64 0.08
40 157.70 156.37 0.84 157.84 0.09
45 200.77 200.05 0.36 200.98 0.10
49 243.57 244.37 0.33 243.82 0.10

The maximum and minimum percentage deviation of S obtained from the presented psy­
chrometric chart, from theoretical S values are found to be 0.10 % and 0.00 %, respectively.
Also, the percentage deviation of predicted S from theoretical S is between 0.32 % to 2.31 %.
This suggests that the values are acceptable for all kinds of underground calculation purposes.

5 CONCLUSION

Based on the fundamental formula given in literatures, a new prediction equation for S is derived.
The prediction results show a high level of agreement with the calculated values with acceptable
error limits. The prediction model also helps in faster calculation of S and consumes less computa­
tional resources. The results suggest that the prediction model may be an acceptable substitute for
the theoretical equation for S in a variety of underground climate simulation software.
Using the available thermodynamic data and psychrometric relations, a chart with constant
Sigma heat lines is plotted in MATLAB for Td of 5 to 50 °C and pressure range of 99 to 120

198
kPa. Comparison with a previously available psychrometric equations shows that the chart
obtained in the current study is equally accurate with an absolute deviation less than 1.84 %.
Moreover, reliance on the data taken from available sources, which are equally accurate, con­
firms the reliability of the developed chart. The success of this work could allow the investiga­
tor to extend this to study at different pressures and at different water-vapor systems.
Although many iterative approaches have been in use for years, presently, there is no direct
equations to determine Tw for a given S and P. The most important part of the present work
is to provide a direct equation for Tw as a function of S and P to avoid this iterative exercise.
The model has an acceptable fit with R2 of 99% and RMSE of 0.11 °C. The prediction equa­
tions presented in the study facilitates a reliable tool to ventilation engineer for the analysis
and faster calculations in various underground heat transfer problems.

REFERENCES

Al-Ismail, A.M. & Al-Azri, N.A. 2016. Simple iterative approach to calculate wet-bulb temperature for
estimating evaporative cooling efficiency. Stress (FITS). 12:13.
ASHARE. 2013. Psychrometrics. In ASHARE fundamental handbook. American Society of Heating
Refrigeration and Air-Conditioning Engineers, Georgia, USA.
ASHRAE, H.O.F. 2001. ASHRAE fundamentals handbook. American Society of Heating Refrigeration
and Air-Conditioning Engineers. Atlanta GA.
Bluhm, S. 1981. Performance of direct contact heat exchangers. Journal of Mine Ventilation Society of
South Africa. 34(Aug.):155–160.
Bluhm, S.J. 1990. Thermal performance of direct-contact water-air heat exchangers. University of the
Witwatersrand, Johannesburg.
Bluhm, S.J. & Whillier, A. 1978. Design of Spray Chambers for Bulk Cooling of the Air in Mines. Jour­
nal of The South African Institute of Mining and Metallurgy. 79(1):1–9.
Carrier, W.H. 1911. Rational psychromatric formulae. Transactions ASME. 33:1309.
Chau, K. V. 1980. Some new empirical equations for properties of moist air. Transactions of the ASAE.
23(5):1266–1271.
Estrada-Jaramillo, M., Vera-Romero, I., Mart\’\inez-Reyes, J., Ort\’\iz-Soriano, A. & Barajas-Ledesma,
E. 2014. Empirical model to calculate the thermodynamic wet-bulb temperature of moist air. Engineer­
ing. 2014.
Ferreira, C. 2006. Gene expression programming: mathematical modeling by an artificial intelligence.
V. 21. Springer.
Hemp, R. 1985. Air Temperature Increases in Airways. Journal of the Mine Ventilation Society of South
Africa. 38(2):13–20.
Jordan, D.W. 1965. The numerical solution of underground heat transfer problems—I Method relating
to dry roadways. International Journal of Rock Mechanics and Mining Sciences & Geomechanics
Abstracts. Pergamon. 2(3):247–270.
MATLAB. 2016.
Maurya, T., Karena, K., Vardhan, H., Aruna, M. & Raj, M.G. 2015. Potential Sources of Heat in
Underground Mines – A Review. Procedia Earth and Planetary Science. 11:463–468. DOI: 10.1016/j.
proeps.2015.06.046.
McPherson, M.J. 1993. Subsurface Ventilation and Environmental Engineering. Dordrecht: Springer Neth­
erlands. DOI: 10.1007/978-94-011-1550-6.
Nelson, H.F. & Sauer, H.J. 2002. Formulation of high-temperature properties for moist air. HVAC&R
Research. 8(3):311–334.
Nie, X., Wei, X., Li, X. & Lu, C. 2018. Heat Treatment and Ventilation Optimization in a Deep Mine.
Advances in Civil Engineering. 2018. DOI: 10.1155/2018/1529490.
Ryan, A. & Euler, D.S. 2017. Heat stress management in underground mines. International Journal of
Mining Science and Technology. 27(4):651–655. DOI: 10.1016/j.ijmst.2017.05.020.
Stull, R. 2011. Wet-bulb temperature from relative humidity and air temperature. Journal of applied
meteorology and climatology. 50(11):2267–2269.
Watson, A.G. 1981. The contribution of conveyed coal to mine heat problems. University of
Nottingham.
Whillier, A. 1969. An improved analysis of counterflow cooling tower performance. Journal of South
African Mechanical Engineer. 18(9):226–232.

199
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Case study: Refrigeration requirements during mineshaft

excavation as a function of heat stress index

K. Tom
Howden, St-Bruno de Montarville, Canada

ABSTRACT: Deep mines often require cooling to comply with local heat stress regulations.
Attempts to mitigate heat strain via improved heat indices are well documented. However, the
impact of the heat-stress index on refrigeration capacity is less explored. The paper examines the
refrigeration capacity necessary to develop a deep mineshaft as a function of both wet bulb globe
temperature (WBGT) and wet bulb temperature (WB). Thermodynamic simulations in Ventsim
indicated chiller capacity sized entirely from a wet bulb or sigma-heat (σ) balance was insufficient
to prevent elevated heat stress, at working depth, according to the WBGT. The study showed
that the primary factor causing increased wet-bulb depressions (the difference between wet bulb
and dry bulb temperatures), was input heat containing an overwhelming percentage of sensible
heat (dry heat) to total heat. Furthermore, the work investigated active engineering controls like
increased airflow quantity, duct insulation, and secondary cooling to help support the refriger­
ation equipment constraints, such as minimum collar WB, and ventilation design criteria. Add­
itional research is necessary to develop empirical methods that could facilitate the prediction of
cooling requirements for mine sites that do not employ WB as the design heat index.

1 INTRODUCTION

1.1 Background
As the world’s reliance on essential metals and minerals increases it will become necessary to
mine deeper to extract the ore. Inherent challenges that workers face in deep underground
mines are autocompression and strata heat, which can result in elevated cases of heat stress
arising from increasing temperature and humidity. NIOSH (2016) defines the former as the
heat load that an individual is exposed to from the aggregate contributions of environmental
and physiological factors, which result in a net increase in heat storage in the body. Mines
employ heat stress indices, which is a single number that predicts the level of heat stress in
a hot environment, to comply with local legislations on heat criteria. However, there is a lack
of universal agreement on a one-size-fits-all index due to the following constraints:
• be accurate and feasible,
• account for both environmental and physiological cursors of heat stress,
• calculations or measurements that are easy to perform,
• threshold limits that function under a wide range of metabolic and environmental conditions
• a range that is physiologically or psychologically weighted to inform of an increased risk to
health and safety (Epstein and Moran, 2006; NIOSH, 2016).
State-of-the-art research on the choice of heat stress indices typically involve metabolic rate.
Brake and Bates (2001) demonstrated a method to compare indices on the basis of the particular
index’s maximum metabolic rate. The results indicated that an increase in air velocity permitted
more strenuous work (i.e. higher metabolic rates), however, the authors were unable to define

DOI: 10.1201/9781003429241-21

200
the optimal index as none of the considered indices spanned the full range of maximum meta­
bolic rates. Moreover, the method becomes increasingly subjective with the inclusion of worker
acclimatization and clothing factors. Instead, Roghanchi, Sunkpal, Carpenter, and Kocsis
(2016) recommended the use of rational (i.e. physiologically based) or empirical indices, such as
wet bulb globe temperature (WBGT), during the design stage and direct indices, like the psy­
chrometric wet bulb temperature (WB), for thermal management. Despite the study’s advances,
it failed to address the practical ramifications that the category of heat stress index would have
on engineering control measures.

1.2 Motivation of present study

Deep mines that wish to minimize disruptions to production, and maintain compliance with occu­
pational legislations on heat stress, typically rely on engineering controls like increased ventilation
and mechanical refrigeration to protect its workers. Beyond a certain depth, artificial cooling
remains the only feasible option. The present work explores how an empirical heat stress index,
like the WBGT, can influence the ventilation and cooling design when compared to a more basic
heat stress index such as the WB. This is assisted by simulations using VentsimTM DESIGN.

2 CASE STUDY

2.1 Model description

A non hard-rock mine plans to sink a deep shaft using a roadheader-like machine. Part of the
due diligence requires a combined ventilation and thermodynamic model to predict the level
of heat stress on workers as the cutting device nears working depth. Figure 1 shows
a schematic of the current application’s push-pull ventilation strategy.

Figure 1. Ventilation schematic with insets from VentsimTM model showing heat and cooling sources.

201
 A dual-stage auxiliary fan supplies fresh air to the shaft bottom through a single duct
column. Downstream of the fan assembly and inline with the duct is a bulk air cooler (BAC)
that chills the air to a specified WB at the shaft collar. The cool air ventilates a number of
service and work decks situated on the cutting machine. The diagram’s inset of the cutter area
indicates the heat sources in the different airways. Notably, air exiting “Airway 03” splits two
ways: a small percentage intakes through another duct that connects to a muck filtration
system (“Airway 08” in the inset), while the remainder meanders through the upper decks.
The two air streams subsequently mix and return to the surface through a ducted, single-stage
exhaust fan. Furthermore, the duty point of this fan is such that ambient air makes up
a percentage of the total exhaust flow.
The heat loads in the model consist of autocompression, fan compression, strata, duct-
airway heat transfer, hydration, and electric machinery. All sources, with the exception of the
strata, are treated as sensible heat. The magnitude of evaporation from the rock depends on
the level of shaft wetness and its age. The shaft-cutter supplier provided the quantity and loca­
tions of electric heat along the machine, while hydration heat calculates from Nykänen’s equa­
tion (Al-Neshawy, 2015).

2.2 Design criteria

Due to accessibility and measurement ease, the WB is often the direct heat stress index of
choice for mine cooling design and routine monitoring of the underground climate (Belle and
Biffi, 2018). Typically, mines ensure the safety of its workers by stipulating a maximum return
or reject WB. One method to define the reject WB is displayed in Figure 2, which specifies an
acceptable level of productivity loss up to 28°C.

Figure 2. Productivity loss as a function of psychrometric wet-bulb temperature (Belle and Biffi, 2018).

Moreover, the WB can assess the air’s heat absorption capacity before requiring refriger­
ation through the cooling power surplus (CPS) described in equation 1,

where ṁ is the mass flow of dry-air (kg/s); σ is the sigma heat (kJ/kg) with subscripts denoting
the reject and inlet WB, respectively. Sigma heat is calculated from equation 2 (McPherson, 1993)

where ω is the absolute humidity (kgv/kga) with the subscripts, “v” and “a”, signifying
water vapor and dry-air, respectively.

202
 Some mines, in particular those located in South Africa and Australia, may impose add­
itional constraints like a maximum DB (i.e. 37°C) or an air velocity (V) of at least 0.25 m/s
within a risk management plan that objectively categorizes an environment as being hot
(Webber, Franz, Marx, and Schutte, 2003). One distinct observation is that these conditions
can primarily be interpreted as working with the WB. Outside the two abovementioned coun­
tries, the WBGT is widely preferred, instead, because it is easy to calculate and better accounts
for both environmental and physiological predictors of heat stress (Epstein and Moran, 2006).
In underground mines with air temperatures measured away from radiant heat sources, the
index can be expressed by equation 3,

where WBn = natural wet bulb (°C), which can be obtained from equation 4 (Carter,
Zaitchik, Gohlke, Wang, and Richardson, 2020),

where C = 0.85 for V < 0.03 m/s; C = 0.96 + 0.069 log10 V for 0.03 m/s ≤ V ≤ 3 m/s; C = 1
for V > 3 m/s.
The replacement of WB with WBn in the WBGTn calculation partially addresses the criti­
cism of the index being insensitive to wind speed. Exposure limits on the WBGT scale can
vary slightly across industrial hygiene organizations due to discrepancies in the cut-off range
for metabolic rates. The limiting values adopted in this study are shown in Table 1,

Table 1. OHSA-recommended WBGT exposure limits for continuous* work on acclimatized

workers (Webber, Franz, Marx, and Schutte, 2003; NIOSH, 2016).
Exposure Limit (°C)
Activity level (maximum metabolic rate, W)
V < 1.5 m/s V ≥ 1.5 m/s

Light (< 234) 30.0 32.2

Moderate (234 – 349) 27.8 30.6
Heavy (> 349) 26.1 28.9

* At least 45 minutes worked per hour

The reject WB and WBGT exposure limits complement the model input variables specified
in Table 2,

Table 2. Model input variables.

Design Criteria Units Value
th
95 percentile WB/DB °C/°C 15.1/19.7
Minimum velocity m/s 0.5
Collar WB °C 6
Geothermal gradient °C/km Considered
Rock density kg/m3 Considered
Rock specific heat kJ/(kg°C) Considered
Rock thermal conductivity W/(mK) Considered
Shaft wetness % 30 – 80
Duct friction factor kg/m3 0.0013
Leakage porosity mm2/m2 25
Duct thickness mm 3
Duct thermal conductivity W/(mK) 0.48
Activity level - Moderate

203
3 RESULTS

3.1 Baseline scenario with reject WB

Table 3 presents the simulated results of a number of key airways from the baseline scenario.
The current installed chiller capacity equates to 1.5 MWr with approximately 1 MWr needed
to cool 30 m3/s to a collar design WB of 6°C. The reduction in air velocity corresponds to the
flow area transition from the duct to the working areas of the cutting machine (i.e. airways 4
to 7, and 9). A downward trend is observed in the CPS as the air absorbs additional heat,
despite a lower or constant entry WB in some airways. This is a consequence of the propor­
tionality between the quantity of mass flow and available cooling power. The baseline refriger­
ation quantity predicts a mostly favorable reject WB profile for heat stress, and should permit
a consistently high level of productivity.
The DB increases more rapidly than the WB as a result of the overwhelming percentage of
dry sensible heat input, with predicted inlet temperatures considered unsafe from a WBGT
point of view. This is further observed by the non-compliance of the inlet WBGT along
most of the cutter, which requires additional engineering means to maintain continuous
production.

Table 3. Baseline airway properties.

Density Flow V Inlet WB/DB Outlet WB/DB Eqn (1) Inlet WBGT Outlet WBGT
Airway 3 3
kg/m m /s m/s °C/°C °C/°C kW °C °C

1 1.19 29.4 20.0 15.1/19.7 18.1/26.5 1600 16.5 20.6

2 1.21 28.2 19.2 17.9/26.5 6.0/6.0 1272 20.5 6.0
3 1.3 21.2 32.1 21.6/41.1 21.5/41.1 621 27.5 27.4
4 1.28 16.2 2.1 21.6/41.2 24.0/47.6 470 27.7 31.4
5 1.27 16.4 1.6 23.9/47.5 24.5/48.1 309 31.4 32.0
6 1.27 16.7 1.4 24.5/48.0 25.0/48.0 277 32.0 32.4
7 1.27 16.7 1.6 25.0/48.1 25.1/47.0 238 32.3 32.1
8 1.18 5.7 4.0 22.3/41.4 34.8/92.2 137 28.0 52.0
9 1.21 23.2 2.9 27.8/57.9 29.7/64.6 24 37.0 40.4

One method to refine the ventilation and cooling strategy is to compute and understand the
heat load distribution like the chart displayed in Figure 3. It can be observed that the electric
equipment and duct heat transfer constitutes more than half of the total heat load. Duct insula­
tion was proposed to reduce this component of heat, however, a review of the shaft fit-out infra­
structure rendered this option impractical. Alternatively, the duct heat transfer can be reduced
indirectly through increased flow, which enhances the overall cooling capacity, at the expense of
a step change in fan motor size. The amount of supply air increase is limited by the performance
envelope of the BAC to maintain an identical collar WB without increasing surface cooling cap­
acity. A secondary chiller, installed mid-shaft to depress the WB back to collar value, was sug­
gested as a contingency plan to implement into the shaft design. These planned measures
prompted the roadheader manufacturer to consider retrofitting the machine with a water cooler
(utilizing the shaft’s existing service and return water pipes) to absorb radiant heat that would
subsequently be pumped back to surface.
Although the strata heat comprises the smallest percentage of the overall heat, the author
surmises, from a simulation perspective, that the wetness factor may play a sensitive role in
the WBGT. For example, a wetter shaft accelerates evaporation, which causes a reduction in
the return air’s DB profile. Consequently, the smaller DB differential, between the shaft’s
return air and the chilled fresh air, dampens the effect of duct heat transfer.

204
Figure 3. Baseline heat load distribution.

3.2 WBGT-informed scenarios

Three additional scenarios informed by the WBGT exposure limit are examined in this sec­
tion. They include increased volume flowrates from a baseline of 30 m3/s to 38 m3/s (hence­
forth identified as enhanced flow), enhanced flow with a secondary chiller installed
approximately mid-shaft, and enhanced flow with water cooling on the cutter that reduces
equipment heat by 190 kW.
Figure 4 Shows the WBGT profile of the entire shaft for the various cases. The peaks in all 4
curves signify the reject temperature above the equipment. The baseline requires some distance up
the shaft for auto-decompression to reduce the WBGT below the threshold. Increasing the flow­
rate highlights the benefit of additional cooling power, which causes an initial larger temperature
reduction within the cutter compared to the rest of the shaft. This is associated with a smaller
temperature rise from the heat sources, which subsequently decreases the thermal driving poten­
tial in duct heat. It is apparent that only the option of a second chiller with more flow complies
fully with the exposure limit. Equipment water cooling produces, arguably, a similar favorable
effect, but with a higher peak WBGT. Thus, it would be important to evaluate the difference in
cost, time and ease of installation between these two solutions. Additional investigations at inter­
mediate depths may assist with lead times (taking advance rates of the cutter into account) regard­
ing either retrofits or procurement.

Figure 4. Shaft WBGT profiles of baseline and WBGT-informed scenarios.

205
3.2.1 Heat balance
The combination chart in Figure 5 illustrates the heat balance for all scenarios. The stacked
column represents the heat sources, while the CPS (interpreted as a negative value) is calculated
prior to the cutter at the shaft bottom. The refrigeration quantity is specified to reach the design
WB at the collar and mid-shaft. Ideally, the predicted refrigeration quantity should be the differ­
ence between the heat sources and CPS. However, the iterative solution behind duct heat causes
an imbalance in the calculations. Autocompression increases proportionally with mass flow, but
has an equal effect on temperature rise. Fan compression and power increases with flowrate to
the third power, and all three enhanced flow scenarios demonstrate a greater value than baseline.
An overlap between CPS and refrigeration quantity is observed in the case of the second chiller
and equipment water cooling. The larger overlap in the former, linked to nearly twice the refriger­
ation, corroborates the above evidence of WBGT exposure limit compliance. On the contrary,
conclusions based on reject WB would argue an overly cooled climate due to excessive engineer­
ing contingencies. Interestingly, the case with the mid-shaft chiller has the greatest quantity of
strata heat, attributed to a higher thermal driving potential between the lower DB and prevailing
virgin rock temperature.

Figure 5. Heat load balance chart.

4 CONCLUSIONS

A study is performed on the ventilation and refrigeration requirements of a mine undergoing

deep shaft excavation as a function of direct and empirical heat stress indices. A baseline solu­
tion informed by the direct WB index predicts an acceptable thermal environment. The advan­
tage of the WB is that it easily predicts the refrigeration requirements by subtracting the CPS
from the known heat sources. The investigation observed a large wet-bulb depression due to
primarily sensible heat driven sources, which caused unfavorably high WBGT. Therefore, the
reliance on reject WB alone in mines with a similar heat profile should be cautioned.
On the other hand, solutions driven by WBGT compliance at the moderate activity scale
indicated nearly twice the amount of required refrigeration with a concurrent increase to
supply flowrate, which required a step change in fan motor power. Thus, it is strongly recom­
mended that measurements be taken, averaged over the work-shift, and subsequently valid­
ated with the model to ensure these engineering measures are not overdesigned.
Additional efforts to understand the effects of sophisticated rational heat stress indices on
refrigeration and ventilation requirements could be the next avenue of research, with a goal to
determine an empirical method for cooling estimation that satisfy the requirements of the
WBGT index.

206
REFERENCES

Al-Neshawy, F. (2015). Estimation of the heat evolution during the hydration of concrete, mix design of hot
concrete, drying and coating of concrete structures. Slideplayer. https://slideplayer.com/slide/8020402/
Belle, B., & Biffi, M. (2018). Cooling pathways for deep Australian longwall coal mines of the future.
International Journal of Mining Science and Technology. 28 (6): 865–875.
Brake, R. & Bates, G. (2001). A valid method for comparing rational and empirical heat stress indices.
Annals of Occupational Hygiene 46 (2): 165–174.
Brake, R. & Bates, G. (2002). Limiting metabolic rate (thermal work limit) as an index of thermal stress.
Applied Occupational and Environmental Hygiene 17 (3): 176–186.
Carter, A.W., Zaitchik, B.F., Gohlke, J. M., Wang, S., & Richardson, M.B. (2020). Methods for estimat­
ing wet bulb globe temperature from remote and low-cost data: A comparative study in Central
Alabama. GeoHealth, 4: 1–16.
Epstein, Y. & Moran, D.S. (2006). Thermal comfort and the heat stress indices. Industrial Health 44:
388–398.
McPherson, M.J. (1993). Subsurface ventilation and environmental engineering. New York, NY: Springer.
NIOSH. (2016). Criteria for a recommended standard: Occupational exposure to heat and hot environments.
U.S. Department of Health & Human Services, Centers for Disease Control & Prevention, National
Institute for Occupational Safety & Health.
Roghanchi, P., Sunkpal, M., Carpenter, K., Kocsis, C. (2016). Application of heat stress indices in under­
ground mines. [Poster session]. https://doi.org/10.13140/RG.2.2.18728.16643
Webber, R.C.W., Franz, R.M., Marx, W.M. & Schutte, P.C. (2003). A review of local and international
heat stress indices, standards and limits with reference to ultra-deep mining. The Journal of The South
African Institute of Mining and Metallurgy: 313–324.

207
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Development of energy efficient and sustainable cooling strategies

for hot underground mines

J.E. Fox
PhD., Senior Consultant, Mine Ventilation Services, a business unit of SRK Consulting, USA

ABSTRACT: As relatively hot mineral deposits are being developed to considerable depth,
the issue of mitigating heat will require monumental cooling capacity in some cases. For deep
deposits to be mined at increasing depths, mechanical cooling systems used to maintain an
acceptable underground climate will most certainly become increasingly complex and expen­
sive. In most methods, the main source of potential cooling energy comes from cold water.
The cold water can come from ice (stowed or transported), in situ reservoirs, or refrigerated
water chillers. This paper investigates the use of in-situ ice as the source of cooling.
Air cooling power is available by passing intake airflow through stowed ice accumulated
during favorable atmospheric conditions. Planning for sustainable cooling potential requires
protected space to successfully maintain ice in storage until needed. Stored ice or cold water
should be judiciously cared for, safely kept from the ill effects of heat, in order to remain
useful for bulk air cooling. Maximizing the production of ice means that atmospheric condi­
tions will be monitored closely during the formation phase. Storage facilities would maintain
restricted airflow while minimizing possibilities for heat transfer during the storage phase.
Successful installations will prioritize environmental management to favor long-timespan
usage of sustainable cooling systems. Sustainable cooling systems could reduce or eliminate
the need for mechanical cooling if installed with favorable conditions.

1 SUSTAINABLE BULK AIR COOLING PROPOSAL

1.1 Background
The current market trend in profitable mining operations involves underground mining activ­
ity below a viable mineral deposit which was initially developed with an open pit. Due to the
scale of this type of operation, a popular method from the surface mining perspective is to use
very large haul trucks bringing ore up a long underground ramp system which daylights in the
open pit. This type of ramp mine portal is driven into the most stable strata surrounding the
open pit. This mining arrangement is the proposed application for bulk air cooling via stowed
ice reservoirs.
Triple gallery entries would be driven approximately parallel to the pit wall ramp portal.
This energy conscious arrangement will cool intake ramp air with natural flow through venti­
lation. Since many northern Nevada underground mines are accessed through large open pit,
the following ice accumulation gallery arrangement could work successfully in Carlin Trend
or other mining operations that experience cold winters. Triple snow storage galleries devel­
oped in bedrock adjacent to the underground ramp would use construction materials to main­
tain a sufficiently cool environment that maximizes the volume of ice preserved until the heat
of summer.

DOI: 10.1201/9781003429241-22

208
1.2 Site considerations
Construction of the underground ice reservoir will require diligence in planning and care during
excavation. Selection of the site where the ramp system is driven will be extremely important if
this type of system is to be employed. Pit wall in which the construction will be excavated needs
to be seismically sound and proven to be devoid of significant failure joints, weakness planes or
ground water draws that will decrease the geological structure rating. In other words finding the
most attractive site, rock mechanically speaking, will be of significant importance during the
planning and subsequently affect associated capital and operating costs.
Maintaining ice reservoir liner integrity during construction is one of the project’s highest
quality control priorities. Floors of the reservoir should be concrete that is substantial enough
to withstand freeze/thaw conditions. Walls of the reservoir should be sprayed with shotcrete
as the minimum. Walls and reservoir bottom should be lined with waterproof material.
A cast-in-place reinforced concrete dam will be constructed at the reservoirs’ low point. Add­
itional sumps should be driven down grade of the reservoir galleries to ensure any melt water
from the reservoirs be captured before allowing overflow into mine workings.
This type of ice generation system would preferably be excavated into the sound and
extremely stable highwall of an open pit. Preference would be given, for the construction area,
to be in close proximity to the ramp system accessing the underground below. In order to con­
tinually verify structural integrity and the host pit wall stability, the snow accumulation gal­
lery system should include monitoring wells drilled into the strata surrounding the galleries.
Additionally, groundwater monitoring and seepage drainpipes would be drilled horizontally
below the cold water bulk air cooling service water reservoir.
Detailed slow slip analyses should be conducted in the immediate vicinity of the planned
site with numerous core drill data in the immediate area where the ice accumulation galleries
are to be excavated. Thorough rock mechanics analysis needs to be performed to determine
all possible failure planes in the construction zone. Careful attention should be given during
the core logging so that localized weakness planes and failure zones can be accurately identi­
fied. This careful analysis will reveal if required structural load bearing exists to store pro­
posed ice banks for 5 months until the ice would be used for cooling.

1.3 Cooling system configuration and operations

Regulators with air locks are installed on either end of each gallery to maintain minimum vel­
ocities during snow and ice accumulation. Using compressed air and water, snow is blown,
during freezing conditions across snow fences arranged in an array perpendicular to the air­
flow. Each fence will be 5 feet high and 10 feet long with the spacing of 25 feet between fences.
Six rows of fences will be arranged in the entry gallery. Commercial high pressure snow blow­
ing nozzles used at ski resorts to broadcast artificially made snow is the type recommended
for this application.
During the beginning of the hot climate months, fresh air will be drawn across the accumu­
lated ice and snow cooling the intake air before it reaches the underground fresh air supply.
Each gallery has efficient drainage with sub-level pumping to allow water recovery. A lower
level cold-water reservoir is lined with hydrophobic geosynthetic liner to keep water from
seeping into the pit or down the ramp into the underground mine.

2 SNOWMAKING HEAT EXCHANGE PROCESS

Snowmaking is a heat exchange process. Heat is removed from snowmaking water by evapora­
tive and convective cooling and released into the surrounding environment. This heat creates
a micro-climate inside the snowmaking plume that is very different from ambient conditions.
Understanding this process can lead to practical benefits for the snowmaker. There are many
variables that affect snowmaking. Three of the most important operational factors are wet
bulb temperature, nucleation temperature and droplet size vs. nucleation reagent mixing.

209
2.1 Wet bulb temperature
The temperature of a water droplet exiting a snow gun is typically between 34 °F and 44 °F.
Once a water droplet passes the nozzle and is released into the air, its temperature falls rapidly
due to expansive and convective cooling and evaporative effects. The droplet’s temperature
will continue to fall until equilibrium is reached. This is the effective ambient wet bulb tem­
perature and it is as important as dry bulb (ambient) temperature in predicting snowmaking
success. For example, snowmaking conditions of 28 °F and 10% humidity are basically
equivalent to those encountered at 20 °F and 90% humidity.

2.2 Nucleation temperature

Once the wet bulb temperature is known, there must be a way to predict whether water drop­
lets will actually freeze at that temperature. Ice is the result of a liquid (water) becoming a solid
(ice) by an event called nucleation. A water droplet must first reach its nucleation temperature
to freeze. There are two types of nucleation, homogeneous nucleation and heterogeneous nucle­
ation. Homogeneous nucleation occurs in pure water with no contact with any other foreign
substance or surface. With homogeneous nucleation, the conversion of the liquid state to solid
state is done by either lowering temperatures or by changes in pressure. However, the primary
influence on the conversion of water to ice or ice to water is temperature.
In homogeneous nucleation, the nucleation begins when a very small volume of water mol­
ecules reaches the solid state. This small volume of water molecules reaches the solid state.
This small volume of molecules is called the embryo and becomes the basis for further growth
until all of the water is converted. The growth process is controlled by the rate of removal of
the latent heat being released. Molecules are attaching and detaching from the embryo at
roughly equal and very rapid rates. As more molecules attach to the embryo, energy is
released causing the temperature of the attached molecules to be lower than the temperature
of the unattached molecules. The growth rate continues until all the molecules are attached.
At this point, you have the solid state (ice). The nucleation event (freezing) for pure water will
take place as low as minus 40 °F. This is mostly likely to occur in laboratory experiments or
high in the upper atmosphere (upper troposphere).
The heterogeneous nucleation process is when ice forms at temperatures above minus 40 °F
due to the presence of a foreign material in the water. This foreign material acts as the embryo
and grows more rapidly than embryos of pure water. The location on which an ice embryo is
formed is called an ice-nucleating site. As with homogeneous nucleation, heterogeneous nucle­
ation is governed by two major factors: the free energy change involved in forming the
embryo and the dynamic of fluctuating embryo growth. In heterogeneous nucleation, the con­
figuration and energy of interaction at the nucleating site become the dominating influence in
the conversion of water to ice. Snowmaking involves the process of heterogeneous nucleation.
There are many materials and substances which act as nucleates, each promotes freezing at
a specific temperature or nucleation temperature. They are generally categorized as high-
temperature (i.e., silver iodide, dry ice and ice nucleating proteins) or low-temperature (i.e.,
calcium, magnesium, dust and silt) nucleates. It is the low temperature nucleators that are
found in large numbers in untreated snowmaking water. The nucleation temperature of snow­
making water is between 15 °F and 20 °F.
But wait a minute! Why do you hear freezing warnings at temperatures around 32 °F? The
answer is that another factor is coming into play with the freezing process. There is an energy
interaction between the ice-nucleating site in the water with the surface. This causes the water
droplets to freeze very near to 32 °F. In snowmaking it is the nucleator having the highest
nucleation temperature that determines when a water droplet will freeze. As a water droplet
cools, heat energy is released into the atmosphere at a rate of one calorie per gram of water.
As it freezes into an ice crystal, the water droplet will release additional energy at a rate of 80
calories per gram of water. This quick release of energy raises the water droplet temperature
to 32 °F, where it will remain while freezing continues. This is one reason why we are accus­
tomed to thinking that water freezes at 32 °F. To be precise, the water will continue to freeze

210
as long as it remains at or below 32 °F, but only after it has first cooled to its nucleation tem­
perature. Any excess energy will be dissipated into the atmosphere.

2.3 Droplet size vs. Nucleation reagent mixing

Since the distribution of various nucleators in a given volume of water is totally random, the size of
the water droplet or the number of high temperature nucleators has a significant effect on the tem­
perature at which freezing occurs (nucleation temperature). In natural water, as the size of the
water droplet decreases, the likelihood that the droplet will contain a high temperature nucleator
also decreases. Conversely, larger water droplets stand a better chance of containing high tempera­
ture nucleators. The optimum situation for snowmakers is one where each and every droplet of
water passing through the snow gun nozzle contains at least one high-temperature nucleator and
where each droplet freezes in the plume (SMI Snowmakers, 2012). It is assumed that the lower res­
ervoirs dedicated to the snowmaking process have sufficient mixing with high temperature nuclea­
tors reagents prior to the reservoir water being pumped into the snowmaking nozzles.
The relationship between the variables of nucleation temperature and droplet size is summar­
ized in two statistically valid conclusions. Firstly, a 50% increase in the droplet size results in
a 1 °F increase in nucleation temperature (SNOWatHome, 2018). Secondly, a 50% decrease in
droplet size results in a 3 °F decrease in nucleation temperature (SNOWatHome, 2018). These
conclusions are based on an average droplet size can be counter-productive to promoting high-
temperature nucleation, unless enough high-temperature nucleators are present to compensate.
Looking at the relationship between droplet size and evaporation, research in cloud seeding
shows that:
• These conclusions further point out the undesirable results from using very small droplets,
especially in areas where water loss is a critical issue;
• Relating droplets size to nucleation temperature, it is possible to increase snowmaking pro­
duction and efficiency by using high-temperature nucleators and larger water droplets;
• This frequently not only allows increased water flow, but also reduces evaporation and
yields more snow on the ground;
• In fact, studies indicate that a 20% increase in water flow can increase snow volume up to
40% if droplet size and nucleation temperature is optimized (SMI Snowmakers, 2012).
Resultant droplet size can be reliably achieved through judicious control of feed water pres­
sure and feed air pressure. There is no way to ensure that each droplet will contain at least one
high temperature nuclear agent reagent, other than ensuring the mixture is well agitated. In
order to ensure adequate mixing of the snowmaking feedwater sump, several centrifugal
pumps should be employed throughout the water management process supporting the snow­
making process. In the best case scenario one pump will be used when any reagent is added to
the reservoir mix. Additionally, one pump should be used to aerate and another pump support
agitation of the water mixture being supplied to the snowmaking nozzles.

3 MODELING COOLING CAPACITY FROM ICE ACCUMULATION RESERVOIRS

A thermodynamic ventilation model is developed in Ventsim© to estimate the cooling potential

capacity of the proposed ice accumulation reservoirs. Ventsim© was never designed to simulate
the physical phenomena of snowmaking or ice melting. Prudent modeling will take into account
the energies created, i.e. stored, or used using simplistic energy balance equations. Ventsim© will,
however, account for heat sources and sinks added to a mine ventilation network. Seasonal vari­
ations of the intake air temperatures are estimated using conditions similar to those experienced
at open pit operations on the Carlin trend. The simulated seasonal low and high wet bulb and
dry bulb temperatures are input to the Ventsim© thermal flywheel effect solution module.
In order to simulate the snow making process, the latent heat of fusion as a heating source
needs to be estimated. Each snowmaking nozzle array will therefore be assigned a small heat­
ing source. In order to simulate the potential cooling power from the ice melting process, the

211
latent heat of melting must be applied to the ventilation model as heat sinks in each ice accumu­
lation gallery as the respective ice stopes melt when exposed to the warm fresh air supply.

3.1 Modeling latent heat of snowmaking

In order to adequately model this snowmaking process, as developed by an individual snow
blowing nozzle, the latent heat of fusion must be quantified. The latent heat of fusion is the
heat generated when water goes from the liquid to the solid phase. Ventsim© has no way of
automatically understanding the snow making process, thermodynamically speaking. There­
fore, the small heat source for each snow blowing nozzle array is added to each ice accumula­
tion gallery during the snow making process. The latent heat of fusion for snowmaking is
assumed (disregarding the heat release of water cooling to the freezing point) at a rate of
approximately 80 cal/g of H2O, which can be calculated at 2465.56 kJ/kg. Given the optimum
snowmaking conditions described above, Mollier’s Chart (Figure 1) estimates humidity ratio
difference (kg H2O/kg dry air) of 0.0025. Assume the density of intake airflow is 1.186 kg/m3
moving past the snowmaking nozzels with a flowrate of 12.9 m3/s.

Figure 1. Mollier’s Chart (https://www.mollier-diagram.com/, 2023).

Latent heat due to the moisture in air can be calculated with the following equation:

where hl = latent heat (kW); Q = volumetric airflow (m3/s); ρ = air density (kg/m3); and hwe =
latent heat of vaporization of water (kJ/kg); and dwkg = humidity ratio difference (kg H2O/kg
dry air).

212
 Therefore, the assumed latent heat of fusion source added for each active snowmaking
nozzle is 94.3 kW (94.3 = (12.9)(1.186)(2465.56)(0.0025)).

3.2 Modeling latent heat of melting ice

In order to adequately model the ice melting process, as delivered to the ventilation network, the
latent heat of melting must be quantified. The mass of ice curb each gallery can easily be calcu­
lated using the volumetric relationship yielding the volume by solving for the product of the
length (15.2 m) times the height (2.4 m) times the width (4.0 m) which has been reduced by
1.5 m on each side to allow safe passage by system audit personnel and maintenance crews. The
resultant curb volume of 146 m3 for each ice curb is then multiplied by 15 ice curbs per gallery,
yields 2,189 m3 of ice per gallery. Since ice has a specific weight of 1000 kg/m3, each ice curb has
an estimated mass of 2,188,800 kg of ice initially before ice melting begins, theoretically speaking.
Calculation of the potential cooling energy provided in each ice accumulation gallery is
based upon the mass of ice in the gallery multiplied by the latent heat of melting. The latent
heat of melting is 334 kilojoules (kJ) per kilogram (kg), the resultant cooling power available
is therefore 334 kJ/kg times the estimated mass of ice which is 2,188,800 kg, would potentially
generate 731.06 GW of cooling power. This is not a reasonable assumption. A more realistic
and practical assumption would assume the outer inch, or 2.5 centimeters, of ice which is
intimately involved with the cooling process due to constant exposure to the ventilating air­
flow, at any given time-step during the modeling exercise.
Two sides on the length and width of each ice curb, plus the ice curb top is exposed to the
warm ventilating airflow during the cooling phase of this proposed mine air conditioning system.
Assuming a one inch thickness for the melting ice interface with the ventilating air flow, 3.8 m3
of ice can be allocated for cooling recruitment by each ice curb in every ice accumulation gallery.
This means a total of 15 ice curbs times 3.8 m3 of ice per accumulation gallery yields 57 m3 of ice
contributing to the cooling system per gallery. This ultimately yields 19.2 kW of cooling power
per gallery when the ice curbs are exposed to the warm, fresh air to which the cooling power, or
coolth, is being applied. This ultimately yields 855 kW of cooling power applied to the ventilation
network on a continual basis, once employed until the ice supply is below 57 m3.

3.3 Model development

The theoretical mine is described in the following modeling example with physical dimensions
and layout as previously described in the text. The model is drawn in AutoCAD© to define
the layout of the 3D network of airway branches and, as far as the calculation of natural ven­
tilation pressure is concerned, elevation coordinates for each airway branch junction. The
model layout is then exported in the DXF format to facilitate import and generation of
a blank ventilation model in Ventsim©.
Preliminary ventilation controls are defined to assign the airflow regime for each galley, assessing
pressure contributions along each network air pathway. Regulators are given a set of resistance
positions that will generate resistances of varying intensities. These resistances are assigned to regu­
lators for the respective ice accumulation gallery, depending on the life cycle of stored ice. Connec­
tion to the intake air side of the ice accumulation galleries at the main ramp is given a different
type of control, specifically a well-constructed vehicle airlock with integrated regulator. Below each
gallery’s dam is positioned a well-constructed vehicle door, also with an integrated regulator.

3.4 Model assessment

Firstly, each airflow regime is described for each of the three initial snow deposition sequences.
Figure 2 shows model airflow designations for each gallery and the main ramp before snowmaking
begins. Airflow temperatures are then described for three distinct snowmaking phases:
1) beginning snowmaking process;
2) midway through snowmaking, resistance responses;
3) snow filled gallery during initial ice conditioning phase.

213
 Figures 3 through 5, describe the temperature characteristics for key snow formation steps.

Figure 2. Ventsim© model of ice accumulation galleries with no snowmaking or bulk air cooling; air­
flow quantities shown in m3/s.

Figure 3. Ventsim© model of ice accumulation Gallery 1 when snowmaking begins; wet bulb/dry bulb
temperatures shown in °C.

Figure 4. Ventsim© model of ice accumulation gallery 2 during initial snowmaking phase; gallery 1 in
initial conditioning phase; wet bulb/dry bulb temperatures shown in °C.

214
Figure 5. Ventsim© model of ice accumulation galleries when snow has fully accumulated in gallery 3;
gallery 1 and 2 in initial conditioning phase; wet bulb/dry bulb temperatures shown in °C.

3.5 Using the stored ice to cool mine intake air

When seasonally warm intake air of nearly 60 m3/s is passed over the three ice accumulation
galleries, the wet bulb and dry bulb temperatures in the ramp are depressed by 2 °C and 4 °C,
respectively, as shown in Figure 6. Almost 20 m3/s with gallery airflow average velocity of
0.6 m/s will pass through each gallery during the cooling phase, as specified by each galleries’
regulator. Individually, each gallery injects airflow below freezing onto the ramp where
mixing occurs in the turbulent ramp air.

Figure 6. Ventsim© model of ice accumulation galleries using stowed ice for bulk air cooling; wet bulb/
dry bulb temperatures shown in °C.

4 CONCLUSIONS AND FUTURE WORK

This innovative air cooling system will help underground mines reduce the operational carbon
footprint by replacing electrical power with renewable cooling energy. Many factors contrib­
ute to the core fundamentals of sustainability. Cultural responsibility to sustainability and
maintaining the environment in a safe and healthy condition lends credibility to mining oper­
ations with extensive mineral reserves that could potentially lead to strategic economic viabil­
ity. Positioning a mine operation with lower capital expenditure and lower operating expenses
will provide long-term economic sustainability through uncertain economic times. Innovative
use of the environment and responsible care of natural resources available at the mine site will

215
inevitably lead to a sustainable mine. Sustainable cooling systems could reduce or eliminate
the need for mechanical cooling if installed with favorable conditions.
Successful installations will prioritize environmental management to favor long time span
usage of sustainable cooling systems.The ice accumulation reservoirs supplying 855 kW of cool­
ing power continuously over 5000+ hours would yield over 4.2 GWh. According to the United
States Environmental Protection Agency (EPA, 2021), 1 kWh is equivalent to 7.09e-4 metric tons
CO2 gas. Conservatively, this renewable cooling system could offset 3,031 tonnes CO2 annually.
Lowering the intake airflow requirement down the ramp will significantly improve the per­
formance of the ice accumulation galleries’ contribution into the overall capacity of the mine
air conditioning system. Sensitivity analyses should be conducted to determine optimum air­
flow regimes for given atmospheric conditions of relative humidity, temperature and baromet­
ric pressure. Additional analyses should be conducted to determine optimum gallery length,
cross sectional area, grade, orientation, trajectory and total number of galleries simulated for
this sustainable mine air conditioning system.
Investigation of practical, operable air cooling systems for underground mining should
include maximum transfer distance, both laterally and vertically, for air cooled from the ice
accumulation galleries. Development of renewable energy air cooling systems for hot under­
ground mines will ultimately benefit any operation that can take advantage of this seasonably
available energy resource. Correct sizing of the air cooling system will determine the success of
this type of air conditioning system. Supplying ice harvested near the surface to underground
air cooling systems is a potential source of immense savings for mines that would require bulk
cooling of the intake airflow of the identified mining scenario.

4.1 Future work

Ultimately, additional model simulations would need to be conducted. Careful estimation of
the thermal flywheel effect for the given location will need to be considered during the course
of these simulations. Snowmaking sensitivity analyses should also be conducted to determine
optimal nozzle types, flow rates and temperature domains for all possible snowmaking condi­
tions. Additional analyses should be conducted to determine optimum underground psychro­
metric conditions including wet bulb temperature and airflow velocity. Intrinsically safe
nucleation agents should also be evaluated by ongoing testing of spray water for total dissolved
solids. Careful chemical evaluation of the interaction between the resident dust in the water
and nucleation agents should also be considered when modeling the snow making process.
The system will use wastewater from the mine and recycle water during operation as the ice
melts into storage sumps. Snow water quality will need to be monitored and maintained.
Nucleation reagents allow for warmer snowmaking conditions that can extend seasonal use­
fulness especially at higher altitudes. Nucleation agents will need to be added to the reservoir
water and sufficiently mixed before spraying application.
Additional geotechnical measures will be necessary to ensure pit portal integrity due to the
addition of the stowed ice reservoirs. In order to maintain the highest level of safety, both sur­
face and underground, continual monitoring of ground conditions will be maintained during
the operation and maintenance of the renewable cooling ice storage facility.

ABOUT THE AUTHOR

Dr. Jon Fox received his BS in Mining Engineering from Missouri School of Mines in 1987.
He was awarded degrees of MS in Mining Engineering and PhD in Mineral Resources Engin­
eering from University of Nevada, Reno in 2012 and 2021 respectively. Jon returned to uni­
versity after 25 years, beginning graduate studies in 2010 and the award of Mackay MS
student of the year was awarded Jon in 2012. Dr. Charles Kocsis acted as Jon’s doctoral
advisor since 2017. Jon has over 32 years industry experience in civil, electrical and mining
engineering and contracting services. Jon consults with Mine Ventilation Services of SRK as

216
a mine ventilation engineer, where dedication to the health and safety of the underground
workforce is his highest priority. Jon’s research topics on innovative renewable energy air
cooling systems are timely and well thought out, thanks to all the input received over the
years from fellow professionals in the mining industry.

REFERENCES

EPA (2021). United States Environmental Protection Agency, www.epa.gov/energy/greenhouse-gases-

equivalencies-calculator-calculations-and-references, accessed 4-22-2021.
McPherson, M.J. (2009). Subsurface Ventilation Engineering. Mine Ventilation Services, Inc., 3rd. ed.,
Fresno, CA.
Mollier-Diagram.com (2023), www.mollier-diagram.com/, accessed 03- 30-2023.
SNOWatHome (2018), www.snowathome.com/snowmaking_science.php, accessed 01- 20-2018.
SMI Smowmakers (2012), www.snowmakers.com/snowmaking-advanced.html, accessed 10-2-2020.

217
Mine dust monitoring and control
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Effect of auxiliary scrubbers on respirable coal mine dust particle

size and composition

F. Animah, A. Greth, C. Keles & E. Sarver

Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA

ABSTRACT: Control of dust in underground coal mines is critical for mitigating both
explosibility and respiratory hazards. For decades, the National Institute of Occupational
Safety and Health (NIOSH) has led research to evaluate the effectiveness of various dust con­
trol technologies in coal mines, including several recent studies related to auxiliary scrubbers.
While these have demonstrated scrubber efficiency to reduce respirable coal mine dust mass
concentrations, detailed dust characterization was not a focus. The current study aims to
expand on NIOSH’s prior work by evaluating the effect of two different scrubbers (one wet
and one dry) on respirable dust composition and particle size distribution. For this, preserved
filter samples were obtained from NIOSH and analyzed by multiple methods. Results indicate
that dust composition was not appreciably affected by either scrubber. However, as expected,
the wet scrubber appeared to decrease the overall respirable particle size distribution.
Although such an effect of the dry scrubber was not consistently observed, this can be attrib­
uted to the fact that the sample collected downstream of the face area cleaned by the dry
scrubber contained mixed dust from the scrubber exhaust and other parts of the mine.

1 INTRODUCTION

Dust in coal mines presents both safety and health hazards. Control of airborne float dust is
a critical strategy for mitigating explosibility hazards (Shahan et al., 2017), and control of res­
pirable coal mine dust is crucial for preventing occupational lung disease (Blackley et al., 2018;
Hall et al., 2019). Indeed, dust control has been a key focus of the mining research portfolio at
the National Institute of Occupational Safety and Health (NIOSH)—including work on scrub­
bers (Colinet et al., 2013; Colinet & Jankowski, 2000; Janisko et al., 2015; Organiscak et al.,
2016; Organiscak & Beck, 2010; Patts et al., 2016). Scrubbers are commonly used on continu­
ous miner machines (CM) to capture dust being generated at the cutting face, and thus prevent
its transport into the return airways and contamination of the ventilating air more broadly
(NIOSH, 2021). CM-mounted scrubbers, referred to as flooded-bed scrubbers, essentially cap­
ture particles by forcing the dust-laden air through a wet filter where they either deposit or
attach to water droplets that are removed by a demister (NIOSH, 2021; Patts et al., 2016).
To improve dust control in other priority areas of underground coal mines (i.e., not neces­
sarily coupled with the CM operation), auxiliary (stand-alone) scrubbers could be another
option. Recently, NIOSH has conducted field studies of a wet auxiliary scrubber (i.e., based
on the traditional flooded-bed type) (Janisko et al., 2015; Patts et al., 2016), as well as a dry
unit that can provide even more flexibility in terms of placement in the mine (Organiscak
et al., 2016). Since the primary objective of these studies was to evaluate the overall dust col­
lection efficiency of each scrubber, detailed dust characterization was not a focus. However,
NIOSH did preserve the dust samples, which now enable further analysis. In the current
work, those samples were used to investigate the effect of the aforementioned wet and dry
auxiliary scrubbers on dust composition and particle size.

DOI: 10.1201/9781003429241-23

221
2 SUMMARY OF NIOSH STUDIES ON AUXILIARY SCRUBBERS

The sections below overview the relevant NIOSH studies and available dust samples that are
the subject of the current work.

2.1 Wet scrubber study

As reported by Janisko et al., (2015) and Patts et al., (2016), NIOSH tested the performance
of an inline wet scrubber (Compact Filter Technic type HCN 600/1 model) to reduce airborne
float dust and respirable dust concentrations downstream of a CM. The CM was operating on
one side of a four-entry super section for longwall development at the study mine (called
“Mine A” from here). An auxiliary fan was used to pull air through ventilation tubing from
the CM to the scrubber, which was located in the return entry (Figure 1a). The airflow
through the scrubber was about 17,000 cfm (8 m3/s), while the overall airflow in the return
was 61,000 cfm (28.8m3/s). To avoid interference of rock dusting with the scrubber evaluation,
rock dust application was halted in the return during sample collection; however, Janisko
et al., (2015) noted that rock dust that had already been applied was re-entrained into the air
by the auxiliary fan which operated along with the scrubber.
Dust samples were collected during each of six separate CM cuts, which were made in different
entries over a total of three shifts. The production rate was consistent across all cuts (about 12.8
tons per meter of advance) and the cut time varied between about 60-80 minutes, with sampling for
the full duration of each cut. On the first shift (Cuts 1-2), the wet scrubber was not in place. On the
later shifts (Cuts 3-6), the wet scrubber was in place and operating; NIOSH noted scrubber power
interruption on Cut 3 and so associated data was excluded from their study (Patts et al., 2016). Sam­
pling was done at the intake and in locations that were spaced 200, 300 and 400 feet downstream of
the wet scrubber location in the return entry. (Only samples from the 200- and 400-feet locations
were included in the current work per Figure 1a). In each location, respirable dust samples were
collected using gravimetric samplers consisting of a standard air pump at 2.0 L/min and 10-mm
nylon cyclone (i.e., D50 of about 4 µm). A personal DataRam (pDR 1000s) sampler was also used
alongside the gravimetric samplers to measure dust concentrations. (Total airborne float dust was
also collected in each location, though these samples were destructed for NIOSH’s low-temperature
ashing technique, to separate out the coal fraction of the sample). While not described in the
NIOSH reports (Janisko et al., 2015; Patts et al., 2016), additional dust samples were collected in
the 200 and 400 feet locations using custom cyclones (Mesa Laboratories Inc., Butler, NJ) designed
for D50 of about 10, 20, and 40 µm, respectively, when used with the same pumps and flow rate as
for respirable sampling. The respirable and 10, 20 and 40 µm samples were collected directly onto
pre-weighed 37-mm polyvinyl chloride filters (5 µm pore size) housed in standard cassettes.
Based on the comparison of dust concentrations (mg/m3) during the two CM cuts without
the wet scrubber versus the three cuts with it operating, results indicated the scrubber yielded
an 86% reduction in respirable dust and more than 92% reduction in float dust in Mine
A (Janisko et al., 2015; Patts et al., 2016).

2.2 Dry scrubber study

Organiscak et al., (2016) tested the performance of a dry scrubber to reduce respirable dust con­
centration downstream of a CM and/or roof bolter (RB) in a different mine (called “Mine B”
from here). Testing was conducted on three separate shifts in each of two super sections in Mine
B (called “Section 1” and “Section 2” from here), yielding a total of six sets of test data. The
scrubber is a self-propelled unit manufactured by J.H. Fletcher & Co of Huntington, WV, which
uses a vane axial fan (with a capacity of 480 V and power of 22.4 kW) to pull air through an on-
board filter (with a 99% efficiency rating for 2µm particles). It was originally designed to protect
the RB operator when working downwind of the CM. However, to simplify the field tests (i.e.,
not move the scrubber to follow CM and RB movements), Organiscak et al., (2016) placed the
scrubber in the last open crosscut of the section being evaluated and allowed it to operate con­
tinually (except for when the CM was mining in the return entry). The scrubber airflow quantities

222
ranged from 2720-4900 cfm (1.28 - 2.31 m3/s), while the initial airflow supplied to the face areas
(i.e., without the scrubber operating) was 1660-4890 cfm (0.784 - 2.31 m3/s). Organiscak et al.,
(2016) reported that during the tests in Section 1 the scrubber intermittently turned off and had
to be restarted. (Notably, they also collected noise measurements and concluded that, under the
conditions tested in Mine B, workers should not be overexposed to noise. However, under higher
airflows and as the scrubber filter becomes loaded with dust, Organiscak et al., (2016) noted that
noise over-exposure could be possible.)
During each test, Organiscak et al., (2016) sampled in three stationary locations as shown in
Figure 1b: just upstream of the scrubber, downstream of the face area cleaned by the scrubber
(i.e., the face in the return entry), and further downstream in the return. Samples were also col­
lected in the RB intake and on the RB, though these are not included in the current work. The
sampling duration for each test was between about 340-400 minutes. In each location, respirable
dust samples were collected using the same equipment, flowrate, filters, and cassettes as used in
the wet scrubber study for respirable dust, and a real-time monitor was also used (pDR-1000,
Thermo Scientific, Waltham, MA). The pDR data had to be calibrated (i.e., using the filter sam­
ples) to estimate respirable dust concentration, which could then be correlated back to time
periods of interest (e.g., to look at specific scenarios of RB and CM placement relative to the
scrubber). A pDR was also used over short periods to monitor right at the scrubber exhaust (i.e.,
before mixing with other air could occur in the face area that should be cleaned).
Comparing pDR-derived respirable dust concentrations just upstream of the dry scrubber
versus at its exhaust indicated an actual collection efficiency of more than 90%. That said,
across all tests in Mine B, the respirable concentration was only reduced by an average of
about 50% in the face area cleaned by the scrubber during times when the CM and/or RB
were operating upstream (Organiscak et al., 2016). The discrepancy was attributed to mixing
of the clean scrubber exhaust with more dusty air in face area being sampled.

Figure 1. Schematics of sampling locations in NIOSH studies of a) an auxiliary wet scrubber in Mine
A (modified from(Janisko et al., 2015)) and b) an auxiliary dry scrubber in Mine B (modified from (Organis­
cak et al., 2016)). Only samples from the locations denoted by green circles were included in the current work.

3 MATERIALS AND METHODS

3.1 Dust sample handling

A total of 34 dust samples from NIOSH’s wet scrubber study were analyzed in the current work,
including all available respirable and 10, 20 and 40 µm samples collected in the 200- and 400-feet
locations (Table 1). A total of 18 samples from the dry scrubber study were analyzed, all of which
were respirable samples from the three stationary locations sampled on each of six shifts (Table 2).
The dust samples were obtained directly from NIOSH in their original sampling cassettes, in which
they had also been stored since collection. Upon receipt, each filter was carefully removed from its
cassette and a stainless-steel trephine was used to cut a 9-mm subsection, which was placed in

223
a clean test tube; these filter pieces from respirable samples were later prepared for dust particle size
analysis by scanning electron microscope with energy dispersive X-ray (SEM-EDX). The rest of
each filter was placed into another clean test tube and prepared for thermogravimetric analysis
(TGA) to estimate the mass fractions of coal, non-carbonate minerals, and carbonates, to loosely
approximate the major sources of dust (i.e., coal and rock strata being mined, and rock dust prod­
ucts being applied, respectively) (Agioutanti et al., 2020; Jaramillo et al., 2022).

3.2 TGA analysis

For all of the filter samples included in the current work, TGA was conducted following the method
described by Jaramillo et al., (2022). Briefly, to each test tube with the available portion of the filter
for this analysis, enough isopropanol (IPA) was added to completely submerge the filter (about
5-10 mL). The tube was capped and sonicated for approximately three minutes to dislodge the dust,
and then the filter was carefully removed. The dust suspension was centrifuged for ten minutes
(2500 rpm) to settle the particles, and then a clean pipette was used to transfer the recovered dust to
a clean, tared TGA pan. After the IPA had completely evaporated, recovered dust was analyzed by
the same TGA instrument used by Jaramillo et al., (2022) and Agioutanti et al., (2020) (Q500, TA
Instruments, New Castle, DE) and using the same thermal routine. The resulting thermogram was
used to estimate the mass fractions of coal, carbonates and non-carbonates following the approach
of Agioutanti et al., (2020), which utilizes a series of mass balance equations to analyze the sample
weight loss in specific temperature regions. Notably, Agioutanti et al., (2020) derived their equations
using lab-generated respirable dust samples collected on polycarbonate (PC) filters (i.e., rather than
PVC, as were available for the current work). Thus, prior to analysis of the mine dust samples in the
current work, a series of lab-generated samples on PVC were used to modify the mass balance equa­
tions—again following the approach of Agioutanti et al., (2020).

3.3 SEM-EDX analysis

Samples selected for particle size analysis were prepared as follows: The 9-mm filter subsec­
tions mentioned above were sonicated in IPA to recover the dust. Then, either a vacuum filtra­
tion unit or a syringe was used to redeposit the particles from the IPA suspension onto a clean
PC filter (track etched, 0.4 µm pore size); the smooth, uniform PC filter is ideal for SEM-EDX
work on respirable-sized particles. Following redeposition, a 9-mm subsection of the PC filter
was cut, mounted on an aluminum stub, and sputter-coated with Au/Pd to render it conduct­
ive. SEM-EDX analysis was conducted using a FEI Quanta 600 FEG environmental SEM
(Hillsboro, OR, USA) equipped with a backscatter electron detector (BSD) and a Bruker
Quantax 400 EDX spectroscope (Ewing, NJ, USA). Bruker’s Esprit software (Version 1.9.4)
was used to run a computer-controlled routine described by Sarver et al., (2021) to identify
and size about 500 particles per sample, with diameters in the range of 1-10µm.

4 RESULTS AND DISCUSSION

4.1 Wet scrubber

Table 1 summarizes the gravimetric and TGA results for the dust samples from Mine A. The rela­
tively small recovered-dust masses for Cuts 4-6 are consistent with the relatively low dust concen­
tration reported in the section return when the scrubber was being operated upstream (Janisko
et al., 2015; Patts et al., 2016). Moreover, the trends in recovered dust mass with cyclone cut size
are consistent with expectations (i.e., more dust was recovered from samples with increasing D50).
With respect to dust composition, the TGA results do not indicate any consistent trends
related to the scrubber operation per se. However, the TGA results do show variability
between different CM cuts. Indeed, the coal and non-carbonate contents, which are sourced
primarily from the coal and rock strata respectively at the production face, appeared to vary
from cut to cut. This can be attributed to variability in the relative heights of coal and rock
mined at the face during each cut, and the coal to non carbonates ratio shown in Table 1 can

224
be used as a crude comparison of these relative heights. For example, Cuts 1, 2 and 6 appear
to show a high coal to non-carbonate content ratio, suggesting more of the mining height was
in coal than in rock during these cuts (as compared to Cuts 4 and 5).
Likewise, the TGA data for the carbonates content, which indicates rock dusting activities
in the mine, show differences between cuts. For instance, Cuts 1, 4 and 5 show more carbon­
ates than Cuts 2 and 6, as seen in Table 1. While rock dusting was halted during the dust sam­
pling by NIOSH, Janisko et al., (2015) did report that operation of the auxiliary fan with the
scrubber appeared to re-entrain rock dust that had been applied earlier. This could explain the
relatively high carbonates content during Cuts 4 and 5, though high carbonates during Cut 1
(when the auxiliary fan and scrubber were not in place) shows that significant rock dust was
also in the -40 µm dust fractions during periods without the fan. This is consistent with obser­
vations of CM section return samples in other mines (Sarver et al., 2019).

Table 1. Summary of dust samples from NIOSH’s wet scrubber study in Mine A that were selected for
analysis here. (NCB = non-carbonate minerals; CB = carbonates).
TGA analysis

Avg. Grav. Mass %

Cut (Scrubber Sample Conc. Recovered Coal:
Sample status) location Cyclone (mg/m3)+ dust mass (mg) Coal NCB CB NCB

1* Cut 1 (NO) Resp 2.6 0.419 71% 13% 16% 5.5

2 10 μm 7.7 1.256 60% 6% 34% 10.0
3 200 ft 20 μm 11.8 1.921 61% 6% 33% 10.2
4 40 μm 14.4 2.353 58% 5% 38% 11.6
5* Resp 3.1 0.439 71% 11% 18% 6.5
6 10 μm 7.8 1.112 63% 6% 31% 10.5
7 400 ft 20 μm 11.9 1.690 71% 8% 20% 8.9
8 40 μm 12.8 1.815 63% 7% 30% 9.0
9* Cut 2 (NO) Resp 3.1 0.360 79% 11% 10% 7.2
10 200 ft 20 μm 11.9 1.368 74% 10% 16% 7.4
11 40 μm 13.8 1.591 74% 9% 17% 8.2
12* Resp 3.7 0.348 84% 8% 8% 10.5
13 10 μm 6.8 0.638 78% 13% 9% 6.0
14 400 ft 20 μm 12.2 1.114 78% 11% 11% 7.1
15 40 μm 13.8 1.219 77% 11% 12% 7.0
16 Cut 4 (YES) 10 μm 1.2 0.150 72% 14% 14% 5.1
17 200 ft 20 μm 2.3 0.272 63% 14% 22% 4.5
18 40 μm 2.7 0.329 48% 5% 47% 9.6
19* Resp 0.4 0.043 80% 16% 4% 5.0
20 10 μm 1.4 0.141 68% 13% 19% 5.2
21 400 ft 20 μm 2.1 0.209 57% 10% 33% 5.7
22 40 μm 2.2 0.228 60% 7% 33% 8.6
23* Cut 5 (YES) Resp 0.6 0.950 63% 19% 19% 3.3
24 10 μm 1.1 0.181 68% 17% 16% 4.0
25 200 ft 20 μm 1.7 0.292 46% 14% 40% 3.3
26 40 μm 2.1 0.355 36% 10% 54% 3.6
27* Resp 0.5 0.770 61% 16% 23% 3.8
28 400 ft 10 μm 1.1 0.167 65% 12% 23% 5.4
29 40 μm 1.8 0.281 39% 3% 58% 13.0
30* Cut 6 (YES) Resp 0.5 0.057 85% 7% 8% 12.1
31 200 ft 20 μm 1.8 0.211 67% 11% 22% 6.1
32 40 μm 2.0 0.237 80% 9% 11% 8.9
33* Resp 0.6 0.064 82% 10% 8% 8.2
34 400 ft 40 μm 1.8 0.840 73% 11% 16% 6.6

* Sample analyzed by SEM-EDX to determine particle size distribution

+ data was obtained from NIOSH with the filter samples, but it was not included in their published
reports (Janisko et al., 2015; Patts et al., 2016)

225
 In general, the TGA results also show that dust sourced from the rock dust application (i.e.,
carbonates) tends to be more concentrated in the larger size fractions, whereas the mine
strata-sourced dust (i.e., coal+non-carbonates) is more concentrated in the finer (including
respirable) fractions. However, no discernable trends could be observed in dust composition
moving from the 200 to 400 feet sampling locations.
With respect to particle size, Figure 2 shows the cumulative size distributions for the respir­
able samples at the 200- and 400-feet locations for all five CM cuts. The distributions were
clearly coarser when the scrubber was not in use and finer when it was. This is consistent with
expectations as the scrubber should be more efficient in removing coarser particle sizes
(NIOSH, 2021). Size distributions did not generally change much between the 200- and 400-
feet locations, which is not surprising since respirable-sized particles should remain entrained
in the mine air over long distances. On the other hand, differences in size distributions between
the 200- and 400-feet locations might be expected for larger size fractions (e.g., 20 or 40 µm
samples), though these were not analyzed in the current work. It is well established that float
dust (containing the larger particle sizes) deposition varies with distance from the mine face
(Courtney et al., 1986; Janisko et al., 2015; Patts et al., 2016; Shahan et al., 2017).

Figure 2. Overall particle size distribution for respirable dust collected in the 200 and 400 ft locations
during all five CM cuts reported in NIOSH’s wet scrubber study.

4.2 Dry scrubber

Table 2 shows the gravimetric and TGA results for the respirable dust samples from Sections 1
and 2 in Mine B. The relative dust concentrations were comparable during all shifts, and the
difference between the upstream (UDS) and downstream (DDS) sampling locations illustrate
the effect of the dry scrubber to improve the air quality in the face area it was supposed to
clean. The dust concentrations in the return (R) location appeared similar to those in the
UDS locations, which illustrates that the cleaning effect of the scrubber was limited to the
DDS face area. As explained by Organiscak et al., (2016), the scrubber has a high dust collec­
tion efficiency, but it only treats a portion of the mine air; in Mine B, the scrubber exhaust
was gradually mixed with other air in the return entry.
The TGA results show that dust composition was generally consistent between all samples
collected during a given shift, including the UDS and DDS samples—meaning no effect of the
dry scrubber could be observed on dust composition. Moreover, the coal to non-carbonates
ratio was relatively consistent across shifts indicating similar ratios in the coal to rock strata
height being mined. Carbonate content in the three sampling locations evaluated here was
relatively low during all shifts in Section 2. In Section 1, dust collected during Shifts 2 and 3
had relatively high carbonates, suggesting significant contribution of rock dust products to
the respirable dust fraction. This is probably due to active dust during or just before these
shifts.
Figures 3a and 3b show the cumulative size distributions for the UDS and DDS location
samples from Section 1 and Section 2, respectively. In Section 2, dust in the DDS location was

226
found to be slightly finer than dust in the UDS location. This is consistent with expectations
(i.e., the scrubber should more efficient at removing coarser particles), and it must be reiter­
ated that the DDS samples do not perfectly represent the dry scrubber exhaust, but rather the
face area that was supposed to be cleaned by the scrubber. Thus, DDS samples represent
a mix of the scrubber exhaust and other mine air in the return entry face area—and the true
particle size reduction yielded by the dry scrubber is likely muted in the DDS samples. On the
other hand, in Section 1, it appears that the dust size distribution was actually coarser in the
DDS location as compared to the UDS location (Figure 3a), which is seemingly inconsistent
with expectations. In this section, it is possible that any size reduction yielded by the scrubber
was completely overwhelmed by coarser particles in the return air that mixed with the scrub­
ber exhaust in the DDS location. Moreover, Organiscak et al., (2016) noted an issue with the
scrubber intermittently turning off and needing to be restarted during the testing in Section 1.
It is also possible that these startup events dislodged coarse particles from the scrubber filter,
which might have affected the size distribution results in the DDS location.

Table 2. Summary of dust samples analyzed from NIOSH’s dry scrubber study in Mine B and the TGA
results. (NCB = non-carbonate minerals; CB = carbonates).
TGA Analysis

Avg. Grav. Recovered Mass%

Sample Conc. dust mass
Section Sample Shift Location (mg/m3) (mg) Coal NCB CB Coal:NCB

1 1* 1 UDS 1.8 1.226 55% 36% 9% 1.5

2* DDS 1.1 0.856 68% 26% 6% 2.6
3 R 1.4 1.118 57% 35% 9% 1.6
4* 2 UDS 1.1 0.884 52% 15% 33% 3.5
5* DDS 0.9 0.672 41% 13% 45% 3.2
6 R 1.2 0.868 44% 17% 39% 2.6
7* 3 UDS 1.4 0.998 47% 30% 23% 1.6
8* DDS 1.0 0.748 51% 29% 20% 1.8
9 R 1.4 1.01 46% 29% 25% 1.6
2 10* 1 UDS 1.5 0.985 61% 30% 9% 2.0
11* DDS 0.8 0.156 83% 13% 4% 6.4
12 R 0.9 0.674 69% 26% 6% 2.7
13* 2 UDS 1.3 0.831 58% 32% 10% 1.8
14* DDS 1.2 0.764 59% 32% 9% 1.8
15 R 1.2 0.793 62% 31% 7% 2.0
16* 3 UDS 1.3 0.804 58% 32% 10% 1.8
17* DDS 0.8 0.694 65% 29% 6% 2.2
18 R 1.0 0.687 58% 33% 9% 1.8

* Sample analyzed by SEM-EDX to determine particle size distribution

The current work demonstrates that detailed dust analysis can be performed on appropriate
samples preserved from prior studies. This approach has obvious merits: it enables further
insights to be gained about the effects of particular dust controls or sampling conditions on
respirable dust characteristics, without the significant time, cost and logistical efforts needed
to conduct new field work. With these benefits, however, come clear limitations too: follow-
up work on preserved samples, and interpretation of results, is constrained by the original
study design. Regarding NIOSH’s original studies that were revisited for the current work,
their sampling designs were geared toward specific research objectives, which did not include
the sort of dust sample analysis performed here. For example, samples were not collected at
the wet scrubber intake or dry scrubber exhaust, which somewhat limits the conclusions that
can be drawn about the direct effects of either scrubber on particle sizes or specific dust
components.

227
5 CONCLUSIONS

NIOSH previously evaluated a wet and a dry auxiliary scrubber and showed that both could
be effective for reducing respirable dust and/or airborne float dust mass concentrations in
underground coal mines. Here, preserved dust samples from those prior studies were further
analyzed to investigate the effect of the scrubbers on dust composition and particle size.
Results indicated that neither scrubber has an appreciable effect on dust composition. Both
scrubbers are expected to more efficiently remove coarser particles from the air that they treat
—meaning that air exiting the scrubber should have a finer size distribution than the air enter­
ing the scrubber. This effect was evident for the wet scrubber, but results were less clear for
the dry scrubber due to mixing of the scrubber exhaust with return air and/or intermittent
issues with the scrubber power during testing.
While the current work implies that scrubbers are not necessarily more efficient on one gen­
eral component of the respirable dust (i.e., dust source from the mine face versus from rock
dusting activities), future studies should focus on specific constituents of interest. For
example, crystalline silica is considered to be the most hazardous constituent and can have
a different size distribution than other dust constituents (Sarver et al., 2021). Moreover, the
effects of other controls, such as water sprays, on particle sizes or specific dust constituents, or
dust generated from specific sources, should be investigated.

Figure 3. Results of Sections 1 and 2 (i.e., the dry scrubber performance evaluation) overall particle size
distributions from SEM-EDX analysis and based on two locations (i.e., Upstream the dry
scrubber-UDS, and Downstream the face area cleaned by the dry scrubber-DDS). Plot a (Section 1) and
plot b (Section 2) represent the overall PSD for all the major dust constituents (i.e., coal, non-carbonate
and carbonates).

ACKNOWLEDGEMENT

The authors are grateful to CDC/NIOSH for funding this research (contract
75D30119C05529). Special thanks are extended to Jay Colinet, Milan Yekich, and George
Luxbacher for their support to identify and provide existing dust samples and related field
data. We also thank Justin Patts for his assistance with data interpretation for dust samples

228
collected in Mine A. We wish to acknowledge all of the other authors involved in both of the
NIOSH studies revisited here.

REFERENCES

Agioutanti, E., Keles, C., & Sarver, E. (2020). A thermogravimetric analysis application to determine
coal, carbonate, and non-carbonate minerals mass fractions in respirable mine dust. Journal of Occu­
pational and Environmental Hygiene, 17(2–3), 47–58. https://doi.org/10.1080/15459624.2019.1695057
Blackley, D. J., Halldin, C. N., & Scott Laney, A. (2018). Continued Increase in Prevalence of Coal
Workers’ Pneumoconiosis in the United States, 1970-2017. American Journal of Public Health, 108(9),
1220–1222. https://doi.org/10.2105/AJPH.2018.304517
CFR. (2021). Code of Federal Regulations-Title 30: Mineral Resources. https://www.govinfo.gov/content/
pkg/CFR-2020-title30-vol2/pdf/CFR-2020-title30-vol2-chapII.pdf
Colinet, J. F., & Jankowski, R. A. (2000). Silica collection concerns when using flooded-bed scrubbers.
Mining Engineering (Littleton, Colorado), 52.
Colinet, J. F., Reed, W., & Potts, J. D. (2013). Impact on Respirable Dust Levels When Operating a
Flooded-bed Scrubber in 20-foot Cuts. DHHS (Niosh) Publication 2014–105, Ri 9693.
Courtney, W. G., Cheng, L., & Divers, E. F. (1986). Deposition of Respirable Coal Dust in an Airway.
Report of Investigations - United States, Bureau of Mines.
Hall, N. B., Blackley, D. J., Halldin, C. N., & Laney, A. S. (2019). Current Review of Pneumoconiosis
Among US Coal Miners. Current Environmental Health Reports, 6(3), 137–147. https://doi.org/
10.1007/S40572-019-00237-5
Janisko, S. J., Colinet, J. F., Patts, J. R., Barone, T., & Patts, L. D. (2015). Field evaluation of an inline
wet scrubber for reducing float coal dust on a continuous miner section. SME Annual Meeting, 1–6.
Jaramillo, L., Agioutanti, E., Ghaychi Afrouz, S., Keles, C., & Sarver, E. (2022). Thermogravimetric
analysis of respirable coal mine dust for simple source apportionment. Journal of Occupational and
Environmental Hygiene, 19(9), 568–579. https://doi.org/10.1080/15459624.2022.2100409
NIOSH. (2021). Best Practices for Dust Control in Coal Mining. In J. By Colinet, Jay F. Halldin, Cara
N. Schall (Ed.), U.S. Department of Health and Human Services, Centers for Disease Control and Pre­
vention, National Institute for Occupational Safety and Health, DHHS (NIOSH) (Second Edi). https://
doi.org/https://doi.org/10.26616/NIOSHPUB2021119
Organiscak, J. A., Noll, J., Yantek, D., Kendall, B., & Fletcher, J. H. (2016). Examination of a newly
developed mobile dry scrubber (DS) for coal mine dust control applications. SME Annual Meeting,
1–7.
Organiscak, J., & Beck, T. (2010). Continuous Miner Spray Considerations for Optimizing Scrubber Per­
formance in exhaust ventilation systems. Society for Mining, Metallurgy, and Exploration (U.S.).
Meeting (2010 : February 28-March 3 : Phoenix, AZ). https://stacks.cdc.gov/view/cdc/9654
Patts, J. R., Colinet, J. F., Janisko, S. J., Barone, T. L., & Patts, L. D. (2016). Reducing float coal dust:
Field evaluation of an inline auxiliary fan scrubber. Mining Engineering, 68(12), 63–68. https://doi.org/
10.19150/me.6883
Sarver, E., Keleş, Ç., & Afrouz, S. G. (2021). Particle size and mineralogy distributions in respirable dust
samples from 25 US underground coal mines. International Journal of Coal Geology, 247, 103851.
https://doi.org/10.1016/J.COAL.2021.103851
Sarver, E., Keles, C., & Rezaee, M. (2019). Beyond conventional metrics: Comprehensive characteriza­
tion of respirable coal mine dust. International Journal of Coal Geology, 207(September 2018), 84–95.
https://doi.org/10.1016/j.coal.2019.03.015
Shahan, M. R., Seaman, C. E., Beck, T. W., Colinet, J. F., & Mischler, S. E. (2017). Characterization of
airborne : Float coal dust emitted during continuous mining, longwall mining and belt transport.
Mining Engineering, 69(9), 61–66. https://doi.org/10.19150/me.7746

229
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Recovery of respirable dust from fibrous filters for particle analysis

by scanning electron microscopy

A. Greth, S. Afrouz, F. Animah, C. Keles & E. Sarver

Virginia Polytechnic Institute and State University, Blacksburg, Virginia, USA

ABSTRACT: The characterization of respirable dust on the basis of constituent fractions and
particle sizes is increasingly of concern for evaluating exposure hazards. For high-resolution par­
ticle analysis, scanning electron microscopy with energy dispersive X-ray (SEM-EDX) can be an
effective tool. However, it requires particles to be deposited on a smooth, uniform substrate such
as a polycarbonate (PC) filter for optimal results. While direct sampling onto PC is possible, this
is not the standard approach in many situations. For example, in coal mines, respirable dust sam­
ples have typically been collected onto polyvinyl chloride (PVC) filters because they are intended
for gravimetric and/or infrared spectroscopy analysis. Such fibrous substrates are not ideal for
SEM-EDX (or other microscopy), but an effective method to recover and redeposit the dust par­
ticles could render such samples suitable for the additional analysis. Here, we present a simple
method and compare SEM-EDX results for paired samples analyzed (1) directly on PC and (2)
following recovery from PVC and redeposition on PC. The results indicate that dust recovery
from PVC is generally representative in terms of mineral constituents, although some changes in
particle size distributions can be observed. Careful control of particle loading density is import­
ant for either direct-on-filter or recovered dust analysis.

1 INTRODUCTION

While routine monitoring of inhalable and respirable particulates typically involves measuring mass
concentration, there is increasingly a focus on particle characteristics to better understand exposure
hazards (Stanek et al. 2011, Shekarian et al. 2021, Kumar et al. 2021). For instance, particle size
can control lung deposition and relative surface area (e.g., Sioutas et al. 2005, Ramgolam et al.
2009, Oberdorster, 2005); and particle surface area and chemistry can control lung response (e.g.,
Dalal et al. 1995, Leung et al. 2012, Ross & Murray 2004, Harrington et al. 2012). In light of the
unexpected resurgence of occupational lung disease among US coal miners (Laney et al. 2010),
a consensus report from the National Academies of Science, Engineering and Medicine recently
called for more in-depth characterization of respirable coal mine dust (RCMD) to enable a clearer
understanding of both dust sources and health implications (NASEM 2018).
Although various instruments are available for generic particle counting and sizing, to con­
duct chemical or mineralogical analysis at the individual particle level, microscopy has been
used (Abbasi et al., 2021). For particles in the respirable range (i.e., on the order of microns
and smaller), scanning electron microscopy with energy dispersive X-ray (SEM-EDX) can be
used to simultaneously acquire particle size and shape data (i.e., from the SEM images) along
with chemistry (i.e., from the EDX spectra); and the chemistry can be interpreted to infer min­
eralogy. The particle data generated by the SEM-EDX can be accumulated to estimate distri­
butions for the entire sample. Due to the time-consuming nature of manually analyzing
individual particles by SEM-EDX, computer algorithms can be used to automate the process
and more easily generate quantitative data on a per sample basis (i.e., which requires analysis
of relatively many particles in a sample). In fact, computer-controlled SEM-EDX routines
specific to RCMD analysis have previously been developed and used by the authors (e.g., see

DOI: 10.1201/9781003429241-24

230
Johann-Essex et al. 2017, Sarver et al. 2019, Sarver et al. 2021) as well as other researchers
(e.g., Pan et al. 2021, Salinas et al. 2022).
One important condition for automated SEM-EDX analysis is that the particles are deposited on
a suitable substrate—meaning one that allows for an image processing routine to accurately identify
particles in contrast to the background and also exhibits uniform composition so as not to interfere
with EDX analysis on the particles. Track-etched polycarbonate (PC) filters are commonly used for
this work because their smooth background, low impurity, and consistently sized pores are ideal for
both sample collection and analysis (Willis et al. 2002). If a dust sample is collected on PC or other
suitable filter media, SEM-EDX analysis can potentially be conducted directly on the sample filter.
However, in many cases, samples are collected for other types of analyses (Jensen & Schafer 1998,
Kulkarni et al. 2011). For example, polyvinyl chloride (PVC) filters are commonly used to sample
respirable dust when the primary aim is gravimetric analysis to determine the mass concentration in
the environment; the fibrous structure of PVC helps trap particles to avoid sample mass loss during
handling, and it is non-hygroscopic which minimizes effects of humidity on weight measurements.
For respirable mine dust, including RCMD, sampling on PVC filters also enables the determination
of quartz using infrared spectroscopy methods such as NIOSH Method 7603 (NIOSH 2003), MSHA
Method P7 (MSHA 2008), or the direct-on-filter Fourier Transform infrared method that NIOSH
has been working to develop (Miller et al. 2012, Miller et al. 2017, Cauda et al. 2016). To use such
PVC filter samples for microscopy analysis though, the dust particles should ideally be recovered and
redeposited on a suitable substrate like PC (Jensen & Schafer, 1998). This is because, even if the PVC
composition itself does not interfere with EDX analysis of the particles, the fibrous nature of the
material can hinder particle identification (versus background), especially for automated work, and
indeed particles embedded in the filter (rather than sitting on the surface) may not be seen at all.
Particulate recovery from air sampling filters has been demonstrated for applications such as
forensic analysis of dust and microbial particles from filters in heating, ventilation, and air condition­
ing (HVAC) systems (Stanley et al. 2008, Mahdavi & Siegel 2020); extraction of particulate matter
mass (PM10 or PM2.5) from PTFE-coated filters for supporting toxicological analysis (Roper et al.
2018); and the removal of PM2.5 from quartz filters to determine the oxidative potential of particu­
late to cause adverse health effects (Fang et al. 2016). Likewise, a simple method for recovery of
respirable dust from standard PVC filters for purposes of microscopy to understand dust sources
and potential health impacts would be quite valuable. However, to our knowledge, such a method
has not been demonstrated heretofore. The current work aims to evaluate the recovery and redepos­
ition of RCMD from a PVC filter onto a PC filter for SEM-EDX analysis. For this, paired PVC
and PC filters were used for dust collection and then SEM-EDX results for particle characteristics
were compared for each pair of directly analyzed (D) and recovered dust samples (R).

2 METHODOLOGY

A total of 10 pairs of respirable dust samples were generated in the laboratory for this study,
and each pair consisted of 1 PC filter and 1 PVC filter. Four powdered source materials were
used in different ratios to create the sample pairs: coal, rock dust, kaolinite, and silica. These
materials were chosen to represent the major particle types observed in real RCMD samples
when using SEM-EDX to classify particles by their mineralogy (Johann-Essex et al. 2017,
Sarver et al. 2021, Pokhrel et al. 2021a). For the coal, a clean coal product was provided by
a mine partner and then pulverized to -230 mesh (<63 µm) before dust sample generation. The
rock dust, which is primarily calcium carbonate, was sourced from a real rock dust product
provided by a mine partner. It did not need to be pulverized as it was already in a powdered
form. To contribute silica dust, MIN-U-SIL® 5 (US Silica, Katy, TX, USA) was used; this is
a high-purity silica product with a topsize of 5 µm. As a surrogate for alumino-silicates, in
general, a high-purity kaolinite powder was used (Ward’s Science, Rochester, NY, USA).
The dust samples were generated in a small enclosure per the method described by (Agioutanti
et al. 2020). Briefly, a small mass (1-10 mg) of each source material was placed at the bottom of the
enclosure; the ratio between materials was varied for each of the 10 sampling events to achieve dif­
ferent ratios of dust constituents in the samples from each event. The dust materials were

231
aerosolized using short pulses of compressed high-purity air. During each event, two respirable dust
samples (i.e., a pair) were collected in 2-piece styrene cassettes using standard equipment consisting
of a 10-mm Dorr-Oliver nylon cyclone, an ELF escort air pump (Zefon International, Ocala, FL)
operating at 2 liters/minute. One sample was collected onto a 37-mm PC filter with track-etched 0.4
µm pores (Zefon International, Ocala, FL, USA, and the other was collected onto 37-mm PVC
filters with a nominal 5.0 µm pore size (Zefon International, Ocala, FL, USA).
The SEM-EDX analysis (described below) was conducted directly on each of the PC filter
samples obtained during dust collection, and these are referred to from here as samples D1-
D10 (to denote the sample pair/sampling event number). For the PVC filter samples, the dust
was recovered and then redeposited on a PC filter for analysis (see Figure 1). Briefly, this pro­
cedure was conducted as follows: (1) the PVC filter sample was submerged in a test tube con­
taining isopropyl alcohol (IPA); (2) the test tube was sonicated for three minutes at 30 C to
dislodge the respirable dust from the PVC filter; (3) the PVC filter was removed from the test
tube; (4) the dust suspension in IPA was pulled through a 47-mm PC filter (0.4 µm pore size)
using a vacuum filtration; and (5) the PC filter was placed in clean cassette and left overnight
to dry completely. The recovered dust samples are referred to as R1-R10 from here.

Figure 1. Sequence for handling each respirable dust sample pair before SEM-EDX analysis.

The D (direct) and R (recovered) samples were prepared for SEM-EDX analysis by carefully
cutting a 9-mm circular subsection from each filter using a stainless-steel trephine. These subsections
were attached with double-sided tape to aluminum stubs and then sputter coated using Au/Pd to
make the samples conductive. SEM-EDX analysis was conducted using an FEI Quanta 600 FEG
environmental scanning electron microscope (Hillsboro, OR, USA) which was equipped with
a backscatter electron detector and a Bruker Quantax 400 EDX spectroscope (Ewing, NJ, USA)
(Sarver et al. 2021). Bruker’s Esprit software (Version 1.9.4) was used to run an automated routine
on each sample to find and analyze about 500 total particles in the range of about 1,000 nm -
10,000 nm. This “supramicron” routine and instrument settings were described in detail by Sarver
et al. (2021). Some other key parameters of the SEM-EDX analysis included a voltage of 15 kV,
1000x magnification (yielding a frame area of 14025 µm2), 5.5 spot size, and 12.5 mm working dis­
tance. For each particle, the length and width were recorded, and the elemental spectra were cap­
tured. Based on its normalized elemental ratios (i.e., for elements C, O, Al, Si, Ca, Mg, Fe, and Ti),
each particle was binned using the criteria reported by Sarver et al. (2021) into one of eight mineral­
ogy classes: carbonaceous (C), mixed-carbonaceous (MC), aluminosilicates (AS), silica (S), other
silicates (SLO), carbonates (CB), heavy minerals (M), or other (O). Then, the per-particle size and
classification data were used to estimate the particle count % in each mineralogy class per sample,
again following the approach by Sarver et al. (2021).

232
3 RESULTS AND DISCUSSION

Table 1 shows a summary of the results for all 10 sample pairs analyzed in this study. In addition
to the mineralogy distributions (particle count % in each class), the D50 for the particles in each
class and all particles analyzed in each sample is shown, as is the particle loading density (PLD).
The D50 represents the size at which 50% of the particles are finer than this particular size. The
PLD represents the number of particles observed under the SEM per unit area (i.e., particles/µm2).

Table 1. Summary of results for each pair of lab-generated respirable dust samples included in this
study.
Estimated Particle Count (%)
D50 (µm)
Pair Sample PLD C MC AS S SLO CB M O All p-value

1 D1 0.0307 62.4 5.2 2.8 0.6 0 28.4 0 0.6 100 0.8049

2.37 2.78 3.07 2.88 - 2.26 - 1.73 2.37
R1 0.0108 70 5.6 3.0 0.6 0 20.6 0 0.2 100
2.12 1.97 3.14 2.21 - 2.44 - 1.50 2.20
2 D2 0.0273 32.0 6.4 21.0 3.0 0.4 36.8 0 0.4 100 0.8979
2.20 1.92 2.65 2.20 3.26 2.69 - 3.86 2.43
R2 0.0462 29.4 9.0 23.4 1.8 0.6 35.4 0 0.4 100
1.87 2.07 2.62 2.69 3.45 2.44 - 2.76 2.30
3 D3 0.0229 16.2 5.6 58.6 4 0 14.6 0.6 0.4 100 0.0096
2.21 2.02 3.10 2.14 - 2.77 1.73 3.33 2.73
R3 0.0579 3.2 2.6 76 5.6 0.6 11.8 0 0.2 100
1.80 1.73 2.59 2.36 2.43 2.53 - 2.07 2.44
4 D4 0.0185 23.2 3.2 11.0 25.6 2.6 34.0 0.2 0.2 100 0.1513
2.04 2.07 2.78 2.77 2.88 2.56 4.99 1.65 2.53
R4 0.0363 10.4 6.6 13.2 32 5.4 32.2 0 0.2 100
1.84 1.73 2.21 2.21 2.69 2.44 - 3.82 2.21
5 D5 0.0061 10 6.6 53 10.2 0.4 18.4 1.0 0.4 100 0.0035
2.07 2.12 3.64 2.57 3.47 2.71 1.96 4.61 3.12
R5 0.0380 3.4 7.6 73 10.6 0.4 4.2 0.6 0.2 100
1.65 1.69 2.65 2.18 1.66 2.21 2.30 1.50 2.43
6 D6 0.0390 4 2 24 27 10.2 32.6 0 0.2 100 0.4843
1.85 2.03 3.35 2.21 2.93 2.37 - 3.93 2.53
R6 0.0450 9.2 4.0 23.0 21.4 5.6 36.8 0.0 0 100
1.96 2.25 2.37 2.20 2.55 2.53 - - 2.30
7 D7 0.0310 9.8 4.0 30.4 1.2 0.2 53.6 0.2 0.6 100 0.0003
2.07 1.77 3.16 2.01 1.29 2.97 1.63 2.93 2.78
R7 0.0147 25.2 13.2 18.6 6.4 0 36.2 0.4 0 100
1.84 1.73 3.14 2.07 - 2.43 4.80 - 2.20
8 D8 0.0412 5.2 2.6 31.6 1.2 0.8 57.2 0.2 1.2 100 0.2330
1.98 1.63 2.55 2.22 2.92 2.21 3.66 1.77 2.30
R8 0.0365 11.2 6.00 28.8 3.8 0 48.8 1.0 0.4 100
1.85 1.65 2.40 1.73 - 2.37 2.21 1.96 2.21
9 D9 0.0218 22.2 7.2 11.8 10.6 1.8 45.8 0.4 0.2 100 0.8279
2.01 2.04 3.57 2.12 3.35 2.57 3.42 1.96 2.40
R9 0.0159 30.2 6.8 12.2 10.8 0.6 39.4 0 0 100
1.85 1.84 3.01 2.29 1.65 2.53 - - 2.30
10 D10 0.0044 19.5 5.6 18.1 12.0 1.9 42.5 0.4 0 100 0.1695
2.44 2.53 3.58 2.34 5.47 2.65 3.15 - 2.73
R10 0.0209 8.0 6.2 26.3 17.7 3.3 37.2 0.5 0.7 100
1.98 1.92 2.93 2.57 2.97 2.65 2.43 3.68 2.57

Figure 2 shows mineralogy distributions for all ten sample pairs. Visually, there appears to be
good agreement between the D and R samples in most pairs. This indicates the method used to
recover particles from the PVC filters and redeposit them for SEM-EDX analysis generally yields

233
a representative sample. To statistically test for the similarity between the D and R sample results
for each pair, the Freeman-Halton exact test of independence was conducted (Agresti 1992). Here,
the null hypothesis is that the sample mineralogy is independent of whether the dust is analyzed
directly or after being recovered and redeposited. P-values for each sample pair are shown in
Table 1 for 95% confidence, and confirm that mineralogy distributions are indeed similar for seven
of the 10 pairs (i.e., p > 0.05). For the other three pairs (Pairs 3, 5, and 7), a close inspection of the
mineralogy results in Table 1 and Figure 2 reveals that disagreements between the D and
R samples are primarily due to shifts in the particle count % reporting to the C, AS and CB classes.
For Pairs 3 and 5, the SEM-EDX analysis indicated the D sample had relatively more C (i.e., coal)
but less AS (i.e., kaolinite) content than its R counterpart—and this shift in mineralogy was concur­
rent with a relatively large increase in the PLD from the D to R sample. On the other hand, for
Pair 7, the D sample had less C but more AS and CB (i.e., rock dust) than its R counterpart—and
the PLD decreased from the D to R sample. Another possibility for the disagreement for pairs 3
and 5 may have been the high amount of kaolinite powder, >50%, used to generate the dust sam­
ples. Because of the SEM-EDX’s tendency to erroneously identify the C present in coal as AS pre­
sent in the kaolinite, the high amount of respirable kaolinite particles surrounding coal particles
may have caused coal particles to be identified as kaolinite.
Figure 3 illustrates the apparent shifts in mineralogy from D to R samples as a function of
the difference in those samples’ PLDs for all 10 pairs. (Note that only the C, MC, AS, S, and
CB classes are shown since these accounted for nearly all particles in each sample.) The shift

Figure 2. Mineralogy distribution results (based on particle counts) for the 10 pairs of directly analyzed
(D) and recovered dust samples (R).

Figure 3. Difference in mineralogy class (i.e., C, MC, AS, S or CB) distributions based on particle
counts versus the difference in PLD (particles/µm2) for the 10 pairs of directly analyzed (D) and
recovered dust samples (R).

234
Figure 4. Cumulative particle size distributions for the directly-analyzed (D) and recovered dust sam­
ples (R) in all 10 sample pairs. The y-axis in each plot shows the % of total particles finer than the respect­
ive particle size shown on the x-axis.

between C and AS is correlated with changes in PLD and can be explained at least in part by
the increased likelihood of misclassifying C particles as PLD is increased. In essence, very
nearby mineral particles (e.g., rich in Al and Si) can interfere with the EDX spectrum collected
on a coal particle (i.e., which, itself, is presumably deficient in Al and Si), such that it might be
classified as mineral (e.g., AS) instead of C. The possibility of particle misclassification due to
high loading has been recently discussed in other studies (Keles et al., 2022, Pokhrel et al.,
2021b) and can potentially affect other particles too (e.g., silica). That said, Figure 3 indicates
changes in PLD between the D and R samples did not have much effect on the classification of

235
particles in the MC, S, or CB classes. Notably, the tendency for D samples to have more CB
than their R counterparts could mean that CB particles were somewhat more susceptible to
loss during the dust recovery procedure than other particle types.
Figure 4 allows a comparison of the D and R samples concerning their overall particle size
distributions. Across all 10 sample pairs, a slight reduction in the size distribution can be
observed for the recovered dust. This is likely due to some deagglomeration during the recov­
ery procedure. A recent study by Gonzalez et al. (2022) showed that respirable-sized agglom­
erates were present in dust samples generated in the laboratory under similar conditions to
those used here and that deagglomeration could be promoted by sonication. Aside from yield­
ing a slightly finer size distribution, deagglomeration might also contribute to the small shift
toward more C and less AS in some of the R samples compared to their D counterparts (e.g.,
if coal-mineral agglomerates break up)—though it is impossible in the current work to separ­
ate this effect from that of the PLD (Figure 3).
When comparing the overall particle size distributions seen in Figure 4 concerning mineral­
ogy class, the observed trends do support that the recovery method used yields a representative
sample of the present particles. The largest observed shift in particle sizes can be seen for the
C and AS mineralogy classes as seen in samples 3, 5, and 7. In these pairs, the R samples are
always finer than the D samples. This is likely because of the deagglomeration which occurs as
a result of sonication when recovering dust from the PVC filters onto a PC filter. Sonication
would cause fine AS particles to separate from coarser C particles. The effects of the sonication
are also evident in the D50 values for C and AS when comparing samples 3, 5, and 7 as the
D50 values for the R samples are lower than the D samples. This reinforces the fact that deag­
glomeration resulting from sonication can have an impact on particle size distribution.

4 CONCLUSION

A simple method for recovering respirable dust from fibrous sampling filters can enable post­
hoc analysis of individual particles. This could be useful in a variety of situations including,
for instance, the case where preserved mine dust samples could be revisited to explore particle
characteristics—even if such analysis was not the original intent of the sample collection.
Overall, the methodology demonstrated here appears reliable for recovering a representative
sample of dust from fibrous PVC filters—especially if particle loading density (PLD) can be
controlled when the recovered dust is redeposited on the PC filter for SEM-EDX analysis. For
the current study, where dissimilarities were observed in mineralogy distributions between dir­
ectly analyzed and recovered dust sample pairs, differences in PLD appear to be the primary
factor. The potential for deagglomeration of dust particles during the recovery procedure must
also be considered with regard to the analytical objectives and interpretation of results.

ACKNOWLEDGEMENTS

The authors would like to thank the Alpha Foundation for Improvement of Mine Safety and
Health (grant AFCTG20-104) and the National Institute of Occupational Safety & Health (con­
tract 75D30119C05529) for funding this work. Additionally, we acknowledge the Institute for
Critical Technology and Applied Sciences’ Nanoscale Characterization and Fabrication Labora­
tory (NCFL), which houses the SEM-EDX facilities used here. The NCFL is supported by the
Virginia Tech National Center for Earth and Environmental Nanotechnology Infrastructure
(NanoEarth), a member of the National Nanotechnology Coordinated Infrastructure (NNCI),
supported by NSF (ECCS 1542100 and ECCS 2025151). Views expressed in this study are those
of the authors and not necessarily the views of sponsors or research partners.

REFERENCES

Abbasi, B., Wang, X., Chow, J. C., Watson, J. G., Peik, B., Nasiri, V., Riemenschnitter, K., &
Elahifard, M. (2021). Review of respirable coal mine dust characterization for mass concentration,
size distribution, and chemical composition. Minerals, 11(4), 426.

236
Agioutanti, E., Keleş, Ç, & Sarver, E. (2020). A thermogravimetric analysis application to determine
coal, carbonate, and non-carbonate minerals mass fractions in respirable mine dust. Journal of Occu­
pational and Environmental Hygiene, 17(2-3), 47–58.
Agresti, A (1992). A survey of exact inference for contingency tables. Statistical Science, 7, 131–153.
Cauda, E., Miller, A. & Drake, P. (2016) Promoting early exposure monitoring for respirable crystalline
silica: Taking the laboratory to the mine site. Journal of Occupational and Environmental Hygiene, 13,
D39–D45.
Dalal, N.S., Newman, J., Pack, D., Leonard, S., & Vallyathan, V. (1995). Hydroxyl radical generation by
coal mine dust: possible implication to coal workers’ pneumoconiosis (CWP). Free Radic. Biol. Med.
18 (1), 11–20.
Fang, T., Verma, V., Bates, J. T., Abrams, J., Klein, M., Strickland, M. J., Sarnat, S.E., Chang, H.H.,
Mulholland, J.A., Tolbert, P.E., Russell, A.G., & Weber, R. J. (2016). Oxidative potential of ambient
water-soluble PM 2.5 in the southeastern United States: contrasts in sources and health associations
between ascorbic acid (AA) and dithiothreitol (DTT) assays. Atmospheric Chemistry and Physics, 16
(6), 3865–3879.
Gonzalez, J., Keleş, Ç, & Sarver, E. (2022). On the occurrence and persistence of coal-mineral microag­
glomerates in respirable coal mine dust. Mining, Metallurgy & Exploration, 39(2), 271–282.
Harrington, A.D., Hylton, S., Schoonen, M.A.A., 2012. Pyrite-driven reactive oxygen species formation
in simulated lung fluid: Implications for coal workers’ pneumoconiosis. Environ. Geochem. Health 34
(4), 527–538.
Jensen, P. A., & Schafer, M. P. (1998). Sampling and characterization of bioaerosols. NIOSH manual of
analytical methods, 1(15), 82–112.
Johann-Essex, V., Keleş, Ç, & Sarver, E. (2017). A computer-controlled SEM-EDX routine for charac­
terizing respirable coal mine dust. Minerals 7 (1), 15. https://doi.org/10.3390/min7010015.
Keles, C., Pokhrel, N., & Sarver, E. (2022). A Study of Respirable Silica in Underground Coal Mines:
Sources. Minerals, 12(9), 1115.
Kulkarni, P., Baron, P. A., & Willeke, K. (Eds.). (2011). Aerosol measurement: principles, techniques, and
applications. John Wiley & Sons.
Kumar, P., Kalaiarasan, G., Porter, A.E., Pinna, A., Klosowski, M.M., Demokritou, P., Chung, K.F.,
Pain, C., Arvind, D.K., Arcucci, R., & Adcock, I.M. (2021). An overview of methods of fine and ultra­
fine particle collection for physicochemical characterization and toxicity assessments. Science of the
Total Environment, 756, 143553.
Laney, A. S., Petsonk, E. L., and Attfield, M. D. 2010. Pneumoconiosis among underground bituminous
coal miners in the United States: Is silicosis becoming more frequent? Occupational and Environmental
Medicine 67(10):652–656.
Leung, C. C., Yu, I. T. S., & Chen, W. (2008). Silicosis. The Lancet, 379(9830), 2008–2018.
Mahdavi, A., & Siegel, J. A. (2020). Extraction of dust collected in HVAC filters for quantitative filter
forensics. Aerosol Science and Technology, 54(11), 1282–1292.
Miller, A. L., Drake, P. L., Murphy, N. C., Noll, J. D. & Volkwein, J. C. (2012) Evaluating portable
infrared spectrometers for measuring the silica content of coal dust. J. Environ. Monit., 14, 48–55.
Miller, A. L., Weakley, A. T., Griffiths, P. R., Cauda, E. G. & Bayman, S. (2017) Direct-on-Filter α-
Quartz Estimation in Respirable Coal Mine Dust Using Transmission Fourier Transform Infrared
Spectrometry. Applied Spectroscopy, 71(5), 1014–1024.
MSHA, 2008. Method No. MSHA P7: Infrared Determination of Quartz in Respirable Coal Mine Dust.
United States Department of Labor-MSHA-Pittsburgh Safety and Health Technology Center.
NASEM (2018). Monitoring and Sampling Approaches to Assess Underground Coal Mine Dust Expos­
ures, Washington, DC, The National Academies Press.
National Institute for Occupational Safety and Health (NIOSH): QUARTZ in coal mine dust, by IR
(NIOSH 7603, NIOSH Manual of Analytical Methods 4th ed.). Cincinnati, OH: NIOSH, 2003.
Oberdörster, G., Oberdörster, E., & Oberdörster, J. (2005). Nanotoxicology: an emerging discipline
evolving from studies of ultrafine particles. Environmental Health Perspectives, 113(7), 823–839.
Pan, L., Golden, S., Assemi, S., Sime, M. F., Wang, X., Gao, Y., & Miller, J. (2021). Characterization of
particle size and composition of respirable coal mine dust. Minerals, 11(3), 276.
Pokhrel, N., Keles, C., Jaramillo, L., Agioutanti, E., & Sarver, E. (2021a). Direct-on-Filter FTIR Spec­
troscopy to Estimate Calcite as A Proxy for Limestone ‘Rock Dust’in Respirable Coal Mine Dust
Samples. Minerals, 11(9), 922.
Pokhrel, N., Agioutanti, E., Keleş, Ç., Afrouz, S., & Sarver, E. A. (2021b). Comparison of mineral con­
tent in respirable coal mine dust samples estimated using FTIR, TGA, and SEM-EDX. In Mine Venti­
lation (pp. 255–267). CRC Press.

237
Ramgolam, K., Favez, O., Cachier, H., Gaudichet, A., Marano, F., Martinon, L., & Baeza-Squiban, A.
(2009). Size-partitioning of an urban aerosol to identify particle determinants involved in the proin­
flammatory response induced in airway epithelial cells. Particle and Fibre Toxicology, 6(1), 1–12.
Roper, C., Delgado, L. S., Barrett, D., Massey Simonich, S. L., & Tanguay, R. L. (2018). PM2. 5 filter
extraction methods: implications for chemical and toxicological analyses. Environmental science &
technology, 53(1), 434–442.
Ross, M. H., & Murray, J. (2004). Occupational respiratory disease in mining. Occupational medicine, 54
(5), 304–310.
Salinas, V., Das, M., Jacquez, Q., Camacho, A., Zychowski, K., Hovingh, M., Medina, A.,
Rubasinghege, G., Rezaee, M., Baltrus. J., Fairley, N., & Roghanchi, P. (2022). Characterization and
Toxicity Analysis of Lab-Created Respirable Coal Mine Dust from the Appalachians and Rocky
Mountains Regions. Minerals, 12(7), 898.
Sarver, E., Keles, C., & Rezaee, M. (2019). Characteristics of respirable dust in eight appalachian coal
mines: A dataset including particle size and mineralogy distributions, and metal and trace element
mass concentrations. Data in brief, 25, 104032.
Sarver, E., Keleş, Ç., & Afrouz, S. G. (2021). Particle size and mineralogy distributions in respirable dust
samples from 25 US underground coal mines. International Journal of Coal Geology, 247, 103851.
Shekarian, Y., Rahimi, E., Rezaee, M., Su, W. C., & Roghanchi, P. (2021). Respirable coal mine dust:
a review of respiratory deposition, regulations, and characterization. Minerals, 11(7), 696.
Sioutas, C., Delfino, R. J., & Singh, M. (2005). Exposure assessment for atmospheric ultrafine particles
(UFPs) and implications in epidemiologic research. Environmental health perspectives, 113(8), 947–955.
Stanek, L. W., Sacks, J. D., Dutton, S. J., & Dubois, J. J. B. (2011). Attributing health effects to appor­
tioned components and sources of particulate matter: an evaluation of collective results. Atmospheric
Environment, 45(32), 5655–5663.
Stanley, N. J., T. H. Kuehn, S. W. Kim, P. C. Raynor, S. Anantharaman, M. A. Ramakrishnan, and
S. M. Goyal. 2008. Background culturable bacteria aerosol in two large public buildings using HVAC
filters as long term, passive, high-volume air samplers. J. Environ. Monit. 10 (4): 474–481.
Willis, R.D. & Conner, T.L. Guidelines for the Application of SEM/EDX Analytical Techniques to Par­
ticulate Matter Samples; US EPA: Washington, DC, USA, 2002.

238
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Effects of vertical air-blocking ring of drill shroud on dust control

for surface mine drilling operation using CFD

Y. Zheng, J.D. Potts & W.R. Reed

Centers for Disease Control and Prevention (CDC), National Institute for Occupational Safety and Health
(NIOSH), Pittsburgh, PA, USA

ABSTRACT: Dust control by a horizontal air-blocking shelf inside a drill shroud was effect­
ive in a previous study by researchers at the National Institute for Occupational Safety and
Health (NIOSH). However, when a medium-sized drilling machine lowers its mast, material
deposited on the shelf can produce a significant dust cloud. To minimize this dust exposure
hazard, researchers applied a 208-L (55-gallon) drum-sized ring, installing it concentrically
with the drill stem under the deck table to act as a vertical air-blocking ring to control the drill
dust leakage. Simultaneously, gravity force was utilized to minimize material buildup. Com­
puter simulation was used in the evaluation. Three groups of ring heights were studied: short
[5.1-cm (2-in), 7.6-cm (3-in), and 10.2-cm (4-in)], medium [15.2-cm (6-in), 30.5-cm (12-in), and
45.7-cm (18-in)], and long [61.0-cm (24-in) and 76.2-cm (30-in)]. The dust leakage was simu­
lated using computational fluid dynamics (CFD) under the commonly encountered field con­
ditions of a 2:1 dust-collector-to-bailing airflow ratio and a 5.1-cm (2-in) shroud-to-ground
gap. Based on this simulation study, the vertical air-blocking ring with medium heights has
the potential to effectively confine the dust inside the drill shroud. Other shorter or longer
height rings may not be as effective in preventing dust leakage from the shroud. The results of
this study can be used to guide future laboratory tests.

1 INTRODUCTION

Drilling operations are common sources of respirable dust generation and liberation at surface
mines. These dust emissions can cause high dust exposures for the drill operator and nearby
mineworkers downwind of the drill if windy conditions are present. When drilling through
silica-bearing materials, such as sandstone and shale, mineworkers can be exposed to high
concentrations of respirable silica dust. Silica dust overexposures can lead to silicosis, an occu­
pational lung disease that has no cure and is often fatal [Lara, 2020].
To reduce or eliminate the likelihood of developing silicosis, many rotary drilling operations
use a dry collection system [Organiscak and Page, 2005]. Figure 1 shows a typical dry dust
collection system schematic for medium and large drills. Compressed bailing air is forced
through the inside of the hollow drill stem and released at the end of the drill bit to flush the
drill cuttings to the surface. On the surface, the bailing air with cuttings are contained by the
drill deck and deck shroud. The larger cuttings are deposited inside the shroud area while fine
particles are pulled into a dust collection system, where they are trapped in filters prior to the
air being exhausted back into the environment through the fan. To prolong filter life, the fil­
ters are back-flushed regularly with compressed air to remove excess particulates, typically
known as the dust filter cake. This dust falls out into the collector chamber and then ultim­
ately drops out of the bottom of the collector onto the mining bench. More detailed descrip­
tions of this operation can be found in studies conducted by the U.S. Bureau of Mines and the
NIOSH [Maksimovic and Page, 1985; Organiscak and Page, 1995; Reed et al., 2008].

DOI: 10.1201/9781003429241-25

239
Figure 1. A basic dry dust collection system on a drill.

Previous laboratory tests and CFD studies have demonstrated that under a 5.1-cm (2-in)
deck shroud-to-ground gap condition, if the dust collector-to-bailing airflow ratios are 3:1,
the dust leakage from the drill shroud will be greatly reduced. If the ratios can be increased to
4:1, the respirable dust can be well-confined within the drill shroud. However, this higher ratio
is difficult to attain at mining sites. The ratio encountered during past NIOSH studies of oper­
ating drills averaged about 2:1 [Zheng et al., 2016].
To better confine the dust cloud inside the drill shroud, the airflow pattern was further
investigated. The Coanda effect was observed [Potts and Reed, 2008]. Two common phenom­
ena of the Coanda effect are: first, the tendency of an air jet approaching a curved surface to
remain attached to the surface even after the surface curves away from the initial jet direction;
and second, the tendency of an air jet to adhere to a nearby surface [Trancossi, 2011]. Within
a drill shroud, the airflow tends to follow the inner surfaces of the drill shroud and leaks out
at any shroud-to-ground gap. An air-blocking shelf was configured to reduce the Coanda
effect and was shown to be effective in both laboratory and field test scenarios [Potts and
Reed, 2008; 2011]. CFD simulations also evaluated the air-blocking shelf and estimated that
a dust-collector-to-bailing airflow ratio as low as 1.75:1 when using the air-blocking shelf may
effectively confine dust inside the drill shroud [Zheng et al., 2018].
During field studies, a considerable amount of debris was deposited on the air-blocking
shelf during the drilling process (Figure 2), and these deposits did not appear to affect the dust
control effects of the air-blocking shelf. However, when the drill mast was lowered to move
from one location to another, the buildup of materials on the air-blocking shelf liberated
a dust cloud when the material fell off the shelf, as shown in Figure 3.
To eliminate debris on top of the air-blocking shelf, NIOSH researchers proposed two types
of inner-shaped drill shrouds, U-shaped and V-shaped, along with a drill shroud with air-
blocking shelves installed at a 45° angle from the horizontal direction. Those three designs
provide a sloped shelf surface with an angle greater than the dust’s angle of repose. This
allows gravity to overcome the material’s internal friction to prevent material accumulations.
The research results revealed that the U-shaped drill shroud and the 45° air-blocking shelves
can potentially confine the dust inside the drill shroud [Zheng et al., 2021].
This study uses a vertical air-blocking ring to eliminate debris accumulation. To simplify
the application, commercially available 208-L (55-gallon) drums (Figure 4) can be utilized
accordingly and cut into different heights. The resulting drum section can then be installed
concentrically with the drill stem under the drill deck to act as a vertical air-blocking ring.
A total of eight cases were studied using CFD analysis with different heights of rings cut from
the commercially available 208-L (55-gallon) drum [0.58-m (23-in) in diameter and 0.86-m
(34-in) in total height]: 5.1-cm (2-in), 7.6-cm (3-in), 10.2-cm (4-in), 15.2-cm (6-in), 30.5-cm (12-
in), 45.7-cm (18-in), 61.0-cm (24-in), and 76.2-cm (30-in). The dust leakage was evaluated
under commonly encountered field conditions of a 2:1 dust-collector-to-bailing airflow ratio
and a 5.1-cm (2-in) shroud-to-ground gap. This current study, along with previously validated
CFD studies, was adopted to evaluate whether the dust-laden airflow can still be effectively
confined inside the drill shroud with minimal leakage when using the vertical air-blocking
ring.

240
Figure 2. The buildup of drill cuttings on the Figure 3. Dust generated during the lowering of the
air-blocking shelves. drill mast.

Figure 4. 208-L (55-gallon) drums manufactured with different materials with 0.58-m (23-in) in diam­
eter and 0.86-m (34-in) in total height [EBK, 2022].

2 CFD MODELING

The ANSYS Fluent Version 18.2 program (ANSYS, Canonsburg, PA) was used to perform
the analysis of dust distribution within the drill shroud and surrounding space. The schematic
of the airflow domain inside the drill table simulator was built according to the geometry
measured from NIOSH’s full-sized test facility. Figure 5 (a) shows the size of the dust cham­
ber: 3.66-m wide by 3.05-m deep by 2.44-m high (12-ft by 10-ft by 8-ft). The shroud, located
in the center of the dust chamber, measures 1.52-m wide by 1.22-m deep by 1.22-m high (5-ft
by 4-ft by 4-ft). More detailed setup of the full-scale facility can be found in previous literature
[Organiscak and Page, 2005; Potts and Reed, 2008, 2011].
A schematic of the computational domain is shown in Figure 5. Only two cases out of the
eight geometric models are shown here. In Figure 5 (a), the segment of the drum is a 5.1-cm
(2-in) height ring under the bottom of the drill deck with a common drum diameter of
58.4-cm (23-in). In Figure 5 (b), the height of the ring shown is 45.7-cm (18-in).
The locations of the boundary conditions for CFD are illustrated in Figure 5 (b). The
boundary conditions applied in the simulation include roof inlet, bailing air inlet, and dust
collector outlet. Intake airflow is pulled into the simulation domain through the three open­
ings on the roof, shown as “Roof inlet” in Figure 5 (b). At the same time, the bailing air with
dust is injected into the simulation domain from a circular face inside the drill pipe, as indi­
cated by the bailing air inlet. The dust collector pulls air from inside the drill shroud and dis­
charges it at the dust collector outlet.
Previously validated CFD models were used in this study to evaluate the effect of these ver­
tical air-blocking rings of the drill shroud on dust control [Zheng et al., 2016; 2018]. Dust was
treated as a gas (CO2), and a species transport model in ANSYS Fluent was used. The bound­
ary conditions used to determine the dust distribution inside the domain can be referred to in
Zheng [2016]. The airflow patterns and dust control capabilities were simulated for the eight

241
Figure 5. Overview of the CFD model with boundary conditions for (a) 5.1-cm (2-in) vertical
air-blocking ring; (b) 45.7-cm (18-in) vertical air-blocking ring.

configurations and evaluated under the same ventilation conditions: 0.24 m3/s (500 cfm) of
bailing airflow, 0.24 m3/s (500 cfm) of intake roof airflow, and 0.48 m3/s (1,000 cfm) of dust
air mixture collected by the dust collector outlet, which equates to a 2:1 collector-to-bailing
airflow ratio.

3 AIRFLOW AND DUST DISTRIBUTION RESULTS FROM CFD MODELS

The vertical air-blocking ring has a 90° slope angle in all eight cases, which is much larger
than the angles of repose of minerals commonly encountered during drilling, which range
from 0° to 50° [Dougherty and Schissler, 2020]. Under the vertical slope angle, the materials
on the ring’s inner walls cannot stay in place and will fall to the ground during the drilling
process. The inner surface of the ring should be smooth enough to not provide any rough sur­
faces that allows any deposition of small dust particles. In addition, the material of the drum
should not offer any electrostatic force to attach small particles. Based on the simulation
results of the eight cases, the phenomena and dust control effects can be separated into three
groups and are described in the following sections:

3.1 Effect of the short vertical air-blocking ring from 5.1-cm (2-in) to 10.2-cm (4-in)
Similar airflow patterns are observed for the vertical air-blocking ring with 5.1-cm (2-in),
7.6-cm (3-in), and 10.2-cm (4-in) height cases. Overall, the ring can divert airflow toward the
ground, but cannot confine the dusty air inside the drill shroud.
Figure 6 shows that after the dust-laden air is released into the bailing air inlet, the air
travels down inside the hollow drill stem, then reverses direction up the gap between the drill
steel and drill hole. As the air flows up, it attaches to the outside surface of the drill steel and
encounters the underside of the drill deck, where it fans out in all directions and flows to
where the vertical air-blocking ring is located.
After the flow encounters the vertical air-blocking ring, the CFD simulation shows that the
flow is blocked by the ring and is diverted downward, as shown in Figure 6 (b). Then, on the
left rear corner of the drill shroud where the inlet of the dust collector is located, the dusty
bailing air is exhausted into the dust collector pipe after the ring. On the right side of the drill
shroud, the bailing air encounters additional air that is drawn into the drill shroud through
the 5.1-cm (2-in) shroud-to-ground gap. The combined air is then recirculated counterclock­
wise toward the upper region of the drill shroud and flows toward the dust collector inlet.
Some bailing airflow at the front right corner of the drill shroud, after being deflected from
the vertical ring, flows outward and is attached to the vertical deck shroud. It then flows
down to the shroud-to-ground gap. Part of this airflow can leak out of the drill shroud after
striking the ground.

242
 The airflow before encountering the vertical air-blocking ring shows a Coanda effect. For the
three cases with 5.1-cm (2-in), 7.6-cm (3-in), and 10.2-cm (4-in) ring heights, the diverted down­
ward flow can still flow outward and attach to the vertical deck shroud wall. This is also
a Coanda effect and can cause dust leakage. It can be observed in Figure 6 that airflow path
lines are shown outside of the shroud and in Figure 7 that dust can leak out of the shroud and
potentially expose any workers around the drill to respirable dust. The amount of dust leakage
can result in potential relative dust concentrations ranging from 1.5 to 15.5 mg/m3. In addition,
the dust leakage occurs on the opposite side of the collector inlet, where the influence of the
dust collector is relatively weak. This side is generally where the drill operator’s cab is situated.

Figure 6. The pathlines of bailing airflow colored by velocity magnitude (0.0–10.0 m/s), 10.2-cm (4-in)
height vertical air-blocking ring: (a) fully developed pathlines; (b) bailing air as it passes by the ring.

Figure 7. Respirable dust concentration distributions under 10.2-cm (4-in) height vertical air-blocking
ring; legend shows the dust levels (1.5–36.5 mg/m3).

3.2 Effect of the medium vertical air-blocking ring from 15.2-cm (6-in) to 45.7-cm (18-in)
For the cases with 15.2-cm (6-in), 30.5-cm (12-in), and 45.7-cm (18-in) height rings, the
increased ring height can confine dusty bailing air more effectively at the center region inside
the drill shroud as shown in Figure 8. The dust-laden air travels down the inside of the hollow
drill stem, then goes up along the outside surface of the drill steel and fans out in all directions
on the underside of the drill deck before the vertical air-blocking ring, similarly to the airflows
depicted in Figure 6 (b).
As the airflow comes out below the ring, there are roughly three groups of flow: 1) the flow
that goes directly toward the dust collector inlet on the left rear corner of the drill shroud; 2)
the flow on the right side of the drill shroud, where the bailing air flows counterclockwise

243
toward the upper region of the drill shroud and then flows toward the dust collector inlet
through the upper front and upper rear regions outside of the ring; and 3) the flow that comes
down at the front and right corner of the drill shroud which generally does not reach the
ground, then flows toward the dust collector inlet. Compared to the flow in Figure 6, the dif­
ference is the third group of bailing airflow which is diverted by the vertical ring closer to the
center of the drill shroud without attachment to the vertical drill shroud wall and therefore
has a smaller chance to leak out of the shroud. This can also be observed in Figure 9 which
demonstrates no respirable dust outside of the shroud.

Figure 8. The pathlines of bailing airflow col­ Figure 9. Respirable dust concentration distributions
ored by velocity magnitude (0.0–10.0 m/s), under 30.5-cm (12-in) height vertical air-blocking ring;
15.2-cm (6-in) height vertical air-blocking ring. the legend shows the dust levels (1.5–36.5 mg/m3).

3.3 Effect of the long vertical air-blocking ring from 61.0-cm (24-in) to 76.2-cm (30-in)
As the vertical air-blocking ring increases to 61.0-cm (24-in) and 76.2-cm (30-in), the airflow
pattern changes. There is a Coanda effect that shows up on the ground, which can be
a problem, as revealed by this study.
The results from the CFD simulation in Figure 10 show that the pathlines of bailing airflow
have the same flow patterns before they leave the bottom edge of the ring. At this point, the
airflow remains directed downward, and the airflow velocity can be in the range of 4.0–5.0 m/
s (787.4–984.3 fpm). Because the distance from the ground to the bottom of the ring is
52.8-cm (20.8-in) for the 61.0-cm (24-in) height ring and 42.7-cm (16.8-in) for the 76.2-cm (30-
in) height ring, the high-velocity downward flow can reach the ground.
Once the airflow reaches the ground, it fans out horizontally in all directions (Coanda effect)
toward the shroud-to-ground gap where it encounters the inflow from outside of the drill
shroud. Although there is incoming flow in the opposite direction, dusty bailing air can still leak

Figure 10. The pathlines of bailing airflow Figure 11. Respirable dust concentration distributions
colored by velocity magnitude (0.0–10.0 m/s), under 76.2-cm (30-in) height vertical air-blocking ring;
76.2-cm (30-in) height vertical air-blocking ring. legend shows the dust levels (1.5–36.5 mg/m3).

244
out of the drill shroud in regions where the inflow is weak. As shown in Figure 11, these regions
are located at the front right corner of the drill shroud, opposite of the dust collector inlet.

4 DISCUSSIONS

It was observed from the eight CFD simulations in this study that when the vertical air-
blocking ring is short (not higher than 10.2-cm or 4-in), it can divert the airflow away from the
bottom of the drill deck. However, it cannot alter the airflow direction completely from horizon­
tal to vertical. The angled downward bailing airflow can still reattach to the vertical deck
shroud walls and flow to the shroud-to-ground gap. Leaking can occur at the front and right
corner where the influence of the dust collector is weak. This location is where the drill oper­
ator’s cab is situated. Therefore, it is important for the drill operator’s cab to have a properly
designed and maintained enclosed cab filtration and pressurization system to prevent the drill
operator from respirable dust exposure during drilling [Organiscak et al., 2018].
When the height of the vertical air-blocking ring increases to the medium range, from
15.2-cm (6-in) to 45.7-cm (18-in), most of the bailing airflow is confined in the central region
of the drill shroud. Very little airflow reattaches to the vertical deck shroud walls contrary to
the short vertical air-blocking ring cases (as shown in Section 3.1). There may be some airflow
that reaches the ground as in longer vertical air-blocking ring cases (as shown in Section 3.3).
However, most of the airflow remains in the center region of the drill shroud and can be con­
fined effectively by the inflow of air from outside the drill shroud.
In the long range, once the heights of the vertical air-blocking ring are increased to 61.0-cm
(24-in) and above, the bailing airflow is redirected vertically down to the ground and fans out
in all directions due to the Coanda effect. As a result of the relatively high speed of the flow
from the vertical ring, the outflow fanning at the ground surface can break through the
shroud-to-ground gap, overwhelm incoming airflow from the opposite direction and increase
the dust exposure to the surrounding environment.
For medium and large drill equipment with similar drill shroud size, this CFD simulation
provides a range of vertical air-blocking ring heights that may be effective for dust control.
However, it should be noted that as the drill cuttings accumulate on the ground inside the drill
shroud (as shown in Figures 1 and 2), the distance between the growing cuttings and the
bottom of the ring can dramatically decrease. The application of the vertical air-blocking ring
may need to be adjusted accordingly based on the field conditions. For example, if the cuttings
gradually accumulated to 25.4-cm (10-in) above the ground, the ring heights from 15.2-cm
(6-in) to 45.7-cm (18-in) that were shown to be effective in this simulation may be required to
be changed to between 15.2-cm (6-in) and 20.3-cm (8-in).

5 CONCLUSIONS

It has been previously demonstrated by NIOSH researchers that horizontally installed air-
blocking shelves can effectively control the respirable dust inside the drill shroud during the
drilling process. These shelf designs have no impact on drill operations or present any add­
itional maintenance requirements. Once installed, the device is invisible, and drilling can con­
tinue as before with no change in productivity. However, the deposits on these shelves can
create dust plumes when the machine changes drilling locations. In this study, various heights
of vertical air-blocking rings were investigated to utilize gravity force to remove the deposit
during drilling operations, preventing the buildup of material.
To simplify the application and minimize cost, the dimension of a commercially available
208-L (55-gallon) drum was built in the simulation and cut into a specific height of ring. The
ring was then put concentrically with the drill stem under the deck table to act as a vertical
air-blocking ring. The dust control capacities of eight rings’ heights were evaluated using
a previously validated CFD model.
Based on this study, the vertical air-blocking ring, categorized as medium, with a height
between 15.2-cm (6-in) and 45.7-cm (18-in), has the potential to effectively confine the dust

245
inside the drill shroud. The cases with shorter heights between 5.1-cm (2-in) and 10.2-cm (4-in)
and longer heights above 61.0-cm (24-in) may have dust-leaking problems at regions where
the dust collector’s influence is weak.
The results of this study can be used to guide future laboratory tests. The vertical air-blocking
ring combined with a different inner-shaped drill shroud design [Zheng et al., 2021] can provide
a solution to dust problems while drilling and when the surface drill location is changed.

ACKNOWLEDGEMENTS

One of the authors of this paper, J. Drew Potts, designed the vertical air-blocking ring in this
study. The authors of this paper would like to sincerely acknowledge Andy B. Cecala, Carin
L. Kosmoski, Liming Yuan, Lihong Zhou, Tim W. Beck, Hua Jiang, and Scott S. Klima for
their technical support.

DISCLAIMERS

The findings and conclusions in this manuscript are those of the authors and do not necessar­
ily represent the official position of the National Institute for Occupational Safety and Health
(NIOSH). Mention of company names or products does not constitute an endorsement by
NIOSH.

REFERENCES

Dougherty, HN, and Schissler, AP. 2020. SME mining reference handbook. Englewood, Colorado: Soci­
ety for Mining, Metallurgy & Exploration (SME).ISBN #: 978–0–87335–435–6.
EBK 2022. 55 Gallon Drums. EBK containers, https://ebkcontainers.com/drums-ibc-totes/drums/55-
gallon-drums/, accessed Oct 28, 2022.
Lara AR. 2020. Silicosis. The Merck Manual, Consumer Version. Kenilworth NJ: Merck & Co. Inc.
Maksimovic SD and Page SJ 1985. Quartz dust sources during overburden drilling at surface coal mines.
Department of the Interior, U.S. Bureau of Mines, Information Circular 9056.
Organiscak JA and Page SJ 1995. Assessment of airborne dust generated from small truck-mounted rock
drills. Department of the Interior, U.S. Bureau of Mines, Report of Investigations 9616.
Organiscak JA and Page SJ 2005. Development of a dust collector inlet hood for enhanced surface mine
drill dust capture. Int J Surf Min Reclamat. Environ 19(1): 12–28.
Organiscak JA, Cecala AB, and Hall RM 2018. Design, testing, and modeling of environmental enclos­
ures for controlling worker exposure to airborne contaminants. Pittsburgh PA: U.S. Department of
Health and Human Services, Centers for Disease Control and Prevention, National Institute for Occu­
pational Safety and Health, DHHS (NIOSH) Publication No. 2018–123, IC 9531. https://doi.org/
10.26616/NIOSHPUB2018123.
Potts JD and Reed WR 2008. Horizontal air-blocking shelf reduces dust leakage from surface drill
shroud. Transactions of the Society of Mining, Metallurgy, and Exploration 324: 55–60.
Potts JD, and Reed WR 2011. Field evaluation of air-blocking shelf for dust control on blasthole drills.
Int J Surf Min Reclamat. Environ 25(1): 32–40.
Reed WR, Listak J, Page S and Organiscak J 2008. Summary of NIOSH research completed on dust
control methods for surface and underground drilling. Transactions of the Society of Mining, Metal­
lurgy, and Exploration 324: 32–40.
Trancossi M 2011. An Overview of Scientific and Technical Literature on Coanda Effect Applied to
Nozzles. SAE Technical Paper: 2011-01-2591.
Zheng Y, Reed WR, Zhou L, and Rider JP 2016. Computational fluid dynamics modeling of a medium-sized
surface mine blasthole drill shroud. Mining Engineering 68(11): 43–49.
Zheng Y, Reed WR, Potts JD, Li M, and Rider JP 2018. Dust control by air-blocking shelves and dust
collector-to-bailing airflow ratios for a surface mine drill shroud. Mining Engineering 70(5): 69–74.
Zheng Y, Reed WR, and Potts JD 2021. Different inner shapes of drill shroud on dust control for surface
mine drilling operation. Tukkaraja, P. (Ed.). 2021. Mine Ventilation: Proceedings of the 18th North
American Mine Ventilation Symposium, 12–17 June 2021, Rapid City, South Dakota, USA (1st ed.).
CRC Press. https://doi.org/10.1201/9781003188476.

246
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Investigation on the effect of water pressure on spray performance

for removal of respirable dust

H. Jiang, S. Klima, T. Beck & Y. Zheng

Centers for Disease Control and Prevention (CDC), National Institute for Occupational Safety and Health
(NIOSH), Pittsburgh, PA, USA

ABSTRACT: Dust control challenges exist in numerous mining applications and is espe­
cially prevalent in confined production spaces like mining faces. Water sprays are the most
economical and technically feasible means of reducing dust concentrations in a wide range of
applications. This study investigated the respirable dust knockdown performance and submi­
cron particle suppression behavior for a typical mining spray nozzle operating at pressures
ranging from 100 to 800 psi. A confined chamber dust removal evaluation approach and an
optical particle counter were used for this study. For respirable dust knockdown, high spray
pressures were more effective in lowering dust concentrations. However, submicron-sized par­
ticles are difficult to remove; at 100-psi water pressure, the knockdown fraction for submicron
particles remained at only 40 % at the end of a 1-hr test. The lowest suppression effective size
boundary was identified from 0.465 to 0.897 µm under a tested pressure range. Resuspension
of submicron material was observed at all water pressure levels after the start of the spray
operation. The resuspended submicron dust remained airborne for different durations under
different spray pressures. Overall, results show a low capture efficiency and a high resuspen­
sion possibility for micron- or nano-scale particles under testing conditions.

1 INTRODUCTION

Dust control problems exist in a variety of industrial sectors, from construction sites to mining
sites (Stacey et al., 2011). It is even more challenging when dealing with respirable dust in
a confined production space, such as the underground mining face. Mining activities can gener­
ate a large amount of dust particles including respirable dust and even submicron dusts (Sarver
et al. 2019). Despite the Mine Safety and Health Administration (MSHA) setting regulations to
improve coal miner health and safety, there has been a dramatic increase in the prevalence of
progressive massive fibrosis among underground coal miners in central Appalachia over the
past decade (Doney et al., 2019). It has long been recognized that the inhalation and deposition
of respirable dust in the lung respiratory tract is the main cause for pneumoconiosis and silicosis
in coal miners (Cecala et al., 2019). The human respiratory tract can be divided into three main
regions—the head, tracheobronchial region, and the gas-exchange or alveolar region (Brown
et al., 2013). Respirable fraction specifically means the mass fraction of inhaled particles pene­
trating into the gas-exchange region. In terms of particle size, the limit for entering the alveolar
region is between 10 and 15 µm (WHO, 1999; European Committee for Standardization, 2014).
These respirable fractions that cause the greatest hazard include quartz and other dusts contain­
ing free crystalline silica produced by grinding or drilling.
Dust mitigation techniques commonly used in the underground mining workplace generally
include three distinct types: (1) ventilation systems to dilute and transport dust away from
workers, (2) isolation of workers by enclosures, and (3) capture of airborne dust using water
sprays. Among these techniques, water sprays are the most economical and technically feasible

DOI: 10.1201/9781003429241-26

247
means of reducing dust concentrations in a wide range of applications (Beck et al., 2018). Sev­
eral water spray applications for controlling longwall face dust have been developed and
adopted in recent years. Drum-mounted, full-cone water sprays are very effective for dust sup­
pression at the point of coal fracture and minimizing dust liberation during coal transport.
Shearer-body-mounted water sprays oriented downwind can promote movement of dust-
laden air along the face side of the shearer to prevent migration into the walkway. There are
also water sprays mounted on the longwall shield to address the dust generated during shield
movement. Underside shield canopy sprays can create a moving water curtain to contain the
dust cloud near the drum area and away from the walkway (Colinet et al. 2010). The atomiza­
tion characteristics and dust removal efficiency of high-pressure sprays were investigated for
optimizing the performance. Cheng et al. (2011) conducted a field test showing that increasing
the longwall shearer and shield spray pressure from 290 to 1,160 psi decreased the dust con­
centration from 536 mg/m3 to 97 mg/m3. This research also recommended the application of
1,160-psi spray pressure for better respirable dust capture efficiency in coal mines. An experi­
mental study concluded that the spray pressure is directly proportional to spray water flow,
while inversely proportional to the water droplet size and spray angle (Han et al., 2020). The
overall dust suppression efficiency for total dust and respirable dust increased with the
increase of pressure from 290 to 1,160 psi.
Dust capture by water droplets can be accomplished by numerous mechanisms including
inertial impaction, interception, diffusion, thermophoresis, and electrostatic charge collection.
These processes usually occur simultaneously, although inertial impaction and interception
are predominant for the capture of coal-mine-generated dusts. The theoretical foundation for
high-pressure suppression performance improvements can be found from the droplet breakup
mode (Majithia et al., 2008), dust suppression efficiency by inertial impaction (Walton and
Woolcock, 1960), and dust suppression efficiency by interception (Fuchs, 1964), illustrated in
equations 1, 2, and 3 respectively:

where

Dw is the diameter of water droplet, m; Wec is the critical Weber number when the droplet
starts to breakup; ρg is the air density, kg/m3; ν is the relative velocity between the dust par­
ticles and the droplets, m/s; σ is the surface tension, N/m; ηimp is the dust suppression efficiency
by inertial impaction; St is Stock number of particle; ηint is the dust suppression efficiency by
interception; Dp is the diameter of dust particle, m; Cf denotes the Cunningham correction
factor; ρp is the density of the dust particles, kg/m3; μ denotes the air dynamic viscosity, Pa·s.
As the spray pressure increases, the water flow rate increases and thereby results in higher
droplet velocity. According to Eq. (1) a higher relative velocity produces smaller water drop­
lets. Eq. (2) and (4) show that a smaller water droplet and larger relative velocity can increase
the dust capture efficiency by inertial impaction. Eq. (3) demonstrates that the interception
efficiency for small dust particulates can be limited but reducing the size of water droplets can
increase the dust interception efficiency. In summary, high-pressure spray generates finer
water droplets and can yield a higher dust suppression efficiency.

248
 There is still skepticism of the water suppression effect for respirable particulates by
researchers (Organiscak and Page 2005). The detailed size-specific behavior within the respir­
able range under suppression was rarely studied. The spray knockdown efficiency for submi­
cron particles was unknown for either low- or high-pressure spray in coal mining. In this
study, the respirable dust suppression characteristics were investigated in a confined chamber
under spray pressure that ranged from 100 to 800 psi. The spray respirable dust knockdown
was monitored and evaluated over a 1-hour testing period. The suppression behavior for sub­
micron particles under different pressures were also investigated.

2 LABORATORY EXPERIMENTAL SYSTEM AND METHODOLOGY

A confined-chamber, dust-removal evaluation approach (McCoy et al., 1985) was used to

investigate the respirable dust suppression effectiveness for a water spray over the pressure
range from 100 to 800 psi. The advantage of this confined-chamber approach compared with
an open-ended duct approach (Han et al., 2020) is that detailed dust particle decay character­
ization in different size ranges can be easily monitored. In addition, dust was only introduced
at the beginning of the test to reach an initial concentration instead of a constant dust source
throughout the test which can create an unstable baseline.
A 512-ft3 dust chamber (8 ft by 8 ft by 8 ft) constructed of plywood was used for testing in
this study, Figure 1. Keystone Mineral Black 325 BA coal dust (Keystone Filler & Mfg. Co.,
Muncy, PA) was introduced through an air inductor into the chamber through a metal pipe
that was grounded to eliminate the risk of imparting any static charge to the particles. A 35-
ft3/s mixing fan located at the upper corner of the chamber ensured the dust was sufficiently
homogenized after injection. A WhirlJet® BDM3 hollow cone spray (Model 3/8-BDM-3,
Spraying Systems, Wheaton, IL) with an orifice of 0.094 inch was mounted at a 58-inch height
on the centerline axis of the chamber; the spray mechanisms provided by the manufacturer are
shown in Figure 2. The water enters the nozzle body through a hole from its side, then the
water spins at high speed in the whirl chamber. The high-speed rotation causes the water to
exit the orifice in a hollow cone pattern. The spray geometry and mechanism are the same
with the brass and stainless steel 3/8 BD3 sprays but manufactured with nylon body material.

Figure 1. Dust chamber system for dust deposition and water spray dust suppression test.

This study is only interested in the respirable particles, over-sized particles in the raw coal
dust were removed by cyclone during the sampling process. In this test, three real-time aerosol
monitors (TSI Dust Trak@ II aerosol monitor model 8530, Shoreview, MN) were connected
to Dorr-Oliver 10-mm nylon cyclones for the purpose of measuring the respirable size fraction
of dust (≤ 10 µm). Sampling points were located 10 inches above the spray. During testing,
one aerosol monitor continuously sampled and recorded the dust concentration, while the

249
other two monitors began sampling and recording when dust concentration reached the start
criteria. The sampling flow rate for each monitor was set to 1.7 L/min, and the data logging
interval was 5 seconds. All reported dust concentrations are the relative mass concentration
measured by an aerosol monitor instrument unless expressed otherwise. One Optical Particle
Sizer (OPS) (TSI model 3330, Shoreview, MN) was used to obtain the particle count results
with a sampling rate of 1 L/min. The OPS operates on the principle of single particle counting.
Particle pulses are counted individually and binned into the 16 channels based on their pulse
heights. Sixteen (16) bin ranges were determined with a refractive index of 1.245 to ensure the
accuracy of the particle count result. A 10:1 aerosol diluter (TSI model 3332, Shoreview, MN)
was installed before the inlet on the OPS to ensure the accuracy of the measurement. It should
be noted that limited water-only testing was conducted to assess the impact of spray water
droplets on OPS particle counting and that the spray droplet had no impact.
Before each test, the chamber was sealed and the spray was operated to wet the interior
walls. Dust was injected into the chamber until the instantaneous concentration reached at
least 100 mg/m3. The spray was tested at designated water pressures and testing was started
once the dust concentration naturally decayed to approximately 100 mg/m3. Water pressures
of 100, 200, 400, 600, and 800 psi were evaluated. At the same time, the aerosol monitors and
particle counter were started and operated continuously throughout the whole test period.
Each test continued for 60 minutes or until the dust concentration reached below 1.0 mg/m3.
A randomized full factorial experimental design with three repetitions for each test condition
was performed.

Figure 2. WhirlJet® BDM3 spray nozzle and working mechanism.

3 ANALYSIS OF THE SPRAY NOZZLE CHARACTERISTICS

The water flow rates and spray angles under different pressures operated in the chamber were
measured and are listed in Table 1. It is obvious as the water pressure increases, the water
volume sprayed out from the nozzle increased, but the spray angle narrowed down. With
a smaller spray angle, a smaller spray coverage (area or volume) in the chamber can be
expected as well. As can be observed from the spray pattern, the plume bottom converged
considerably starting from 400 psi. This convergence started closer to the nozzle tip under
higher pressures. This divergence effect can reduce the actual spray coverage from the spray
angle based on theoretical coverage calculations.

4 RESPIRABLE DUST KNOCKDOWN PERFORMANCE EVALUATION

Figure 3 shows the respirable dust concentration (solid line) and cumulative water consump­
tion (dashed line) results for tests under different pressures over the testing period. The dust
concentration dropped sharply for all pressures within the first 10 minutes, then it slowly
decreased to the ending concentration. Under spray pressure 100 and 200 psi, the final dust
concentration reached 8.70 and 2.77 mg/m3, respectively. While under higher pressure, the
final dust concentrations were able to reach the 1.0 mg/m3 stop criteria, and the time it takes
to reach this criterion for tests with spray pressures of 400, 600, and 800 psi are 30.8, 13.7, and
8.3 minutes, respectively.

250
Table 1. Spray specifications and characteristics of WhirlJet® BDM3 hollow cone spray under different
water pressure.

These results were anticipated because more water was sprayed into the chamber under
higher spray pressure. Evidence can be found from the dashed lines in Figure 3 that the cumu­
lative water consumption for higher pressure is always higher than lower pressure tests. How­
ever, since high-pressure spray is more effective on capturing dust, the total water
consumption to realize the same dust reduction for high pressure is much smaller. For
example, to realize 90 % dust reduction, the total water consumption under 100, 200, 400, 600,
and 800-psi spray pressures is 41.7, 20.1, 10.2, 7.4, and 5.9 gallon, respectively.

Figure 3. Respirable dust concentration for BDM3 under different pressures.

5 SUPPRESSION EFFECTIVENESS EVALUATION FOR SUBMICRON

PARTICULATES

A previous study (Jiang et al., 2022) addressed the issue of low capture efficiency for submi­
cron particulates by water sprays. The results revealed that under 100-psi spray pressure, par­
ticles under 1 µm can remain airborne through the 60-minute test, and particles under 0.5 µm
can be considered irremovable particles under the same condition. Therefore, this section

251
analyzes the submicron particulate suppression performance under different spray pressures.
Since size channels for the particle counter were calculated with a refractive index, the cap size
was not rounded at 1 micron. It should also be noted that each size bin is represented by its
top size range, which means bin 0.374 µm indicates a particulate size in the 0.3 to 0.374 µm
range.
Figure 4 plots the sub-1.117-µm particle counts during the whole testing process under dif­
ferent spray pressures. A particle count peak can be found under each pressure within the first
10 minutes of spray. These peaks were likely caused by resuspension of the residual dust in the
chamber by the spray action. Based on the ending counts from different pressures, it appears
that submicrons can be suppressed by higher spray pressures in a relatively short period of
time. It took around 6, 10, and 24 minutes for spray pressure of 800, 600, and 400 psi to
reduce 80 % of submicrons from their peak. The final reduction percentage of submicrons
under 200 and 100 psi are 69.4 and 41.6 %, respectively. These results agree with findings from
other studies that the wettability of submicrons is lower than larger sizes (Yu et al., 2020).
Considering the short interaction period of spray and particulate in the underground face,
and it can be expected that most submicrons are likely to escape from low-pressure open
sprays and enter the mine atmosphere.

Figure 4. Sub-1.117-µm particle count results under different spray pressures.

In addition to the sub-1.117-µm particle count results, the detailed particle count results for
7 size bins at different time stamps (shown by various colors) under different pressures are
plotted in Figures 5, 6, and 7. The y-axis is the dust counter result percentage at the time com­
pared to its initial count.

Figure 5. Sub-1.117-particle count results under 100-psi spray water pressure (mean ± SD, n=3).

Resuspension of particles were observed after 3 minutes of spray under all pressure levels.
The observed size range for particle resuspension is narrower for higher pressure spray than

252
Figure 6. Sub-1.117-particle count results under 200-psi spray water pressure (mean ± SD, n=3).

lower pressure spray. At 100-psi spray pressure, the size range for resuspension is 0 to 0.721
µm, while the range under 800 psi is below 0.374 µm. These figures also show that when the
peak of particle resuspension is reached at different times after spraying begins, it takes longer
to reach the peak under low-pressure spray—at 100 psi it took almost 30 minutes to fully
resuspend the sub-0.3-µm particles. It should be noted that under a 100-psi and 200-psi spray
pressure condition, there is a substantial amount of dust remaining above the initial level after
a 60-minute test. The knockdown efficiency for that size range under these pressures is so low
that after 60-minute test, the ending particle count is still higher than when the spray begins.
Because the spray water droplet and dust interaction period is less than 3 minutes in the
production face, the size boundary for effective suppression can be identified from the dust
counter results. The lower boundary can be found when the counter results at 3 minutes in
that size bin are lower than the starting level. Using this criterion, the lowest effective suppres­
sion size boundary for this spray under different pressures can be identified as 0.897 µm for
both 100 and 200 psi, 0.721 µm for 400 psi, and 0.465 µm for both 600 and 800 psi.
The above results indicate a low capture efficiency and high resuspension possibility for par­
ticles under micron scale or nano-scale. In recent years, nano-scale dust particles and their
aggregates or agglomerates have obtained significant attention due to their potential influence
on the health of workers. Based on the analysis of coal dust samples collected from several
underground coal mines, the presence of Coal Dust NanoParticles (CD-NPs, smaller than 500
um) have been confirmed (Keles, et al., 2022). Studies have examined and compared the differ­
ences in physicochemical properties, the acute pulmonary toxicity, and the chronic pulmonary

Figure 7. Sub-1.117-particle count results under 800-psi spray water pressure (mean ± SD, n=3).

253
fibrosis-inducing ability of Coal Dust Micron Particles (CD-MPs, smaller than 5 µm) and
CD-NPs as well as their adverse effects of coal dust size on the lung (Zhang, et al., 2022).
Findings indicate that CD-NPs are more chronically toxic than CD-MPs due to a lower
oxygen-to-carbon ratio, smaller size, and larger surface area based on characterization ana­
lysis of particles. Other researchers also conclude that CD-NPs pose a higher risk to human
health and cause a larger surface area with a different toxicology compared to the same mass
concentration of dust of a relatively large size (Fan and Liu, 2021). Though it has been proven
that submicron particles are difficult to capture by water sprays, their flow path can be easily
manipulated by air stream. This means that these particles can be easily isolated from workers
by ventilation design. In this case, there will be advantages to using the air-moving properties
of the spray rather than the suppression ability.

6 CONCLUSION

Laboratory tests were conducted in a 512-ft3 dust chamber with the WhirlJet® BDM3 hollow
cone spray to study its respirable dust removal performance at five different water pressure
levels ranging from 100 to 800 psi. For each test, the respirable dust concentration and particle
count were continuously monitored during the 60-min spray application or when the dust con­
centration dropped below 1.0 mg/m3. Results show that the total water consumption to
achieve the same dust reduction under high pressure is smaller than a low-pressure spray. The
suppression effectiveness of submicron particulates was also evaluated in this study. The final
reduction percentages of submicrons under 100 and 200 psi are 41.6 % and 69.4 %, respect­
ively. The lowest suppression effective size boundary from 100 to 800 psi is between 0.897 to
0.465 µm.
The resuspension of submicron particles was observed under all spray pressure levels, and
the cap size for particle resuspension is lower for a higher-pressure spray than for a lower-
pressure spray. There is a substantial amount of submicron particles that remain airborne
after a 60-minute spray test, especially for the low-pressure condition. Overall, results show
a low capture efficiency and high resuspension possibility for micron- or nano-scale particles
under the testing conditions. Considering the adverse effect of smaller particles on lung func­
tion and the higher risk to human health compared to the same mass concentration of dust
with a relatively larger size, other mitigation methods or ventilation tools should be used to
address this problem.

DISCLAIMER

The findings and conclusions in this report are those of the author(s) and do not necessarily
represent the official position of the National Institute for Occupational Safety and Health
(NIOSH), Centers for Disease Control and Prevention (CDC). Mention of any company or
product does not constitute endorsement by NIOSH.

REFERENCES

Stacey P., Andrew Thorpe, and Paul Roberts. 2011. Levels of respirable dust and respirable crystalline
silica at construction sites, RR878 research report, Health and Safety Executive.
Sarver E, Keles C, Rezaee M. 2019. Characteristics of respirable dust in eight Appalachian coal mines:
A dataset including particle size and mineralogy distributions, and metal and trace element mass
concentrations. Data Brief. 2019;25:104032.
Doney BC, Blackley D, Hale JM. 2019. Respirable coal mine dust in underground mines, United States,
1982–2017. Am J Ind Med. 62(6):478–485.
Cecala A. AD O’Brien, J Schall et al. 2019. Dust control handbook for industrial minerals mining and
processing, Second Edition. National Institute for Occupational Safety and Health, DHHS (NIOSH)
Publication No. 2019-124 (RI 9701)

254
Brown JS, Gordon T, Price O, Asgharian B. 2013. Thoracic and respirable particle definitions for human
health risk assessment. Part Fibre Toxicol. 2013 Apr 10;10:12. doi: 10.1186/1743-8977-10-12.
WHO, OEH. 1999. Hazard prevention and control in the work environment:Airborne dust. Geneva.
European Committee for Standardization. 2014. EN 13205 Part 1–6, Workplace exposure Assessment of
sampler performance for measurement of airborne particle concentrations. Brussels.
Cheng WM, Nie W, Zhou G and Zuo QM. 2011. Study of dust suppression by atomized water from
high-pressure sprays in mines. Journal of China University of Mining and Technology 40(2):185–189.
Han, H., Wang, P., Liu, R. et al. 2020. Experimental study on atomization characteristics and
dust-reduction performance of four common types of pressure nozzles in underground coal mines.
Int J Coal Sci Technol 7, 581–596. doi.org/10.1007/s40789-020-00329-w
Majithia A.K., Hall S., Harper L., Bowen P.J. 2008. Droplet Breakup Quantification and Processes in
Constant and Pulsed Air Flows, Paper ID ILASS08-4-4, Institute for Liquid Atomization and Spray
Systems, Sep. 8–10, Como Lake, Italy.
Walton W. and Woolcock A. 1960. The suppression of airborne dust by water sprays. Int. J. Air. Pollut.
3. 129-1.53.
Fuchs N.A. 1964. The Mechanics of Aerosols, p. 165. Macmillan, New York.
Organiscak JA, Page SJ. 2005. Development of a dust collector inlet hood for enhanced surface mine
drill dust capture. Int J Surf Min Reclam Environ. 19(1):12–28.
McCoy, J.F., Schroeder, W.E., Rajan, S.R., Ruggieri, S.K., and Kissell, F.N. 1985. New Laboratory
Measurement Method for Water Spray Dust Control Effectiveness, American Industrial Hygiene Asso­
ciation Journal, 46:12, pp:735–740, DOI:10.1080/15298668591395652.
Yu H, Wang Y, Liu E, Zhang G, Huang L, Liu L, et al. 2020. Experimental study of the influences of
temperature on the properties of particles in a gasifier during coal-water slurry gasification. ACS
Omega. 2020;5:5698–703.
Keles C, Taborda MJ, Sarver E. 2022. Characteristics of respirable dust in eight Appalachian coal mines:
A dataset including particle size and mineralogy distributions, and metal and trace element mass con­
centrations with expanded data to cover a total of 25 US mines. Data Brief. 2022;42:108125.
Zhang Yinci, Amin Li, Jiafeng Gao et al. 2022. Differences in the characteristics and pulmonary toxicity
of nano- and micron-sized respirable coal dust, Respiratory Research, 23: 197. doi.org/10.1186/s12931-
022-02120-8.
Fan, L., Liu, S. 2021. Respirable nano-particulate generations and their pathogenesis in mining work­
places: A review. Int J Coal Sci Technol 8, 179–198. doi.org/10.1007/s40789-021-00412-w.
Colinet JF, Rider JP, Listak JM, Organiscak JA, Wolfe AL. 2010. Best Practices for Dust Control in
Coal Mining. U.S. Department of Health and Human Services, Centers for Disease Control and Pre­
vention, National Institute for Occupational Safety and Health. 2010 Information Circular 9517,
DHHS (NIOSH) Publication No. 2010–110.
Beck T.W., C.E. Seaman, M.R. Shahan, and S.E. Mischler. 2018. Open-air sprays for capturing and con­
trolling airborne float coal dust on longwall faces, Min Eng., 70(1): 42–48. doi:10.19150/me.7978.
Han, H., Wang, P., Liu, R. et al. 2020. Experimental study on atomization characteristics and
dust-reduction performance of four common types of pressure nozzles in underground coal mines.
Int J Coal Sci Technol 7, 581–596. doi.org/10.1007/s40789-020-00329-w.
Jiang H., . Zheng Y, Klima S., Seaman C.E. and Beck T.W. 2022. A Laboratory Study of the Dust
Deposition and Suppression Process for Respirable Coal Dust in a Confined Chamber, Preprint for
SME annual conference, Feb 27- Mar 02, Salt Lake City, UT.

255
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Real-time measurements of respirable crystalline silica, kaolinite,

coal, and calcite

William Arnott
Atmospheric Science Program, Department of Physics, University of Nevada, Reno

Charles Kocsis
Mining Engineering Department, University of Utah, Salt Lake City

Xiaoliang Wang
Desert Research Institute, Reno, Nevada

Bankole Osho & Pedro Nascimento

Department of Mining and Metallurgical Engineering, University of Nevada, Reno

Samuel Taylor, Bjoern Bingham, Chauntelle Murphy & Michael Sandink

Atmospheric Science Program, Department of Physics, University of Nevada, Reno

ABSTRACT: Real-time measurements of coal mine speciated dust mass concentration

(SDMC) are needed for health and ventilation applications. The current method requires captur­
ing aerosol on filters by vacuum pump and Fourier Transform Infrared Spectroscopic (FTIR)
analysis, providing end-of-work-shift measurements. We are developing a photoacoustic spec­
trometer (PAS) equipped with a tunable quantum cascade laser (QCL) to measure SDMC in
real-time. The QCL is sequentially tuned to four unique wavelengths for each compound to
quantify the SDMC of respirable crystalline silica (RCS), kaolinite, coal, and calcite. Each dust
type also absorbs a little laser light at the other three wavelengths. A matrix of measurements
developed quantifies the relationship between SDMC and aerosol light absorption at the four
wavelengths. The inverse of this matrix provides SDMC from PAS measurements for a mixture
containing all four dust types. Individual and mixed dust measurements demonstrate the current
state of the art in PAS sampling of SDMC.

1 INTRODUCTION

Respirable (Vincent, 2005) coal dust is produced when coal is extracted from the seam by drill­
ing, blasting, cutting, loading and transporting operations. In underground mines it consists
of aerosol with diameters (0.01 - 20 μm) of coal, silica, kaolinite, and other minerals which can
be inhaled and deposited into the lungs, to a worker’s alveoli where oxygen enters the blood
stream. This is a serious health concern because in this region of the lungs, the human body
has few means of removing hazardous particles. As a result, respirable mine dust, which
includes coal as well as carbonate (i.e. rock dust (Pokhrel, Keles, Jaramillo, Agioutanti, &
Sarver, 2021)) and non-carbonate material (i.e. RCS) can generate serious health hazards if
these elevated dust concentrations are not adequately controlled (Kachuri et al., 2014).
The long-term exposure of coal miners to high levels of respirable coal dust can cause
pneumoconiosis, emphysema, and other occupational lung diseases such as the chronic
obstructive pulmonary disease, which is collectively known as black lung disease. Overexpos­
ure to respirable silica dust can lead to silicosis. Short-term, but elevated respirable silica dust

DOI: 10.1201/9781003429241-27

256
concentrations are considerably more hazardous than the same level of cumulative exposure
at lower concentrations (e.g. recommend10-hour average exposure to RCS of < 0.05 mg m-3)
(Howard, 2011). Consequently, it is important to design, manufacture, and commercialize
a new personal real-time monitoring instrument, which has the ability to measure and monitor
both respirable coal dust and respirable silica dust concentrations in the production workings
and throughout the mine.
Our goal is to develop, fabricate and test a new real-time personal coal and silica dust monitor­
ing instrument, based on photoacoustic absorption spectroscopy, as opposed to time averaged
filter-based design principles such as end-of-shift FTIR analysis (Pampena, Cauda, Chubb, &
Meadows, 2020) or quantum cascade laser filter analysis (Wei, Kulkarni, Ashley, & Zheng, 2017)
or aerosol light scattering that is not sensitive to aerosol composition (Zhang, Nie, Liang, Chen,
& Peng, 2021). The instrument will have the ability to continuously measure respirable coal dust
and respirable silica dust concentrations in the production areas and throughout the mine with
no user intervention and minimal equipment maintenance needs. As opposed to the current
filter-based technology, the new instrument is based on photoacoustic principles; the effects of
the absorbed laser light by the carbon and silica particles are measured with an acoustic detection
system. Details are in the following section. The photoacoustic measurement theory has been
developed (Taylor, Nascimento, Arnott, & Kocsis, 2022) and experimental demonstration in the
laboratory has occurred for dust containing coal, RCS, and kaolinite (Nascimento et al., 2022).
This manuscript extends our previous work to include calcite dust (i.e. rock dust), and to
give the theory and results for measurements of the SDMC that are needed for autonomous
instrument operation.

2 METHODS

Laboratory measurements of SDMC are obtained using our Marple-like (Marple & Rubow,
1983) dust resuspension chamber shown schematically in Figure 1 (Nascimento et al., 2022).

Figure 1. Schematic of the aerosol resuspension chamber used for measuring SDMC and evaluating
and calibrating the photoacoustic instrument. The chamber is 2 meters long and the plexiglass window is
centered. The current PA instrument is on a 2ʹ x 4ʹ optical table and the instrument enclosure fits in a 19”
equipment rack.

Dust is dispensed from a container, measured at ¼ teaspoon, and placed in the metal bowl.
A five second duration burst of compressed air suspends the dust. The dust mixes to
a uniform state in about six minutes, its concentration diminishing slowly over about a three-
hour interval. The inlets for each instrument are close together and are approximately 20 cm
above the chamber floor.
The photoacoustic instrument measures aerosol light absorption by means of an acoustic detec­
tion system. A schematic of the instrument is shown in Figure 2, taken from (Nascimento et al.,
2022) where details may be found. Briefly, sample air enters the inlet and is drawn through the
instrument with a vacuum pump. The aerosols interact with laser light in the resonator section,

257
Figure 2. Schematic of the photoacoustic instrument.

creating the sound that is measured to quantify light absorption. Aerosol heats during light absorp­
tion and transfers heat to the surrounding air, thereby creating a very small pressure wave, sound,
at the resonant frequency of the chamber since the laser power is modulated at this frequency
(Arnott, Moosmuller, Rogers, Jin, & Bruch, 1999), then detected by a microphone. The instrument
gains spectroscopic capability by use of a tunable quantum cascade laser (QCL), as the wavelength
can be changed quickly. Laser power is measured with a mercury cadmium telluride (MCT)
detector salvaged from a Fourier Transform Infrared Spectrometer (FTIR). The piezoelectric trans­
ducer is used to measure the resonance frequency, quality factor, and microphone calibration at
resonance. Measurements are reported for both 1 and 16 second integration time. Noise equivalent
light absorption measurements are obtained by integrating the microphone signal away from reson­
ance (Lewis, Arnott, Moosmuller, & Wold, 2008).
The Aerodynamic Particle Sizer Spectrometer (APS, model 3321, TSI Incorporated, Shore­
view, MN, USA) instrument optically detects the time of flight of accelerated aerosol to
obtain aerodynamic diameter for individual aerosol. A 20:1 dilution ratio for sample air
versus clean air was achieved using the TSI Aerosol Diluter model 3302A. The mass distribu­
tion for laboratory generated RCS is shown in Figure 3. Distributions for other dust types
were similar. Most of the aerosol mass is below 4 μm, especially after the middle of the experi­
ment, as the larger particles settle. The mass distributions are integrated to obtain real-time
respirable dust mass concentrations. Because the user must input the dust density for mass
concentration measurements, the APS is best suited for experiments involving homogeneous
dust. Average mass distribution near coal mining fronts have peak diameters of 3-6 μm, and

Figure 3. Mass distribution measurements from the APS for silica. Curves are for the indicated times.
The vertical line is at diameter of 4 μm.

258
have more coarse mode aerosol than our laboratory aerosol, indicating a need to test our
instruments in mines (Skubacz, Wojtecki, & Urban, 2016).
The SPS30 is a low cost air quality sensor manufactured by Sensirion (Sensirion, 2020). It is an
optical particle counter that reports aerosol number concentrations for 4 size ranges in addition to
PM1, PM2.5, PM4, and PM10. The recommended mass concentration range of PM2.5 is up to
1 mg m-3. We have found that SPS30 PM4 values correlate well with gravimetric mass measure­
ments, even with concentrations up to around 50 mg m-3. Laboratory evaluation showed that the
SPS30 over (under) reported PM2.5 for woodsmoke (Arizona road dust) compared with filter
measurements, and that the SPS30 operated properly after prolonged sampling at concentrations
as high as 33 mg m-3 (Tryner, Mehaffy, Miller-Lionberg, & Volckens, 2020).
Gravimetric and FTIR sampling were performed using PVC filters to obtain total respirable
mass and silica concentrations using the NIOSH FAST technique. SKC aluminum cyclones (Zefon
International, Catalog NO. ZA0060) selected respirable dust that was deposited through use of per­
sonal sampling cassettes (Zefon International, REF 745PVC-CF-FTIR) on 37 mm PVC filters
(Zefon International, REF FSP37R). A vacuum pump external to the aerosol chamber was con­
nected to the filter samplers with tubing, and operated nominally at 2.5 LPM as measured with
a flow meter. The cyclone sampler operates at 2.5 LPM to obtain respirable dust. The exposed area
of the filter was 6.92 cm2. A custom filter holder was developed for FTIR measurements.

3 RESULTS

As previously mentioned, laboratory dust simulants may have different composition and aerosol
size than dust found in mining environments. Many grades and type of coal are found in nature,
though all absorb light throughout the 11-13 μm measurement range. Pre-dispersed silica dust can
coagulate into larger, spherical clumps, perhaps due to static generated by the plastic storage con­
tainer. Our research has identified key wavelengths to use to measure absorption by four coal-mine
-relevant dust species. We have compared the SPS30 and APS measurements of total dust mass
concentration with gravimetric measurements, and have developed a method to calculate SDMC
from multispectral photoacoustic aerosol light absorption measurements.
FTIR analysis of individual filters collected during dust sampling procedures confirms the
use of unique wavelengths for each components’ absorption spectrum to obtain SDMC. The
PAS can also be used to find the absorption spectrum of any sampled air; however, FTIR
analysis is used as the industry’s standard method to compare the accuracy of the PAS, as

Figure 4. FTIR analysis of dust mixture and individual dusts, demonstrating the absorption spectra for
each component, and how they come together to form the absorption spectra for the mixture. Also illus­
trated are laser lines for the wavelengths best suited for quantifying dust constituents.

259
well to provide supporting evidence for the wavelengths chosen to quantify SDMC. Figure 4
shows FTIR absorption spectra obtained as described in (Nascimento et al., 2022) of the indi­
vidual dust types (kaolinite, calcite, coal and silica) along with the dust mixture of all four
dust types in equal parts. This figure nicely illustrates the dust mixture’s absorption as
a combination of its constituents, represented by “Mixture PVC9 Babs.”
Figure 4 also illustrates the wavelengths at which kaolinite, calcite, coal and silica absorb
strongly, with the least amount of interference from water vapor and other minerals, and were care­
fully chosen to represent each mineral. Each dust type also absorbs a little laser light at the other
three wavelengths. The wavelengths that will most effectively represent and quantify each mineral
in question are 11.040 μm for kaolinite, 11.408 μm for calcite, 11.826 μm for coal, and 12.495 μm
for silica. These wavelengths will be used in the further development of the PAS control software to
automatically obtain concentrations of each mineral. The process of obtaining SDMC through ana­
lysis is elaborated on in Section 4. Strong light absorption by water vapor at 12.53 μm, for example,
can be used along with relative humidity measurements to periodically evaluate the instrument cali­
bration (Nascimento et al., 2022; Taylor et al., 2022).

Figure 5. Photoacoustic measurements of kaolinite alongside PM4 concentrations measured by the

APS and SPS30, demonstrating the absorption of laser light by kaolinite at the designated wavelength,
11.04 μm, compared to absorption at the other minerals’ designated wavelengths.

Measurements of aerosolized kaolinite in the dust chamber are shown in Figure 5 to further
demonstrate the absorption theory for SDMC. We can see how kaolinite has the greatest
absorption at the ‘Kaolinite λ’=11.04 μm, the wavelength chosen to quantify it. We can also
see how kaolinite absorbs at the other wavelengths designated for calcite, silica and coal. The
behavior of the dust as it is aerosolized can also be observed in this figure, from 14:27 for
about 60 seconds. Once the dust is aerosolized, there is a large spike in absorption measure­
ments from the PAS as well as PM4 measurements from the other instruments. Immediately
afterwards, there is a sharp dip, representative of the dust plume as it is initially lifted, and
then rises out of reach of the sensors and vacuum pumps in the chamber. For this reason, to
more accurately represent mine conditions, data analysis and filter sampling doesn’t com­
mence until about six minutes after aerosolization. Data analysis for determining SDMC can
be normalized using either data from the SPS30 or the APS. The APS is the standard for
quantifying particulate matter concentrations however, our findings suggest the SPS30 might
be a more suitable instrument for this purpose.
Gravimetric and optically determined aerosol mass concentrations by the SPS30 and APS
are shown in Figure 6 for all four dust types. APS data like that shown in Figure 3 are inte­
grated over diameter until 4 μm, and the SPS30 directly reports PM4. “PM4” is used to repre­
sent respirable gravimetric aerosol mass concentration. Figure 3 shows that most of the

260
Figure 6. Scatter plot of filter mass concentration graphed against PM4 measurements for both the
SPS30 and APS for all four dust types measured separately. The SPS30 has a much better fit to the gravi­
metric mass concentration, with an R-squared value of 0.92.

aerosol mass diameter is below 4 μm for our laboratory generated aerosol. The SPS30 data
most closely resembles the filter mass, with a coefficient of determination of around 0.92, com­
pared to the APS data, with a coefficient of determination of 0.55. Both however also have
a large offset. These findings support the use of the SPS30 PM data for use in the matrix inver­
sion calculations.

4 SPECIATED DUST MASS FROM PHOTOACOUSTIC MEASUREMENTS

There are two main equations that allow for the PAS to measure aerosol concentrations from
light absorption. The first is an expression for aerosol light absorption, Babs. It is given by

where Pm is the acoustic pressure measured with the microphone, PL is the laser power used,
about 5 mW in our case, Ares is the cross-sectional area of the resonator, γ=1.4 is the is the
ratio between isobaric and isochoric dry air specific heats, f0 is the resonance frequency of the
chamber, about 1500Hz, and Q is the quality factor, which is about 60. Babs has units of Mm–1,
or inverse megameters. All of these values are either controlled by experimental parameters or
measured with the PAS. The other equation vital to the success of our method is given by

where PM has units of [μg m–3] and MAE, or mass absorption efficiency, has units of [m2g–1].
In determining the SDMC, which here are silica, kaolinite, black coal and as a new addition
calcite, we must turn PM and Babs into vectors, with dimensions defined by the number of
species, which would subsequently turn MAE into a square matrix also with dimensions
defined by species count. For this case, there are four species present, corresponding to a 4x4
MAE matrix, and 4x1 PM and Babs vectors. The MAE matrix is separated into both species
and wavelength used to measure a species to be able to quantify contributions from other

261
species at wavelengths where a given species was strongly absorbing. For clarity, below is
a more detailed look at Equation 2:

Of course, the end goal here is measuring PM for different species. So, multiplying the left
of both equations with the inverse of the MAE matrix gives MAE 1 � Babs ¼ PM. The inver­
sion of the matrix is done after all elements are obtained.
To inspect the relationship between PM and Babs, values of SPS30 PM4, SPS30 PM2.5, and
APS PM4 were plotted with Babs measurements as time series on separate axes for a given species
at a given wavelength. To obtain MAE matrix elements, PM values were plotted against Babs
value as a scatter plot, and a linear regression was performed. For four dust species, each meas­
ured at four separate wavelengths, there are 16 total charts like this for each PM type. The slopes
of the regressions for PM2.5 were used to obtain MAE matrix elements. The completed inverse
MAE matrix, along with the example of all species combined together, is included as Equation 4.

The inverse MAE matrix, Equation 4, has units of [g m–2]. This matrix uses SPS30 PM2.5
values instead of the APS PM4 because one: SPS30 has proven to function with high-fidelity
compared to the APS, as shown in Figure 6; and two: PM2.5 appears to have more accur­
acy in its prediction than does PM4. This is due to a more steady and consistent flow of
PM2.5 for the PAS. These Babs entries were calculated by fitting the PM2.5 time series for
the filter sample of the dust mixture, containing equal parts by volume of coal, kaolinite,
calcite and silica, to a second-order polynomial, applying this curve fit to Babs to renormalize
Babs such that there is no drop in PM for a given time interval (30 minutes in this case),
mimicking measuring all four species at once with four different lasers, instead of in
sequence. It will be necessary to perform the measurements sequentially at each wavelength
as quickly as possible (1-16 seconds per measurement) for a fully automated PAS, and to
include a SPS30 to normalize the PM measurements.

5 CONCLUSION

We have developed a method for obtaining speciated dust mass concentration (SDMC)
through mass absorption efficiency matrix inversion, found unique wavelengths at which to
quantify absorption spectra of kaolinite, calcite, coal and silica dusts and provided evidence to
support the use of SPS30 measurements in lieu of APS data. This was achieved with the use of
the real-time personal coal and silica dust monitoring instrument, based on photoacoustic
absorption spectroscopy. The procedure for obtaining SDMC was developed through use of
measurements with the laser manually set to the four wavelengths used for each dust type.
The current instrument is mounted on a 2ʹ x 4ʹ optical table and the instrument enclosure
fits in a 19” equipment rack. Future development of the photoacoustic spectrometer involves
refinement and further testing in both laboratory and mine settings. The instrument will have
the ability to continuously measure respirable coal and silica dust concentrations in the pro­
duction areas and throughout the mine with no user intervention and minimal equipment

262
maintenance needs. Areas in need of further development include: automatic calibration with
relative humidity, measurement of speciated dust movement in-flight with rapid (1-16 second)
measurements at the four laser wavelength, reducing the size of the acoustical resonator,
replacing the liquid-nitrogen-cooled integrating sphere laser power detector with a smaller
device—all to reduce instrument size and to improve commercial viability.
The relevance and importance of this research cannot be stressed enough—the health and
safety of the men and women procuring natural resources, the very backbone of the develop­
ing and developed world, depends on our ability to monitor the hazards they are exposed to
during a normal work day. Without the health and safety of our miners, the industry risks
productivity and financial sustainability, not to mention loss of life.

REFERENCES

Arnott, W. P., Moosmuller, H., Rogers, C. F., Jin, T. F., & Bruch, R. (1999). Photoacoustic spectrometer
for measuring light absorption by aerosol: instrument description. Atmospheric Environment, 33(17),
2845–2852.
Howard, John. (2011). Coal Mine Dust Exposures and Associated Health Outcomes: A Review of Informa­
tion Published Since 1995. Retrieved from https://arlweb.msha.gov/s&hinfo/blacklung/2011-
172NIOSH.pdf
Kachuri, L., Villeneuve, P. J., Parent, M. E., Johnson, K. C., Harris, S. A., & Canadian Canc, Registries.
(2014). Occupational exposure to crystalline silica and the risk of lung cancer in Canadian men. Inter­
national Journal of Cancer, 135(1), 138–148. doi:10.1002/ijc.28629
Lewis, K., Arnott, W. P., Moosmuller, H., & Wold, C. E. (2008). Strong spectral variation of biomass
smoke light absorption and single scattering albedo observed with a novel dual-wavelength photoa­
coustic instrument. Journal of Geophysical Research-Atmospheres, 113(D16), 14. doi:10.1029/
2007jd009699
Marple, V. A., & Rubow, K. L. (1983). AN AEROSOL CHAMBER FOR INSTRUMENT EVALU­
ATION AND CALIBRATION. American Industrial Hygiene Association Journal, 44(5), 361–367.
doi:10.1080/15298668391404978
Nascimento, P., Taylor, S. J., Arnott, W. P., Kocsis, K. C., Wang, X. L., & Firouzkouhi, H. (2022).
Development of a Real-Time Respirable Coal Dust and Silica Dust Monitoring Instrument Based on
Photoacoustic Spectroscopy. Mining, Metallurgy & Exploration, 9. doi:https://doi.org/10.1007/s42461-
022-00653-6
Pampena, J. D., Cauda, E. G., Chubb, L. G., & Meadows, J. J. (2020). Use of the Field-Based Silica
Monitoring Technique in a Coal Mine: A Case Study. Mining Metallurgy & Exploration, 37(2),
717–726. doi:10.1007/s42461-019-00161-0
Pokhrel, Nishan, Keles, Cigdem, Jaramillo, Lizeth, Agioutanti, Eleftheria, & Sarver, Emily. (2021).
Direct-on-Filter FTIR Spectroscopy to Estimate Calcite as A Proxy for Limestone ‘Rock Dust’ in
Respirable Coal Mine Dust Samples. Minerals, 11(9), 922.
Sensirion. (2020). SPS30 Specification Statement. In: Sensirion.
Skubacz, K., Wojtecki, L., & Urban, P. (2016). The influence of particle size distribution on dose conver­
sion factors for radon progeny in the underground excavations of hard coal mine. Journal of Environ­
mental Radioactivity, 162, 68–79. doi:10.1016/j.jenvrad.2016.05.020
Taylor, S. J., Nascimento, P., Arnott, W. P., & Kocsis, C. (2022). Real-time Photoacoustic Measure­
ments of the Mass Concentration of Respirable Crystal Silica Dust: Theory. Mining Metallurgy &
Exploration, 10. doi:10.1007/s42461-022-00657-2
Tryner, J., Mehaffy, J., Miller-Lionberg, D., & Volckens, J. (2020). Effects of aerosol type and simulated
aging on performance of low-cost PM sensors. Journal of Aerosol Science, 150, 17. doi:10.1016/j.
jaerosci.2020.105654
Vincent, J. H. (2005). Health-related aerosol measurement: a review of existing sampling criteria and pro­
posals for new ones. Journal of Environmental Monitoring, 7(11), 1037–1053. doi:10.1039/b509617k
Wei, S. J., Kulkarni, P., Ashley, K., & Zheng, L. N. (2017). Measurement of Crystalline Silica Aerosol
Using Quantum Cascade Laser-Based Infrared Spectroscopy. Scientific Reports, 7, 8. doi:10.1038/
s41598-017-14363-3
Zhang, H., Nie, W., Liang, Y., Chen, J. G., & Peng, H. T. (2021). Development and performance detec­
tion of higher precision optical sensor for coal dust concentration measurement based on Mie scatter­
ing theory. Optics and Lasers in Engineering, 144, 15. doi:10.1016/j.optlaseng.2021.106642

263
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Comparing respirable dust characteristics from full-scale cutting

tests of three rock samples with conical picks at three stages of wear

S. Slouka
Colorado School of Mines, Golden, CO, USA

E. Sidrow
University of British Columbia, Vancouver, Canada

C. Tsai
University of California Los Angeles, Los Angeles, CA, USA

J. Brune
Colorado School of Mines, Golden, CO, USA

ABSTRACT: In environments where excavation is used to break rock, airborne rock dust
poses respiratory health threats to workers. This includes mining with mechanical cutting
machines, which are common in many mines and tunneling operations. The purpose of this
study is to compare characteristics of rock dust generated by pick cutters at different wear
conditions such that results could be used in the evaluation of proper bit management and
dust suppression controls in the underground environment. The study included full-scale
laboratory cutting of concrete, limestone, and sandstone samples. Each rock was cut with
three conical picks each at new, moderately worn, and fully worn stages of wear. Dust samples
were analyzed to reveal the concentrations, mineralogy, particle size distributions, and particle
shapes. The results show that for all rock types, the worn picks consistently generated the
highest concentration of dust, all picks consistently contained quartz, and all three picks con­
sistently generated dust particles of similar shapes. Additionally, although there are slight
shifts in the particle size distributions between the various pick wears, all the particle sizes
reside in the respirable range of less than 10µm in aerodynamic diameter for all the rock types
cut in this experiment.

1 INTRODUCTION

Exposure to airborne rock dust particles containing silica, coal, or other minerals, can cause
irreversible diseases such as coal worker’s pneumoconiosis (Department of Health and
Human Services, 2011), silicosis (Ross & Murray, 2004), and other lung complications (Pope
III et al., 2002; Stansbury, 2018). However, even with the regulations implemented by the US
Mine Safety and Health Administration (MSHA) to limit the concentration exposure and
exposure to silica, there has been an unfortunate increase in lung disease cases in the United
States since the 1990s with a continued rise in numbers in recent years (Blackley et al., 2016;
Doney et al., 2019; Hall et al., 2019; Lee et al., 2014; Zosky et al., 2016).
It is not fully understood why modern miners’ lung diseases are increasing with the MSHA
concentration and silica exposure limits in place (Centers of Disease Control and Prevention,
1974; MSHA, 1977). Therefore, various current research investigations are analyzing the pres­
ence of nanoparticles, particle size distributions, mineralization analysis, and how particle

DOI: 10.1201/9781003429241-28

264
shape affects deposition into the human lung (Fan & Liu, 2021; Khattak et al., 2019; Lab­
ranche et al., 2021; Sarver et al., 2019; Zellnitz et al., 2019). Additionally, there is limited
understanding and implementation of dust suppression management in relation to the pick
cutter life.
With the most dominant machines used for soft and moderate strength rock excavation
being roadheaders, continuous miners, or drum shearers, the primary rock cutting tools are
conical picks. Rock dust is created and released at the interface between the conical pick and
rock surface where some of the deposited dust generated can be transported and re-
introduced into the air during the muck handling process. Conical picks create dust at the car­
bide tip because the concentrated loads they apply penetrate the rock surface and fragment
the rock into smaller pieces and dust particles. The geometry of the pick tip will change over
time as they are used because they wear down (Plis et al., 1988; W. Roepke et al., 1976), but it
is uncertain if the changing of the pick tip changes dust characteristics. The limited evidence
connecting pick wear to dust characteristics is that coal dust concentration increases with the
wear of pick tips (Fowell & Ochei, 1984; Hanson & Roepke, 1979; Qayyum, 2003;
W. W. Roepke & Hanson, 1983; Zhou et al., 2020). There are limited further investigations on
other rock types, nor is there a clear quantitative relationship of the concentration of dust gen­
erated to the geometry of the pick tip.
Therefore, the following research investigates the concentration, mineral composition, par­
ticle size distributions, and particle shapes of suspended and deposited rock dust generated
from three different conical pick wears during laboratory full scale cutting of a concrete, lime­
stone, and sandstone rock block. The conical pick is used for cutting because these picks are
regularly implemented in excavation operations of concrete, limestone, and sandstone in
a variety of applications (Roxborough et al., 1981; Shao, 2016; Su & Akkaş, 2020; Talbo &
Sevigné, 1986). The purpose of the work is to compare the dust particle characteristics gener­
ated from rock cutting with three pick tip geometries as it wears out during operation. This
can eventually guide operators to strategize bit management in relation to production rate
and machine utilization, with consideration of dust suppression measures when cutting soft to
medium-strength rock.
The overarching conclusions reveal that as the pick wear increases, the suspended respirable
dust concentration increases in all rock types. Independent of pick wear, all three picks in all
the rock types generated dust containing silica. Additionally, there are slight trends acknow­
ledged between the pick wear and the particle size distributions for all the rock types, however,
all particle size distributions reside in the respirable size range. Lastly, all the picks generated
suspended respirable particles with similar particle shapes in all rock types in terms of round­
ness, aspect ratio, and roughness measures.

2 RESEARCH METHODS

For full-scale cutting tests, a fine-grained concrete, Indiana limestone, and sandstone rock
block were used with the Linear Cutting Machine (LCM) at the Earth Mechanics Institute
(EMI) at the Colorado School of Mines. Each of the three samples share similar rock strength
properties and provide uniform and homogenous material throughout the specimens as seen
in Figure 1. For example, none of the samples contained any apparent joint sets or discontinu­
ities in structure, which made cutting as consistent as possible. The rock strength properties
measured in the rock mechanics lab at EMI include unconfined compressive strength (UCS),
Brazilian tensile strength (BTS), Cerchar abrasivity index (CAI), and the punch penetration
energy slope index. These properties determined for each sample are given in Table 1.
The cutting methodology, dust collection setup, pick wear quantification, and particle char­
acterization parameters are detailed in a previous study of the limestone block experiments
(Slouka, Brune, Rostami, et al., 2022). The same testing equipment and parameters are used
for all three rock samples. Therefore, when cutting rock samples, 0.76 cm (0.3 in) penetration
and 3.81 cm (1.5 in) spacings at 250 mm/sec (10 in/sec) are used on all samples. When collect­
ing dust, the same Dorr-Oliver cyclones, Tsai Diffusion Samplers (TDS), and vacuums are

265
Figure 1. Concrete (left), limestone (center), and sandstone (right) samples cast into concrete and cut
with multiple lines of cuts along the surface.

utilized to collect suspended and deposited dust. When quantifying pick wear, the same U92
4.0 conical picks were used with the new, moderately worn, and fully worn pick tip radii at
1.8 mm (0.07 in), 3.7 mm (0.15 in), and 5.2 mm (0.2 in), respectively. The concentration, min­
eralogy, particle size distribution, and particle shape characteristics are analyzed with the
same equipment and methods, such as utilizing the NIOSH manual of analytical methods
(NMAM), field emission scanning electron microscope (FE-SEM), and laser diffraction.

Table 1. Rock properties of the three samples tested.

UCS BTS Punch Pen. Energy slope index
Rock-type (MPa) (MPa) CAI (kN/mm)

Concrete 34.2 3.1 1.13 6.0

Limestone 43.4 3.0 0.57 9.4
Sandstone 80.3 6.4 1.7 13.9

3 RESULTS

3.1 NIOSH manual of analytical methods

The NIOSH analytical method 0600 was used to determine the concentration of suspended
airborne respirable dust generated from each of the pick wears when cutting each of the rock
types. Figure 2 reveals the concentration values obtained for each scenario. Additionally, the
NIOSH manual of analytical methods 7500 was used to determine traces of silica present in
the suspended respirable dust. Figure 3 reveals the percent silica determined in each dust
sample collected during tests for the three pick wears when cutting each of the rock types.
A two-way ANOVA statistical test was conducted using MATLAB® to test if rock type and
pick wear affect dust concentration and if there is a meaningful interaction effect between rock
type and pick wear. The p-value for rock type was 0.0128, the p-value for pick wear was 0.0064,
and the p-value for their interaction was 0.7023. The p-value threshold for this analysis is 0.05,
where values below 0.05 correspond to sufficient evidence that the output variable is affected by
the input variable. Therefore, the research team concludes that pick wear and rock type affect
the concentration of dust generated while cutting. However, there is weak statistical evidence
that the rock type and the pick wear interact to affect the dust concentration.
A two-way ANOVA statistical test was performed to test if rock type and pick wear affect
silica percent generated and if there is a meaningful interaction effect between rock type and
pick wear. The p-value for rock type was 6.7E-10, the p-value for pick wear was 0.86, and the
p-value for their interaction was 0.91. The p-value threshold for this analysis was also 0.05,
where values below 0.05 correspond to sufficient evidence that the output variable is affected
by the input variable. Therefore, statistical evidence suggests that the rock type affects
the percent silica generated in dust, but that the pick wear does not influence the percent silica

266
Figure 2. Dust concentrations of suspended respirable dust generated with three pick wears and deter­
mined from NMAM 0600 method. A) Concrete. B) Limestone. C) Sandstone. Note that the scale of the
y-axis changes between sub-figures.

generated in dust while cutting. Additionally, there is weak statistical evidence that the rock
type and the pick wear interact to affect the percent silica generated within the dust.

Figure 3. Percent silica determined in the suspended respirable dust generated with three pick wears
and determined from NMAM 7500 method. A) Concrete. B) Limestone. C) Sandstone. Note that the
scale of the y-axis changes between sub-figures.

267
3.2 Particle size distributions
Laser diffraction was used to determine the particle size distribution for the suspended and
deposited particles, as outlined in previous studies. The “percent channel” used for analysis
results refers to the bins during particle counting, where each particle is individually analyzed
and placed into bins dependent on their size. The physical diameters obtained were then cor­
rected to aerodynamic diameters using the Cunningham correction factor and slip correction
factor (Hinds, 1999). Additionally, each colored curve displayed in Figures 4 and 5 is the aver­
age of three duplicate sample runs.
Figure 4 reveals the suspended particle size distributions for the three pick wears generated
from the three rock types. Visually analyzing the graphs, the particle size single-mode distribu­
tions generated from each pick wear in every rock type lay close to one another without any
shifts or trends, which is expected when cutting rock materials. However, the sandstone gener­
ated slightly more particles in smaller sizes compared visually to the concrete and limestone
samples. Therefore, Kolmogorov-Smirnov (KS) statistical tests were performed between par­
ticle size distributions generated by the three pick wears to test the null hypothesis that the
airborne distributions of particle sizes generated from each pick are the same. A significance
level of 0.05 was used in the statistical analysis. Therefore, p-values below 0.05 provide strong
statistical evidence to reject the null hypothesis. The results suggest that the particle size distri­
butions are statistically different between different pick wears with the trend confirmed for
each rock type cut. For example, all p-values determined between particle size distributions
generated from the three different pick wears were 1.5E-10 or less.
Additionally, Figure 5 reveals the deposited particle size distributions for the three pick
wears generated from the three rock types. For concrete, limestone, and sandstone, the aver­
age values for each distribution increase with the increase in wear. For example, concrete aver­
age deposited particle aerodynamic diameter from the new, moderately worn, and fully worn
pick was 2.47, 2.74, and 2.80 respectively with standard deviations between 1.9 and 2.2 for
each of the pick wears.
KS statistical tests were also performed between the deposited particle size distributions
generated by the three pick wears. Of the about 10,000 particles analyzed for each pick wear
of each rock, the recorded sizes were input into the KS test to test the null hypothesis that the
distributions of particle sizes generated from each pick are the same. A significance level of
0.05 was used in the statistical analysis. Therefore, p-values below 0.05 provide strong statis­
tical evidence to reject the null hypothesis. The results suggest that the particle size distribu­
tions are statistically different between size distributions generated by different pick wears.
This trend is confirmed for each rock type cut. For example, all p-values determined between
particle size distributions generated from the three different pick wears were 9.1E-27 or less.

3.3 Particle shapes

A field emission scanning electron microscope was used to obtain images of the dust particles
as seen in Figure 6, where the images were then processed through a program to calculate the
particle shapes in terms of roundness, aspect ratio, and roughness. Details of image capture,
analysis, and shape calculations are provided in prior reports of experiments on limestone and
igneous rock cutting (Slouka, Brune, & Rostami, 2022; Slouka, Brune, Rostami, et al., 2022).
Figure 7 reveals the average roughness, aspect ratio, and roughness values of the dusts gener­
ated from cutting the various rock types with the three stages of pick wear and the results
from two-way ANOVA statistical tests.
A two-way ANOVA test was conducted to test if rock type and pick wear affect dust par­
ticle shapes. The p-value threshold for this analysis is 0.05, where values below 0.05 corres­
pond to sufficient evidence that the output variable of particle shape is affected by the input
variable. The results of the two-way ANOVA statistical tests between the datasets provide
weak evidence that the particle shapes differ between pick wears. In other words, with all the
pick wear p-values above the 0.05 threshold as seen in Figure 7, the statistical test fails to pro­
vide strong evidence that the pick wears effect the roundness, aspect ratios, or roughness

268
Figure 4. Suspended particle size distributions generated from three different pick wears. A) Concrete.
B) Limestone. C) Sandstone.

Figure 5. Deposited particle size distributions generated from three different pick wears. A) Concrete.
B) Limestone. C) Sandstone.

269
Figure 6. FE-SEM images of suspended dust generated from the moderate pick while cutting concrete
(left), limestone (center), and sandstone (right).

Figure 7. Suspended respirable particle shapes generated from the three rock types. A) Roundness
values. B) Aspect ratios. C) Roughness values. D) Two-way ANOVA p-value results of particle shapes.

particle shapes. On the other hand, with the rock type p-values all below the 0.05 threshold,
the statistical test provides evidence that the rock type causes an effect to the roundness,
aspect ratio, and roughness particle shapes.

4 DISCUSSION

4.1 NIOSH manual of analytical methods

As revealed in Figure 2 and with the ANOVA statistical tests, the concentration of suspended
respirable dust consistently increases as the wear of the pick increases for all three rock types
cut. The trend is repeated and confirmed in all the duplicate test sets and confirms previous
studies of coal samples cut with various pick wears (Hanson & Roepke, 1979; W. W. Roepke
& Hanson, 1983; Zhou et al., 2020). Therefore, as picks get older and wear down with a more
rounded tip, increased measures should be taken to mitigate and suppress dust.
As revealed in Figure 3 and with the two-way ANOVA test, the percent silica in suspended
respirable dust is inconsistent and appears unaffected by pick wear. Additionally, the two-way
ANOVA test provides evidence that the percent silica generated is affected by the rock type
cut. The purpose of the NIOSH manual of analytical methods 7500 analyses was to observe
the presence of silica, which is true for all the rock types, whether in the form of cristobalite,
quartz, or tridymite. Therefore, regardless of the material being cut with any wear of pick,

270
dusts need to be mitigated and suppressed to limit silica exposures when cutting concrete,
limestone, or sandstone.

4.2 Particle size distributions

Observing Figure 4 reveals that the suspended particle size distributions generated with the
new pick, the moderately worn pick, and the worn pick are all similar for all rock types cut.
Each particle size distribution contains a single-model peak, where the distributions lay on
top, or close to one another. The KS statistical tests provide some evidence that the distribu­
tions are not similar, which is most likely due to the peak value variance. With this, all the
airborne particle size distributions reside within the respirable range, where dust can pene­
trate deep into the lungs. Therefore, the immediate dust released into the environment with
all pick wears while cutting concrete, limestone, or sandstone should be mitigated and
suppressed.
The KS statistical tests performed between datasets provide evidence that the deposited par­
ticle size distributions generated from various pick wears are statistically different, as p-values
are below 0.05 in all cases. Figure 5 and the averages with standard deviations of each distri­
bution further reveal that the deposited particle size distributions slightly shift in response to
pick wear used to cut each rock. As the pick wear increases, the single-modal particle size dis­
tributions shift to the right, therefore suggesting that the deposited particle sizes increase with
pick wear. It is theorized that the new pick generated smaller deposited respirable particles
due to the pin-pointed surface contact area that occurs during rock cutting with newer picks.
It is possible that particles become larger as the pick wears down because the worn pick tips
are rounder and can distribute the load more during cutting (Dogruoz et al., 2016; Dogruoz &
Bolukbasi, 2014; Evans, 1965; Hurt & Evans, 1981). The smaller point of contact might lead
to more rock crushing under the pick tip due to a higher force concentration in the pressure
bubble beneath the pick tip. Therefore, even though there is a shift in the particle size distribu­
tions of the deposited particles as the pick wears, the shifts are not significant amounts and
the particle sizes generated from all the picks are still in the respirable region and can cause
health complications.

4.3 Particle shapes

Figure 7 with the two-way ANOVA test fails to provide evidence that the pick wear variable
affects the suspended dust particle shapes. Therefore, during operations, as the pick wears, the
shapes of the particles do not appear to drastically change. However, there is some evidence
that changing the rock type slightly changes the particle shapes in terms of roundness, aspect
ratio, and roughness.
When averaging the new, moderate, and worn particle shape values for all the rock types,
the total average particle roundness is 0.57, the aspect ratio is 1.54, and the roughness is
0.96, as seen in Table 2. This quantitative data explains that the particles generally have
a slightly oval shape and mostly smooth edges. Additionally, this information, along with
current studies on particle shape influencing deposition in the lung (Kleinstreuer & Feng,
2013; Kori & Pratibha, 2019; Ohsaki et al., 2019; Sturm, 2010; Zellnitz et al., 2019), will aid
in determining how harmful dust particles could be and when they will need to be
suppressed.

Table 2. Average particle shape values for the three various rock types cut.
Shape characteristic Concrete Limestone Sandstone Overall average

Roundness 0.56 0.57 0.57 0.57

Aspect Ratio 1.55 1.52 1.56 1.54
Roughness 0.96 0.94 0.97 0.96

271
5 CONCLUSION

Suspended and deposited dust was generated from concrete, limestone, and sandstone and
collected while cutting full-scale rock samples with a new, moderately worn, and worn conical
pick. The concentrations, silica contents, particle size distributions, and particle shapes were
determined for each dust generated. The analyses reveal that suspended respirable dust con­
centrations increased as the pick wear increased for all rock types. In terms of silica, all three
picks cutting all the rock types generated suspended dust containing quartz or cristobalite. In
addition, the suspended particle size distributions are similar for all pick types, whereas the
deposited particle sizes slightly increase as the pick wear increases. Although there are slight
trends between the pick wear and the deposited particle size distributions for all the rock
types, there are no significant differences as all particle size distributions reside in the respir­
able size range. Lastly, the pick wear does not influence any change in the particle shapes. All
the picks consistently generated suspended respirable particles with similar particle shapes
that are slightly oval with mostly smooth edges.

REFERENCES

Blackley, D. J., Crum, J. B., Halldin, C. N., Storey, E., & Laney, A. S. (2016). Resurgence of progressive
massive fibrosis in coal miners — Eastern Kentucky, 2016. Morbidity and Mortality Weekly Report, 65
(49), 1385–1389. https://doi.org/10.15585/mmwr.mm6549a1
Centers of Disease Control and Prevention. (1974). Recommendations for an occupational exposure stand­
ard for crystalline silica (pp. 1–23). https://www.cdc.gov/niosh/pdfs/75-120a.pdf
Department of Health and Human Services. (2011). Coal Mine Dust Exposures and Associated Health
Outcomes: A Review of Information Published Since 1995. Current Intelligence Bulletin, 64.
Dogruoz, C., & Bolukbasi, N. (2014). Effect of cutting tool blunting on the performances of various
mechanical excavators used in low- and medium-strength rocks. Bulletin of Engineering Geology and
the Environment, 73(3), 781–789. https://doi.org/10.1007/s10064-013-0551-y
Dogruoz, C., Bolukbasi, N., Rostami, J., & Acar, C. (2016). An Experimental Study of Cutting Perform­
ances of Worn Picks. Rock Mechanics and Rock Engineering, 49(1), 213–224. https://doi.org/10.1007/
s00603-015-0734-x
Doney, B. C., Blackley, D., Hale, J. M., Halldin, C., Kurth, L., Syamlal, G., & Laney, A. S. (2019). Res­
pirable coal mine dust in underground mines, United States, 1982-2017. American Journal of Industrial
Medicine, 62(6), 478–485. https://doi.org/10.1002/ajim.22974
Evans, I. (1965). The force required to cut coal with blunt wedges. International Journal of Rock Mechan­
ics and Mining Sciences And, 2(1), 1–12. https://doi.org/10.1016/0148-9062(65)90018-5
Fan, L., & Liu, S. (2021). Respirable nano-particulate generations and their pathogenesis in mining
workplaces: a review. International Journal of Coal Science and Technology, 8(2), 179–198. https://doi.
org/10.1007/s40789-021-00412-w
Fowell, R. J., & Ochei, N. N. (1984). A comparison of dust make and energy requirements for rock cut­
ting tools. International Journal of Mining Engineering, 2(1), 73–83. https://doi.org/10.1007/
BF00880859
Hall, N. B., Blackley, D. J., Halldin, C. N., & Laney, A. S. (2019). Current Review of Pneumoconiosis
Among US Coal Miners. Current Environmental Health Reports, 6(3), 1–11. https://doi.org/10.1007/
s40572-019-00237-5.Current
Hanson, B. D., & Roepke, W. W. (1979). Effect of Symmetric Bit Wear and Attack Angle on Airborne
Respirable Dust and Energy Consumption. Report of Investigations - United States, Bureau of Mines,
8395.
Hinds, W. C. (1999). Aerosol technology: Properties, behavior, and measurement of airborne particles.
In Library of Congress Cataloging in Publication Data. https://doi.org/10.1016/0048-9697(83)90099-2
Hurt, K. G., & Evans, I. (1981). Point Attack Tools: an Evaluation of Function and Use for Rock
Cutting. Mining Engineer, 140(234), 673–683.
Khattak, J., Theisen, D., & Tsai, C. S. J. (2019). Sampling Techniques on Collecting Fine Carbon Nano­
tube Fibers for Exposure Assessment. Scientific Reports, 9(1), 1–8. https://doi.org/10.1038/s41598-019-
43661-1
Kleinstreuer, C., & Feng, Y. (2013). Computational analysis of non-spherical particle transport and
deposition in shear flow with application to lung aerosol dynamics - A review. Journal of Biomechan­
ical Engineering, 135(2). https://doi.org/10.1115/1.4023236

272
Kori, J., & Pratibha. (2019). Simulation and modeling for aging and particle shape effect on airflow
dynamics and filtration efficiency of human lung. Journal of Applied Fluid Mechanics, 12(4),
1273–1285. https://doi.org/10.29252/jafm.12.04.29290
Labranche, N., Keles, C., Sarver, E., Johnstone, K., & Cliff, D. (2021). Characterization of particulates
from australian underground coal mines. Minerals, 11(5), 1–10. https://doi.org/10.3390/min11050447
Lee, K., Lee, E. J., Lee, S. Y., In, K. H., Kim, H.-K., & Kang, M.-S. (2014). Resurgence of
a Debilitating and Entirely Preventable Respiratory Disease among Working Coal Miners. American
Journal of Respiratory and Critical Care Medicine, 190(6), 708–709.
MSHA. (1977). Federal Mine Safety & Health Act of 1977 (Mine Act).
Ohsaki, S., Mitani, R., Fujiwara, S., Nakamura, H., & Watano, S. (2019). Effect of particle-wall inter­
action and particle shape on particle deposition behavior in human respiratory system. Chemical and
Pharmaceutical Bulletin, 67(12), 1328–1336. https://doi.org/10.1248/cpb.c19-00693
Plis, M., Wingquist, C., & Roepke, W. (1988). Preliminary Evaluation of the Relationship of Bit Wear to
Cutting Distance, Forces, and Dust Using Selected Commercial and Experimental Coal- and
Rock-Cutting Tools. US Bureau of Mines: Report of Investigations, 9193.
Pope III, C. A., Burnett, R. T., Thun, M. J., Calle, E. E., Krewski, D., & Thurston, G. D. (2002). Lung
Cancer, Cardiopulmonary Mortality, and Long-term Exposure to Fine Particulate Air Pollution. The
Journal of the American Medical Association, 287(9), 1132–1141. http://jama.jamanetwork.com/article.
aspx?doi=10.1001/jama.287.9.1132
Qayyum, R. A. (2003). Effects of bit geometry in multiple bit-rock interaction.
Roepke, W., Lindroth, D. P., & Myren, T. A. (1976). Reduction of dust and energy during coal cutting
using point-attack bits. Bureau of Mines, 8185.
Roepke, W. W., & Hanson, B. D. (1983). Effect of asymmetric wear of point attack bits on coal-cutting
parameters and primary dust generation. US Department of the Interior, Bureau of Mines, 8761.
Ross, M. H., & Murray, J. (2004). Occupational respiratory disease in mining. Occupational Medicine, 54
(5), 304–310. https://doi.org/10.1093/occmed/kqh073
Roxborough, F. F., King, P., & Pedroncelli, E. J. (1981). Tests on the Cutting Performance of
a Continuous Miner. Journal of The South African Institute of Mining and Metallurgy, 81(1), 9–25.
https://doi.org/10.1016/0148-9062(81)90237-0
Sarver, E., Keles, C., & Rezaee, M. (2019). Characteristics of respirable dust in eight appalachian coal
mines: A dataset including particle size and mineralogy distributions, and metal and trace element
mass concentrations. Data in Brief, 25, 104032. https://doi.org/10.1016/j.dib.2019.104032
Shao, W. (2016). A study of rock cutting with point attack picks. The University of Queensland, School of
Mechanical and Mining Engineering, Australia.
Slouka, S., Brune, J., & Rostami, J. (2022). Characterization of Respirable Dust Generated from
Full-Scale Laboratory Igneous Rock Cutting Tests with Conical Picks at Two Stages of Wear.
Mining, Metallurgy & Exploration. https://doi.org/10.1007/s42461-022-00625-w
Slouka, S., Brune, J., Rostami, J., Tsai, C., & Sidrow, E. (2022). Characterization of Respirable Dust
Generated from Full Scale Cutting Tests in Limestone with Conical Picks at Three Stages of Wear.
Minerals, 12(8), 930. https://doi.org/10.3390/min12080930
Stansbury, R. C. (2018). Progressive massive fibrosis and coal mine dust lung disease: The continued
resurgence of a preventable disease. Annals of the American Thoracic Society, 15(12), 1394–1396.
https://doi.org/10.1513/AnnalsATS.201809-598ED
Sturm, R. (2010). Theoretical models for dynamic shape factors and lung deposition of small particle
aggregates originating from combustion processes. Zeitschrift Fur Medizinische Physik, 20(3),
226–234. https://doi.org/10.1016/j.zemedi.2010.04.001
Su, O., & Akkaş, M. (2020). Assessment of pick wear based on the field performance of two transverse
type roadheaders: a case study from Amasra coalfield. Bulletin of Engineering Geology and the Environ­
ment, 79(5), 2499–2512. https://doi.org/10.1007/s10064-019-01712-x
Talbo, H., & Sevigné, C. (1986). Longwall mining system U.S. Patent No. 4,382,633.
Zellnitz, S., Zellnitz, L., Müller, M. T., Meindl, C., Schröttner, H., & Fröhlich, E. (2019). Impact of drug
particle shape on permeability and cellular uptake in the lung. European Journal of Pharmaceutical
Sciences, 139(September), 105065. https://doi.org/10.1016/j.ejps.2019.105065
Zhou, W., Wang, H., Wang, D., Zhang, K., Du, Y., & Yang, H. (2020). The effect of geometries and
cutting parameters of conical pick on the characteristics of dust generation: Experimental investigation
and theoretical exploration. Fuel Processing Technology, 198 (October2019), 106243. https://doi.org/
10.1016/j.fuproc.2019.106243
Zosky, G. R., Hoy, R. F., Silverstone, E. J., Brims, F. J., Miles, S., Johnson, A. R., Gibson, P. G., &
Yates, D. H. (2016). Coal workers’ pneumoconiosis: An Australian perspective. Medical Journal of
Australia, 204(11), 414–418.e2. https://doi.org/10.5694/mja16.00357

273
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Accuracy of low-cost particulate matter sensor in measuring coal

mine dust- a wind tunnel evaluation

Mirza Muhammad Zaid, Guang Xu & N.A. Amoah

Department of Mining Engineering, Missouri University of Science and Technology, Rolla, Missouri, USA

ABSTRACT: Increased exposure to respirable coal mine dust leads to irreversible lung dis­
eases in mine workers. MSHA mandates a maximum dust concentration of 1.5 mg/m3 in
underground coal operations. Current ways of monitoring concentrations using a filter-based
method or expensive PDM3700 monitors cannot provide real-time concentration. Low-cost
light-scattering particulate matter (PM) sensors have the potential to overcome these chal­
lenges. Tests were performed using factorial design to evaluate the impact of wind velocity
and sensor direction on the sensor response. Three levels of wind velocity 3.0 m/s, 1.5 m/s, and
0.5 m/s, and sensor direction towards the stream, perpendicular to the stream and opposite to
the stream and were evaluated. The results showed that the direction of the sensor has
a significant impact on their response. For each input concentration of 0.0-1.0 mg/m3,
1.0-2.0 mg/m3, and 2.0-3.0 mg/m3, PSM 5003 sensor has a superior response than any other
sensor. This study will provide valuable information on the feasibility of using these sensors to
make real-time decisions.

Keywords: Low-cost PM senors, Coal mine dust monitoring, Particulate matter

1 INTRODUCTION

Coal mining generates dust whose exposure for longer periods results in various health dis­
eases including coal workers’ pneumoconiosis (CWP) and silicosis (Ishtiaq et al., 2018).
Increased exposure leads to various lung diseases that increase miners’ mortality (Zhang et al.,
2007). The permissible coal mine dust exposure in underground and surface coal mines of the
USA is 1.5 mg/cm3 given by Mine Safety and Health Administration (MSHA). CWP has
steadily increased in underground coal miners since 2000, especially in young miners in the
Appalachian region (Blackley, David J. Laura E., Reynolds Connie Short, Ron Carson,
Eileen Storey, Cara N. Halldin, 2022; Potera, 2019), (Sellaro et al., 2015). Even with the
advancement in underground coal excavation methods, the coal workers in an underground
mine are at risk of inhalation of respirable coal mine dust (dos S Antao et al., 2005). There­
fore, due to the threat posed by the coal mine dust to the miners, it is critical to constantly
monitor the dust concentration.
Current ways of monitoring coal mine dust concentration are either filter-based methods or
using the MSHA-certified PDM3700 monitor. Some instruments use filter-based gravimetric
methods to measure PM concentration. FLIR Airtec is a filter-based DPM monitor that
measures the elemental carbon (EC) to determine the exposure of workers in underground
mines to diesel particulate matter. One disadvantage associated with these monitors is that
EC’s standard varies in different countries. EC: TC ratios must be set on the instrument to get
the conversion of EC to TC and hence the DPM concentration. Due to drastic changes in
temperatures during experiments, vapor condensation can increase on the sampling cassette
which results in an overestimation of particle concentration (Janisko and Noll, 2010; Takiff

DOI: 10.1201/9781003429241-29

274
and Aiken, 2010). Additionally, the measured concentration is based on the average value of
high and low concentrations during the shift (Wang et al., 2016). A high concentration spike
for a short time can be misleading (Chung et al., 2001). The second type of instrument, like
the PDM3700, uses a micro-balanced-based method using a tapered element oscillating micro­
balance (TEOM). The PDM measures dust concentration every minute. Research has shown
that transport losses and particle blow-off of the filter may result in underestimation of the
results (Halterman et al., 2018; Volkwein et al., 2006a, 2004). The PDM’s high cost also pro­
hibits it to be used by each miner during the shift. Another disadvantage reported in PDM is
reporting negative mass concentration due to the loss of volatile material due to heating of the
air inlet and tapered element to ensure stability (Abbasi et al., 2021). Therefore, these instru­
ments are economically expensive to be utilized frequently to collect the PM concentration.
Low-cost light-scattering PM sensors have the potential to overcome these drawbacks. The
particles are exposed to the light beam in the sensing area, and the intensity of the light scattered
by particles is detected by a photodetector. These sensosor does not directly measure the mass
concentration rather the microprocessor converts the intensity into mass or number concentra­
tion (Masson et al., 2015; van den Bossche et al., 2017). Recently, three different low-cost PM
sensors (Shinyei PPD42 NS, Sharp GP2Y1010AU0F, and Lase SEN0177) were tested for coal
mine dust and their response was compared with a reference monitor DustTrack. These sensors
yielded a fair response when compared to the reference monitor. The mean square error (MSE)
was the least for the laser sensor (0.0206) and MSE was 0.0232 and 0.0311 for Shinyei and
Sharp sensors (Ghamari et al., 2022). In another study, an optical sensor was used for coal mine
dust concentration measurement. Results showed that this sensor can measure coal mine dust
concentration accurately and also has good stability. (Zhang et al., 2021a).
There are limited studies available on the performance of low-cost PM sensors when meas­
uring coal mine dust. A study compared the low-cost sensor performance but their responses
were compared to DustTrack as a reference monitor (Ghamari et al., 2022). DustTrack is not
a certified reference monitor to measure coal mine dust unlike a PDM 3700. In another study,
two low-cost sensors, Plan tower PMS1003 and Alphasense CO–B4, were tested as an alterna­
tive for long-term air quality monitoring. These sensors were located at four locations in Aus­
tralia and China and one of the testing sites in China had coal mine dust as one of the
pollution sources (Liu et al., 2020). The author did not explain the response of these sensors
when challenged with the coal dust. Based on these studies, it is important to evaluate the per­
formance of low-cost sensors in coal mine dust measurement. Recently, an optical sensor for
higher-precision coal mine dust measurement was developed. The results indicated that this
showed considerable precision and stability while measuring coal mine dust. The mean meas­
ured error was found to be 3.0-4.0 % in short-term tests and 4.0-5.0% in long-term tests. the
coal concentration used in this research ranged between 200.0-800.0 mg/m3 (Zhang et al.,
2021b). This concentration is exceptionally high compared to the limit imposed by MSHA.
This requires more studies on its detection performance near exposure limit concentrations.
Wind velocity and sensor direction were shown to have an impact on the PM sensors’
response. However, their impact on coal mine dust measurement is unknown. A few studies
have reported the impact of low wind velocities on the performance of low-cost PM sensors.
A wind velocity of less than 0.5 m/s was used for low-cost sensor evaluation (Alfano et al.,
2020; Zikova et al., 2017). In underground mining, an airflow velocity ranges between
1.0-2.0 m/s to provide a comfortable environment for the workers (Roghanchi et al., 2016).
Although most pollutants can be diluted with an airflow velocity of 0.3 m/s. However, the vel­
ocity usually varies between 1.0-3.0 m/s in the production working areas (Christensen et al.,
1984; Fobelets, 1987). A wind velocity of more than 4.0 m/s can enhance the amount of dust
produced (Fanger, 1977). Another study evaluated the low-cost PM sensor with a wind vel­
ocity range of 2.0-7.0 m/s (Mukherjee et al., 2017). In a recent study, three low-cost plan
tower PSM3003 sensors were evaluated that concluded that changing wind direction has an
impact on the performance of these low-cost PM sensors (Bulot et al., 2019; Liu et al., 2020).
There is a need to assess the performance of low-cost PM sensors with the lowest and highest
range of air velocity in underground mining. The impact of the low-cost PM sensor inlet direc­
tion on the accuracy of the measurements needs to be investigated as well.

275
 Thus, this paper aims to evaluate the impact of wind velocity and sensor direction on the
performance of low-cost PM sensors. Evaluation will be performed using Keystone mineral
black 325. Factorial design of experiments will be performed at three dust concentration
levels. At each level, there will be two factors (wind velocity and sensor direction) with three
levels for each factor. Three levels of wind velocities are, 0.5 m/s, 1.5 m/s, and 3.0 m/s. The
sensor direction will be towards the stream, perpendicular to the stream, and opposite to the
stream. The response of each low-cost sensor will be compared with the reference monitor.
This study investigates the factors commonly encountered in an underground mining environ­
ment, which provides valuable information on the feasibility of using these low-cost sensors in
the mining industry where coal mine dust is a concern.

2 EXPERIMENTAL METHODS

In this study, four low-cost PM sensors were evaluated. These sensors include the Gaslab,
CM-505 (Farrar, n.d.) referred to as the GLS multi-gas sensors, Plantower PMS5003 (LTD,
2023), referred to as PMS low-cost PM sensor SPS30 sensor (Sensiron, n.d.), and the Airtrek
sensor(“AerosolWorks LLC,” n.d.) . To observe a better precision response between them,
two units of each type of sensor, except Airtrek, were evaluated. The Gaslab sensors can meas­
ure PM2.5 and PM10 together with other gases like carbon dioxide (CO2), carbon monoxide
(CO), and Oxygen (O2) concentration. The second type PMS is affordable, lightweight, and
gives real-time concentration. These PMS sensors characterize PM by size into PM1 PM2.5
and PM10. SPS 30 sensor can read PM 1.0, 2.5,4 and 10.
An MSHA-approved instrument called a personal dust monitor (PDM 3700) is used as
a reference monitor in this study (Volkwein et al., 2006b). This instrument uses the principle
of tapered element oscillation microbalance (TEOM). The cut-off diameter of the coal mine
dust particle which can enter this instrument is 4.0 µm. The wind tunnel used in this study is
a custom-built wind tunnel made with metal frames and acrylic glass panels. The wind tunnel
has a particle generator made up of an air blower connected to a venturi feeder. A vibratory
feeder is used to feed the coal to the venturi feeder. The particle generation setup is shown in
Figure 1.
The coal mine dust used for this experiment is the Keystone mineral black 325 with a density
of 1,220 kg/m3 and particle sizes in the range of 0.04 µm to 35.00 µm(Kumar, 2018). Before the
injection of dust particles PDM was placed outside the tunnel, and their inlet nozzles along with
two PMS, two SPS30, two GLS, and one Airtrek sensor were placed inside the tunnel.

Figure 1. Particle generation setup used in this study.

A full factorial experimental design was used to determine the accuracy of the low-cost sen­
sors for measuring coal mine dust. Three concentration levels of coal mine dust,0-1.0, 1.0-2.0,
and 2.0-3.0 mg/m3 were used. A total of 27 experiments were performed with a set of 9

276
experiments for each concentration of coal. Each concentration of dust was evaluated in
one day. On each concentration level, the effect of three levels of wind velocity 0.5, 1.5, and
3.0 m/s and three directions of sensors (towards the stream, perpendicular to the stream, and
opposite to the stream) were tested. The response of low-cost sensors will be compared to ta
reference monitors i.e., PDM 3700. It will also be evaluated how these sensors respond to coal
mine dust, and the sensor direction and wind velocity impact the linearity of these sensors
with the reference monitor.

3 RESULTS

The objective of this study is to evaluate the impact of sensor direction and wind velocity on
the performance of low-cost PM sensors. A pair of each low-cost PM sensor i.e., PMS5003,
SPS30, Gaslab, and one Airtrek sensor was tested during this study. The PDM utilizes the
BGI HD cyclone with a precise D50 cutoff point of 4.37 µm.

3.1 Sensor direction

To evaluate the effect of sensor directions on the performance of low-cost PM sensors, the
linearity of each sensor was evaluated by comparing the sensor response to the actual con­
centration. When the output from the Gaslab sensors was compared with the actual concen­
tration measured with respect to time, it showed very low linearity with PDM when the
direction of the sensor was towards the stream. When the actual concentration on day 3 was
between 2.0-3.0 mg/m3, the R-squared value for the Gaslab sensor was 0.0003 which is very
low. This is due to the limited range of Gaslab sensors which makes it report a maximum
value of 1.0 mg/m3 even when the actual concentration is higher. When the actual concentra­
tion is kept between 1.0-2.0 mg/m3, the Gaslab sensor still reported a very low linearity
which is evident from its R-squared value of 0.22 (See Figure 2(b)). Similarly, when the
actual concentration was kept between 0-1.0 mg/m3, the Gaslab sensor showed a very infer­
ior response (R-squared value= 0.0159 (See Figure 2(c)). This low R-squared value shows
that even within the effective range of this sensor (0.0-1.0 mg/m3), the performance of this
sensor renders it ineffective. Therefore, Gaslab sensors will not be evaluated for any further
analysis.
Airtrek sensor had slightly better linearity as compared to Gaslab sensors but still, these
sensors showed very low correlation with the reference monitor when the direction of the
sensor was towards the stream. When the actual concentration goes beyond 2.0-3.0 mg/m3,
the R-squared value is 0.36 which is very low Figure 3(a). At this concentration, the Airtrek
effectively read a concentration between 1.5-2.0 mg/m3.
Airtrek has a measurement range of up to 1.0 mg/m3 even though it can report concentrations
to 2.0 mg/m3 according to the manufacturer datasheet (“AerosolWorks LLC,” n.d.). When the
actual concentration was between 1.0-2.0 mg/m3 on day 2, the Airtrek showed very low linearity.
The R-squared value is as low as 0.119 (See Figure 3 (b)). At a concentration between 1.0-2.0 mg/
m3, the Airtrek mostly reported concentration between 0.8-1.5 mg/m3. Airtrek showed a slightly
better response within its effective concentration range was 0.0-1.0 mg/m3, with an R-squared
value of 0.46 (Figure 3 (c)). This still falls in the category of low correlation in PM monitoring
studies so the response of Airtrek will not be analyzed for further analysis of other sensor
directions.
The SPS30 sensors showed an extremely poor response at a concentration higher than
1.0 mg/m3 which is a confirmation of the manufacturer’s datasheet. It states that the effective
mass concentration range for the SPS30 sensor is between 0.0-1.0 mg/m3. The sensors will still
read the concentration beyond this limit but with very less accuracy. When the actual concen­
tration was between 1.0-2.0 mg/m3 and 2.0-3.0 mg/m3, the R-squared value was 0.0005 and
0.0519 respectively as shown in Figure 4 (a & b). These very low R-squared values indicate
that at a concentration above 1.0 mg/m3, these sensors cannot be utilized for coal mine dust
monitoring.

277
Figure 2. Correlation between PDM reference Monitor and Gaslab Sensor at (A) 2.0-3.0 concentration,
(B) 1.0-2.0 concentration, (C) 0-0.1 Concentration.

Sensor direction has a significant effect on the linearity of SPS30 sensor response within
their effective range of concentration (0-1 mg/m3). At a wind velocity of 3.0 m/sec inside the
tunnel, the sensor showed moderate linearity with the reference monitor which is evident from
its R-squared value of 0.71. The R-squared values decreased to 0.36 and 0.18 when the sensor
direction was changed to perpendicular and opposite respectively as shown in Figure 5(a).
A similar trend was observed for the sensor response for sensor direction at a wind velocity of
1.5 m/s and 0.5 m/s and it can be seen in Figure 5 b & c. A p-value of 0.01 shows the statistical
significance of sensor direction and sensor response.

Figure 3. Correlation between PDM reference Monitor and Airtrek Sensor at (A) 2.0-3.0 mg/m3 con­
centration, (B) 1.0-2.0 mg/m3 concentration, (C) 0-0.1 mg/m3 concentration.

278
Figure 4. The correlation between the PDM reference monitor and SPS30 sensor at (a) Concentration
1.0-2.0 mg/m3 and (b) Concentration 2.0-3.0 mg/m3.

The effect of the sensor direction can be explained in two ways. A moderately high R-squared
value for sensor direction towards the stream shows that particles have more chances to enter
the sensor inlet and be detected. On contrary, when the direction of the sensor is perpendicular
or opposite to the stream, the coal particles have lesser chances to enter the inlet. This is espe­
cially true when the direction is opposite to the stream. Inferior R-square values for perpendicu­
lar and opposite sensor directions also indicate that the fan at the sensor inlet is not very
powerful to draw the same number of particles when facing the stream. This could be the result
of low to bad linearity of the sensor while changing the direction even when the concentration
in the tunnel is within the same range as the sensor direction toward the stream.

Figure 5. Correlation between PDM reference monitor and SPS30 sensor at concentration 0-1.0 mg/m3
(a) velocity 3.0 m/sec, (b) velocity 1.5 m/sec, (c) velocity 0.5 m/sec.

Sensor direction also has a significant impact on the linearity of PMS sensors. As can be
seen from the plots in Figure 6, PMS sensors showed a high correlation with the reference
monitors than any other sensor in this study when the inlet direction of the sensor was
towards the stream. At an actual concentration between 0-1.0 mg/cm3 and a wind velocity of

279
3.0 m/s, the PMS sensor showed very high linearity with an R-squared value of 0.91 Figure 6.
R-square values of 0.85 and 0.8 were obtained when the sensor direction was toward the
stream at wind velocities of 1.5 and 0.5 m/s in the tunnel respectively. These high correlation
values agree with previous studies which also obtained high linearity values for PMS sensors
(Amoah et al., 2023; Sayahi et al., 2019; Wang et al., 2015). When the direction of the sensor
is perpendicular to the stream, a considerable drop in the linearity is observed at all three
wind velocities. An R-square value of 0.31, 0.26, and 0.21 is observed in Figure 6 (a, b, & c).
Similarly, for the sensor direction opposite to the air stream, the PMS sensor showed the
lowest correlation. In Figure 6, R-square values of 0.008, 0.0123, and 0.08 were recorded for
wind velocities of 3.0, 1.5, and 0.5 m/s. Statistical analysis confirmed the significance of sensor
direction on sensor response with a P value of 0.01.
The same trend for sensor direction was observed in the response of PMS sensors with
a higher concentration between 1.0-2.0 mg/m3 as the previous concentration range which
can be seen in Figure 7 a, b & c. PMS sensors showed an interesting trend for the linearity at
higher input concentrations in the 2.0-3.0 mg/cm3 range. At high wind velocity at 3.0 m/s,
the PMS sensor showed a moderately high correlation with, an R-square value of 0.75, with
the reference monitor when the sensor inlet direction is toward the stream Figure 7(a).
Unlike the previous concentrations, the PMS sensor showed very low linearity with the ref­
erence monitor at an inlet velocity of 1.5 and 0.5 m/s. At a wind velocity of 0.5 m/s, when
the sensor direction was toward the stream, an R-square value of 0.38 was recorded (see
Figure 7(c)). At a wind velocity of 1.5 m/s, an R-square value of 0.3 was recorded. Figure 7
(b). These low values of linearity when the inlet sensor direction is towards the stream are
due to the coincidence error. It can be observed in Figure 7 that the actual concentration is
more than 3.0.

Figure 6. Correlation between PDM reference monitor and PMS5003 sensor at concentration 0-1.0 mg/
m3 (a) velocity 3.0 m/sec, (b) velocity 1.5 m/sec, (c) velocity 0.5 m/sec.

At higher concentrations when there are many particles in the sensing area, the sensor reads
multiple particles as one large and heavy particle which results in an incorrect interpretation
of the mass concentration as compared to the reference monitor.

280
3.2 Wind velocity
Wind velocity showed a significant impact on the performance of the low-cost sensors used in this
study. In Figure 5 (a), (b) & (c) it can be seen that velocity does impact the sensor response. At
a concentration of 1.0 mg/cm3, an R-squared value of 0.71, 0.66, and 0.62 were recorded at
a velocity is 3.0, 1.5, and 0.5 m/s. Similarly, the same trend was observed in the perpendicular
direction. PMS sensors also followed the same trend, but their correlation was better than the
SPS30 sensor. A P-value of 0.1 shows that the impact of the velocity on the linearity of the sensor
is significant. Similarly, in Figure 7 and Figure 8, it can be seen that when the velocity is highest
at 3.0 m/s and the sensor direction toward the air stream, a better correlation was observed.
A better correlation at the highest velocities indicates that the fan inlet velocity of the sensor is
close to 3.0 m/. This can be explained by the concept of isokinetic sampling (Wilcox, 1956). For
example, if the wind tunnel velocity is as same as the fan inlet velocity (3 m/sec), then particles
will just follow their streamlines and enter the sensor inlet without much deviation. In the case of
a wind tunnel velocity of 1.5 and 0.5 m/sec, this is different than the velocity at the fan inlet. In
this case, the particles will not follow the streamlines and will deviate from their path at the fan
inlet which results in incorrect sampling entering the sensor and hence lower sensor response.

4 PRECISION

The results of the precision test are displayed in Figure 9 where the response of pair of each
kind of PMS and Sps30 was cpmpared to each other. It can be seen from the Figure 9 that
sps30 sensors have shown the highest degree of precision with an R-squared value of 0.94. The
manufacturer datasheet says that sps30 sensors can effectively read the concentration up to
1.0 mg/m3. These sensors can report higher concentrations, but these concentrations are
imprecise. After sps30, the PMS5003 has shown the second-highest precision.

Figure 7. Correlation between PDM reference monitor and PMS5003 sensor at concentration
1.0-2.0 mg/m3 (a) velocity 3.0 m/s, (b) velocity 1.5 m/s, (c) velocity 0.5 m/s.

Unlike sps30 sensors, these sensors did not experience excessive peak,which means that
sometimes these snosrs reported concetrtaion as high as 15 mg/m3,and reported the concentra­
tion in good agreement with the reference monitors. For both SPS30 and PMS5003 sensors,
the precision is decreasing with increasing concentration.

281
Figure 8. Correlation between PDM reference monitor and PMS5003 sensor at concentration
2.0-3.0 mg/m3 (a) velocity 3.0 m/s, (b) velocity 1.5 m/s, (c) velocity 0.5 m/s.

5 CONCLUSION

The aim of this study was to evaluate the impact of sensor direction and wind velocity on the
performance of low-cost PM sensors. It was found that PM 5003 sensor showed the best lin­
earity as compared to all other sensors evaluated in this study. The impact of three sensor dir­
ections, with three wind velocities and three different concentration levels was evaluated.

Figure 9. Intra-mode correlation between two models of each sensors type (a) PMS5003 (b) SPS30.

Sensor direction has a significant impact on the sensor response. The R-squared values
drastically changed for different sensor directions. The sensors face towards, perpendicular,
and opposite to the wind direction. When the sensor direction was towards the stream, PMS
5003 sensor showed an excellent linearity which is evident from an R-squared value of as high
as 0.91, 0.85 and 0.8 at different wind velocity levels. SPS-30 sensors also showed a very good
response when the sensor direction is toward the stream with an R-squared value of 0.71, 0.66
and 0.62. when the sensor direction is changed to perpendicular and opposite to the stream,
a significant drop was observed in the R-squared value which showed the impact of sensor
direction. It is found that PMS sensor can be utilized to measure the coal mine dust accurately
upto 3 mg/m3.
The impact of the wind velocity is statistically significant on the linearity of these Low-cost
PM sensors. Apparently, R-squared value at different levels of wind velocity did not show

282
a huge difference. We performed a statistical analysis and did the effect test and it turned out
that wind velocity does have a significant effect on the linearity of these low-cost PM sensors
while measuring coal mine dust. It is evident from the F-ratio value <0.05.
During this study, it was found that SPS-30 sensors can only read the concentration effect­
ively with 0-1.0 mg/m3 concentration range. Even though these sensors can measure high con­
centration, but the accuracy of these sensors is compromised over a concentration of
1.0 mg/m3.
Airtrek and Gaslab sensors have very low linearity with the reference monitors at any con­
centration level. Due this this reason the author decided not the evaluate these sensors further
for other factors.
It is concluded from this research that PMS 5003 sensor can be utilized for accurate dust
monitoring up to a concentration level of 3 mg/m3, and SPS-30 sensor can used for coal mine
dust monitoring at lower concentration level up to 1.0 mg/m3.

REFERENCES

Abbasi, B., Wang, X., Chow, J.C., Watson, J.G., Peik, B., Nasiri, V., Riemenschnitter, K.B.,
Elahifard, M., 2021. Review of respirable coal mine dust characterization for mass concentration, size
distribution and chemical composition. Minerals 11, 1–36.
AerosolWorks LLC [WWW Document], n.d. URL https://aerosolworks.com/ (accessed 7.3.23).
Alfano, B., Barretta, L., Giudice, A. Del, De Vito, S., Francia, G. Di, Esposito, E., Formisano, F.,
Massera, E., Miglietta, M.L., Polichetti, T., 2020. A review of low-cost particulate matter sensors
from the developers’ perspectives. Sensors (Switzerland) 20, 1–56.
Amoah, N.A., Xu, G., Kumar, A.R., Wang, Y., 2023. Calibration of low-cost particulate matter sensors
for coal dust monitoring. Sci. Total Environ. 859, 160336.
Blackley, David J. Laura E., Reynolds Connie Short, Ron Carson, Eileen Storey, Cara N. Halldin, A.S.
L., 2022. Progressive Massive Fibrosis in Coal Miners From 3 Clinics in Virginia 2017–2018.
Bulot, F.M.J., Johnston, S.J., Basford, P.J., Easton, N.H.C., Apetroaie-Cristea, M., Foster, G.L.,
Morris, A.K.R., Cox, S.J., Loxham, M., 2019. Long-term field comparison of multiple low-cost par­
ticulate matter sensors in an outdoor urban environment. Sci. Rep. 9, 1–13.
Christensen, N.K., Albrechtsen, O., Fanger, P.O., Trzeciakiewicz, A., 1984. Air movement and draught.
In: Indoor Air. Proceedings of the 3rd International Conference on Indoor Air Quality and Climate.
pp. 301–308.
Chung, A., Chang, D.P.Y., Kleeman, M.J., Perry, K.D., Cahill, T.A., Dutcher, D., McDougall, E.M.,
Stroud, K., 2001. Comparison of real-time instruments used to monitor airborne particulate matter.
J. Air Waste Manag. Assoc. 51, 109–120.
dos S Antao, V.C., Petsonk, E.L., Sokolow, L.Z., Wolfe, A.L., Pinheiro, G.A., Hale, J.M., Attfield, M.
D., 2005. Rapidly progressive coal workers’ pneumoconiosis in the United States: geographic cluster­
ing and other factors. Occup. Environ. Med. 62, 670–674.
Fanger, P.O., 1977. Discomfort due to air velocities in spaces. In: Proc. of Meeting of Commission B1,
B2, E1, of Instit. Refrig. pp. 289–296.
Farrar, P., n.d. No TitleGASLAB PLUS CM-500 MULTI-GAS DETECTOR SERIES [WWW Docu­
ment]. URL https://www.co2meter.com/collections/cm-500-gaslab-plus-collection (accessed 7.3.23).
Fobelets, A.P.R., 1987. Subjective human response to low-level air currents and asymmetric radiation.
ASHRAE Trans. 93, 497–523.
Ghamari, M., Soltanpur, C., Rangel, P., Groves, W.A., Kecojevic, V., 2022. Laboratory and field evalu­
ation of three low-cost particulate matter sensors. IET Wirel. Sens. Syst. 12, 21–32.
Halterman, A., Sousan, S., Peters, T.M., 2018. Comparison of respirable mass concentrations measured
by a personal dust monitor and a personal DataRAM to gravimetric measurements. Ann. Work
Expo. Heal. 62, 62–71.
Ishtiaq, M., Jehan, N., Khan, S.A., Muhammad, S., Saddique, U., Iftikhar, B., Zahidullah, 2018. Poten­
tial harmful elements in coal dust and human health risk assessment near the mining areas in Cherat,
Pakistan. Environ. Sci. Pollut. Res. 25, 14666–14673.
Janisko, S., Noll, J.D., 2010. Field Evaluation of Diesel Particulate Matter Using Portable Elemental
Carbon Monitors. 13th United States/North Am. Mine Vent. Symp. 1–6.
Kumar, A.R., 2018. Dust Control Examination using Computational Fluid Dynamics Modeling and
Laboratory Testing of Vortecone and Impingement Screen Filters 49.

283
Liu, X., Jayaratne, R., Thai, P., Kuhn, T., Zing, I., Christensen, B., Lamont, R., Dunbabin, M., Zhu, S.,
Gao, J., Wainwright, D., Neale, D., Kan, R., Kirkwood, J., Morawska, L., 2020. Low-cost sensors as
an alternative for long-term air quality monitoring. Environ. Res. 185, 109438.
LTD, P., 2023. PMS5003 Particulate Matter Sensor with Cable [WWW Document]. URL https://shop.
pimoroni.com/products/pms5003-particulate-matter-sensor-with-cable (accessed 7.3.23).
Masson, N., Piedrahita, R., Hannigan, M., 2015. Approach for quantification of metal oxide type semi­
conductor gas sensors used for ambient air quality monitoring. Sensors Actuators, B Chem.
Mukherjee, A., Stanton, L.G., Graham, A.R., Roberts, P.T., 2017. Assessing the utility of low-cost par­
ticulate matter sensors over a 12-week period in the Cuyama valley of California. Sensors (Switzer­
land) 17.
Potera, C., 2019. Black Lung Disease Resurges in Appalachian Coal Miners. Am. J. Nurs. 119, 14.
Roghanchi, P., Kocsis, K.C., Sunkpal, M., 2016. Sensitivity analysis of the effect of airflow velocity on
the thermal comfort in underground mines. J. Sustain. Min.
Sayahi, T., Kaufman, D., Becnel, T., Kaur, K., Butterfield, A.E., Collingwood, S., Zhang, Y.,
Gaillardon, P.E., Kelly, K.E., 2019. Development of a calibration chamber to evaluate the perform­
ance of low-cost particulate matter sensors. Environ. Pollut. 255, 1–22.
Sellaro, R., Sarver, E., Baxter, D., 2015. A standard characterization methodology for respirable coal
mine dust using SEM-EDX. Resources 4, 939–957.
Sensiron, n.d. SPS30 [WWW Document]. URL https://sensirion.com/products/catalog/SPS30/ (accessed
7.3.23).
Takiff, L., Aiken, G., 2010. A Real-Time, Wearable Elemental Carbon Monitor for Use in Underground
Mines. 13th United States/North Am. Mine Vent. Symp. 137–141.
van den Bossche, M., Rose, N.T., De Wekker, S.F.J., 2017. Potential of a low-cost gas sensor for atmos­
pheric methane monitoring. Sensors Actuators, B Chem.
Volkwein, J.C., Vinson, R.P., McWilliams, L.J., Tuchman, D.P., Mischler, S.E., 2004. Performance of
a New Personal Respirable Dust Monitor for Mine Use. Rep. Investig. 9663, Natl. Inst. Occup. Saf.
Heal. Pittsburgh Res. Lab.
Volkwein, J.C., Vinson, R.P., Page, S.J., McWilliams, L.J., Joy, G.J., Mischler, S.E., Tuchman, D.P.,
2006a. Laboratory and Field Performance of a Continuously Measuring Personal Respirable Dust
Monitor. Usbm Ri 9669.
Volkwein, J.C., Vinson, R.P., Page, S.J., McWilliams, L.J., Joy, G.J., Mischler, S.E., Tuchman, D.P.,
2006b. Laboratory and field performance of a continuously measuring personal respirable dust
monitor.
Wang, Y., Li, J., Jing, H., Zhang, Q., Jiang, J., Biswas, P., 2015. Laboratory Evaluation and Calibration
of Three Low-Cost Particle Sensors for Particulate Matter Measurement. Aerosol Sci. Technol. 49,
1063–1077.
Wang, Z., Calderón, L., Patton, A.P., Sorensen Allacci, M.A., Senick, J., Wener, R., Andrews, C.J.,
Mainelis, G., 2016. Comparison of real-time instruments and gravimetric method when measuring
particulate matter in a residential building. J. Air Waste Manag. Assoc. 66, 1109–1120.
Wilcox, J.D., 1956. Isokinetic flow and sampling. J. Air Pollut. Control Assoc. 5, 226–245.
Zhang, H., Nie, W., Liang, Y., Chen, J., Peng, H., 2021a. Development and performance detection of
higher precision optical sensor for coal dust concentration measurement based on Mie scattering
theory. Opt. Lasers Eng. 144.
Zhang, H., Nie, W., Liang, Y., Chen, J., Peng, H., 2021b. Development and performance detection of
higher precision optical sensor for coal dust concentration measurement based on Mie scattering
theory. Opt. Lasers Eng. 144.
Zhang, R., Li, G., Fan, J., Wu, D.L., Molina, M.J., 2007. Intensification of Pacific storm track linked to
Asian pollution. Proc. Natl. Acad. Sci. 104, 5295–5299.
Zikova, N., Masiol, M., Chalupa, D.C., Rich, D.Q., Ferro, A.R., Hopke, P.K., 2017. Estimating hourly
concentrations of PM2.5 across a metropolitan area using low-cost particle monitors. Sensors (Switz­
erland) 17, 1–19.

284
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Parametric studies to maximize the dust protection performance

of the two-level manifold canopy air curtain and computational
fluid dynamics modeling

N.A. Amoah & G. Xu

Department of Mining and Explosives Engineering, Missouri University of Science and Technology, Rolla,
Missouri, USA

A.R. Kumar
Department of Energy and Mineral Engineering, Pennsylvania State University, University Park,
Pennsylvania, USA

ABSTRACT: Prolonged exposure to high concentrations of respirable coal mine dust causes
coal workers’ pneumoconiosis and silicosis. Underground coal mine roof bolter operators are
more prone to these ailments due to their elevated exposure to dust. The canopy air curtain
(CAC) was developed by NIOSH to alleviate their exposure. The CAC supplies filtered air
over their breathing zone that dilutes the high coal dust concentrations and provides for an
impenetrable air curtain. Many studies attempted to improve the CAC design, but field tests
show variable dust control efficiencies. The non-uniform airflow distribution across the
plenum and ineffective perimeter flow are the major impediments. This study, therefore, sys­
tematically redesigns the CAC for an optimized flow distribution. The Simplex Evolutionary
Operational (EVOP) optimization algorithm was applied to the geometrical parameters. Com­
putational fluid dynamics (CFD) simulations were run at each iteration of the algorithm. The
combination of these methodologies led to an enhanced uniformity of the airflow distribution
across the plenum to achieve the best possible dust suppression efficiency. A lab experiment
using a physical model was used to validate the computer models and validate the efficacy of
this parametrically designed CAC.

1 INTRODUCTION

Extended exposure to elevated respirable coal mine dust concentrations causes coal workers’
pneumoconiosis (CWP), a potentially fatal lung disease with no known conclusive cure (Blackley
et al., 2018; Laney & Attfield, 2010; Mine Safety and Health Administration, 2020). This has
also been linked to the onset of silicosis if the silica content in the coal dust is high (Cohen et al.,
2016; Sato et al., 2018). Underground coal mine roof bolter operators are prone to a higher risk
of both excessive coal dust and silica exposure due to their proximity to the source (Mine Safety
and Health Administration, 2020; Pollock et al., 2009; Reed et al., 2020a). Miners’ position
downwind of the continuous miner, improper and inadequate ventilation, and dust escaping
from the drill shrouds due to faulty or unmaintained dust collectors add to the severity. Occa­
sionally, roof bolting downwind to the continuous miner can substantially increase bolter
operators’ coal dust exposure level up to 7.0 mg/m3 (Listak & Beck, 2012a). These levels exceed
the maximum concentrations allowed by MSHA outlined in the New Dust Rules of 2014
(MSHA, 2014). The rule requires the mines to maintain a time-weighted average (TWA)
respirable coal dust concentration at or below 1.5 mg/m3 at the shift end. The permissible
limit is further reduced if the mine’s respirable dust contains more than 5% silica.

DOI: 10.1201/9781003429241-30

285
 To reduce dust exposure levels of coal mine roof bolters, a canopy air curtain (CAC) was
developed by the National Institute of Occupational Safety and Health (NIOSH). The CAC is
mounted below the roof bolter’s protective canopy. The CAC uses an exhaust fan to draw air
from the mine atmosphere through a high-efficiency particulate air (HEPA) filter. The filtered air
is then delivered through tubing to the plenums which are located underneath the canopy of the
bolting machine. The plenums supply filtered air over the breathing zone of the operator during
the bolting operations. This offers a level of coal dust protection without affecting operator
safety, comfort, and mobility (Listak & Beck, 2012a). Laboratory assessment carried out by
NIOSH on their initial design yielded an efficiency of 62.0 % (Goodman & Organiscak, 2001).
Many improvements were made in the CAC since its invention. A new prototype with an
improved design covered the entire operating canopy area (Listak & Beck, 2012b). It was then
optimized to provide uniform airflow over the entire plenum, as proven by NIOSH for
enhanced efficiency, using computer simulations and a trial-and-error process. Its dust reduc­
tion efficacy exceeded 67.0 % for entry velocities up to 0.61 m/s (120 fpm). Another design
was then made with a much thinner profile with slotted openings to provide perimeter airflow
(Reed et al., 2017). However, NIOSH gallery tests indicated its low efficiency of about 24 %.
Later, a 3rd generation CAC was designed with a higher perimeter air velocity to prevent con­
taminated airflow from infiltrating the protected zone. Although its lab test dust reduction
has improved up to 49.3 % (Reed et al., 2019), the field test has shown variable efficiencies.
This necessitates further design improvements (Reed et al., 2020b). Therefore, it is important
to ensure an even distribution of airflow across the plenum at a velocity higher than 0.51 m/s
(100 fpm) to protect the entire canopy area.
The objective of this research is to optimize the design of the CAC using a two-level mani­
fold. The manifold consists of the main flow header and several outlets. However, it is challen­
ging for a typical manifold with a constant cross-sectional area header to achieve uniform flow
distribution. This is due to the static pressure build-up toward the end. Studies have shown
that the manifold with a tapered longitudinal section can achieve nearly uniform flow distribu­
tion under the effects of the increased resistance but decreasing cross-section area (Hassan
et al., 2014). Air is firstly distributed horizontally evenly using the main manifold. Its outlet is
connected to a series of sub-manifolds that uniformly distribute air downward. Computational
Fluid Dynamics (CFD) simulations are carried out using the existing peripheral design used in
a previous NIOSH study but with a redesigned internal structure to incorporate the two-level
manifold system. A CFD-based parametric study was performed to determine the optimum
manifold design. The Simplex Evolutionary Operation (EVOP) optimization algorithm was
used to optimize the CAC geometrical features. Each iteration in the optimization process
required a CFD simulation. A CAC prototype was then built based on this optimum design
configuration and tested in laboratory experiments to validate the CFD models.

2 MODEL DESCRIPTION

An efficient CAC must discharge a high volumetric flow of air through the outlets toward the
operator’s breathing zone. The CAC structure must also allow for an elevated and uniform air
velocity at the outlets resulting in an air-curtain formation that protects the operator from
exposure to particulate matter. The following sections describe the geometry and associated
CFD models.

2.1 Geometry
This research builds on a J.H. Fletcher and Co. CAC model developed by NIOSH. Figure 1
illustrates the peripheral geometry used in this study (Listak & Beck, 2012b). Previous designs
have added various designs of baffles, inflow vanes, and flow straighteners to the CAC
internal structure. In the existing design, a single angled plate with an adjustable louver is
installed inside the plenum to regulate the airflow distribution. This current design has an
uneven flow of air beneath the plenum with approximately 30.0 % of the plenum area being

286
unprotected. Areas with low to no airflow are the trapezoidal region as well as the transition
zone between the square and the trapezoidal region. The currently used third-generation CAC
only reduces coal dust exposure by 34.6 % - 49.3 % (Reed et al., 2019).
The schematic of the manifold structure used in this study is demonstrated in Figure 2. The
internal design of our model consists of a two-level manifold system as shown in Figure 3. The
outlets of the first-level manifolds serve as the inlets to the second-level manifolds. The first level
manifold connected to the fan has a cross-sectional dimension of 10.2 cm × 4.0 cm at its entry
and is tapered at an angle at two points along its length. The fan duct measures 10.2 cm in diam­
eter and is considered the inlet for CFD analysis. The CAC is required to have a height restriction
of 5.1 cm (2 in.) to provide head clearance for operators. This restricts the overall entry height of
the first level manifold of the CAC to approximately 5.0 cm. Air flows through the first level
manifold and distributes into the plenum area containing the second level manifolds. Internally,
the plenum area is made up of branches whose parameters are designed in this study. The heights
of these manifolds were fixed at 4.0 cm due to the height restrictions for the CAC. This also maxi­
mized the use of the available space. The widths of these second-level manifolds are, however,
included in the optimization process to be determined. The design constraints restricted the
number of second-level manifolds to ten. The separation distance was also optimized in this
study. These second-level manifolds are linearly tapered at an angle along the header for con­
trolled airflow distribution. Their ends’ dimensions were also optimized to generate the optimum
tapering angle. The use of the two-level manifold gives a significantly higher control over airflow
distribution across the plenum by controlling the quantity of air being sent into each branch.

2.2 Optimization method

The Simplex Evolutionary Operation (EVOP) algorithm was used in this research. This algo­
rithm first defined the parameters to be optimized, their initial conditions, and their paramet­
ric boundaries as defined in Table 1. Ten second-level manifolds are designed for practical
considerations when fabricating the CAC. The constraints for each of the parameters are
determined based on geometry limitations and the parameter in relation to other parameters.
As can be seen, there are k=6 variables to be optimized. The Simplex EVOP method first
determines the initial k+1=7 conditions to be simulated. The conditions for each initial simu­
lation cases are shown in Table 2, where the value of p and q are defined in Equation 1 and
Equation 2. For each iteration in the optimization process, a CFD simulation of the CAC
model was performed.

Figure 1. Button-up view of the original geometry of the CAC showing the perforated plate.

287
Figure 2. Schematic of (a) first level manifold and (b) second-level manifold with labels of parameters
to be optimized.

Figure 3. Two-level manifold CAC.

Table 1. CAC design parameters.

Parameters Initial Value (cm) Constraints Comments

D1 5.0 2 < D1 < 10 First level manifold middle size

D2 2.0 0 < D2 < D1 First-level manifold end size
R 5.0 1 < R < 10 First level manifold rectangular outlet width (height
is fixed as 5.0 cm)
d 2.5 0 =< d ≤ 5 Second-level manifold end size
w 5.0 1 < w < 10 Second-level manifold hole size
s 5.0 1 < s < 15 Second-level manifold hole space

Table 2. Initial conditions.

Vertex D1 (cm) D2 (cm) L (cm) D (cm) W (cm) s (cm)

1 x1 = 5.0 x1 = 2.0 x1 = 5.0 x1 = 2.5 x1 = 5.0 x1 = 5.0

2 x1 + p x1 + q x1 + q x1 + q x1 + q x1 + q
3 x1 + q x1 + p x1 + q x1 + q x1 + q x1 + q
4 x1 + q x1 + q x1 + p x1 + q x1 + q x1 + q
5 x1 + q x1 + q x1 + q x1 + p x1 + q x1 + q
6 x1 + q x1 + q x1 + q x1 + q x1 + p x1 + q
7 x1 + q x1 + q x1 + q x1 + q x1 + q x1 + p

288
 Once the above seven models were developed, the dimensionless nonuniformity coefficient was
calculated for each case and are ranked in decreasing order. The objective of this exercise was to
obtain the case with the lowest non-uniformity (elaborated described in section 2.3). The coordin­
ate of this worst response was assigned the ‘w’ variable and the next to worst as the ‘n’ variable.
These denote the coordinate matrix of all responses that are better than ‘w’ by ‘c’. The inflection
point ‘r’ is calculated using Equation 3. Point ‘r’ is the next model that needs to be simulated. This
is expected to move away in the opposite direction from the one with the worst non-uniformity.
This case will replace the previous worst case, and the procedure is repeated. If the inflection point
is outside the constraints of feasible dimensions of any parameter, it is treated as having the worst
non-uniformity. This will force the procedure to reflect cases with feasible parameters.

2.3 Objective function of the non-uniformity coefficient

The objective function of the optimization process is to minimize the dimensionless non-
uniformity coefficient, Φ as defined in Equation 4. This was calculated for every CAC design
considered at each iteration of the optimization process. In the equation, βi is a dimensionless
parameter defined in Equation 5 which represents the flow ratio in the ith outlet, � β represents
the average flow ratio for all outlets defined in Equation 6. The parameter, N denotes the
total number of parallel outlets in the manifold. In Equation 5, Qi denotes volume flowrate
for the ith outlet (m3/s) and Q represents total flowrate (m3/s). A higher value of Φ indicates
a high nonuniformity in the flow distribution. Iterations were terminated when either (i) Φ
does not change anymore, or (ii) the required pressure reached the maximum practical pres­
sure that a fan can provide.

2.4 Boundary conditions and numerical schemes for steady-state simulations

The simulations were developed using Ansys Fluent CFD software. An inlet velocity of 29.5 m/s
was assigned at the duct connection. The outlets of the plenum were assigned a 0 Pa. static pres­
sure. All other impermeable surfaces were assigned a stationary wall condition. A steady-state
solver for incompressible, turbulent flow, using the semi-implicit algorithm (SIMPLE) algorithm
was used for these steady-state flow simulations. The standard k – ε turbulence model was imple­
mented, which is a common model used in mining turbulent flow applications including NIOSH
CAC studies (Zheng & Reed, 2020). This is a two-equation model that solves Reynold’s stresses
by solving two additional transport equations for turbulence kinetic energy (k) and dissipation
rate of turbulence (ε) (Xu et al., 2017). Second-order discretization numerical schemes were used
for better accuracy.

2.5 Governing equations

Airflow within the domain was considered incompressible and turbulent with no heat transfer.
The Navier-Stokes equations were solved using the software. These are vector equations

289
obtained by applying Newton’s law of motion applied to the discretized fluid element. The
equations represent the conservation of mass (Equation 7) and momentum (Equation 8). The
equations were solved iteratively using the Fluent solver until convergence was achieved.
A threshold residual value of 0.0001 in velocity components the and mass flow rate was set.

where ρ is the density of the simulation fluid (kg/m3), t is the time (seconds), and ~
v is the vel­
ocity vector.

τ is the viscous stress tensor (N) given by Equation 4 below for a Newtonian fluid, ~
~
where ~ b is
body force and μ is molecular viscosity coefficient.

2.6 Mesh independence

The mesh independence study was performed by generating coarse, medium, and fine meshes
with progressively lower characteristic dimensions. Figure 4 shows one of the meshes with infla­
tion layers. Richardson’s method of uncertainty quantification in CFD was used for the study
(Roache, 1997). The simulations were run until convergence and the variation in the total pressure
on the inlet was computed across all three meshes. The relative standard deviation between the
total pressure values obtained from the three meshes should be within 5.0 % for grid independ­
ence. The mesh parameters and measured percentage deviations are reported in Table 3 and
Table 4 respectively. The results obtained from the study fall within the criteria set out in Richard­
son’s method. Therefore, the medium meshes were adopted for all other simulations in this
research.

Figure 4. Computational mesh for the CFD simulations.

Table 3. Grid parameters for mesh independence studies.

Parameters Coarse Mesh Medium Mesh Fine Mesh

Volume (m3) 0.0188 0.0188 0.0188

Number of elements, N (million) 3.25 7.07 14.77
Average cell size (mm) 1.81 1.40 1.09
Cell size ratio – 1.3 1.28

290
Table 4. Calculation of mesh convergence indices.
Parameters Coarse Mesh Medium Mesh Fine Mesh

Static pressure at inlet (Pa) 2,623.15 2,695.61 2,774.3

Average static pressure at inlet (Pa) 2,697.69
Deviation from mean pressure (%) 2.76 0.08 2.84
Average deviation from mean pressure (%) 1.89

3 RESULTS

There are six design factors covering both the primary and secondary level manifolds of the
CAC. The results obtained from the simulations were used in combination with the Simplex
EVOP optimization algorithm to determine the optimum configuration of these design fac­
tors. The results are discussed in the sections that follow.

3.1 Design optimization

The results of the CAC optimization using the simplex EVOP are shown in Figure 5. In general,
the optimization resulted in a progressively more uniform configuration with each iteration. This
trend was expected since the nature of the EVOP algorithm is to generate a better configuration
with each iteration. As a result, the non-uniformity coefficient, Φ, generally decreased throughout
the iterations. The optimum configuration for the CAC was determined to be at the 22nd iteration.
This decision was based on the satisfaction of the double termination criteria of the EVOP algo­
rithms used in this study. As can be seen from Figure 5, the nonuniformity of the cases continued
to decrease beyond the 22nd iteration. However, the static pressure drops for each iteration remain
somewhat constant within the CAC fan performance capabilities. The static pressure then began
to rapidly increase rapidly out of the range of the fan after the 22nd iteration. As a result, all
designs generated after the 22nd iteration were not practical for field implementation due to
extremely high pressure drop and therefore, power requirements. These conditions satisfy both
termination criteria making the 22nd iteration design the optimum design.
The high uniformity generated in this model is attributed to the systematic design process. Since
the canopy of the roof bolter is trapezoidal, the first level manifold needed to be designed to allow
the lengths of the branches to decrease towards the trapezoidal end. Distributing equal quantities
of air into each branch would mean there would be higher air volumes delivered to the area under
the trapezoidal part of the plenum as compared to the square region. To achieve the desired uni­
formity across both regions, the first-level manifold was tapered at two sections along its length.
This method is established as one of the most effective manifold uniformity techniques as dem­
onstrated in (Hassan et al., 2014; Tong et al., 2009). Applying this technique to the first level mani­
fold caused different rates of air quantity distribution along these two adjoining headers. This
made it possible to adjust the airflow distribution within the trapezoidal region independently of
the airflow distribution within the squared region. The goal was to send progressively lesser quan­
tities of air towards the shorter branches to achieve a uniform quantity per unit area across all
branches of the first level manifold. A combination of two linear tapering angles along the length
of the first level manifold was simulated to determine their optimum combination to uniformly
distribute the airflow exiting the outlets across all the branches. This technique greatly enhanced
the uniformity of our designs. The detailed configuration with the dimensions of the optimized
CAC design is illustrated in Figure 6. The optimized model fits the existing design of the roof
bolter’s canopy while meeting the head clearance requirement. Therefore, this design required no
equipment redesign to be mounted.

3.2 CFD results

The airflow velocity and quantity contours for the steady state of the simulation for the CAC
and the outlets are shown in Figure 7 and Figure 8. In Figure 7, the contours in red color repre­
sent areas with a relatively higher velocity of greater than 3.5 m/s while the blue areas represent

291
areas with lower air velocities below 1.0 m/s. As can be seen from Figure 7, airflow from the fan
forces its way into the first-level manifold of the CAC which is uniformly distributed among
the second-level manifolds. The optimized tapered design of the first-level manifold in this study
confirms that the optimum taper design generated nearly no static pressure build-up within any
part of the manifold causing a perfectly uniform efflux throughout the second-level manifolds.
Airflow contours of the CAC outlets shown in Figure 5 moreover indicate that there is a high
degree of improved outflow uniformity across the plenum. This shows a significant improve­
ment in the existing third-generation CAC model. However, a small region of relatively lower
flow quantities can still be observed towards the center of the CAC. This, however, is of no sig­
nificant influence on performance and the overall uniformity and performance of the model as
the results give a standard deviation of 1.03 m/s. Moreover, this will contribute to enforcing that
polluted airflows do not penetrate the perimeter airflow region. Analysis of the quantity vectors
of this simulation moreover showed that there was no backflow of air into the CAC.

Figure 5. Optimization results.

Figure 6. Optimum CAC configuration.

292
Figure 7. The contour of internal airflow pattern in CAC.

Figure 8. Contours for the CAC outlets.

4 CONCLUSION

Roof-bolters operators are exposed to elevated dust concentrations that make them prone to the
onset of irreversible diseases such as CWP and silicosis. To control their exposure levels, the
canopy air curtain was designed by NIOSH to deliver filtered air on the roof bolter operators
breathing zone to dilute coal dust within their working area. Additionally, this design and subse­
quent improvements also had a provision for perimeter flow. A high-velocity air jet emancipating
from the slits downstream serves as the air curtain. This is impassable to dust particles present in
the ventilation air stream lowering his exposure to respirable dust. This paper summarizes the
parametric studies of the geometry of the CAC aimed at allowing near-equal airflow through all
the branches. The authors investigated the optimum design parameters for the proposed two-level
manifold-based CAC which will achieve the best possible uniformity and best efficiency.
The original perimeter profile of the CAC was used in this study. A two-level manifold system
was introduced into the CAC peripheral design to carry air toward the operator’s breathing zone.
Steady-state CFD models were developed to investigate the impact of four major parameters on
the airflows. This study focused on optimizing the uniformity of the CAC by tapering the two-
level manifolds to redistribute airflow across their outlets. Systematically adjusting the tapering
angle resulted in the desired airflow redistribution. A parameter called the non-uniformity index,
Φ, was calculated for all the models generated. A low Φ value indicates an even volumetric airflow
rate through all the outlets. The design which resulted in the least Φ was determined to be the
optimum design of the two-level manifold-based CAC. This design is critical for the effective dilu­
tion of dust-laden air close to the breathing zone of the miner.

293
 Analysis of all the parameters showed that an inlet size (D1) dimension of 5.0 cm × 5.0 cm,
tapered header end size (D2) dimension of 5.0 cm × 0.2 cm, outlet size (d) of 1.27 cm (1/2 in.)
in diameter, outlet separation of 6.35 mm (1/4 in.) and a header length of 56.52 cm was the
optimum configuration for the second level manifolds. Additional studies show that an inlet
size (D1) of 10.2 × 6.0 cm, tapered header end size (D2) of 2.0 × 6.0 cm, header length (L) of
101.6 cm, outlet size (d) of 5.0 × 5.0 cm and outlet separation (S) of 0.5 cm was the optimum
configuration for the first level manifold. The combination of these parameters yields the best
possible design to achieve the most uniform airflow distribution across the plenum of a CAC.
Further studies are focusing on building a physical model based on these parameters to be
evaluated in a mining setup to validate the CFD results and to determine the efficiency of this
model. Future studies will focus on optimizing additional parameters to investigate the pos­
sible improvement in the uniformities in airflow distribution.

REFERENCES

Blackley, D. J., Halldin, C. N., & Scott Laney, A. (2018). Continued increase in prevalence of coal work­
ers’ pneumoconiosis in the United States, 1970-2017. In American Journal of Public Health (Vol. 108,
Issue 9, pp. 1220–1222). American Public Health Association Inc. https://doi.org/10.2105/
AJPH.2018.304517
Cohen, R. A., Petsonk, E. L., Rose, C., Young, B., Regier, M., Najmuddin, A., Abraham, J. L.,
Churg, A., & Green, F. H. Y. (2016). Lung pathology in U.S. coal workers with rapidly progressive
pneumoconiosis implicates silica and silicates. American Journal of Respiratory and Critical Care Medi­
cine, 193(6), 673–680. https://doi.org/10.1164/rccm.201505-1014OC
Goodman, G. V. R., & Organiscak, J. A. (2001). Laboratory Evaluation of a Canopy Air Curtain for
Controlling Occupational Exposures of Roof Bolters. Seventh International Mine Ventilation Congress,
Proceeding, 299–305.
Hassan, J. M., Mohamed, T. A., Mohammed, W. S., & Alawee, W. H. (2014). Modeling the Uniformity
of Manifold with Various Configurations. Journal of Fluids, 2014(April), 1–8. https://doi.org/10.1155/
2014/325259
Laney, A. S., & Attfield, M. D. (2010). Coal workers’ pneumoconiosis and progressive massive fibrosis
are increasingly more prevalent among workers in small underground coal mines in the United States.
Occupational and Environmental Medicine, 67(6), 428–431. https://doi.org/10.1136/oem.2009.050757
Listak, J. M., & Beck, T. W. (2012a). Development of a canopy air curtain to reduce roof bolters’ dust
exposure. Mining Engineering, 64(7), 72–79.
Mine Safety and Health Administration. (2020). Open Government Initiative Portal - MSHA. https://
arlweb.msha.gov/OpenGovernmentData/OGIMSHA.asp
MSHA. (2014). Lowering Miners’ Exposure to Respirable Coal Mine Dust, Including Continuous Personal
Dust Monitors (Vol. 79). http://www.cdc.gov/niosh/docs/2008-143/pdfs/
Pollock, D. E., Potts, J. D., & Joy, G. J. (2009). Investigation into dust and exposures and mining practices
in mines in the southern appalachian region. Society for Mining, Mettalurgy, and Exploration, Incorp­
orated, 2009.
Reed, W. R., Joy, G. J., Kendall, B., Bailey, A., & Zheng, Y. (2017). Development of a roof bolter
canopy air curtain for respirable dust control. Mining Engineering, 69(1), 33–39. https://doi.org/
10.19150/me.7010
Reed, W. R., Joy, G. J., Shahan, M., Klima, S., & Ross, G. (2019). Laboratory results of a 3rd gener­
ation roof bolter canopy air curtain for respirable coal mine dust control. International Journal of Coal
Science and Technology, 6(1), 15–26. https://doi.org/10.1007/s40789-019-0237-6
Reed, W. R., Shahan, M., Klima, S., Ross, G., Singh, K., Cross, R., & Grounds, T. (2020a). Field study
results of a 3rd generation roof bolter canopy air curtain for respirable coal mine dust control. Inter­
national Journal of Coal Science and Technology, 7(1), 79–87. https://doi.org/10.1007/s40789-019-
00280-5
Reed, W. R., Shahan, M., Klima, S., Ross, G., Singh, K., Cross, R., & Grounds, T. (2020b). Field study
results of a 3rd generation roof bolter canopy air curtain for respirable coal mine dust control. Inter­
national Journal of Coal Science and Technology, 7(1), 79–87. https://doi.org/10.1007/s40789-019-00280-5
Roache, P. J. (1997). Quantification of uncertainty in computational fluid dynamics. Annual Review of
Fluid Mechanics, 29, 123–160. https://doi.org/10.1146/annurev.fluid.29.1.123
Sato, T., Shimosato, T., & Klinman, D. M. (2018). Silicosis and lung cancer: Current perspectives. Lung
Cancer: Targets and Therapy, 9, 91–101. https://doi.org/10.2147/LCTT.S156376

294
Tong, J. C. K., Sparrow, E. M., & Abraham, J. P. (2009). Geometric strategies for attainment of identical
outflows through all of the exit ports of a distribution manifold in a manifold system. Applied Thermal
Engineering, 29(17–18), 3552–3560. https://doi.org/10.1016/j.applthermaleng.2009.06.010
Xu, G., Luxbacher, K. D., Ragab, S., Xu, J., & Ding, X. (2017). Computational fluid dynamics applied
to mining engineering: a review. International Journal of Mining, Reclamation and Environment, 31(4),
251–275. https://doi.org/10.1080/17480930.2016.1138570
Zheng, Y., & Reed, W. R. (2020). Effects of Roof Bolter Canopy Air Curtain on Airflow and Dust Disper­
sion in an Entry Using Exhaust Curtain Ventilation. August. https://doi.org/10.1007/s42461-020-00294-7

295
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Respirable coal mine dust research: Characterization and toxicity

analysis based on dust sources

V.P. Salinas, M.C. Das, G. Rubasinghege & P. Roghanchi

New Mexico Institute of Mining and Technology, Socorro, NM, USA

K. Zychowski
College of Nursing, University of New Mexico-Health Sciences Center, Albuquerque, NM, USA

ABSTRACT: This study investigates the respirable coal mine dust (RCMD) characteristics
and toxicity based on the mine’s sources of dust particles. Dissolution experiments in simu­
lated lung fluids (SLFs) and in-vitro response were conducted to determine the toxicity level
of samples collected from the coal seam, and the host rock (roof and floor) of an underground
coal mine. Dust characteristics were investigated using various dust characterization studies.
Significantly higher mineral and elemental contents were found in samples from the host rock,
suggesting higher toxicity. Si, Al, Fe, Cu, Ba, Sr, and Pb were found in dissolution experi­
ments. Host rock displayed a much higher bio-accessibility than coal when exposed to artifi­
cial lysosomal fluid (ALF), but not with Gamble’s solution (GS). Additionally, neutrophilic
and lung epithelial cells showed major affinity to activate inflammatory pathways when
exposed to host rock dust. Contrary, macrophage cells were more like to activate inflamma­
tion from the coal, but with low expressions.

1 INTRODUCTION

Mineral coal continues nowadays to be important in industry and life because it still provides
a significant percentage of the energy consumed in the U.S. (EIA, 2022). RCMD toxicity
gained considerable relevance in coal mining in the last 50 years after encountering thousands
of coal mine workers suffering from lung diseases and even dying as a result of years of con­
tinued exposure to RCMD.
Although efforts have been made to identify the characteristics and root causes of RCMD
toxicity, they remain unclear and are a subject of investigation (Rahimi et al, 2023).
The toxicity of RCMD has been related to several factors, such as particle size, shape, com­
position, silica content, and mine practices, among others (Shekarian et al., 2021). The higher
rates of reported miners diagnosed with coal workers’ pneumoconiosis in the eastern region of
the U.S. have also suggested that the RCMD coming from this area may exhibit higher tox­
icity, but until now, the reason why has not been clarified (Salinas et al., 2022). Some studies
have suggested that the seam thickness may influence due to the high silica in the air resulting
from cutting the rock strata of the floor and roof in the production areas (Sarver et al., 2019;
NAS, 2018).
Silica has been pointed out as an enhancer of toxicity in RCMD (Porter et al., 2002; Baron
et al., 2002). It is classified as a carcinogen material, so it is believed that higher silica content
in the RCMD may make it more toxic (Abbasi et al., 2021). Quartz is the most common crys­
talline silica polymorph and the second most abundant mineral in the continental crust. It is
one of the most important minerals in forming rocks and soil and is often found in coal mines
and as a component of RCMD (Tarbuck & Lutgens, 2005).

DOI: 10.1201/9781003429241-31

296
 Also, several studies have tried to determine whether the content of Fe in RCMD can
increase its toxicity, but until now, there have been findings indicating both that it is related
and not (Shangguan et al., 2022; Trechera et al., 2020)
Considering that different factors may influence the toxicity of RCMD, the primary research
focused on three approaches or variables to determine RCMD toxicity: geographic location,
source, and particle size. The geographic location influence was previously evaluated, and
results can be found in Salinas et al. (2022). In this work, samples taken directly from the coal
seam of five mines from the Appalachian and Rocky Mountains regions were characterized and
analyzed regarding their toxicity using dissolutions in simulated lung fluids and in-vitro experi­
ments. Still, no trends were found indicating that the geographic location influenced the toxicity
based on these experiments (Salinas et al., 2022).
Regarding the source influence, the dust particles in the mine can come from different
sources, such as the host rock, rock dust, the coal seam, diesel particulate matter (DPM), and
inlet air, among others (Shekarian et al., 2022). All these particles coming from different
sources have different compositions, mineralogic and elemental wise. These differences may
make the dust particles display different toxicity when exposed to the lungs. The hypothesis is
that the RCMD generated by cutting host rock is more toxic than the dust coming from cut­
ting coal seam. This variable is further analyzed in this document.
RCMD is composed of particles with a size below 10 µm (Brown et al., 2013). In this range,
particles of different sizes may expose different toxicities. Smaller particles may have a more
extensive specific area exposed to be in contact with the lungs when inhaled, which may help
the elements that compound the samples to dissolve more. So, the hypothesis regarding the
influence of the particle size is that smaller dust particles have higher metal/elemental dissol­
ution than larger particles in simulated lung fluids. The evaluation of this last variable is pro­
posed for future work.

2 MATERIALS AND METHODS

Samples were collected from an underground coal mine located in the state of Pennsylvania,
belonging to the coal basin of the Appalachian region. Bulk samples were taken from the coal
seam and the host rock surrounding the seam, including one sample from the roof and
another from the floor of the production area. The information about the mine from where
the samples were collected and the sample IDs used in this document are shown in Table 1.

Table 1. Information about the sample collection.

Information:

Location Pennsylvania
– CO: Coal seam
Samples collected – HR: Host roof
– HF: Host floor
Mine type Underground
Mine method Room and pillar
Coal rank Sub-bituminous
Seam height Low seam

The methodology followed comprised 3 phases: sample preparation, characterization, and

toxicity assessment. All the experiments carried out in this work were done following the same
procedure and with the same equipment used in the previous work (Salinas et al., 2022). Also,
the sample from the coal seam (CO) was used in this previous work to compare the geographic
location’s influence on toxicity.
Samples were prepared by crushing and grinding the bulk samples until getting very fine
material, which was subsequently sieved and separated with a cascade impactor in its less than

297
10 µm fraction. The sample particle size distribution was verified using a scanning electron
microscope (SEM). The resulting samples were used for further analysis.
Characterization included X-ray diffraction (XRD) to identify the mineral components in
the samples, total microwave digestion to determine the elemental composition, and Fourier
Transformed Infrared spectrometry to identify the functional groups in the samples before
and after dissolution experiments.
The toxicity assessment was performed using dissolution in simulated lung fluids (SLFs) and
in-vitro analysis. The objective of the dissolution experiment was to analyze the elements dis­
solving when the samples were in contact with two different SLFs after 24 hours. The SLFs
used were Gamble’s solution (GS) and artificial lysosomal fluid (ALF). GS and ALF are some
of the most used SLFs and have been widely used to assess the bio-accessibility of different
metals from urban and industrial airborne particulates. GS mimics the fluid present in the lung
interstitium and ALF mimics the acid fluid that engulfs particles in the lung macrophages.
When particles are inhaled and reach the alveolar region, can get in contact with these SLFs
giving them the chance of being totally or partially digested. To simulate the process and the
lung conditions, dust samples were put in contact with the SLFs for 24h, the experiment was
conducted in a dark room (to simulate the absence of light in the lungs) at a constant tempera­
ture of 37°C (to simulate body temperature), and oxygen was added to the solution before start­
ing. Additional SLFs models have been used before for other specific purposes like to model
interactions of particles with extracellular lung fluids (Mareno et al., 2019).
The in-vitro experiments looked at the inflammatory response of different human cells
involved in lung disease development (macrophages, epithelial, and neutrophilic) after dust
exposures. For each cell, 10 cytokines were analyzed. Cytokines were selected based on the
cell type exposed and their potential role in coal workers’ pneumoconiosis (CWP), occupa­
tional lung injury, or inflammation, according to the literature (Kilpatrick & Harris, 2011;
Salinas et al., 2022).

3 RESULTS AND DISCUSSION

3.1 Characterization

3.1.1 XRD
The XRD results are shown in Table 2 and Figure 1. The analysis showed that all the samples
have quartz in their mineral composition. As mentioned before, this mineral is very common
in rocks and coal mines. However, it is noticeable that the quartz peaks (e. g. the one at ~26.7°
in Figure 1) in the host rocks (HR and HF) are considerably more significant than the quartz
peak for the coal sample (CO), which means that the host rock has a much higher content of
quartz, behavior also observed in the peaks for muscovite and kaolinite, indicating a more sig­
nificant amount of minerals in the host rock. These results are expected since the CO sample
might be compound mainly by organic material. This organic material is evidenced in the CO
pattern with the amorphous hump that can be observed at the beginning of the graph for this
sample. Regarding the rest of the mineral components, it was observed that the CO compos­
ition was limited to kaolinite, pyrite, and calcite, while the HR and HF had in their compos­
ition a variety of minerals that are usually found in the latest stages of rock disintegration
forming sedimentary rocks such as shales and limestone, which are typically found interlaying
the seams in coal deposits.
The following abbreviations represent the different minerals in Figure 1: Quartz (Qz), Kao­
linite (Kln), Pyrite (Py), Muscovite (Mu), Siderite (Sd), and Calcite (Ca). Microcline and
chlorite are present in the samples in very small quantities, so they cannot be observed without
zooming in on the patterns.

3.1.2 Total microwave digestion

The microwave digestion provided information regarding the elemental composition of the
samples. In the same way as the XRD results, this analysis showed a significantly higher

298
 Table 2. Minerals, compound names, and chemical formu­
las of the crystalline phases present in CO. HR, and HF.
Sample ID
Mineral CO HR HF

Quartz X X X
Kaolinite X X
Pyrite X X
Muscovite X X
Microcline +/-
Chlorite Trace X
Calcite X X

+/- Under detection limit (2.5%). The sample may or may not
contain the mineral.

Figure 1. XRD patterns of CO, HR, and HF samples.

concentration of most of the elements analyzed in the host rock (HR and HF), especially for
Si, Ti, Ba, V, and Cr. The major elements found in the samples were aluminum (Al), silicon
(Si), iron (Fe), potassium (K), and titanium (Ti), and the trace elements were strontium (Sr),
magnesium (Mg), barium (Ba), lead (Pb), manganese (Mn), nickel (Ni), copper (Cu), zinc
(Zn), arsenic (As), vanadium (V), and chromium (Cr). Many of these elements have been asso­
ciated with the production of reactive oxygen species which have been related to RCMD tox­
icity (Moreno et al., 2019). Table 3 shows the concentration of each element analyzed.

299
Table 3. Elemental concentrations in CO, HR, and HF samples.

Sample Al Si Fe K Ti Sr Mg Ba Pb Mn Ni Cu Zn As V Cr
ID % % % % % µg/g µg/g µg/g µg/g µg/g µg/g µg/g µg/g µg/g µg/g µg/g

CO 0.90 3.53 0.9 N.A. 0.11 41.2 N.A. 56.6 16.6 24.4 21.8 20.4 52.7 27.8 36.9 26.3
HR 0.6 23.8 1.1 2.35 0.63 256.7 237 442.9 27.5 57.5 46.5 48.3 51.2 18.9 179.0 92.2
HF 0.5 25.0 1.7 2.04 0.72 53.4 125 341.7 1.1 160.2 42.9 35.1 138.7 3.3 170.1 94.9

Figure 2. Major and trace element components in the samples.

3.1.3 FTIR
The FTIR analysis was carried out on the samples before and after dissolution to determine
the functional groups in the samples’ composition and their changes after exposure to SLFs.
Figure 3 and Table 4 show the FTIR spectra obtained and the peak assignments respectively.
In the first graph (a), which corresponds to the CO sample, it can be observed that on the left
side, numerous peaks are missing in the samples HR and HF. These peaks belong to
antisymmetric–CH3 deformation (1445 cm−1) and the benzene C=C stretching (1602 cm−1)
which are characteristics of the carbonaceous material that is present in CO. Also, it can be
observed that most of the peaks decrease after the dissolution experiment, indicating that
those functional groups are actually dissolving, especially the ones for quartz (779 and
799 cm−1) and kaolinite (913 cm−1). These peaks are pretty much gone after dissolution. This
behavior was also observed in the previous work where the influence of geographic location
was evaluated [4].

Figure 3. FTIR spectra of (a) CO, (b) HR, and (c) HF.

300
 Table 4. Peak assignments from FT-IR (Lin et al., 2019).
Wavenumber (cm−1) Peak Assignment

779 and 799 Quartz

913 Kaolinite
1000–1200 Si=O, Si-O-Si, Si-O-C, C-O-C
1445 Antisymmetric–CH3 deformation
1602 Benzene C=C stretching

3.2 Toxicity assessment

3.2.1 Dissolution in SLFs

Dissolution experiments were carried out for 24 h in Gambles’ solution (GS) and artificial
lysosomal fluid (ALF). The results of the dissolution experiments in GS and ALF are shown
in Figures 4 and 5, respectively. In GS, only Si and Al were dissolving, including a little Cu
from the CO sample. On the other hand, in ALF there is a long list of elements dissolving,
including Al, Si, Fe, Cu, Ba, Sr, Ni, Pb, and Mn. This indicates that the elements in dust sam­
ples are more bio-accessible when they are exposed to ALF. The dissolution of elements in GS
showed to be negligible for most of the elements, representing lower risk compared to the
exposure to ALF.
The dissolution of elements in GS was low in general, similar to the ones obtained in the
previous work for the CO sample, even whit significantly higher initial availability of most of
the elements. For example, in the case of the Si, the initial availability in HR and HF was
around 7 times the initial availability of CO, but the dissolutions of HR and HF were just
slightly higher in GS.
In contrast, the dissolutions in ALF had a significant increase. The final concentrations
after 24 hours increased in an order of 10, which means that the initial availability of elements
in the samples played an important role in the dissolutions in ALF. This also indicates that
the elements in dust samples are more bio-accessible when they are exposed to ALF. Addition­
ally, these higher dissolutions obtained in ALF indicate that the dust coming from the host
rock (HR and HF) is more bio-accessible that the one coming from the coal seam (CO). Still,
even with these higher concentrations, the chances of intoxication or to present adverse effects
due to metal overload remain questionable. However, the release of these elements in the
human body may activate the production of reactive oxygen species, which cannot be dis­
missed as an enhancer of toxicity.

Figure 4. Mass normalized dissolution of metals in GS vs time from (a) CO, (b) HR, (c) HF.

3.2.2 In-vitro experiments

In vitro experiments were conducted to assess the inflammatory response in different human
cells involved in lung disease development. In total, 10 cytokines were analyzed based on lit­
erature for each one of the cells HL-60, A549, and THP-1, as follows:

301
Figure 5. Mass normalized dissolution of metals in ALF vs time from (a) CO, (b) HR, (c) HF.

• HL-60 and A549 cells: IFN-γ, IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IL-13, and TNF-α.
• THP-1 cells: GM-CSF, IL-1α, IL-5, IL-7, IL-12/IL-23p40, IL-15, IL-16, IL-17A, TNF-β,
and VEGF-A

Figure 6. Results of dust exposures to (a) HL-60 cells using low (5 µg/mL), medium (10 µg/mL), and
high (20 µg/mL) concentrations, and (b) THP-1 cell and (c) A549 cells using low (10 µg/mL) and high
(100 µg/mL) concentrations.
302
 These cells play an important role in the body’s immunological response to external agents.
Neutrophilic cells are some of the first cells to react against harmful agents such as viruses and
bacteria, and their main role is to fight infection. Epithelial cells make up the tissue of the
internal and external surfaces of the body and are the first to come in contact with external
agents, such as dust. They contribute to the immunological response mainly with the produc­
tion of cytokines. Macrophages are cells that act against infection and injury by engulfing and
eliminating foreign agents, damaged or dead cells, and other harmful organisms. These cells
are also involved in other immunologic processes.
The results of selected cytokines for cells HL-60, A549, and THP-1 are shown in Figure 6.
In general, cytokines are proteins released from cells that are involved in multiple processes,
such as signaling the immune system to act against foreign agents, cell growth, and pro-
inflammatory and anti-inflammatory responses. In the experiment, an upregulated response
(positive values) indicates a pro-inflammatory response, while a downregulated response
(negative values) may indicate an immunosuppressive response. Immunosuppression is still
important because an immunosuppressed lung may be more susceptible to infection (influ­
enza, etc.), especially when considering occupational dust exposure.
In HL-60 cells (neutrophilic), 6/10 cytokines had at least one proinflammatory response
among the 3 samples and the different concentrations of exposure (IL-1β, IL-2, IL-4, IL-8, IL-
12p70, and TNF-α). In general, almost all the reactions were obtained from HR. In A549 cells
(lung epithelial), 8/10 cytokines had at least one proinflammatory response (IFN-γ, IL-1β, IL-
2, IL-4, IL-6, IL-8, IL-13, and TNF-α), and 2 of them had a reaction from all the samples
under all the concentrations tested (IL-4 and IL-6). In these last 2 cytokines, the reactions
were mostly higher from the samples HR and HF, and for the other 6 cytokines, the reaction
came from either HR or HF, displaying a major affinity to display a proinflammatory
response with exposures of the host rock (HR and HF). In THP-1 cells (macrophage), 8/10
cytokines had at least one proinflammatory response, from which 7 had a reaction from the
CO sample. This indicates, opposite to the other cells, that the coal sample (CO, even with less
element and mineral content) is the activator of inflammation in the macrophage cells,
although the expressions were low. Additionally, HF was the sample with the highest content
of Fe. It has been mentioned in previous research that Fe may enhance the toxicity of RCMD
(Trechera, et al., 2020), but in this case, only some of the lung epithelial cytokines displayed
an inflammatory response (IL-1β, IL-4, IL-6, IL-8, and TNF-α), but not with a remarkable
trend, and in the macrophage and neutrophilic cells, almost any cytokine reacted to HF.
Thus, this study does not demonstrate whether Fe influences RCMD toxicity.
In summary, neutrophilic and epithelial cells had a greater inflammatory response from the
host rock samples in general, while macrophages had activation of inflammation produced
mainly by the coal sample. This indicates that the host rock is more toxic than coal for the
neutrophilic and epithelial cells, but the opposite for macrophage cells.

4 CONCLUSIONS

This study compared the characteristics and toxicity of dust samples from different sources in
the mine: coal seam and host rock, with the objective of determining if the dust coming from
the host rock is more toxic. The following are the main conclusions found:
• When the composition was compared, as expected, it was found that the host rock has
a significantly higher content of elements (especially Si, Ti, Ba, V, and Cr) and minerals
(quartz, kaolinite, and muscovite), which may influence in its final toxicity.
• Dissolution experiments showed no big differences between the dissolutions in coal and
host rock samples when GS was used. The element concentrations remained similar regard­
less of the initial bioavailability. However, when tested with ALF, more elements were dis­
solved, and the dissolutions increased in an order of magnitude of 10 (from ×104 to ×105)
when comparing coal and host rock samples. This indicates that the elemental bio-
accessibility is higher in samples from the host rock, and thus, their toxicity.

303
• In the neutrophilic cells the inflammatory response was moderate across the cytokines ana­
lyzed, and the responses were coming mainly from HR. Lung epithelial cells displayed
major affinity to respond to host rock samples, while the macrophage cells’ response was
associated with the CO samples, but low. This indicates that the host rock is more likely to
activate the cytokines from the neutrophilic and lung epithelial cells, but not much from the
macrophages. Macrophage cells were more likely to be activated by the coal samples that
were considered less toxic regarding composition and elemental dissolution in SLFs.
• Further research is needed using different sets of samples from different mines to determine
if the trends found in this research are consistent from mine to mine. Also, additional exam­
ination of the implications of metal release in the body is required to better understand
their interactions with the body processes and their relationship with RCMD toxicity.

REFERENCES

EIA. (2022). U.S. energy facts explained. 2022. https://www.eia.gov/energyexplained/us-energy-facts/

(accessed Oct. 04, 2022).
Rahimi, E., Shekarian, Y., Shekarian, N., & Roghanchi, P. (2023). Investigation of respirable coal mine
dust (RCMD) and respirable crystalline silica (RCS) in the US underground and surface coal mines.
Scientific Reports, 13(1), 1767.
Shekarian, Y., Rahimi, E., Rezaee, M., Su, W. C., & Roghanchi, P. (2021). Respirable coal mine dust:
a review of respiratory deposition, regulations, and characterization. Minerals, 11(7), 696.
Salinas, V., Das, M., Jacquez, Q., Camacho, A., Zychowski, K., Hovingh, M., . . . & Roghanchi, P.
(2022). Characterization and Toxicity Analysis of Lab-Created Respirable Coal Mine Dust from the
Appalachians and Rocky Mountains Regions. Minerals, 12(7), 898.
National Academies of Sciences, Engineering, and Medicine. (2018). Monitoring and sampling
approaches to assess underground coal mine dust exposures.
Sarver, E., Keles, C., & Rezaee, M. (2019). Beyond conventional metrics: Comprehensive characteriza­
tion of respirable coal mine dust. International Journal of Coal Geology, 207, 84–95.
Porter, D. W., Barger, M., Robinson, V. A., Leonard, S. S., Landsittel, D., & Castranova, V. (2002).
Comparison of low doses of aged and freshly fractured silica on pulmonary inflammation and damage
in the rat. Toxicology, 175(1-3), 63–71.
Baron, P. A., Rice, F. L., Key-Schwartz, R., Bartley, D., & Schlecht, P. (2002). Health effects of occupa­
tional exposure to respirable crystalline silica.
Abbasi, B., Wang, X., Chow, J. C., Watson, J. G., Peik, B., Nasiri, V., . . . & Elahifard, M. (2021).
Review of respirable coal mine dust characterization for mass concentration, size distribution and
chemical composition. Minerals, 11(4), 426.
Trechera, P., Moreno, T., Córdoba, P., Moreno, N., Zhuang, X., Li, B., . . . & Querol, X. (2020). Min­
eralogy, geochemistry and toxicity of size-segregated respirable deposited dust in underground coal
mines. Journal of Hazardous Materials, 399, 122935. [10] E.
Trechera, P., Moreno, T., Córdoba, P., Moreno, N., Amato, F., Cortés, J., . . . & Querol, X. (2021). Geo­
chemistry and oxidative potential of the respirable fraction of powdered mined Chinese coals. Science
of the Total Environment, 800, 149486.
Shekarian, Y., Rahimi, E., Shekarian, N., Rezaee, M., & Roghanchi, P. (2021). An analysis of contribut­
ing mining factors in coal workers’ pneumoconiosis prevalence in the United States coal mines, 1986–
2018. International Journal of Coal Science & Technology, 8, 1227–1237.
Brown, J. S., Gordon, T., Price, O., & Asgharian, B. (2013). Thoracic and respirable particle definitions
for human health risk assessment. Particle and fibre toxicology, 10, 1–12.
Marques, M. R., Loebenberg, R., & Almukainzi, M. (2011). Simulated biological fluids with possible
application in dissolution testing. Dissolution Technol, 18(3),15–28.
L. E. Kilpatrick L, and Harris M, (2011). Cytokines and Inflammatory Response in the Fetus and Neo­
nate, Fetal and Neonatal Physiology, Elsevier, 2011, pp. 1652–1671.
Moreno, T., Trechera, P., Querol, X., Lah, R., Johnson, D., Wrana, A., & Williamson, B. (2019). Trace
element fractionation between PM10 and PM2. 5 in coal mine dust: Implications for occupational
respiratory health. International Journal of Coal Geology, 203, 52–59.
Lin, S., Liu, Z., Zhao, E., Qian, J., Li, X., Zhang, Q., & Ali, M. (2019). A study on the FTIR spectra of
pre-and post-explosion coal dust to evaluate the effect of functional groups on dust explosion. Process
Safety and Environmental Protection, 130, 48–56.

304
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Comparison of respirable coal and silica dust monitoring systems

for underground mining applications

A. Medina, A. Vanegas, E. Madureira & P. Roghanchi

New Mexico Institute of Mining and Technology, Socorro, NM, USA

R. Rajapaksha, L. Uecker, T. Rawson & C. Harb

Ring-IR, Albuquerque, NM, USA

ABSTRACT: Respirable coal mine dust (RCMD) and respirable crystalline silica (RCS) are
hazardous materials that must be controlled in underground coal mines to reduce miners’ expos­
ure. For this, monitoring systems are used to identify high dust concertation spots and apply crit­
ical controls. Using RingIR’s patented rtCRDS technology, a portable rtCRD spectrometer was
built to use as a dust monitor in mining environments to measure coal and silica respirable dust
levels rapidly and accurately. This research compares a PDM3700, a TSI aerodynamic particle
sizer 3321, an SPS30 particulate matter sensor, and NIOSH methods for silica analysis with the
new dust monitor developed by Ring-IR to test its functionality in simulated underground coal
mine conditions. Simulated mine conditions are created inside a dust chamber using a Topas
solid aerosol generator for dust supply and a humidity control station for humidity conditions.

1 INTRODUCTION

Mine workers encounter several health and safety hazards during daily job activities, increas­
ing the risk of illnesses and injuries (Akintoye & MacLeod, 1997; Chu & Muradian, 2016;
Domínguez et al., 2019; Mitchell et al., 1998; Sanmiquel et al., 2015; Stewart, 2020). One of
these dangerous situations is exposure to respirable dust particles in working areas. Dust is
generated daily by underground and surface mining activities, such as grinding, cutting, and
transporting materials. Nonetheless, underground mines’ conditions could be more dangerous
due to limited ventilation, confined spaces, and other environmental conditions (Jiskani et al.,
2020; Strzemecka et al., 2019; Sunkpal et al., 2018; Yang et al., 2021). Consequently, there are
more chances of finding high dust concentration areas in underground coal mines, which
leads to the need for more engineering and administrative controls to keep a safe workplace.
In underground coal mines, dust particles’ respirable fraction is called respirable coal mine dust
(RCMD), which mineral content depends on the specific geology of the deposit, but the most
common minerals associated include quartz, illite, calcite, pyrite, and kaolinite (National Institute
for Occupational Safety and Health (NIOSH), 1995). It can also include diesel particulate matter
(DPM), among other materials (Pan et al., 2021). Some of these materials, such as crystalline silica
and DPM, are carcinogens (Calas et al., 2017). In addition, respirable crystalline silica (RCS) is
a critical component of RCMD and its content should be considered in workers’ exposure limits.
Long-time exposure to RCMD and RCS can lead to multiple lung diseases, including coal
worker’s pneumoconiosis (CWP), mixed dust pneumoconiosis, silicosis, dust-related diffuse
fibrosis (DDF), and progressive massive fibrosis (PMF) (Centers for Disease Control and Pre­
vention, 2002; Cullinan et al., 2017; NIOSH, 2021; Salinas et al., 2022; Shekarian et al., 2021).
Dust is controlled in mines through several mitigation techniques involving water spray systems,
filters, and equipment to reduce dust concentrations in working areas. OSHA and MSHA defined
the permissible exposure limits (PEL) to control exposure in work environments, also known as

DOI: 10.1201/9781003429241-32

305
dust standards (Centers for Disease Control and Prevention, 2022). If corrections action, engineer­
ing, and administrative controls cannot eliminate the hazard or reduce concentrations below PEL,
PPEs are needed. Having all this in mines, continuous personal dust monitors (CPDM) will finally
alert miners to move from high concentration zones and avoid breathing these particles. Continu­
ous dust monitoring across the mines can significantly help in the decision-making to reduce
miners’ exposure in working areas.
The approved CPDM device to measure RCMD in underground coal mines is the PDM3700,
which can provide dust concentration in near real-time. However, it cannot distinguish difference
components in RCMD, and therefore it does not give silica content in RCMD samples. Coal mine
dust personal sampling units or equivalent are approved devices commonly used to measure RCMD
and RCS. The process of measuring RCS can take more time and involve lab techniques that require
taking the sample off-site, which has led to the need to create a CPDM for silica analysis.
This research aims to create laboratory conditions to simulate underground coal mine
environments and test monitoring systems to measure RCMD and RCS. Available technology
in the market will be used to test the effectiveness of monitoring systems under different
humidity conditions, and a near real-time dust monitor developed by Ring-IR will be tested to
measure RCMD and RCS using cavity-enhanced direct absorption methods.

2 EQUIPMENT

The development of this project included two main parts before experimentation. The first
was to create a platform to simulate underground coal mine conditions, modifying humidity
and concentration inside a dust chamber. The second was to complete the Ring-IR dust moni­
tor development for testing in the platform. The experimental part of this research will involve
using the following equipment.

2.1 Humidity Control Station (HCS)

The humidity control station (HCS) shown in Figure 1 is an in-house-made instrument that pro­
duces relative humidity (RH) values from 10 % to 80% in the dust chamber. The system has
a supply of compressed air and is portable if it needs to be moved. The air initially passes through
a water and oil filter to reduce contaminations. After that, the air is passed through a moisture
trap, considerably reducing moisture content for dry testing conditions. If humid air is needed, the
next step is passing the air through water bubblers. Also, when higher humidity values want to be
reached, the water bubbler is wrapped with a heating band that increases the temperature of the
water and produces faster high humidity conditions inside the chamber.

Figure 1. Humidity control station.

306
2.2 PDM3700
The personal dust monitor developed by Thermo Fisher is an approved device to measure
RCMD in underground coal mines. It will be used as a reference for comparison when testing
the Ring–IR dust monitor.

2.3 TSI Aerodynamic Particle Sizer (APS) 3321

The APS 3321 could be used to measure aerodynamic particle size and particle concentration
inside the chamber. Also, this device can be used to confirm that the prepared samples are
between the allowable limits for the experiments once they are airborne inside the chamber.

2.4 Sensirion SPS30 particulate matter sensor

The sensor is a low-cost PM monitor with multiple uses in some industries. This low-cost
sensor can provide PM concentration values in a controlled environment inside the dust
chamber for research purposes. The device can measure particles in the air as soon as it starts
working; it is very small and provides continuous readings for the needed experiment time.
Multiple sensors can be used simultaneously from all corners of the chamber to get an average
reading of particle concentration in the air.

2.5 Topas solid aerosol generator SAG 410 U

This device can work with different dust materials in a range of sizes between 0 to 100
microns. It will produce the needed airborne particles inside the chamber so that the other
devices can read more accurate values. Usually, for experiments inside dust chambers, the
feed rate is used in very low percentages since the instrument can produce high concentrations
in short periods.

2.6 Coal mine dust personal sampling unit or equivalent

It consists of an ELF Escort pump, a 10-mm Dorr-Oliver cyclone, and a 37-mm 5-µm PVC
filter. The unit will be used to measure silica content through NIOSH 7500, NIOSH 7603, and
NIOSH FAST methods.

2.7 Ring- IR dust monitor prototype instrument

The main goal of this project is to test the Ring-IR dust monitor prototype instrument and
compare it with monitoring systems to measure RCMD and RCS. The device is shown in
Figure 5 and it was created to monitor RCMD and RCS in near real-time.

3 DUST CHAMBER

A customized static dissipative PVC dust chamber was created in collaboration with Terra
Universal Inc. and will work as the testing body to simulate underground coal mine condi­
tions. Dimensions are 35” L x 24” W x 25” H with a total volume of 12.5 ft3. The chamber
has a door to access it, openings to introduce tubing from equipment, and a removable sheet
to introduce more extensive equipment. The chamber was designed to include multiple acces­
sories for the intended experiments, such as a transfer chamber or airlock to pass equipment
or instruments during testing, automatic RB valves to maintain positive pressure inside,
a humidity module that works with the HCS to keep the desired RH values, a differential
pressure gauge, and gloves to access the chamber during experiments. Also, it includes
a recirculating system with a filter for air cleaning. The final product can be seen in Figure 2.

307
Figure 2. Dust chamber.

4 RING-IR DUST MONITOR

The rtCRDS is a cavity-enhanced direct absorption method. For this cavity-enhanced

method, a high-resonant optical cavity was built using four mirrors: two input coupling mir­
rors and two highly reflective mirrors (Figure 3). Quantum cascade lasers (QCLs) are used as
the light source. These mirrors are optically aligned in a way to generate a ringdown signal
which is generated due to the resonance of light inside the optical cavity. The CRD signal
shows exponential decay because the coherent light transmitted by the optical cavity decays
exponentially as e^((-t)/τ), where τ is the characteristic ringdown time of radiation inside the
cavity at time t. Optical alignment of the cavity is validated and monitored by generating
a CRD signal and calculating the ringdown time constant, τ.

Figure 3. Optical cavity and optical layout.

The CRD signal from MCT detector is fed into a Field Programmable Gate Array (FPGA) based
lock-in amplifier and hence respective absorption spectrum will be generated. Preliminary tests were
done using a coal dust sample (New Mexico mine, <104μm), and a silica dust sample (Australian
mine, <10μm) for this analysis. The dust sample was measured in a dust sampling jar and the cap was
replaced. Cyclone was set up with inlet and outlet tubing as shown in Figure 4. The inlet tubing was
inserted into the sampling jar from the opening of the cap and this tube was set near the upper portion
of the jar without any contact with dust in the bottom of the jar. In this way, only small size particles
will get through the instrument. The outlet tubing was connected to the inlet of the optical cavity and
the cavity outlet was connected to a particle filter inlet. Then the particle filter outlet was connected to
the pump. The sampling jar was continuously shaken throughout the experiment.

308
Figure 4. Sampling jar attachment to the spectrometer.

Figure 5. Ring-IR dust monitor prototype instrument.

After each dust sample, room air was pumped through the system for one minute and
checked for the existence of dust particles by capturing a spectrum. That spectrum was com­
pared against the initial background spectrum. The next sample was carried out only when the
system was free from dust particles and the background spectrum was matched to the initial
background spectrum. Dust samples were tested from lowest to highest concentrations.

309
5 DISCUSSIONS

The preliminary dust results obtained for silica dust samples are shown in Figure 6. This
data demonstrates the potential of rtCRDS as a dust monitoring in mine environments.
Further testing is underway with the dust chamber simulating underground coal mine
conditions.

Figure 6. RING-IR dust preliminary results.

Simulated underground coal mine conditions were created inside a dust chamber to com­
pare monitoring systems to measure RCMD and RCS. The humidity control station (HCS)
constantly supplied the dust chamber to produce low or high RH values. In addition, an aero­
sol generator was used to keep airborne particle concentrations during experiments.
The HCS was initially tested using a model with a volume of 0.7 ft3 and a bigger chamber
with a volume of 18 ft3; this last one was different from the one shown in Figure 2. The model
was a smaller testing body and took about 1 to 5 minutes to reach very dry or humid values.
On the other hand, experiments in the chamber usually took more than an hour to get 80%
RH conditions. After reaching desired RH conditions, humidity maintains very similar for
more than 1 hour inside the chamber.
For dust concentration testing, dust was conducted from the aerosol generator to the
chamber with tubing to the back top middle (TM) area to disperse from top to bottom, and
an APS 3321 was used to measure concentration values. Area distributions are shown in
Figure 7. Experiments were performed in six internal areas to better understand concentra­
tion conditions inside the chamber, measuring changes in particle concentration with differ­
ent materials and various feed rates from the aerosol generator. After analyzing results and
comparing changes over different experimental scenarios, the bottom middle (BM) area
was defined as the more reliable internal area to do readings during experiments with moni­
toring systems due to consistency in readings and similarity with average and medium
values. Figure 8 shows the concentration over time graph for each internal area. After
twelve initial tests to define the best experimental setup, a final test was completed with
three reading intervals of 10 minutes for each area using coal mine dust to measure the
total concentration inside the chamber in number of particles per cubic centimeter
(#pt /cm³).

310
Figure 7. Internal area distributions dust chamber.

Figure 8. Time vs concentration inside the chamber on different positions.

6 FUTURE WORK

More research should be conducted under several scenarios inside the dust chamber, changing
humidity and concentration values to test the Ring-IR dust monitor response to typical
changes in underground coal mines. The next set of experiments will compare the PDM3700
and the SPS30 sensor with values from the Ring-IR device. Once consistent results have been
found measuring RCMD, the following experiments will involve using a coal mine dust per­
sonal sampling unit to measure silica content using NIOSH 7500, NIOSH 7603, and NIOSH
FAST methods and compare results with the Ring-IR dust monitor.

REFERENCES

Akintoye, A. S., & MacLeod, M. J. (1997). Risk analysis and management in construction. International
Journal of Project Management, 15(1), 31–38. https://doi.org/10.1016/S0263-7863(96)00035-X

311
Calas, A., Uzu, G., Martins, J. M. F., Voisin, D., Spadini, L., Lacroix, T., & Jaffrezo, J.-L. (2017). The
importance of simulated lung fluid (SLF) extractions for a more relevant evaluation of the oxidative
potential of particulate matter. Scientific Reports, 7(1), 11617. https://doi.org/10.1038/s41598-017-11979-3
Centers for Disease Control and Prevention. (2022, October 30). Methods: Exposure. https://Wwwn.Cdc.
Gov/Eworld/Appendix/Exposure.
Centers for Disease Control and Prevention, N. I. for O. S. and Health. (2002). Health Effects of Occupa­
tional Exposure to Respirable Crystalline Silica.. Retrieved from https://www.Cdc.Gov/Niosh/Docs/
2002-129/.
Chu, C., & Muradian, N. (2016). Safety and environmental implications of coal mining. International
Journal of Environment and Pollution, 59 (2/3/4),250. https://doi.org/10.1504/IJEP.2016.079899
Cullinan, P., Muñoz, X., Suojalehto, H., Agius, R., Jindal, S., Sigsgaard, T., Blomberg, A., Charpin, D.,
Annesi-Maesano, I., Gulati, M., Kim, Y., Frank, A. L., Akgün, M., Fishwick, D., de la Hoz, R. E., &
Moitra, S. (2017). Occupational lung diseases: from old and novel exposures to effective preventive
strategies. The Lancet Respiratory Medicine, 5(5), 445–455. https://doi.org/10.1016/S2213-2600(16)30424-6
Domínguez, C. R., Martínez, I. V., Piñón Peña, P. M., & Rodríguez Ochoa, A. (2019). Analysis and
evaluation of risks in underground mining using the decision matrix risk-assessment (DMRA) tech­
nique, in Guanajuato, Mexico. Journal of Sustainable Mining, 18(1), 52–59. https://doi.org/10.1016/J.
JSM.2019.01.001
Jiskani, I. M., Han, S., Shahani, N. M., Ali, M., Dianat, I., & Chalgri, S. R. (2020). Evaluation of phys­
ical and environmental working conditions of underground coal mines within the framework of
ergonomics. International Journal of Mining and Mineral Engineering, 11(3), 240. https://doi.org/
10.1504/IJMME.2020.109639
Mitchell, R. J., Driscoll, T. R., & Harrison, J. E. (1998). Traumatic work-related fatalities involving
mining in Australia. Safety Science, 29(2), 107–123. https://doi.org/10.1016/S0925-7535(98)00012-5
National Institute for Occupational Safety and Health (NIOSH). (1995). Criteria for a Recommended
Standard. Occupational Exposure to Respirable Coal Mine Dust.
NIOSH. (2021). Best practices for dust control in coal mining, second edition. https://doi.org/10.26616/
NIOSHPUB2021119
Pan, L., Golden, S., Assemi, S., Sime, M. F., Wang, X., Gao, Y., & Miller, J. (2021). Characterization of
Particle Size and Composition of Respirable Coal Mine Dust. Minerals, 11(3), 276. https://doi.org/
10.3390/min11030276
Salinas, V., Das, M., Jacquez, Q., Camacho, A., Zychowski, K., Hovingh, M., Medina, A.,
Rubasinghege, G., Rezaee, M., Baltrusaitis, J., Fairley, N., & Roghanchi, P. (2022). Characterization
and Toxicity Analysis of Lab-Created Respirable Coal Mine Dust from the Appalachians and Rocky
Mountains Regions. Minerals, 12(7), 898. https://doi.org/10.3390/min12070898
Sanmiquel, L., Rossell, J. M., & Vintró, C. (2015). Study of Spanish mining accidents using data mining
techniques. Safety Science, 75, 49–55. https://doi.org/10.1016/J.SSCI.2015.01.016
Shekarian, Y., Rahimi, E., Rezaee, M., Su, W.-C., & Roghanchi, P. (2021). Respirable Coal Mine Dust:
A Review of Respiratory Deposition, Regulations, and Characterization. Minerals, 11(7), 696. https://
doi.org/10.3390/min11070696
Stewart, A. G. (2020). Mining is bad for health: a voyage of discovery. Environmental Geochemistry and
Health, 42(4), 1153–1165. https://doi.org/10.1007/s10653-019-00367-7
Strzemecka, J., Goździewska, M., Skrodziuk, J., Galińska, E., & Lachowski, S. (2019). Factors of work
environment hazardous for health in opinions of employees working underground in the ‘Bogdanka’
coal mine. Annals of Agricultural and Environmental Medicine, 26(3), 409–414. https://doi.org/
10.26444/aaem/106224
Sunkpal, M., Roghanchi, P., & Kocsis, K. C. (2018). A Method to Protect Mine Workers in Hot and Humid
Environments. Safety and Health at Work, 9(2), 149–158. https://doi.org/10.1016/j.shaw.2017.06.011
Yang, L., Birhane, G. E., Zhu, J., & Geng, J. (2021). Mining Employees Safety and the Application of
Information Technology in Coal Mining: Review. Frontiers in Public Health, 9, 709987. https://doi.
org/10.3389/fpubh.2021.709987

312
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Evaluation of different surfactants’ performance in varying coal

dust concentration through logistic regression analysis

Z. Zhao, A. Ghosh & P. Chang

WA School of Mines: Minerals, Energy and Chemical Engineering, Curtin University, Kalgoorlie, WA,
Australia

Y. Liu
Henan Joint International Research Laboratory of Mine Safety and Occupational Health, Henan Polytechnic
University, Jiaozuo, Henan, China

ABSTRACT: Coal dust causes a range of diseases and health problems worldwide, such as
Coal Workers’ Pneumoconiosis. Numbers of static studies showed that surfactants could
increase the coal dust wettability effectively in water spraying. However, coal dust does not
remain in a static state in underground mines. Therefore, studies should be conducted in
a dynamic state where coal dust and surfactant droplets will get a shorter contact time. In this
study, wind tunnel tests have been conducted to capture the dynamic process. Logistic regres­
sion modelling has been done to evaluate the suppression performance in terms of efficiency
and quantity of dust suppressed with respect to three dependent variables: surfactant types,
surfactant quantity, and coal dust concentration. The results showed that at low coal dust
concentrations, the suppression efficiency is greater compared to high dust concentrations.
Moreover, the surfactant quantity of 0.05 % to 0.20 % is more efficient than above. Although
the surfactant type did not emerge as significant when low and high dust concentrations were
combined, it becomes significant at low dust concentrations where SDS and CTAB are more
efficient than SDBS and TX100. The results will help in selecting surfactants with the appro­
priate quantity in different coal dust concentrations.

1 INTRODUCTION

Coal dust is a fine powdered form of coal, which is created by the crushing, grinding, or pulver­
izing of coal during mining, transportation, or by mechanically handling coal. It is a form of
fugitive dust and poses the acute hazard of forming an explosive mixture in the air and the
chronic hazard of causing pulmonary illness in people who inhale excessive quantities of it. Coal
workers’ pneumoconiosis (CWP), also known as the black lung disease, is one of the most
famous occupational diseases, which is usually caused by overexposure to respirable coal dust.
In 2012, there were 27,420 cases that were reported as occupational diseases globally, and the
CWP occupied 45.24 % of that (Organization, 2013). In 2013, CWP has resulted in 25,000
deaths globally (Abubakar et al., 2015). In the U.S., the prevalence of CWP decreased from 3.3 %
to 0.3 % from 1970 to 1998, then rose to 3.3 % as of 2012 (Blackley et al., 2014, Blackley et al.,
2016). In the United Kingdom, although the prevalence of pneumoconiosis decreased from 12 %
to 0.2 % during 1994-1997, it increased to 0.8 % from 1998 to 2000 (Scarisbrick and Quinlan,
2002). In Australia, 20 new cases of CWP had shown within only a fortnight in 2019, and 31 cases
were confirmed in 2020 (Hendry, 2019, Burt; and McGhee, 2020). All the data illustrated
above indicated that the CWP has shown a resurgence tendency. Hence, an effective method for
controlling the coal dust in coal mines is a necessity for the prevention of CWP occurrence.

DOI: 10.1201/9781003429241-33

313
 Generally, there are three methods for controlling coal dust for underground mining,
including ventilation, foam technology, and water spray (Chang et al., 2019, Ren et al., 2012,
Rider and Colinet, 2007, Chang et al., 2020). Water spray is still a primary method that has
been undertaken in the past decades. This method is readily applicable to mining industries,
no customized equipment is required, and the cost of consumables and equipment is relatively
low. Although the efficiency of untreated water is usually lower than expected, surfactants can
dramatically increase the efficiency of coal dust suppression. Surfactants usually contain both
hydrophobic groups and hydrophilic groups. While surfactants diffuse in water, the hydro­
philic groups remain in the water phase, and the hydrophobic groups extend out of the bulk
water phase for capturing (Xu et al., 2018, Zhao et al., 2022). In the water spray system, sur­
factant solution will be sprayed out into the air then hydrophobic groups will catch dust par­
ticles. Previous studies claimed that the surfactant-aided solution could dramatically reduce
the coal dust by up to 90 % (Chander et al., 1990, Kilau, 1996, Tien and Kim, 1997). For
example, in Zhou et al.’s study, it was shown that adding sodium dodecyl benzenesulfonate
(SDBS) into untreated water can increase the suppression efficiency for total dust from 55 %
to 76 % in a field test (Zhou et al., 2019). In another study by the same research group, SDBS
was evaluated with additives to improve performance (Zhou et al., 2017). It was found that
the suppression efficiency of both total dust and respirable dust can be improved from 53 %
and 38 % to 75 % and 70 %, respectively. Similarly, another compounding surfactant was also
evaluated by a field test in Wang et al.’s study (Wang et al., 2019). It was claimed that the
suppression efficiency of total dust could be increased from 50 % to 90 %, while the same of
respirable dust can be dramatically improved from 39 % to 88 %. Therefore, the evaluation of
surfactants is significant for improving the suppression efficiency.
The evaluation methods of surfactants can be divided into two categories, including static
tests and dynamic tests. In previous studies, static tests, such as the surface tension test and
the sink test, generally measure interactions between particles and surfactant solutions, such
as the surfactant adsorption ability and coal wettability (Xu et al., 2018, Zhao et al., 2021,
Chang et al., 2021). In contrast, dynamic tests, such as the wind tunnel test and the field test,
can directly test the results of coal dust suppression and then analyze the efficiency of suppres­
sion. Moreover, compared to static tests, where the coal particles and the solution are firmly
contacted over the test period, dynamic tests take into consideration of particle contacts, colli­
sions, and fluid dynamics (Xu et al., 2018). Although the result of dynamic tests is more reli­
able, however, many studies conducted static tests the most when evaluating surfactants due
to time-saving and economic reasons. For instance, Chen et al. evaluated SDBS in static tests
such as the sink test, the surface tension test, and the zeta potential test (Chen et al., 2017).
The result showed that SDBS ranging from 0.2 % to 0.4 % had the best performance in static
tests. Another study by Wang et al. evaluated several surfactants by using static tests, such as
the surface tension test and the contact angle test, and the wetting agent A was optimized with
the concentration and thus had excellent performance in both tests (Wang et al., 2014). Add­
itionally, in Chang et al.’s study, three surfactants, including SDBS, Triton X-100 (TX100),
and cetyl trimethylammonium bromide (CTAB), were tested by the surface tension test and
the sink test. It was found that TX100 performed the greatest in terms of coal dust wettability,
while CTAB performed the weakest.
Despite the fact that static tests usually are more convenient than conducting dynamic tests,
static tests and dynamic tests sometimes may provide discrepant results. For example, the per­
formance of TX100 and sodium dodecyl sulphate (SDS) was evaluated and compared by
static tests in Glanville and Haley’s study (Glanville and Haley, 1982). It was concluded that
SDS performed better and showed a higher wettability than TX100. Nevertheless, another
study of dynamic tests illustrated a different result, that TX100 could improve the suppression
efficiency greater than SDS in dynamic tests (Tessum and Raynor, 2017). Moreover, the coal
dust concentration underground usually is fugitive and may differ with coal mines. Although
static tests can easily evaluate surfactants with different types or concentrations, the coal dust
concentration usually cannot be taken into consideration. Thus, the wind tunnel test was con­
ducted in this study to evaluate surfactants with different coal dust concentrations. This paper
aims to evaluate the suppression efficiency of surfactants with different surfactant

314
concentrations and initial coal dust concentrations. The significance of surfactant type, surfac­
tant concentration, and initial coal dust concentration will be analyzed and discussed by
a factorial experiment design. A wind tunnel apparatus was applied to conduct all experiments
in this study. Four surfactants were evaluated, including CTAB, SDS, SDBS, and TX100.
Seven concentrations of surfactants were tested, including 0.00 %, 0.05 %, 0.10 %, 0.15 %,
0.20 %, 0.25 %, and 0.30 %. Two initial coal dust concentrations were selected, including Low
(13.35 mg/m3 (95 % CI: 12.21-14.50)) and High (101.62 (95 % CI: 99.23-104.02)). The findings
of this study can help coal dust control for industries in terms of the determination and opti­
mization of surfactants.

2 METHOD

2.1 Experimental apparatus and procedures

In this study, all the tests would be conducted through the wind tunnel apparatus, which is
shown in Figure 1. The dimensions of this apparatus are 0.5 m (H) × 0.5 m (W) × 4.50 m (L).
Coal dust, which is prepared in advance of the experiment, would be delivered through the dust
generator. The dust concentration can be controlled based on the purpose. The exhaust fan can
provide airflow within the tunnel at 0.68m/s and collect dust. Solutions would be sprayed into
the tunnel through the nozzle and powered by the water pump at a 5L/min flow rate. The dust
concentration is measured by a portable dust monitor (model: DustTrak DRX Aerosol Monitor
8534), which can detect the particle concentration ranging from 0.001 mg/m3 to 150 mg/m3. The
procedure of one wind tunnel test includes several steps. Firstly, the coal dust and surfactant
solutions were prepared to regard the size, type, and concentration based on the purpose of this
study. In this experiment, sub-bituminous coal was used for testing throughout the whole
experiment. The exhaust fan and dust generator were kept running until the airflow achieved
a steady state. Then coal dust will be put into the dust generator. The dust concentration should
be measured after the airflow achieves a steady state. After measuring the dust concentration,
surfactant solutions would be sprayed into the wind tunnel by the spray system. The dust con­
centration should be measured again after the airflow keeps steady. The efficiency of this dust
suppression can be calculated by the following equation (Zhou et al., 2017):

Where the R represents the suppression rate, C (Initial) and C (After) represent the coal
dust concentration prior to and after the surfactant solution action, respectively.

2.2 Experiment design

This experiment has three factors, including dust concentration, surfactant type, and surfac­
tant concentration, based on the factorial design. The three factors would be settled as
follows:
a. Dust concentration (2 levels): Low (13.35 mg/m3 (95 % CI: 12.21-14.50)) and High
(101.62 mg/m3 (95 % CI: 99.23-104.02)).
b. Surfactant type (4 levels): SDS, SDBS, CTAB, and TX100.
c. Surfactant solution concentration (7 levels): 0.00 %, 0.05 %, 0.10 %, 0.15 %, 0.20 %, 0.25
%, and 0.30 %.
This three-factor factorial experiment runs two replicates entirely with 84 runs totally. This
experiment was conducted in an air-conditioned laboratory, where the temperature was kept
stable.

315
Figure 1. Wind tunnel apparatus (1) dust generation system, (2) spray system, (3) monitor (4) dust col­
lection system.

3 RESULTS

Table 1 and Figure 2 show the data summary of this experiment. Logistic regression is a kind
of data analysis model that can evaluate data from different factors. In this study, this method
can evaluate the relations between the suppression efficiency and surfactant type, surfactant
concentration, and coal dust concentration, respectively. The overall logistic regression results
are shown in Table 2.

Table 1. Summary of the coal dust efficiency.

Surfactant types
SDBS SDS TX100 CTAB
Initial coal dust Initial coal dust Initial coal dust Initial coal dust
concentration concentration concentration concentration
Solution levels (wt. %) Low High Low High Low High Low High
36.51 % 19.38 % 36.51 % 19.38 % 36.51 % 19.38 % 36.51 % 19.38 %
0
42.67 % 18.83 % 42.67 % 18.83 % 42.67 % 18.83 % 42.67 % 18.83 %
40.37 % 22.94 % 50.81 % 29.75 % 34.48 % 23.34 % 59.33 % 34.91 %
0.05
36.74 % 31.82 % 39.27 % 19.59 % 41.06 % 14.63 % 60.58 % 34.09 %
35.25 % 21.93 % 58.97 % 33.20 % 29.67 % 19.08 % 63.35 % 35.43 %
0.10
43.94 % 29.02 % 50.62 % 37.67 % 25.14 % 22.28 % 57.57 % 20.42 %
41.35 % 29.80 % 42.69 % 31.57 % 40.00 % 34.79 % 35.93 % 28.92 %
0.15
34.78 % 19.65 % 48.48 % 40.11 % 43.50 % 25.67 % 29.42 % 27.21 %
40.78 % 28.12 % 60.55 % 32.90 % 33.97 % 32.61 % 44.54 % 19.52 %
0.20
47.93 % 32.82 % 55.35 % 24.90 % 43.61 % 27.01 % 49.12 % 14.40 %
28.13 % 24.25 % 39.78 % 20.49 % 41.37 % 28.83 % 29.77 % 24.54 %
0.25
30.35 % 17.48 % 51.25 % 18.02 % 34.00 % 24.66 % 25.66 % 32.26 %
41.52 % 25.57 % 59.81 % 31.73 % 24.44 % 35.31 % 44.64 % 24.95 %
0.30 29.20 % 28.83 % 46.55 % 18.02 % 30.86 % 23.80 % 33.53 % 23.05 %

316
Figure 2. Surfactant concentration type interaction.

Table 2. Analysis results of logistic regression.

95 % C.I. for
EXP(B)
B S.E. Wald Df Sig. OR Lower Upper

Coal dust concentration 4.356 1.135 14.733 1 0.000 77.934 8.428 720.617
Surfactant type .928 .813 1.303 1 0.254 2.530 .514 12.447
Surfactant concentration 2.430 1.116 4.739 1 0.029 11.362 1.274 101.337
constant -3.857 1.256 9.425 1 0.002 .021

As resulted in Table 2, the significance of coal dust concentration ranked at the top among
the three factors, which was less than 0.001. The Odd Ratio (OR) was 77.93 with 95 % Confi­
dence Interval (CI) between 8.43 and 720.62, which indicated that the suppression efficiency
can be improved significantly when the coal dust concentration decreased. In addition, the sig­
nificance of the concentration of surfactants was 0.029, which was the second significant
factor. The OR was 11.362 with 95 % CI between 1.274 and 101.337, showing that the sup­
pression efficiency can be easily impacted by the surfactant type. The surfactant concentra­
tions were divided into two groups, including a lower group (0.05 %-0.15 %) and a higher
group (0.20 %-0.30 %). As a result of this analysis, the surfactant solution in the lower group
was more efficient than the higher group. Figure 2 presents the average suppression efficiency
of four surfactants with different concentrations. It was found that 0.05 % CTAB had the
highest average suppression efficiency, while 0.10 % TX100 had the lowest. SDS had the best
overall suppression efficiency among all concentrations. For a better result of the analysis, the
data can be divided and analyzed into two groups, including the low initial coal dust concen­
tration group and the high initial coal dust concentration group. The statistical analysis of
both these two groups is shown in Table 3 and Table 4.

317
Table 3. Analysis results at low initial coal dust concentration.
95 % C.I. for EXP(B)
B S.E. Wald Df Sig. OR Lower Upper

Surfactant type 3.383 1.252 7.303 1 .007 29.472 2.533 342.875

Surfactant concentration -2.281 1.243 3.366 1 .067 .102 .009 1.168
constant -1.219 .807 2.283 1 .131 .295

Table 3 shows the logistic regression results of the low initial coal dust concentration group.
Surfactant types were classified by type A (CTAB and SDS) and type B (SDBS and TX100).
As observed, the surfactant type significantly impacted the result of suppression efficiency due
to the conspicuous significance of 0.007, which was less than 0.05. Type A (CTAB and SDS)
had a higher effect on suppression efficiency than type B (SDBS and TX100). However, the
surfactant concentration only had a less relevant coefficient against suppression efficiency,
although it had a more considerable significance towards the entire experiment. The signifi­
cance was 0.067, and the 95 % CI was from 0.009 to 1.168, which indicated that there was no
obvious evidence to prove the relationship between the concentration of surfactants and the
suppression efficiency. Figure 3 showed the average suppression efficiency at the low initial
coal dust concentration. Similarly, SDS had the best overall suppression efficiency among
these concentrations. Interestingly, CTAB at 0.05 % and 0.10 % concentrations had great per­
formance than at higher concentrations. For other surfactants, the average suppression effi­
ciency was similar at each concentration. Hence, at lower initial coal dust concentrations, the
surfactant type had the dominant impact on the suppression efficiency of coal dust.

Figure 3. Surfactant concentration-type interaction at low initial coal dust concentration.

Table 4 shows the logistic regression results of the high initial coal dust concentration
group, and the interaction between surfactant concentration and type is illustrated in
Figure 4. It is noteworthy that both surfactant type and surfactant concentration had less cor­
relation towards the suppression efficiency. The significances of surfactant type and surfactant
concentration were 1.000 and 0.133, while the 95 % CI of both makeups were from 0.213 to
4.7 and 0.693 to 16.022, respectively. This could occur mainly because of insufficient surfac­
tant, which cannot suppress a higher volume of coal dust. Therefore, it is necessary to analyze

318
the impacts of the surfactant type and the surfactant concentration towards the quantity of
coal dust suppression instead of the efficiency rate, which is shown in Table 5.

Table 4. Analysis results at high initial coal dust concentration.

95 % C.I. for
EXP(B)
B S.E. Wald Df Sig. OR Lower Upper

Surfactant type .000 .790 .000 1 1.000 1.000 .213 4.700

Surfactant concentration 1.204 .801 2.259 1 .133 3.333 .693 16.022
constant -.693 .729 .905 1 .341 .500

Figure 4. Surfactant concentration-type interaction at high initial coal dust concentration.

Table 5. Analysis results of suppression quantities.

95 % C.I. for
EXP(B)
B S.E. Wald Df Sig. OR Lower Upper

Surfactant type .000 .839 .000 1 1.000 1.000 .193 5.174

Surfactant concentration 1.887 .858 4.843 1 .028 6.600 1.229 35.438
constant -1.099 .788 1.946 1 .163 .333

Table 5 shows the impacts of the surfactant type and the surfactant concentration towards
the quantity of coal dust suppression. The significance of the surfactant concentration was
0.028, which indicated that the surfactant concentration can significantly impact the quantity
of coal dust suppression. Meanwhile, the expectation of the surfactant concentration was 6.6,
which means the surfactant concentration had a higher suppression efficiency from 0.05 % to
0.15 %.

319
 From the above data analysis and discussion, the suppression efficiency of four common
surfactants was evaluated with different surfactant concentrations and different initial coal
dust concentrations. It was found that, increasing surfactant concentration generally can
improve the suppression efficiency, while the performance of surfactants would be weakened
if the initial coal dust concentration was high. As for the surfactant type, SDS and CTAB had
greater average suppression efficiency than TX100 and SDBS. It is noticeable that SDS
showed a great performance on overall surfactant concentrations, while CTAB performed
better at lower surfactant concentrations. It was suggested that, SDS ranging from 0.10% to
0.20% should be applied for controlling coal dust, because it can significantly increase the sup­
pression efficiency of coal dust in both low and high initial coal dust concentration conditions.
Additionally, CTAB at lower surfactant concentrations could be an alternative option for
controlling coal dust, especially for suppressing coal dust with lower concentrations. Surpris­
ingly, the concentration of surfactants was a crucial factor impacting the suppression effi­
ciency at lower coal dust concentrations. Nevertheless, it was insignificant in improving the
suppression efficiency of the high initial coal dust concentrations. An inadequate amount of
surfactant spraying could be the main reason for this phenomenon. In future studies, the
impact of surfactant spraying rate on the suppression efficiency should be investigated.

4 CONCLUSION

The main target of this study is to evaluate how three factors, including the initial coal dust con­
centration, the surfactant type, and the surfactant concentration, impact the coal dust suppression
efficiency. From the Logistic Regression modelling results, the initial coal dust concentration had
a significant impact overall towards the coal dust suppression efficiency among three factors.
Moreover, the surfactant concentration also showed a dramatic impact, whereas the surfactant
type was relatively less critical towards the suppression efficiency. However, for the lower initial
coal dust concentration, there was a different situation for the surfactant type. The surfactant type
had a higher effect than the surfactant concentration. The CTAB and SDS were more significant
for the coal dust suppression than that of SDBS and TX100. At the higher initial coal dust concen­
tration, it is noteworthy that both the surfactant type and surfactant concentration were less sig­
nificant towards the suppression efficiency. This is probably caused by insufficient surfactant
quantities for dust suppression. Thus, it is crucial to evaluate the quantities of coal dust suppres­
sion rather than efficiency at the higher initial coal dust concentration. The result clearly showed
that the surfactant concentration can substantially affect the suppression quantities of the coal
dust, even though the surfactant type is still of unimportance. In conclusion, the initial coal dust
concentration had the highest impact on coal dust suppression efficiency. At the lower initial coal
dust concentration, the surfactant type showed a higher impact on the suppression efficiency than
that of the surfactant concentration. In contrast, at the higher initial coal dust concentration, the
surfactant concentration can substantially impact the suppression quantities of the coal dust.

REFERENCES

Abubakar, I., Tillmann, T. & Banerjee, A. (2015) Global, regional, and national age-sex specific all-cause
and cause-specific mortality for 240 causes of death, 1990-2013: a systematic analysis for the Global
Burden of Disease Study 2013. Lancet, 385, 117–171.
Blackley, D. J., Crum, J. B., Halldin, C. N., Storey, E. & Laney, A. S. (2016) Resurgence of progressive
massive fibrosis in coal miners—Eastern Kentucky, 2016. Morbidity and mortality weekly report, 65,
1385–1389.
Blackley, D. J., Halldin, C. N. & Laney, A. S. (2014) Resurgence of a debilitating and entirely prevent­
able respiratory disease among working coal miners. American journal of respiratory and critical care
medicine, 190, 708–709.
Burt;, J. & Mcghee, R. (2020) Cases of mine-dust lung disease and silicosis increasingly found in Queens­
land coal mine workers. Retrieved from https://www.abc.net.au/news/2020-02-25/silicosis-and-black-
lung-cases-rise-queensland-workers/11998404.

320
Chander, S., Alaboyun, A. & Aplan, F. (1990) On the mechanism of capture of coal dust particles by
sprays. Proceedings of the Third Symposium on Respirable Dust in the Mineral Industries.
Chang, P., Chen, Y., Xu, G., Huang, J., Ghosh, A. & Liu, W. V. (2019) Numerical study of coal dust
behaviours and experimental investigation on coal dust suppression efficiency of surfactant solution
by using wind tunnel tests. Energy Sources, Part A: Recovery, Utilization, and Environmental Effects,
1–16.
Chang, P., Xu, G., Chen, Y., Ghosh, A. & Moridi, M. A. (2021) Improving coal powder wettability
using electrolyte assisted surfactant solution. Colloids and Surfaces A: Physicochemical and Engineering
Aspects, 613, 126042.
Chang, P., Zhao, Z., Xu, G., Ghosh, A., Huang, J. & Yang, T. (2020) Evaluation of the coal dust sup­
pression efficiency of different surfactants: A factorial experiment. Colloids and Surfaces A: Physico­
chemical and Engineering Aspects, 124686.
Chen, Y., Xu, G. & Albijanic, B. (2017) Evaluation of SDBS surfactant on coal wetting performance
with static methods: Preliminary laboratory tests. Energy Sources, Part A: Recovery, Utilization, and
Environmental Effects, 39, 2140–2150.
Glanville, J. O. & Haley, L. H. (1982) Studies of coal dust wetting by surfactant solutions. Colloids and
Surfaces, 4, 209–212.
Hendry, M. (2019) Black lung advocates say 20 Queenslanders diagnosed with coal dust diseases in a
fortnight. Retrieved from https://www.abc.net.au/news/2019-02-26/dozens-of-new-black-lung-cases-
qld-advocates-say/10851482.
Kilau, H. W. (1996) Suppression of longwall respirable dust using conventional water sprays inoculated
with surfactants and polymers.
Organization (2013) National Health and Family Planning Commission of the People’s Republic of
China. 9th Global Conference on Health Promotion, Shanghai.
Ren, X. W., Wang, M. D., Kang, H. Z. & Lu, X. X. (2012) Engineering case report: A new method for
reducing the prevalence of pneumoconiosis among coal miners: Foam technology for dust control.
Journal of occupational and environmental hygiene, 9, D77–D83.
Rider, J. P. & Colinet, J. F. (2007) Current dust control practices on US longwalls.
Scarisbrick, D. & Quinlan, R. (2002) Health surveillance for coal workers’ pneumoconiosis in the United
Kingdom 1998–2000. Annals of Occupational Hygiene, 46, 254–256.
Tessum, M. W. & Raynor, P. C. (2017) Effects of Spray Surfactant and Particle Charge on Respirable
Coal Dust Capture. Safety and Health at Work, 8, 296–305.
Tien, J. C. & Kim, J. (1997) Respirable coal dust control using surfactants. Applied occupational and
environmental hygiene, 12, 957–963.
Wang, K., Jiang, S., Wu, Z., Shao, H. & Pei, X. (2014) Experimental study on complex wetting agent
enhancing the dusting effect and its application. Electron. J. Geotechnical Eng., 19, 9784–9798.
Wang, X., Yuan, S., Li, X. & Jiang, B. (2019) Synergistic effect of surfactant compounding on improving
dust suppression in a coal mine in Erdos, China. Powder Technology, 344, 561–569.
Xu, G., Chen, Y., Eksteen, J. & Xu, J. (2018) Surfactant-aided coal dust suppression: A review of evalu­
ation methods and influencing factors. Science of the Total Environment, 639, 1060–1076.
Zhao, Z., Chang, P., Xu, G., Ghosh, A., Li, D., & Huang, J. (2021) Comparison of the coal dust suppres­
sion performance of surfactants using static test and dynamic test. Journal of Cleaner Production,
129633.
Zhao, Z., Chang, P., Xu, G., Xie, Q. & Ghosh, A. (2022) Comparison of static tests and dynamic tests
for coal dust surfactants evaluation: A review. Fuel, 330, 125625.
Zhou, Q., Qin, B., Ma, D. & Jiang, N. (2017) Novel technology for synergetic dust suppression using
surfactant-magnetized water in underground coal mines. Process Safety and Environmental Protection,
109, 631–638.
Zhou, Q., Qin, B., Wang, F. & Wang, H. (2019) Experimental investigation on the performance of
a novel magnetized apparatus used to improve the dust suppression ability of surfactant-magnetized
water. Powder Technology, 354, 149–157.

321
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Process ventilation solutions for mitigation of combustible and

non-combustible dust hazards at mining operations

J. Finn
Englo Inc

ABSTRACT: Process ventilation involves using airflow to control dust generated in process
operations. Understanding of process ventilation design practice is important for mining engin­
eers and operations personnel responsible for mitigating the risk of handling combustible and
non-combustible dust hazards. Applications such as crushing, screening, conveying, loading,
and other bulk material handling processes often require use of exhaust hoods and ductwork to
contain and then convey dust emissions to an air cleaning device. This paper is intended to be
a short summary of nine key steps to follow when developing a process dust control solution.

1 INTRODUCTION

Dust emission control at surface and underground mining operations is an important issue for
several reasons. Among these are respirable hazards such as pneumoconiosis (black lung) at
coal handling sites and silicosis where silica dust is present. In addition to these health hazards
there are issues such as accumulation of dust in electrical controls and premature failure of
operating equipment. Combustible dust at coal handling operations presents a huge hazard
that requires specific risk mitigation design practices.
Process dust control solutions include housekeeping cleanup, dust suppression misting
sprays and exhaust ventilation dust control systems. Operations design should also include
consideration for making process modifications to minimize dust generation from the
outset.
Process ventilation guidelines detailed in Industrial Ventilation: A Manual of Recom­
mended Practice, published by ACGIH, are used for engineering of dust control sys­
tems in the USA. NFPA National Fire Protection Association publishes standards for
design of combustible dust control systems. NFPA 652 Standard on the Fundamentals
of Combustible Dust 2016 provides basic principles of and requirements for identifying
and managing the fire and explosion hazards of combustible dusts and particulate
solids.
NFPA 652 requires completion of a Dust Hazard Analysis (DHA) as a necessary starting
point for defining the scope of combustible dust control needed when combustible dusts are
involved. The publication, Guidelines for Combustible Dust Hazard Analysis, by The Center
for Process Safety of the American Institute of Chemical Engineers, is an excellent resource
for this topic.
NIOSH Mining Program Information Circular IC 9532, Best Practices for Dust Control in
Coal Mining, is another important resource that covers a much broader view of mining dust
control than provided by the very limited scope of this paper.
Designing for mining dust control based on the resources and practices mentioned
above ensures that best practice is achieved for safety, maintenance, and effective system
operation.

DOI: 10.1201/9781003429241-34

322
2 DUST MANAGEMENT SOLUTIONS

Approaches for control of fugitive dust include process design change, housekeeping cleanup,
misting suppression, roadway dust containment, unducted exhaust ventilation, machine mounted
control equipment and exhausting with hoods through ductwork to an air control device.
Three general ventilation examples of unducted ventilation are as follows:
Figure 1 shows a wet dust extractor with a water recirculation system. The unducted extractor
was installed at an underground drift mine being developed by drill and blast. The equipment is
placed outby of the blast zone and left off for the blast. Following the blast the extractor is
started and quickly clears the dust and blasting odors by recirculation of ambient air.

Figure 1. Underground limestone mine. Figure 2. Machine mounted dust extractor.

Another version of unducted ventilation is this example of a wet dust extractor mounted on
a roadheader mining machine being readied for delivery to a tunneling project (Figure 2).
Lastly, Figure 3 shows one of two river diversion tunnels developed using roadheaders. In
this project an inadequate ventilation plan resulted in excessive dust and the tunnel develop­
ment was shut down. Exhausting air at the face was increased with larger ductwork connected
to Englo wet dust extractors located at the tunnel entrance. The project went into compliance,
and hundreds of men resumed work resulting in completion in a tight deadline time frame.

Figure 3. Exhaust Ventilation at the face.

3 PROCESS VENTILATION SYSTEM DESIGN

Nine steps for developing a process dust control system using exhaust ventilation systems are
presented in the following sections. They are based in general on design procedure taught in

323
the Industrial Ventilation Manual, although not always in the same order. The examples
given are mostly from evaluation of existing systems. However, the same design approach
applies when developing solutions for new systems.
In addition to the design procedure, it is important to keep in mind that regulatory require­
ments must be considered so that the completed project achieves these objectives.
The Industrial Ventilation manual has extensive information about exhaust design prin­
ciples with comprehensive details about each of the design steps. The description provided
here is a very brief overview of the design approach.

3.1 Identify emission sources and hood designs

Recently a mine operator emailed the author a simple question, “How do I determine the
ACFM?” This is a loaded question, meaning that the question cannot be answered simply.
That is what this presentation is about: understanding information needed to answer the real
question: “How do I develop a successful dust control system for my mining operation?”
Step one is to define each dust emission source and what control approach is needed to con­
tain and capture the dust being generated. Every source has its own characteristics and that is
where the process ventilation engineer starts.
What is the dust being handled? Define those properties that affect design of a dust control
system. Is the dust combustible? How does it react in an air cleaning device? Can a dry system
be used? Or would a wet dust collector work best? What are the health issues involved with
handling the dust? These are among the many important considerations needed for developing
a successful dust control system.
Each dust source must be evaluated to determine how the dust is generated, what capture
hood design works best and what service access needs must be considered. Housekeeping
cleanup of combustible dust accumulations is also an important factor.
Figure 4 shows a wet dust extraction system at a coal mine screening plant. The design
began by considering dust at the vibrating screen decks plus at the conveyor load zones. The
screen decks were enclosed so exhaust hoods were centered on these enclosures. Exhaust
hoods were installed at conveyor load zones below the screens.

Figure 4. Coal mine screening plant.

3.2 Define control volume at each source

What volume of exhaust air is required to contain and convey the captured dust? The control
air volume is determined by considering Displaced Air, Induced Air and In-Draft Air. Other
factors such as temperature and external wind effects may need to be considered.

324
 Displaced air is the volume of air caused by the flow of ore thru a process location. Calcu­
late the volume of bulk material being processed. Multiply density of crushed ore times the
flow rate to get the product displacement volume.
Calculation of displacement for transferring material from a belt conveyor can be deter­
mined by multiplying the ore cross section times belt speed. Once the bulk material displace­
ment volume is determined, a multiplication factor is then used to calculate a displaced air
volume for design. This factor may range from 1.2 to 2.0 depending on process conditions.
The equation, Q = VA, is used in many ways during the design process. Q equals volume of
air expressed as cubic feet per minute, cfm, or cubic meters per second, m3/s. V equals velocity
of the air in feet per minute, fpm, or meters per second, mps.
Figure 5 illustrates a conveyor skirt box exhausted with a tapered hood. Consider a 48”
conveyor running at 500 fpm. The air displacement, Q=VA is 4ʼ width x 1ʼ material average
height x 500 fpm belt speed equals 2000 cfm. The average height factor of one foot is conser­
vative as the actual number will be less than that in most instances. This approach also covers
the minimal factor of Induced Air. Displaced Air is then determined by multiplying the cfm
by a factor ranging from 1.2 or more if needed. Displaced Air equals 3000 cfm.

Figure 5. Exhaust hood for ventilating a conveyor load zone.

Given this information we can determine the exhaust hood design. The first item to decide
is the area of the hood at the capture point. With a belt conveyor the general rule is to keep
the hood inlet face velocity below 500 fpm. Therefore, the hood inlet area is 3000 cfm/500 fpm
equals six sq ft.
Next, calculate the duct diameter for the hood connection. Key to this is determining min­
imum duct transport velocity, V. Transport velocity depends on multiple factors and can
range from very low, below 2000 fpm for an HVAC clean air duct to above 10,000 fpm in
a high vacuum system. For combustible coal dust V equals 4000 to 5000 fpm. Using Q of 3000
cfm, you then calculate A = Q/V, or A = 3000/4500 = 0.67 sq ft, which lets you select the duct
diameter.
Engineering Toolbox has a wonderful duct design calculation tool at “Air Ducts - Major
Friction Head Loss, Online Calculator.” Enter 3000 cfm and iterate to an 11” diameter duct
with a velocity of 4546 fpm. There is also a metric tabulation in this calculator.
Note, in Figure 5, a bleed-in branch is used in combination with a butterfly damper to regu­
late indraft at the conveyor hood. The purpose of this design approach is to enable setting the
draft at the capture zone while adjusting air volume going to the main ducting using the
bleed-in branch. This ensures that duct velocity is kept in balance with the hood draw. Main­
tained air velocity is fundamental to preventing dust accumulation in the ductwork.

325
3.3 Provide plan and elevation views
Preparing layout drawings describing each hood location is the next step in the design process.
These will typically be preliminary sketches including plan views, elevation views and section
views when working with an existing process. Photos and video are a great help as are equip­
ment and facility drawings.
Field sketches showing hood and equipment locations in reference to column centers are very
important. A plan view field sketch of exhaust hoods for the screening plant from Figure 4 is
shown in Figure 6. Plan, elevation and section views were made for creating the engineering
drawings.

Figure 6. Field sketch of screen and conveyor hoods and duct.

The sketch includes hood and ducting diameters and a duct routing layout. Additional
details including elevation and section views as well as hood details are used to prepare
a SolidWorks 3D design model. Hood exhaust volumes are combined to determine duct and
branch sizing for use in Step Four.

3.4 Size ductwork diameter and lengths

Duct sizing is based on air volume required at the capture hoods and duct velocity necessary
to maintain particulate transport. Material selection to mitigate abrasion and corrosion elem­
ents is another typical concern.
The screening plant system shown above in Figure 6, illustrates design elements for a ducting
system. Hoods 4 and 5 were sized for a 10” diameter duct connection based on the design hood
volume of 2500 cfm. The target minimum duct velocity for this system is 4500 fpm. Using the
Air Duct tool yields an air volume of 2500 cfm at 4584 fpm. The combined volume at Hoods 4
and 5 is 5000 cfm. The rest of the system is developed in a similar manner.

3.5 Develop line sketches of duct routing

Detailed concept sketches of duct routing, hood locations and equipment general arrange­
ments are required so that system design calculations can be completed. Again, Figure 6 illus­
trates a system field sketch for design.

3.6 Calculate ductwork pressure losses

Ductwork static pressure is calculated by considering hood loss factors, duct friction loss and
equipment differential pressure requirements. Static pressure is used to size the exhaust fan
and determine motor horsepower.

326
 A spreadsheet is an alternative to the Air Duct tool for sizing duct diameters as well as for
calculating system static pressure. The spreadsheet in Figure 7 calculated a system volume of
17,600 cfm at 8.61 inches of static pressure loss. This information is then used for selection of
the dust collector and fan.

Figure 7. Duct Design Spreadsheet.

3.7 Select equipment

Selection of safe and effective dust control equipment is critical to successful system operation.
A properly designed system will deliver consistent air volume, high filtration efficiency and
peak operating performance. Maintenance costs and service requirements are also important
factors.
Cartridge dust collectors deliver ultra-high filtration efficiency and are often used for filter­
ing fine particulates such as silica dust in mining applications. The cartridge dust collector in
Figure 8 was installed at a brick grinding facility where dust from grinding, screening and con­
veying operations was captured and conveyed by ductwork. Cartridge filters have a smaller
footprint than baghouse units due to the large filtration area provided by the pleated media.

3.8 Determine equipment locations

Consider factors including equipment physical size, utility requirements, installation issues,
and discharge of collected dust. Can the equipment be located near process operations? Will it
need to be located outside of the immediate operating area or of the building? Regulatory and
safety issues for combustible dust control systems may be important factors. These may
include items such as explosion venting, backdraft venting controls, exhausting outside and
other factors.
The tall round tower in Figure 9 is a reverse air baghouse that was used for coal dust con­
trol at multiple silos being loaded by belt conveyor. Dust collectors for coal handling are

327
 Figure 9. Reverse air baghouse.
Figure 8. Cartridge dust collector.

subject to NFPA 652 guidelines since combustible dust is involved. Provisions for using dry
collectors may include explosion venting, fire suppression, backdraft dampers and other
safety protections.
The reverse air baghouse was eventually replaced by a wet dust extractor to mitigate fire
and explosion risk. The extractor, including integral fan, was installed where the original fan
had been located. Dust transported at 4500 fpm velocity over a 400 ft length enters the wet
zone of the extractor in less than six seconds. Wet dust extractors instantly mitigate the fire
and dust explosion potential in the dust collector and are recognized as BDT Best Determined
Technology for coal processing and handling operations.

3.9 Develop a detailed system design

Detailed system drawings need to be prepared showing final plan and elevation views of
hoods, ductwork and equipment. Care must be taken to include clearances for interference
obstacles such as piping, electrical conduits, equipment and structures. Design drawings are
initially submitted for approval. Final drawings for fabrication and installation are then com­
pleted. Figure 10 shows the 3D model for this system. Figure 11 is an example of design draw­
ings prepared for the sample room installation.

Figure 10. Sample room 3D model.

328
Figure 11. Sample room design drawings.

4 ENDING COMMENTS

Readers will have seen an emphasis in this paper on installations where wet dust extractors were
used instead of baghouse or cartridge dust collectors. This is not intended to say that dry dust
collectors should not be considered in a system design. My experience with dust control systems
began in the 1970’s, as a member of an engineering team that developed cartridge dust collectors
for downhole drill rigs used in strip mining. I have worked with hundreds of installations based
on dry collection. However, wet dust extraction system design has been the primary approach
used in my work over the past twenty years and is the main source of my examples.
Designing a safe and effective process ventilation dust control system is a process involving
many considerations. Having team members who understand this process is important to
achieving a successful installation.

REFERENCES

ACGIH 2010, Industrial Ventilation: A Manual of Recommended Practice, Cincinnati OH

CCPS 2017, Guidelines for Combustible Dust Hazard Analysis, John Wiley & Sons, Inc., Hoboken NJ
Engineering Toolbox “Air Ducts - Major Friction Head Loss, Online Calculator.” https://www.engineer
ingtoolbox.com/duct-friction-pressure-loss-d_444.html
NFPA 652 Standard on the Fundamentals of Combustible Dust 2016, Quincy MA
NIOSH Mining Program Information Circular 9532, 2021, Best Practices for Dust Control in Coal
Mining, Pittsburgh PA

329
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Development of VR-CFD-based training tool for dust control

in gateroad development

M. Qiao & T. Ren

School of Civil, Mining, Environmental and Architectural Engineering, Wollongong, New South Wales

J. Roberts
School of Mechanical, Material, Mechatronics and Biomedical Engineering, University of Wollongong,
Wollongong, New South Wales

J. Hines
School of Health and Society, University of Wollongong, Wollongong, New South Wales

C. Chow
School of Computing and Information Technology, University of Wollongong, Wollongong, New South
Wales

A. Clayton
Coal Services, Wollongong, New South Wales

ABSTRACT: This paper details the development of an immersive simulation tool that can be
used to provide enhanced 3D visualization of monitoring and computational modeling datasets
of dust and ventilation flow in an interactive, virtual environment, and clearly demonstrate the
exposure impact of certain working condition changes and dust mitigation practices in gateroad
development panels. To achieve this aim, onsite dust monitoring with AM520i and PDM3700
dust monitors and computational fluid dynamics (CFD) modeling were conducted to gain an
improved understanding of dust and airflow characteristics around the continuous miner (CM)
and evaluate different practices for dust controls. An algorithm was then developed to process
large amounts of datasets exported from CFD into a form that can be visualized in a virtual
reality (VR) environment using standalone VR headsets. A VR-CFD-based training tool in the
form of VR headsets has been developed and used as an educational platform to create an
effective communication of WH&S awareness of dust exposure in the workplace for CM oper­
ators and provide industry workforce (CM operators and development crews) training that
enhances knowledge learning and operational practices for dust exposure reduction.

1 INTRODUCTION

Long-term exposure to respirable coal dust and respirable silica dust in underground mines
continues to present an occupational health and safety hazard to mine workers in under­
ground coal mines (Wang et al., 2019, Ren et al., 2018, Zosky et al., 2016). Airborne dust
monitoring data collected by Coal Services (CS) show that Development/Continuous Miner
(CM) operators have the second highest workplace exposures next to longwall (LW) workers
when compared to the workplace exposure standards (WES) (Boyne, 2022). A similar trend
has also been reported in Queensland underground coal mines. The solution to effective miti­
gation of dust exposure depends not only on effective engineering controls, but also on on-going

DOI: 10.1201/9781003429241-35

330
innovative education and training programs for the employees on-the-job, to be able to identify
changing conditions, understand the potential impacts of operational changes, and know how
they should react to correct the situation in a timely manner.
Currently, the majority of mining machines for developing gateroad is 12CM30’s. Focusing
on the 12CM30 only will reduce the variability in the existing data while still maintaining
a large data set that can be interrogated. An initial review of the existing NSW data from
2018- 2021 shows that there are 910 (respirable coal dust) and 700 (respirable crystalline silica)
sample results taken from the breathing zone of bolting rig operators and miner drivers while
conducting operations on 12CM30 single pass miners (Coal Services, 2021). During 2021,
there were 2544 respirable crystalline silica (RCS) samples collected. Of these, 862 were taken
from the breathing zone of bolting rig operators and miner drivers while conducting oper­
ations on 12CM30 single-pass miners. 5% of the total samples exceeded the workplace expos­
ure standard (WES) of 0.05 mg/m3 and 39% of these were on the CM. Dust exposure within
the development panels where the CMs operate can be attributed to a number of different fac­
tors. As shown in Figure 1, the top three contributing factors to high dust exposures are high­
lighted in a red rectangle, including cutting stone, operator positioning and poor ventilation
standards, which will be used to develop the major dust scenarios of this study.

Figure 1. NSW Order 42 12CM30 Airborne dust exposure standard exceedance contributing factors
July 2018-June 2021 (Coal Services, 2021).

The evaluation of dust exposures within underground mines from historical data allows the def­
inition of three overarching scenarios in development panels with due consideration of the top
three contributing factors to be extensively studied in mine site dust monitoring, computational
fluid dynamics (CFD) modeling, and virtual reality (VR) development process. A VR-CFD-based
simulation tool is finally developed and inbuilt into a VR headset that has been taken to mine
sites to train users on the hazards of dust emissions and the best practices for dust control.

2 DUST MONITORING

Mine site dust monitoring was conducted in an underground coal mine situated near Sydney,
Australia, where the development panels are tunneled by a 12CM30-type CM, and cutting
coal is transported out of the heading face by a shuttle car (SC), with the layout of develop­
ment gateroad being shown in Figure 2. The gateroad is 5.2 m in width and 2.9 m in height.
The fresh air is provided from the travel road, with some streams of air flowing towards the
heading face of the travel road and some air passing through cut-through (CT) 15 and then
diverting into the belt road. The harmful dust produced in the process of coal cutting, trans­
porting and dumping process is sucked into the exhausting ventilation tube.

331
 Real-time monitors are specific types of direct-reading instruments that detect, react and
measure a specific contaminant. The measurements are usually data logged and are recorded in
short periods of time with the appropriate response time needs to be defined by the application
(Cauda 2021). The use of real-time monitors is becoming more common for specific monitoring
situations. Real-time monitoring is particularly useful in assessing more accurately where con­
trols are needed and the effectiveness of implemented controls. TSI AM520i SidePak™ Personal
Aerosol Monitor (AM520i) and the PDM3700 were determined in this study, both of which are
characterized by battery-operated portable monitoring devices designed for personal monitoring
(to be worn on a worker). However, these two main instruments can also be used for area moni­
toring (placed in a location) to gain a measurement of dust concentration.

Figure 2. Layout of development gateroad.

Dust monitoring instruments were placed in the vicinity of the CM, with the exact locations
shown in Figure 3. Location 1 and 4 were in front of the miner left and right respectively, 2
and 5, mid miner left and right respectively, and 3 and 6 rear of the miner left and right
respectively. The alignment of each sample head was located to be in the same direction and
at the same height.

Figure 3. Dust monitoring locations in the development gateroad.

The continuous data collected from the AM520i was used directly in the CFD validation as
the data was in a format that was easier to manipulate, and more data was collected from this
instrument with seventeen (17) sets of sampling results compared to 7 for the PDM3700,
which allowed a more robust interpretation and modeling of the data. The PDM 3700 results
were limited in number and variation in location.

332
3 CFD MODELING

Based on the mine site condition, a three-dimensional (3D) computational model consisting of
CM, SC and ventilation tube was built, as shown in Figure 4. The fresh air intake marked
with a blue arrow is provided via a cut-through connecting with the heading of interest, while
the dusty air is marked with a red arrow on the return side (away from the CM and SC).
In order to reduce total computational elements and preserve high accuracy, the model was
then meshed using the polyhedral meshing method, and the mesh element size was progressively
reduced closer to the heading face such that the required CFD modeling could be completed with
sufficient accuracy and detail while maintaining computationally appropriate simulation times.
The simulations were optimized to be completed in less than 48 hrs so that future work can allow
further scenarios to be generated quickly with mine-specific setups or scenarios. A mesh independ­
ence study was conducted to eliminate the influence of mesh size on simulation results, and the
total elements of the selected meshed model were approximately 2.4 million, as shown in Figure 5.
The minimum orthogonal quality of the meshed model was 0.20, which was of good quality
according to the standard released by the ANSYS Meshing User’s Guide (ANSYS, 2022).

Figure 4. CFD model of development gateroad.

Figure 5. Meshed model.

To be able to model any arbitrary scenario for visualization in the VR system, it was neces­
sary to verify the CFD model against the real-time dust monitoring data collected from the
mine site. The simulation results were validated against the site data based on a 1-min average
dust concentration. As illustrated in Figure 6, the CFD model correlated well with onsite dust
monitoring data, meaning that the CFD would be sufficient to simulate other scenarios with
confidence in its accuracy. A key aspect of the good correlation between site data and simula­
tion data was the inclusion of multiple dust sources to match observations of dust generation

333
during face cutting and loading of the SC; this included from the face, the throat of the miner,
the conveyor, and at the loading point in the SC. This was a significant improvement on previ­
ous simulations that have been conducted in the past for this purpose.

Figure 6. Model validation results.

With the validated CFD model., twelve scenarios were simulated, including coal cutting at
different sections of the development gateroad (roof, middle and floor) under two different
ventilation conditions where the exhausting ventilation duct was advanced to a standard pos­
ition or in a poor position, stone dust cutting at the roof section under standard or poor venti­
lation condition, sudden ventilation fan tripping-off, more dust from outbye of the SC,
breaking-away the cut-through as well as cut-through holing into another development gate­
road. Only two scenarios are introduced in this paper where the full study can be found in the
final ACARP report for project C33030 (Ren Ting et al., 2022).

Figure 7. Dust dispersion characteristics with coal or stone cutting at the roof section under standard
ventilation.

334
 The dust dispersion characteristics with coal or stone cutting at the roof position under
standard ventilation conditions are shown in Figure 7. At t=5 s, it is noted that dust produced
by cutting becomes caught in one of the two recirculating zones around or below the cutting
drum and a portion of high-concentration dust at the left-hand-side (LHS) of the CM is
sucked into the ventilation duct. In addition, in the middle of the cutting drum with stone cut­
ting, some dust falls onto the floor due to the entrainment effect of air coming off the drum
and the downward gravity force, which is different from the coal-cutting counterpart. At
t=10 s, dust from the shovel and conveyor belt passes through the gap between the top cover
and beam of the CM and is sucked into the duct with a mass concentration of approximately
35 mg/m3. In comparison to the scenario of coal cutting at the roof section under standard
ventilation, more stone dust at the right-hand side (RHS) of the CM drops underneath the
cutting drum at t=10 s, which is attributed to a larger density of stone dust than coal dust. At
t=15 s, dust from the SC mixes with dust from the shovel and conveyor belt at the top cover
of the CM and then is drawn into the ventilation duct. After 5 s without dust releasing
(t=25 s), it is noted that dust concentration at the LHS of the CM is still higher than 100 mg/m3
for both scenarios, which means more time is required to dilute dust concentration below the
statutory limit.

4 VR-CFD SYSTEM

In order for further visualization in the VR system, simulation data was required to be
exported from the CFD software and reprocessed into a form suitable for VR. To create
awareness of dust and ventilation flow from the viewpoint of mineworkers and to allow mine­
workers to better understand their real surroundings and make critical decisions during
mining operations, the real-time visualization of actual dust particles and the ability to inter­
act with the data and the virtual environment are required. However, the real-time visualiza­
tion of dust particles from the CFD datasets generated for this project is a challenging
problem. This is due to the fact that the data produced by the CFD models for each respective
scenario is enormous. This makes it impossible to render the raw data in a VR environment in
real time. To overcome this problem, extensive research was required to examine context-
preserving sampling techniques to convert and filter CFD data into a form suitable to be
imported into a VR platform, along with intuitive visual representations of the data in an
interactive immersive VR system, with overall workflow shown in Figure 8.
An algorithm was therefore developed for processing large volumes of CFD datasets. The data
processing method included four steps, namely filtering raw data, applying clustering algorithms,
random sampling, and finally visualizing the data in the VR system. Firstly, to filter the raw data,
a dust density threshold was set to remove unneeded data; Secondly, the K-means method was
used to group the remaining data into different clusters by choosing a centroid that minimized the
within-cluster sum-of-squares. Thirdly, a pre-defined amount of data in each cluster was randomly
sampled, and data that was selected in the last frame is given a high probability of reselection to
help stabilize the visualization. Finally, the pre-processed data was installed into a headset and the
particle system of Unity was applied for visualization where each measurement in the data set
corresponds to a particle.

Figure 8. Overall workflow from CFD modeling to visualization in VR.

335
 Outcomes were significant after data pre-processing, as illustrated in Figure 9. It was evi­
dent that although the density of the dataset was reduced with the following processing, the
results still preserved the required fidelity for adequate visualization. Processing of the data
required approximately 10 hours to complete on an i9-10940X CPU where the number of
data points was reduced from the initial 3 million down to 5000 per frame which results in
a reduction in total file size from approximately 230GB to around 0.18GB, dropping by
approximately 99.92%. This process was required for all scenarios that were to be visualized
in the VR environment.

Figure 9. Examples comparing the filtered data with the original raw data.

The final purpose of the proposed VR tool was a self-guided learning experience where the
user could fit the headset and be taken through a predefined learning experience with moder­
ate user interaction. The VR tools guided users through a number of scenarios that demon­
strate safe and unsafe situations. The dust simulations clearly showed where the dust was
going under different circumstances. Voice-over was used to educate users about what they
were seeing and encouraged them to engage with the content. Users were able to teleport
around the room and view the miniature development panel from many different angles. The
user could also enter the mine at a 1:1 scale to see the simulation from the point of view of an
operator. Finally, the menu showed a quiz button that tested users on what they had learned.
The VR system developed is shown in Figure 10.

Figure 10. Visualization of the VR system.

5 INDUSTRY SHOWCASES

As a significant task of this ACARP project, several industry showcases and technology trans­
fers have been made to project meetings, industry workshops and conferences by using stan­
dalone Meta Quest 2 headsets. This enables stakeholders or interested parties to gain a better
understanding of dust and ventilation behavior in various development operations and cutting
scenarios and provide feedback on the developed intuitive VR training tool. Significant
improvements have been made according to feedback and an improved VR training tool has
been developed for stakeholders to start to use in their operations.

336
Figure 11. Industry showcases made to key industry stakeholders and a wider community in the industry.

6 CONCLUSIONS

The aim of this study was to develop an immersive simulation tool for the 3D of dust and
ventilation flow in an interactive virtual environment to demonstrate the exposure impact of
certain working conditions. The advantage of the interactive VR simulation tool is that it can
be configured and customized to various site conditions and equipment, and provide the
industry workforce with training that enhances knowledge and learning of operational prac­
tices for dust exposure reduction. To achieve this aim there were various key stages, enabling
technologies and research innovations that were required to produce the end product of the
VR tool; in addition to the VR tool there were various other important and interesting learn­
ing outcomes that came from the project which may be useful to the industry.
Overall, the study has resulted in the development of a VR visualization and educational tool
that can more intuitively and comprehensively communicate the results of CFD simulation data
to mine workers. In doing so, a number of key technological achievements were made that can be
used in the future to more quickly translate real-world mine site particulate readings into a VR
visualizer. Along the way, we were able to ensure that we aligned the VR solution to known best
practices in the field, as well as also identify areas of future research where there is more to learn.

ACKNOWLEDGEMENT

This study was supported by ACARP project C33030. The authors acknowledge the support
from various mine sites in NSW.

REFERENCES

Ansys, I. (2022) Ansys Meshing User’s Guide-2022r1, Canonsburg, Pa 15317.

Boyne, L. (2022) 2021 Order 42 Airborne Dust Results Review. Coal Services, Coal Services.
Coal Services (2021) Order 42 Continuous Miner Test Results Summary June 2018 – June 2021.
Ren, T., Wang, Z. & Zhang, J. (2018) Improved Dust Management At A Longwall Top Coal Caving
(Ltcc) Face – A Cfd Modelling Approach. Advanced Powder Technology, 29, 2368–2379.
Ren Ting, Roberts Jon, Hines Jennifer, Chow Casey, Qiao Ming, Zong Wei, Su Guoxin, Hai Faisal,
Susilo Willy, Sugden Adrian, Clayton Andrew, Shepherd Mark, Farrelly Matthew & Gareth, K. (2022) Vr
Cfd Based Simulation And Training Tool For Dust Control In Gateroad Development Panels. Acarp.
Wang, Z., Li, S., Ren, T., Wu, J., Lin, H. & Shuang, H. (2019) Respirable Dust Pollution Characteristics
Within An Underground Heading Face Driven With Continuous Miner – A Cfd Modelling
Approach. Journal Of Cleaner Production, 217, 267–283.
Zosky, G. R., Hoy, R. F., Silverstone, E. J., Brims, F. J., Miles, S., Johnson, A. R., Gibson, P. G. &
Yates, D. H. (2016) Coal Workers’ Pneumoconiosis: An Australian Perspective. The Medical Journal
Of Australia, 204, 414–418.

337
Mine fans
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Stall impact on axial fans and testing of anti-stall rings

J. Fernandez
Deputy Chief Executive Officer, Zitron S.A

ABSTRACT: A fan’s operational reliability is a critical factor in ensuring that an under­

ground mine’s primary ventilation system enables and sustains the mine’s continuous
operation.
At a time when underground mining operations are required to access ore bodies located at
ever greater depths, primary ventilation systems are having to deliver volumes of fresh air
over longer distances, thereby increasing the operating pressures applied to the ventilation
fans. The search is therefore on for ventilation systems with a higher degree of operational
flexibility and reliability to cope with constantly changing mine conditions.
In this context, the fan stall phenomenon is the subject of considerable discussion in rail
tunnel ventilation systems, where fans are often subject to varying conditions. A typical metro
tunnel fan is designed with an anti-stall feature to protect the fan against train induced pres­
sures (i.e., the fan pressure rise increasing continuously from the full flow/zero static pressure
point to the maximum pressure rise of the fan).
However, the different ventilation circuit scenarios in a rail tunnel are relatively easier to
predict than the changing conditions in underground mines. Increased production levels
require larger volumes of air and increasing the ventilation system resistance or, alternatively,
adding more primary ventilation fans to the ventilation network, which may result in the
fan(s) entering the unstable part of their performance curve (i.e., the fan stall region).
It therefore follows that the use of fans with an anti-stall feature may be more justified in
the underground mining sector than in the infrastructure tunnel sector.
This paper describes the phenomenon of fan stall and provides insights into some of the
anti-stall ring (ASR) design parameters, discussing the mechanical and aerodynamic consider­
ations that arise when fitting an axial fan with an anti-stall ring. It also compares the fan per­
formance of an axial fan with and without the stabilisation ring fitted, highlighting the
importance of the fan manufacturer’s system design and testing capabilities, and analyses the
advantages and disadvantages of axial fans with fitted with anti-stall devices.

1 INTRODUCTION

Energy consumption by mine ventilation systems is higher than ever due to mining operations
being conducted at ever deeper levels. Whilst the increased use of Ventilation On Demand sys­
tems (VODs) will in many cases produce significant savings in energy consumption as
a consequence of an enhanced understanding of the ventilation equipment’s performance
under real-time underground conditions, the primary ventilation systems will be required to
deliver higher volumes of fresh air over longer distances, which will result in generating higher
pressures and more powerful fans being required.
To address these challenges, it is vital that mine ventilation practitioners carry out accurate
ventilation calculations, clearly defining all possible system resistances that the mine’s fans
will have to overcome during their design life. Providing an adequate pressure margin to the
fan stall point is a critical consideration during the design process. However, consideration

DOI: 10.1201/9781003429241-36

341
should be given to the fact that excessive pressure margins will impact fan efficiency and there­
fore fan energy consumption.
Anti-stall rings (ASR) can severely impact fan efficiency and therefore the use of ASR has
not always been regarded as the preferred solution. However, if the ASR system is designed
correctly, it can enable the safe operation of a fan under severe pressure fluctuations, without
a significant impact on fan efficiency, thereby making the use of an ASR a good choice when
setting a mine’s fan selection strategy.
This paper provides a brief description of the axial fan stall phenomenon, commenting on
the alternatives available to control and manage the fan stall, and reviewing the data from
two axial fan performance tests:
1. Test #1: An aerodynamic fan test, where different blade pitch angles are tested with and
without an anti-stall ring fitted to the fan.
2. Test #2: A strain gauge test, measuring the mechanical and electrical impact of pressure
fluctuations on axial fans with and without an anti-stall ring fitted.

2 DESCRIPTION OF THE AXIAL FAN STALL PHENOMENON

When a fan is operating in a stalled condition, severe separation of the flow lines along the
profile of the blade occurs, and this induces a strong decrease in the increasing pressure pro­
duced by the fan, negatively impacting fan efficiency, increasing radiated noise and the mech­
anical stress on the blades, which may lead to catastrophic fan failure. Some of the main
problems arising from the fan stall condition are described below:
• The fan cannot deliver design-duty performance and generates more noise.
• Fan efficiency is significantly impacted, which in turn increases the fan’s energy
consumption.
• Increased vibration levels could lead to a fan/motor premature failure.
• Stall operation may result in fan blade rupture due to metal fatigue.
The effects of fan stall are not the same for all fans. Stall severity depends on two main
factors: (i) pressure developing capacity of the fan; and (ii) impeller pitch angle.
Impellers with a higher hub-to-tip ratio and a larger number of blades can develop higher
pressure. However, they are subjected to a more severe stall operation, with a significant dip
in the stall region (Figure 1).

Figure 1. Typical unstable region high-pressure fans.

Low hub-to-tip diameter ratio fans with low pitch angle settings are the preferred option
for minimizing the impact stall in axial fans (Figure 2). However, it must be noted that select­
ing fans with a low pitch angle will impact fan efficiency, which in turn increases energy con­
sumption. Selecting these fans as a strategy to minimize the risks associated with fan stall may

342
result in oversized fan diameters and higher motor speeds to compensate for the lower pres­
sure developing the capacity of the fan. Caution must be taken with these fan selections, as
they have higher centrifugal forces and tip speeds which may impact the life of the fan.

Figure 2. Typical unstable region low-pressure fans.

The impeller pitch angle is also an important contributing factor to stall severity. Figure 3
shows the aerodynamic tested performance of a medium-pressure axial fan. The tested model
has a 2.5-m impeller diameter, 1,250-mm-diameter hub, and 450 kW motor-rated power. As
reflected in Figure 3, the 2.5-m-diameter tested fan will show some kind of ‘non-stall’ charac­
teristic with a -10° pitch setting.

Figure 3. 2.5-m-diameter fan without ASR.

3 OPTIONS TO MANAGE AND CONTROL THE AXIAL FAN STALL

A commonly used strategy to avoid fan stall is to select a fan with high-pressure developing
capacity, allowing for a significant pressure margin with the design duty point(s). Typically,
mine primary fans are medium to high-pressure types, which is relevant because higher-
pressure fans normally have their peak efficiency in areas closer to their stall limit. Therefore,
selecting a fan with a large pressure margin to the stall point may be less efficient at the design
duty point(s). Three (3) options available to avoid a fan’s operation in stall are described in
this paper.

343
3.1 Option 1: Using stall detectors
A simple and cost-effective system for fan stall detection is the Petermann probe system
described in Figure 4. The pressure variations in the stall zone of an axial flow fan can cause the
airflow to be blocked or flow in the opposite direction. The Petermann probe detects the stall
condition of the fan by measuring the pressure difference between the total air pressure in the
opposite direction of rotation of the fan impeller and a reference pressure (static pressure at the
wall of the air duct upstream of the fan blades). This pressure difference is negligible in the
stable range but increases significantly when the fan enters the stall condition. The graph in
Figure 4 shows the pressure difference between the probes at different air flow rates, indicating
a characteristic change at the junction between the stable and unstable ranges of the fan.

Figure 4. Petermann probe, axial fan stall detection system.

3.2 Option 2: Bringing the fan out of the stall while the fan is in operation
The mine ventilation system resistance may deviate from the expected due to miscalculations
or unforeseen mine developments. In Figure 5, a primary axial fan was designed for a duty
point of 300 m3/s and 4.2 kPa, corresponding to a system resistance of 0.046 Ns2/m8. If the
actual resistance is over twice as high, the fan would stall. By reducing the pitch angle, such as
to +2° or -4°, fan stall operation can be avoided.
It is important to note that it is not possible to take a fan out of its stalled operation just by
reducing the motor speed (Figure 5). Sample fan performance curves at various speeds for the
same blade angle are shown in Figure 6.
Reducing the system resistance or reducing the impeller pitch angle is the only solution. The use
of a stall detector system in combination with a variable pitch in impeller motion and a variable
speed drive to adjust the motor speed as required, provide a significant degree of flexibility to the
ventilation system and is a good solution to protect the fan against fan-stalled operation.

3.3 Option 3: Using anti-stall rings (ASR)

Most modern primary fan installations are variable speed drive (VSD) operated, and PLC
managed, which allows for continuous monitoring of the ventilation system’s performance. In
some cases, ventilation practitioners rely on the fan performance (i.e, vibration monitoring
system) to detect when the fan is operating in a stall. However, there are cases when the fan
operates in a stall and vibration levels are not significantly high.
To mitigate the fan stall effects an anti-stall ring device may be the best option. The anti-
stall ring consists of a peripherical duct-shaped ring integrated with the fan impeller casing. In
the mining unidirectional fans, the ASR is located just upstream of the impeller blades
(Figure 7).

344
Figure 5. Excessive system resistance, fan stalled operation.

Figure 6. Fan performance curve at various motor speeds.

The effect of the ASR system on the fan performance curve is to eliminate the severe dip in
the stall region. Once the severe separation of the flow lines along the blade profile of the
blade occurs, the ASR guide vanes will re-inject the flow in the axial direction. The images in
Figure 8 show the flow lines at the blade tip for the 2.5-m-diameter impeller fan described in
this paper. Three different fan duty point scenarios have been studied, each with and without
the ASR fitted to the fan:
• Scenario 1: fan operating duty point within the normal fan operating range.
• Scenario 2: fan operating in an early stage of a stall.
• Scenario 3: fan operating under a very developed stalled condition.

345
Figure 7. Axial fan impeller with ASR.

From the figures below, it is possible to observe the area impacted by the stall phenomenon
(flow in the reverse direction) and how part of this flow is injected into the ASR chamber and
re-injected in the correct direction (left to right in the figures below).
In the case of a very developed stall (scenario 3), it is possible to observe that a big area is
impacted by the stall, while with the ASR fitted the stall-affected area is significantly
smaller.

Figure 8. CFD simulation axial fan in operation with and without ASR.

346
4 TESTING THE STALL IMPACT ON AXIAL FANS

In order to analyze the impact of the stall effect on axial fans, two (2) tests have been carried
out. The first test is aimed at understanding how the ASR impacts fan aerodynamic perform­
ance and is carried out in the Zitrón test tunnel at its manufacturing operations in Gijon,
Spain.
Measuring 100 meters in length and 52 square meters in cross-section, this is the largest cer­
tified test tunnel in the world, allowing axial fans up to 5.8 meters in diameter to be tested at
full power, including up to 2,000 kW, whilst reproducing any real ventilation circuit
resistance.

4.1 Test #1: Measuring the aerodynamic benefits of anti-stall rings (ASR)
Once the aerodynamic performance test of the 2.5-m impeller diameter fan described in this
paper is undertaken with the ASR casing fitted, a steel cover plate is used (Figure 9) to test the
performance of the axial fan without ASR.

Figure 9. Axial fan impeller with and without ASR.

The graph in Figure 10 below shows the fan aerodynamic test results when ASR is fitted.
Test results conducted without ASR (Figure 3) are now compared with the test results in
Figure 10. For the particular fan tested, the ASR system provides significant benefits when
the impeller is set to +10° pitch. However, the benefits are not so significant when the impeller
pitch is set to -10° (Figure 11).
It is also important to highlight that the impact on peak efficiency is higher at -10° pitch. It
must be noted that the ASR system is designed to operate at its maximum efficiency on
a specific blade pitch angle, therefore fan pitch adjustments on site (i.e., to accommodate for
future mine resistances) may impact the ASR effectiveness. The ASR efficiency impact in low-
pitch angles will be different from other ASR designs.
Other important considerations when designing ASR systems for mining primary fans are
the air speed through the ASR and its size so that potential blockages due to dust, and other
typical mine pollutants are avoided.

4.2 Test #2: Measuring the mechanical and electrical impact of pressure fluctuations on axial fans
This second set of tests is aimed to analyse the impact of pressure fluctuations on axial fans
with and without a fitted ASR. In a primary mine fan, there are several possible scenarios that
could impact the axial fan performance and its mechanical strength. Important natural

347
Figure 10. A 2.5-m-diameter fan with ASR.

Figure 11. A 2.5-m-diameter fan, +10 & -10 pitch with and without ASR.

348
ventilation effects, piston effects generated by the circulation of large vehicles through
declines, or the influence of other ventilation fans in the system, may induce dangerous pres­
sure fluctuations for mine primary axial fans.
Fitting an anti-stall ring will minimize the impeller fatigue stress suffered by the impeller
blades during the fan stalled operation. However, it must be noted that although the ASR will
significantly reduce the fatigue stress on the blades, it does not eliminate it completely.
The fan pressure fluctuation test setup is described in Figure 12. It is composed of
a ventilation circuit including flow straighteners, a closing valve (guillotine), and fan perform­
ance monitoring instrumentation (flow, pressure strain gauges on blades, etc.). The change in
the system resistance is achieved by the operation of an auxiliary fan and a control damper.

Figure 12. Test bench arrangement for pressure fluctuations on axial fans.

The main objective of the test is to measure the impact of pressure fluctuations on axial
fans. The axial fan chosen for the strain gauge test has a 2.2m impeller diameter and the test­
ing facility is operated to analyze pressure fluctuations of +35%/- 70% of the fan design duty
point (Figure 15).

Figure 13. Tested pressure fluctuations on the 2.2-m impeller diameter axial fan.

The strain-stress is measured (Figure 13) by installing two (2) strain gauges on the blade
pressure side and another two on its opposite side (pull/push forces). The alternating stress
produced by the variations in pressure is measured (Figure 14) under four (4) possible fan
operation scenarios: fan running at full and half nominal speeds, with and without ASR
fitted.

349
Figure 14. Blade fatigue stress measurement.

The fan tested was originally designed to operate with a duty point of 100 m3/s and 1.5 kPa
total pressure. Several fan operating scenarios are tested, with system resistances of +500 Pa
and – 675 Pa over the system design resistance (Figure 15).

Figure 15. Fan performance curves at different system resistances.

The fan performance and mechanical stress of the fan when running at full and half nom­
inal speed and under higher and lower system resistances were analyzed. Below is a summary
of the analysis of the mechanical stress test results:
• The average stress on the blades is proportional to the pressure variation on the fan.
• The amplitude and fluctuations of stress are dependent on the fan operating duty point.
• The stress fluctuations frequency has two components (Figure 16): the fan rotation speed
(16,6 Hz – 993 rpm) and the blade natural frequency (112 Hz).
The speed change cycles impact blade fatigue life, therefore reducing the amplitude of the
stress fluctuations is very important to assure a long design life of the equipment. The test
results show that the anti-stall ring system reduces the amplitude of the stress fluctuations,

350
Figure 16. Frequency of stress fluctuations.

particularly when the fan is running in the unstable region of the fan performance curve.
Table 1 below shows the summary of the peak alternating stress measured during the tests.

Table 1. Peak alternating stress at various fan speeds.

Test number Axial Fan speed Fan type Peak alternating stress

#1 993 rpm Without ASR 8.1 MPa

#2 993 rpm With ASR 6.1 MPa
#3 500 rpm Without ASR 10.5 MPa
#4 500 rpm With ASR 6.2 MPa

When the fan is running at full speed (993 rpm) and with the ASR fitted, the peak alternat­
ing stress is 6.1 MPa (Figure 17). In this test, the blade’s peak alternating stress increased to
8.1 MPa (33% increase) when the fan was tested without a fitted ASR.

Figure 17. Alternating stress - fan running at full speed 993 rpm with ASR.

According to fan laws, the pressure developed by the fan is proportional to the square of its
running speed. It is therefore evident that a +500 Pa pressure increase will have a higher
impact on the blades alternating stress when it is operated at lower speeds.

351
 When the fan was operated at half of its nominal speed (500 rpm) and without a fitted ASR,
the increase in alternating stress ranged from 8.1 MPa (993 rpm) to 10.5 MPa (500 rpm). This
is a 30% increase in alternating stress produced by the system resistance increase.
When the fan with a fitted ASR was operated at 500 rpm, and under the +500 Pa increased
system resistance, the peak alternating stress was 6.2 MPa, which represents a 40% reduction
of maximum stress when compared to the test without ASR (10.5 MPa).
For this paper, only the alternating stress on the fan is analyzed. However, from the vari­
able speed drive electrical readings, we can observe that significant negative pressure fluctu­
ations, particularly when the fan is running at lower speeds, could make the fan operate in
a negative pressure part of its performance curve. This effect could damage the VSD (DC
link) and the use of regenerative VSDs may be needed.

5 CONCLUSION

The impact of a stall is not the same on all axial fans. Fan aerodynamic design is very relevant
for its behavior under stalled conditions. However, its impeller setting is also important. While
the higher-pressure fans, with higher hub-to-tip ratios and shorter blades, are typically sub­
jected to more severe effects on a stall, any axial fan configured with a low blade pitch angle
will have a much better response to the stall than a fan with a higher blade pitch angle setting.
It must be noted that the results presented in this paper are only applicable to the specific
fans tested. However, they may be used to establish best practices when selecting and operat­
ing axial fans.
There are several alternatives to avoid fan stalled operation. Selecting fans with a higher-
pressure developing capacity and therefore bigger pressure margins to the stall area, may be an
option. However, it must be noted that this selection strategy impacts the fan operating efficiency
and therefore the power consumed by the ventilation system. Sensors based on the Petermann
probe are very effective to detect the fan stalled operation, shutting down the fan as required.
However, for those cases when the fan cannot be stopped during its stalled operation or for
scenarios when the axial fans are subjected to severe pressure fluctuations, the anti-stall ring
may be a good solution.
From an aerodynamic point of view, ASR systems will allow the continuous rise of the fan
characteristic back toward zero flow. However, there are cases where the impeller design angle
is so low, that the aerodynamic benefits of the ASR are negligible. From a mechanical point
of view, the ASR systems do not eliminate the impact on the impeller blades. However, they
reduce it. ASR systems may be particularly important for those cases where the fans are sub­
ject to high-pressure fluctuations while running at reduced speeds.
Fan factory testing, in combination with the manufacturer’s system calculations and fluid
dynamic simulations, results in a very accurate prediction of any possible fan operating condi­
tion. Consideration must be given to the requirement for full fan testing, including the fan
unstable area of operation at the factory. Fan tests at various speeds and pitch blade angles in
a controlled environment, provide very useful information, allowing ventilation system practi­
tioners to make any necessary improvement to the system on-site as required.

ACKNOWLEDGEMENTS

The author would like to acknowledge Roberto Arias, Susana Rodriguez, Jorge G. Norniella,
and Adrian Salvado, for carrying out the fan aerodynamic testing and strain gauge testing, as
well as finite element analysis that has allowed the author to prepare this paper.

REFERENCES

Eurovent 1/11 – 2007. Fans and System Stall: Problems and Solution.
Piston effect testing results 2021, Zitron S.A.
Factory acceptance test Model ZVN 1-25-450/6 (manufacturing order 23444), Zitron S.A.

352
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Practical values for the evaluation of fan system efficiencies

John Bowling
SRK Consulting, USA

Garrett Schult & J. Van Diest

Nevada Gold Mines, USA

ABSTRACT: Fans purchased for mines are typically specified for the projected maximum
operating point that they are expected to perform at or close to it. It is common to use an
overall efficiency for the fan installation to predict power consumption. While the fan manu­
facturer recommends a fan based upon an efficient operating range, there are, however, sev­
eral factors that can affect the fan system efficiency such as poor installation practices,
operating the fan outside the predicted operating range, and poor maintenance. In this study,
a set of data collected from main and booster fan installations is used to provide additional
information with the intent of improving the understanding of how to determine the range of
fan system efficiencies. The study discusses how data were measured and sorts the data so that
one may compare the differences between fan installations at coal and metal operations.

Keywords: mine ventilation, planning, fans, efficiency, coal, metal/nonmetal, fan installa­
tions, main fans, booster fans

1 INTRODUCTION

Primary fan installations are often among the most energy consuming fixed equipment at an
underground mine. Power costs are largely influenced by fan efficiency. That is why it is
important for planning purposes to be able to correctly estimate power consumption for fans.
In most cases, a fan is sized for the maximum operating duty over its life span, which is in
most cases based on life-of-mine (LOM) planning and matched to a fan that operates in an
efficient range at its LOM or peak duty. This operating point is generally developed by
a ventilation engineer using mine projections to predict the required airflow and the total pres­
sure difference required to overcome the losses occurring in the airways and at the intake and
discharge. This pressure loss often doesn’t include any losses associated with components of
the fan installation itself such as ducting, dampers, screens, silencers, etc. Losses due to the
components are in fact part of the “fan system efficiency” (ASHRAE, 2020). Fan system effi­
ciency is typically assumed to be 65% to 75% for planning purposes, or efficiency values from
manufacturers’ fan performance curves are used without considering the additional losses
associated with the fan installation. Using measured data from 84 unique fan installations,
this study examines fan system efficiency calculated from real-world examples.

2 BACKGROUND INFORMATION

2.1 Fan power calculations

There are two parts in calculating the system efficiency of a fan installation. First, the air power
of the system must be determined. Air power is calculated from Equation 1 (McPherson, 2009).

DOI: 10.1201/9781003429241-37

353
This is done by measuring the airflow quantity through the fan and the fan total pressure.
The second part is determining the actual absorbed power of the fan. This is accomplished by
measuring the voltage, current, and, if only apparent current and voltage are measured, the
system power factor. Using these values, the operating power of the fan can be calculated using
Equation 2 (Feick, 2005). System efficiency is calculated by dividing the air/output power by
motor absorbed power, as in Equation 3.

Where:
PO = output power or air power (kW)
Q = quantity of airflow (m3/s)
PFT = fan total pressure (kPa)

Where:
PM = motor absorbed power (kW)
Vavg = average phase voltage (V)
Aavg = average phase current (A)
PF = power factor (unitless)

Where:
η = efficiency (unitless)
PO = output power or air power (kW)
PM = motor absorbed power (kW)
This measurement is system efficiency, which for the purpose of this analysis includes all
the components of the fan installation losses, not just the fan’s aerodynamic efficiency (which
is reported on the manufacturer’s performance curve). The components that make up the fan
system losses are illustrated in Figure 1. Component losses included in fan system efficiency
can be described as follows (ASHRAE, 2020):
• Inlet Losses: Shock losses due to duct inlet and other duct components upstream of the
impeller, e.g., silencers, dampers, changes in duct orientation, size, or shape.
• Motor Loss: Electrical inefficiency of the motor operating at the given speed and torque/
load.
• Drive Loss: Mechanical losses in the drive train, i.e., slippage and bearing friction.
• Mechanical/Aerodynamic Loss: Loss of useful fluid energy due to aerodynamic factors of
the fan impeller and guide vanes, including drag and vortex induction.
• Discharge Losses: Shock losses due to duct outlet and other duct components downstream
of the impeller, e.g., silencers, dampers, changes in duct orientation, size, or shape.
A detailed treatment of calculation or estimation of these losses, e.g., computation of shock
losses, explanation of electrical motor and drive efficiencies, can be found in many compre­
hensive reference texts[1,3]. Estimation of the contributions of each of these loss types was not
considered in the scope of this study. Note: for any given flow rate, pressure loss and power
loss are used interchangeably as “loss”, as the two variables are linked by Equation 1.
Fan measurement ports are often unavailable or nonideally located for measuring the oper­
ating pressure of an installed mine fan. Therefore, the actual fan total pressure in Equation 1
is often higher than indicated by typical in-situ measurement, but for the purpose of this ana­
lysis, any measurable fan system pressure is treated as the “fan total pressure” for the purpose
of computing output/air power. For underground installations, applied pressure from the fan

354
Figure 1. Components of the fan system losses.

bulkhead is often used. In-situ measurements typically do not allow for discrete quantification
of the internal losses of the fan and other system components (e.g., silencers, diffusers). The
motor efficiency is also incorporated into the power reading which results in additional losses
being unaccounted for. The combined system efficiency, if plotted on a generalized fan per­
formance curve, might look like that shown in in Figure 2.

Figure 2. Idealized axial fan performance curve (total pressure and efficiency vs. airflow).

2.2 Field measurement data

The fan measurements for this study were taken during ventilation surveys conducted by
Mine Ventilation Services, Inc. and SRK Consulting (U.S.), Inc. engineers from 2003 through
2020. Several of these fans have been measured during multiple consecutive surveys but only the
most recent readings were used to create a dataset of 84 unique fans. These fan measurements

355
were collected from a wide range of underground mine sites, with measurements representing
a variety of fan sizes and duties installed both on surface and underground. Mine sites repre­
sented include coal, metal, and nonmetal mines primarily in North America with some in South
America and Europe. The range of operating points of these fans are shown in Figure 3.

Figure 3. Measured operating points of selection of fans reviewed in this study.

For underground fans installed in bulkheads, typically the static pressure was measured
with a digital manometer by placing a tube through the bulkhead. This measures pressure
applied to the ventilation network and therefore the fan system applied pressure.
For fans installed on surface in ductwork, a digital manometer connected to a pitot tube
was used to measure pressure through ports in the fan housing. For an exhausting fan, the
total pressure was measured at (as near as practically measured) the inlet side of an exhausting
fan was measured with a digital manometer. For practical reasons, this measurement was trea­
ted as the fan total pressure. A similar method was used to measure forcing fans, though the
measurement ports were located at the fan outlet. These measurements are not perfect read­
ings but provide enough resolution to provide an estimation of the fan operating conditions.
Power measurements for volts and amps were typically collected from the fan motor PLC
display at the time of fan measurement or from site electricians during the site visit. The
power factor was often estimated for the entire mine site – not necessarily the actual power
factor of the fan or simply unavailable. For other fans, the PLC displays power factor and
only active amps and volts are displayed. For fans driven by variable frequency drives
(VFDs), drive input power is measured and VFD inefficiency is accounted for in the power
consumption. For fixed-speed fans (those not using a VFD), the total absorbed power may
include the reactive power if the fan current and voltage data recorded were corrected to
active current and voltage.
United States coal mines must locate their primary fans on surface, a requirement which
tends to concentrate the total fan power into fewer larger fans compared with metal/non-
metal mines.

356
3 RESULTS AND ANALYSIS

3.1 Overall
The overall results combine fans of all motor sizes and installation types and from both coal
and metal mines. A total of 84 unique fans were used in the total. A visual representation of
the results is shown in Figure 4. The results show that there is a very wide spread of fan effi­
ciencies and lower values than the defaults provided by popular ventilation network simu­
lators. The software values are meant to represent only typical aerodynamic efficiency, except
where labelled specifically. Simple linear regression shows a weak trend of increasing efficiency
with increasing fan output/air power (dotted lines on Figure 4) for both coal and metal fans,
though a typical efficiency of nearly 10% higher for coal mine fans.

Figure 4. Plot of fan system (static efficiency) vs. output/air power, by commodity type.

3.2 Comparison of measured system efficiency by commodity type

Calculated fan system efficiencies were grouped by mine commodity type, shown in Figure 5.
The values demonstrate a wide distribution, with the average fan system efficiency found to be
48% with a standard deviation of 18%. The 38 coal mine fans among the dataset show an
overall average fan system efficiency at 57% ±15%. The 46 metal and non-metal primary fans
considered show an average fan system efficiency of 43% ±17%.
The authors expect the higher average efficiency of coal mine fans is due to the larger, higher-
duty, more permanent fan systems being more heavily engineered. Other contributing factors
include better maintenance of fans located on surface or less blade wear due to lack of blasting,
less abrasive dust, and the prevailing shaft-drive configuration of the larger fans. This could also
be attributed to better definition of the fan output/air power the measurements were more likely
completed using a pitot tube traverse close to the fan blades of an exhausting fan.
Metal mines demonstrated a lower average fan efficiency, likely attributable to smaller size/
duty receiving less engineering and optimization and due to measurement differences. The most
likely contribution to this is that many of the fan total pressure measurements were collected
from measuring the static differential pressure across the bulkhead in which the fan is installed.

357
Figure 5. Comparative histogram of fan system efficiencies.

3.3 Analysis by motor power

Fan system efficiencies were also categorized by motor size for each commodity group. The
comparison of efficiencies grouped by motor size, shown in Figure 6, do not appear to have
a definitive trend when based on motor sizing or absorbed motor power.

Figure 6. Average fan system efficiency by motor power.

4 CONCLUSIONS

When planning for the power requirement of a fan installation it is important to understand the
contributors to fan system efficiency and what assumptions the value includes. For mining appli­
cations, the fan system static efficiency is the relevant metric; except for jet fans, air momentum is

358
regarded as contributing no useful work to the ventilation system (Jorgensen, 1999). The fan
system (static) efficiency is comprised of duct and other system losses associated with the fan
installation as well as motor and drive efficiency and the aerodynamic efficiency of the fan (impel­
ler). The fan system efficiency is calculated by dividing the resulting air power the fan applies to
the ventilation system divided by the total electrical power absorbed by the fan system.
Default simulation software values, published fan curves, and manufacturers’ quoted fan effi­
ciencies all contribute to confusion regarding what these values represent. Fan efficiency as read
from the manufacturer’s performance curve typically represents only aerodynamic efficiency,
one component of overall fan system (static) efficiency. The fan system (static) efficiency is
affected by many factors such as the inlet and diffuser accessories, silencers, dampers, duct tran­
sitions, and/or unfavorable velocity distribution at the fan inlet. Performance and efficiency can
also change over time due to blade wear, deformation of duct components, and wear or damage
to motors and drive linkages and vary as a function of the fan speed/percent of full load.
Operating points and power consumption data were gathered from 84 unique primary fan
installations as part of ventilation surveys to create or update ventilation system simulations.
These data were reviewed collated to provide a guide for planning for the power demand for
main fan installations. The results in Table 1 show that the actual system efficiency for main
fans is lower than what would traditionally be expected. Planning software has default ranges
between (65% and 76%) depending on how the fan was modeled. The fan system efficiency is
important in planning for the total power demand the ventilation system has on the power
grid and directly relates operating power costs.

Table 1. Combined fan system efficiency results.

Type Number Measured Average Fan System Efficiency Standard Deviation

Overall 84 49% 18%

Coal 38 57% 15%
Metal/Non-metal 46 43% 17%

A notable difference is observed between coal and Metal/Non- metal efficiencies with coal
averaging about 13% higher. The main contributing factors between the two is that typically
only one or two fans at coal mines are installed on surface with a well-engineered ducting
system that is easily accessible to maintain. Metal mines often install fan underground in bulk­
heads where additional losses can be generated due to an under-engineered installation (e.g.,
due to lack of space for an inlet bell or diffuser) or subsequent damage to duct components.
Metal mines often install several fans.
Shortcomings recognized in this study include a relatively limited dataset, including lack of
exact description of fan measurement points and method for each fan. The authors acknow­
ledge that further detailed descriptions of the measured fans would provide for more consist­
ent comparison among fan installations.
The collected data suggest that many primary fans are operating at a lower efficiency than
expected. It is common for design purposes to expect a fan operating with a system efficiency
of 75% and fan manufacturers often quote expected operating efficiencies of 80% or higher.
These values account for the aerodynamic and perhaps motor/drive losses, but typically do
not account for other losses necessary in a fan system. Possible reasons for this discrepancy
include the following:
• Fans are not operating at the designed operating point. The fan may have been specified
during an early phase of the mine’s development and the expected total mine resistance
curve is not as expected.
• The fans may need maintenance. Blade and bearing wear can attribute to additional mech­
anical and aerodynamic losses.
• Fans/fan system components are poorly installed. The fan system could be installed such
that excessive losses are experienced at the inlet or discharge.

359
 Ventilation simulations may be hastily performed for the purpose of specifying fan duties
and may not include all these system components, which often reside in a “gray area” between
the modeler’s assumptions and the fan manufacturer’s proposal. Duct and other system losses
may be missed in the specification of fan systems if the system purchaser does not understand
the battery limits. The authors recommend that:
• Purchasing mine engineers review/commission a third-party review of assumed fan efficien­
cies when these assumptions are used for motor sizing and power requirement estimation.
• If a less-detailed model is used (e.g., at an earlier study stage), a lower system efficiency
should be assumed for estimating power requirements to allow tolerance for poorly defined
inlet and discharge losses.
• As study detail progresses, the fan system efficiency value should increase to approach
the manufacturer’s aerodynamic efficiency, especially as the fan model and more of the
components comprising inlet and discharge losses are defined and represented separately
in simulations.

REFERENCES

ASHRAE. 2020. 2020 ASHRAE Handbook: HVAC Systems and Equipment, SI Edition. Atlanta, Geor­
gia: ASHRAE.
Feick, T. 2005. Fan Primer, 2nd Edition. Zweibrücken. Bad Hersfeld, Germany: TLT-Turbo GmbH.
Jorgensen, R, Ed. 1999. Fan Engineering, Ninth Edition. Buffalo, New York: Howden Buffalo.
McPherson, M.J. 2009. Subsurface Ventilation Engineering. Fresno, CA: Mine Ventilation Services, Inc.

360
Mine fires and explosion prevention
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Longwall maingate and tailgate proactive sponcom and gas

management strategy- An operational safety share on risk
management

B. Belle
Anglo American Steelmaking Coal, Australia
University of New South Wales, Sydney, Australia
University of Queensland, Brisbane, Australia
University of Pretoria, South Africa

R. Balusu
CSIRO Mineral Resources, Australia

ABSTRACT: The evolution of major coal oxidation and resulting sponcom incidents are
sudden and may result in catastrophic negative safety outcome or result in the withdrawal of
persons and closure of longwall panels/mines. Historically, gassy longwall workings in Austra­
lian Goonyella Middle (GM) seam (late-1990’s to mid-2010’s), experienced increasing trend in
CO levels associated with coal oxidation and sponcom indicator gases and major safety inci­
dents were due to oxygen ingress on the maingate side. The original Australian active longwall
goaf gas drainage system designs are based on the past work of the CSIRO, supported by the
operational experiences. The CSIRO based studies in gassy and hot coal mines had carried
out numerical and field data investigations on goaf hole gas flow mechanisms and proactive
inertisation strategies for preventative spontaneous combustion and goaf gas management.
This critical foundational knowledge work contributed to the original goaf gas drainage and
sponcom management strategies in other Australian longwall mines and potentially extended
to rest of the world. Considering the risks associated with sponcom, GM seam operations
were the first operations in Australia to introduce proactive N2 injection along the MG in
mid-to-late-2000’s, to manage sponcom fire and explosion risks in an active goaf, despite not
having field data readily available to calibrate the models prior to its implementation.
Over two decades ago, active goaf gas drainage flow rates were moderate (2,000 l/s to 3,000 l/s)
and the oxygen ingress on TG side was not a major concern. However, with increasing goaf gas
drainages rates and manual or automated mode operation of goaf wells to extreme flow rates to
address higher longwall goaf gas emissions, TG oxygen ingress and air wash zones became
a major issue recently, necessitating the introduction of TG inertisation strategies now to address
this emerging issue. Introduction of MG proactive inertisation strategy had ultimately reduced the
number of high CO or intensive oxidation incidents over two decades. This paper provides prac­
tical safety benefits of longwall tail gate (TG) inertisation supported by the original computational
fluid dynamics (CFD) modelling studies carried out by the CSIRO. The field verification with
both MG and TG inertisation using proactive N2 injection during various phases of longwall pro­
duction and stoppages in an active longwall provides reasonable technical and operational justifi­
cations on gas and sponcom management strategy for worker’s safety.

1 INTRODUCTION

The original and modified longwall seam gas drainage system design evolutions are essentially
based on the joint industry led historic initiatives and work of the CSIRO, supported by the

DOI: 10.1201/9781003429241-38

363
GM seam operations (Balusu et al. 2001, 2002, 2004, 2005, 2011, 2017, 2019). The historic
review of goaf drainage introduction in Australia and the rest of the world, drainage design
and operational practices are summarised elsewhere (Belle, 2015, 2017). In addition, the
impact of longwall and TG hole positioning study by the CSIRO for the GM seam operations
is summarised elsewhere (Khanal et al, 2021). The science-based CSIRO studies in gassy,
sponcom prone hot coal mines had been carried out through numerical and field data investi­
gations on goaf hole gas flow mechanisms and proactive inertisation strategies for preventa­
tive spontaneous combustion management. This critical operational knowledge contributed
significantly to the original goaf gas drainage and sponcom management strategies in Austra­
lian underground longwall mines and potentially extended to rest of the coal mining world.
The contradictory nature of the GM seam sponcom led fire and gassy mine requiring maxi­
mised goaf drainage capacity systems and the need to reduce oxygen ingress into the MG and
TG active goaf required careful operational strategies. The term ‘sponcom’ used in this paper
is used to discuss the various stages of coal oxidation to the development of fully uncontrol­
lable combustion resulting in fire when large quantities of coal left in the goaf due to geo-
technical and mining safety considerations. Considering the risks associated with sponcom,
GM seam mines were the first operations in Australia to introduce proactive N2 injection to
manage sponcom related fire and explosion risks in active or sealed goaf areas. In this context,
it is important to note that inappropriate use of ventilation driven controls for longwall tail­
gate gas management that are practiced elsewhere in Australia or in the world, are not neces­
sarily apt for gassy, steep geo-thermal gradient and known sponcom prone GM seam mines.

2 BACKGROUND TO GAS AND SPONCOM MANAGEMENT STRATEGIES

Gassy and known sponcom prone GM seam longwall operations require a greater under­
standing the goaf gas behaviour, post gas drainage control strategies for TG gas management,
and well balanced gas and sponcom management strategies of high magnitude gas reservoir
mines. Following paragraphs below highlight the spectrum of scientific applied research based
engineering controls and monitoring systems developed and improved over the last two dec­
ades in the Australian longwall operations.

2.1 Longwall goaf gas distribution patterns

In order to provide a visual understanding of goaf gas flow patterns for coal mine workers,
operators and ventilation engineers, CFD models of operating longwalls were developed
with operating panel geometries of longwall panel (Balusu et al. 2001) covering 1.0 km
length of longwall goaf using actual floor contours for 2 gate and 3 gate road development
scenarios. A typical longwall schematic with U ventilation system with the total ventilation
quantity of 50 m3/s flows across the longwall face is used in CFD modelling studies are
shown in Figure 1.

Figure 1. Typical 2 gateroad Longwall ventilation system (Balusu, et al. 2002).

364
 At the tailgate (TG) return, an outflow boundary condition was specified in the modelling
simulations. The longwall panel width is 300 m and the roadway width on both maingate
(MG) and tailgate (TG) sides of the face is 5.4 m. The goaf height up to 80 m above the work­
ing seam and the floor strata down to 10 m below the working seam is included in all the
CFD models. The CFD models incorporated MG and TG cut-throughs of 5 m in width and
cut-throughs spaced at 100 m intervals along the panel and goaf drainage holes replicated the
drainage conditions of the operating site. The total number of finite volume cells used for
meshing are around 2.0 million, for obtaining the grid independent solutions in simulations.
For visual understanding purposes, methane and oxygen gas distributions patterns in long­
wall goafs under two different conditions using operational longwall panel gas emissions and
goaf gas drainage conditions with total gas emissions into the longwall goaf of around 9,000 l/s
with 98% methane (CH4) is shown Figure 2. The total goaf gas drainage rate was around
8,000 l/s, with gas concentration in different vertical goaf holes varying between 80% and
95% with adjacent sealed panel goaf drainage of 800 l/s for a typical U ventilation system in
a 2 gateroad panels. In the methane and oxygen gas distribution color contours below, the
red colour indicates higher gas concentration and the blue colour indicates lower methane or
oxygen gas concentration. As noted herein, the presence of oxygen and continued methane emis­
sions are contradictory controls requiring finer balance and continued vigilance in goaf drainage
operations and highly reliable gas trend monitoring.

Figure 2. Methane and oxygen gas distributions patterns in longwall goafs under two different
conditions.

Various parametric studies by the CSIRO verified by the operational data from Australian
mines had indicated that the gradient of the seam, both across the face and along the panel,
had a major effect on the distribution of gasses within the goaf. Similarly, results showed that
gas emission rate and face airflow have a substantial effect on oxygen ingress into the goaf,
particularly into the deep goaf (> 1 to 2 km behind the longwall face). Results show that
intake airflow influenced airwash zone in the goaf with over 10% oxygen concentration levels
has extended further into the goaf with increase in intake airflow. In base case simulations the
high oxygen level zone extended up to 150 - 200 m behind the face, whereas in the case of high
intake airflow the high oxygen zone extended up to 300 m into the goaf (Figure 2).
In the recent years, there is often a preferential emphasis put on the “pressure differentials” or
loosely termed “pressure” in an active goaf for ventilation and gas management. Historic work
in relation to the static pressure distribution in the goaf with traditional practice of gas drainage
from two goaf holes near the longwall face is presented visually in Figure 3 for surface goaf holes
closest to the face operating at the total flow rate of about 1,500 l/s. Results show that with this
type of goaf hole drainage, gas static pressure in the goaf measured at the MG seal builds up to

365
180 Pa, that is sensitive various other mining and natural factors. The time based static pressure
distribution model indicates that all the goaf gas migrates towards the tailgate corner of the goaf
and a major proportion of the goaf gas may escape into the tailgate return airway, particularly
during low barometric pressure periods. Therefore, both the goaf gas drainage strategy and its
operation are paramount in addition to the goaf hole design.
In order to improve the gas drainage system efficiency, deep goaf holes gas drainage strat­
egy was introduced into the modelling simulations. Deep goaf hole drainage in this paper
refer to those vertical goaf drainage holes from surface that drain the goaf gas located greater
than 1 to 2 km behind the operating longwall face and yet times even near the longwall start
up face area for very long panels. The static goaf gas pressure distribution in the goaf with
maximized gas drainage strategy is presented in Figure 3. Results showed that the static pres­
sure development in the goaf with this strategy is only 80 Pa, compared with 180 Pa in the
traditional strategy scenario. Results indicated that goaf gas migrates to a wider area towards
the tailgate side and only a minor proportion of goaf gas escapes towards the tailgate return.
In addition, the low goaf gas pressure development in the goaf helps in reducing the effects of
changes in barometric pressure on return gas levels. These results indicate that the optimum
gas drainage strategy should incorporate goaf holes near the face as well as deep goaf holes
with optimized increased numbers in the panel in order to improve the gas drainage system
efficiency with large longwall block well retreated with large goaf gas reservoir size.

Figure 3. Goaf gas pressure distribution in the goaf with few operating wells behind the face (Left) to
maximized goaf drainage system (right).

2.2 Gas management strategies

As a leading traditional practice, highly gassy mines are to be managed through extensive pre-
drainage techniques long before the actual longwall mining to take place. During the active long­
wall mining, goaf drainage systems are used as the primary control for gas management with
adequate drainage capacity for maximized drainage along with ventilation as the secondary con­
trol for gas management as a dilution control (Belle, 2015). Longwall gas emissions have increased
significantly in recent years in some Australian longwall mines due to increased seam gas reservoir
size with multiple upper and lower seams, higher production rates and increase in mining depths.
In addition, there have been mines, previously deployed 3 gateroad systems in their longwall
panels for continued access to diesel vehicles during maintenance periods and for gas manage­
ment. With the greater understanding with extensive and flexible goaf drainage systems and cap­
acity, Australian coal mines use 2 gateroad U ventilation system for longwall panel development
and extraction. Extensive scientific and field work has been carried out previously to develop opti­
mum gas and spontaneous combustion control strategies for 2 gateroad longwall panels (Balusu
et al. 2001, 2002, 2004, 2005, 2011, 2017, 2019; Belle, 2014, 2015, 2017; Balusu, Belle and Tangu­
turi, 2017, Balusu and Tanguturi, 2019).
Although 3 gateroad system provides more ventilation capacity during gateroad develop­
ment and assists in providing more ventilation dilution capacity in tailgate during longwall
extraction, its effect on goaf gas distribution and explosive fringe gas profiles in the longwall
goaf areas was historically unknown. There is a continued perception that as the 3 gateroad
system provides more ventilation capacity for gas dilution in the longwall tailgate return, it

366
would also reduce the explosive fringe gas distribution profile near the tailgate area in the
longwall goaf to manage the explosion risk. The results of the CSIRO CFD modelling simula­
tions calibrated with field conditions indicated that there is a significant difference in the
spread of explosive fringe gas distribution profiles in the longwall goaf under 2 gateroad and 3
gateroad conditions, i.e. a significant increase in the spread of explosive fringe (or close to
explosive range) zone in the goaf under 3 gateroad conditions. Based on the results of these
investigations, appropriate strategies have been developed for gas control and minimization of
the spread of explosive fringe gas distribution in the longwall goaf (Balusu, Belle, Tangutiri,
2021). Therefore, based on the extensive operational gas management experience, it is the pre-
drainage and longwall goaf gas drainage management is the primary gas management control
rather than ventilation engineering controls to manage the major gas hazard.
One of the major difficulties in the ventilation and gas flow dynamics is our inability to visualize
the complex likely gas concentration profiles, i.e., methane or oxygen in the active goaf with time
and non-constant retreating longwall. The advances in the CFD numerical calculations have
enabled the industry by providing an understanding of gas management or the extent of ‘air wash
zone’ that is often used colloquially during risk assessment or emergency situations. In this paper,
air wash zone is typically referred to as the concentration of relative oxygen in the goaf atmos­
phere aiding the left over coal oxidation, and typically referred to with oxygen concentration
values of 3 % to 21%. Fresh air is the concentration that is representative of longwall fresh air
intake. It is to be noted that due to almost no relative airflow movement that can be measured in
the goaf may also mean the presence of oxygen even at levels of 2 % to 3 %. The CSIRO studies
(Balusu, 2002) provided a visual scenario of potential oxygen distribution in an active LW goaf
and goaf well nearer to the LW face. The field studies have noted that (Balusu et al, 2006) the
oxygen concentration was above 19 % for up to 100 m behind the longwall face and reduced to 6
% at 250 m behind the face in the absence of any inertisation control. This air penetration dis­
tances of 250 m to 350 m may be mainly attributed due to poor MG brattice control practices
along with the increased longwall airflow rates for gas dilution purposes and inadequate gas drain­
age. The tracer gas studies have revealed that the goaf at 300 m behind the face is highly consoli­
dated and does not allow direct travel of air fom the intake side to return side of the TG.
Similarly, methane gas concentration distribution profiles at the tailgate region of different
mines vary significantly depending on the geological, gas, mining and operational conditions. Gas
concentration distribution profiles (Balusu, 2020) at the tailgate area under three mining condi­
tions are presented in Figure 4. The white box in the plots show the TG motor area and white line
replicating the Bretby across the face. In addition, the gas distribution profiles even at the same
mine can vary significantly depending on number of conditions, including changes in barometric
pressures, goaf falls, gas emission rates, goaf gas drainage efficiency, face location with respect to
goaf holes and cut-throughs, face creep, floor contours, caving conditions behind the face and in
gateroads, coal production rates, face ventilation, face cutting and chock advance sequences. Gas
concentration distribution profiles and potential flammable gas mixture zones near the tailgate
roadway area of the longwall face are dynamic and complex in nature and varies widely depend­
ing on the changes in above parameters during mining operations. Thus the behavior of goaf gas
composition is complex and influenced by mining engineering and fluid dynamics and are to be
assessed by suitably skilled, qualified and experienced expert to provide appropriate guidance to
safe operation of coal mines.
An example comparison of the methane and oxygen gas concentration distribution patterns
in the longwall goaf near the tailgate area in a 3 gate longwall retreat scenario are presented in
Figure 5. Results of this simulations indicate that the methane gas distribution inbye of the
longwall face is close to the explosive range in both cases which reflects the goaf drainage hole
design and their operational effectiveness. The contour scale provides the methane and oxygen
distribution. For example, 0.207 % equal to 20.7 % oxygen and 0.40% equal to 40 % methane.
In the Figure 5, “closed at location 1” signifies the temporary roadway seal. The results of the
CFD modelling simulations indicate that there is a significant difference in the spread of explo­
sive fringe gas distribution profiles in the longwall goaf under 2 gateroad and 3 gateroad con­
ditions, i.e. a significant increase in the spread of explosive fringe (or close to explosive range)
zone in the goaf under 3 gateroad conditions or partial 3 gate road LW operations.

367
Figure 4. A snapshot of methane gas distributions profiles near the tailgate area of longwalls.

Figure 5. Close up views of methane and oxygen distribution inbye of the longwall face (Balusu, Belle,
Tangutiri, 2021).

2.3 Sponcom management strategies

Based on the two decades of close collaboration between the coal operators, ACARP and
CSIRO, extensive scientific and field-based studies have been carried out. The learning from
these studies, have resulted in the following summary and context behind the proactive pre­
ventative sponcom and gas management goaf inertisation strategy:
• Major events in Queensland demonstrates the critical importance of proactive sponcom
management for underground coal mines extracting/working in known sponcom prone
Moranbah region GM seams.
• Widely referred low sponcom propensity (R70) of coal risk ratings in Principal Hazard
Management Plan (PHMP) documents and frequency of their testing may be misleading
the likely initiation or risk frequency estimations. For example, both German Creek and
Goonyella Middle Seam R70 values are similar in magnitude but the left-over roof coal in
GM seam goaf increase the oxidation risk with increasing depth due to steep geo-thermal
gradient.
• Historically, gassy longwall workings in Australian Goonyella Middle (GM) seam (late-
1990’s to mid-2010’s), experienced increasing trend in CO levels associated with coal ox-
idation and sponcom indicator gases and major safety incidents were due to oxygen in-
gress on the maingate side. To address this issue, MG proactive inertisation strategy was
introduced at GM seam operations mid-to-late-2000’s, which ultimately reduced the
number of high CO incidents over the next decade.

368
• When MG proactive inertisation strategy developed by the CSIRO and the operators was
first introduced, there was no precedence in Australia and there was no field data to valid­
ate its effectiveness (prior to its implementation). However, it is to be noted that during
longwall operations of GM seams, it’s the additional proactive N2 inertisation strategy that
was essential to successfully manage the sponcom and resulting major fire risks during long
periods of production stoppages due to geo-technical and mining related matters.
• The evolution of major coal oxidation and sponcom incidents are sudden and may result in
catastrophic negative safety outcome or result in the withdrawal of persons and closure of
panels/mines.
• In view of the recent incidents in a number of mines working in GM seam (irrespective of
the cause of the incidents), elevated oxidation reaction may potentially become an ignition
source and inadequate control may result in the undesirable safety outcome.

2.4 Maximised goaf drainage Strategies -Lessons Learned

To develop optimum and effective goaf gas drainage strategies for any new or operating mine,
an extensive goaf gas monitoring scheme should be implemented in at least one or two panels
to obtain detailed information on gas flow patterns and goaf gas distribution under various
operating circumstances for the site conditions and geometry. In many cases, the recent stand­
ard practice of draining gas from 2 to 5 goaf holes near the face operating its peak capacity
would not solve the tailgate gas problems, but exacerbate the oxygen ingress into the deeper
portion of tailgate area of an active LW goaf. A number of factors including goaf gas emission
flow rates and composition, panel ventilation, coal seam gradients, overlying and underlying
coal seams, face retreat rates, caving characteristics, and goaf gas flow patterns need to be
considered during development of goaf gas drainage strategy and goaf hole operations.
Based on the results of various CSIRO studies and investigations supported by the coal
operators over the last two decades, the following practical guidelines are recommended for
optimum maximized goaf gas drainage strategies at highly gassy mines:
• Surface goaf holes for gas drainage provide the highest capacity, flexibility and lowest total
cost option for goaf gas drainage under most circumstances.
• Goaf holes should be drilled on the return side of the goaf, preferably at 20 to 70 m from
gateroad depending on the longwall caving conditions.
• Goaf holes are to be drilled 80 m to 100 m away from faults/dyke areas to overcome coal
oxidation and spontaneous combustion risks.
• Uniform, stable and continuous operation of goaf holes are essential for effective longwall
ags management (sudden peaks and lows in goaf drainage flow rate increases the coal oxi­
dation potential resulting in sponcom risk).
• Goaf gas drainage hole diameter should be in the range of 250 to 400 mm for optimum
flow rates and the goaf holes may be drilled at 50 m to 200 m spacing depending on the
longwall gas reservoir size, goaf gas emissions and other mining production conditions.
• The total capacity of the goaf gas drainage plants should be around 2 to 3 times the expected
goaf gas emissions to cater for deep goaf holes gas drainage, shifting of goaf plants or goaf
hole connection changes and reduced plant efficiencies due to high pressure losses. Provision
of a high-capacity and flexible gas drainage system allows optimisation of goaf gas drainage
strategies, maximized goaf drainage with retreating longwall, flexibility, improves the overall
efficiency and provides greater gas management control on the longwall face.
• The goaf gas drainage system should include a combination of goaf holes near the face and
deep goaf holes in the panel in order to improve the overall gas drainage efficiency and to
reduce the effects of barometric pressure changes on tailgate gas levels with increasing
active goaf gas reservoir size.
• The strategy of continuous operation of deep goaf holes at low to moderate flow rate
should be implemented i.e., intermittent operation of deep goaf holes at high capacity may
not improve the overall efficiency and may lead to unexpected problems such as oxygen
ingress into deep goaf.

369
• Goaf gas drainage should be carried out from around maximised number of goaf holes
with the retreating longwall goaf in the panel (including deep holes), instead of the standard
practice of gas drainage from just few goaf holes closest to the face for gassy high produc­
tion longwall mining.
• Application of increased suction pressure to drain more gas from goaf holes closest to the
face might result in increased air dilution, without any net increase in gas drainage flow rates.
• The ventilation system in the panel should be designed to minimise oxygen ingress into the
goaf, including immediate sealing-off all the cut-throughs behind the face, MG tight brat­
tice control in order to improve overall gas drainage efficiency.
• Oxygen concentration level in the goaf hole flow should be less than 5% for extended
periods of time in goaf holes beyond 100 m from the LW face line to reduce sponcom risk
in the longwall goafs during normal production and long periods of stoppages.
• Gas drainage from adjacent old goafs should also be carried out wherever possible, depend­
ing on the goaf gas emission flow rates and adjacent seal strengths (Belle, 2014).

3 LW GAS AND PROACTIVE MG AND TG INERTISATION AND MONITORING

STRATEGY

In order to manage the elevated oxidation levels with increased goaf gas drainage in high pro­
duction gassy mines to effectively manage the active longwall TG gas levels, following pro­
active inertisation strategy was implemented in an operating mine. The optimum locations of
the TG and MG holes for both gas and sponcom management with proactive inertisation for
high gassy mines were based on the original and fundamental goaf gas and sponcom manage­
ment work by the CSIRO, calibrated with the operational experiences over the last decade
(Balusu, 2021).

3.1 Data collation and limitations

The study had analysed extensive longwall panel data, that had implemented for the first time,
a strategy that incorporated MG and TG nitrogen (97%N2) inertisation and goaf gas manage­
ment. The limitations and characteristics of the data used for analyses are summarized briefly
below:
1. The goaf drainage hole data which recorded CO, CO2, CH4, flow were averaged daily for
individual holes collected for the entire longwall panel.
2. Active goaf gas composition was based on the daily individual average goaf hole data com­
prising, flow, CH4, O2, CO, CO2 composition.
3. Goaf gas flow data with flow rates of > 200 l/s are used in the final analyses, as some of the
new goaf holes were usually checked for validating the goaf connections once the longwall
had retreated 20 m to 30 m behind the longwall face line.
4. It is to be noted that the data flow rate < 200 l/s is usually associated with the start-up of
the goaf hole prior to LW intersections with higher levels of O2 (> 14%) for short periods.
5. Negative O2 readings were removed from the data set considering the likely data flaws or
sensor measurement errors.
6. It was noted that the data set yet times showed that during the normal operation of goaf
wells probably based on TARP trigger levels, resulted in undesirable gas composition
mixtures.
7. The data contained over 90 vertical surface goaf drainage holes with a combination of
deep goaf, adjacent goaf, MG and TG goaf week operations of the entire longwall panel
for effective longwall TG gas management.

370
3.2 LW gas management and proactive MG and TG inertisation assessment
An extensive data analyses was carried out for the entire longwall panel for the gas and spon­
com management effectiveness and are discussed hereafter. Figure 6 shows the profiles of
active longwall goaf gas composition with retreating longwall with likely less than effective
inertisation and the absence of TG inertisation.
Key observations from the analyses are as follows:
• The effectiveness of the N2 injection and the operation of goaf gas flow rate management is
reflected in the gas composition of the active longwall goaf gases, viz., CO, O2, CH4.
• The composition of higher levels of oxygen behind the longwall face in the TG region (upto
300 m behind the face) may be attributed to the very higher flow rates and goaf holes
coming online at the time of goaf formation.
• Similarly, higher daily average CO levels, demonstrate the presence of the excessive and
conducive oxygen rich environment as a result of deeper air wash zone in the TG area of
the goaf contributing towards early oxidation of left over coal in the goaf.
• It is equally noted that the methane concentration behind the active longwall in overall
terms increases in the purity as a result of continued desorption of overlaying undrained
coal seams with greater gas reservoir size. The failure to drain the deeper portion of the
active longwall goaf will eventually travel towards the tail gate region as a result of increas­
ing goaf gas reservoir size with the longwall retreat and definitively contributes in higher
gas levels during the steep drop in barometric pressures.
• It is equally important to note that the uncontrolled deep goaf hole operation at very high
flow rates, may further exacerbate the oxygen rich environment deeper in the TG goaf area
in the absence of effective TG inertisation.

Figure 6. Goaf hole gas composition without the TG inertisation with retreating LW face- Oxygen
(Top Left), CH4 (Top right), CO (Bottom Left); Flow rate (Bottom right).

Similarly, Figure 7 shows the profiles of active longwall goaf gas composition with retreat­
ing longwall with effective MG inertisation and with the introduction of TG inertisation for
the first time.
Key observations from the analyses are as follows:

371
Figure 7. Goaf hole gas composition with MG and TG inertisation with retreating LW face- Oxygen
(Top Left), CH4 (Top right), CO (Bottom Left); Flow rate (Bottom right).

• Combined MG (1500 L/s) and TG inertisation rate of 750 L/s with greater goaf gas man­
agement is clearly reflected in the favourable gas composition of the active longwall goaf
gases, viz., CO, O2, CH4.
• The composition of reduced levels of oxygen behind the longwall face in the TG region
(upto 100 m behind the LW face) may be attributed to the N2 injection in the tailgate min­
imising the oxygen ingress into the goaf. There is a significant difference in the oxygen
levels at less than 5 % beyond 100 m of the inbye goaf assisting in minimising the risk of
coal oxidation and oxygen presence in the deeper portion of the goaf.
• Similarly, daily average CO levels in the active goaf have significantly reduced, demonstrat­
ing the effectiveness of the N2 injection and less than adequate conducive oxygen presence
in the TG airwash zone, thus potentially minimising early oxidation.
• As before, the methane concentration behind the active longwall in overall terms increases
in the purity as a result of continued desorption of overlaying undrained coal seams with
greater gas reservoir size and aided by the N2 presence. This environment enables the
deeper goaf drainage practices at low to moderate flow rate for maximised goaf gas man­
agement and drainage efficiency, to reduce the gas reservoir as well as ventilation air
methane (VAM) and greenhouse gas (GHG) management.

3.3 Discussions of MG and TG inertisation and gas management strategy

Presence of oxygen in the active goaf is unavoidable when carrying out goaf drainage activities
to manage tail gate gas levels of an active longwall. Extensive goaf seal monitoring activities
by the industry have provided greater understanding of the goaf dynamics and input to the
goaf gas composition studies for over two decades. Figure 8 shows one such historic work
that informs the impact of oxygen profile extent as a result of proactive N2 injection on both
TG and MG areas. The reduced air wash zone is clearly evident in the active goaf thus minim­
ising the conducive environment for any elevated oxidation events (Balusu, 2005). Figure 9
provides the latest CSIRO study using calibrated CFD model of MG and TG inertisation
strategy for an operating high gassy longwall mine.
Based on the aforementioned gas composition analysis of field goaf hole production data,
the goaf gas distribution under intensive gas drainage is summarised as a conceptual model is
shown in Figure 10 (Xiang et al, 2021). This drawing may be very subjective—specifically air­
wash zone contours can shift depending on the complex goaf hole operations for

372
Figure 8. Historic MG and TG inertisation conceptual strategy for longwalls (Balusu, 2005).

Figure 9. Latest MG and TG inertisation strategy for an operating gassy longwall mine (Balusu, 2021).

U ventilation system with varying goaf hole operational controls and designs without pro­
active inertisation. The goaf drainage operation suggests that greater oxidation may be pos­
sible in the TG, while there is minimal occurrence of it at MG area with greater ventilation
controls such as tight brattice controls and timely build up of seals.

Figure 10. A conceptual model of goaf gas environment under intensive goaf gas drainage impact.

The historic field studies (including tracer gas and goaf hole shut off studies) of CSIRO
have shown that reducing and increasing the flow capacity of goaf holes, including complete
shut-off of the holes and measuring corresponding changes in return gas levels and response
times have shown that 80 % to 90% of goaf gas migrated to longwall return within 1 to 3.0 hrs
when goaf holes were 100 to 400 m from the longwall face. Response times varied between
two minutes (150 m behind the face) to an hour (1000 m behind the face) along the MG and
on the TG goaf holes with few minutes to several hours depending on the goaf hole designs

373
and caving characteristics. Field studies have shown that even when goaf holes located more
than 1,000 m from the face, had a substantial effect on gas flow dynamics and on longwall
return gas levels. The recommended strategy was that the deep goaf holes are to be operated
at low to moderate flow rates continuously as long as the oxygen levels below 3 % and no
increased trend in the CO levels.
One of the additional challenges often faced during an unfortunate gas event leads to pleth­
ora of suggestions without having appropriate science and data based studies or specialist
expertise in the field. One such view is that the goaf drainage flow rate has no impact on
oxygen in goaf well or active goaf. Figure 11 below shows the relationship between goaf hole
flow rates and goaf hole oxygen levels at various distances from the active longwall face.

Figure 11. A conceptual model of goaf gas environment under intensive goaf gas drainage impact.

It is observed from the above plot that when the goaf holes are deep, the O2 levels are typic­
ally very low and lower flow rates help to minimize the goaf reservoir size with the retreating
longwall with major rate of barometric pressure drop. The general observation is that the
increase in goaf flow rate to the extreme regions causes significant increase in the O2 levels in
the goaf, which would require inertisation to minimize the potential oxidation. It’s the moderate
flow rate with increased number of operating wells against few number of wells operating at
extreme flow rates assists in managing the TG gas levels. While the sole intent is to minimize the
TG gas levels, the risk of operating few wells at maximum flow rates behind the longwall face
would certainly bring the fresh air or oxygen into the active goaf of a retreating longwall. It
must be noted that the goaf hole design including its distance from the working seam (historic
design of ideally 10 m or 40 ft as in the USA) will have the impact on TG gas levels as well as
oxygen levels in the active goaf. Therefore, it’s the maximized goaf drainage of active goaf with
moderate goaf flow rates and increased number of holes supported by the active inertisation
would assist in the effective management of gas and sponcom management risks.

4 CONCLUSIONS AND RECOMMENDATIONS

Conjointly managing the gas and sponcom risk is fundamental to securing a safe underground
place of work at GM seam longwall operations. With the increasing gassy and known spon­
com prone coal seams and a working depth with steep geothermal gradient is contributing
towards the step changes in controls required for gas and sponcom management in order to
be compliant with the safe TG gas limits as well as reduced ventilation air methane (VAM)
emissions. This strategy reinforces the fundamental importance of the pre-drainage systems
with long lead drainage time prior to longwall mining. Over two decades ago in Australia,
goaf gas drainage rates in Qld and NSW were low to moderate (1,000 l/s to 3,000 l/s) and the
ingress of airwash zone on longwall TG side was not a major concern. However, in the recent

374
years, with increasing goaf gas drainages rates upto 6,000 l/s 10,000 l/s and manual or auto­
matic mode operation of goaf wells to the extreme flow rates to address higher goaf gas emis­
sions, TG oxygen ingressing deeper in the goaf has become a major issue, necessitating the
introduction of both TG and MG inertisation strategies now to address this emerging coal
oxidation risk. Furthermore, contrary to the views in relation to elevated coal oxidation and
sponcom events related to goaf hole spacing and maximized goaf drainage practices, it is pru­
dent to note that there have been historic cases of sponcom events with 400 m to 200 m goaf
hole spacing, and even with no goaf drainage practices.
Following safety benefits reasoned with adequate technical and operational justifications
for gas and sponcom management strategy aided with both MG and TG inertisation using
proactive N2 injection during various phases of longwall production and stoppages:
• In the absence of active and continuous proactive N2 inertisation, maintaining 5 % to 8 %
O2 levels in the active goaf would be very difficult and may exacerbate the oxidation in an
active goaf when the longwall retreat rates slows down or stops for weeks and months. On
the contrary, sub optimal goaf drainage to manage the O2 ingress in deep goaf area of TG
region will significantly increase the longwall TG gas levels. This approach would put oper­
ations in a dangerous position from gas management perspective.
• Based on the data analyses evidence of manual daily data on O2, it is noted that there is no
clear trend in terms of goaf well spacing justifying the oxygen management, rather the
lower O2 levels in deep goaf or when the LW face is further away from the goaf well, are
conducive to operate at lower flow rates for steady deeper goaf drainage holes for gas man­
agement for retreating longwall of growing goaf gas reservoir size.
• It is the operational management of goaf wells (not sudden or automatic operation of the
goaf wells to the extreme flow rates, rather stepwise increase) with proactive LW active
goaf inertisation for sponcom management will enable the appropriate maximised goaf
drainage gas management to manage the longwall return gas levels and oxidation risks.
• The introduction of TG inertisation assists in reducing the airwash zone. In addition, long­
wall operations need to continue with the leapfrogging of well-established MG seal inertisa­
tion strategy [including quicker and timely MG seal build, and tight MG brattice control],
along with the dedicated tube bundle monitoring points for goaf gas monitoring to under­
stand the goaf flow dynamics and sponcom management.
• As a general long-term strategy, known sponcom prone longwall operations to ensure flexible
and contingency inertisation infrastructure is readily available and to remain in a state that it
is able to be recommissioned within a single shift at any time during the monitoring period.
• The association of slow retreat due to known or unknown geological structures in the long­
wall hazard plan and known historic oxidation related incidents, have further reinforced
the active inertisation system maintaining at least 2,200 l/s of inert flow into the active
panel [around 1200 l/s on the MG side and 1000 l/s on the TG side using dedicated vertical
N2 holes drilled 5 to 10 m from the working seam height].
• Considering the various uncertainties associated with the LW operations in sponcom prone
seams, recommendation of inertisation holes at 200 m spacing on TG side is essential and
appropriate for long term risk management planning and design purposes. If the evidence
suggests otherwise, i.e., constant increased retreat or no stoppages, then TG proactive iner­
tisation may be carried out through inertisation holes at increased spacing (i.e. alternate
inertisation holes) in those areas. This would mean that the intermediate dedicated holes
and other old/deep inertisation holes can be equally used for oxygen ingress or airwash
zone monitoring and sponcom monitoring on the TG side of the goaf as per the oxygen
limit recommendations of an active goaf.
• It is to be noted that these gas and sponcom management strategies are not merely based
on the LW retreat rate, but also includes the inherent nature of seam propensity for spon­
com despite it being equally rated as “low” risk sometimes, geo-thermal gradient, existence
of faults and structures, amount of coal left behind, changes to the goaf hole designs, delay
in operational related building of MG seals, inadequate MG brattice leakage control for
long periods. other engineering related uncertainties associated with the LW equipment,

375
 strata control uncertainties associated with moisture/water in TG roadways, and uncertain­
ties associated with cavity control measures.
• The implemented TG and MG proactive N2 inertisation strategy of active longwall goaf at
the sponcom prone seam has resulted in the new technical/empirical data generation in veri­
fying the fundamental understanding of goaf gas drainage maximization and sponcom
management.
• Finally, strengthening the proactive inertisation strategy on both MG and TG with flexible
inertisation capacity and responding to the up to date trigger response values of the oxidation
scenario rapidly developing into an advanced stage using appropriate early monitoring strat­
egy is essential for the future proofing of underground sponcom risk.

REFERENCES

Balusu, R., Deguchi, G., Holland, R., Moreby, R., Xue, S., Wendt, M. & Mallett, C. (2001). Goaf gas
flow mechanics and development of gas and sponcom control strategies at a highly gassy coal mine.
Australia-Japan Technology Exchange Workshop, 3-4, December 2001. Hunter Valley, Australia. 18
pp.
Balusu R, Mallett C, Xue S, Wendt M, and Worrall R (2002). Mine Gas Control – JCOAL, CSIRO and
Dartbrook Collaborative project Final report. CSIRO Exploration and Mining Report No. 1012C,
March/October 2002, 183 pp.
Balusu, R., Tuffs, N., Peace, R., Harvey, T., Xue, S., Ishikawa, H. (2004). Optimisation of Goaf Gas
Drainage and Contro Strategies, ACARP Project C10017, CSIRO Exploration and Mining Report
1186F, 149 pp.
Rao Balusu, Ting X Ren and Patrick Humphries,(2005). Proactive Inertisation Strategies and Technol­
ogy Development, ACARP Project C12020, CSIRO Exploration and Mining Report P2006/26, 114
pp.
Balusu, R., Schiefelbein, K., Ren, T, O’Grady, P, Harvey T. (2011). Prevention and control of fires and
explosions in underground coal mines, The Australian Mine Ventilation Conference, Sydney, Austra­
lia, September 2011, pp 69–77.
Balusu, R., Belle, B., Tanguturi, K. (2017). Development of gas and spontaneous combustion control
strategies for 6.0km long longwall panels, 16th US Mine Ventilation Symposium, Golden, Colorado,
USA, June 2017, pp 18–11 – 18-18.
Balusu, R., Belle, B., Tanguturi, K. (2019). Development of Goaf Gas Drainage and Inertisation Strat­
egies in 1.0-km- and 3.0km-Long Panels. Mining, Metallurgy and Exploration. 2019; 36(6):1127–1136.
https://doi.org/10.1007/s42461-019-0071-9.
Balusu, R., Tanguturi, K. (2019). Gas management and risk mitigation strategies for longwalls, ACARP
Project C25066, CSIRO Energy Report EP191657, March 2019, 210 pp.
Balusu, R, Tanguturi, K, Belle, B. (2021). Goaf gas distribution profiles near the longwall tailgate area,
Proceedings of the 18th North American Mine Ventilation Symposium, 12-17 June, 2021, Rapid City,
South Dakota, USA, Edited ByPurushotham Tukkaraja.
Balusu, R. (2020). Longwall Tailgate Gas Compositions, Internal Industry Note, Australia
Balusu, R. (2021). Unpublished CFD Model Study on MG and TG inertisation Strategies, CSIRO.
Belle, B, 2014. Evaluation of barometric pressure and cage effect on longwall tailgate gas levels for pre­
vention of explosion in coalmines, IMVC, Johannesburg.
Belle, B. (2015). Innovative tailgate mobile goaf gas management in two gateroad longwall panels – con­
cept to implementation”, The Australian Mine Ventilation Conference, Sydney, Australia.
Belle, B. (2017). Optimal goaf hole spacing in high production gassy Australian longwall mines – oper­
ational experiences, The Australian Mine Ventilation Conference, Brisbane, Australia.
Khanal, M, Poulsen, B, Adhikary, D, Balusu, R, Wilkins, A, Belle, B. (2021). Numerical Study of Stabil­
ity and Connectivity of Vertical Goaf Drainage Holes. Geotechnical and Geological Engineering.
2021; 39(3):2669–2679. https://doi.org/10.1007/s10706-020-01650-6.
Xiang, Z, Si, G, Wang, Y, Belle, B, Webb D. (2021). Goaf gas drainage and its impact on coal oxidation
behaviour: A conceptual model, International Journal of Coal Geology, Volume 248, 103878.

376
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Real-time methane prediction with small dataset in underground

longwall coal mining using AI

D.C. Demirkan, H.S. Duzgun, A. Juganda & J. Brune

Mining Engineering Department, Colorado School of Mines, Golden, CO, USA

G. Bogin
Mechanical Engineering Department, Colorado School of Mines, Golden, CO, USA

ABSTRACT: Detecting the development of explosive methane–air mixtures on a longwall

face remains a difficult task. Even when atmospheric monitoring systems and computational
fluid dynamics modeling are used to inspect methane concentrations, they are insufficient as
a real-time warning system in crucial areas, such as near cutting drums. In our previous work,
the long short-term memory algorithm has been adopted in order to predict and manage
explosive gas zones in longwall mining operations prior to explosions. It was found that faster
and more prominent coverage of accurate real-time explosive gas accumulation predictions is
possible using the proposed algorithm. However, the data that was used for training and test­
ing was more than 12TBs. In this work, we evaluated the prediction accuracy of the proposed
methodology with a smaller data set. The data was developed and extracted based on CFD
model outputs with similar locations and operational characteristics. The results indicate that
the algorithm can forecast explosive gas zones in 3D with an overall accuracy of 82.3 percent
for a small size dataset (n=1000).

1 INTRODUCTION

Despite alternative energy sources, worldwide coal production is still increasing each year
(Dudley and others 2018). Longwall mining is the most utilized coal mining method, due to its
high productivity and safer operating conditions (Peng 2019). However, in usual mining oper­
ations and conditions, coal mining still faces serious challenges (Qiao et al. 2022). Despite the
advancements in technology and safety management, longwall face explosions from accumu­
lated methane gas are known to be the most common cause of methane explosions (Juganda
2020a). Existing industry practices depend on point-type methane sensors in critical regions to
prevent explosive gas accumulations (Juganda et al. 2020). However, point sensors are not
reliable at spotting and warning about explosion hazards, especially in crucial areas, such as
near the cutting drum, tailgate, and headgate areas of the longwall face (Davis, Engel, and
van Wingerden 2015). One catastrophic example of a methane explosion accident was the
2010 Upper Big Branch Mine in West Virginia, U.S. (Davis et al. 2015). Although atmos­
pheric monitoring systems can report real-time methane concentrations, they fall behind due
to their limited number of sensors and locations (D C Demirkan et al. 2022), which lack full
coverage of the whole longwall face. Computational fluid dynamics (CFD) was employed to
simulate ventilation conditions in longwall faces to reproduce airflow aerodynamics and the
formation of hazardous gas mixtures, which are not detectable with conventional monitoring
and ventilation inspection practices (Juganda et al. 2020). Although CFD modeling can accur­
ately predict explosive gas zones, high computational power and time requirements diminish
its use for real-time ventilation monitoring purposes (D C Demirkan et al. 2022).

DOI: 10.1201/9781003429241-39

377
 In 2008 Karacan et al. proposed principle component analysis and an artificial neural network-
based approach to predict methane emission rate throughout 63 longwall mines. The study
showed that the volume of daily methane emission from each mine could accurately be predicted
(Karacan 2008). In the following years Dougherty and Karacan utilized the prediction model in
Karacan’s previous work (Karacan 2008) and developed software that can predict ventilation
emissions with elastic properties (Dougherty and Özgen Karacan 2011). In 2018 Duda and Krze­
mień proposed a framework for forecasting methane emissions from seams to goafs; and predicted
the average volume of methane per minute for each year of mine life (Duda and Krzemień 2018).
In 2021 Sidorenko et al. provided the necessary parameters to predict methane emissions from
seams to goafs (Sidorenko, Dmitriev, and Sirenko 2021). Although these studies discussed predic­
tion, these predictions were neither real-time nor spatial outputs.
Previous studies demonstrated the successful implementation of Artificial Intelligence (AI) in vari­
ous fields with real-time predictions. For example, Nyanteh et al. implemented an AI for real-time
fault detection (Nyanteh et al. 2013). Dong et al. used the AI model for real-time monitoring and
prediction of slope failures (Dong et al. 2021). Rodríguez-Rangel et al. incorporated big data analy­
tics for autonomous vehicles’ speed estimation (Rodríguez-Rangel et al. 2022). Wahyono et al. pro­
posed combining AI with data mining for real-time forest fire detection (Wahyono et al. 2022).
Real-time prediction of methane in a longwall face requires predicting explosive gas zone forma­
tion in time, and its location in 3D (x, y, and z coordinates). Predictions in time are mainly con­
ducted using time series classification/analysis (Bagnall et al. 2017). Spatiotemporal AI models and
time series classification are relatively new to the field. Moreover, the vast majority of successful
models take into account only one or two spatial dimensions, such as x and/or y coordinates, and
the data of interest (Dempster, Petitjean, and Webb 2020; Deng et al. 2013; He et al. 2016; Ismail
Fawaz et al. 2020; Li and Principe 2017; Lines, Taylor, and Bagnall 2018; Lucas et al. 2019; Schä­
fer and Leser 2017; Shifaz et al. 2020). For example, climate change studies that track carbon emis­
sions use latitude, longitude, and carbon content (Alléon et al. 2020), whereas water quality
studies use one-dimensional distance of the intersections and water content (Li 2011)
Current explosive gas accumulation monitoring practices in longwall coal mines rely on two
methods, namely, point sensors and CFD modeling. Point sensors take real-time measurements
along the face and provide methane concentration values at the installed locations. CFD modeling
provides methane prediction for the whole longwall face. However, due to computational cost and
load, predictions take days or weeks, depending on the resolution of the study area. This study is
a continuation of our previous works, i) benchmarking and analyzing the suitability of the dataset
and off-the-shelf algorithms (Demirkan 2022), and ii) adopting a modified long-short term
memory (LSTM) algorithm and methodology as a real-time explosion hazard warning system
(Doga Cagdas Demirkan et al. 2022). This study explores the use of a 3D spatiotemporal predic­
tion model for real-time prediction of methane in the coal face with a small data set.

2 DATA PREPARATION

Version 18.2 of the Ansys Fluent software was used to model and simulate a longwall face. The
modeled longwall face had a length of 300 meters, a mining height of 3 meters, and a depth of 6
meters. Also represented were two basic pieces of equipment: i) support equipment (shields) and
(ii) cutting equipment (shearers). Each shield was 2 meters long and was affixed to the model. One
10 m long shearer was set in the headgate (HG) part of the long mine and between 265 to 275
meters. The region that was modeled was then covered with about one thousand hexagonal and
octagonal meshes. The simulation exemplified a transient CFD model of methane (CH4) gas emis­
sion from the coal face based on a bleeder ventilation system with a tailgate (TG) back return
setup. The transient model was run for 20 seconds with 1-second recording intervals. Raw data
from the Fluent software were preprocessed, which included data conversion into CSV files,
removal of empty fields, the addition of cutting directions, and correction of all fields’ two signifi­
cant numbers. The LSTM network, which had been modified in our previous work (Doga Cagdas
Demirkan et al. 2022), was trained and tested with new CFD outputs. The inputs to the modified
LSTM model were; x, y, and z coordinates, airflow velocity, methane concentration, and the

378
volume of each cell for 20 seconds, which were recorded at 1-second intervals. In Figure 1, long­
wall mining model along with the shearer location and data collection locations are given.

Figure 1. Sampling locations.

3 RESULTS

This research utilized a high-end gaming PC with the following specifications: CPU: 10th Gen
Intel Core i7-10875H (8-Core, 16MB Cache, up to 5.1GHz with Turbo Boost 2.0), GPU:
NVIDIA(R) GeForce RTX(TM)2070 8GB GDDR6, Memory:32GB DDR4 2666MHz, Stor­
age: 2TB PCIe M.2 SSD. The training of the algorithm took around six hours, and the time
required to make a prediction was twenty seconds.
Depending on the dataset, the usual practice of separating data into train, validation, and
test sets can range from 50–50 percent to 80–20 percent. When there are fewer data points (n
< 10,000), 80%–20% splitting is utilized. With higher numbers of recordings, however (n >
1,000,000), the significance of split ratios diminishes and ratios can be 50%-50% . Overall,
good data representation in the train and test sets is the most critical aspect of partitioning
a dataset (Dubbs 2021; Gholami, Torkaman, and Dalir 2019; Gholamy, Kreinovich, and
Kosheleva 2018; Joseph 2022; Owusu, Bansah, and Mensah 2019). In this study, there were
one thousand recordings; hence, training and testing were split as 80% to 20% for each
instance using the stratified K-folds cross-validation method. This resulted in a balanced data
division that maintained the proportion of samples for each methane content. Eighty percent
of the data was used to train the model.
Validation data were separated from training data, which validated the AI’s performance.
Training and validation accuracy helps users evaluate their mode. Figure 2 illustrates
a commonly used metric for assessing algorithm performance, validation, and training accur­
acy versus an epoch.
In the training and validation accuracy graph, the curves’ slopes approached horizontal
after the 15th epoch, which indicates that data did not make a significant learning process
over the algorithm. At approximately the 19th epoch, the learning curve became almost hori­
zontal. This indicates that no further training was necessary after the 20th epoch, as the accur­
acies did not change considerably and ranged from 83.8% to 85.8%. Lastly, the validation
curve above the training curve indicates that the model was fed a good representation of data,
was ready for testing and was able to provide reliable predictions.
After training and validation, test data were used to evaluate the AI model’s performance in
predicting methane. The model’s performance was analyzed using the actual and predicted
methane content of the testing data as a non-explosive and explosive zone. The results are
given in Table 1.

379
Figure 2. Validation and training accuracy of the algorithm.

Table 1. Confusion matrix.

Explosive Zone Non-Explosive Zone

Explosive Zone 372 68 Sensitivity 84.55%

Non-Explosive 19 451 Specificity 95.96%
Precision Recall
95.14% 86.90% Overall Accuracy: 82.30%

4 DISCUSSION AND CONCLUSION

As was the case with the Upper Big Branch accident, a longwall mine methane explosion can
be fatal. Current explosive gas zone management practices are carried out either with point
sensors or CFD modeling. Whereas explosive gas zone monitoring relies on point sensors, the
critical regions of the longwall face cannot be tracked in real time. Although CFD modeling
can overcome the sensors’ coverage, the required prediction time could be days to weeks,
depending on the resolution of the simulation. Moreover, training the AI with whole mine
data needs intensive training for accurate results.
Leveraging the power of AI might be crucial for monitoring explosive methane concentra­
tion. The primary objective of this study was to combine the advantages of current methane
monitoring practices and reduce their disadvantages. For this purpose, this study utilized
modified LSTM architecture for real-time methane prediction with a small data set and with
the proposed methodology in our previous studies.
Although each mine ventilation condition is unique, previously conducted studies
(KRICKOVIC S and FINDLAY C 1971; Marts 2015; Marts et al. 2014; Schatzel et al. 2012;
Schatzel, Krog, and Dougherty 2011) prove that simulated data (CFD modeling) represents
the actual ventilation that includes but is not limited to “(i) continuous leakage of fresh air
from the face to the gob, and the higher accumulation of methane as the supplied air travels from
the headgate to tailgate side of the face; (ii) higher leakage around the headgate and tailgate
corners of the face due to the high porosity and permeability around the edge of the gob; (iii)
Methane accumulation seems to follow linear regression based on ventilation surveys done in sev­
eral longwall operations.” (Juganda 2020b).Therefore this study utilized the CFD data as val­
idation while calculating accuracies. As a result, the overall accuracy was 82.30% with a small
data set, n=1000, when compared to CFD results. This concludes that the modified LSTM

380
algorithm could combine compelling aspects of CFD modeling (full coverage and representa­
tion) and point sensor measurements (real-time). AI algorithms could achieve 3D coverage of
CFD modeling and real-time point sensor data measurements.,
However, the AI’s prediction capabilities depend on the simulated CFD model results.
Therefore, predictions can only be as accurate as similar longwall face models. Training the
AI model by using different longwall face models can increase the AI model’s capacity. If the
AI model can be trained with more data, it might be used for methane monitoring purposes.
Although the proposed methodology successfully predicted methane concentration throughout
the longwall face, the results only contained numbers representing location, time, and methane
content. Results consisted of lines of numbers that could not be interpreted or used by engineers
and/or miners to determine if the explosive gas accumulation was hazardous.
As future research associated with this study, the authors will train the algorithm by using
other mines’ methane emission models. This is expected to increase the algorithm’s prediction
capabilities and enable its usage and implementation in more longwalls in the world.

ACKNOWLEDGMENTS

This study was supported by the National Institute for Occupational Safety and Health
(NIOSH), grant number 0000HCCR-2019-36403.
This study utilized resources from Colorado School of Mines’ CyberInfrastructure and
Advanced Research Computing (CIARC) team and the University of Colorado Boulder
Research Computing Group, which is supported by the National Science Foundation (awards
ACI-1532235 and ACI-1532236), the University of Colorado, Boulder, and Colorado State
University RMACC Summit.
The authors declare no conflicts of interest. The funders had no role in the design of the
study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or
in the decision to publish the results.

REFERENCES

Alléon, Antoine, Grégoire Jauvion, Boris Quennehen, and David Lissmyr. 2020. “PlumeNet: Large-Scale
Air Quality Forecasting Using A Convolutional LSTM Network.”
Bagnall, Anthony, Jason Lines, Aaron Bostrom, James Large, and Eamonn Keogh. 2017. “The Great
Time Series Classification Bake off: A Review and Experimental Evaluation of Recent Algorithmic
Advances.” Data Mining and Knowledge Discovery 31(3):606–60.
Davis, Scott G., Derek Engel, and Kees van Wingerden. 2015. “Complex Explosion Development in
Mines: Case Study—2010 Upper Big Branch Mine Explosion.” Process Safety Progress 34(3):286–303.
Demirkan, D C, S. Duzgun, A. Juganda, J. Brune, and G. Bogin. 2022. “Evaluation of Time Series Arti­
ficial Intelligence Models for Real-Time/near-Real-Time Methane Prediction in Coal Mines.” CIM
Journal 0(0):1–10.
Demirkan, Doğa Çağdaş. 2022. “NEAR-REAL-TIME/REAL-TIME METHANE PREDICTION
AND VISUALIZATION IN UNDERGROUND LONGWALL COAL MINING by Doga Cagdas
Demirkan.”
Demirkan, Doga Cagdas, H. Sebnem Duzgun, Aditya Juganda, Jurgen Brune, and Gregory Bogin. 2022.
“Real-Time Methane Prediction in Underground Longwall Coal Mining Using AI.” Energies 15
(17):1–12.
Dempster, Angus, François Petitjean, and Geoffrey I. Webb. 2020. “ROCKET: Exceptionally Fast and
Accurate Time Series Classification Using Random Convolutional Kernels.” Data Mining and Know­
ledge Discovery 34(5):1454–95.
Deng, Houtao, George Runger, Eugene Tuv, and Martyanov Vladimir. 2013. “A Time Series Forest for
Classification and Feature Extraction.” Information Sciences 239:142–53.
Dong, Mei, Hongyu Wu, Hui Hu, Rafig Azzam, Liang Zhang, Zengrong Zheng, and Xiaonan Gong.
2021. “Deformation Prediction of Unstable Slopes Based on Real-Time Monitoring and Deepar
Model.” Sensors (Switzerland) 21(1):1–18.
Dougherty, Heather N. and C. Özgen Karacan. 2011. “A New Methane Control and Prediction Software
Suite for Longwall Mines.” Computers and Geosciences 37(9):1490–1500.
Dubbs, Alexander. 2021. “Test Set Sizing Via Random Matrix Theory.” (December).

381
Duda, Adam and Alicja Krzemień. 2018. “Forecast of Methane Emission from Closed Underground
Coal Mines Exploited by Longwall Mining – A Case Study of Anna Coal Mine.” Journal of Sustain­
able Mining 17(4):184–94.
Dudley, Bob and others. 2018. “BP Statistical Review of World Energy.” BP Statistical Review, London,
UK, Accessed Aug 6(2018):116.
Gholami, V., J. Torkaman, and P. Dalir. 2019. “Simulation of Precipitation Time Series Using
Tree-Rings, Earlywood Vessel Features, and Artificial Neural Network.” Theoretical and Applied Cli­
matology 137(3–4):1939–48.
Gholamy, Afshin, Vladik Kreinovich, and Olga Kosheleva. 2018. “Why 70/30 or 80/20 Relation Between
Training and Testing Sets : A Pedagogical Explanation.” Departmental Technical Reports (CS) 1–6.
He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. “Deep Residual Learning for Image
Recognition.” Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern
Recognition 2016-Decem:770–78.
Ismail Fawaz, Hassan, Benjamin Lucas, Germain Forestier, Charlotte Pelletier, Daniel F. Schmidt,
Jonathan Weber, Geoffrey I. Webb, Lhassane Idoumghar, Pierre Alain Muller, and
François Petitjean. 2020. “InceptionTime: Finding AlexNet for Time Series Classification.” Data
Mining and Knowledge Discovery 34(6):1936–62.
Joseph, V. Roshan. 2022. “Optimal Ratio for Data Splitting.” Statistical Analysis and Data Mining
(February):531–38.
Juganda, Aditya. 2020a. “Evaluation of Point-Based Methane Monitoring and Proximity Detection for
Methane Explosive Zones in Longwall Faces of Underground Coal Mines.” Colorado School of Mines.
Juganda, Aditya. 2020b. “Evaluation of Point-Based Methane Monitoring and Proximity Detection for
Methane Explosive Zones in Longwall Faces of Underground Coal Mines.” 255.
Juganda, Aditya, Claire Strebinger, Jürgen F. Brune, and Gregory E. Bogin. 2020. “Discrete Modeling
of a Longwall Coal Mine Gob for CFD Simulation.” International Journal of Mining Science and
Technology 30(4):463–69.
Karacan, C. Özgen. 2008. “Modeling and Prediction of Ventilation Methane Emissions of U.S. Longwall
Mines Using Supervised Artificial Neural Networks.” International Journal of Coal Geology 73(3–
4):371–87.
Krickovic S and Findlay C. 1971. “Methane Emission Rate Studies in a Central Pennsylvania Mine.”
U S Bur Mines, Rep Invest 7591.
Li, Kan and Jose C. Principe. 2017. “Transfer Learning in Adaptive Filters: The Nearest Instance
Centroid-Estimation Kernel Least-Mean-Square Algorithm.” IEEE Transactions on Signal Processing
65(24):6520–35.
Li, Lei. 2011. “Fast Algorithms for Mining Co-Evolving Time Series.” (September):217.
Lines, Jason, Sarah Taylor, and Anthony Bagnall. 2018. “Time Series Classification with HIVE-COTE:
The Hierarchical Vote Collective of Transformation-Based Ensembles.” ACM Transactions on Know­
ledge Discovery from Data 12(5).
Lucas, Benjamin, Ahmed Shifaz, Charlotte Pelletier, Lachlan O’Neill, Nayyar Zaidi, Bart Goethals,
François Petitjean, and Geoffrey I. Webb. 2019. “Proximity Forest: An Effective and Scalable
Distance-Based Classifier for Time Series.” Data Mining and Knowledge Discovery 33(3):607–35.
Marts, J. A. 2015. “Nitrogen Injection in Progressively Sealed Longwall Gobs and the FORMATION
OF A COMPLETE AND DYNAMIC SEAL.”
Marts, J. A., R. C. Gilmore, J. F. Brune, G. E. Bogin, J. W. Grubb, and S. Saki. 2014. “Dynamic Gob
Response and Reservoir Properties for Active Longwall Coal Mines.” Mining Engineering 66(12):41–48.
Nyanteh, Yaw, Chris Edrington, Sanjeev Srivastava, and David Cartes. 2013. “Application of Artificial
Intelligence to Real-Time Fault Detection in Permanent-Magnet Synchronous Machines.” IEEE
Transactions on Industry Applications 49(3):1205–14.
Owusu, Obed, Kenneth Joseph Bansah, and Albert Kobina Mensah. 2019. “‘Small in Size, but Big in
Impact’: Socio-Environmental Reforms for Sustainable Artisanal and Small-Scale Mining.” Journal of
Sustainable Mining 18(1):38–44.
Peng, Syd S. 2019. Longwall Mining. CRC Press.
Qiao, Ming, Ting Ren, Jon Roberts, Xiaohan Yang, Zhongbei Li, and Jianming Wu. 2022. “New Insight
into Proactive Goaf Inertisation for Spontaneous Combustion Management and Control.” Process
Safety and Environmental Protection 161:739–57.
Rodríguez-Rangel, Héctor, Luis Alberto Morales-Rosales, Rafael Imperial-Rojo, Mario
Alberto Roman-Garay, Gloria Ekaterine Peralta-Peñuñuri, and Mariana Lobato-Báez. 2022. “Ana­
lysis of Statistical and Artificial Intelligence Algorithms for Real-Time Speed Estimation Based on
Vehicle Detection with YOLO.” Applied Sciences (Switzerland) 12(6):1–20.

382
Schäfer, Patrick and Ulf Leser. 2017. “Fast and Accurate Time Series Classification with WEASEL.” Pp.
637–646 in Proceedings of the 2017 ACM on Conference on Information and Knowledge Management,
CIKM ’17. New York, NY, USA: Association for Computing Machinery.
Schatzel, S. J., R. B. Krog, and H. Dougherty. 2011. “A Field Study of Us Longwall Coal Mine Ventila­
tion and Bleeder Performance.” SME Annual Meeting and Exhibit and CMA 113th National Western
Mining Conference 2011 65–70.
Schatzel, Steven J., C. Özgen Karacan, Heather Dougherty, and Gerrit V. R. Goodman. 2012. “An Ana­
lysis of Reservoir Conditions and Responses in Longwall Panel Overburden during Mining and Its
Effect on Gob Gas Well Performance.” Engineering Geology 127:65–74.
Shifaz, Ahmed, Charlotte Pelletier, François Petitjean, and Geoffrey I. Webb. 2020. TS-CHIEF:
A Scalable and Accurate Forest Algorithm for Time Series Classification. Vol. 34. Springer US.
Sidorenko, Andrey Alexandrovich, Pavel Nikolaevich Dmitriev, and Yury Georgiyevich Sirenko. 2021.
“Predicting Methane Emissions from Multiple Gas-Bearing Coal Seams to Longwall Goafs at Russian
Mines.” ARPN Journal of Engineering and Applied Sciences 16(8):851–57.
Wahyono, Agus Harjoko, Andi Dharmawan, Faisal Dharma Adhinata, Gamma Kosala, and Kang
Hyun Jo. 2022. “Real-Time Forest Fire Detection Framework Based on Artificial Intelligence Using
Color Probability Model and Motion Feature Analysis.” Fire 5(1).

383
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Evaluation of different suppression techniques for lithium-ion

battery fires

Liming Yuan, Wei Tang, R.A. Thomas & John Soles

Pittsburgh Mining Research Division, National Institute for Occupational Safety and Health (NIOSH),
Pittsburgh, PA, USA

ABSTRACT: Lithium-ion (Li-ion) batteries are finding more use as power sources in the
mining industry. However, they are known to pose significant fire and explosion hazards.
When a Li-ion battery is exposed to excessive operating conditions, its internal temperature
may exceed a normal operating range, allowing the active component materials to decompose
or react with each other, eventually leading to thermal runaway. A Li-ion battery contains
certain oxidizing agents making suppression of a battery fire very challenging. A series of Li-
ion battery fire suppression tests were conducted by researchers at the National Institute for
Occupational Safety and Health (NIOSH) to evaluate the effectiveness of different fire sup­
pression test systems including dry chemical, water spray/mist, and Class D extinguisher
powder. The batteries tested are commercial nickel manganese cobalt (NMC) and lithium iron
phosphate (LFP) battery packs. The results indicated that dry chemical and Class D powder
could extinguish the fire temporarily, but a reignition occurred. Water mist was able to extin­
guish the battery fire completely with continuous cooling of the battery to prevent the reigni­
tion. The suppression results for both NMC and LFP chemistries were also compared. These
test results can be used to develop appropriate firefighting strategies for safe and effective sup­
pression of battery fires in a mine.

1 INTRODUCTION

Li-ion batteries are known to pose significant fire and explosion hazards as the use of these
batteries has become more widespread, ranging from consumer electronics, battery electric
vehicles, to energy storage systems. The mining industry has also seen increased utilization of
Li-ion batteries for powering equipment in both surface and underground mining operations.
The safety issues related to Li-ion batteries pose a great challenge to their wider application of
Li-ion batteries in large mining equipment which may consist of thousands of battery cells.
Several battery fire incidents have occurred in underground mines in the early stages of bat­
tery electric vehicle deployment (Gillet, 2021). A battery fire is usually caused by thermal run­
away that occurs when heat generated from exothermic reactions inside a battery outpaces
heat dissipated from the battery. If unmitigated, thermal runaway will proceed to cell rupture
and the venting of hot, toxic, and highly flammable gases, leading to a possible fire or explo­
sion. These potential fire hazards need to be assessed and techniques developed to effectively
suppress Li-ion battery fires.
Many laboratory-scale experiments have been conducted on the battery fire suppression
techniques (Xu et al., 2021; Zhang et al., 2021; Cui et al., 2022; Said et al., 2022). Li et al.
(2015) studied the fire-extinguishing efficiency of ABC powder, carbon dioxide, aqueous film-
form foam (AFFF), and water mist on the 18650-type LiCoO2 Li-ion battery pack. The ABC
powder, carbon dioxide, and 3% aqueous film-form foam were able to extinguish the open
flames of the Li-ion battery pack, but the reignition could not be avoided. The water mist

DOI: 10.1201/9781003429241-40

384
system could not suppress the battery fire in the experiments because of low water flow rate.
The laboratory experimental results of Xu et al. (2020) indicate that water mist/spray has
a better suppression effect than CO2 and HFC-227ea. The Federal Aviation Agency (FAA)
conducted experiments to screen fire-extinguishing agents for battery fire involving consumer
electronic products powered by Li-ion batteries (Maloney, 2014). The experimental results
demonstrated that water-based extinguishing agents were the most effective, while nonaqu­
eous extinguishing agents were the least effective. Russo et al. (2018) conducted experiments
to evaluate the effectiveness of different fire extinguishing agents for single Li-ion cells and
a battery pack. Out of the five fire-extinguishing agents including water, foam, dry powder,
carbon dioxide, and water mist, water and foam were the most effective by rapidly reducing
the temperature of the cell and extinguishing the fire.
There are few studies on the suppression of large Li-ion battery pack fires. Egelhaaf et al.
(2014) carried out firefighting tests of Li-ion traction batteries, and the test results showed
that the amount of water required to extinguish such a fire is a lot larger amount than that
used for firefighting of conventionally driven vehicles. In National Transportation Safety
Board (NTSB) investigations of battery fires in electric vehicles, the total amount of water
used to suppress high-voltage Li-ion battery fires ranged from 300 to 20,000 gallons (NTSB,
2020). These studies indicated that water can be an effective suppression agent for fighting
battery fires if a sufficient amount of water is available. For battery electric vehicles used in
underground mines, water may not be readily available, or the amount of water may be
limited. Therefore, research is needed to investigate the characteristics of different Li-ion bat­
tery fires in mines and effective fire suppression techniques. In this study, fire suppression
experiments were conducted for both NMC and LFP battery packs using water mist, dry
chemical, and Class D extinguishing powder, and the suppression results for these two battery
chemistries were compared. Experiments were also conducted to study the effect of battery
size on the suppression of NMC battery pack fires. The experimental results from this study
may help develop appropriate firefighting strategies for safe and effective suppression of Li-
ion battery fires in underground mines.

2 EXPERIMENTAL

Li-ion battery fire suppression experiments were conducted in a container with dimensions of
12-m long, 2.4-m wide, and 2.85-m high (40 ft × 8 ft × 9.5 ft) and using a data acquisition and
test observation trailer located nearby the container. A fan is installed at one end of the con­
tainer and the opposing end is open as shown in Figure 1. The ventilation velocity was set at
0.2 m/s (40 fpm) for all tests. The container is equipped with three fire suppression systems:
dry chemical, water spray/mist, and Class D powder. The fire suppressant discharge nozzle
direction is located above the battery. The suppression system is manually activated from out­
side the container. Thermocouple trees and gas sampling tubes were installed at the exit of the
container to measure exit gas temperatures and collect gas samples. Two cameras were
installed inside the container to monitor testing.
The batteries used for the suppression tests were a 12V 30Ah NMC battery pack consisting
of 36 cylindrical cells and a 12V 35Ah LFP battery pack consisting of 24 cylindrical cells.
Before each test, the battery pack was charged to the 100% state. During the tests, the battery
pack was placed onto two heater strips for heating up with each of two 750-Watt heaters. In
addition, 12 to 18 K-type thermocouples were attached to the surfaces of the battery cells of
the battery pack to measure the battery temperature as shown in Figure 2. During the tests,
when the battery reached thermal runaway, the heaters were turned off, and then, after
a stable flame was established, the fire suppression system was activated. The suppression
nozzle was placed 1.5 m above the battery pack. For the dry chemical and Class D powder
suppression systems, the release time was between 30 and 50 seconds. For the water mist
system, the flow rate was 3.6 gallon per minute and water was on for 2 to 5 minutes. Battery
surface temperatures were measured by the data acquisition system. Air exiting the container
was sampled using an infrared gas analyzer to determine concentrations of carbon monoxide

385
Figure 1. Battery fire suppression test setup.

Figure 2. Battery packs and attached thermocouples. NMC (left), LFP (right).

(CO) and carbon dioxide (CO2). Cameras were used to observe and record the fire and sup­
pression behaviors. Observations were made on a monitor in the trailer to determine the times
for the first appearance of smoke and flame and the final extinguishment of the fire if there
was one.

3 RESULTS AND ANALYSIS

3.1 Free burn tests

Free burn tests were conducted to understand the fire characteristics for both 12V NMC and
LFP battery packs. During the free burn, the battery pack was heated the same way as for the
suppression tests, and the same ventilation airflow was used. As no fire suppressant was
applied, the battery was left to burn all the way to self-extinguishment. For the NMC battery
pack, flames appeared after the battery smoked for more than 6 minutes, and then the heaters
were turned off. The first cell explosion occurred about 1.5 minutes after the first flame. Mul­
tiple cells exploded during the free burn, and the battery pack was disintegrated. The total
burning time was 4.5 minutes with constantly violent explosions. For the LFP battery pack,
flames appeared after the battery smoked for 5.5 minutes, then the heaters were turned off.

386
Compared to the NMC battery pack, significantly heavier smoke was generated before the
first flame. The first cell explosion also occurred about 1.5 minutes after the first flame. How­
ever, very few cells exploded, and the battery pack was not disintegrated. The total burning
time was over 19 minutes. The battery did not burn as violently as the NMC battery. Instead,
the LFP battery only burned violently for 1-2 minutes, and then maintained a local small
flame for 4-5 minutes, followed by another violent but short burn Figure 3 (a) shows the vio­
lent burn with cell explosion for the NMC battery pack, and Figure 3 (b) shows the flaming of
the LFP battery.
The apparently different combustion behaviors of NMC and LFP batteries can be further
examined in Figure 4. Figure 4 (a) shows the exit gas temperature increase above the ambient
temperature, while Figure 4 (b) shows the measured CO2 concentrations at the exit for the
two batteries. The NMC battery had a slightly higher maximum gas temperature increase, but
a lower maximum CO2 concentration than the LFP battery. As observed, the NMC battery
burned violently lasting for only 4.5 minutes. This can be seen clearly in the gas temperature
increase and CO2 concentration data. For the LFP battery, the gas temperature increase and
CO2 concentration data show clearly that there were a few peaks after the first peak indicating
the existence of small flames and burning for more than 19 minutes. This is consistent with the
observation on the camera monitor.

Figure 3. 12V battery pack free burn (a) NMC (left) and (b) LFP (right).

Figure 4. (a) Comparison of exit gas temperature increase (left) and (b) CO2 concentrations (right).

3.2 NMC and LFP battery pack fire suppression using dry chemical and Class D powder
Dry chemical is a commonly used fire suppressant for mine equipment fires. In this study,
a dry chemical suppression system was used to suppress fires from both 12V NMC and LFP
battery packs under the same experimental conditions. For the NMC battery, the dry chem­
ical suppression system was activated after the first cell explosion occurred, and the

387
suppressant discharge lasted for 10 seconds. The flame was extinguished by the suppressant.
Figure 5 (a) shows the burning of the NMC battery before the discharge of the suppressant,
and Figure 5 (b) shows the NMC battery pack after the flame was extinguished. The battery
pack was dislocated a little and broke into two sections by the cell explosion. After 2.5 min­
utes, the reignition occurred and the NMC battery exploded again as shown in Figure 5 (c).
Eventually, the burning slowed down as shown in Figure 5 (d). The reignition lasted for about
8 minutes. Class D extinguishing powder was also tested for suppression of the battery pack
fires. Similar to the dry chemical, the Class D powder was able to quench the flame quickly
for the NMC battery pack; however, the flame came back in a few seconds after the discharge
of the suppressant. The NMC battery continued to burn for about 5 minutes to self-
extinguishment.
Figure 6 shows the LFP battery fire and suppression process with the dry chemical suppres­
sion system. Figure 6 (a) shows the flame before the discharge of the dry chemical. Figure 6
(b) shows the battery after the flame was extinguished by the suppressant with everything
covered by the white chemical. After roughly 2 minutes, reignition occurred as shown in
Figure 6 (c) because the dry chemical was not able to cool the LFP battery to stop the exother­
mic internal reactions. The LFP battery continued to burn all the way to self-extinguishment
lasting for more than 10 minutes as shown in Figure 6 (d). During the reignition period, there
was a cell explosion, and the flame went out for about half a minute and came back again.
For the LFP battery pack, the Class D powder was not able to extinguish the flame, and the
battery continued to burn for 10 minutes.

Figure 5. NMC battery fire suppression using dry chemical: (a) flame before suppression; (b) after sup­
pressant discharge; (c) explosion following reignition; (d) flame near burnout.

3.3 NMC and LFP battery pack fire suppression using water mist
A water mist system was applied to suppress the battery fires from both the NMC and LFP
battery packs. For the NMC battery fire, water mist was turned on one minute after the first
flame, and the fire was extinguished in 65 seconds. There was one cell explosion during the
suppression. Figure 7 (a) shows the burning of the NMC battery before water suppression

388
Figure 6. LFP battery fire suppression using dry chemical: (a) flame before suppression; (b) after sup­
pressant discharge; (c) flame after reignition; (d) flame near burnout.

was turned on, while Figure 7 (b) shows the NMC battery after the fire was extinguished.
Water was on for 2 minutes and there was no reignition. For the LFP battery fire, the water
mist extinguished the flame in 43 seconds, but the flame occurred again in half a minute. As
water was still on, the fire was extinguished in 14 seconds. The total extinguishing time was 86
seconds, and there was no reignition. Figure 8 (a) shows the burning of the LFP battery
before the release of water mist, and Figure 8 (b) shows the LFP battery pack after the fire
was extinguished. The results indicate that water mist was effective in extinguishing both the
NMC and LFP battery pack fires as water could cool the battery continually to prevent the
reignition of the battery.

Figure 7. NMC battery fire suppression with water mist; (a) flame before suppression (left) and (b)
flame out after water mist (right).

389
Figure 8. LFP battery fire suppression with water mist; (a) flame before suppression (left) and (b) flame
out after water mist (right).

3.4 Effect of battery size on suppression of NMC battery pack fires

The effectiveness of a fire suppression system for a Li-ion battery can be affected significantly
by the battery size. In this study, fire suppression tests were also conducted for a larger NMC
battery pack. This larger NMC battery pack was 24V 40Ah consisting of 108 cylindrical cells
in three rows. The test conditions and procedures were the same as for the 12V battery pack.
With the water mist system, it took 172 seconds to completely extinguish the fire. During this
process, the battery burnt violently with over 20 cell explosions. There was no reignition as
water was on for 6 minutes to cool the battery. Compared to the 52-second extinguishing time
for the 12V NMC battery pack, the larger size pack needed significantly longer extinguishing
time with the water mist system.
With the dry chemical system, the flame was extinguished after the discharge of the suppres­
sant. However, reignition occurred after 15 minutes, and the battery continued to burn for
over 5 minutes. Compared to the 2.5 minutes before the reignition occurred for the 12V NMC
battery pack with the dry chemical system, it took a significantly longer time for the reignition
to occur for the larger NMC pack. With the Class D powder system, the flame was not com­
pletely extinguished, and a small flame persisted after the discharge of the powder. The battery
continued to burn violently with multiple cell explosions for 12 minutes. For the 12V NMC
battery pack fire suppression with the Class D powder system, the reignition occurred in a few
seconds. These results demonstrate that Class D powder is ineffective for extinguishing the
NMC battery pack fires.

Figure 9. Comparison of exit gas temperature increases between different NMC battery pack sizes
using (a) dry chemical (left) and (b) Class D powder (right).

390
Figure 10. Scattered NMC cells after flame was out with (a) dry chemical system (left) and (b) Class
D powder system (right).

Figure 9 (a) and Figure (b) show exit gas temperature increases for both the 12V and 24V
NMC battery packs during the fire suppression process using the dry chemical and Class
D powder system, respectively. It can be seen that the dry chemical was able to extinguish the
flame immediately after the discharge of the suppressant. But reignition occurred after 15 min­
utes as the dry chemical was not able to cool the battery. Dry chemical smothered the battery
completely and stopped the radiative heat transfer from the flame to the battery as the flame
was extinguished. The chemical reactions inside the battery were still ongoing, but there was
no external heat transfer any longer. This may be the major reason that it took 15 minutes for
the reignition to occur, and the battery burned very violently with many cell explosions. Using
the dry chemical suppressant, 24V NMC battery pack was completely disintegrated, and cells
were scattered as shown in Figure 10 (a). The maximum exit gas temperature increase was
about 38° C for the 24V NMC battery pack fire while using the dry chemical suppressant,
Figure 9 (a). On the other hand, the Class D powder was not able to extinguish the flame
completely, but just reduced it to a small flame. The small flame continually radiated heat to
the adjacent battery cells. The chemical reactions inside those cells were accelerated by the
24V NMC battery pack flame heat, and it took 70 seconds for those cells to explode produ­
cing a peak exit gas temperature increase of 20° C, Figure 9 (b). About 5 minutes later, there
was another round of cell explosions with another peak exit gas temperature increase of 21°
C. As shown in Figure 10, more of the 24V NMC battery pack cells exploded with the dry
chemical system than with the Class D powder system.

4 CONCLUSIONS

Experiments were conducted to study the fire behaviors of 12V NMC and LFP Li-ion battery
packs. Under the same conditions, the NMC battery pack burned more violently with more
cell explosions than the LFP pack, while the LFP pack burned for a longer time and generated
more CO2. Fire suppression experiments were then conducted for both battery chemistries
using dry chemical, Class D powder, and water mist systems. The experimental results demon­
strate that water was more effective in extinguishing the battery pack fires than dry chemical
and Class D powder because water was able to cool the battery continually. With dry chemical
and Class D powder, the flame was extinguished temporarily, but reignition occurred after 2
to 2.5 minutes with the dry chemical, and a few seconds with the Class D powder.
The effect of battery size on the suppression of NMC battery pack fires was investigated
through suppression tests using a larger 24V battery pack. The experimental results show that
even though water was still able to extinguish the larger 24V battery pack fire, the extinguishing
time increased significantly as compared to the smaller 12V battery pack, from 65 to 172 seconds
indicating a large amount of water is needed to extinguish a large format Li-ion battery fire. With
the larger 24V battery pack and the dry chemical, the flame was also extinguished temporarily,
but it took a much longer time for the reignition to occur as compared to the smaller 12V battery
pack, from 2.5 to 15 minutes, indicating the larger the battery size, the longer time for the

391
reignition to occur. With the Class D powder system, the flame was not completely extinguished,
and the battery continued to burn, indicating that the Class D powder is least effective for sup­
pressing the Li-ion battery fires. The experimental results also demonstrate that the longer time
before the reignition results in a more violent explosion, as more cells become involved.

5 DISCLAIMERS

The findings and conclusions in this report are those of the authors and do not necessarily
represent the official position of the National Institute for Occupational Safety and Health,
Centers for Disease Control and Prevention. Mention of any company or product does not
constitute endorsement by NIOSH.

REFERENCES

Cui, Y., Liu, J., Han, X., Sun, S., Cong, B. (2022). Full-scale experimental study on suppressing
lithium-ion battery pack fires from electric vehicles, Fire Safety Journal, 129, 103562.
Egelhaaf, M., Wolpert, D., Lange, T. Fire Fighting of battery electric vehicle fires. Proceedings of the
Third International Conference on Fire in Vehicles, October 1-2, 2014, Berlin, Germany. pp. 221–230.
Gillett, S., 2021. Battery Electric Vehicle Emergency Response Incident Review and Best Practices.
Workplace Safety North Virtual Symposium: Battery Electric Vehicle Safety in Mines, January 20,
2021.
Li, Y., Yu, D., Zhang, S., Hu, Q., Liu, X., Wang, J. (2015) On the fire extinguishing tests of typical lith­
ium ion battery. Journal of Safety Environment, 15:120–125.
Maloney, T. (2014). Extinguishment of lithium-ion and lithium-metal battery fires, US Department of
Transportation; Federal Aviation Administration: Washington, DC, pp.46–51.
NTSB, 2020. Safety risks to emergency responders from lithium-ion battery fires in electric vehicles,
Safety Report NTSB/SR–20/01.
Russo, P., Di Bari, C., Mazzaro, M., De Rosa, A., Morrielo, I. (2018). Effective fire extinguishing sys­
tems for lithium-ion battery. Chemical Engineering Transactions, 67, 727–732.
Said, A.O., Garber, A., Peng, Y., Stoliarov, S.I. (2022) Experimental investigation of suppression of
18650 lithium ion cell array fires with water mist, Fire Technology, 58, 523–551.
Xu, J., Guo, P., Duan, Q., Yu, X., Zhang, L., Liu, Y., Wang, Q. (2020). Experimental study of the effect­
iveness of three kinds of extinguishing agents on suppressing lithium-ion battery fires. Applied Ther­
mal Engineering, 171, 115076.
Xu, J., Duan, Q., Zhang, L., Liu, Y., Zhao, C., Wang, Q. (2021). Experimental study of the cooling
effect of water mist on 18650 lithium-ion battery at different initial temperatures, Process Safety and
Environmental Protection, 157, 156–166.
Zhang, L., Duan, Q., Liu, Y., Xu, J., Sun, J., Xiao, H., Wang, Q. (2021). Experimental investigation of
water spray on suppressing lithium-ion battery fires, Fire Safety Journal, 120, 103117.

392
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Characterization and preliminary assessment of diesel fire prior to

setting up large size battery fire experiment

Robert Ilango Pushparaj, Guang Xu*, Amir Iqbal & O.B. Salami
Department of Mining and Explosives Engineering, Missouri University of Science and Technology, Rolla,
MO, USA

ABSTRACT: Any fire could quickly destroy human lives including casualties. Fire risk
exposure by diesel-powered vehicles in confined area is critical. Also, turning to battery
powered vehicles, fire risk is worse. Li-ion battery fire produces intense heat, smoke, and com­
plex toxic gasses. Among them, toxic gases constitute a significant threat to human health.
Therefore, it is necessary to study the detailed identification of gas emissions from the battery
fire. Previous studies reported the emission of toxic gases after firing the different types of bat­
teries, still unknown gases are not studied or quantified. Remarkably, real-time gas analysis is
not explored very well which motivates us to conduct this work. In this primary study, we
conducted a large-scale diesel fire test at the predesigned fire station. The fire was recorded
using the IR radiation from the thermal camera. The total fire lasted about 20 min with
a maximum fire temperature of 601ºC. Thermocouple 1 (T1) measured the maximum
temperature of 39ºC which is comparatively lower than Thermocouple 2 (T2) due to the con­
siderable distance between them about 2 ft. This primary fire characteristics investigation will
provide a new way to conduct large scale battery fires and a detailed understanding of battery
fire and associated health risks.

1 INTRODUCTION

Rechargeable Li-ion batteries (LIB) play a vital role in the energy storage system due to their
high gravimetric and volumetric energy density (Scrosati and Garche, 2010). Besides, electrical
vehicles are emerging in the mining industry to substitute the current extensively used mobile
equipment operated primarily by diesel internal combustion engines owing to its low energy
consumption and very less heat emission (Stampatori et al., 2020). Furthermore, the mainten­
ance and ventilation infrastructure cost are lower than diesel powered vehicles(Skidmore,
2021). As an example, Newmont is a famous gold mine industry have agreement with caterpil­
lar to achieve zero emission by 2027 through investing primarily $100 billion on battery elec­
trification of mining vehicles (Release, 2021).
Nevertheless, using Li-ion battery is not safe due the thermal insecurity that leads to Li-ion
battery fire and explosion (Sun et al., 2020). Generally, the battery fire releases hazardous
gases and heat that consequently impact the potential health risks. For example, ‘B747ʹ cargo
flights crashed and reported two pilots death in 2011 due to the LIB fire (Brett, n.d.). Also,
the Occupational Safety and Health Administration report says twenty-five thousands of Li-
ion battery fire incidents happened over 5 years (Yuan et al., 2020). Therefore, understanding
the risks of BEV fire enable us to prepare for preventing BEV incidents that enables safe mas­
sive use of LIB.
State of charge (SOC) may impact the battery fire characteristics such as heat, flame, and
gas emissions. Recent studies focused thermal and fire behaviors of 38 Ah and 243 Ah LIBs
with different SOC and analyzed the relationship of thermal runaway and gas generations.

DOI: 10.1201/9781003429241-41

393
For example, M. Chen et al, conducted experimental Li-ion fire study with primary batteries
first and later with modern battery like 18650 battery type and reported that the combustion
efficiency and yield of CO2 varied from single to bundle primary battery fire. Further, solid
electrolyte interface layer breakdown and the separator drop-off are the main key reasons for
the Li-ion battery fire (Chen et al., 2016)(Chen et al., 2015). Afterwards, V. Somandepalli
et al, examined quantification of combustion hazards, but they only suggested the gas com­
position highest over pressure and flammability that helps to design vehicles (Somandepalli
et al., 2014). Also, P. Huang et al, studied combustion behavior using large scale lithium titan­
ium battery, revealed risk increases according to SOC (Huang et al., 2015). Very recently,
F. Larson et al, investigated the fluoride gas emission with different types of batteries and this
investigation is limited to small or very few batteries gas emission observations (Larsson et al.,
2017). Though several reports identified the toxic gases emissions and quantification, still
other hazardous gasses not yet clearly studied or quantified. Moreover, previous studies are
tested with very few batteries (less than 100 g) and estimated the thermal runway and predic­
tion of fire suppression that is not enough for the understanding of Li-ion battery fire safety,
particularly underground mine battery vehicles.
In the past, diesel pool fire has been studied and their impact has been examined. For
example, L.Yuan et al (Yuan et al., 2018) studied diesel fire and reported the sensor response
time to early detect the fire risk. Later, R. Hansen (Hansen, 2019) researched a number of
design fire situations which were created from risk evaluations. Also, a deterministic approach
was executed and where the findings from previous full-scale fire experiments in underground
mines. Very recently, G.V Kuznetsov et al (Kuznetsov et al., 2022) classified fire and analyzed
fire characteristics, gas analysis and finally recommended fire prevention systems.
The objective of this paper is to determine the diesel pool fire characteristics before setting
up a large battery fire station to identify and quantify toxic gases emitted in LIB fires and
evaluate their risk in underground mines. This diesel pool fire characteristic includes fire sta­
tion setup, positioning the fire pool, measuring the surface temperature, and flame character­
istics. This preliminary work will help to improve the battery fire station and control external
factors such as air direction and velocity before battery fire risk analysis.

2 EXPERIMENTAL SET UP

Figure 1(a) describes the simple schematic representation of a full-scale diesel fire test station.
A few significant devices that have been used to test such as a thermal camera, thermocouple,
and exhausting fan. In brief, a pan container (12-inch diameter) and a 0.5-gal diesel pool were
used in this experiment. The thermal camera was placed 10 ft away from the pool fire. We
used two thermocouples (T1, T2), and they measured the instantaneous temperature at two

Figure 1. (a) Schematic representation of diesel fire test (b)Photograph of actual fire station.

394
different locations with a distance of 2 ft. The approximate length of thermocouple from the
center of the pool fire is 5 ft. The fire was ignited using an external source, and the fire was
slowly started and stabilized over 5 min, and total fire lasted about 20 min. At the end of the
duct the fan was connected, and it was able to pull some amount of smoke out of the gas sta­
tion indicating that this fan could pull enough smoke. This smoke will be transferred the gas
cell for the further toxic and toxic gas evaluation using the FITR shown in the Figure 1(b).
The same procedure will be used to conduct the large-scale battery fire test and gas evaluation
which will be discussed in our future research.

3 RESULTS AND DISCUSSION

Figure 2 illustrates the combustion behavior of the diesel fire corresponding to the various time­
stamps. Noted that, the diesel fire was not stable due the uncontrolled air direction and velocity.
According to the burning characterization, the fire was classified in to three types. (i) ignition,
(ii) stable combustion and (iii) abatement (Liu et al., 2021). The fire was visible after 2 minutes
of ignition. The fire intensity reached to its maximum value at the reported time. Therefore,
t = 2.0 min was chosen as the representative time for stage I of combustion. The images depict­
ing the 3.0 min somewhat higher temperature. It can be seen that stage I has low flame intensity
with a temperature of 199 ºC after 2 min which further burned and reached to stage II represent­
ing stable combustion. The stage II was lasted about 10 min with a high intensity of fire and
read the maximum temperature of 601ºC. Finally, the fire reached to the stage III which called
as abatement where the fire last 2 min and temperature was recorded about 150ºC indicating
the burning was slow down. In comparison with each stage, the stage II has major vertical fire
than others. However, the fire was going all the way to all the direction due to the air flow direc­
tion. This experiment was set up in a place where too many factors were associated.

Figure 2. Combustion behavior of diesel fire.

From this research finding, we can understand the fire temperature and direction. This
research will be extended at the place where you can control the external factors to determine
the full fire characteristics. The thermal camera was used to record the video and then classify
the burning behaviors of the fire. The thermal camera uses IR radiation to measure the fire

395
temperature. IR camera visualizes the surface’s temperature gradient by converting changes in
electrical resistance made by infrared radiation into a thermogram (Spieß et al., 2021). In
general, thermal camera will work based on the Stefan-Boltzmann law. This law states that
the total energy radiated per unit surface area of a black body across all the wavelengths per
unit time is s directly proportional to the fourth power of its absolute temperature. Previously,
similar studies were conducted to analyze the 2.9Ah powered Li-ion battery burning behaviors
(Ping et al., 2015). Therefore, this case study will help to further research on large scale battery
fire classification and gas quantification.
Surface temperature is critical to trigger the battery thermal runaway. Hence, it should be
measured using an appropriate tool. Thermo couple is powerful tool that has been used to
detect the battery surface temperature during the thermal runaway process (Chen et al., 2019).
Primarily, we calculated the diesel fire the axial temperature using thermocouple. Figure 3 rep­
resents the temperature vs time profile of the diesel fire. T1 and T2 representing the two differ­
ent locations. Position T1 shows clearly that the maximum temperature reached to 39ºC while
T2 shows only 23ºC. The temperature difference is due to the location and no uniform air
inlet and external turbulent. Due to external factors such as air direction, air velocity, climate
conditions, the temperature may not be reproduced at open conditions. In our future study,
we will use this temperature logger to determine the maximum surface temperature of the bat­
tery instead of diesel fire. Also, increasing the thermocouples at different locations around the
fire may ensure temperature detection accuracy. In the past, many researchers have been used
1-16 thermocouple to determine the axial temperature from various fire sets up (Xu and Lei,
2022). Hence, using many thermocouples, we can predict the average temperature ranges
around the battery fire. We extend the use of thermocouple on the battery surface to know the
surface temperature to trigger the thermal runaway. This case study will provide the risk
evaluation during thermal management design.

Figure 3. Temperature vs Time profile of diesel fire.

4 FUTURE WORKS

To execute the battery fire research, we will use the diesel fire test station with high-accuracy
analytical instruments, as shown in Figure 4. The dimension of the fire station is 20 x 10 x 10
(L x H x W in ft). The fire station is covered with a metal roof to avoid fire, or any explosion.
Primarily, the fire station is closed on three sides. A gap of 1 ft from the ground was left to
allow air venting into the chamber. The smoke duct dimension is 8.0 cm in diameter and
1.0 m in length in vertically. It will be connected horizontally with a 3.0 cm duct, and the total

396
length will be 5 ft from the center. The duct end will be connected to the exhaust fan (AC
Infinity, model AI-CLA4) to pull the fumes.

Figure 4. Schematic diagram for the future battery fire test.

To collect and transfer the gas, we will use a vacuum pump (Dia-Vac, model R222-At-
AA1-C), which will be connected to the outlet of the FTIR gas cell (Antaris IGS gas
analyzer) to create the negative pressure. The gas cell size is 2 M in length which will
maintain temperature and pressure to resist gas from condensation. Finally, K-type
(model TP-01A) thermocouples will be used to measure the temperature and SD data
logger will be used to record the temperature data. We will use thermal camera (FLIR
E86, wavelength 650 mn, temperature range º0C-650ºC). Commercial large-sized Li-ion
battery pack or module (Lithium-Ion phosphate, 500-600 kWh) will be used first to iden­
tify the battery fire risk. We will also use different gas collection methods, such as gas
washing bottles (Iplusmile,60 ml glass) and gas cell (thermo scientific) collection to check
the offline gas measurement. Ion chromatography (thermo scientific) will be used to ana­
lyze the offline gas measurement and then the data will be compared with the real-time
measurement.

5 CONCLUSIONS

In this study, a large-scale diesel fire (0.5gal capacity) was tested and measured the maximum
surrounding temperatures at different locations. Also, the maximum temperature achieved by
full-scale diesel fire temperature using IR radiation. The maximum fire temperature was
601ºC during the burning. This study suggests that air direction and velocity to be controlled
to reproduce the data. Overall, this fire can be distinguished with respect to time which opens
a way to understand more about fire characteristics. The primary analysis of the combustion
fire and changes in temperature reveals that the thermocouple and thermal camera will be
handy in determining the fire-related properties. Also, this study, provides the basic informa­
tion of fire behaviors air direction, and gas collection. This research will be conducted using
a real battery used in mining vehicles to identify toxic gases, suppression techniques and ther­
mal management systems.

397
REFERENCES

Brett, D., n.d. airportwatch:Another cargo aircraft carrying batteries crashes – off South Korea.
Chen, M., Dongxu, O., Liu, J., Wang, J., 2019. Investigation on thermal and fire propagation behaviors
of multiple lithium-ion batteries within the package. Appl. Therm. Eng. 157, 113750.
Chen, M., He, Y., De Zhou, C., Richard, Y., Wang, J., 2016. Experimental Study on the Combustion
Characteristics of Primary Lithium Batteries Fire. Fire Technol. 52, 365–385. https://doi.org/10.1007/
s10694-014-0450-1
Chen, M., Zhou, D., Chen, X., Zhang, W., Liu, J., Yuen, R., Wang, J., 2015. Investigation on the ther­
mal hazards of 18650 lithium ion batteries by fire calorimeter. J. Therm. Anal. Calorim. 122, 755–763.
https://doi.org/10.1007/s10973-015-4751-5
Hansen, R., 2019. Design of fire scenarios for Australian underground hard rock mines–Applying data
from full-scale fire experiments. J. Sustain. Min. 18, 163–173.
Huang, P., Wang, Q., Li, K., Ping, P., Sun, J., 2015. The combustion behavior of large scale lithium
titanate battery. Sci. Rep. 5, 1–12. https://doi.org/10.1038/srep07788
Kuznetsov, G. V, Volkov, R.S., Sviridenko, A.S., Strizhak, P.A., 2022. Fast detection of compartment
fires under different heating conditions of materials. Process Saf. Environ. Prot. 168, 257–274.
Larsson, F., Andersson, P., Blomqvist, P., Mellander, B.E., 2017. Toxic fluoride gas emissions from
lithium-ion battery fires. Sci. Rep. 7, 1–13. https://doi.org/10.1038/s41598-017-09784-z
Liu, P., Li, Y., Mao, B., Chen, M., Huang, Z., Wang, Q., 2021. Experimental study on thermal runaway
and fire behaviors of large format lithium iron phosphate battery. Appl. Therm. Eng. 192, 116949.
https://doi.org/10.1016/j.applthermaleng.2021.116949
Ping, P., Wang, Q., Huang, P., Li, K., Sun, J., Kong, D., Chen, C., 2015. Study of the fire behavior of
high-energy lithium-ion batteries with full-scale burning test. J. Power Sources 285, 80–89.
Release, N., 2021. Newmont and Caterpillar Announce Revolutionary Strategic Alliance to Achieve
Zero Emissions Mining1–4.
Scrosati, B., Garche, J., 2010. Lithium batteries: Status, prospects and future. J. Power Sources 195,
2419–2430. https://doi.org/10.1016/j.jpowsour.2009.11.048
Skidmore, Z., 2021. Battery electric vehicles – facilitating low-carbon mines.
Somandepalli, V., Marr, K., Horn, Q., 2014. Quantification of combustion hazards of thermal runaway
failures in lithium-ion batteries. SAE Int. J. Altern. Powertrains 3, 98–104. https://doi.org/10.4271/
2014-01-1857
Spieß, B., Metzsch-Zilligen, E., Pfaendner, R., 2021. Mechanistic evaluation of flame retardants during
UL94 standard testing via IR-camera. Polym. Test. 103, 107320.
Stampatori, D., Raimondi, P.P., Noussan, M., 2020. Li-ion batteries: A review of a key technology for
transport decarbonization. Energies 13. https://doi.org/10.3390/en13102638
Sun, P., Bisschop, R., Niu, H., Huang, X., 2020. A Review of Battery Fires in Electric Vehicles, Fire
Technology. Springer US. https://doi.org/10.1007/s10694-019-00944-3
Xu, T., Lei, P., 2022. Experimental study on flame height and heat release rate estimation of
diesel-wetted wood powder fire. Case Stud. Therm. Eng. 33, 101906.
Yuan, L., Dubaniewicz, T., Zlochower, I., Thomas, R., Rayyan, N., 2020. Experimental study on ther­
mal runaway and vented gases of lithium-ion cells. Process Saf. Environ. Prot. 144, 186–192. https://
doi.org/10.1016/j.psep.2020.07.028
Yuan, L., Thomas, R.A., Rowland, J.H., Zhou, L., 2018. Early fire detection for underground diesel fuel
storage areas. Process Saf. Environ. Prot. 119, 69–74.

398
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Fire-induced temperature attenuation under the influence of

a single ceiling smoke extraction point in a bifurcated drift

O.B. Salami & G. Xu

Department of Mining and Explosives Engineering, Missouri University of Science and Technology, Rolla,
Missouri, USA

A.R. Kumar
Department of Energy and Mineral Engineering, The Pennsylvania State University, University Park, PA, USA

R.I. Pushparaj & A. Iqbal

Department of Mining and Explosives Engineering, Missouri University of Science and Technology, Rolla,
Missouri, USA

ABSTRACT: Experimental investigation of temperature decay and maximum smoke tem­

perature beneath an underground mine drift (54 m long, 2.6 m wide, and height of 3.2 m) was
carried out to examine the effect of ceiling smoke extraction on fire evolution in an under­
ground mine. A single non-centralized smoke extraction with an extraction flow rate between
0.24-5.42 m3/s was considered, the measured longitudinal velocity was between 0.012-0.220 m/s,
and the fire heat release rate (HRR) was between 85-425 KW. The results show that the
maximum temperature decreases with the increase in exhaust air volume under the same
HRR. Furthermore, an empirical correlation was developed to predict smoke temperature
decay beneath the ceiling due to the effect of a single exhaust smoke extraction point. In this
study, a comparison of temperature models for different extraction points is further ana­
lyzed to investigate the effect of the number of smoke extraction points on the temperature
attenuation coefficient. The model developed can be applied to other practical solutions to
predict temperature decay beneath the ceiling for axis-symmetric fires in an underground
mine drift for a single-point smoke extraction system.

Keywords: Full-scale fire, underground mines, temperature attenuation, heat release rate

1 INTRODUCTION

Fires in confined underground spaces are very deadly due to their impending death (Fernán­
dez-Alaiz et al., 2020). The complexity of underground mines owing to their layout, ventila­
tion limitations, and the presence of hazardous and flammable materials in such environments
make underground fire incidents particularly perilous (De Rosa, 2004, Fernández-Alaiz et al.,
2020). Studies have shown that mine fires produce significant levels of toxic gases and smoke
which can be transported through the ventilation network to various parts of the underground
workings. These noxious gases released during fires further reduce visibility and escape
chances of the underground personnel, and research has shown that most fatalities due to
underground fires can be attributed to the inhalation of poisonous gases which is mainly
carbon monoxide (Bahrami et al., 2021, Zhou, 2009, Zhou et al., 2018).
Contemporary studies have investigated temperature decay and maximum gas temperature
beneath the ceiling for different ceiling extraction strategies (Tang et al., 2018a, Tang et al.,

DOI: 10.1201/9781003429241-42

399
2018b, Tao et al., 2022). In one of the studies, Tang et al. (Tang et al., 2018a), examined the
combined effect of longitudinal wind and single smoke ceiling extraction on the temperature dis­
tribution in a model-scale horizontal channel with a similarity scale of 1:66 based on Froude
correlation. They proposed an empirical model for the set experimental conditions which could
be applied to similar practical situations. The research showed that the longitudinal temperature
decays faster in upstream positions than in downstream positions. In another study, Tang et al.
(Tang et al., 2018b) argued that most of the previously suggested correlations were developed
using a square or circular fire source. They opined that it is important to extend the geometry of
the fire source to account for other practical tunnel fire scenarios such as rectangular fire
source. Hence, a new empirical correlation for temperature evolution with combined smoke ceil­
ing extraction induced by a rectangular fire source was established. Tao’s team (Tao et al.,
2022) on the other hand, used a model scale tunnel with a Froude similarity criterion of 1:10 to
explore the temperature distribution in tunnels with the combined effect of a single-side central­
ized smoke extraction system. Their study showed that the maximum temperature initially
increases and then decreases when the smoke extraction rate is increased with fires having the
same HRR. A new temperature decay model for the single-side centralized smoke extraction
system was established based on the experimental results.
Most of our knowledge on fire and toxic gases spread in underground mines environment is
based on previous studies from straight tunnel fires or scaled model tests. This makes it impossible
to preserve all the dimensionless terms derived from the scaling theory (Li and Ingason, 2012).
Also, most existing studies do not account for the multiple bifurcations and environmental condi­
tions common in underground mines. Studies showed that the thermal properties of the material
used for building the model scale tunnel may significantly influence temperature and heat transfer
(Li and Ingason, 2012). Hence, there is a need to conduct full-scale fire tests for a deeper under­
standing of fire behavior and temperature decay in real underground mining situations.
This study investigates the interactive influence of ventilation conditions induced by ceiling
smoke extraction rate and fire size on temperature attenuation, and maximum excess gas tem­
perature in an experimental underground mine. The schematic of the experimental rig and
data communication system used in this study is illustrated in Figure 1. The scale was first set
up and tarred to zero. The fire pan was then placed on a 30.0 cm by 30.0 cm frame installed
on the scale to alleviate the exposure of the sensitive scale electronics to high temperatures.
Diesel fuel was then poured into the pan. The monitors and sensors were turned on and syn­
chronized to log the data from a remote computer before the fire pool is ignited. Full-scale
fire tests were then carried out in the main drift of the mine to evaluate the effect of varying
ventilation conditions on temperature and toxic gas distribution. Based on the results, an
empirical correlation for temperature decay in the mine drift was developed. A new model for
a single-side non-centralized smoke extraction system in an underground space was proposed.
The obtained data from this study will also be very useful to validate existing mine fire compu­
tational fluid dynamics (CFD) models and model scale test results based on Froude similarity.

Figure 1. Schematic of the experimental rig and data communication system used for this study.

400
2 FULL-SCALE FIRE TEST

2.1 The experimental mine

Full-scale fire experiments were carried out in the main drift of the Missouri S&T experimen­
tal mine. The layout of the mine is presented in Figure 2. The mine airways are approximately
rectangular with an average width of 2.6 m and height of 3.2 m along the major drift. The
mine is a typical room and pillar underground mine with four pillars and a main drift about
54.0 m long. Air enters the mine from the entrance at the lower part of the map in Figure 2
and is exhausted through a main fan on the surface.

Figure 2. Plan view of the study location and the exhaust fan.

2.2 Instrumentation and measurements

2.2.1 Full-scale fire setup

The airflow quantity and longitudinal ventilation velocity were determined at two cross-
sectional stations in the mine before each fire test using the moving and point traverse
methods. The first station was located 1.0 m from the mine inlet. The other station was
located along the major drift in the mine 3.0 m downstream of the fire source. Both traverses
were carried out with a rotating vane electrical anemometer. For the point traverse, the meas­
urement was done by dividing the cross-section into nine equal rectangular areas. Velocities
were determined at the center points of each area denoted as P1 to P9 as shown in Figure 3.
Two factors, the fire pool size, and ventilation rate were used for the design. Each factor had
three levels and, therefore, a total of nine experimental runs were conducted using a 3k factor­
ial design as shown in Table 1.

Table 1. Randomized experimental test cases.

Experiment 1 2 3 4 5 6 7 8 9

Ventilation rate (%) 100 50 0 0 50 50 100 100 0

Exhaust flow rate (m3/s) 5.42 2.97 0.24 0.24 2.97 2.97 5.42 5.42 0.24
Fire Size (m) 0.7 0.5 0.5 0.7 0.7 0.3 0.3 0.5 0.3
Ambient Temperature (°C) 27.5 22.5 22.5 22.5 19.5 19.5 19.5 19.5 19.5

The diameters of the fire pool containers were 0.3 m, 0.5 m, and 0.7 m as shown in
Figure 4 (a). Three ventilation conditions were studied. They were varied by adjusting the ceiling
smoke extraction rate between 0.0-100.0 %. The fan was turned off to eliminate mechanical venti­
lation; this case was denoted as the ‘0.0 %’ ventilation condition. The mine was under natural
ventilation in this condition. The second case is the ‘50.0 %’ ventilation rate which was realized by

401
turning the fan on but blocking half of the fan duct cross-section area using a circular wooden
board as shown in Figure 4 (b). Whereas for the third case, the fan was turned on and the duct
was kept fully open to realize the ‘100.0 %’ ventilation condition (as depicted in Figure 4 (c)).
A constant diesel volume, 0.5 gals, was used for each test. After each test, the fan was run at
100.0 % ventilation condition for 10.0 mins, to remove the smoke and allow the underground
temperature to reach ambient conditions. The fire pool was placed at 13.0 m from the in-by
portal along the main drift (indicated by a fire icon in Figure 2). A ventilation door is located at
about 12.0 m from the fire pool downstream which separates the testing region from the blind
heading of the main drift in the underground and thus, the fire could be considered axisymmetric
within the test region in the underground mine. The smoke temperature at the mine ceiling was
determined using K-type thermocouples with an accuracy of 0.025°C and a diameter of 0.6 mm
connected to a data logger. At the roof of the mine, thermocouples were installed at all seven
locations (S1-S7) as depicted in Figure 5.

Figure 3. Velocity measuring points at the station’s cross-section.

Figure 4. Fire pans and the fan opening.

3 TEMPERATURE DISTRIBUTION

Understanding the temperature distribution in the underground environment is critical for

fire safety design factors such as the selection of appropriate tunnel lining material, deploy­
ment of proper smoke control strategies, fire suppression systems, and emergency evacuation
planning Two main temperature indicators explain the distribution well. They are, (a) the tem­
perature attenuation and, (b) the maximum excess gas temperature beneath the tunnel ceiling.
Temperature attenuation may either be longitudinal or transverse (Li & Ingason, 2018). Lon­
gitudinal attenuation indicates the temperature decay along the underground drift or tunnel

402
while transverse attenuation measures the temperature decay between the walls of the tunnel.
However, the temperature variation along the tunnel or mine airways is more important as
occupants tend to evacuate along the tunnel instead of moving zigzag between its walls. In
this study, only the longitudinal temperature attenuation and maximum excess smoke tem­
perature beneath the ceiling were considered. The theoretical background of temperature
decay beneath the ceiling is presented in section 3.1.

Figure 5. Experimental setup and drift instrumentation.

3.1 Longitudinal temperature attenuation

To develop a reliable emergency evacuation plan and firefighting technique for underground
space, it is imperative to understand the temperature decay pattern in such environments.
Many researchers have investigated longitudinal temperature attenuation in tunnels and
underground structures (Delichatsios, 1981, Hu et al., 2007, Hu et al., 2013, Huang et al.,
2018, Tang et al., 2020, Zhao et al., 2018). In an earlier study, an empirical model for tempera­
ture distribution beneath a tunnel ceiling was developed for centrally located fire in beamed
channels as follows (Delichatsios, 1981):

where ΔT0 is the average temperature rise near the ceiling over the fire source, ΔT is the
average instantaneous temperature rise at the specified location from the fire axis, lb represents
the beam depth, H is the separation between the fire surface and the ceiling, and St is the Stan­
ton number which was estimated to be 0.03 from the correlation of experimental data
obtained from the study. The temperature attenuation was found to conform with exponential
decay laws and subsequent investigations have corroborated this finding. A later study by Hu
et al. (Hu et al., 2013, Hu et al., 2007) which involved a series of full-scale tests in large tunnels
and vehicular tunnels showed that temperature beneath the ceiling in a longitudinally venti­
lated tunnel can be expressed as an exponential decay given by:

403
where ∆Tx is the maximum increase in temperature x m away from a reference point, ∆Tt is the
maximum increase in temperature at the reference point, x is the distance from the coordinate
origin, xr denotes the distance of separation between the fire and the coordinate origin, and β is
the temperature attenuation factor. More recently, Zhao et al. (Zhao et al., 2019) investigated
the ceiling temperature distribution in a 1:20 reduced-scaled tunnel model tunnel and estimated
the attenuation factor to be 0.536 for a tunnel fire with a two-point extraction ventilation
system. Thus, using their attenuation factor, a modified Hu’s equation was obtained:

In another study which involved a total of twelve tests in a scaled-model tunnel (1:23), Inga­
son and Li (Ingason and Li, 2010), demonstrated that the distribution of excess gas tempera­
ture beneath the ceiling without mechanical ceiling smoke extraction can be expressed as
a sum of two attenuations given by:

Therefore, the generalized temperature decay model beneath the tunnel/drift ceiling can be
modified using the exponential model as follow:

where A, B, F1, and F2 are coefficients.

Although Hu’s model, with the assumption of zero entrainments of air from counter flow,
remains the most widely used model for tunnel ceiling temperature attenuation, most equations
presented above could only be applied to the one-dimensional spreading phase (see Figure 6)
downstream of the fire (Zhao et al., 2018). However, it is important to understand how the
temperature evolves along the tunnel from a reference point upstream. This will help firefight­
ers and rescue teams to perceive the fire situation in the confined space based on the observed
temperature upstream of the fire. This study becomes even more significant as most rescue
teams usually approach the fire from the upstream direction. Figure 6 illustrates an emergency
scenario where miners try to escape from fire and the firefighters/rescue team advances toward
the fire to put out the fire to enable the evacuation of the trapped occupants.

Figure 6. Illustration of the region of applicability of existing temperature attenuation models.

404
4 RESULTS AND DISCUSSION

4.1 Airflow/ Longitudinal Ventilation velocity

The airflow measurement during the experiments was obtained following several steps. Firstly,
the cross-sectional traverse was done 1.0 m from the air inlet to determine the airflow in the
mine for each of the ventilation scenarios considered. Ventilation flow rates were set at 0.0 %,
50.0 %, and 100.0 % for the exhaust shaft as described earlier in section 2.2.1. The calculated
airflow at the fan exhaust was 0.24 m3/s, 2.97 m3/s, and 5.42 m3/s for the three ventilation
conditions. For each of the three ventilation cases, the moving traverse was repeated three
times until an acceptable error limit of < 5.0 % is obtained. The average velocity was calcu­
lated, and the results are presented in Table 2. Similarly, the airflow was measured
3.0 m downstream of the fire using the point traverse method for each test before starting
fires. This was done for the three different ceiling extraction rates to determine the longitu­
dinal airflow across the fire plume. The results are presented in Figure 3. The average velocity
for each case is then obtained by calculating the mean of the nine values. The calculated aver­
age longitudinal ventilation velocity in the main drift is 0.220 m/s, 0.093 m/s, and 0.012 m/s
for 100.0 %, 50.0 % smoke extraction rate, and natural ventilation conditions respectively.

Table 2. Moving traverse ventilation measurement results.

Test 1 2 3 4 5 6 7 8 9

Ventilation size (%) 100 50 0 0 50 50 100 100 0

Calculated average velocity (m/s) 0.84 0.41 0.21 0.21 0.41 0.41 0.84 0.84 0.21
Error (%) 1.60 0.81 3.74 3.74 0.81 0.81 1.6 1.6 3.74

Table 3. Point traverse airflow upstream of the fire for different experimental cases.
S/N P1 P2 P3 P4 P5 P6 P7 P8 P9

Exp 1 0.47 0.33 0.27 0.32 0.12 0.00 0.33 0.16 0.04
Exp 2 0.16 0.16 0.21 0.00 0.00 0.07 0.07 0.11 0.06
Exp 3 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.04
Exp 4 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.04
Exp 5 0.16 0.16 0.21 0.00 0.00 0.07 0.07 0.11 0.06
Exp 6 0.16 0.16 0.21 0.00 0.00 0.07 0.07 0.11 0.06
Exp 7 0.47 0.33 0.27 0.32 0.12 0.00 0.33 0.16 0.04
Exp 8 0.47 0.33 0.27 0.32 0.12 0.00 0.33 0.16 0.04
Exp 9 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.04

4.2 Heat Release Rate (HRR) calculations

The heat release rate (HRR) is arguably the most important fire parameter. It is also a very
critical input for CFD simulations of fire. The HRR is used to estimate the fire intensity. It
can be determined by measuring the changes in the mass of the fuel to the burning time as
given in equations (6) and (7) (Zhou et al., 2018, Zhang, 2012, Iqbal et al., 2004, Marková
et al., 2020, Yuan et al., 2016).

where m'' is the burning rate or mass loss rate per unit area per unit time (kgm-2s-1), ∆Hc,eff is
the effective heat of combustion (kJ.kg-1), kβ is the empirical constant (m-1), and D is the

405
diameter of the burning area (m). The equation could be simplified by eliminating the term in
the parentheses and the HRR could be obtained from the product of mass loss rate and heat
of combustion as given by the equation (7) (Yuan et al., 2016, Zhou et al., 2018).
For this study, the HRR was determined from the mass loss rate of the diesel using the sim­
plified equation (7). The mass loss rate and the corresponding HRR results in the order of
decreasing maximum temperature are presented in Table 4. Table 4 shows that the fire size
has a significant influence on the burning rate. Longitudinal ventilation has a negligible effect
on the burning rate due to the very low ventilation velocity observed in the mine drift during
the fire tests. For instance, considering test experimental cases 5 and 6, a decrease in fire size
to about half may cause the burning time to increase with a factor of up to 4.5 if the longitu­
dinal ventilation is kept constant.

4.3 Temperature attenuation factor

The temperature evolution curves measured for the center point experiment (i.e., experiment with
0.5D and 50.0 % ventilation) during the full-scale fire test (Test 2) are shown in Figure 7.
A quasi-steady stage was reached at about 7.0 mins for the different measuring stations after igni­
tion. In Figure 7, the negative distance signs represent upstream locations while the positive dis­
tance signs indicate downstream locations. The maximum measured temperature decreases as the
distance increases from the fire source. Also, from the summary of exponential fittings presented
in Table 5, the temperature attenuation was observed to be greater downstream (attenuation
factor of 0.17) than upstream (attenuation factor of 0.12). This may be due to smoke rollback
upstream and smoke mixing faster with fresh air as they travel downstream towards the exhaust.
Since the study focused on the maximum temperature rise for each station to determine the
attenuation factor, the maximum rise in smoke temperature (as shown in Figure 8) within the
quasi-steady stage at the different measuring stations was employed for subsequent calculations.

Table 4. Heat release rate values obtained for the experimental cases.
S/N Test Case Pool diameter (m) Ventilation Opening (%) Mass loss rate (g/s) Max HRR (KW)

1 Exp 4 0.7 0.0 0.4313 425.0

2 Exp 5 0.7 50.0 0.1927 425.0
3 Exp 1 0.7 100.0 0.2307 425.0
4 Exp 3 0.5 0.0 0.4500 212.5
5 Exp 2 0.5 50.0 0.4420 212.5
6 Exp 8 0.5 100.0 0.0869 212.5
7 Exp 9 0.3 0.0 0.0765 85.0
8 Exp 6 0.3 50.0 0.0217 85.0
9 Exp 7 0.3 100.0 0.0600 85.0

For the calculation model of temperature attenuation upstream and downstream of fire, an
improved technique (Tao and Zeng, 2022) that incorporates a two-dimensional spreading
phase for temperature distribution was introduced instead of the conventional one-
dimensional temperature decay model found in many works of literature. First, normalized
the smoke temperature Tx was normalized, measured by the thermocouples from a distance
x from the fire to a dimensionless temperature ∆TND,x by a reference value Tr and the ambient
air temperature Ta as follows:

The reference temperatures Tr for the upstream and downstream positions were taken at
station S3 and station S5 (as shown in Figure 9 and Figure 10), 2.0 m upstream and

406
Figure 7. Temperature curves measured at different distances for Test 2 (Centre point experiment).

Figure 8. The maximum rise in smoke temperature at mine roof for the different ventilation conditions.

downstream of the fire source respectively. The dimensionless normalized temperatures are
plotted against the distance of separation of the thermocouples measuring stations to the fire
source as seen in Figure 9 to Figure 11. The experimental data from each test was fitted to an
exponential function using the following equations:

The dimensionless normalized temperatures can be properly fitted using the exponential
function (Zhao et al., 2019). The average value of the attenuation factors was determined and
used for developing the empirical model in this study. Finally, a two-dimensional spreading
phase empirical model was developed that could predict the temperature distribution for an
axisymmetric fire in a mine drift with multiple bifurcations. From the results in Table 5, the
semi-empirical calculation model is summarized as follows:

The exponential fitting of dimensionless smoke temperature distribution along the under­
ground drift is presented in Table 5. The average values of a, b, and R2 are calculated. The R2
value for upstream and downstream are well-fitted which indicates a strong correlation of
temperature with the distance of separation from the fire source.
A comparison of the proposed temperature attenuation model with existing models was
done to highlight the difference between the temperature decay models. Ingason and Li

407
Figure 9. Correlation of temperature evolution in the mine drift for 425.0 kW (0.7D) fire under different
ventilation conditions.

Figure 10. Correlation of temperature Evolution in the mine drift for 212.5 kW (0.5D) fire under differ­
ent ventilation conditions.

Figure 11. Correlation of temperature Evolution in the mine drift for 85.0 kW (0.3D) fire under differ­
ent ventilation conditions.

Table 5. Summary of temperature

attenuation correlation.
Upstream a 0.6755
b -0.1219
R2 0.8901
Downstream a 0.9827
b -0.1703
R2 0.9813

408
(Ingason and Li, 2010) developed a model using a reduced-scale tunnel with no mechanical
smoke extraction. Zhao et al. (Zhao et al., 2019) developed a temperature attenuation model
using a reduced-scale model tunnel too but with a two-point mechanical smoke extraction
point. However, the attenuation model developed in this study was based on a single-point
smoke extraction system. It can be deduced from

Figure 12. Comparison of the temperature attenuation model developed and existing typical tempera­
ture attenuation models.

5 CONCLUSIONS

Full-scale fire tests were conducted in underground mine drift. The analysis of temperature
attenuation and maximum excess gas temperature beneath a mine drift is presented in this
study using experimental and theoretical approaches. The temperature evolution upstream
and downstream of the mine drift were fitted using exponential correlation and a new empir­
ical model of temperature attenuation given by Equation (10) was proposed. The attenuation
model upstream of the fire could predict the temperature rise as we approach the fire source
from a reference point upstream of the fire. This correlation is valid for HRR value between
85-425 KW and maximum excess gas temperature of up to 160 °C. This type of prognostica­
tive attenuation model will further improve firefighting and suppression strategy as responders
move towards the fire zone during an emergency. Finally, the findings from this study could
serve as fire-field data in building and optimizing emergency evacuation plans, in addition to
validating existing CFD models for underground mines and other confined space
environments.

REFERENCES

Bahrami, D., Zhou, L. & Yuan, L. (2021) Field Verification of an Improved Mine Fire Location Model.
Mining, Metallurgy & Exploration, 38, 559–566.
De Rosa, M. I. (2004) Analysis of mine fires for all US metal/nonmetal mining categories: 1990–2001.
Delichatsios, M. A. (1981) The flow of fire gases under a beamed ceiling. Combustion and Flame, 43,
1–10.
Fernández-Alaiz, F., Castañón, A. M., Gómez-Fernández, F. & Bascompta, M. (2020) Mine fire behav­
ior under different ventilation conditions: Real-scale tests and CFD modeling. Applied sciences, 10,
3380.

409
Hu, L., Chen, L., Wu, L., Li, Y., Zhang, J. & Meng, N. (2013) An experimental investigation and correl­
ation on buoyant gas temperature below ceiling in a slopping tunnel fire. Applied Thermal Engineering,
51, 246–254.
Hu, L., Huo, R., Wang, H., Li, Y. & Yang, R. (2007) Experimental studies on fire-induced buoyant
smoke temperature distribution along tunnel ceiling. Building and Environment, 42, 3905–3915.
Huang, Y., Li, Y., Dong, B., Li, J. & Liang, Q. (2018) Numerical investigation on the maximum ceiling
temperature and longitudinal decay in a sealing tunnel fire. Tunnelling and Underground Space Tech­
nology, 72, 120–130.
Ingason, H. & Li, Y. Z. (2010) Model scale tunnel fire tests with longitudinal ventilation. Fire safety jour­
nal, 45, 371–384.
Iqbal, N., Salley, M. H. & Weerakkody, S. (2004) Fire dynamics tools (FDTs): quantitative fire hazard
analysis methods for the US Nuclear regulatory commission fire protection inspection program,
United States Nuclear Regulatory Commission, Office of Nuclear Regulatory . . ..
Li, Y. Z. & Ingason, H. (2012) The maximum ceiling gas temperature in a large tunnel fire. Fire safety
journal, 48, 38–48.
Marková, I., Lauko, J., Makovická Osvaldová, L., Mózer, V., Svetlík, J., Monoši, M. & Orinčák, M.
(2020) Fire size of gasoline pool fires. International journal of environmental research and public health,
17, 411.
Tang, F., Cao, Z., He, Z., Ling, X. & Wang, Q. (2018a) Thermal plume temperature profile of
buoyancy-driven ceiling jet in a channel fire using ceiling smoke extraction. Tunnelling and Under­
ground Space Technology, 78, 215–221.
Tang, F., Cao, Z., Palacios, A. & Wang, Q. (2018b) A study on the maximum temperature of ceiling jet
induced by rectangular-source fires in a tunnel using ceiling smoke extraction. International Journal of
Thermal Sciences, 127, 329–334.
Tang, F., Chen, L., Chen, Y. & Pang, H. (2020) Experimental study on the effect of ceiling mechanical
smoke extraction system on transverse temperature decay induced by ceiling jet in the tunnel. Inter­
national Journal of Thermal Sciences, 152, 106294.
Tao, L. & Zeng, Y. (2022) Effect of single-side centralized exhaust on smoke control and temperature
distribution in longitudinal ventilation tunnel fires. Tunnelling and Underground Space Technology,
119, 104241.
Tao, L., Zeng, Y., Li, J., Yang, G., Fang, Y. & Li, B. (2022) Study on the maximum temperature and
temperature decay in single-side centralized smoke exhaust tunnel fires. International Journal of Ther­
mal Sciences, 172, 107277.
Yuan, L., Zhou, L. & Smith, A. C. (2016) Modeling carbon monoxide spread in underground mine fires.
Applied Thermal Engineering, 100, 1319–1326.
Zhang, X. (2012) Underground mine fire simulation using multiscale modeling approach, Missouri Univer­
sity of Science and Technology.
Zhao, P., Yuan, Z., Yuan, Y., Yu, N. & Yu, T. (2019) A study on ceiling temperature distribution and
critical exhaust volumetric flow rate in a long-distance subway tunnel fire with a two-point extraction
ventilation system. Energies, 12, 1411.
Zhao, S., Liu, F., Wang, F. & Weng, M. (2018) Experimental studies on fire-induced temperature distri­
bution below ceiling in a longitudinal ventilated metro tunnel. Tunnelling and Underground Space
Technology, 72, 281–293.
Zhou, L. (2009) Improvement of the mine fire simulation program MFIRE.
Zhou, L., Yuan, L., Bahrami, D., Thomas, R. A. & Rowland, J. H. (2018) Numerical and experimental
investigation of carbon monoxide spread in underground mine fires. Journal of Fire Sciences, 36,
406–418.

410
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Application of pressure balancing techniques at the West Elk coal

mine

Charles Kocsis, Felipe Calizaya, Jeffrey Johnson, Tulio Dias & Natanna Nunes
Department of Mining Engineering, University of Utah

Elizabeth Lindgren & Garren Atchley

Mountain Coal Company, West Elk Mine, Somerset, Colorado

John Poulos
John Poulos, Geosyntec Consultants Inc., Lakewood, Colorado

ABSTRACT: Spontaneous combustion is a serious safety hazard in underground coal

mines, particularly in mined-out areas (e.g., gob). It is estimated to be the cause of more than
20% of coal mine fires. Pressure balancing is a ventilation technique that can be used to
reduce or eliminate the ingress of oxygen to the mine gob (in-gassing), thus reducing the risk
of spontaneous combustion. At the same time, pressure balancing can also be used to reduce
the risk of methane seeping into the mine ventilation circuits from the gob (out-gassing), due
to barometric pressure changes on the surface. Two pressure balancing chambers have been
constructed at the West Elk Mine, Colorado, operated by the Mountain Coal Company
(MCC): one in a mined-out area, and another close to the active area, near a longwall mine
gob. Each chamber is equipped with an isolation stopping, safety doors, a nitrogen injection
system, and a set of pressure and environmental monitoring sensors. Several pressure balan­
cing tests for different ventilation conditions have been conducted in these chambers. This
study presents a summary of a procedure used to conduct these tests, the results achieved, and
the steps taken to reduce the risk of spontaneous combustion.

1 INTRODUCTION

In 2016, the University of Utah completed a research project on “Control of Spontaneous

Combustion Using Pressure Balancing Techniques”, funded by the Alpha Foundation. The
main objective of this project was to construct a physical ventilation model in which different
types of pressure-balancing systems can be tested and evaluated. The model included
a network of ventilation ducts assembled to represent a coal mine, a simulated mine gob,
a CO2-based pressure chamber, and an atmospheric monitoring and control system. The
model was used to conduct pressure balancing tests designed to reduce pressure differentials
between the chamber and the simulated mine gob. In a coal mine, this process would reduce
or eliminate the inflow of oxygen that leads to the self-heating of coal, thus reducing the possi­
bility of a spontaneous combustion (Sponcom) event. The results of these tests showed that
pressure balancing can be used effectively for the control of spontaneous combustion condi­
tions in U.S. coal mines. Upon the conclusion of the project, a recommendation was to “test
the operation principles of a pressure balancing system in the field.”
Testing the above principles in the field required an underground coal mine where pressure
balancing tests can be conducted. Several mine operators were contacted requesting help with
the project. The West Elk Mine operated by Mountain Coal Company (MCC) showed interest
in participating. The mine operates one longwall and up to three development sections. The

DOI: 10.1201/9781003429241-43

411
mine uses a blower-type ventilation system equipped with two surface fans. The longwall
panel is ventilated by a bleederless ventilation system where the mined-out areas are isolated
by means of high-pressure seals and the gob is inertized with nitrogen gas that is injected from
two to five cross-cuts inby to the face from the headgate side (Calizaya et.al. 2016).
The aim of this project is to construct two pressure chambers at the West Elk mine: one in
a mined-out (passive) area, and another near an active longwall mine gob. Each chamber would
be equipped with a nitrogen injection system, flow control valves, and pressure sensors. The
chambers would be tested under different ventilation conditions. This study presents a summary
of the results achieved, and the steps taken to minimize the risk of starting a fire in a mine gob.

2 PRESSURE BALANCING TECHNIQUES

Pressure balancing is a ventilation technique used to neutralize the pressure differences

around and across caved areas (e.g., gob) of an underground coal mine. Ideally, if these pres­
sure differences can be reduced to zero, then there will be no leakage through the stoppings
and seals and surrounding strata thus there will be no oxygen to start and sustain the self-
heating of coal. In a mine, pressure balancing can simply be accomplished by either increasing
the air pressure on the gob side of the seal or by decreasing the pressure on the ventilated side
of the seal until the leakage is reduced to zero. It can also be accomplished by constructing
pressure chambers and pressurizing them with inert gas, such as nitrogen. Pressure balancing
has been used in many coal mining countries, but not much in US coal mines. Australia, the
United Kingdom, South Africa, India, and some European countries have been utilizing these
techniques to combat and prevent fires in underground mines for many years (Ray 2007,
Chalmers 2008 & Grubb 2008). Except for a few cases, this technique has not been used
within the United States (Smith & Lazzara 1987 & Bessinger et al. 2005).
There are two types of pressure balancing systems: passive and active. The main difference
between these two systems is the amount of pressure difference that can be neutralized. In
a passive system, the ventilation pressure available in the area to be neutralized limits its appli­
cation. In an active system, an inert gas, such as nitrogen, is delivered to an area at very high
pressure; therefore, it can be used to neutralize large pressure differences.

2.1 Passive pressure balancing

Passive pressure balancing is achieved by changing airway and regulator resistances, pressure
differentials across the stoppings, and flow quantities near the gob. It can also be achieved by
using atmospheric pressure to pressurize the chambers. This practice is used mainly when the
overburden of the caved area is shallow and fractured where barometric pressure fluctuations
can influence air leakage and cause airflow reversals in the sealed area (Bhowmick 1992 &
Moreby 2009). Usually, in a passive pressure balancing system, the chamber is pressurized by
using ventilation air from existing airways in a controlled manner. Alternatively, boreholes
driven from the surface to the gob area are used to pressurize the chamber.

2.2 Active pressure balancing

Active pressure balancing is achieved by using an external pressure source usually in the form
of inert gas that is injected into the chamber in a controlled manner. It requires the construc­
tion of a pressure chamber, the installation of an inert gas injection system, and the operation
of gas flow control valves when abnormal conditions are detected. Barometers and manom­
eters are used to monitor atmospheric conditions and pressure differentials across the stop­
pings and seals and to determine the direction of the leakage flow. When the leakage flow is
from the gob to the chamber (outgassing of explosive gases) or from the chamber to the gob
(ingress of air), the chamber is pressurized by using an inert gas. This method can be used to
overcome high-pressure differentials such as those caused by sudden changes in barometric
pressure on the surface. When this pressure decreases, the gas volume in the gob will increase,

412
causing emissions of the air-gas mixture from the gob. Similarly, when the barometric pressure
increases the gas volume will contract to allow the ventilation air to migrate into the gob
(Francart & Beiter 1997, & Schatzel et al. 2012). Pressure changes of this kind can be con­
trolled by operating an active pressure-balancing system.

3 CONSTRUCTION OF PRESSURE CHAMBERS

Following a mine visit, the University of Utah and the mine personnel decided to develop an
action plan to start the construction of the 1st chamber at the West Elk mine. The chamber would
be established by constructing a Kennedy stopping at about 3 m (10 ft) in front of an isolation
seal in a mined-out area. The standard stopping design used in the mine would be modified to
include two personnel doors and curtains to ventilate the chamber and to allow the mine person­
nel to inspect the seal. The chamber would be equipped with a set of manometers, pressure trans­
ducers, gas sampling ports, and a nitrogen gas injection system, connected through control valves
to the mine’s nitrogen pipeline.

3.1 Satisfying MSHA requirements

Because the project involves modifications to current stopping construction practices at the
mine, we were advised by MSHA to amend our action plan and to include safety features to
operate the chambers. Specifically, we included the following requirements in the plan:
a) Technical drawings of the chamber: Drawings must specify the type of stopping to be con­
structed, the type and size of man-doors, details of the nitrogen injection system, the loca­
tion of gas sampling points, and pressure taps.
b) Communication system: Phones and warning signs must be posted at all accesses to the
area and must be linked to the mine’s communication system to provide the mine operator
with the necessary information before, during, and after the test.
c) Location of monitors and control devices: The chamber must be equipped with oxygen
sensors upwind and downwind of the chamber with respective alert and alarm levels set at
19.7% and 19.5% O2. All control valves must be located upwind of the chamber.
d) Ventilation: A minimum air volume of 1.0 m3/s (2,000 CFM) is to be provided at the
chamber to allow enough air to dilute the contaminates from leakage.
e) Procedure to operate the chamber: There must be a written protocol to operate the pres­
sure chamber, to open and close the man doors, to purge the enclosed area, and to allow
the mine personnel to complete the weekly examination of the seals.
Based on the above requirements, technical drawings and procedures were revised, the com­
munication and ventilation systems were modified, and a protocol to operate the chamber was
developed. Figure 1 shows the revised drawing for the 1st pressure chamber. The modified
plan, including the protocol to operate the chamber, was re-submitted and approved by the
MSHA District Manager on January 22, 2020.

3.2 Construction of the 1st pressure chamber

The first chamber is located in the “B Seam in the Box Canyon North Mains” area of the mine. It
was established by installing a Kennedy stopping at about 3 m (10 ft) in front of an isolation seal.
The chamber is equipped with two safety doors, ventilation curtains, a nitrogen injection system,
pressure tubes, and a set of ventilation and environmental monitors. Once installed, the Kennedy
stopping was tested and sealed to minimize leakage. A 5 cm (2-in) HDPE pipeline was extended
from the main nitrogen source to the balance chamber location (~770 m). The pipeline is equipped
with a set of regulators, pressure gauges, and flow and control valves. Two pressure sampling
tubes were extended from the chamber to the control room. In addition, the chamber is equipped
with a set of pressure differential and oxygen sensors. Figure 2 shows two photographs of the
Kennedy stopping and a nitrogen injection system fitted with control valves, gages, and two gas
flow regulators.

413
Figure 1. Plan view of the modified pressure chamber.

Figure 2. West Elk mine pressure chamber construction details: (a) Kennedy stopping and digital man­
ometers, and (b) Nitrogen injection pipeline and flow control valves.

3.3 Construction of the 2nd pressure chamber

The second chamber is located at the “Sunset 1 Headgate, 2E” cross cut, near an active long­
wall panel. A visual inspection of this area showed that the roof, floor, and ribs are in excel­
lent condition. The existing joint patterns and stress-induced cracks are still tight. The seal is
set back at approximately 11.6 m (38 ft) from the intersection. As in the previous case, the
chamber was established by installing a Kennedy stopping at about 3 m in front of the 2E seal
(about 8.5 m from the intersection). The same MSHA-approved plan shown in Figure 1 was
used to construct and equip the chamber. The floor and walls were trenched and cleaned
before installing the Kennedy panels. The construction job was completed with the installation
of the nitrogen injection system and four electronic devices to monitor the following variables:
barometric pressure, oxygen concentration downwind of the chamber, and pressure differen­
tials across the seal and chamber respectively. Figure 3 shows two photographs of the 2nd
pressure chamber depicting the details of the Kennedy stopping, the nitrogen gas injection
system, and electronic manometers used to monitor pressure differentials.

414
Figure 3. Second pressure chamber: (a) Kennedy stopping and nitrogen injection system and (b) Hand-
held pressure sensors tapped to mine’s AMS.

4 PRESSURE BALANCING TESTS

In a mine ventilated by a blower fan, such as at the West Elk mine, a negative pressure differ­
ence across the isolation seal (P3 – P2 < 0) means that the ventilating air (oxygen) is flowing to
the gob through or around the seal. This may be sufficient to start the Sponcom process. The
risk can be reduced by injecting pressurized nitrogen into the chamber.
Several pressure balancing tests were conducted on each pressure chamber. Once the cham­
ber was constructed and equipped with a nitrogen injection system and pressure sensors, pres­
surized nitrogen was injected into the chamber and the pressure differentials were monitored.
Figure 4 shows a schematic of the nitrogen injection system. The system consists of a 5 cm
HDPE pipe extending between the main nitrogen source and the pressure chamber. Valves V1
and V2, regulators R1 and R2, and pressure gages G1 through G3 were all used in the 1st
chamber. However, for practical reasons, regulator R2 and gage G3 were removed in the 2nd
chamber. During each test, nitrogen flow rate and pressure differentials across the seal and
stopping were monitored continuously. While the seals held high-pressure differentials effect­
ively, the stoppings did not, particularly in the 1st chamber. To overcome the problem, the
Kennedy stopping was reinforced and sealant material was applied between the panels from
the high-pressure side. This helped but was not sufficient to hold a pressure difference of 1,000
Pa. Lessons learned from the 1st chamber were applied in the construction, testing, and moni­
toring of the 2nd chamber. The results of each test are presented below.

Figure 4. Nitrogen injection system - general layout.

4.1 Tests on the 1st chamber

Several climate monitoring and pressure differential tests were conducted on the 1st chamber.
Initially, tests were performed to investigate the effect of barometric pressure and temperature

415
variations on the pressure chamber, then, the chamber was filled with pressurized nitrogen and
evaluated for leakage through the Kennedy stopping. When a fixed flow rate of nitrogen was
injected into the chamber, the chamber pressure increased to a higher level. When the flow was
shut off, the chamber pressure dropped to its initial level rapidly. This was mainly due to gas
leakage through the Kennedy stopping. The results of the two experiments are presented below.

4.1.1 Test 1 – September 1- 30, 2021

This test was performed to determine the effect of atmospheric pressure on the chamber walls.
Figure 5 shows the results of this test. It shows three-time series: one for barometric pressure
and two for pressure differentials. Barometric pressures (green line) were recorded by a digital
barometer, located on the surface at the West Elk mine’s main office. The pressure differen­
tials (dp) were monitored by two digital manometers: one for pressure differentials across the
seal and another for pressure differentials across the Kennedy stopping. A quick evaluation of
these graphs shows that the barometric pressures vary on a daily basis and are inversely pro­
portional to changes in surface temperature. Figure 5 also shows two pressure differential
series: one for the isolation seal (orange line) and another for the Kennedy stopping (blue
line). An evaluation of these graphs shows that while the pressure differentials across the seal
varied between -2.36 in w.g. and +3.24 in. w.g. the pressure differentials across the stopping
did not show any significant change.
Another important finding is that the pressure differentials across the permanent seal are
inversely proportional to changes in barometric pressure (see highlighted area).

Figure 5. Barometric pressure and differential pressure variation across the seal and stopping.

4.1.2 Test 2 – December 2-3, 2021

This test was performed to determine the ability of the chamber to hold pressurized nitrogen.
The test started at approximately 9:07 AM. The control valves V1 and V2 (see Figure 4) were
open and the regulators R1 and R2 adjusted until the pressure gages G1 and G2 displayed 41
and 14 kPa (6 and 2 psi) respectively. Just after the start, the following DPs were recorded (see
Figure 6): + 400 Pa (+1.60 in-wg) for the isolation seal and -25 Pa (-0.10 in-wg) for the stop­
ping (outgassing in both cases). The test was stopped when a significant gas leakage through
the stopping was detected. During this period, the DP across the seal increased slowly to reach
a maximum of 875 Pa (3.5 in-wg) in about four hours. The DP across the stopping did not
change significantly. The DP across the seal increased from -100 to 870 Pa (-0.4 to +3.48 in-
wg) with a combined DP = +970 Pa (3.88 in-wg). Figure 6 also shows the barometric pressure
variations on the surface which started to decrease from 103.4 to 102.3 kPa (30.42 to 30.24 in-
Hg) during this period (highlighted by dotted lines). This represents a decrease of 614 Pa (2.5
in-wg). Consequently, the increase in DP across the seal is the result of a combined influence
of a decrease in barometric pressure and pressure loss in the chamber.

416
Figure 6. Pressure differentials series after nitrogen gas injection (December 2-3, 2021).

4.2 Tests on the 2nd chamber

Two pressure balancing tests were conducted on the 2nd chamber between August 9 and 10. In
each case, a test started by closing and locking the personnel doors, opening the nitrogen flow
control valves, V1 and V2, and changing the flow rate by means of regulator R1 (Figure 4).
Initially, valves V1 and V2 were open to full pressure, and the opening of the regulator R1
was set to allow very low pressure to the chamber. Then, the opening of this regulator
increased in steps of about ½ turn. Each test lasted for less than 30 min. While the nitrogen
flow rate was increased in steps, the barometric pressure and pressure differentials were moni­
tored continuously. The results of each test are presented below.
4.2.1 Test 1- August 9, 2022
The test started by opening the pressure control valves V1 and V2 and fixing the gas flow rate
to a set value by means of regulator R1. This was followed by recording pressure differentials
across the seal and stopping and monitoring for nitrogen gas leakage into the access drift. The
process was repeated for different gas flow rates. The test was stopped when the nitrogen gas
leakage from the chamber reduced the oxygen level in the surrounding area to less than 19.5%.
Chronology of Events – August 9, 2022

11:25 AM. For regulator R1 set to minimum flow, Valves V1 and V2 were open. Pressure
gages: G1 = 400 kPa (58 psi), G2 = 0, DP2-1 = 42 Pa (0.17 in w.g)
11:30 AM. Regulator position was increased by ½ turn. Pressure gages recorded: G1 = 385
kPa (56 psi), and G2 = 20.7 kPa (3 psi), and DP2-1 = 243 Pa (0.97 in-wg)
11:37 AM. Regulator position increased by another ½ turn. Gage pressures: G1 = 365 kPa
(53 psi), and G2 = 20.7 kPa (3 psi), and DP2-1 = 492 Pa (1.97 in-wg)
11:41 AM. Regulator position increased by another ½ turn. Gage pressures: G1 = 379 kPa
(55 psi), and G2 = 34 kPa (5 psi), and DP2-1 = 630 Pa (2.52 in-wg)
11:42 AM. Regulator position increased by ½ turn. Gage pressures: G1 = 379 kPa (49 psi),
and G2 = 28 kPa (4 psi), and, DP2-1 = 808 Pa (3.23 in-wg)
11:48 AM. Significant leakage of nitrogen through door gaskets was detected. The nitrogen
control valves were shut off. It took 20 sec. to depressurize the chamber.

Figure 7 shows barometric pressure variations (actual, not standardized), and pressure differ­
entials (chamber-entry and gob-chamber) recorded during this test. Notice that when the gas
flow rate increased, the pressure differentials across the seal and stopping increased rapidly
and remained constant thereafter, indicating that both walls can hold pressurized nitrogen
with reduced leakage. Table 1 summarized the differential pressures across the chamber walls.

417
Figure 7. Pressure differentials across the chamber stopping and seal – Test 1.

Table 1. Pressure differentials across the chamber stopping and seal – Test 1.

Chamber-Entry Gage Average DP2-1 Chamber-Gob Gage Average DP2-3

Pressure (DP2-1) Pa in wg Pressure (DP2-3) Pa in wg

Δ1 42 0.17 Δ6 –51 –0.20

Δ2 241 0.97 Δ7 –229 –0.92
Δ3 471 1.89 Δ8 –467 –1.88
Δ4 628 2.52 Δ9 –624 –2.51
Δ5 804 3.23 Δ10 –805 –3.23

4.2.2 Test 2- August 10, 2022

Following a visual inspection of the chamber and the nitrogen gas injection system, the per­
sonnel doors were closed and locked, and the 2nd test started on August 10. As in the previous
case, the chamber was pressurized in steps by changing the regulator setting and stopped
when significant nitrogen leakage through the stopping was detected. The highlights of these
tests are presented below.
Chronology of Events – August 10, 2022
12:15 PM. For regulator R1 set to minimum flow, valves V1 and V2 were open. Pressure
gages: G1 =372 kPa (54 psi), and G2 = 0. and DP2-1 = 33 Pa (0.13 in-wg)
12:17 PM. Regulator position was increased by a full turn. Pressure gages recorded: G1 = 359
kPa (52 psi), and G2 = 21 kPa (3 psi), and DP2-1 = 645 Pa (2.58 in-wg
12:22 PM. Regulator position increased by ½ turn. Pressure gage p recorded: G1 = 359 kPa
(52 psi), and G2 = 21 kPa (3 psi), and DP2-1 = 1218 Pa (4.87 in-wg)
12:27 PM. Regulator position was increased by ½ turn. Pressure gages recorded: G1 = 345
kPa (50 psi), and G2 = 21 kPa (3 psi), and DP2-1 = 1560 Pa (6.24 in-wg)
12:29 PM. Significant leakage of nitrogen through door gaskets was detected. The nitrogen
control valves were shut off. It took 45 sec. to depressurize the chamber.

418
 Figure 8 shows barometric pressures and pressure differentials across the stopping and seal
during the test. Notice a slight decline when the differential pressure across the stopping
reached 6.41 in w.g. Table 2 shows the differential pressures across the chamber walls.

Figure 8. Pressure differentials across the chamber stopping and seal – Test 2.

Table 2. Pressure differentials across the chamber stopping and seal – Test 2.

Chamber-Entry Gage Average DP2-1 Chamber-Gob Gage Average DP2-3

Pressure (DP2-1) Pa in wg Pressure (DP2-3) Pa in wg

Δ1 33 0.13 Δ5 –26 –0.10

Δ2 641 2.58 Δ6 –643 –2.58
Δ3 1212 4.87 Δ7 –1200 –4.82
Δ4 1600 6.42 Δ8 –1586 –6.37

The above results show that the balance chamber will hold pressurized nitrogen as long as
a steady flow of nitrogen gas is supplied to the chamber. Because of leakage though the per­
sonnel doors and fractured coal around the periphery of the stopping, the chamber was not
able to hold pressurized gas for a long time. In test 1, when the flow control valves were shut
off, it took only 20 seconds to depressurize the chamber. When the leakage points were fixed
for the second test, the chamber took approximately 45 seconds to depressurize. To improve
the quality of the stopping and to reduce leakage, it is recommended to reverse the position of
the access doors so that the door gaskets are compressed when pressurizing the chamber.

5 CONCLUSIONS AND DISCUSSIONS

Two pressure chambers were constructed at the MCC’s West Elk Mine. Each chamber was
established by installing a Kennedy stopping at about 3 m (10 ft) in front of an existing seal
equipped with a nitrogen injection system and a set of monitors linked to the mine’s AMS.

419
 Several pressure balancing tests were conducted on each chamber. Each test started by
injecting pressurized nitrogen into the chamber and monitoring for changes in barometric
pressure, along with the pressure differentials across the chamber’s seal and stopping. Signifi­
cant leakage of nitrogen through the Kennedy stopping was detected.
Changes in barometric pressure in an underground coal mine can cause “in-gassing”, or
“out-gassing” conditions at the permanent seals. At the West Elk Mine, the barometric pres­
sure varies daily as the air temperature on the surface increases or decreases. In this mine, on
average, this daily change is in the order of 2.5 in w.g.
When a fixed rate of nitrogen was injected into the 1st chamber, the pressure differentials
across the seal increased from -0.4 in. w.g. to +3.48 in. w.g. No major gage pressure changes
across the stopping were observed. The results showed that while the permanent seal is doing
the intended work, the Kennedy stopping still needs additional work to reduce leakage.
Once established and equipped, the 2nd chamber was tested for different nitrogen flow
rates. The results showed that the balance chamber will hold pressurized nitrogen as long as
a steady flow of nitrogen is supplied. Under the existing conditions, when the nitrogen flow to
the chamber is stopped, the chamber will depressurize in less than one minute.

ACKNOWLEDGMENTS

The authors would like to acknowledge the Alpha Foundation for providing financial support
for this project. The authors would also like to express their gratitude to Mr. John Poulos,
Mining Consultant at Geosyntec Inc., Jessica Wilczek, Manager of Engineering and Environ­
mental Affairs at MCC, Elizabeth Lindgren, Ventilation Engineer at MCC, Rick Chermak,
Project Construction Manager at MCC, and Garren Atchley, Mine Engineer at MCC for
their support in developing this project.

DISCLOSURE

This study was sponsored by the Alpha Foundation for the Improvement of Mine Safety and
Health, Inc. The views, opinions, and recommendations expressed herein are solely those of
the authors and do not imply any endorsement by the ALPHA FOUNDATION, its Direct­
ors, and its staff.

REFERENCES

Bessinger, S.L., Abrahamse, J.M., Bahe, K.A., McCluskey, J.E. & Palm, A.T. 2005. Nitrogen inertization
at San Juan Coal Company’s longwall operation. SME Annual Meeting, Salt Lake City, UT.
Bhowmick, B.C., et al. 1992. Dynamic Balancing of Pressure for Control of Fire in Underground Coal
Mines. Journal of Mining Research. 1:1. New Delhi.
Calizaya F., Nelson M. G., Bateman, C., and Jha A. 2016. Pressure Balancing Techniques to Control
Spontaneous combustion. SME Annual Meeting, Preprint 16–050. Phoenix, AZ.
Chalmers, D.R. 2008. Sealing design. Proceedings of the 12th North American/US Mine Ventilation
Symposium, Reno, NV: University of Nevada. pp. 219–223.
Francart, W.J. & Beiter, B.A. 1997. Barometric Pressure Influence in Mine Fire Sealing. Proceedings of
the 6th International Mine Ventilation Congress. Littleton, CO:S ME. pp. 341–342.
Grubb, J.W., 2008. Preventative Measures of Spontaneous Combustion in Underground Coal Mines.
Ph.D. dissertation, Colorado School of Mines, Golden, CO.
Moreby, R. 2009. Ventilation. Australasian Coal Mining Practice. Carlton, Victoria: Australasian Insti­
tute of Mining and Metallurgy.
Ray, S.K. & Singh, R.P. 2007. Recent Developments and Practices to Control Fire in Underground Coal
Mines. Fire Technology (Central Mining Research Institute, Dhanbad, India). 43: 285–300.
Schatzel, S.J., Krog, R.B. & Dougherty H. 2012.Methane Emissions and Airflow patterns on a Longwall
Face. SME Annual Meeting, Preprint 12–016. Seattle, WA.
Smith, A.C. & Lazzara, C.P. 1987. Spontaneous Combustion Studies of U.S. Coals, USBM RI 9079.

420
Mine gases
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Underground coal methane gas forecasting using multivariate time

series with one and two auxiliary variables

J.C. Diaz, Z. Agioutantis & S. Schafrik

University of Kentucky, Lexington, Kentucky, USA

D.T. Hristopulos
Technical University of Crete, Chania, Crete, Greece

ABSTRACT: Prediction of methane gas concentrations in underground coal mines has been
a topic of interest for academia and the mining industry for decades as methane gas is con­
sidered a significant hazard to mining safety and productivity. This study utilizes multivariate
time series forecasting approaches to predict methane emissions in an active mine. The multi­
variate method incorporates the past values of methane gas concentrations as a predictor of
its future values but also integrates the correlations between gas concentration and barometric
pressure as well as barometric pressure and coal production rate in predictive models of
methane gas emissions. The data used in this study were collected from the atmospheric moni­
toring systems of an underground coal mine operation in the eastern USA.
The data were stored, pre-processed, and statistically analyzed to develop multivariate fore­
casting models based on the Vector Autoregression (VAR) and Autoregressive Integrated
Moving Average with Exogenous Input (ARIMAX) methods. The optimal model for each
method was selected based on the Akaike Information Criterion. The forecasting performance
and accuracy of each model were assessed using cross-validation to establish the best overall
model. It was concluded that both VAR and ARIMAX could, in most cases, adequately fore­
cast methane gas concentrations in underground coal mines.

1 INTRODUCTION

The number of fatalities due to methane gas explosions in underground coal mines has signifi­
cantly decreased in recent decades, mainly due to the extraordinary effort made by different
organizations such as the mining industry, governmental agencies, and academia to develop
new methods and techniques to monitor and more efficiently control methane gas emissions
and concentrations in underground coal mine operations (Taylor and Karacan, 2010; Diaz
et al., 2022). However, methane gas explosions still represent the most feared threat in the coal
mining industry worldwide. For example, in 2009, more than 750 coal miners lost their lives in
over 150 methane gas explosions in China (Cheng, 2018). Furthermore, in the U.S. coal mining
industry, methane gas explosions are also considered the most dangerous hazard; since 1900,
more than 11,600 underground coal mine workers have died in more than 500 mining disasters
(NIOSH, 2022). Other examples include the catastrophe at the Soma Coal Mine (Turkey) in
2014, which killed 300 underground coal miners and has been documented as the worst disaster
in the recent history of the mining industry (Düzgün and Leveson, 2018).
Methane gas disasters have been the primary motivation for exploring different methodolo­
gies that could accurately predict methane gas concentrations in underground coal mine oper­
ations. As a result, different approaches have been implemented to tackle this problem. For
example, these include empirical, numerical, and statistical techniques. Indeed, a statistical
approach based on time series analysis has recently been employed to develop different

DOI: 10.1201/9781003429241-44

423
forecast models capable for methane gas concentrations in underground coal mines using uni­
variate and multivariate time series data (Diaz et al., 2021a, b; 2022).
Past research has demonstrated that methane gas concentrations are autocorrelated and
that the past values of methane gas time series can be used to predict its future values allowing
the implementation of univariate time series analyses (Dixon, 1992; Dixon and Longson,
1993). On the other hand, other studies have established that methane gas concentrations cor­
relate with barometric pressure (BP) and coal production rate, permitting the use of multivari­
ate time series analysis (Wasilewski, 2014; Hemp, 1994; Xu et al., 2014). For example,
methane gas and coal production rate are positively correlated, while methane gas and baro­
metric pressure are negatively correlated (Diaz et al., 2021a, b; 2022).
The research presented in this paper explains the main steps for developing different multivari­
ate forecasting methods to predict methane emissions at an active mine. The multivariate method
incorporates the past values of methane gas concentrations as a predictor of its future values but
also integrates barometric pressure as well as barometric pressure and coal production rate as
predictors of methane gas concentration. The mine data (methane gas and coal production rate)
were collected by the atmospheric monitoring system (AMS) of an active, underground coal
mine in the eastern US. The weather data (barometric pressure) were retrieved from the closest
public weather station to the case study. Furthermore, various statistical techniques were imple­
mented to assess the association (cross-correlation) between the methane gas concentration and
the exogenous variables. Such associations were employed to develop two multivariate forecast­
ing approaches. Finally, the optimal model for each approach was selected using the Akaike
Information Criterion (AIC). The results obtained from the two forecasting approaches were
compared using cross-validation metrics to determine the best model overall.

2 MATERIALS AND METHODS

2.1 Data gathering

Three different time series data were employed to conduct the research presented in this
paper. The first dataset consists of methane gas concentrations collected from the atmospheric
monitoring system of an active underground coal mine that is designated as Mine A. ­
The second dataset comprises daily coal production values (tons/day) provided by the mine
personnel. Finally, the third dataset contains barometric pressure values gathered from the
closest to the mine site public weather station (Weather Underground).
The data frequency is not the same for all three time series. For example, the methane gas
time series and barometric pressure time series are recorded approximately every hour (hourly
frequency). On the other hand, the frequency of coal production rate data is daily. Therefore,
data homogenization was used to ensure that the different datasets (time series) share
a standard common (daily) time stamp.
The process of data homogenization and pre-processing (e.g., identification of missing
values, faulty records, and outliers) was performed using an Atmospheric Monitoring Ana­
lysis and Database mAnagement (AMANDA) system, a custom relational database that was
created to manage AMS data. Detailed information concerning the characteristics (e.g.,
source, frequency, and units) of the data collected, the pre-processing stage, and AMANDA
have already been discussed elsewhere (Diaz et al., 2021a, b; 2022).

2.2 Forecasting methodologies

A time series can be defined as a set of observations Xt, where each observation is recorded
during a specific time t over a set of discrete time points. Time series analysis consists of statis­
tical techniques and methods to study a sequence of data points arranged in chronological
order. The main objective of the analysis is to fill gaps in the time series and forecast future
values (Box et al., 2015; Brockwell and Davis, 2016; Diaz et al., 2022).
Forecasting methods can be classified into two main categories: qualitative and quantita­
tive. Qualitative forecasting methods, also known as judgmental forecasting techniques, are

424
based on someone’s intuition, knowledge, and experience. Quantitative forecasting methods,
on the other hand, are based on mathematical modeling and are the focus of the study pre­
sented in this paper (Shumway and Stoffer, 2017). Univariate and multivariate forecasting
methods are the most common quantitative forecasting models employed.
Univariate techniques estimate the future values of a process utilizing past values of the pro­
cess (time series). In contrast, multivariate forecasting methods incorporate more than one
correlated time series and utilize past values from all the variables that are mutually analyzed.
In addition, multivariate methods offer a more comprehensive understanding of the studied
process and, in most cases, higher accuracy than univariate forecasting methods (Chatfield,
2004; Shumway and Stoffer, 2017). Therefore, multivariate forecasting approaches with one
and two auxiliary variables are employed herein to forecast methane gas concentrations.

2.3 Multivariate forecasting approach

The two multivariate forecasting approaches used to predict methane gas concentrations from
Mine A are discussed in this section. They involve (i) the autoregressive integrated moving
average with exogenous input, ARIMAX(p,d,q), and (ii) the vector autoregressive, VAR(p),
methods. Both methods were implemented using the programming and numeric computing
platform, MATLAB®, using functions from the econometrics toolbox.
The cross-validation approach (Cerqueira et al., 2020) was used to assess the forecasting
models’ performance. The datasets are split into two disjoint sets: the “training set” and the
“validation set.” The training set contains 95% of the time series records and is used to esti­
mate the model parameters. The validation set consists of the remaining 5% of the data and
provides the ground truth against which the model forecasts are compared.
The best possible model for each approach is established using the Akaike Information Cri­
terion, which assesses the model’s fit on the training data and includes a penalty term for the
complexity of the model. The optimal model is the one that returns the lowest AIC value
(Shumway and Stoffer, 2017). The optimal model is then employed to forecast methane con­
centration one step ahead (i.e., for the next point in time). The probabilistic forecasts include
an estimate of the 95% prediction intervals (C.I.), which are essential for two reasons: (i) they
permit to verify if the true values are at least contained within the prediction intervals, and (ii)
they allow assessing the precision of the forecast (Diaz et al., 2022; Hristopulos, 2020).
The multivariate time series models are applied in one-step-ahead forecast mode. In this
approach, the model is first estimated based on the data in the training set. Then, in the fore­
casting stage, the data used to predict the time series are continuously updated by incorporat­
ing in the sample the most recent value of the time series. The value of the process at the next
point in time is then forecasted using the augmented sample. The performance and accuracy
of the ARIMAX(p,d,q) and VAR(p) forecasting models are assessed by means of cross-
validation statistics which measure the agreements between the values in the validation set and
the respective forecasts. The cross-validation statistics used herein include the linear (Pear­
son’s) correlation coefficient (R), the mean error (ME), the root mean squared error (RMSE),
and the mean absolute error (MAE).
2.3.1 The ARIMAX(p,d,q) model
The Autoregressive Integrated Moving Average with Explanatory Variable(s) model is an
extended version of the univariate ARIMA model, which uses one or more time series as inde­
pendent predictors to forecast the dependent variable. ARIMAX models are composed of
four main terms: (i) the autoregressive (AR) term (p), which includes lagged values of the
dependent variable (ii) the moving average (MA) term (q), which involves lagged innovation
terms (residuals) representing stochastic factors, (iii) the order of differencing (d) which is used
to render a time series stationary (usually, d = 0, 1 is adequate but for certain datasets, d ≥ 2
may be necessary), and (iv) the exogenous term (X), which incorporates dependence on one or
more exogenous variables (Andrews et al., 2013).
In this research, the ARIMAX(p,d,q) model is utilized to take advantage of the positive cor­
relation between methane gas and coal production rate and the negative correlation between

425
methane gas and barometric pressure in order to predict methane gas concentrations. Further­
more, The ARIMAX(p,d,q) model is employed herein using d = q = 0 in order to compare its
performance and accuracy with the VAR(p) model. The autoregressive order p* of the optimal
ARIMAX model is selected based on the lowest AIC.

2.3.2 The VAR(p) model

The Vector Autoregressive model (VAR) treats a collection of time series as a vector entity. This
allows the incorporation of correlations between different time series in the model. In contrast to
ARIMAX, which separates the target (dependent) variable from the exogenous (auxiliary) vari­
ables, in VAR, all the variables are treated on an equal footing. As a result, VAR(p) models can
be represented as systems of equations containing as many equations as variables. VAR(p)
models are widely used due to their ability to capture the interconnected dynamics between
cross-correlated datasets (time series) (Lütkepohl, 2005; Tsay, 2015). More information about
the VAR(p) model is given in (Lütkepohl, 2005; Kirchgässner and Wolters, 2007; Tsay, 2015).

3 RESULTS AND DISCUSSION

This section introduces the results of methane gas concentration forecasting using the ARIMAX
(p, d, q) and the VAR(p) models. Four datasets (from 1 to 4) with daily averaged time series were
employed to run both models. The time series in datasets 1 and 3 correspond to three months
(from October 2019 to December 2019), while those in datasets 2 and 4 correspond to six months
(from October 2019 to March 2020). Datasets 1 and 2 consist of two time series: (i) the methane
concentration time series (dependent variable) and (ii) the barometric pressure time series (inde­
pendent variable). Datasets 3 and 4, on the other hand, include three time series: (i) the methane
concentration time series, (ii) the barometric pressure time series, and (iii) the coal production
rate time series. The last two time series are considered auxiliary predictors.
Figures 1 and 2 consist of four plots (a to d). Graphs (a) and (b) are the forecasts obtained
using the multivariate (VAR and ARIMAX) one-step-ahead approaches with one auxiliary
(barometric pressure) variable, while plots (c) and (d) are the forecasts developed using two
auxiliary (barometric pressure and coal production) variables, respectively. Additionally, the
blue line represents the training data (methane gas time series), the upper and lower black
dashed lines signify the 95% prediction confidence interval, and the red line represents the
forecast. The validation and forecast periods contain five percent (5%) of the total points
(data records) used in the training time series.
Figure 1 presents the forecasting results using the ARIMAX(p,d,q) model. Figure 1a shows the
forecast for dataset 1; the optimal autoregressive order is p = 4 based on the lowest AIC. Visual
inspection demonstrates that the forecast (red line) is close to the actual values in the validation
period (blue line). Furthermore, the correlation coefficient (henceforth, R) is calculated as R = 0.78,
implying a strong correlation between the validation data and the forecast. Similarly, Figure 1b dis­
plays Forecast 2; the best possible autoregressive order is p = 9 according to the AIC and R = 0.67,
indicating similar variations between the validation data and the forecast. Likewise, Figures 1c and
1d show forecasts for the gas concentration time series over the same periods presented in Figures 1a
and 1b, respectively, using two auxiliary (independent) variables. Figure 1c shows that the optimal
autoregressive order is p = 1 based on the lowest AIC. In this case, the forecast (red line) is closer to
the actual values in the validation period (blue line), with a higher correlation coefficient R = 0.87
than the one obtained in Figure 1a (see Table 1). Similarly, in Figure 1c, the best possible autoregres­
sive order is p = 2 and R = 0.76, also higher than the one in Figure 1b. Finally, in all cases, the
actual values and the forecasts are captured by the 95% prediction confidence interval.
Figure 2 introduces the forecasting results using the VAR(p) model. Figures 1a and 1b show the
forecast for datasets 1 and 2, respectively. The best model in Figure 1a has p = 2. As a result, the
correlation coefficient between the forecasts and the validation data is R = 0.63, which is lower than
the value obtained with the ARIMAX model with one exogenous variable (Figure 1a). Similarly, in
Figure 1b, the best model is obtained for p = 2. The correlation coefficient is R = 0.64, slightly lower
than the one shown in Figure 1b. Furthermore, Figures 1c and 1d present the forecast for datasets 3

426
and 4 correspondingly. In both cases, the best model is obtained for p = 2 according to the lowest
AIC. The correlation coefficients are R = 0.66 and R = 0.65, respectively, which are lower than
those shown in Figures 1c and 1d, as presented in Table 2.
Tables 1 and 2 demonstrate that the multivariate ARIMAX(p,d,q) and VAR(p) forecasting
methods with one and two auxiliary variables can predict methane gas concentrations. In
most cases, the methane gas concentration forecasted by both models matches the direction
(increase/decrease) of the validation data. Furthermore, the observed methane gas levels are
continuously captured by the 95% prediction intervals, as illustrated in Figures 1 and 2. In
addition, the CV correlation between the forecasts and the validation data is strong and posi­
tive, and the values of other CV measures are similar (see Tables 1 and 2).
The results obtained herein show that the forecasting model is improved (a higher correl­
ation between the forecast and the validation data is achieved) by using two auxiliary (baro­
metric pressure and coal production rate) variables than one (barometric) pressure auxiliary
variable. Furthermore, in all cases presented in Figure 2 (VAR model), the results were infer­
ior to those shown in Figure 1 (ARIMAX model).

Figure 1. ARIMAX(p,d,q) one-step-ahead CH4 concentration forecasts using one and two exogenous vari­
ables: (a) forecast for dataset 1; (b) forecast for dataset 2; (c) forecast for dataset 3; (d) forecast for dataset 4.

4 CONCLUSIONS

Implementing barometric pressure time series and coal production time series as auxiliary pre­
dictors provided better methane gas concentration forecasts than those obtained with only
one auxiliary variable (i.e., barometric pressure).
The ARIMAX(p,d,q) one-step-ahead forecast model offered better cross-validation per­
formance (higher correlation between the forecast and the validation data) than the ones

427
obtained using the VAR (p) one-step-ahead forecast. On the other hand, both approaches
gave similar results for the other cross-validation measures (i.e., RMSE, MAE, and ME). The
performance of both forecasting approaches should be further assessed with different datasets
and operating conditions.
The ARIMAX and VAR forecast methods proposed in this research provide a potential
solution to the critical problem of forecasting methane gas concentrations in underground
coal mines. Accurate forecasts of methane concentration can lead to a warning system that
will improve the safety conditions of underground coal miners worldwide.

Table 1. Summary of methane gas forecasting results for the ARIMAX(p,d,q) one-step-ahead model
(using one exogenous (barometric pressure) and two exogenous (barometric pressure and coal production
rate) time series as independent variables). p*: Optimal autoregressive order of ARIMAX model based
on the lowest AIC.
Data Length (days)
Number of Auxiliary Training Validation
Variables Data Data p* R RMSE ME MAE

Forecast 1 One (BP) 90 5 4 0.78 0.16 0.05 0.15

Forecast 2 One (BP) 180 9 9 0.67 0.13 0.07 0.10
Forecast 3 Two (BP and Coal 90 5 1 0.87 0.14 -0.11 0.13
Prod.)
Forecast 4 Two (BP and Coal 180 9 2 0.76 0.09 0.05 0.08
Prod.)

Figure 2. VAR(p) one-step-ahead CH4 concentration forecasts using one and two exogenous variables:
(a) forecast for dataset 1; (b) forecast for dataset 2; (c) forecast for dataset 3; (d) forecast for dataset 4.

428
Table 2. Summary of methane gas forecasting results for the VAR(p) one-step-ahead model (using
a multivariate vector of methane gas concentration, barometric pressure, and coal production rate).
p*: Optimal autoregressive order of VAR model based on the lowest AIC.
Data Length (days)
Number of Exogenous Training Validation
Variables Data Data p* R RMSE ME MAE

Forecast 1 One (BP) 90 5 2 0.63 0.16 0.09 0.15

Forecast 2 One (BP) 180 9 2 0.64 0.13 0.10 0.10
Forecast 3 Two (BP and Coal Prod.) 90 5 2 0.66 0.14 0.07 0.13
Forecast 4 Two (BP and Coal Prod.) 180 9 2 0.65 0.12 0.09 0.09

ACKNOWLEDGMENT
This study was sponsored by the Alpha Foundation for the Improvement of Mine Safety and
Health, Inc. (ALPHA FOUNDATION). The views, opinions, and recommendations
expressed herein are solely those of the authors and do not imply any endorsement by the
ALPHA FOUNDATION, its Directors, and staff.

REFERENCES

Andrews, B.H. Dean, M.D. Swain, R. Cole, C. 2013. Building ARIMA and ARIMAX models for pre­
dicting long-term disability benefit application rates in the public/private sectors. Society of Actuaries.
University of Southern Maine, Maine, USA.
Box, G.E. Jenkins, G.M. Reinsel G.C. Ljung, G.M. 2015. Time Series Analysis: Forecasting and Control,
5th edn. John Wiley & Sons, Hoboken, NJ, USA.
Brockwell, P. J. and Davis, R. A. 2016 Introduction to Time Series and Forecasting. 3rd edition,
Springer. https://doi.org/10.2307/2965440.
Cerqueira, V. Torgo, L. Mozetič, I. 2020. Evaluating time series forecasting models: an empirical study
on performance estimation methods. Machine Learning. https://doi.org/10.1007/s10994-020-05910-7.
Chatfield, C. 2004. The Analysis of Time Series - An Introduction. 6th Edition. Chapman & Hall/CRC,
New York, USA.
Cheng, J. 2018. Explosions in Underground Coal Mines - Risk Assessment and Control. Springer.
https://doi.org/10.1007/978-3-319-74893-1.
Diaz, J. Agioutantis, Z. Hristopulos, D.T. Schafrik, S. 2021a. Managing and utilizing big data in atmos­
pheric monitoring systems for underground coal mines. Materials Proceedings 5(1):78. https://doi.org/
10.3390/materproc2021005078.
Diaz, J. Agioutantis, Z. Hristopulos, D.T. Schafrik, S. 2021b. Towards atmospheric monitoring data
analysis in underground coal mines. In: Tukkaraja P (ed) Proceedings of the 18th North American
Mine Ventilation Symposium (NAMVS 2021), June 21–23. CRC Press, Boca Raton, FL, pp 498–506,
https://doi.org/10.1201/9781003188476-51.
Diaz, J. Agioutantis, Z. Hristopulos, D.T. Schafrik, S. Luxbacher, K. 2022. Time Series Modeling of
Methane Gas in Underground Mines. Mining, Metallurgy & Exploration. Springer, https://doi.org/
10.1007/s42461-022-00654-5.
Dixon, D.W. 1992. A statistical analysis of monitored data for methane prediction. PhD thesis, Univer­
sity of Nottingham, URL http://eprints.nottingham.ac.uk/12977/1/334922.pdf
Dixon, D.W. Longson, I. 1993. A statistical method for methane prediction and improved environmental
control. In: Proceedings of the 6th US Mine Ventilation Symposium, Salt Lake City, UT, USA.
Düzgün, H.S. Leveson, N. 2018. Analysis of Soma mine disaster using causal analysis based on systems
theory (CAST). Saf. Sci. Volume 110, Part A, December 2018, 37–57. https://doi.org/10.1016/j.
ssci.2018.07.028.
Hemp, R. 1994. The effect of changes in barometric pressure on mines in the highveld of South Africa.
Journal of the Southern African Institute of Mining and Metallurgy 94(6):133–146.
Hristopulos D.T. 2020. Random Fields for Spatial Data Modeling: A Primer for Scientists and Engin­
eers. Springer, Netherlands, https://doi.org/10.1007/978-94-024-1918-4
Kirchgässner G, Wolters J. 2007. Introduction to Modern Time Series Analysis. Springer, Berlin,
Germany.

429
Lütkepohl, H. 2005. New Introduction to Multiple Time Series Analysis. Edited by Springer Science &
Business Media. Berlin, Germany.
Shumway, R.H. Stoffer, D.S. 2017. Time series analysis and its applications, 4th edn. Springer.
Switzerland.
Taylor, C. Karacan, C.Ö. 2010. Historical development of technologies for controlling methane in under­
ground coal mines. In: Extracting the Science: A century of mining research. Brune JF (ed), Littleton,
CO, Pittsburgh, USA.
NIOSH. 2022. NIOSH data and statistics. https://wwwn.cdc.gov/NIOSH-Mining/MMWC/MineDisa
sters/Count. The National Institute for Occupational Safety and Health Accessed August 02, 2022.
Tsay, R.S. 2015. Multivariate Time Series Analysis with R and Financial Applications. First Ed. New
Jersey, USA: John Wiley & Sons, Inc.
Wasilewski, S. 2014. Influence of Barometric Pressure Changes on Ventilation Conditions in Deep
Mines. Archives of Mining Sciences, 59(3), pp. 621–639. https://doi.org/10.2478/amsc-2014-0044.
Xu, L. Lin, X. Amen, J. Welding, K. McDermitt, D. 2014. Impact of changes in barometric pressure on
landfill methane emission. Global Biogeochemical Cycles, 28(7), pp. 679–695. https://doi.org/10.1002/
2013GB004571.

430
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Complete degasification of longwall panels in U.S. coal mines

P.C. Thakur
West Virginia University, Morgantown, WV, USA

ABSTRACT: Longwall mining is the safest and most economical way to mine coal in under­
ground mines. At present 37 longwall panels in 11 states are mining more than 50% of under­
ground mined coal. Proper degasification of the panels prior to mining and post mining
promotes mine health and safety and improves productivity. Produced gas can be sold for
added profit. For proper degasification, all coal seams are divided as shallow, medium-depth
and deep coal seams. The degasification scheme for a typical mine in each category is dis­
cussed in detail. Special attention is given to floor gas capture that has led to some recent coal
mine disasters.

1 INTRODUCTION

1.1 Background
Majority of global coal production comes from underground mining which is done by two
methods: (1) room and pillar and (2) longwall mining. As shallow coal seams are mined out
and mining reaches greater depths, longwall mining becomes the preferred method of mining.
In United States, more than 50% of all underground mined coal is mined by the longwall
method. Table 1 shows the distribution of 37 longwall panels in U.S. coal mines.

Table 1. Longwall installations by parent company (2021-2022).

Company AL CO IL MT NM OH PA UT VA WV WY Total

Alliance Resource Partners 1 2 3

ACNR 1 1 5 7
Arch Resources 1 2 3
Blue Mountain Energy 1 1
CONSOL Energy 5 5
Iron Senergy 1 1
Coronado Coal 1 1
Foresight Energy (ACNR) 4 4
Hatfield Met (ACNR) 1 1
Peabody Energy 1 1 2
Signal Peak Energy 1 1
Solvay Chemicals 1 1
Genesis Alkali 1 1
Warrior Met Coal 3 3
Westmoreland Mining 1 1
Wolverine Fuels 2 2
Total 5 3 5 1 1 1 6 3 1 9 2 37

Reference: Coal Age, January-February 2022. (www.coalage.com)

DOI: 10.1201/9781003429241-45

431
 The trend for panel sizes and mining equipment in the coal industry is to continue to go
upward pushing production capacities and productivity to higher levels. Another reason for
increasing the longwall panel size is that development sections cannot keep up with the rate of
longwall advance. Increasing the width of the longwall panel slows down the rate of advance.
Today it is quite realistic to plan longwall panels in mildly gassy coal seams that are 1500 ft
wide and 15,000 ft long containing nearly 5 to 7 million tons of raw coal. Such large longwall
panels offer many benefits as follows:
• Improved safety and reduced injury rate
• Improved recovery of in situ coal reserve
• Improved productivity and reduced cost/ton
On the other hand, large longwall panels cause some concerns, such as follows:
• Ventilation and methane control
• Respirable dust control
• Increased likelihood of geological anomalies, e.g., faults, washouts, etc.
These problems become more serious when the gas content of coal seams increases with the
increase in mining depths. Costs of mine ventilation and coal seam degasification go up, and
it becomes necessary to optimize the two processes to keep the mine environment safe and the
combined cost at the minimum. Drilling process for degasification can also locate any geo­
logical anomalies in the panel prior to mining and helps in respirable dust control.
The optimized plan for mine degasification and ventilation with added benefit of coal seam
exploration can be summarized as follows:
1. Recover as much coal bed methane (CBM) as possible prior to mining and postmining to
minimize ventilation requirements.
2. Optimize longwall panel sizes and the size of the district depending on the gas content of
the coal seams to keep the combined cost of degasification and ventilation down.
3. Plan mine ventilation to maximize ventilation efficiency measured as the ratio of air horse­
power delivered to face to the total air horsepower created at the fans.
4. Seal all mined out areas with approved design of stoppings and recover as much CBM as
possible without making the atmosphere in the sealed area explosive.
5. Measure coal seam thickness while drilling and locate geological anomalies if any.
6. Gather all produced gases and market them after processing to defray the cost of degasifi­
cation and ventilation.

2 SOURCES OF GASES IN LONGWALL MINING

In the development phase of mining, the only source of gas is the coal seam being mined.
Ideally, gas content should be reduced to 100 ft3/t for safe and efficient mining. This is
achieved by pre-mining methane drainage as discussed later in the paper.
Most of gas in longwall mining comes from the mined-out area, called ‘gob’. Longwall
mining subsides the overlying coal seams and makes the underlying coal seams heave. Various
estimates of the area called ‘gas emission space’ is shown in Figure 1.
The vertical dimension of the gas emission space can range up to 1000 feet above the mined
coal seam and about 250 feet in the floor.
Factors which govern the rate of gas emission in the longwall gobs can be summarized as
follows:
1. The density and proximity of gas-bearing zones in the gas emission space.
2. The gas content of coal seams and other gas-bearing horizons.
3. The extent to which overlying and underlying coal seams have been mined.
4. The rate of coal face advance. This is the most important factor.
5. The presence of geological disturbances and leftover coal pillars.
6. The subsidence characteristics of the area which determines the permeability of the gas
emission space.

432
Figure 1. Vertical limits of the gas emission space.

For efficient degasification, the following reservoir properties must be measured:

1. Gas content of coal seams.
2. Permeability and porosity.
3. Reservoir pressure.
4. Sorption time.
Detailed discussion of these parameters is available in literature (Thakur 2016, Thakur
2018). In addition, the following parameters are the most important for planning a successful
mine degasification.

2.1 Degree of gassiness

It is safe to assume that all coal seams are gassy. They only vary in their degree of gassiness,
i.e., gas contained per ton of coal. For planning, all coal seams can be classified as shown in
Table 2.

Table 2. Gassiness of coal seams.

Category Gas Content (ft3/t) Depth (ft)

Mildly gassy <100 <600

Moderately gassy 100 – 300 600 – 1500
Highly gassy 300 – 700 1500 – 3000

The depth of a coal seam is only a rough indication of its gas content. Direct measurement
of gas contents is highly recommended.

2.2 Specific gas production

The second reservoir parameter that is equally important is the “specific gas production
rate” of the coal seam. It is a measure of how much gas will be produced per day if an
opening (a horizontal borehole or a vertical slit) is created in the coal seam by a given dega­
sification technique. It is measured in MCFD/100 ft (e.g., 15 MCFD/100 ft for Pittsburgh

433
coal seam). This parameter is a combination of reservoir pressure, permeability, porosity,
gas content, and diffusivity of the coal seam. Table 3 shows specific gas production for some
U.S. coal seams.

Table 3. Specific gas emissions for coal seams.

Coal Seam Depth (ft) Rank Specific Gas Production (MCFD/100 ft)

Pittsburgh 500 – 1000 High Vol. Bituminous 15.00

Pocahontas No. 3 1400 – 2000 Low Vol. Bituminous 8.00
Blue Creek/Mary Lee 1400 – 2000 Low Vol. Bituminous 9.00
Pocahontas No. 4 800 – 1200 Medium Vol. Bituminous 5.00
Sunnyside 1400 – 2000 High Vol. Bituminous 9.00

2.3 Specific gob gas production

The third reservoir parameter of importance is the “specific gob gas production” for the
mine which is a measure of total gas production per unit area of the gob and is measured in
MMCF/acre. It represents the total influx of gases into the gob from all overlying and
underlying coal seams that are disturbed by longwall mining. For moderately and highly
gassy mines in the United States, the specific gob gas production ranges from 8 to 30
MMCF/acre.

3 MILDLY GASSY COAL SEAMS

These coal seams are shallow in depth and rarely have thick coal seams overlying or under­
lying the coal seam being mined.

3.1 Pre-mining degasification

Coal seams with less than 100 ft3/ton of gas content usually do not need any pre-mining degasi­
fication. Shallow mines in Ohio and Illinois are mining without any pre-mining degasification.
Horizontal drilling of longwall panels is recommended for discovering any geologic anom­
aly within the panel.

3.2 Post mining degasification

Shallow coal mines usually do not have thick coal seams above or below the coal seam being
mined. The gob gas emission is therefore low, and in most cases, gas can be safely diluted by
ventilation air. It is, however, desirable to install one gob well within 1000 ft of the set-up
entry and another one in the middle of the panel to minimize the load on ventilation system.
Construction of gob wells is described in literature (Thakur 1981).

4 MODERATELY GASSY COAL SEAMS

As shown in Table 2, these coal seams have a depth of 600 – 1500 ft and a gas content of
100 – 300 ft3/ton. They need to be degassed pre-mining and post-mining. Such mines are
located in the Northern Appalachian Basin of Pennsylvania and West Virginia.

4.1 Pre-mining degasification

Most coal seams in this category need pre-mining degasification. Figure 2 shows
a degasification scheme that degases both the development heading and the longwall panel.

434
In-mine horizontal boreholes are about 1000 – 1500 feet long. Two boreholes are drilled from
each location: one to degas the development headings; the other across the longwall panel.
A well-designed drilling program can remove 50 – 60% of the in-situ gas making mining much
safer and highly productive.

Figure 2. Coal seam degasification with in-mine horizontal drilling.

4.2 Post mining degasification

Working coal seams in this depth range (600 – 1500 ft) usually have several overlying and
a few underlying coal seams. They all contribute to gob gas emission. Total gob emissions
range from 3 – 10 MMCFD (million cubic feet per day).
Ventilation air generally cannot cope with this much gas and gob wells are needed to bypass
50 – 70 percent of the gob gas directly to the surface. The first gob well is put within 500 feet
from the set-up entry for the longwall face. Other gob wells are installed every 1000 feet to
2000 feet. The diameter of these gob wells ranges from 4 to 9 inches. Gas production is always
assisted by vacuum pumps, popularly called blowers. They have a capacity of 0.5 to 1.0
MMCFD and create a vacuum of 5 to 20 inches of water gauge.
On a typical 1200 ft x 10,000 ft longwall panel six to eight gob wells may be needed for
adequate degasification. Depending on the size and number of gob wells, 50 – 70% of the total
gob emissions can be captured for utilization.

5 HIGHLY GASSY COAL SEAMS

These coal seams are 1500 feet to 3000 feet deep with very high gas contents as shown in
Table 2. Extensive efforts are needed for both pre-mining and post-mining degasification.
These mines are located in Central and Southern Appalachian Basins of USA.

5.1 Pre-mining degasification

For adequate degasification, these longwall panels are drilled vertically at 10-15 acre spacing
and hydraulically fractured (Figure 3). Fracture wings (2) are 500-1000 feet long and follow
the direction of major horizontal stress. Properly designed frac-wells can remove 50% of the
gas in-situ. Additional, in-mine, horizontal drilling is needed to bring down the gas content

435
below 200 ft3/ton. The longwall panel is drilled at every 100- or 200-feet interval across the
panel. Gas pipelines are laid to take all gases to surface for utilization. A typical 1000 ft
x 10,000 ft panel may need 10-12 frac-wells. The fracking process is fully descried elsewhere
(Thakur 2016).

Figure 3. Vertical hydrofracked wells.

5.2 Post mining degasification

Gobs in highly gassy mines produce large volumes of gases ranging from 10-30 MMCFD/
acre. Vertical gob wells with 9 to 12” diameters are exclusively used to degas them. Seventy to
eighty percent of gob emissions must be captured so the remaining gases can be safely handled
by the ventilation air.
Spacing of gob wells and their locations on a typical deep mine is shown in Figure 4. For
best efficiency all gob wells must be located between the centerline of the longwall panel and
the tailgate. Best results are obtained when the gob wells are located about 100 feet from the
centerline towards the tailgates as shown in Figure 4.

Figure 4. Optimum layout of gob wells on a longwall face.

436
 Gas emissions rate in gobs is highly dependent on the specific gob emission and the rate of
mining in ft/day. A typical longwall panel 750 ft wide x 10,000 feet may need 30-35 gob wells
of 12” diameter. Blowers on these gob wells have a capacity of 2-4 MMCFD with a suction of
120 inches of water gauge. The width of the longwall panel is limited by the number of gob
wells needed. The maximum width is reached when the spacing between the gob wells equals
half the width of the panel (Thakur 2008).

6 SPECIAL CASES OF HIGH GAS EMISSIONS FROM FLOOR

6.1 Special construction of gob wells

It is common that the working coal seam is underlain by one or more coal seams, and they fall
in the gas emission space. Gob wells are generally completed 50 to 100 ft above the working
coal seam. Many disasters (mine explosions) have been caused by the heavy floor emission of
gases. To avoid disasters, gob wells are drilled to a depth of 240 feet below the working coal
seams as shown in Figure 5.

Figure 5. Gob well completed 240 ft below working seam.

The portion of the borehole below the mineable coal seam is filled with coarse sand and
small gravel, and a solid cement plug is put across the working coal seam. When longwall face
is mined through the borehole and moved 100-150 ft outby, the floor heaves. The partially
filled gob well provides a channel for gas to come out in the gob at a point that is 100-150 ft
inby of the longwall face. Such gob wells can be placed at close intervals (150-500 ft apart) on
the longwall panels that are prone to floor gas emissions. Gas production is assisted by
vacuum pumps creating 1-5 psi negative pressure.

437
7 SUMMARY AND CONCLUSIONS

Longwall mining is the safest and the most efficient method of mining underground coal. In
USA, more than 50% of all underground coal production is done by longwall mining. With
increasing depth, it is becoming necessary to degas the coal seams both before mining and
post mining. Coal seams are divided into three categories depending on their gas contents in
ft3/ton.
Pre-mining and post mining degasification of a typical coal seam in each category is
discussed. Modification of gob well construction to capture excessive floor emissions is
also discussed. Some deep coal mines can produce 70 – 80 MMCFD from gob wells
alone. Combining this production with gas production from frac-wells drilled in advance
of mining can create a gas flow of 200 – 250 MMCFD. At the current price of natural
gas at $6-9/MMCF, it can create a sizeable additional income to the coal mines defraying
all cost of degasification. Increased productivity and reduced cost of mining are added
benefits.

REFERENCES

Thakur, P.C. 1981. Methane Control for Longwall Gobs, Longwall-Shortwall Mining, State-of-the-Art,
Chapter 10, SME, AIME, pp 81–86.
Thakur, P.C. 2008. Optimum widths of longwall panels in highly gassy mines – Part II, 12th U.S. Mine
Ventilation Symposium, University of Nevada, Reno, Nevada.
Thakur, P.C. 2016. Advanced Reservoir and Production Engineering, Elsevier, pp. 210. Chapters 2-5; pp
17–73.
Thakur, P.C. 2016. Advanced Reservoir and Production Engineering, Elsevier, Chapter 8, pp 105–133.
Thakur, P.C. 2018. Advanced Mine Ventilation. Elsevier, pp 512. Chapter 14: pp 227–246.

438
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Study of in-situ coal seam gas content for Australian coal and gas
outburst management: A field data analysis and laboratory
experiment

Z.B. Li, T. Ren & M. Qiao

School of Civil, Mining, Environmental and Architectural Engineering, Wollongong, New South Wales

D. Black & J. Juric

South32 – Illawarra Metallurgical Coal, Wollongong, New South Wales

ABSTRACT: The gas content and composition are widely utilized in underground coal
mining to evaluate the potential for coal and gas outbursts. This study investigates the in-situ
coal seam gas content and gas sorption capacity in three coal seam sections (A1, A2, and A3)
in Sydney basin to understand their implications for gas management and outburst risk in
Australian coal mines. The fast direct desorption method and associated devices were
described in detail and employed to measure the in-situ coal seam gas components (Q1, Q2,
and Q3) and gas composition. This method offers a reliable and accurate approach for assess­
ing coal seam gas content and composition, with an error factor of ± 15%. The results
revealed gas contents in the A1 and A3 sections exceeding Level 2 Threshold Limit Values
(TLV), necessitating gas drainage measures. High-pressure isothermal gas sorption experi­
ments were conducted on coal samples from the A1, A2, and A3 seam sections to investigate
their gas sorption capacity. The Langmuir model demonstrated excellent performance in
describing CO2 sorption behavior, with fit coefficients R² greater than 0.99. The study
observed a strong positive correlation (Pearson’s coefficient >0.855) between in-situ gas con­
tent and Langmuir volume, suggesting that higher sorption capacities correspond to higher in-
situ gas content in coal seams. The findings of this study highlight the importance of under­
standing the relationship between in-situ gas content and gas sorption capacity in managing
gas emissions and mitigating outburst risks in underground coal mines. The insights derived
from the study can contribute to the development of effective gas management strategies and
enhance the safety and efficiency of coal mining operations in Australia.

1 INTRODUCTION

Coal mine gas emissions and outbursts have been long-standing challenges in the coal mining
industry, posing considerable threats to the safety of miners and the efficiency of mining oper­
ations (Thakur 2016; Lei et al. 2023). The gas content within coal seams is a vital factor for
comprehending and addressing these concerns (Liu et al. 2020). Accurate measurement and
analysis of this parameter help to estimate coal seam gas reserves, assess the risk of coal and
gas outbursts, and develop appropriate gas management strategies (Black 2018).
Gas content measurement techniques can be classified into two primary categories: indirect
and direct (Diamond and Schatzel 1998). Indirect approaches depend on gas adsorption prop­
erties under particular pressure and temperature conditions or empirical data linking gas con­
tent with other factors such as gas pressure, emission rate, coal rank, and overburden depth
(Hou et al. 2020). Nevertheless, indirect methods present various drawbacks, including diffi­
culties in acquiring precise in-situ seam gas pressure measurements, the presumption of com­
plete gas saturation in coal seams, and the complexity in determining the composition of

DOI: 10.1201/9781003429241-46

439
multi-component seam gases (Xue and Yuan 2017). In contrast, direct methods assess the real
amount of gas liberated from coal specimens and can be categorized into slow and fast desorp­
tion techniques. The slow desorption method requires weeks or months for a coal sample to
reach a low desorption rate or to stop desorbing, while the fast direct desorption method
expedites the process by crushing the sample upon arrival at the laboratory, obtaining gas
content results within days (Xue, Xie, and Wang 2013).
Since the introduction of the direct gas content measurement method by Bertard et al. (Ber­
tard, Bruyet, and Gunther 1970), various systems have been developed for direct measurement
of coal gas content. And these methods and their subsequent modifications have been effective
in measuring coal seam gas content in cored coal samples. Bertard and his colleagues used an
arrangement of three distinct containers (allocated for initial measurement, transportation,
and crushing), a U-tube manometer, and a crusher to assess the volume of gas desorbed from
coal samples collected from operational mining sites (Bertard, Bruyet, and Gunther 1970).
Kissell et al. tested the applicability of the “direct method” for determining the methane con­
tent of virgin coalbeds in the Vesta No. 5 coal (Kissell, McCulloch, and Elder 1973). They
conducted experiments using exploration cores to assess the method’s effectiveness in estimat­
ing gas content and predicting the total gas emissions from a potential mine situated at the
exploration site. The results indicated that the direct method, when applied to exploration
cores, successfully provided accurate estimates of coalbed gas content and reasonably pre­
dicted the overall gas emissions from the prospective mine. Xue and Yuan developed an
adapted direct method to measure gas content in coal seams by employing coal cuttings from
underground boreholes, specifically in cases of soft, friable, or geologically intricate coal
seams where acquiring cores may pose significant challenges (Xue and Yuan 2017; Xue et al.
2014). The modified method showed satisfactory results when applied in a soft seam at Huai­
nan, China, demonstrating its potential as a viable alternative to the standard procedure for
determining gas content when cores are not available. These findings suggest that the direct
method can be a valuable tool in gas content analysis and mine planning, contributing to
improved safety and efficiency in coal mining operations.
The Sydney Basin in Australia is known for its abundant coal resources and complex geo­
logical conditions, which make the assessment of gas content and sorption capacity particu­
larly crucial for local coal mining operations (Faiz et al. 2007). Previous studies have
primarily focused on the measurement of gas content and composition using various direct
and indirect methods. However, there is a lack of comprehensive research on the relationship
between in-situ gas content and gas sorption capacity in the context of coal seams from the
Sydney Basin.
In this study, we aim to address this gap by investigating the in-situ coal seam gas content
and gas sorption capacity in three coal seam sections (A1, A2, and A3) in the Sydney Basin.
The fast direct desorption method and associated devices were employed to measure the in-
situ coal seam gas components (Q1, Q2, and Q3) and gas composition. High-pressure isother­
mal gas sorption experiments were conducted on coal samples from these seam sections to
investigate their gas sorption capacity. The relationship between in-situ gas content and gas
sorption capacity is analyzed to understand its implications for gas management and outburst
risk in Australian coal mines.

2 METHODOLOGY

2.1 Measurement of coal seam gas content and gas composition

There are two commonly used methods to estimate the gas content in a coal seam: the direct
method and the indirect method. The direct method measures the total gas volume directly
from a coal sample, while the indirect method estimates the gas content from sorption iso­
therms or empirical correlations under given temperature and pressure conditions. The fast
and slow desorption methods are two types of direct methods, with the fast desorption
method taking less than five days, while the slow desorption process takes over five days for

440
gas to desorb from the intact core before the final crushing stage. Coal gas content is exam­
ined using the quick direct desorption method at the Coal Gas Laboratory A, and the gas
composition is assessed employing a 4-channel Micro Gas Chromatograph.

2.1.1 Lost gas component (Q1)

The lost gas component (Q1) refers to the portion of gas that escapes from the coal sample
during collection, prior to being sealed into the desorption canister. Q1 needs to be estimated
using the gas emission gathered after sealing the sample into the desorption canister. It is
widely acknowledged that the volume of gas released during initial desorption is proportional
to the square root of the desorption time (Zhang 2013), as shown in Figure 1 (a) and detailed
in AS3980:2016 (Australia 2016). By projecting the best-fit line representing the initial gas
emission from the time the core was sealed into the gas desorption canister to the midpoint
between the start and end of coring the sample, an estimation of the gas volume lost during
core sample recovery can be obtained. Q1 can be estimated based on the following equation:

where q = the accumulated volume of desorbed gas since coal is placed in the desorption can­
ister and measurement commenced, m3/t; t = the time since the start of gas desorption in the
hole, min; k = the slope of a regression line passed through canister desorption data.
Note: the desorbed volumes are expressed at STP conditions (20 ℃, 101.325kPa).

Figure 1. Lost gas content (Q1) calculation illustration (a) and its measuring apparatus (b) (Australia
2016).

2.1.2 Desorbed gas component (Q2)

The desorbed gas component (Q2) represents the volume of gas released from a coal sample
while it is contained in a desorption canister. The fast desorption test typically lasts less than
five days, often taking less than one day. Generally, gas released from a core sample is meas­
ured by water displacement using a graduated glass or plastic measuring flask. As shown in
Figure 2, the measurement apparatus can be set up so that the gas liberated from the core
sample within the desorption canister enters the measuring flask via a tube connected to either
the top or bottom of the flask. Connecting the tube to the top of the cylinder is preferred, as it
prevents the desorbed gas from bubbling through the water column, thereby reducing the risk
of gas loss through dissolution, particularly when the seam gas has a high CO2 concentration
(AS3980:2016) (Australia 2016). The desorbed gas (Q2) is calculated by summing all volume

441
increments determined over the measurement period. If there are n steps to complete the
measurement, the Q2 value can be expressed as:

where m = the mass of sample, g. ΔVi = the incremental volume of gas desorbed, which is the
difference between the current corrected gas volume at instant i and the previous corrected
volume in the system at instant i - 1 (provided the cylinder has not been refilled). i = 1 of the
measurement corresponds to the beginning of measurement and associated V is the volume of
void in the system at this instant.

Figure 2. Desorbed gas volume measurement apparatus for estimation of Q2.

2.1.3 Residual gas component (Q3)

The residual gas component (Q3) quantifies the gas released from a coal sample after crushing.
The equipment for this process is illustrated in Figure 3, and the procedure involves the fol­
lowing steps:
a) Perform a desorbed gas test on the coal sample to measure the gas that is released from the
coal under a particular set of conditions.
b) Remove the coal core from the desorbed gas test apparatus and collect a representative
sub-sample of the coal.
c) Seal the sub-sample into a crushing or grinding mill.
d) Crush the coal to a particle size such that 95% of the coal is less than 212µm.
e) Measure the volume of gas released from the coal sample using a water column similar to
the one employed in the desorbed gas (Q2) test.
The result of this procedure is the Q3, which represents the quantity of gas that remains
within the coal matrix after desorption, under the particular conditions applied in the
desorbed gas test.

2.1.4 Measured total gas content (QT)

The total gas content is calculated by adding the three gas components (Q1, Q2, and Q3), as
demonstrated in Equation (3), and determined at a standard temperature of 20°C and
a pressure of 101.325 kPa.

442
Figure 3. Residual gas content measurement apparatus for estimation of Q3.

2.1.5 The limit of this method and its controls

The precision of the coal seam gas content measurement method outlined in Section 2.1 is sub­
ject to a variety of factors. These factors encompass system leakage, desorption rate (influen­
cing the Q1 calculation), gas solubility in water (especially CO2), sample moisture content,
partial pressure (impacting the equilibrium endpoint for desorption), temperature (affecting
the desorption magnitude), and barometric pressure (altering the desorption magnitude). To
mitigate errors stemming from these factors, a combination of procedural, measurement, and
calculation controls can be implemented during the gas content determination process.
These controls comprise methods for detecting system leaks, evaluating samples based on
an “as-received” condition, and utilizing a measurement apparatus designed to prevent gas
loss through dissolution, particularly when dealing with seam gas containing high CO2 con­
centrations. By adopting these controls, the overall gas estimation accuracy using this method
is considered to have an error factor of ± 15%, aligning with the targeted maximum relative
error for gas content measurements in the coal mining sector (Xue and Yuan 2017; Moore
2012). This degree of accuracy guarantees that the method delivers dependable and consistent
outcomes, rendering it an appropriate choice for evaluating gas content and composition in
Australian coal mining industries.

2.2 Lab experiment of gas sorption isotherms

In this study, coal samples were obtained from a representative coking coal seam with a depth
of cover about 400m in Australian Sydney basin. The collected samples were prepared into
particles with particle sizes of <0.212mm, as shown in Figure 4.
A purpose-built high-pressure isothermal sorption system at University of Wollongong was
used to test the gas sorption isotherms, as shown in Figure 5, which consists of six modules:
gas supply and control module, adsorption and desorption module, pressure acquisition
module (accuracy: 0.001 Pa), vacuum degassing module, temperature control and measure­
ment module (accuracy: 0.001 °C), and high-precision weight measurement module (accuracy:
0.001 g) (Li et al. 2023).
a) Dry the samples in a vacuum drying oven at 100°C for 1 hour and cool in a desiccator.
b) Conduct airtightness test of the sorption canister by injecting 5.0 MPa helium and keep for
24 hours; re-seal if pressure drop is higher than 20 kPa.

443
Figure 4. Sample preparation.

Figure 5. Structural diagram of adsorption and desorption test system.

c) Degas the sorption canister for 2 hours using a vacuum pump till relative pressure reaches
-100kPa, and let it stand for 24 hours; inject helium to measure dead space volume.
d) Charge CO2 into the canister with samples, and test adsorption and desorption isotherms
at pressure gradient of 0.5 MPa, maximum pressure of 5.0 MPa, and equilibrium time of
24 hours.
e) Maintain thermostatic water bath and room temperature at 30°C during sorption test.
The Langmuir equation is a mathematical model that is commonly used to describe the
adsorption of gases onto a porous surface.

where V = the adsorption volume under equilibrium pressure P, cc/g; VL = the Langmuir
volume, which represents the limit gas adsorption capacity when the gas pressure is infinite,
cc/g; PL = the Langm uir pressure, which represents the pressure at which the adsorption
volume V reaches half of the VL, MPa.

3 RESULTS AND ANALYSIS

3.1 Analysis of in-situ coal seam gas content data

3.1.1 Coal seam gas content and outburst threshold limits (TLV)
The Australian outburst TLV varied linearly based on gas composition, increasing from
a minimum in CO2 rich conditions (5 m3/t) to a maximum in CH4 rich conditions (9 m3/t), as

444
shown in Figure 6, with the presence of CO2 seam gas considered a significantly higher out­
burst risk than CH4 seam gas. The Level 1 TLV for ‘normal’ mining was 9.0 m3/t in pure CH4
conditions and 5.0 m3/t in pure CO2 conditions. If gas content was not reduced below the
Level 1 TLV, mining allowed only under outburst mining procedures. The Level 2 TLV for
outburst mining was 12.0 m3/t in pure CH4 conditions and 10.0 m3/t in pure CO2 conditions.
If gas content did not fall below the Level 2 TLV, mining was only permissible using remote-
operated equipment, with all personnel staying clear of the outburst risk zone (Black 2019;
Black and Aziz 2010).
Based on the measuring results from coal mine A, the gas content of the upper A1, mid-
section A2, and lower section A3 of the same seam has been analyzed. Figure 6 demonstrates
that the CO2 composition in all three seam sections exceeds 95%. The virgin gas contents in
both A1 and A3 sections surpass the Level 2 Threshold Limit Value (TLV), while the gas con­
tent in the A2 seam section falls between Level 1 and Level 2 TLV. Hence, is crucial to imple­
ment gas drainage measures to reduce the gas content below the respective TLVs in each seam
section.

3.1.2 Q1, Q2, Q3 and QT gas content component

Figures 7 and 8 display the distribution of the Q1, Q2, Q3 and QT gas content components for
the A1, A2, and A3 seam sections. The total gas content (QT) in the lower section A3 ranges
from 10.9 to 13.7 m3/t, which is higher than that of the upper section A1 (10.17-12.69 m3/t)
and the mid-section A2 (7.16-8.8 m3/t). Among the three gas components, the residual gas
content (Q3) varies from 4.26 m3/t in the A3 seam section to 8.58 m3/t in the A1 seam section.
Q3 contributes the most to the total gas content (QT), with proportions ranging between 39%
and 88%.
The desorbed gas component (Q2) spans from 0.67 m3/t in the A2 seam section to
4.60 m3/t in the A3 seam section. The lost gas content (Q1) contributes the least to the total
gas content (QT), with proportions ranging from 3% in the A2 seam section to 27% in the
A3 seam section. Since Q1 and Q2 gas contents are measured during the early stages of
desorption, they are more closely associated with coal and gas outburst propensity (Black
et al. 2009).
Among the three sections, the Q1 and Q2 gas contents of the A3 seam section are higher
than those of the A1 and A2 seam sections, as well as their contribution proportions. Conse­
quently, it is reasonable to suggest that the outburst propensity of the A3 seam section is
higher than that of the A1 and A2 seam sections. Moreover, for the A1 and A2 coal seam
sections, the residual gas content (Q3) is higher than that of A3, indicating that these seam
sections are more difficult to drain, and the gas drainage efficiency will be lower.

Figure 6. In-situ coal seam gas content of Mine A and outburst threshold limits (TLV).

445
Figure 7. Q1, Q2, Q3 and QT gas content of different seam sections of coal seam A.

Figure 8. The proportion contribution of Q1, Q2, Q3 to QT of each seam section.

3.2 CO2 sorption isotherms and hysteresis

It is reasonable to focus on pure CO2 sorption isotherms in this study, given that the gas com­
position of the tested coal seam is more than 95% CO2. Figure 9 and Table 1 present the CO2
adsorption and desorption isotherms and Langmuir model fitted results of samples from three
sections of the same coal seam (A1, A2, and A3). The Langmuir model demonstrates excellent

446
performance in describing the sorption behaviour of CO2, with fit coefficients R2 greater than
0.99, emphasizing the model’s suitability for representing the CO2 sorption process in these
coal samples.
Analysis of the experimental and characterization results reveals variations in adsorption
and desorption capacities for CO2 among the tested coal samples from the different seam sec­
tions. Notably, the desorption Langmuir volume of CO2 (de-VLCO2) is smaller than ad-VL
CO2, varying between 18.11 m3/t and 26.63 m3/t, with a considerable proportion of residual
gas content Vr (5.38-6.49 m3/t). Sample A3 has the highest Langmuir volume and a moderate
Langmuir pressure for adsorption, indicating a strong adsorption capacity, which could be
attributed to the presence of developed micropores in the coal reservoir. The highest residual
gas content and high hysteresis degree of A3 could be related to the presence of poorly con­
nected pores, which may contribute to the trapping of CO2 gas.

Figure 9. CO2 sorption isotherms and hysteresis.

Table 1. Model fitted parameters for CO2 ad-/desorption isotherms and its hysteresis index.
Adsorption parameters Desorption parameters

Residual gas Sorption hyster­

Sample content (Vr) esis index (HI)
ID VL (m3/t) PL (kPa) R 2 3
VL (m /t) PL (kPa) (m3/t) R2 (%)

A1 28.97 867.06 0.997 23.63 1239.50 4.81 0.997 22.49

A2 23.02 980.57 0.996 18.11 1322.97 4.39 0.995 25.22
A3 31.38 932.37 0.996 26.63 1438.51 4.10 0.998 18.12

447
3.3 The relationship between in-situ gas content and gas sorption capacity
The in-situ gas content and gas sorption capacity of a coal seam are interconnected factors
that significantly impact the determination of coal seam gas reserves and the propensity for
outbursts in underground mines. In this section, we examined the relationship between in-situ
gas content and gas sorption capacity, utilizing the data obtained from the A1, A2, and A3
seam sections. So as to comprehend how variations in sorption capacity affect the in-situ gas
content and to identify factors that could impact gas management and outburst risk.
Figure 10 presents a scatter plot comparing in-situ gas content (QT) with the Langmuir
volume for the coal samples from the three seam sections. A strong positive correlation (Pear­
son’s coefficient >0.855) is observed between gas content and Langmuir volume. This suggests
that higher sorption capacities correspond to higher gas content within coal seams.

Figure 10. The relationship between in-situ gas content and gas sorption capacity.

The previous data show that the A3 seam section has the highest in-situ gas content (QT) and
the highest adsorption Langmuir volume. Conversely, the A2 seam section exhibits the lowest
in-situ gas content and the lowest adsorption and desorption Langmuir volume. These results
further reinforce the positive correlation between in-situ gas content and gas sorption capacity.
The isotherm-based residual gas content (Vr) also plays a critical role in understanding the rela­
tionship between gas content and sorption capacity. As mentioned earlier, the A1 and A2 seam
sections have higher residual gas content (Q3) than the A3 seam section, indicating that these sec­
tions are harder to drain and exhibit lower gas drainage efficiency. This could be attributed to the
coal’s pore structure, which may influence the connectivity of pores and the trapping of gas within
the coal matrix.
In summary, he connection between gas content and gas sorption capacity is evident.
Higher sorption capacities correspond to higher in-situ gas content in the coal seams, and
variations in residual gas content can impact gas management strategies and outburst risk.
Understanding these relationships is crucial for developing effective gas drainage techniques
and ensuring safe mining operations in coal mines with high gas content.

4 CONCLUSIONS

In this study, we examined the in-situ coal seam gas components (Q1, Q2 and Q3) and corres­
ponding measurement methods and associated devices, analyzed their limitations and controls.

448
Utilizing the methods, the coal seam gas content and gas composition of the A1, A2, and A3
seam sections are determined. Furthermore, the gas sorption capacity for samples taken from
the A1, A2, and A3 seams are investigated through high-pressure isothermal gas sorption
experiments. Based on the results and discussions presented, the following four key conclusions
can be drawn:
The direct desorption method, specifically the fast direct desorption method, proves to be
an effective and reliable approach to measure the in-situ coal seam gas content and gas com­
position for the coal seams investigated in this study. By implementing various procedural,
measurement, and calculation controls, this method ensures accurate results with a low error.
The method can be a well-suited for assessing coal seam gas content and gas composition in
Australian coal mining industries.
The in-situ gas content and gas composition of the coal seams exhibit significant variations
among the A1, A2, and A3 seam sections. The A3 seam section shows the highest in-situ gas
content, while the A2 seam section has the lowest. The CO2 composition of all three seam
sections is higher than 95%, highlighting the importance of understanding CO2 sorption
behavior in these coal seams.
The relationship between in-situ gas content and gas sorption capacity is evident, with higher
sorption capacities corresponding to higher in-situ gas content. The A3 seam section demon­
strates the strongest adsorption capacity, which could be attributed to the presence of developed
micropores in the coal reservoir. The higher residual gas content in the A1 and A2 seam sections
indicates that these sections are harder to drain and exhibit lower gas drainage efficiency, pos­
sibly due to the coal’s pore structure and the trapping of gas within the coal matrix.
The analysis of gas content components (Q1, Q2, and Q3) reveals that the A3 seam section
has a higher outburst propensity than the A1 and A2 seam sections, due to its higher Q1 and
Q2 gas contents. This finding underscores the need for effective gas management strategies
and outburst prevention measures in coal mines with high-gas content and varying gas sorp­
tion capacities.
These conclusions offer valuable insights into the relationship between in-situ gas content
and gas sorption capacity in coal seams, the importance of accurate measurement methods,
and the factors influencing gas management and outburst risk in coal mines. Future research
could explore the impact of different coal properties, such as pore structure, sorption hyster­
esis and coal rank, on the relationship between in-situ gas content and gas sorption capacity
to optimize gas drainage techniques and ensure safe mining operations.

REFERENCES

Australia, Standards. 2016. “Determination of gas content of coal and carbonaceous material - Direct
desorption method.” In. Sydney: SAI Global Limited.
Bertard, C., B. Bruyet, and J. Gunther. 1970. ‘Determination of desorbable gas concentration of coal
(direct method)’, International Journal of Rock Mechanics and Mining Sciences & Geomechanics
Abstracts, 7: 43–65.
Black, Dennis. 2018. “Control and management of outburst in Australian underground coal mines.” In
18th Coal Operators’ Conference, edited by Naj Aziz and Bob Kininmonth, 246–63. Mining Engineer­
ing, University of Wollongong.
Black, Dennis, and Naj Aziz. 2010. “Outburst threshold limits-current research outcomes.” In 2010 Coal
Operators’ Conference, edited by Naj Aziz and Bob Kininmonth. Mining Engineering, University of
Wollongong.
Black, Dennis, Naj Aziz, Matt Jurak, and Raul Florentin. 2009. ‘Gas content estimation using initial
desorption rate’.
Black, Dennis J. 2019. ‘Review of coal and gas outburst in Australian underground coal mines’,
Int J Min Sci Technol, 29: 815–24.
Diamond, William P., and Steven J. Schatzel. 1998. ‘Measuring the gas content of coal: A review’, Inter­
national Journal of Coal Geology, 35: 311–31.
Faiz, M., A. Saghafi, N. Sherwood, and I. Wang. 2007. ‘The influence of petrological properties and
burial history on coal seam methane reservoir characterisation, Sydney Basin, Australia’, International
Journal of Coal Geology, 70: 193–208.

449
Hou, Xiaowei, Shimin Liu, Yanming Zhu, and Yun Yang. 2020. ‘Evaluation of gas contents for a
multi-seam deep coalbed methane reservoir and their geological controls: In situ direct method versus
indirect method’, Fuel, 265: 116917.
Kissell, Fred N., Charles M. McCulloch, and Curtis H. Elder. 1973. “The direct method of determining
methane content of coalbeds for ventilation design.” In.: US Department of Interior, Bureau of
Mines.
Lei, Yang, Yuanping Cheng, Liang Wang, Ting Ren, and Qingyi Tu 2023. ‘Mechanisms of Coal and Gas
Outburst Experiments: Implications for the Energy Principle of Natural Outbursts’, Rock Mechanics
and Rock Engineering, 56: 363–77.
Li, Zhongbei, Ting Ren, Xiangchun Li, Ming Qiao, Xiaohan Yang, Lihai Tan, and Baisheng Nie. 2023.
‘Multi-scale pore fractal characteristics of differently ranked coal and its impact on gas adsorption’,
International Journal of Mining Science and Technology.
Liu, Yanwei, Yang Du, Zhiqiang Li, Fajun Zhao, Weiqin Zuo, Jianping Wei, and Hani Mitri. 2020.
‘A rapid and accurate direct measurement method of underground coal seam gas content based on
dynamic diffusion theory’, International Journal of Mining Science and Technology, 30: 799–810.
Moore, T. A. 2012. ‘Coalbed methane: A review’, International Journal of Coal Geology, 101: 36–81.
Thakur, Pramod. 2016. Advanced reservoir and production engineering for coal bed methane (Gulf Profes­
sional Publishing).
Xue, Sheng, Jun Xie, and Yu Cang Wang. 2013. ‘A Direct Measurement System to Estimate Gas Con­
tent in Coal’, Applied Mechanics and Materials, 313-314: 604–07.
Xue, Sheng, and Liang Yuan. 2017. ‘The use of coal cuttings from underground boreholes to determine
gas content of coal with direct desorption method’, International Journal of Coal Geology, 174: 1–7.
Xue, Sheng, Liang Yuan, Jun Xie, and Yucang Wang. 2014. ‘Advances in gas content based on outburst
control technology in Huainan, China’, International Journal of Mining Science and Technology, 24:
385–89.
Zhang, Lei 2013. ‘Study of coal sorption characteristics and gas drainage in hard-to-drain coal seams’,
University of Wollongong.

450
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Airflow patterns and blast fume dispersion in different mining

methods

S. Jayaraman Sridharan, A. Adhikari, P. Tukkaraja & J. Connot

Department of Mining Engineering and Management, South Dakota Mines, Rapid City, USA

ABSTRACT: Airflow behavior is a complex phenomenon and depends on the geometrical

configuration, pollutant source and ventilation parameters. The knowledge of the airflow pat­
terns and pollutant dispersion in the underground working environment is useful in designing
an effective ventilation system. Numerical simulation of blast fume dispersion in large open­
ings can help understand the distribution of fumes in the underground working areas, the role
of auxiliary ventilation on the fume dispersion, and air flow patterns. In this research, the air­
flow pattern and blast fume dispersion in working panels particularly for large openings are
analyzed using CFD. The results show that the blast fumes can accumulate in stagnant regions
such as non-active, dead-end headings and could be a potential hazard if not monitored and
controlled in a timely manner. The dead-end zone where there is a lack of sufficient ventilation
should be monitored for fumes before entering such areas.

1 INTRODUCTION

As the near surface mineral resources deplete and mineral consumption increases, the under­
ground will become more extensive and deeper. Underground mining can pose several challenges
such as support, drainage, transportation, subsidence, illumination, and ventilation. The mining
method largely determines these challenges. A well-thought-out plan is necessary to manage these
challenges in an underground mining operation. One of the main aspects of underground mining
that is often overlooked is ventilation. Ventilation is required to control the level of air contamin­
ants, such as dust, fumes, and gases, and to provide a comfortable working environment for
miners. In addition to maintaining air quality, ventilation can also be used to control the tempera­
ture and humidity in the mine, which can affect worker comfort and safety. The ventilation
system can also play a critical role in emergency situations, such as fires or explosions, by helping
to control smoke and toxic gases and facilitating evacuation. Mine health and safety depend on
ventilation, and it requires significant capital and operating cost. Given the importance of ventila­
tion in underground mining, it is essential that mining companies invest in high-quality ventilation
systems and ensure that they are properly maintained and monitored.
In mining, when the geological conditions are favorable, methods that involve large openings
are often preferred. This can provide several advantages over other techniques, including
improved access, increased safety, higher production rates and lower costs. Some of the challenges
associated with large openings could be complex airflow patterns, pressure differentials, heat and
energy loss, pockets of stagnant air and gas accumulation in enclosed areas. These potential issues
need to be studied and understood for better design, planning and health and safety of the miners.
Blasters and ventilation engineers must ensure that adequate ventilation is provided to
dilute pollutants below threshold limit values before miners can enter the working face. Trad­
itionally, most underground mines estimate post-blast dilution time based on observations
and experiences, often resulting in an erroneous selection of post-blast dilution time. Few
mines use a theoretical approach; however, as they are unable to account for changes in blast­
ing conditions (Tiile, 2019). Most underground mines perform blasting operations at the end

DOI: 10.1201/9781003429241-47

451
of the working shift to allow sufficient time to clear the blast fumes. Due to prolonged waiting
time, a significant production delay can occur in mines when multiple blasts are required in
a day (Gillies et al., 2004).
The waiting time or re-entry time depends on a number of factors such as the air velocity at
the inlet, vent duct diameter, and the amount of gases or fumes produced during blasting.
While there are different algorithms and models available to calculate the post-blast dilution
time in a blind heading, they lack accuracy for some conditions and geometries (Torno and
Torano, 2020). An exponential function was obtained for calculation of gas clearance after
blasting with a correct statistical adjustment by means of experimental values. Area of the
face advance, mass of explosive used in the blast, time for the air to clear, distance from the
end of the ventilation duct to the face and quantity of air (Torno and Toraño, 2020).
Airflow behavior is a complex phenomenon and depends on the geometrical configuration,
pollutant source and ventilation parameters. Computational Fluid Dynamics (CFD) can help
predict the airflow behavior, visualize the airflow patterns, and measure the performance of
the ventilating system. Gillies et al. (2004) developed assessment tools to minimize safe after
blast re-entry time in an underground metal mine in Australia. Lu (2011) studied blast fume
diffusion in eight different conditions of different drift lengths and outlet positions. The results
showed that the fume discharge time had a linear relationship with the drift length and pipe
outlet position. Stewart (2014) performed experimental investigation to evaluate the re-entry
prediction methods. Akash et al. (2023, 2022) investigated the effect of trapped fumes on the
re-entry time using CFD and developed a machine learning model to predict the dilution time.
Many researchers have successfully used the CFD techniques to model the ventilation
system and evaluate the performance of fume and dust removal (Qian et al., 2020). Purusho­
tham et.al (2021) investigated the airflow and radon dispersion for block cave mining method
using CFD studies. Numerical simulation of blast fume dispersion in large openings can help
understand the distribution of fumes in the underground working areas, role of auxiliary ven­
tilation on the fume dispersion, and air flow patterns.
Large-opening mine ventilation is unique for the following reasons: (1) it is challenging to
keep airflow velocities high enough to effectively remove or dilute airborne contaminants, (2)
large air volumes can be moved through the mines with little static pressure drop, and (3)
stoppings to direct ventilation airflows are costly to construct and maintain (Grau III et al.,
2006). The knowledge of the airflow patterns and pollutant dispersion in the underground
working environment is useful in designing an effective ventilation system (Krog et al., 2004).
Using CFD techniques, this study examines air flow patterns and blast fume dispersion in
large openings:
i) Sanford Underground Research Facility (SURF) cavern,
ii) large room and pillar working panel, and
iii) longhole open stope.

2 METHODOLOGY

A brief overview of the methodology is presented in this section. Three geometry models are
considered for the airflow pattern and fume dispersion study.
Blast fume is a mixture of toxic gases like carbon monoxide (CO), carbon dioxide, nitrogen
oxides, etc. Out of these gases, CO is of major concern due to the amount released in a blast and
the permissible exposure limit (PEL). The fluid domain in CFD (air) is assumed to be a mixture
of Nitrogen, Oxygen and CO. The amount of CO released from the blast is calculated using the
NIOSH study (Rowland and Mainiero, 2000) to determine the effect of ANFO fuel oil content on
CO production i.e., 1 Kilogram of ANFO produces 16 liters of CO. Since a forcing ventilation
system is a highly efficient and commonly used auxiliary ventilation system, a forcing ventilation
system was considered to study the blast fume dispersion problem in large openings.
First, the airflow patterns and blast fume dispersion from an auxiliary ventilated drift in
a regular room and pillar panel is investigated using CFD. Next, a large room and pillar panel
is investigated and compared with the regular room and pillar airflow characteristics.

452
2.1 Room and pillar
Room and pillar mining is one of the earliest and most common mining methods. Figure 1
shows the dimensions and extraction sequence of a typical room and pillar panel. In this
study, the panel is assumed to be at the extraction stage 11.

Figure 1. Excavation sequence in a room and pillar mine (Kohler, 2022).

Dimensions of an actual large room and pillar limestone mine is considered for this study.
The mine has a single-level room and pillar operation with a 7.9 m (26 ft) high roof,
13.7 m (45 ft) wide entries, and 13.7 m by 13.7 m (45 ft by 45 ft) pillar dimensions. The mine is
accessed via a 396 meter (1300 ft) long decline and 100 meter (330 ft) deep ventilation shaft.
Part of the mining layout investigated in this study is shown in Figure 2.

2.2 SURF
SURF is located at the site of the former Homestake Gold Mine in Lead, South Dakota. This
facility has since been repurposed and extensively modified to accommodate numerous under­
ground science experiments. The site is owned and operated by the South Dakota Science and
Technology Authority (SDSTA). A global neutrino physics community is developing a leading-
edge, dual-site experiment for neutrino science and proton decay studies, known as the Deep
Underground Neutrino Experiment (DUNE). The facility required for this experiment, the Long-
Baseline Neutrino Facility (LBNF), is constructed at the Sanford Underground Research Facility
(SURF) in Lead, South Dakota. The construction project will excavate three caverns a mile
underground as shown in Figure 3. Two of the caverns will house the detector modules

453
Figure 2. Layout of the large room and pillar geometry modeled in the CFD model.

(approximately 28-m H x 20-m W x 145-m L) and a central cavern (approximately 11-m H x

19.5-m W x 190-m L) will house the utilities needed to support the detectors. The excavation of
each cavern proceeds from the top to the bottom using drill and blast.

Figure 3. Caverns constructed at the SURF for the DUNE experiment (Fermilab, 2022).

The final stage of cavern development considered for this study is shown in Figure 4. At
this stage, the opening volume is at the largest.

2.3 Longhole open stoping

Long-hole sublevel stoping, often referred to as sublevel open stoping or blast hole stoping is
a commonly used method in large scale mining. It is a versatile and productive method that is

454
Figure 4. SURF Geometry with auxiliary ventilation.

primarily used for extraction of large ore bodies with a steep dip, regular shape, and defined ore
boundary.
Development of sublevel stoping includes a haulage drift, raises to access sublevels, under­
cuts below the stope, draw points, and a slot raise at the end of the stope (Figure 5). The ore

Figure 5. Typical long-hole stoping operation (Copco, 2000).

455
body is divided into parallel sublevels of typically 20-40 metres vertically apart. Once this is
completed, a slot raise is developed and subsequently increased by blasting and drilling
accesses that are developed into the ore, filled with explosives and blasted sequentially.

3 RESULTS AND DISCUSSION

Boundary conditions used in the simulation of different mining methods are presented in
Table 1.

Table 1. Explosive and ventilation conditions for each scenario.

Amount of Explosive CO produced in Auxiliary Ventilation Main ventilation
Mining Method in the blast (kg) the blast (l) quantity (m3/s) quantity (m3/s)

Room and pillar 183 2928 7.85 35.4

Large Room 440 7040 15.7 94.4
and pillar
SURF 150 2400 7.85 38
Stope 2260 36160 15.7 37.5

3.1 Room and pillar

The results from the regular room and pillar airflow simulation are presented in this section.
For this study, the layout is assumed to be at the excavation sequence 11 shown in Figure 1.
At this stage, the working face is at the farthest end. The velocity distribution and the turbu­
lence energy are shown in Figure 6. Velocities higher than 1 m/s and concentration more than
50 ppm are shown in red.

Figure 6. Velocity magnitude and velocity vectors at the blind heading.

The blast fumes are released from the central room as shown in Figure 7. The other two
headings do not have any active blasting going on and hence there is no auxiliary ventilation
in those headings. The blast fume dispersion in the panel is shown at different times in the
Figure 7. The results show that the blast fumes accumulate in the headings without auxiliary
ventilation. The asymmetry in the fume dispersion is due to the auxiliary ventilation and the
negative pressure at the auxiliary duct inlet.

456
Figure 7. CO dispersion in the room and pillar panel at time 10 s, 90 s, 300 s, 720 s, 900 s, and 1200 s.

Figure 8. Velocity distribution in the large room and pillar panel.

457
 The velocity vectors in Figure 6 shows that auxiliary ventilation duct is not recirculating the
fumes. However, the CO dispersion plots in Figure 7 show that the duct inlet is positioned in
the region where the blast fumes could diffuse and reach higher concentrations. This demon­
strates the need for identifying the appropriate location for the auxiliary ventilation system.

3.2 Large room and pillar

The velocity distribution in a large room and pillar working panel is shown in Figure 8. The
CO concentration after the blast at different time are shown in Figure 9.

Figure 9. CO dispersion in the room and pillar panel at time 10 s, 60 s, 180 s, 300 s, 480 s, 900 s, 1200 s,
and 1500 s.

458
3.3 SURF
A steady state simulation of airflow and gas dispersion is performed for the SURF cavern
geometry to identify the regions of dead spaces. To visualize regions of gas accumulation in
steady state condition, high gas release rate at the working face is assumed as shown in Table 2.

Table 2. Boundary conditions for steady state analysis.

Location Type Value

inlet_cavern Velocity inlet 1 m/s

duct-inlet Velocity inlet 10 m/s
outlet_cavern Outflow NA
working_face MASS-FLOW-INLET 0.006 kg/s of CO

The duct inlet is at high velocity (10 m/s) and the velocity inside the cavern is less than 1 m/s.
The color coding for the streamline of velocity shown in Figure 10 is customized to clearly visual­
ize low velocities i.e. velocity magnitude more than 1 m/s are shown in red. Similarly for the
volume rendering of gas concentration in Figure 10, gas concentrations more than 50 ppm of CO
are shown in red. In Figure 11, the dead spaces immediately below the inlet can be clearly seen.

Figure 10. Streamlines of airflow inside the cavern. Figure 11. CO concentration in the cavern.

Figure 12. Velocity distribution in the drifts and crosscuts of the longhole stope (left); CO distribution
in the longhole stope after 20 minutes of ventilation (right).

459
3.4 Longhole open stoping
The velocity distribution in the drifts, crosscuts, and stope are shown in Figure 12. The
broken rock in the stope is modeled as a porous media hence the low velocities in the porous
media. Auxiliary ventilation in the upper and lower levels help in the dilution of the fumes in
the stope. Figure 12 shows the high CO concentration within the stope due to the fumes
trapped in the broken rock.

4 CONCLUSIONS

This study shows that CFD is a useful tool in analyzing the blast fume dispersion and airflow
patterns under various airway configurations used in different mining methods. While this
study focuses on underground mining methods, understanding of airflow pattern and pollu­
tant distribution in large underground space will be beneficial to the study of the flow and
explosion of methane, mine fires, and diesel particulate matter.
The results show that the change in airway dimensions (large openings) changes the airflow
behavior and gas dispersion. The large openings are characterized by the lower velocities in the
airway and slower removal of gases in low velocity areas. The study shows that air velocity plays
a major role in the dilution of the fumes. Strategic positioning of the auxiliary duct and low-
pressure propeller fans can help in ventilating large underground openings. Because of the porous
media (broken ore), stopes take longer for dilution. The methodology and the study results will be
helpful in investigating the gas dispersion and airflow patterns of other underground bulk-mining
methods such as caving and stoping methods. The extra-large dimensions make the cavern one of
a kind and the study results will be pioneer for such large excavations in the future.

REFERENCES

Adhikari, A., Jayaraman Sridharan, S., Tukkaraja, P., Sasmito, A. & Vytla, S. (2022) The Effect of
Trapped Fumes on Clearance Time in Underground Development Blasting. Mining, Metallurgy &
Exploration, 39, 1811–1820.
Adhikari, A., Tukkaraja, P., Sridharan, S. J. & Verburg, A. (2023) Machine learning model for predicting
blast fume dilution time in underground working areas. CIM Journal, 14, 64–73.
Copco, A. (2000) Sublevel Stoping Operation Atlas Copco Group.
Fermilab (2022) DUNE at LBNF. Department Of Energy.
Gillies, A. D. S., Wu, H. & Shires, D. (2004) Development of an assessment tool to minimize safe after
blast re-entry time to improve the mining cycle. Proceedings of the Tenth US/North American Mine
Ventilation Sm. Alaska, USA, AA Balkeme publishers.
Grau Iii, R., Krog, R. & Robertson, S. (2006) Maximizing the ventilation of large-opening mines. Pro­
ceedings of the 11th US/North American Mine Ventilation Symp. University Park, PA.
Kohler, J. (2022) Room and Pillar Method. PennState.
Krog, R., Grau, R., Mucho, T. & Robertson, S. (2004) Ventilation planning layouts for large opening mines.
Lü, W.-S., He, L. & Yang, P. (2011) Simulation study of blast fume diffusion characters of driving face.
Journal of Coal Science and Engineering (China), 17, 47–51.
Qian, J., Wang, J., Liu, H. & Xu, H. (2020) CFD Modeling of Ventilation and Dust Flow Behavior in
Polishing and the Design of an Innovative Wet Dust Removal System. International Journal of Envir­
onmental Research and Public Health, 17, 6006.
Rowland, J. H. & Mainiero, R. (2000) Factors Affecting ANFO Fumes Production. Proceedings of the
26th Annual Conference on Explosives and Blasting Technique, Vol. 1.
Stewart, C. (2014) Practical prediction of blast fume clearance and workplace re-entry times in develop­
ment headings.
Tiile, R. N. (2019) Investigating blast fume propagation, concentration and clearance in underground
mines using computational fluid dynamics (CFD). Mining Engineering. Missouri University of Science
and Technology.
Torno, S. & Torano, J. (2020) On the prediction of toxic fumes from underground blasting operations
and dilution ventilation. Conventional and numerical models. Tunnelling and Underground Space
Technology, 96.

460
Torno, S. & Toraño, J. (2020) On the prediction of toxic fumes from underground blasting operations
and dilution ventilation. Conventional and numerical models. Tunnelling and Underground Space
Technology, 96, 103194.
Tukkaraja, P., Bhargava, R. & Jayaraman Sridharan, S. (2021) Radon in Underground Mines. Mining
Technology.

461
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

The oxiperator for Ventilation Air Methane (VAM)

E. Prabhu & M. Prabhu

Prabhu Energy Labs

J.E. Fox
Ph.D., Senior Consultant, Mine Ventilation Services, a business unit of SRK Consulting, USA

ABSTRACT: One of the largest sources of methane (CH4) emissions is ventilation air
methane (VAM) from underground coal mines. Prabhu Energy Labs is developing the “Oxi­
perator,” intended to be a significant step in achieving low-cost, scalable VAM destruction via
oxidation. The Oxiperator is a patented, all-metal, porous heat exchanger for oxidizing CH4.
For VAM applications, Oxiperator units will eliminate CH4 at concentrations as weak as
0.2%, all the way up to 100% (by volume in air). For CH4 concentrations 1.5% or higher, the
Oxiperator can be coupled to a gas turbine to generate power. A 40,000 CFM Oxiperator
operated out of a 40-foot shipping container can be quickly delivered and installed. Compared
to a Regenerative Thermal Oxidizer (RTO), the Oxiperator will have a much smaller foot­
print, cost less, and perform better. Oxiperators VAM air always flows continuously in the
same direction, avoiding the complicated pipes and poppet valves needed for RTO flow rever­
sal every few minutes. This paper discusses Oxiperator design, the benefits of high temperature
3D additive manufacturing, lab-scale/field-scale experimental results and future plans.

1 INTRODUCTION

Underground coal mine VAM accounts for roughly 40 million tons of CH4 a year globally .
Using a modest GWP (global warming potential) of 25, this is equivalent to over a gigaton of
CO2e (carbon dioxide equivalent). VAM concentration in air varies considerably from mine to
mine and hour to hour. For many mines, VAM is emitted in the range of 0.3% to 1.0% in air
(30,000 - 100,000 ppm by volume) . It is imperative that as much VAM as possible be eliminated
to meet global greenhouse gas targets. Several studies have documented the magnitude of the
issue in the US and elsewhere. These include the United States Environmental Protection Agency
(US EPA, 2021), and the International Journal of Coal Geology (C.Ö. Karacan et.al, 2011).
The only technology currently available for VAM destruction is oxidization using
a Regenerative Thermal Oxidizer (RTO). RTOs have seen limited adoption despite growing
international demand for coal VAM reduction as governments and stakeholders recognize the
damaging effect of CH4 on regional air quality and global climate change. Several papers have
been published on oxidation. The International Journal of CH4 Geology addresses RTO tech­
nologies (C.Ö. Karacan et al, 2011).
RTOs are large vessels filled with ceramic shells. The ceramic shells are maintained at about
1500° to 1600° F. Every few minutes the VAM flow directions must be reversed to keep the
temperatures hot enough for oxidation to occur. The flow reversal is achieved by large poppet
valves that must operate at high temperatures. RTOs take years to install, and the flow rever­
sal is challenging. They are difficult to move to other locations. By contrast, the Oxiperator
will be delivered and operated in a cargo container, ready of rapid deployment, or relocated
after new bleeder wells are brought into service.

DOI: 10.1201/9781003429241-48

462
 Prabhu Energy Labs is developing the Oxiperator, a new technology for oxidizing VAM.
Oxiperators have the potential to oxidize VAM in the range of 0.3% CH4 to 1.0% CH4 and
higher, at a lower cost and smaller footprint with relatively minimal operating costs.

2 THE OXIPERATOR

The Oxiperator is the first all-metal, high-temperature porous heat exchanger for oxidation of
low-concentration (weak) CH4 emissions mixed in air, protected by two US patents. Unlike
other oxidizers, the Oxiperator uses no catalyst or ceramics. The Oxiperator is made up of
two all-metal sections: a conventional heat transfer section and a porous section. The porous
section of the heat exchanger is the key to continuous oxidation. Due to its all-metal construc­
tion and modular design, the Oxiperator has the potential to take full advantage of the bene­
fits of current 3D printing technology using extreme temperatures.
Figure 1 is a schematic of the Oxiperator processing a mixture of 0.3% CH4 in air, which is
a widely measured VAM concentration from many coal mine shafts. The intake air dry bulb
temperature is assumed, for this illustration, to be 60° F.

Figure 1. Basic Oxiperator schematic for 0.3% CH4.

Incoming air is heated by passing through processed hot gas mixture in the traditionally
designed section of the first heat exchanger. At this point the temperature has been raised to
1500° F, at which point the CH4 is already oxidizing rapidly. As the CH4 laden intake air oxi­
dizes, the first heat exchanger temperature is raised commensurately. The gas mixture then
enters the porous section where fresh, intake gas mixture mixes through pores with return hot
gas, promoting remaining gas mixture oxidation. All processed gas mixture passes thusly
through the porous section, reaching 1650° F, thereby oxidizing any remaining CH4.
The hot gas mixture returns through the non-porous side and exits the Oxiperator at 210°
F. When 0.3% CH4 in air is oxidized, it raises the temperature by 150° F. The gas mixture
exhausts the system at 210° F, which is 150° F hotter than the intake gas mixture. At higher
CH4 concentrations, the Oxiperator can generate power without using previously researched
catalysts. At 1.5% CH4 the Oxiperator could potentially generate power without a catalyst.
An experiment with a catalytic combustor was reported in Energy (S. Su et. al. 2015). Other
operators have used RTOs to produce steam and run a steam turbine for downstream energy
recuperation.

463
3 TECHNOLOGY DEVELOPMENT AND SCALE-UP

A 25 cfm Oxiperator was successfully tested in California. Plans are underway for rapid scale-
up in size. Prabhu Energy Labs is currently manufacturing a 3D printed 100 cfm Oxiperator
that will be used to oxidize CH4 slip from a lean-burn gas engine. CH4 slip from lean burn
engines refers to the fugitive fuel gas that escapes from the exhaust of lean burn engines. It is
a major concern for the oil and gas industry where tens of thousands of lean burn engines are
employed in gathering and boosting stations, all of which have CH4 slip that is about 0.3% con­
centration. The work on CH4 slip will provide valuable information for VAM Oxiperators.
Upon successful testing, a 1,000 cfm Oxiperator will be manufactured for a larger lean burn
engine, then a 3,000 cfm Oxiperator for a one-megawatt lean burn engine.
The work on CH4 slip will provide valuable information for VAM Oxiperators. A modified
version of the 3,000 cfm Oxiperator will then be tested on VAM from a coal mine in the US.
The next Oxiperator will be 10,000 cfm, followed by a 40,000 cfm Oxiperator for demonstration
at a coal mine. At the commercial scale a 40,000 cfm Oxiperator system, capable of oxidizing
2,000 tons of CH4 a year, will be installed inside a 40 foot shipping container including the fan,
filter and controls. The Oxiperator system would be delivered to the mine site and commissioned
in several days, operating from the container installation using instrumentation and controls for
safety concious remote operation. Several 40,000 cfm Oxiperators could be placed around the
mine exhaust, which may be exhausting more than 300,000 cfm of airflow. As mining produc­
tion changes, Oxiperators can be moved to other mine site objectives, or swapped out if neces­
sary. Figure 2 is a conceptual description of an Oxiperator working at a coal mine.

Figure 2. Conceptual Oxiperator in storage container.

4 3D PRINTING

Oxiperators will be manufactured from a 3D printing process forming high temperature

metals. Rapid advancement of innovative 3D printing processes for complex metal compo­
nents is evolving quickly. Larger printing systems are being successfully developed and associ­
ated printing costs are declining as well. 3D printing allows for complex shapes and far
thinner wall thicknesses, which make heat transfer more efficient while reducing material
costs. Scaling up to larger VAM airflow volumes is possible since exchanger components can
be stacked together.

464
 Heat transfer rates using microchannels are much higher than traditional heat exchangers.
Ultra-thin (0.3mm thick) walls are 1/10th the thickness (and mass) of conventionally manu­
factured units but have been shown to withstand pressures up to 80 bar (1200 psi). 3D printing
will reduce high temperature materials by more than 90% compared to conventional manufac­
turing. The original Oxiperator prototype had wall thicknesses of 3 mm, with a total mass of
60 kg. Preliminary designs for an Oxiperator made with 3D printing have wall thickness of
0.3 mm, for a total mass of 4 kg. Figure 3 shows the potential order of magnitude difference
between traditional manufacture and 3D printing.

Figure 3. Oxiperator heat exchanger scale difference.

5 CONCLUSIONS AND SAFETY CONSIDERATIONS

Oxiperators can be used for many other applications. They can be installed at abandoned
mines to tackle residual gas emissions. For gas concentrations 1.5% or higher, Oxiperators
can be coupled to a gas turbine to generate electricity. If there is drainage gas or other CH4
available, VAM could be sweetened to 1.5% strength, still well below explosive range, and
used to generate power.
Applications are not limited to coal mines. For example, one of the challenging issues for
natural gas distribution networks is CH4 slip from thousands of lean burn gas engines that are
used to drive compressors. The CH4 slipping into the exhaust is commonly about 0.3%. The
project is already underway and will reduce the technology risk before the scale up in size and
before the coal mine demonstrations.

465
 Methane is explosive in concentrations between 5% and 15%. Coal mines must adhere to
stringent safety standards set by MSHA, the US Mine Safety and Health Administration. It is
essential to obtain prior MSHA approval of all ventilation specific safety systems before an
Oxiperator is installed at a coal mine. It is planned to keep the Oxiperator at a safe distance
from the mine exhaust, with a sufficient “safety” gap between the mine exhaust and the Oxi­
perator inlet. This distance will allow some of the VAM to escape, but it is considered neces­
sary to maximize safety when workers are underground. Several monitors will also sense CH4
concentrations and the Oxiperator will be rapidly isolated as concentrations approach to the
explosive range. Methane sensors would be installed at the bottom of the shaft, if possible,
providing early warning of a gas concentration rise. This early warning will allow precious
moments for protective safety action at the Oxiperator.
According to the United States Environmental Protection Agency (EPA, 2021), 1 tonne
CH4 in equivalent to 25 tonnes of CO2. Also, the EPA has recommended that coal mining
operators should be charged an empirical CH4 emission (metric tons of CH4 released per ton
of coal mined) rate multiplied by tons of coal mined per year. The suggested cost for each
metric ton of CO2 released into the open atmosphere is currently $51. The EPA has intro­
duced legislation that will nearly quadruple that cost to $190 per tonne of CO2 equivalent by
the end of 2024! If one 40,000 cfm VAM Oxiperator destroys 2,000 tonnes of CH4 a year,
which would potentially save the coal mine operator US$9,500,000 annually!

REFERENCES

C. Ö. Karacan, F. A. Ruiz b, M. Cotè, S. Phipps: “Coal mine methane: A review of capture and utiliza­
tion practices with benefits to mining safety and to greenhouse gas reduction”. International Journal
of Coal Geology, 2011.
S. Su et al.: “A 25 kWe low concentration methane catalytic combustion gas turbine prototype unit”.
Energy Magazine, 2015.
US EPA Report: “Ventilation Air Methane (VAM) Projects in the United States: Barriers and Potential
Opportunities”, 2021.

466
Mine heat
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Findings and learnings from thermal parameter studies at four

LKAB sites

F.K.R. Klose, A.L. Martikainen & T.H. Jones

Luossavaara-Kiirunavaara Aktiebolag, Malmberget, Sweden

ABSTRACT: As part of its strategic transformation, Luossavaara-Kiirunavaara AB (LKAB)

is currently evaluating an expansion of activities at four of its sites located in northern Sweden.
The project encompasses production rate and depth increases at the existing underground
mines, a surface to underground transition, as well as a potential greenfield operation. Ventila­
tion modeling is part of the evaluation process and requires an extensive set of inputs. This
paper details the collection and assessment of geothermal gradient and thermal rock properties
for each site. These parameters were not previously available despite being essential inputs for
ventilation modeling and impacting ventilation designs. Geothermal gradients were determined
through depth-temperature logging of select boreholes. Core samples extracted during explor­
ation were analyzed using various analytical approaches, with multiple samples for each of the
main site-specific lithologies. Results show depth and location-dependent differences between
geothermal gradients as well as differences in thermal properties obtained, which are compared
to values calculated using modal composition data for a limited set of lithologies. The potential
causes of these differences and general implications for ventilation modeling are discussed in
addition to recommendations for future thermal parameter studies.

1 INTRODUCTION

State-owned Luossavaara-Kiirunavaara Aktiebolag (LKAB) currently operates mines at three

sites in northern Sweden (Figure 1). Two mines, located in Malmberget and Kiruna, are sublevel
caving operations, while Svappavaara consists of two pits, Leveäniemi and Gruvberget. Only
Leveäniemi is presently operated. LKAB is currently evaluating the expansion of its operations
as part of its transformation process. While Malmberget and Kiruna are to expand to depth
and increase production rate, the Svappavaara underground transition and the development of
the Per Geijer deposit north of Kiruna are being evaluated. Ventilation modeling and simulation
are an essential part of the process to evaluate future needs for mine heating and cooling. Mine-
specific input parameters are required for accurate simulations. Previous work by Parasnis
(1975, 1982) on geothermal gradients in the nearby areas was considered too limited to serve as
reliable inputs for climate modelling. The thermal parameters of site-specific rocks were essen­
tially non-existent. Given the targeted sizes, depths, and production rates of the (future) mines,
it was deemed necessary to determine site-specific parameters. This paper outlines the work
done and shares findings of the parameter studies completed for all four sites.

2 GEOTHERMAL GRADIENT

The geothermal gradient is derived from virgin rock temperature (VRT) measurements. Such
measurements can be performed in short boreholes underground or in long exploration boreholes.
VRT measurements at the Malmberget mine showed that short boreholes were less suitable for

DOI: 10.1201/9781003429241-49

469
Figure 1. Overview of approximate site locations. Modified from OpenStreetMap (CC BY-SA 2.0).

obtaining a geothermal gradient, so only the long borehole method was continued at all other
sites. The quasi-continuous depth-temperature logs were obtained using a commercial tempera­
ture logging device in a manner identical to that described in Klose, Jones and Martikainen,
2022.

2.1 Borehole selection and measurements

A borehole selection methodology was established that looked at factors such as inclination and
potentially weak borehole sections, as outlined in Klose, Jones and Martikainen (2022). The
selected boreholes were all underground and were taken mainly from levels 600, 970 and 1250.
A similar strategy was used to pre-select underground exploration boreholes at Kiruna.
A general pre-selection was made from a borehole database. The selection procedure used cri­
teria similar to those used for Malmberget, though using site-specific selection criteria to better
estimate hole stability. Candidate boreholes were visualized to check for their exact location and
any significant deviation. Measurements in deep exploration boreholes in Kiruna were restricted
to levels 1375 and 1385. Measurements from higher levels, as in Malmberget, was not possible
due to caving. To minimize time and effort, measurement sites were identified based on the
availability of steeply dipping boreholes and the location within the area of interest.
This approach was continued for the surface boreholes of Per Geijer and Svappavaara,
which are mostly located in difficult to access areas. The primary criterion for Svappavaara
and Per Geijer was therefore accessibility, spread, and date of drilling, with boreholes from
recent campaigns assumed to have a higher probability of a successful measurement. An add­
itional criterion for Svappavaara was to have boreholes targeting both the Leveäniemi and
Gruvberget deposits. No measurements were conducted inside the pits.

2.2 Results
In general, data was processed as described by Klose, Jones and Martikainen (2022). The first
40m of borehole data was removed to eliminate surface temperature effects. The geothermal

470
gradient was then determined as a line of best fit through the remaining downhole and uphole
data. True depth corrections were applied to all measurements using data on the encountered
water levels and borehole collar elevations. Collar elevations of the Malmberget and Kiruna
boreholes refer to each mine’s reference coordinate system. This was corrected during post-
processing to instead use sea level as a reference elevation, ensuring comparability between
sites (Figure 2, expanded in Figure 3). Borehole data for surface boreholes in Svappavaara
and Per Geijer referred to the true elevation above sea level.

Figure 2. Compiled depth-temperature profiles at 40 m below collar and referenced to sea level.

Data was collected from a total of 13 boreholes on levels 600, 970 and 1250 in Malmberget.
A good spread of data was achieved despite some targeted areas no longer being accessible and
blockages in boreholes preventing full-length measurements. Three boreholes were measured from
level 600 to complement measurements on lower levels and to help establish a gradient trend.
A total of seven boreholes were successfully surveyed at Kiruna. Five of these boreholes are
located on the hanging wall side with boreholes directed towards the orebody. The remaining
two boreholes are on the foot wall and are dipping away from the orebody. Measurements
spanning up to 1000 m in borehole length were possible at nearly all selected sites and differ­
ent gradients were observed for the hanging wall and foot wall, which indicates that geology,
and thus distinct thermal properties, can play a greater role than initially thought.
At Svappavaara, a total of six boreholes were surveyed, with the Leveäniemi and Gruvber­
get deposits each targeted by three boreholes. Four boreholes were located outside of the
mine’s industrial area, while the two remaining boreholes were located close to the active pit
but targeting Leveäniemi and Gruvberget, respectively. Given that the exploration holes are
depth-limited according to the depth of the planned mine extension, the measurements were
limited as well and only the final sections of the boreholes were used to determine
a geothermal gradient. The start of these final sections was determined by visually identifying
the depth where a change in gradient occurred.
The same approach was used for the three boreholes surveyed in the Per Geijer exploration
area. All three boreholes are located on the hanging wall side. Poor accessibility is the main
reason for the limited number of boreholes surveyed.

471
Figure 3. Site-specific depth-temperature profiles at 40m below collar and referenced to sea level.

2.3 Discussion
Boreholes closer to the pit had significantly lower water levels than those further away. Just as
it is important in an underground mine to measure thermal gradient away from the influence
of the underground mine infrastructure, in a surface mine, it is important to intersect depths
that are below the currently active pit. Specifically, it is important to measure an area
unaffected by the cone of depression resulting from pit dewatering.
Consistencies in temperatures between different sites are noted when viewed with sea level
as the common reference elevation (Figure 2). The depth-temperature profiles of the Svappa­
vaara measurements appear consistent with that of from level 600 at Malmberget. The Per
Geijer measurement profiles seem consistent with hanging wall measurements at Kiruna.
These are four different mineralizations found in the four similar, yet variable, geologies
located up to 75 km apart. Although the deposits are at different stages of mine development,
use different mining methods, and have varying depths, the overall geothermal gradients are
the same for all of them.
This indicates that on a macro-scale, geothermal gradient is quite independent of human
influence, except for the role of groundwater, and that the gradient between the different sites
is less variable than originally thought. Mines spaced so closely together can use the same

472
geothermal gradient, generally. This will be true unless there is a clear boundary condition
that changes the base-state in which the mine is developed. Examples of this might be presence
of a localized highly radioactive geology, known geothermal activity limited by a geological
boundary, or other such factors.
However, on a micro-scale, used for mine ventilation planning and simulation, the role of
geology becomes more important. On this scale, variable geothermal gradients within one
mine can exist and are influenced by geological properties, which are discussed in greater
detail in the next section.

3 THERMAL PARAMETERS

Apart from the analyses of thermal conductivities by Parasnis (1975, 1982), little information
is available on the thermal properties of the host rocks in LKAB mines. For a higher degree
of confidence and more accurate consideration of site-specific conditions, thermal properties
need to be determined. The parameters are related to one another through density as follows:

Where α is thermal diffusivity (m2 s-1), λ is thermal conductivity (W m-1 K-1), cp is specific
heat (J kg-1 K-1), and ρ is density (kg m-3).
Conductivity is typically measured using the divided bar (DB) method. Specific heat can be
measured using calorimetric methods. With the density value obtained through Archimedes’
principle, thermal diffusivity of the sample can be determined using the Equation 1. Conduct­
ivity can also be measured using Thermal Conductivity Scanning (TCS). This method requires
minimal and non-destructive sample preparation for core sections. Thermal Diffusivity Scan­
ning (TDS) can be obtained concurrently, which is why the method will be referred to as
“TCS/TDS”. Specific heat can be calculated once the density value is known by rearranging
Equation 1.

3.1 Sampling methodology

A clear distinction can be made between the foot wall and hanging wall lithologies for Kiruna
and Per Geijer. Only a simplistic geological distinction is possible for Svappavaara, which is
why orebody boundaries were used to indicate the hanging and foot walls. This approach also
applies to Malmberget mine but is complicated by the mining of multiple orebodies and the
potential future location of infrastructure. The geology for Malmberget is therefore simply
referred to as “host”.
Sampling for the thermal parameter analysis was influenced by the variability in operating
procedures, the availability and practicality of acquiring the samples, and concern for safety
during sampling. Sampling of exploration core in Malmberget focused on the 5-6 most
common lithologies, while in Kiruna the major focus was on obtaining data for the hanging
and foot wall rock types. Core for Per Geijer and Svappavaara was selected by geologists and
exploration technicians to sufficiently represent the most common lithologies of each mine.
Multiple samples for each lithology were selected from both the foot and hanging wall. The
analyzed core was generally of BQ or NQ diameter, with a small number of AQ-sized core in
the sample set.

3.2 Sample analysis and results

This study considers a total of 71 samples that were analyzed externally by two laboratories,
one employing the TCS/TDS method and the other the DB method. The first laboratory used
the TCS/TDS method to analyze 20 Malmberget host rock samples and 15 Kiruna samples

473
(nHW = 3; nFW = 12). The second laboratory used the DB method to analyze 20 Svappavaara
samples (nHW = 10; nFW = 10) and 15 Per Geijer samples (nHW = 5; nFW = 10). The results are
summarized as boxplots in Figure 4, with Malmberget, Svappavaara, Kiruna and Per Geijer
respectively indicated with “M”, “S”, “K” and “P”.

Figure 4. Comparison of laboratory results across all sites and methods, showing distribution of sample
a) thermal conductivity, b) thermal diffusivity, c) specific heat and d) density.

The reported standard procedure for the TCS/TDS method was for each sample to be
scanned at least two times, with the sample being rotated or shifted the second time. Mean
values were calculated for each sample. When measurements were potentially affected by
inhomogeneities such as inclusions or banding, samples were reported being split into separate
scanning sections or the scans being repeated on a different side of the specimen. Grain size
and cavities was another reported issue. Scanning biases were only reported for core and are
presumed lower for massive, sawed samples due to the more ideal surface and depth.
For the divided bar method, disks were cut from homogeneous sections of the core. At
times, the cut samples were found to not meet the method specifications. Lack of homogen­
eity, smoothness, thinness, and breakage reported as the most common issues. In these cases,
new samples were cut that met the required testing criteria. It is important to note that all
disks were cut to have the same thickness, regardless of the diameter of the core.

3.3 Comparison of methods and sample dimensions

A total of 10 Malmberget samples were reanalyzed using the DB method to provide an initial
comparison of the methods. The comparison in Figure 5 shows that thermal conductivities
obtained using the divided bar method are generally higher, at nearly identical density.

474
 The core diameter (BQ or NQ) had no noticeable influence on thermal parameters as deter­
mined by the divided bar method as shown in Figure 6.

Figure 5. Comparison of results for the same Malmberget samples analyzed using the TCS/TDS and
DB methods, showing distribution of a) thermal conductivity, b) thermal diffusivity, c) specific heat and
d) density.

Figure 6. Comparison of results for different core diameters for the DB method, showing distribution
of a) thermal conductivity, b) thermal diffusivity, c) specific heat and d) density.

A direct comparison of results for different drill core diameters analyzed for identical litho­
logical groups indicated that measured values of thermal conductivity and specific heat are
within ±5%. The value spread can be explained by slight variations in mineralogical compos­
ition, as the samples originate from different sections and boreholes. A similar assessment of
size effects for the TCS/TDS method was not possible with such a limited sample set.

3.4 Comparison with calculated parameters

Rock thermal parameters can also be estimated through calculations. This is possible if the
constituent minerals, their volume or mass fractions, and individual thermal properties are
known. It was possible to estimate parameters for a limited number of samples for the Kiruna
mine, the only mine with the required data.
Sandberg (2018) conducted modal analyses of 19 thin sections of Kiruna foot and hanging
wall samples and divided the identified minerals into more commonly and less commonly

475
identified ones. Sandberg also provided a possible range of volume fractions for each mineral
identified in the thin sections. The lower and upper boundaries of each range were separated
by 10 percentage points in most cases. Only the more common minerals listed by Sandberg
(2018) were used for the purpose of calculating the thermal properties because the volume
fraction was greater than 1% for only a few of the more uncommon minerals. The properties
of the high purity mineral samples were based on data compiled by Waples and Waples
(2004), Čermák and Rybach (1982) and Boulanger-Martel, Bussière and Poirier (2018), citing
Côte and Konrad (2005) as seen in Table 1.

Table 1. Compiled properties of component minerals for calculation of thermal parameters.

Mineral ρ* cp* λ† α* Comment
-3 -1 -1 -1 -1 -6 2 -1
- kg m J kg K Wm K 10 m s -

Amphibole
Actinolite 3200 705 3.483 1.54
Hornblende 3080 710 2.85 1.30
Amphibole
Actinolite 3200 705 3.483 1.54
Hornblende 3080 710 2.85 1.30
Anhydrite 2900 585 4.76 2.81
Biotite 3000 770 1.17 0.51
Calcite/“Carbonate” 2745 815 3.57 1.60 Calcite assumed
Chlorite 2800 600 5.14 3.06
Clinopyroxene Augite 3380 670 3.82 1.69 Augite assumed
Epidote 3400 780 2.82 1.06
Feldspar Mix of feldspar minerals assumed
Albite 2600 730 2.31 1.22
Orthoclase 2600 628 2.31 1.41
Hornblende 2580 700 2.49 1.38
Hematite 5150 620 11.3 3.54
Ilmenite 4750 849 2.38§ 0.59
Magnetite 5175 586 5.10 1.68
Pyrite 5010 510 19.2 7.51
Quartz 2648 740 7.69 3.92
Titanite 3500 810 2.33 0.82

* Waples (2004), †Cermak and Rybach (1982), §Boulanger-Martel, Bussière and Poirier (2018), citing
Côte and Konrad (2005).

Using the Sandberg (2018) data and the compiled thermal parameter values for different
minerals, average volume fractions were then calculated for each mineral with the reported
lower and upper boundaries as input. The individual volume fractions were then cumulated.
This cumulative value was used to correct each individual volume fraction so that the new
cumulation would reach exactly 100%.
Thermal conductivity was calculated using Equation 4, after information compiled by
Robertson (1988). All other properties, including density, were calculated using Equation 2
after substituting thermal conductivity with the desired property.

476
 Where λ is thermal conductivity (W m-1 K-1), n is volume fraction (1), p is parallel and s is
serial mineral grain arrangement.
Two comparisons can then be performed. The first is a direct comparison between calcu­
lated (“Calc.”) and measured (“Meas.”) values for four thin sections and their corresponding
core samples, seen in Figure 7. The second, indirect comparison in Figure 8 focuses on the
properties of the foot and hanging wall and includes a greater number of samples, leading to
an averaging effect.

Figure 7. Direct comparison of calculated and measured properties for four Kiruna thin sections and
their core samples, showing distribution of a) thermal conductivity, b) thermal diffusivity, c) specific heat
and d) density.

Figure 8. Comparison of TCS-measured and calculated foot wall and hanging wall properties for
Kiruna samples, showing distribution of a) thermal conductivity, b) thermal diffusivity, c) specific heat
and d) density.

3.5 Discussion
Two different methods of analysis were used to obtain thermal parameters for the selected
samples. The TCS/TDS method is applied lengthwise to a core sample while the divided bar

477
method uses a disk cut from a homogenous core section. Especially the latter method can be
considered firmly established, while core is a readily available source of information. Differ­
ences in laboratory results could be explained by a size effect, with TCS/TDS results being
seen as more representative of a rock mass due to the inclusion of defects. However, this is
a cautious assessment. In general, laboratory analyses and calculated values do not reflect the
conditions of the rock mass in-situ. However, in-situ studies are potentially more costly and
time-consuming than laboratory analyses. For this reason, further study of the impact of core
diameter and sample geometry on the results is suggested. This would improve overall confi­
dence in the results for simulation applications. For the divided bar method at least, core
diameter does not appear to impact results.
Average thermal conductivities highlight similarities and differences between the geologies
of each site. Conductivity is the only property that is measured directly by both the TCS/TDS
and DB methods. The results for this property show that similarities appear to exist between
samples from Malmberget, Svappavaara and the Kiruna foot wall. The Kiruna hanging wall
and Per Geijer foot wall samples are also somewhat similar, although the low number of sam­
ples for the former influences this assessment. The Per Geijer hanging wall is distinct from the
other sample sets.
Meanwhile, the large range of densities for Malmberget points towards distinct properties
for a number of the sampled lithologies. This is also seen in the extended range seen for the
Svappavaara foot wall, which in terms of sample size and sampled lithologies, is similar to the
hanging wall. The otherwise consistent average densities and ranges suggests more homoge­
neous geologies contained within the sample sets.
Calculations of thermal parameters seem to be influenced by several factors. The first factor
is the identification of minerals and volume fractions as well as the choice of mineral proper­
ties to represent generic groupings such as amphibole and feldspar, which contribute to large
volume fractions. Figure 7 appears to show inconsistencies between the measured and calcu­
lated parameters, but having only four data points for comparison, sample size is likely to
play a significant role in this case. This changes when the sample populations are increased,
leading to an averaging effect visible when comparing Figure 7 with Figure 8. The results
show that the calculation of parameters from modal values, if available, could be a valid
option. It is an example of how data that is routinely collected in a mineral processing setting
can be used to generate input parameters for mine climate models.
The calculation of thermal parameters assumes a homogeneous rock mass that has no pores
or pore fluids. The DB method, which analyses a uniform thin section may provide results
that are more similar than the TCS/TDS method, which is conducted on a full piece of rock
core. This is supported by the comparisons resulting from this work and is evidenced in Fig­
ures 5 and 8. While the results seem to support this conclusion, the small sample size means
that further investigation regarding the comparison between the TCS/TDS method, the DB
method, and thermal parameter calculation is required for validation.

4 APPLICATIONS AND IMPLICATIONS FOR CLIMATE MODELING

The VentsimTM DESIGN (Howden Group Ltd 2022) software package is used for climate
modeling of underground mines. Global settings for geothermal gradient and thermal param­
eters can be overridden by applying custom rock types and localized changes in the geother­
mal gradient to individual airways using VentsimTM DESIGN presets. The filtering of airways
greatly assists this process.
Two gradients were applied in Malmberget. The first of these was applied for elevations
above level 1000 and consists of the only valid measurement for this elevation range, which is
also offset from the other measurements. Due to their consistency (Figure 4), these were used
to generate a second gradient that was applied to elevations below 1000. This was seen as
a conservative measure, since part of the downcast air in the model would interact with a rock
mass with a higher VRT (Klose, Jones and Martikainen 2022).

478
 Different gradients were found for the Kiruna hanging wall and foot wall, which simplifies
the application of a climate simulation. A global gradient was applied to Svappavaara and Per
Geijer as no underground infrastructure is currently present and surface borehole data was col­
lected to only a limited depth. For Svappavaara, the mean was calculated from four gradients
determined from the bottom-most stable sections, while the other two were excluded due to
insufficient length. As only two of three Per Geijer measurements are without major surface
influence and show different gradients, the Kiruna hanging wall gradient was used instead. This
was deemed appropriate given the proximity of the deposit to Kiruna, the overall gradient
macrotrend as shown in Figure 3, and similar thermal conductivities as shown in Figure 4.
A rock surface temperature was calculated as a model input as a final step in gradient
implementation. This is used as an input to calculate VRT. The rock surface temperature is
hypothetical and can be interpreted as the intersection of the geothermal gradient line with the
mine grid surface datum. The VRTs calculated from these inputs should coincide with actual
measurements. This is applicable to Malmberget and Kiruna. While the hanging wall gradient
and surface rock temperature of the Kiruna hanging wall was applied for Per Geijer, the aver­
age surface rock temperature referring to sea level was implemented for Svappavaara.
After the implementation of the gradients for each site, thermal parameters were applied.
For Malmberget, general thermal properties were calculated based on logged meters available
in a core logging database using Equation 5 (Klose 2022).

Where P is the rock property to be calculated, xi is the fraction (1) of lithology i out of the
n main lithologies, pi is the property of lithology i, and p̄ is the arithmetic average of the prop­
erty for all lithologies. The units of P, pi and p̄ are identical and depend on the property to be
calculated.
The same approach and database were used to determine more individual thermal proper­
ties for two target areas of the mine expansion, Area 1 and Area 2, by filtering the database
accordingly. These properties were applied to airways located in the specific area only, while
all other airways were assigned the general values. For Malmberget, the results of the reana­
lyzed samples were included.
In the case of the Kiruna mine, nearly all infrastructure is located in the foot wall. The clear
geological distinction from the hanging wall greatly simplified the assignment of thermal prop­
erties to airways, as the hanging wall could essentially be excluded from further consideration.
For the hypothetical Svappavaara and Per Geijer operations, average foot wall and hanging
wall properties appeared adequate at this early stage of evaluation.
The properties implemented in each climate simulation model are summarized in Table 2
below.

Table 2. Implemented rock mass properties.

Surface Datum
Site Location Gradient λ cp α
Rock Temp
- - - °C/100m °C W m-1 K-1 J kg-1 K-1 10-6 m2 s-1

Malmberget Area 1 Below 1000 1.59 -5.2 2.40 685 1.24

Area 2 Below 1000 1.59 -5.2 2.33 684 1.21
Other Above 1000 1.35 -1.0 2.35 685 1.22
Below 1000 1.59 -5.2 2.35 685 1.22
Kiruna FW FW 1.67 -5.9 2.43 691 1.33
HW HW 1.47 -3.3 2.94 655 1.48
Svappavaara FW General 1.39 6.9 2.39 756 1.14
HW General 1.39 6.9 2.22 744 1.09
Per Geijer FW General 1.47 -3.3 3.10 688 1.68
HW General 1.47 -3.3 3.93 690 2.12

479
4.1 Case study: Malmberget mine
The impact of implementing new parameters was evaluated through a case study, for which
the most recently updated VentsimTM model for Malmberget mine was selected. The primary
focus of this section is to highlight the approach taken towards this case study as modelling
work and calibration is ongoing.
The Malmberget ventilation system features several ventilation shafts and a total airflow
capacity of 2300 m3 s-1. Approximately 18% of this capacity is currently directed from two
shafts (Shaft 1 and Shaft 2) towards Area 2. Area 1 is supplied by its own shaft, which has
a capacity of 540 m3 s-1.
Before implementing the parameters, the model settings were evaluated. Particular attention
was paid to the mine grid elevation, which acts as reference elevation to parameters relevant
to the implementation of the geothermal gradient as well as the input surface conditions
(Howden Group Ltd n.d.). In the original model, this elevation was set to the approximate
elevation of the nearest active weather station operated by the Institute of Meteorology and
Hydrology of Sweden (Sveriges meteorologiska och hydrologiska institut, SMHI). To allow
appropriate comparisons to be made, the mine grid and surface datum rock temperatures
were adjusted to the Shaft 1 collar elevation. The virtually identical VRT calculated for differ­
ent airways indicated the success of the mine grid adjustment.
A dataset from summer 2022 was used in this case study. Near-shaft dry bulb temperature
DBT and wet bulb temperatures WBT (°C) and barometric pressures BP (kPa) were collected
using a Kestrel 5500 weather meter in logging mode. Data was logged every 20 minutes (Shaft 1)
and 30 minutes (Shaft 2). An identical instrument was used to perform underground measure­
ments within the same timeframe at locations shown in Figure 9. The measurements of the dataset
focused on airflow directly from the local shaft connections. Considering the depth and compli­
cated ventilation route, it is likely that DBT (°C) and WBT (°C) are over- and underestimated,
respectively, compared to the airway values. No mining or development activity was observed in
the area during the measurements and all measurements were performed the same way.
As the times of surface and underground measurements did not align directly, surface condition
inputs were calculated using weighted averages by assigning underground measurements to the
closest surface measurement time. Weighted averages were calculated for DBT (°C), WBT (°C)
and BP (kPa), which were then used as model inputs. This was deemed acceptable because of the

Figure 9. Partial overview of Area 2 in VentsimTM, showing locations of underground spot measurement.

480
steady surface conditions during the measurement period, the depth of the underground measure­
ment locations and the thermal flywheel effect contributing to the stability of the underground
climate. The calculated values for surface conditions at Shaft 1 were entered into the model with
the surface temperature adjustment setting activated. The success of this approach indicated by
simulated inlet temperatures at Shaft 2 being in general agreement with actual measurements at
that location.
Heat simulation results for dry bulb and wet bulb temperatures (°C) were compared following
the adjustment of input parameters, which is summarized in Table 3. The comparisons covered the
adjustment of rock thermal parameters and gradient individually and in combination. Only airways
with airflow on levels within the depth range of performed geothermal gradient measurements were
part of the assessment. Simulation results were cross-checked with underground spot measurements
taken between levels 1000 and 1250 in Area 2. The expectation was an improved agreement
between actual measurements and simulated values at these locations. Measured and simulated
values would be identical if the absolute difference between them were 0°C, whereas a positive dif­
ference would indicate that the simulated temperatures are higher than those measured.

Table 3. VentsimTM parameters used in the original and adjusted models, with reference to values used
during the parameter implementation.
Adjusted
Global Parameter Unit Original Adjusted (parameter implementation)

Geothermal gradient °C/100m 1.0 1.0 *

Rock Density kg m-3 2700 2700 †

Rock Specific Heat 10-6 m2 s-1 790.0 790.0 *

Rock Thermal Conductivity W m-1 K-1 2.00 2.00 *
Rock Thermal Diffusivity 10-6 m2 s-1 0.938 0.938 *
Surface Datum of Mine Grid m -264 -35 -35
Surface Datum BP kPa 101.0 95.2 95.2
Surface Datum Rock Temp °C 15.0 12.7 *
Surface Datum DBT °C 15.0 9.1 9.1
Surface Datum WBT °C 10.0 7.3 7.3

* Multiple parameters used, see Table 2 for details.

† Multiple, calculated using Equation 1 and values from Table 2 to achieve respective specific diffusivity value.

The simulation results (Figure 10) show that the rock thermal parameters themselves affect nei­
ther DBT (°C) nor WBT (°C) beyond 0.1°C. Likely explanations for this behavior are similarities
in certain values such as thermal conductivity, as well as the applied airway age reducing the heat
flow into the opening due to cooling of the surrounding rock. This was not investigated further.
On the other hand, changing the geothermal gradient generally decreases both DBT (°C) and
WBT (°C). Where the simulation approaches the measured DBT (°C) values, such as at Site 3 and
Site 5, WBT (°C) generally increases. This could be explained by increased evaporation rates.
In general, the temperature changes can be explained by changes in simulated VRT. To
illustrate this, the surface rock temperature and presumedly linear gradient for the adjusted
model result in a theoretical VRT of 24.9°C. On the other hand, extending the other gradients
down to the current main level of 1250 leads to the respective theoretical VRT values of 15.4°C
using the gradient “Above 1000” and 14.1°C for the other. Changing the gradient therefore has
consequences for the heat transfer between the air and the rock mass from the moment air
enters the mine. This indicates that closer scrutiny of near-surface rock mass temperatures is
necessary, possibly requiring a seasonal adjustment.
Finally, some improvement of temperature values can be observed for Sites 1, 2 and 4
when both thermal parameters and gradients are implemented. This emphasizes not only the
importance of the gradient as a key influencing value, but also that the implementing of
mine-specific parameters and gradients should not be considered separately.

481
 Anomalies such as DBT (°C) being lowered significantly such as at Site 2 and Site 4 when
applying the gradient only, or the greatly increased WBT (°C) at Site 2 compared to the actual
measurements, require further investigation and suggest the need for recalibrations following
the implementation of new parameters.

Figure 10. Difference in simulated a) DBT (°C) and b) WBT (°C) relative to measured values.

5 SUMMARY AND RECOMMENDATIONS

Generalized thermal gradient values taken from other nearby operations are suitable for use
during the first stages of ventilation planning for a new mine with no pre-existing data, as
long as extenuating circumstances are not present. For existing operations with experience,
data, and a need for more precise local modeling for expansion purposes, geothermal gradient
measurements should be conducted in boreholes intersecting rock where future infrastructure
is expected to be located. This ensures the highest relevancy of geothermal gradients and sim­
plifies application in climate models. Thermal gradients determined from surface borehole
data can be used to infer a temperature at a mine-relevant depth, although the extrapolation
of the gradient is affected by uncertainty in measurements. As such, measurements should go
as deep as possible and data collection near active open pits should be avoided, unless meas­
urements beyond the extent of the pit and the cone of depression can be guaranteed.
The collection of thermal parameters is a complex endeavor, as illustrated by the case studies
of the two existing and two potential underground mines. Knowledge of the present and

482
potential future mining situations can help in obtaining relevant parameters. Use of core sam­
ples to obtain thermal parameters for individual rock types allows the use of large sample col­
lections and later manual adjusting of possible rock mass properties. This is advantageous for
large mines with complex geologies.
The choice of laboratory method to obtain simulation parameters matters less than obtaining
representative samples of the main rock types of the rock mass in which infrastructure will be
placed. Most borehole meters originate from exploration boreholes that are drilled from the sur­
face or from underground exploration tunnels towards the orebody. Borehole location and direc­
tion need be considered during sample selection, as the approach of identifying the most common
lithologies through drilled meters alone can introduce a bias towards hanging wall lithologies.
Using a mine climate simulation software allows the assignment of multiple gradients and
rock types in the same mine. As the software is a tool that can be used to predict underground
conditions, it is important to know the software capabilities and limitations. These would
guide the application of the available data and can guide parameter collection efforts as well.
Implementing thermal parameters is potentially more complex than geothermal gradients
and largely depends on the level of detail required in the model. However, the Malmberget
case study shows that neither thermal parameters nor gradient should be viewed in isolation.
Ultimately, replacing placeholder thermal parameters and gradients with site-specific ones
can improve reliability and accuracy of climate simulation results for strategic planning pur­
poses. This is of particular value for potential greenfield sites such as Svappavaara and Per
Geijer. Existing sites such as Malmberget and Kiruna benefit first and foremost from the
transparency and understanding that comes with the updating process and require subsequent
model calibration checks to use effectively.

ACKNOWLEDGEMENTS
The authors gratefully acknowledge the assistance from LKAB personnel, particularly from the
R&D and Exploration departments, in realizing this work. The authors would also like to extend
their thanks to the anonymous reviewers for their valuable suggestions and improvements.

REFERENCES

Boulanger-Martel, V., Poirier, A., Côté, J. & Bussière, B. 2018. Thermal conductivity of Meadowbank’s
mine waste rocks and tailings. In GeoEdmonton 2018 – 71st Canadian Geotechnical Conference, Edmon­
ton, 23-26 September 2018. 8pp.
Čermák, V. & Rybach, L. 1982. Thermal properties: Thermal conductivity and specific heat of minerals and
rocks. In G. Angenheister (Ed.), Landolt-Börnstein Zahlenwerte und Funktionen aus Naturwissenschaften
und Technik, Neue Serie, Physikalische Eigenschaften der Gesteine. Berlin, Heidelberg, New York:Springer.
Howden Group Ltd 2022. VentsimTM DESIGN (Version 5.4) [Computer software]. Retrieved from
https://ventsim.com/.
Howden Group Ltd n.d. VentsimTM DESIGN User Guide Version 5.4. Retrieved from https://ventsim.
com/files/VentsimManual.pdf.
Klose, F.K.R. 2022. Heating and cooling in LKAB underground mines. MS thesis. University of
Leoben, Austria.
Klose, F.K.R., Jones, T.H. & Martikainen, A.L. 2022. Geothermal gradient determination for ventila­
tion and air conditioning modelling at Malmberget mine. In Proceedings of the Australian Mine Venti­
lation Conference, Gold Coast, 10-12 October 2022: 66–80.
Parasnis, D. S. 1975. Temperature phenomena and heat flow estimates in two precambrian ore-bearing
areas in north Sweden. Geophysical Journal International 43(2): 531–554.
Parasnis, D. S. 1982. Geothermal flow and phenomena in two Swedish localities north of the arctic circle.
Geophysical Journal International 71(3): 545–554.
Robertson, E.C. 1988. Thermal Properties of Rocks. Open-File Report 88–441.
Sandberg, H. (2018): A petrological investigation of the host rocks for the Kuj-Kiirunavaara ore. BS
thesis. University of Uppsala, Sweden.
Waples, D.W. & Waples, J.S. 2004. A review and evaluation of specific heat capacities of rocks, minerals,
and subsurface fluids. Part 1: Minerals and nonporous rocks. Natural Resources Research 13(2): 97–122.

483
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Scenario-driven evaluation of heat sources in underground

production scheduling

J.A. Buaba
South Dakota School of Mines and Technology, Mining Engineering and Management, Rapid City,
South Dakota, USA

E. Udofia & A. Newman

Colorado School of Mines, Mechanical Engineering, Golden, Colorado, USA

A.J. Brickey
South Dakota School of Mines and Technology, Mining Engineering and Management, Rapid City,
South Dakota, USA

ABSTRACT: Heat loads, associated with autocompression, geothermal gradient activity,

and mining equipment, can lead to high temperatures underground that require mitigation
and/or refrigeration in order to create a safe working environment for miners. In this paper,
temperature levels associated with heat generated from equipment and other sources are
incorporated into a medium-term production scheduling optimization model (Ogunmodede
et al, 2022). The resulting schedule considers temperature by production level and determines
refrigeration requirements, if applicable, based on the ventilation capacity, production rates,
and equipment usage. This tool can be used to evaluate numerous production and ventilation
scenarios to make production-related decisions. Using a realistic dataset from an underground
mine, numerous scenarios are evaluated to understand the impacts of various equipment heat
loads and ventilation settings to illustrate the impacts of heat on the production schedule.

1 INTRODUCTION

Heat build-up in underground mines, i.e., mines below the surface, is a safety concern for mine
operators because it may impact a miner’s health. The human body’s ability to cool down is con­
trolled by air temperature and humidity levels in an environment, i.e., thermoregulation; high air
temperature and humidity levels increase the potential for heat-related illnesses such as heat faint­
ing, heat exhaustion, heat rash, and heat stroke (Ryan & Euler, 2017). Further studies show
a negative correlation between heat stress (total heat load placed on the body) and cognitive per­
formance which may cause workplace safety incidents and affect worker productivity (Yeoman
et al., 2022). Heat is an innate part of all mining activities in both surface and underground mines.
Mining activities such as drilling, blasting, loading, and hauling use mechanized equipment, or by
the nature of the activity, generate heat which is transferred to the mine atmosphere. In surface
mines, the working areas are exposed to the surface atmosphere, and the heat generated is dis­
persed into the environment whereas in underground mines, a ventilation system is required to
control and manage the mine environment in order to create a safe working condition for the
miners. The Mine Safety and Health Administration (MSHA) defines a “hot” work site as
a combination of high air temperatures, humidity, radiation, and relatively low airspeed that
exceeds a wet bulb globe temperature of 26°C (79°F) (U.S. Department of Labor, 2012). This
temperature limit captures the convective, evaporative, and radiative heat exchange around the
work site. Apart from the heat generated by mechanized equipment, heat from the surrounding

DOI: 10.1201/9781003429241-50

484
rocks (strata heat) and autocompression, i.e., the rise in the temperature of air due to the increase
in pressure at lower elevations, increase the ambient air temperature in underground mines. Heat
is transferred between and/or within mediums through conduction, convection, and radiation in
the presence of a thermal gradient (McPherson, 1993a). In underground mines, the airflow is sup­
plied using large primary fans that could be on the surface or underground, and the air is circu­
lated through the mine drifts, i.e., the airways, to remove heat via convection. Booster and
auxiliary fans in conjunction with ventilation controls such as ventilation doors, stoppings, and
regulators are also used to distribute the airflow to the working areas. As the air moves from the
surface to the lower levels, some of its potential energy is converted to heat which increases the air
temperature (autocompression). The air temperature can also increase due to the heat emitted by
the strata and mechanized equipment via thermal convection and radiation as it circulates through
the mine drifts. The wet bulb temperature of the air, i.e., the temperature at which air can be
cooled by evaporation, increases at the rate of about 4°C (7.2°F) per 1000 m vertical descent
(Ramsden, et al. 2007). The maximum working temperature for most mining operations ranges
from 27.5°C (81.5°F) to 28.5°C (83.3°F) wet bulb temperature (Ramsden, et al. 2007). The wet
bulb temperature limit used in this study is 28 °C. This is based on the regulations of the host
country for the case study mine. Heat transfer from the strata to the air and vice versa varies by
the virgin rock temperature present in the mine. The heat generated from the equipment depends
on the rated engine capacity, fuel consumption, and the percentage of time it is utilized (van der
Walt, et al. 1983). Heat in underground mines may be reduced by (i) engineering controls, such as
mine planning, ventilation, and air conditioning, (ii) administrative controls and work practices
such as developing a work-rest routine, rotating personnel on hot jobs, no idling of equipment,
and (iii) providing personal protective clothing and equipment such as reflective clothing to stop
the body from absorbing radiant heat (U.S. Department of Labor, 2012). As mines become
deeper, the problems of heat become more critical, requiring an air conditioning system to control
the air temperature and humidity levels in the mine environment.
In this paper, we evaluate engineering controls, i.e., mine planning, ventilation, air condi­
tioning, and optimization to understand heat emissions and the impact on production in
underground mines. Temperature levels associated with heat generated from equipment and
other sources are modeled and incorporated into a medium-term production scheduling opti­
mization model developed by Ogunmodede et al. (2022). The scheduling model maximizes the
net present value while adhering to precedence, resource, and heat constraints. The resulting
schedule considers temperatures by production level based on ventilation capacity, production
rates, and equipment usage. We evaluate the usefulness of this model by analyzing various
production and ventilation scenarios to determine if heat can be managed via production
scheduling or if modifications to the ventilation system is needed.

1.1 Background
Underground mine planning is an iterative process to determine the economic viability of
a mineral deposit. The planning process consists of resource modeling, mine design, production
scheduling, ventilation planning, equipment selection, infrastructure design, reconciliation, and
financial evaluation which continues throughout the life of the mine operation (Bullock, 2011).
The information needed to develop the mine plans is typically based on historical operational
data and/or engineering estimates. As new information becomes available, the mine plans are
updated accordingly. One outcome of mine planning is a production schedule that considers the
time value of money to verify the economic viability of the mineral deposit. The production
schedule determines the start time and extraction sequence of the mineral deposit. The schedule
is subject to precedence constraints and resource constraints such as production (mining equip­
ment) and processing (crushing, milling) capacities. This process can be used to generate strategic
(long-term), tactical (medium-term), and operational (short-term) production schedules (Hall,
2017). Long-term production schedules define the overall business objective for the mining oper­
ation, e.g., capital projects and life-of-mine production based on corporate goals. Medium-term
production schedules guide the operation to achieve the long-term goals up to a five-year time
horizon at monthly or quarterly fidelity. Short-term production schedules define the day-to-day

485
operation of the mine and focus on achieving the targets set in the medium-term schedule. The
time horizon typically ranges from weekly up to two years at hourly, by shift, or daily fidelity.
This integrated approach to mine planning requires the use of complex tools such as operations
research techniques to make informed decisions to maximize the value of the mineral deposit.
Operations Research is a scientific approach using mathematical models to represent
a problem and analyzing the solution to make better decisions (Rardin, 2017). The process
begins with formulating or modeling the problem as close to reality as possible. The decision
variables to be determined are defined together with the parameters, objective function, and
constraints. The parameters are a set of known information about the problem, the objective
function maximizes or minimizes a function of decision variables, and the constraints set
a limit on the solution. Various algorithms are used to solve the model and the results are
analyzed to make meaningful conclusions that can be applied. The common types of models
used in operations research and applied in underground mining include linear programming
(LP), integer programming (IP), mixed integer programming (MIP), networks, non-linear pro­
gramming (NLP), constraint programming, and stochastic programming.

2 LITERATURE REVIEW

Operations research techniques were first applied in underground mine production scheduling to
determine a feasible schedule to meet the annual contract agreement for a copper mine (Williams
et al. 1973). The authors formulated a linear programming model to produce a “good” schedule;
however, it was difficult to define the qualities of a “good” schedule. They proceeded to develop
a tool that allowed a mine planner to evaluate multiple scenarios and select the “best” option.
Though optimality was not achieved, the resulting schedule showed significant improvement
over manual scheduling. Over the past few decades, improved computational capability, new
solution algorithms, and mathematical modeling techniques have allowed researchers to develop
integer programs that incorporates binary variables and allow for “yes-no” decision-making.
Trout (1995) formulated a mixed integer programming (MIP) multiperiod production scheduling
model for a sublevel stoping copper operation located in Mt Isa, Australia. The objective of the
model was to maximize net present value (NPV) by scheduling stopes for production over a two-
year horizon. The solution could not solve to optimality due to insufficient memory; however,
the model improved the NPV by 23%. Trout’s model was improved by Nehring and Topal
(2007) who formulated an additional constraint function to limit multiple fill mass exposures
without violating other constraints. This made the model more applicable to the sublevel stoping
method. The model was tested on a conceptual mine stoping operation. Their model was then
compared to a manual schedule which showed a 0.66% increase in NPV. Carlyle and Eaves
(2001) used an integer programming model to plan a production schedule for a sub-level stoping
operation at Stillwater Mining Company for a 10-quarter planning period, to maximize revenue
from mining platinum and palladium. The flexibility in their formulation allowed engineers to
investigate multiple production scenarios quickly. McIsaac (2005) used an MIP model to gener­
ate a long-term mine plan for a polymetallic mine to maximize NPV over four years at quarterly
fidelity. The model solved for 30 minutes; however, no mention is made if the solution obtained
was optimal. Little et al., (2008) revised Nehring’s and Topal (2007) model by reducing the
number of binary decision variables, which significantly improved solution time by 92% without
altering the solution. Smith et al. (2003) formulated an MIP model to produce a life-of-mine
schedule at Mt Isa. Their solution showed continuity of production and improvement in NPV
during the mine’s life when compared to the traditional method of planning using spreadsheets
and other commercial mine schedulers. Kuchta et al. (2003) implemented an MIP model to
schedule activities at LKAB’s Kiruna Mine in Sweden. The model minimizes deviation from
monthly planned production quantities while adhering to operational constraints. Other opti­
mization models implemented at LKAB used aggregation and decomposition heuristics to
improve model tractability (Kuchta et al. 2004; Martinez & Newman, 2011; Newman & Kuchta,
2007). O’Sullivan & Newman (2014) scheduled the underground Lisheen Mine, located in Ire­
land, with an MIP model. The objective maximizes discounted metal (zinc and lead) with

486
complex precedence constraints. Aggregation techniques were used to reduce variables and devel­
oped methods for assigning precedence to the various activities; however, aggregation was limited
because of the complicated precedence structures. Lisheen Mine utilized the model to make end-
of-mine-life production scheduling decisions.
In recent years, several researchers have made attempts to standardize the underground mine pro­
duction scheduling problem as a Resource Constrained Project Scheduling Problem (RCPSP)
(Brickey, 2015; Fava et al. 2012; Maybee, 2011). The RCPSP optimization model consists of an
objective function that maximizes NPV or minimizes cost, precedence constraints, and resource con­
straints. The decision variables can be binary, i.e., either 0 or 1, which determine when to start
a given activity. A novel algorithm for solving the linear programming (LP) relaxation of the
RCPSP problem has allowed researchers to model and solve more complex mine production sched­
uling problems (Bienstock & Zuckerberg, 2010). Terblanche and Bley (2015) formulated an under­
ground mine scheduling optimization problem as an MIP using the generic RCPSP structure. Their
formulation applies to selective mining where ore grade is highly variable. Extracting only the high-
grade areas in the mine reduces cost and improves mine profitability. Brickey (2015) formulated
a generalized underground mine production scheduling model using the RCPSP structure to maxi­
mize discounted gold ounces mined and determines the optimal sequence of activities in relation to
development, extraction, and backfilling of an underground mine with ventilation as a consumable
resource. The model formulation is flexible in that constraints can be customized and used in other
underground mines with similarly structured data. The model was solved using the OMP Solver
(Rivera et al. 2016) which utilizes the Bienstock-Zuckerberg algorithm (Bienstock & Zuckerberg,
2010) and a TopoSort heuristic (Chicoisne et al. 2012) to find the global optimal solution. Lopes
(2017) formulated a binary integer programming mine production scheduling model for a large-scale
underground hard-rock mine at daily fidelity to maximize net present value (NPV) using the model
structure developed by Brickey (2015). The author determined solutions for the model using the
OMP Solver and the sliding window heuristics to expedite solutions for scheduling problems. Huang
et al. (2019) developed a mixed integer linear programming (MILP) production scheduling model
for an underground cut-and-fill mine. The objective function maximizes NPV subject to mining and
processing constraints. The model was applied to two case study mines and the results showed a 9%
to 17% increase in NPV, respectively, when compared with the manual schedule. Brickey et al.
(2020) applied the RCPSP production scheduling model to determine a medium-term production
schedule for the Barrick Turquoise Ridge mine. Ogunmodede et al. (2022) formulated an MIP using
the RCPSP structure, to maximize NPV subject to precedence and capacity constraints with consid­
eration of heat generated by equipment, the strata, and autocompression. The authors generated
more realistic schedules and indicated when refrigeration was activated to mitigate heat emissions.
We extend the work of Ogunmodede et al. (2022) by analyzing various production and ventilation
scenarios to make better investment decisions and describe their usefulness to mine planning engin­
eers. More detail on the application of operations research in underground mine planning can be
found in Newman et al., (2010), Acuña and Lowndes (2014), and Chowdu et al. (2020).

3 METHODOLOGY

We apply the mathematical model developed by Ogunmodede et al. (2022) to a case study mine,
herein referred to as Mine X, to evaluate the impact of heat on the production schedule. Ogun­
modede’s model, i.e., model (R+), quantifies heat from three main sources in underground mines
using thermodynamic principles. These heat sources— (i) diesel equipment, (ii) autocompression,
and (iii) strata-rock— are modeled given the following assumptions: (i) the medium-term produc­
tion schedule with heat constraints is solved at steady state, meaning that heat from the previous
time period is not carried over to the current time period; (ii) the air velocity is constant, assum­
ing there are auxiliary fans on each level, (iii) the mass flow rate, ṁ , is constant on each mining
level; (iv) the relative humidity is constant throughout the mine; and, (v) sensible heat from
equipment is absorbed into the air via convection on any given mine level.
The heat from diesel equipment is quantified using GT-SUITE software (Gamma Technolo­
gies, 2021), a one-dimensional modeling software that utilizes unsteady nonlinear Navier-Stokes

487
equations based on engine specifications and ambient conditions on a mine level. The result from
GT-SUITE is the heat emitted from a given piece of equipment in a specific cycle stage, i.e., trav­
eling up or down a ramp, loading or idling, and/or traveling on the surface. The total heat from
the equipment is determined by the proportion of time the equipment operates in a specific cycle
stage. Heat from autocompression is determined using Equation 1 (McPherson, 1993b).

where:
Tfinal = final air temperature on a mine level, °C
Tinitial = initial air temperature from surface, °C
g = acceleration due to gravity, m/s2
z1 = surface elevation of the mine, m
z2 = mine level elevation, m
cp = specific heat capacity of air, J/kg °C
Strata heat from newly excavated airways is quantified using Equation 2 (Mutmansky et al.
2012).

where:
l = length of airway, m
p = perimeter, m
k = thermal rock property, W/m °C
r = hydraulic radius, m
Tv = virgin rock temperature on a mine level, °C
Tair = air temperature on a mine level, °C
ω = Goch-Patterson coefficient on a mine level, unitless
All the heat sources are quantified a priori and included as parameters in the production sched­
uling model. The medium-term production scheduling model used in this study is formulated as
an RCPSP optimization model. We present the generalized mathematical formulation of an
RCPSP in Appendix A. We ask our audience to refer to the detailed formulation in Ogunmodede
et al. (2022). The RCPSP model is solved using the Bienstock-Zuckerberg algorithm (BZ) (Bien­
stock & Zuckerberg, 2010) and the TopoSort heuristic (Chicoisne et al., 2012). The BZ algorithm
provides continuous solutions to the RCPSP model by relaxing it as a linear programming
model. These continuous solutions are then converted into integer-feasible solutions using the
TopoSort heuristic. The open mine planner (OMP) solver (Rivera, Goycoolea, Moreno, & Espi­
noza, 2021) is used to determine solutions to the RCPSP model by executing the BZ and Topo­
Sort algorithms. The OMP solver process begins by exporting the mine schedule from Deswik
(Deswik Mining Consultants Pty Ltd, 2022) as a CSV file to a python script to extract the time
periods, precedence, and resource constraints into OMP files. The OMP solver outputs the time
periods for each activity, the quality of the solution, and the constraints which were violated.
This information is imported back into Deswik® for analysis.

4 CASE STUDY

Mine X is a multi-deposit underground hardrock operation, with several major high-grade,

low sulfidation, epithermal gold or silver vein deposits. The operation produces < 300,000
ounces (oz) of gold annually (Mine X Corporation, 2021). The mining method used at Mine
X is a combination of transverse and longitudinal long-hole sublevel stoping with cemented
rock backfill. Ore is mined from the three active mine areas which are accessed by a decline.
The current mining fleet consist of a mix of underground loaders, underground haul trucks
and surface haul trucks split between the active mines. The mine ventilation system consists of

488
a decline and several raises. Total airflow is approximately 210 m3/s (445,000 cfm). We obtain
the medium-term and ventilation plan for one of the active mines as a Deswik CAD and
scheduler, and Ventsim™ file format, respectively. Activities in the medium-term plan from
the mine are scheduled at a daily fidelity over a period of 5 years. From this, we obtain infor­
mation on the types of activities, their duration, resource assignment, and precedence struc­
ture. Table 1 shows the resource and equipment capacities used in the model. The operation
would like to optimize their ventilation resource to create a safe working environment for
miners, most especially heat, while optimizing production.

Table 1. Resource and equipment capacity constraints.

Resource (units) Daily maximum capacity

Ore tonnes (t) 1000

Backfill tonnes (t) 1500
Lateral development meters (m) 25
Ore drive development meters (m) 10
Vertical development meters (m) 7
Production drill units (quantity) 2
Drift development units (quantity) 2
Electrical and ventilation station unit (quantity) 1
Ramp development unit (quantity) 1

5 RESULTS AND DISCUSSION

Using the data from the case study mine, we solve and analyze two production scheduling
models. The first model, defined here as base model, is constrained by the precedence structure
and the resource capacities outlined in Table 1. The heat generated from the activities is uncon­
strained in this model. The second model, defined as heat-constrained model, includes the con­
straints defined in the base model with additional constraints to limit the heat generated from
activities per level. The heat constraints ensure that the wet bulb temperature on a mine level is
below the allowable limit. We use these models to analyze the following scenarios in this study:
1. Normal operating conditions
2. Increase in equipment heat load
3. Decrease in ventilation
Under normal operating conditions, the initial temperature, relative humidity, and airflow
per level are 18.9 °C (66°F), 76.5%, and 52.5 m3/s respectively. We assume that four levels are
active daily, and the airflow is evenly distributed to the levels. We vary the equipment heat
load and ventilation in the other two scenarios to observe how changes in these parameters
affect the production schedule. The following sections present the results for each scenario.

5.1 Base model versus heat-constrained model under normal operating conditions
The result for this scenario using the base model and heat-constrained model is shown in
Figure 1. The graph shows the annual ore tonnes produced and development meters completed.
The base model ramps up production in the early years of the mine life. Production peaks in
2023 and gradually declines towards the end of the mine life. There is a decline in development
meters in the same year but gradually peaks in 2025. The objective function of the base model
maximizes NPV hence, it schedules development activities just in time for ore production. The
schedule from the heat-constrained model provides ore tonnes at a slower rate and gradually
peaks later in the mine life in year 2025. For development meters, the heat-constrained model
completes development activities faster than the base model. This is because of the lower heat
load from development activities compared to stope extraction and backfill activities. The equip­
ment used for development and production drilling are mostly electric powered. Given the lower
heat load from development and production drilling activities, the heat-constrained model

489
schedules much of these activities at the lower levels where temperatures are high due to auto­
compression and strata heat, while adhering to precedence and resource constraints. The heat-
constrained model optimizes NPV by scheduling stope extraction activities on the upper levels
until stopes are available to mine on the lower levels. The heat constraints included in the model
ensures that the total heat load from activities do not exceed the allowable limit for each level in
the mine. Figure 2 shows the wet bulb temperature on the active mine levels for a 6-month
period in 2025 where we first observe a temperature violation. The solution obtained for the
base model violates the maximum allowable temperature, i.e., 28 °C (82.4°F) wet bulb, while the
heat-constrained model adheres to the heat limits.

Figure 1. Annual ore production and development meters using the base model and heat-constrained
model.

Figure 2. Wet bulb temperature on an active mine level under normal operating conditions using the base
model.

5.2 Base model versus heat-constrained model with an increase in equipment heat load
The equipment heat load in this scenario is raised at 10% increments under normal operating
conditions, i.e., 18.9 °C (66°F) initial temperature, 76.5% relative humidity, and 52.5 m3/s

490
airflow per level. This modification to the parameters reflects an underestimation of the equip­
ment heat load used in the previous scenario (normal operating conditions). The production
schedules generated in this scenario are similar to the previous scenario hence we focus our
attention on the heat generated from these activities. We observe an increase in the number of
active mine levels that violate the allowable temperature limit as the equipment heat load
increases using the base model. Figure 3 shows the maximum and average wet bulb temperature
for the 50% increase in equipment heat load using the base model for a 6-month period in 2023.
We observe that Levels 400 and 425 violate the allowable temperature limit multiple times.
Data from the production schedule show that several stope extraction, backfill, and develop­
ment activities are happening on these levels. Information from the temperature helps the mine
planner to identify mining areas that are likely to violate the allowable temperature limit and
adopt strategies to mitigate high temperatures on the level. Figure 4 shows that the wet bulb
temperature for the same period using the heat-constrained model is below the allowable limit.

Figure 3. Wet bulb temperature on active mine level for a 50% increase in equipment heat load using
the base model.

Figure 4. Wet bulb temperature on active mine level for a 50% increase in equipment heat load using
the heat-constrained model.

491
5.3 Base model versus heat-constrained model with a decrease in ventilation scenario
In this scenario, we maintain all the parameters used under normal operating conditions, i.e.,
the equipment heat load, initial temperature, and relative humidity, but vary the airflow to the
mine levels. This scenario simulates situations where damage to the ventilation fan may dis­
rupt production. Similar to the scenario discussed in Section 5.2, we observe an increase in the
number of mine levels that violate the allowable temperature as the airflow is reduced at 10%
increments. Figure 5 shows the wet bulb temperature for a 50% decrease in ventilation. We
observe that the wet bulb temperature per level increases as high as 45°C (113°F) and multiple
levels violate the allowable temperature limit in the same period. This implies that the ensuing
production schedule is unachievable. This information helps the mine planner to make
informed decisions such as reducing the number of active levels, and focusing on high priority
areas, i.e., areas that contribute significantly to NPV.

Figure 5. Wet bulb temperature on active mine level for 50% decrease ventilation using the base model.

6 CONCLUSION AND FUTURE WORK

In this study, we model heat associated with auto compression, geothermal gradient, and
mining equipment, and incorporate it into a medium-term production schedule developed by
Ogunmodede et al. (2022). We define two models: the base model, where the heat generated
from activities are unconstrained, and the heat-constrained model. Using data from a case
study mine, we demonstrate the strength of this model by simulating scenarios including an
increase in equipment heat load and a decrease in ventilation that could disrupt production.
The result show that the heat-constrained model generates operationally feasible schedules that
manage heat within a mine. The heat-constrained model schedules activities such that the heat
generated is below the allowable wet bulb temperature limit.
As underground mine extraction depths increase, the natural sources of heat, i.e., autocom­
pression and strata can increase to uncontrollable levels. The controllable factor is to manage
the heat from mining activities. The heat-constrained model, evaluated in this study, can serve
as a tool for mine planners to evaluate both the production schedule and the environmental
conditions associated with the schedule to ensure safety while efficiently managing produc­
tion. Also, the model can help mine planners perform further analysis on the impact various
disruptions could have on the schedule. Furthermore, this model can be used for financial
analysis in the feasibility stage of a mining project to make better investment decisions. Future
research would incorporate other contaminants such as diesel particulate matter and respir­
able dust to evaluate their impact on the production schedule and ventilation system.

492
ACKNOWLEDGMENTS

We thank the National Institute for Occupational Safety and Health (NIOSH) for funding
this research under grant 0000HCCR-2019-36404. We also thank Aaron Swift and John Aya­
buri for their assistance in modeling the heat values for diesel equipment.

REFERENCES

Acuña, E. I., & Lowndes, I. S. (2014). A review of primary mine ventilation system optimization. Inter­
faces, 44(2), 163–175. https://doi.org/10.1287/inte.2014.0736
Bienstock, D., & Zuckerberg, M. (2010). Solving LP relaxations of large-scale precedence constrained
problems. In F. Eisenbrand & F. B. Shepherd (Eds.), Integer Programming and Combinatorial Opti­
mization 2010 (pp. 1–14). https://doi.org/10.1007/978-3-642-13036-6_1
Brickey, A. J. (2015). Underground production scheduling optimization with ventilation constraints.
(Doctoral Thesis). Colorado School of Mines. Arthur Lakes Library.
Brickey, A. J., Chowdu, A., Newman, A., Goycoolea, M., & Godard, R. (2020). Barrick’s Turquoise
Ridge gold mine optimizes underground production scheduling operations. INFORMS Journal of
Applied Analytics. https://doi.org/10.1287/inte.2020.1027
Bullock, R. L. (2011). Introduction to underground mine planning. In P. Darling (Ed.), SME Mining
Engineering Handbook (3rd ed., Vol. 1, pp. 1135–1141). Littleton: Society for Mining, Metallurgy &
Exploration.
Carlyle, W. M., & Eaves, B. C. (2001). Underground planning at Stillwater Mining Company. Interfaces,
31(4), 50–60. https://doi.org/10.1287/inte.31.4.50.9669
Chicoisne, R., Espinoza, D., Goycoolea, M., Moreno, E., & Rubio, E. (2012). A new algorithm for the
open-pit mine production scheduling problem. Operations Research, 60(3), 517–528. https://doi.org/
10.1287/opre.1120.1050
Chowdu, A. A., Nesbitt, P., Brickey, A. J., & Newman, A. (2020). Operations research in underground
mine planning : A review. INFORMS Journal on Applied Analytics. Forthcoming.
Deswik Mining Consultants Pty Ltd. (2022). Deswik - Delivering mining efficiency. Retrieved from
https://www.deswik.com/
Fava, L., Maybee, B., & Millar, D. (2012). Decision support for an underground gold mining operation -
A case study using the Schedule Optimisation Tool. Proceedings Project Evaluation 2012, (May),
41–48. Melbourne: Australasian Institute of Mining and Metallurgy.
Hall, B. (2017). Long-term mine planning and the mine-planning process. Retrieved October 18, 2022,
from https://www.amcconsultants.com/experience/long-term-mine-planning-and-the-mine-planning-
process
Huang, S., Li, G., Ben-Awuah, E., Afum, B. O., & Hu, N. (2019). A robust mixed integer linear pro­
gramming framework for underground cut-and-fill mining production scheduling. International Jour­
nal of Mining, Reclamation and Environment, 1–18. https://doi.org/10.1080/17480930.2019.1576576
Kuchta, M., Newman, A., & Topal, E. (2003). Production scheduling at LKAB’s Kiruna Mine using
mixed-integer programming. Mining Engineering, 55(4), 35–40.
Kuchta, Mark, Newman, A., & Topal, E. (2004). Implementing a production schedule at LKAB’s
Kiruna Mine. Interfaces, 34(2), 124–134. https://doi.org/10.1287/inte.1030.0059
Little, J., Nehring, M., & Topal, E. (2008). A new mixed-integer programming model for mine produc­
tion scheduling optimisation in sublevel stope mining. In P. Knights (Ed.), Australian Mining Technol­
ogy Conference 2008. Twin Waters.
Lopes, T. V. F. (2017). Underground mine production scheduling using OMP solver: analysis of modifica­
tions to an integer programming model and implementation of a sliding time window heuristic. South
Dakota School of Mines & Technology.
Martinez, M. A., & Newman, A. M. (2011). A solution approach for optimizing long- and short-term
production scheduling at LKAB’s Kiruna mine. European Journal of Operational Research, 211(1),
184–197. https://doi.org/10.1016/j.ejor.2010.12.008
Maybee, B. (2011). A risk-based evaluation methodology for underground mine planning. ProQuest Disser­
tations Publishing.
Mcisaac, G. (2005). Long-term planning of an underground mine using mixed-integer linear
programming. CIM Bulletin, 98(1089), 89.
McPherson, M. J. (1993a). Background to subsurface ventilation and environmental engineering. In Sub­
surface Ventilation and Environmental Engineering (pp. 1–11). https://doi.org/10.1007/978-94-011-1550-
6_1

493
McPherson, M. J. (1993b). Subsurface ventilation and environmental engineering. In Subsurface Ventila­
tion and Environmental Engineering. https://doi.org/10.1007/978-94-011-1550-6
Mine X Corporation. (2021). 2021 Annual report (Vol. 2021).
Mutmansky, J. M., Hartman, H. L., Ramani, R. V., & Wang, Y. J. (2012). Mine ventilation and air condi­
tioning. Germany: Wiley.
Nehring, M., & Topal, E. (2007). Production schedule optimisation in underground hard rock mining
using mixed integer programming. In R. Flanagan (Ed.), Project Evaluation Conference 2007 (pp.
169–175). Melbourne: The Australasian Institute of Mining and Metallurgy.
Newman, A. M., & Kuchta, M. (2007). Using aggregation to optimize long-term production planning at
an underground mine. European Journal of Operational Research, 176(2), 1205–1218. https://doi.org/
10.1016/j.ejor.2005.09.008
Newman, A. M., Rubio, E., Caro, R., Weintraub, A., & Eurek, K. (2010). A review of operations
research in mine planning. Interfaces, 40(3), 222–245. https://doi.org/10.1287/inte.1090.0492
O’Sullivan, D., & Newman, A. (2014). Extraction and backfill scheduling in a complex underground
mine. Interfaces, 44(2), 204–221. https://doi.org/10.1287/inte.2013.0730
Ogunmodede, O., Lamas, P., Brickey, A., Bogin, G., & Newman, A. (2022). Underground production
scheduling with ventilation and refrigeration considerations. Optimization and Engineering, 7(3),
219–242. https://doi.org/10.1007/s11081-021-09682-4
Radin, R. L. (2017). Optimization in operations research (2nd ed.). Hoboken, NJ: Pearson Higher Educa­
tion Inc.
Ramsden, R., Branch, A.R. and Wilson, R. (2007). Factors influencing the choice of cooling and refriger­
ation systems for mines. Journal of the Mine Ventilation Society of South Africa, 60(3), 92–98.
Rivera, O., Goycoolea, M., Moreno, E., Alexandra, N., & Brickey, A. (2016). The OMP user’s guide.
Unpublished material.
Rivera, O., Goycoolea, M., Moreno, E., & Espinoza, D. (2021). The Open Mine Planner Guide.
Ryan, A., & Euler, D. S. (2017). Heat stress management in underground mines. International Journal of
Mining Science and Technology, 27(4), 651–655. https://doi.org/10.1016/j.ijmst.2017.05.020
Smith, M., Sheppard, I., Karunatillake, G., & Camisani-Calzolari, F. (2003). Using MIP for strategic
life-of-mine planning of the lead/zinc stream at Mount Isa Mines. 31st International APCOM Sympo­
sium Proceedings, Capetown, South Africa, (Figure 3), 465–474. Retrieved from http://www.saimm.co.
za/Conferences/Apcom2003/465-Smith.pdf
Technologies, G. (2021). GT-SUITE. Retrieved from https://www.gtisoft.com/gt-suite/
Terblanche, S., & Bley, A. (2015). An improved formulation of the underground mine scheduling opti­
misation problem when considering selective mining. ORiON, 31(1), 1. https://doi.org/10.5784/31-
1-422
Trout, L. P. (1995). Underground mine production scheduling using mixed integer programming. The
25th International APCOM Symposium, (July), 395–400. Retrieved from https://ausimm.com/product/
underground-mine-production-scheduling-using-mixed-integer-programming/
U.S. Department of Labor. (2012). Heat stress in mining. In National Mine Health and Safety Academy
Safety Manual Series (SM 6). Retrieved from http://www.ergonomics.org.au/downloads/EA_Jour
nals/Heat_Stress_in_Mining_-_leveritt.pdf%5Cnpapers2://publication/uuid/F2AECEBC-BA0E-4924-
94E6-3A9E247C01A8
van der Walt, J., de Kock, E. M., & Smith, L. K. (1983). Analysis of ventilation and cooling requirements
for mines. Journal of The South African Institute of Mining and Metallurgy, 83(2), 25–34. https://doi.
org/10.10520/AJA0038223X_1371
Williams, J. K., Smith, L., & Wells, P. M. (1973). Planning of underground copper mining. 10th Interna­
tational Applied Computer Operations Research in Mineral Industries, 251–254.
Yeoman, K., Weakley, A., DuBose, W., Honn, K., McMurry, T., Eiter, B., Poplin, G. (2022). Effects of
heat strain on cognitive function among a sample of miners. Applied Ergonomics, 102, 103743. https://
doi.org/10.1016/j.apergo.2022.103743

494
APPENDIX A – GENERALIZED RESOURCE CONSTRAINED PROJECT
SCHEDULING PROBLEM (RCPSP)

Sets:
A set of activities
R set of resources
P set of activity precedence relationship
T set of time periods
Parameters:
d^a duration of activity a 2 A
cat objective value to be derived if activity a is initiated in time period t
δt discount factor for objective function in time period t
d^a;a delay between activity a0 and activity a
^rrt quantity of resource r available in time period t
qar quantity or resource r consumed by executing activity a
Decision variables:

Objective function:

Constraints:

The objective function (3) maximizes the discounted total value of parameter c from all
activities over the entire time horizon. This could represent the NPV from production oper­
ations or quantity of mineral commodity produced. The objective function could as well be
altered to minimize parameters such as production costs or deviations from production target.
Constraint (4) ensure that no activity is executed more than once. For instance, a given stope
cannot be mined more than once. Constraint (5) impose precedence relationships on activities,
while constraint (6) enforces resource consumption limits. Constraint (7) ensure that the deci­
sion variables are binary.

495
Mine ventilation and automation
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Permeability determination for potential interaction between shale

gas wells and the coal mine environment due to longwall-induced
deformations under deep cover

M.L. Harris, S.J. Schatzel, K.M. Ajayi, M. Van Dyke, P. Zhang, V. Gangrade &
J.D. Addis
National Institute for Occupational Safety and Health, Pittsburgh, PA, USA

H. Dougherty & E. Watkins

formerly National Institute for Occupational Safety and Health, Pittsburgh, PA, USA

ABSTRACT: This paper summarizes the changes in permeability for two boreholes located
above an abutment pillar in a longwall coal mine to characterize potential interaction between
shale gas wells and the coal mine operations under deep cover. To determine the safety of the
mine environment in case of a potential well breach, fracture network characteristics are
needed to conduct a comprehensive hazard assessment. Permeability was measured using
a falling-head slug test and calculated in accordance with the Hvorslev model during the
mine-by of a longwall panel on one side of the pillar. The two boreholes had screened lengths
at different depths to evaluate stratigraphic zones of interest above the active mining seam.
The permeability at each borehole increased from pre-mining to post-mining and was highest
while the face was close to and during mine-by of the test site.

1 INTRODUCTION

Within the past 15 years, unconventional shale gas wells have been drilled through current and
future coal reserves in Pennsylvania, West Virginia, and Ohio. Impacts on the mechanical
integrity of these wells becomes a concern when mining occurs in and around these wells. The
shale gas wells penetrate the coal seams and the coal seams are subsequently mined. The coal
and shale gas industries need to know what the impacts of longwall mining are on the shale gas
wells, what the potential deformation may be, and what stresses are imposed on the gas wells.
In 2012, the Pennsylvania Department of Environmental Protection (PADEP) recognized
the 1957 Pennsylvania Gas Well Pillar Regulation (Commonwealth of Pennsylvania, 1957)
was created without current data from modern day longwall mining and called for research to
revise the outdated regulation. Also, the 1957 Pennsylvania Gas Well Pillar Regulation has
been widely used by the Mine Safety and Health Administration (MSHA) along with CFR
75.1700 to manage gas well pillar stability issues. Therefore, given the posed questions and the
need for further guidance given modern mining technologies and practices, the National Insti­
tute for Occupational Safety and Health (NIOSH) initiated research to address these issues.
One of the first steps in quantifying the interaction between the two processes is to determine the
change in ground permeability at depths where longwall-induced deformations are of highest con­
cern. If longwall-induced subsurface strata deformations and stresses should affect shale gas well
casings and if a breach occurs, the changes in the ground permeability could potentially increase
the rate of shale gas infiltration into an underground coal mine. Permeability field measurements
have been conducted under various differing conditions. Boreholes placed in the barrier pillar and
in the longwall gob areas under differing depths of cover have been monitored during completion
of the longwall coal seam. Previously, three monitoring boreholes were monitored to experimen­
tally measure borehole permeability and changing water head under shallower cover (deepest at

DOI: 10.1201/9781003429241-51

499
127 m or 417 ft). Watkins et al. (2021) assessed permeability changes around boreholes above an
abutment pillar under shallower cover. This paper will discuss a similar scenario except under
deeper cover conditions (up to 238 m or 1,047 ft) and during the first longwall panel mine-by.

2 TEST METHOD

The study included drilling two boreholes to various depths to monitor the Sewickley and Union­
town horizons and performing field experiments to gather permeability data. Although Marcellus
shale gas reservoirs are at much greater depths than the Pittsburgh coal bed longwall mines, the
risk of a hypothetical casing failure is thought to be greatest where the maximum zone horizontal
movement occurs in response to mining. This zone is typically shallower than the mined coalbed
and within a few hundred meters of the unit. Additionally, the areas above abutment pillars are
particularly important as shale gas wells are required to be drilled through them.

2.1 Borehole and mine layout

The most commonly sought-after shale gas reserve in the southwestern Pennsylvania region is
the Marcellus Shale Reserve which can be located at a depth of about 2,100 m (7,000 ft) to
2,750 m (9,000 ft). Commonly, several wells can be positioned on a single drill pad to decrease
drilling costs and reduce the surface footprint of the well.
The Pittsburgh seam, a relatively flat deposit, is longwall mined in this region due to its con­
sistent quality and thickness. The stratigraphic zones monitored for possible changes in per­
meability due to mining-induced deformation were the Sewickley coal bed and the Uniontown
coal bed, whose thicknesses vary in the region (Figure 1). These coal beds typically have lower
compressive strengths than the adjacent limestone, shale, and siltstone/sandstone units
(Rusnak & Mark, 2000; Mohan et al., 2001). A field study showed that horizontal fractures
formed between layers with differing uniaxial compressive strengths and seam thicknesses
(Palchik, 2005). This stratigraphy is found at the study site with the thick limestone units and
thin coal beds (Figure 1). Consequently, zones of high ground movement would be associated
with the coal horizons due to mining-induced deformation.

Figure 1. Generalized stratographic profile for the VEP boreholes (not to scale).

The primary factors creating the gas transport network associated with longwall gobs are ground
movements related to subsidence. Previous ground control studies (Su et al., 2019) indicate the pri­
mary movement component that potentially deforms well casings in longwall pillars is subsidence or
horizontal movement. FLAC3D™ modeling estimations show an insignificant amount of vertical
movement for borehole locations above abutment pillars. However, higher magnitudes of displace­
ment perpendicular to the panel are typical and are shown for this area in Figure 2. Although strati­
graphic zones of maximum horizontal movement in the region can be variable, previous modeling
shows that the Uniontown and Sewickley coal seams are among the likely locations (Su et al., 2019).

500
Most of the horizontal movement was estimated to be perpendicular to the longwall panel due to
the ground subsidence pulling the strata above the pillar toward the caved gob zone, which is per­
pendicular to the panel direction. Figure 2 presents data for a nearby location and is representative
for this monitoring site. Here, a horizontal displacement of about 0.7 cm (0.3 in) with a predicted
casing diameter reduction of 0.15 cm (0.16 in) is predicted. A depth of approximately 155 m (509 ft)
is when horizontal displacement increases to about 2.0 cm (0.8 in) and the predicted casing diameter
reduction increases to 0.70 cm (0.3 in). These zones of maximum movement are also important loca­
tions for fluid transport as these are typically high permeability zones for the transport of gas from
a hypothetical breach.

Figure 2. Shale gas well horizontal displacement by depth perpendicular to the panel from field meas­
urements and FLAC3D™ modeling.

In southwestern Pennsylvania, monitoring boreholes used to examine longwall-induced per­

meability changes under deeper cover were drilled in the overburden directly above an abut­
ment pillar between two planned longwall panels. The field test site is located on the top of
a steep hill and has an overburden depth to the Pittsburgh seam of about 370 m (1,214 ft).
The monitoring boreholes are labeled as VEP-S and VEP-U to monitor the Sewickley and Union­
town formations respectively. VEP-S was drilled to a depth of 328 m (1,076 ft) ending at the Sewick­
ley formation and VEP-U was drilled to a depth of 291 m (955 ft) ending at the Uniontown
formation equivalent. According to Figure 2, at the depths of 291 m (955 ft) and 328 m (1,076 ft) the
predicted horizontal displacement is expected to be 0.6 cm (0.2 in) and 0.3 cm (0.1 in), respectively.
The borehole pattern was centered above a 38 m (125 ft) by 84 m (275 ft) abutment pillar that
was part of a three-entry gate road system (Figure 3) with 457 m (1,500 ft) wide longwall panels that
had lengths of about 3,660 m (12,000 ft) on both sides. The distance from the center of a borehole
collar to the edge of the pillar adjacent to an active panel was at most 18 m (60 ft). Using the guid­
ance from the 1957 PADEP study, this distance may be as small as 15 m (50 ft). Measurements
place the Pittsburgh seam to be at a depth of 370 m (1,214 ft) from the surface at this location.
The VEP-S and VEP-U boreholes were drilled and cased with a steel casing and grout baskets
except for the test intervals, which had screen lengths of 3 m (10 ft) and 5 m (15 ft), respectively.
Grout or bentonite were used to cement the borehole annulus outside of the casing.

2.2 Falling-head slug tests

Throughout the mining of the first longwall panel, researchers conducted falling-head slug
tests to measure the permeability of the slotted casing lengths. VEP borehole slug testing
began when the longwall face was 273 m (897 ft) away for the first panel and will continue
until the completion of mining for both panels. To date, only the first longwall panel has been

501
Figure 3. Top view of a three-entry longwall layout. Test boreholes and mining direction are shown.

completed. These test zones were chosen to correspond to probable locations of high ground
movement where increases in permeability were considered most likely. An INW PT2X (Sea­
metrics™) piezometer tracked the long-term equilibrium water height and water slug height
(Figure 4). For the falling-head slug tests, a water slug height of up to 3 m (10 ft) was added
to the boreholes. The previous study (Watkins et al., 2021) recorded the fall of water head at
5, 30, or 60 s intervals, depending on the expected drainage rate of the water. However, during
the current study the rate of loss was at a much slower rate and, therefore, the data was
recorded every minute, and the falling rate was later determined from the downloaded data.

Figure 4. The sideview of the VEP-S and VEP-U monitoring boreholes (not to scale).

When the longwall face was 364 m (1193 ft) away from the monitoring site, the slug tests
were conducted on each of the boreholes twice a week. The frequency increased to three times
per week when the longwall face was within 100 m (328 ft) of the boreholes. When the active
longwall face passed the borehole locations and was further away, the sampling rate for slug
tests was approximately once per week. Once the longwall panel was completely mined, long-
term continuous water pressure readings were recorded every eight hours. The piezometer
readings, water slug height (Hw), and initial slug height (Ho) were converted to Hw/Ho values.
For the calculation of permeability from falling-head slug tests, Hw/Ho values were plotted on
a semi-log graph to determine the T37 time, which is the time, in minutes, in which the water slug
drained to 37% of the initial water slug height (Dawson & Istok, 1991; Watkins et al., 2021). For
a 10-ft water slug, the T37 time would be calculated as the time it took for the slug to drain to 3.7 ft.
If the T37 value did not fall within the given data range, two points were chosen along the most
linear section of the semi-log plot of the slug test, and the slope of those two points estimated the
T37 time. Watkins et al. (2021) details the application of T37 in the Hvorslev method to determine
the hydraulic conductivity. The hydraulic conductivity is dependent upon the well casing radius, the
well screen radius, and the length of the well screen. For both VEP boreholes, the well casing radius

502
is 0.052 m (2 in), and the well screen radius is 0.104 m (4 in). The well screen length is 4.6 m (15 ft)
for VEP-U and is 3.0 m (10 ft) for VEP-S.
As with the previous study, multiple assumptions about the slug test and ground aquifer are
applied when conducting the Hvorslev method to calculate permeability. This method assumed that
the water slug is added instantaneously, the groundwater flow is described by Darcy’s Law, and the
volume of water that flows into the aquifer is equivalent to the change in water volume within the
well casing (Watkins et al., 2021). The aquifer is assumed to be incompressible, lithologically homo­
geneous, isotropic, and vertically confined by aquicludes. The injection well is assumed to have
a negligible radius in relation to the size of aquifer, the well has a screen with a negligible head loss,
and the water slug flow travels horizontally away from the well in all directions.

3 RESULTS AND DISCUSSION

3.1 VEP-U
Figure 5 displays the changes in water head from experimental data in VEP-U. The data was col­
lected when the longwall panel was at varying distances from the monitoring boreholes. In this
figure, the negative distance (– 54 m or -177 ft) in the legend refers to an approaching longwall
face and the positive distance (+61 m or +200 ft) refers to the longwall face moving away. Zero
distance is where the longwall face passed the center of the test abutment pillar and is the basis for
position measurements. As seen in Figure 5, the permeability was 3.19 x 10-15 m2 (3.2 mD) when
the longwall was 54 m (177 ft) from mining by the monitoring holes. The permeability de-creased
slightly to 2.51 x 10-15 m2 (2.51 mD) during the mine-by before rising to 5.33 x 10-15 m2 (5.33 mD)
when approximately 61 m (200 ft) beyond the monitoring boreholes.

Figure 5. VEP-U water height curves for slug tests at various longwall positions. The negative value
refers to the approaching longwall face and the positive value refers to the face moving away from the
test location.

Figure 6 shows the permeability, m2, and pressure measurements, kPa, at the Uniontown hori­
zon from VEP-U. The numbers above the bars are the distances (m) of the longwall panel from
the monitoring boreholes. The black negative numbers above the bars indicate the longwall panel
approaching the monitoring location and the red positive numbers above the bars indicate the dis­
tance as the longwall panel passes the monitoring location. Prior to mining, the permeability in
VEP-U was 0.19 x 10-15 m2 (0.19 mD). The borehole water level steadily rose prior to 08/23/2021.
This is seen in the rise in associated pressure measurements (orange plots) on Figure 6. The longwall
panel was mined by the abutment pillar on 08/25/2021 when the permeability was 2.51 x 10-15 m2

503
(2.54 mD). The water levels and pressures were also the highest at this time with pressures just over
910 kPa (132 psi). When the longwall panel passes the monitoring point, the permeability increases
and reaches a maximum of 5.33 x 10-15 m2 (5.4 mD) and the pressure in the hole drops to 585 kPa
(85 psi). After this maximum, the permeability drops to a low of 8.78 x 10-16 m2 (0.89 mD) before
slightly rising. The downhole pressure is relatively steady during the first half of September and
hovers around 500 kPa (73 psi) before dropping at the end of September.

Figure 6. Permeability values and pressure measurements from VEP-U at the Uniontown horizon.
Black numbers on the graph indicate the distance (m) of the longwall face advancing toward the monitor­
ing point, and red numbers indicate the distance (m) of the longwall.

3.2 VEP-S
Figure 7 displays the water changes before, during, and after the longwall panel mine-by of VEP-S.
When the longwall face was approaching the monitoring site (54 m or 177 ft away), the permeabil­
ity was 1.41 x 10-15 m2 (1.43 mD). The permeability then increased to 2.70 x 10-15 m2 (2.74 mD)
during the longwall panel mine-by before increasing again to 1.27 x 10-14 m2 (12.8 mD).

Figure 7. VEP-S water height curves for slug tests at various longwall positions. The negative value refers
to the approaching longwall face and the positive value refers to the face moving away from the test location.

504
 Figure 8 shows the pressure measurements and the corresponding permeability values. The
highest permeability, 1.755 x 10-14 m2 (17.8 mD), occurred two days after the mine-by when
the panel was 19 m (62 ft) past the monitoring point (Figure 8). The associated water pressure
at this point was 885 kPa (128 psi). After the longwall panel mine-by, the pressure decreased
and held steady at around 500 kPa (73 psi), while the permeability values varied in the range
of 0.97 x 10-15 m2 – 1.93 x 10-15 m2 (0.98 mD – 1.96 mD). Towards the end of September, the
pressure started decreasing again as the permeabilities slightly increased.

Figure 8. Permeability values and pressure measurements from VEP-S at the Sewickley horizon. Black
numbers on the graph indicate the distance (m) of the longwall face advancing toward the monitoring
point and red numbers indicate the distance (m) of the longwall.

3.3 Comparison
Table 1 lists the minimum and maximum measured permeability values after the first longwall
panel mine-by as well as the predicted horizontal displacements. As with the previous study,
the deeper Sewickley horizon saw higher permeability values since it is closer to the mining
horizon and would be most impacted by the compression of the fractured rock. The maximum
permeabilities under deeper cover are smaller than for shallower cover by an order of magni­
tude of two for both horizons. As seen in Table 1, the maximum permeability for the Sewick­
ley horizon under deeper cover was 1.75 x 10-14 m2 (17.8 mD) and 1.34 x 10-12 m2 (1358 mD)
under shallower cover. For the Uniontown horizon, the maximum permeability under deeper
cover was 5.33 x 10-15 m2 (5.4 mD) and 2.42 x 10-13 m2 (245 mD) under shallower cover.

Table 1. Minimum and maximum permeability values for VEP-S and VEP-U resulting from this study
and for corresponding FEB-1 and FEB-2 (Watkins et al., 2021).
Longwall Panel 1
Measured Permeabilities Predictions
Horizontal
Borehole Horizon Overburden Minimum Maximum Displacement
ID of Interest Depth m (ft) m2 (mD) m2 (mD) cm (in)

VEP-S Sewickley 328 (1,076) 4.64E-16 (0.47) 1.75E-14 (17.8) 0.6 (0.2)
VEP-U Uniontown 291 (955) 1.84E-16 (0.19) 5.33E-15 (5.4) 0.3 (0.1)
FEB-1 Sewickley 127 (417) 4.06E-13 (411) 1.34E-12 (1358) 17 (6.7)
FEB-2 Uniontown 76 (249) 2.16E-14 (22) 2.42E-13 (245) 4 (1.6)

505
 The previously stated trend also correlates with the predicted horizontal displacements. The
predicted displacements for both shallower and deeper covers are larger for the deeper Sewick­
ley horizons than for the Uniontown. The horizontal displacement prediction was 0.6 cm (0.2
in) for the deeper cover and 17 cm (6.7 in) for the shallower cover and, for the Uniontown
seam, it was 0.3 cm (0.1 in) and 4 cm (1.6 in) for the deeper and shallower covers, respectively.
The predicted horizontal displacements are also 1-2 orders of magnitude in difference with the
shallower cover having higher values than that of the deeper cover.

4 CONCLUSIONS

At a mine site in southwestern Pennsylvania, permeability measurements were conducted over

a 2-month period on two boreholes monitoring the Uniontown and Sewickley coal horizons
under deeper cover for these seams. The permeability values were higher for the Sewickley hori­
zon (maximum 1.75 x 10-14 m2) than for the Uniontown horizon (maximum 5.33 x 10-15 m2).
This also corresponds with the predicted horizontal displacement values for each.
When comparing the previous work by Watkins et al. (2021) that was conducted under
shallower cover, the Sewickley unit also exhibited higher permeability compared to the Union­
town horizon. However, there is a significant difference in that the maximum deeper cover
values are 2 orders of magnitude smaller than the maximum measured for previous shallower
cover.
Future work will include monitoring permeabilities at this deeper cover site during an idle
mining period between panels and during a second panel mine-by. The future monitoring
efforts will be compared with the previous shallower cover data, contributing to the current
knowledge and database.

DISCLAIMER

The findings and conclusions in this report are those of the author(s) and do not necessarily
represent the official position of the National Institute for Occupational Safety and Health,
Centers for Disease Control and Prevention. Mention of any company or product does not
constitute endorsement by NIOSH.

REFERENCES

CFR. Code of Federal Regulations. Washington, DC: U.S. Government Printing Office, Office of the
Federal Register.
Commonwealth of Pennsylvania, Department of Mines and Mineral Industries, Oil and Gas Division.
1957. Joint Coal and Gas Committee, Gas Well Pillar Study. Harrisburg, Pennsylvania, 28 pp.
Dawson, K. J., & Istok, J. D. (1991). Aquifer testing: Design and analysis of pumping and slug tests (pp.
297–305). Lewis Publishers, Inc.
Mohan, G. M., Sheorey, P. R., & Kushwaha, A. 2001. Numerical estimation of pillar strength in coal
mines. International Journal of Rock Mechanics & Mining Sciences, 38, 1185–1192.
Palchik, V. 2005. Localization of mining-induced horizontal fractures along rock layer interfaces in over­
burden: Field measurements and prediction. Environmental Geology, 48, 68–80.
Rusnak, J., & Mark, C., 2000. Using the point load test to determine the uniaxial compressive strength
of coal measure rock. In Proceedings of the 19th international conference on ground control in mining
(pp. 362–371). West Virginia University.
Su, D. W. H., Zhang, P., Schatzel, S. J., Gangrade, V., Watkins, E., et al. 2019. Longwall-induced sub­
surface deformations and permeability changes shale gas well casing integrity implication. In Proceed­
ings of the 38th international conference on ground control in mining (pp. 49–59). Morgantown, WV:
West Virginia University.
Watkins, E., Karacan, C. Ö., Gangrade, V., & Schatzel, S. 2021. Assessing Gas Leakage Potential into
Coal Mines from Shale Gas Well Failures: Inference from Field Determination of Strata Permeability
Responses to Longwall-Induced Deformations. Natural Resources Research, 30(3), 2347–2360.

506
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Analysis of variation in longwall-induced permeability under

different mining depths

Kayode M. Ajayi, Zoheir Khademian, Steven J. Schatzel, Marcia L. Harris &

James D. Addis
Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health,
Pittsburgh, PA, USA

ABSTRACT: Longwall-induced deformation could compromise the stability of shale gas

wells positioned in the abutment pillars of current and future coal reserves. Consequently, gas
from the casing(s) could flow towards the mine increasing the risk of explosive gas accumulation
beyond the mandated limits. To assess the impact of this hypothetical scenario, the permeabil­
ities of the surrounding strata are required to be quantified for potential gas flow to the mine.
However, varying mining depths for different coal reserves could significantly impact permeabil­
ity. Therefore, this study presents an analysis of longwall-induced permeability under shallow,
<152 m (<500 feet), and deep cover, >274 m (>900 feet), using measurements obtained from
different study sites in southwestern Pennsylvania along with discrete fracture network (DFN)
modeling in 3DEC and Fracture Flow Code (FFC). The field study measured permeability
changes for specific strata, and the numerical model predicted the permeability changes for all
the strata in the overburden. At the study site, the maximum permeability measured over the
abutment pillar at the Uniontown horizon is 2.17×10-14 m2 (22 mD) and 1.75×10-14 m2 (18 mD)
for deep and shallow cover sites, respectively. These findings provide a measure for comparing
the potential risk of a hypothetical breach under a shallow and a deep cover.

1 INTRODUCTION

The integrity of an unconventional shale gas well positioned in the abutment pillar of a longwall
mine could be compromised due to mining-induced deformation. Consequently, gas from the
compromised casing(s) is released in the overburden and could flow towards the mine. This
increases the risk of an explosion or accumulation of explosive gas beyond the mandated limits
(Mine Safety and Health Administration, 2018). This is a novel problem as the coexistence of
coal and shale gas is an emerging field in current and future coal reserves in Pennsylvania, West
Virginia, Ohio, Virginia, and Tennessee. Currently, it is required that shale gas wells near mining
should be plugged or adequately protected by a designed pillar following the design criteria devel­
oped from the 1957 Pennsylvania Gas Well Pillar Study (Commonwealth of Pennsylvania, 1957),
which predates longwall mining. This current study requires an update to consider new mining
methods such as longwall mining and the impact of deep-cover mining. Hence, researchers at the
National Institute for Occupational Safety and Health (NIOSH) have been conducting research
to provide engineering guidelines for shale gas wells influenced by longwall mining under shallow,
medium, and deep cover. Findings from these studies have predicted and measured horizontal
displacement within the abutment pillar where the unconventional gas wells are positioned (Su
et al., 2019, Su et al., 2018, Zhang et al., 2019a, Zhang et al., 2019b). For a medium and deep-
cover site in southwestern Pennsylvania, Su et al. (2019) predicted that the horizontal displace­
ment under deep cover (at 361-m depth) is one order of magnitude smaller than the medium-
cover site. Even though this observation might be site specific, the findings show that the impact
of mining-induced deformation could vary with depth.

DOI: 10.1201/9781003429241-52

507
 The Pennsylvania longwall mine data shows that mining depth could be as shallow as
30.48 m (100 ft) in stream locations and as deep as 366 m (1,200 ft) (Pennsylvania Department
of Environmental Protection, 2022). Similarly, the depth of the target Marcellus Shale could
vary from 1,500 to 2,700 m (Zhang et al., 2019b). Therefore, these unconventional shale gas
wells have been drilled through coal seams, and the impact of mining-induced deformation on
the gas well casing could vary due to the mining depth. Even though the current research by
NIOSH is aimed at providing engineering principles to ensure a safe co-existence of both oper­
ations (coal mining and shale gas extraction), it is also important to consider the impact of
a hypothetical shale gas well casing breach. The possibilities that the gas could flow towards the
mine is dependent on the permeability of the surrounding strata from the breach location to the
mine. Schatzel et al. (2012) observed that mining-induced permeability could change up to 7
months after mining with a hundred or thousands millidarcy increase. Hence, this research
group has conducted field studies to measure mining-induced permeability at study sites in
southwestern Pennsylvania under shallow and deep cover (Watkins et al., 2021). Due to the
limitations of field studies in the number of strata that could be monitored, a discrete fracture
network (DFN) model was developed using the core log data of the study sites (Ajayi et al.,
2022, Khademian Z. et al., 2022). DFN models are preferred because the fractures are modelled
explicitly similar to field conditions unlike the continuum approach. It assumes that the fracture
permeabilities are notably greater than the permeability of the rock matrix such that it can be
ignored . The model is stochastic because it is impractical to exactly represent the fracture net­
work in the overburden. It is used to predict the permeability of all the strata in the overburden,
and the results are validated with the field measurements. Previous numerical models for
a shallow-cover site (Ajayi et al., 2021) and a deep-cover site (Khademian Z. et al., 2022) have
focused on the permeability over the abutment pillar. However, this study presents the perme­
ability for the overburden from the center of the abutment pillar to the gob (Figure 1). The
difference in the induced permeability is used to analyze the impact of a hypothetical casing
breach on the mine under different covers. The research approach is presented in Section 2.0
and the findings are presented in Section 3.0 with conclusions in Section 4.0.

Figure 1. Layout of the first and second panel.

2 RESEARCH APPROACH

2.1 Field study and site geology

Figure 2 summarizes the method used in this study to predict mining-induced permeability for
a shallow and deep-cover site in southwestern Pennsylvania. Both sites are mining the Pitts­
burgh coal bed in the Northern Appalachian Basin using the longwall mining method. Both
sites have two adjacent longwall panels with a width of 457.2 m (1,500 ft) each with a three-
entry gateroad system, and a 38.1-m × 83.82-m (125-ft × 275-ft) center abutment pillar as
illustrated in Figure 2. The lithology of the overburden consists of mainly sandstone, sandy

508
shale, limestone, shale, coal, and shaley limestone. The mining horizon for the shallow-cover
site is 146.9-m (482-ft) and four monitoring boreholes were drilled over the abutment pillar.
Three of the boreholes monitored mining-induced permeability changes during the mining of
both panels, and the fourth borehole monitored the ground movement due to longwall
mining. The boreholes monitored permeability changes at specific horizons: the first borehole
(tagged FEB 1) monitored the Sewickley coal horizon at 124.97-m (410-ft) depth, the second
borehole (tagged FEB 2) monitored the Uniontown coal horizon monitored at 79.29-m (260-
ft) depth, and the third borehole (tagged FEB 3) monitored 41.15-m (135-ft) depth, which tar­
gets the projected top of the fractured zone. The choice of these monitoring locations is based
on an earlier geomechanical study for the shallow-cover site, which identified that the Sewick­
ley and Uniontown horizons are the primary zones of horizontal displacement of 17 cm and
4 cm, respectively. The deep cover site is a longwall mine with an overburden depth of
341 m (1,119 ft). Tables 1 and 2 present a summary of the field permeability measurements
conducted using a falling-head test for the shallow and deep-cover sites, respectively.

Table 1. Field permeability measurements over the abutment pillar for shallow cover (Watkins E, 2020).
Borehole ID First Panel Range m2 (mD) Second Panel Range m2 (mD)
13 12 12 12
FEB 1 4:07 � 10 1:35 � 10 (411–1,360) 1:09 � 10 5:03 � 10 (1,100–5080)
14 13 14 13
FEB 2 2:17 � 10 2:43 � 10 (21.9–245) 1:14 � 10 3:82 � 10 (11.5–386)
12 11 11 10
FEB 3 2:70 � 10 3:26 � 10 (2,730–32,900) 3:81 � 10 1:31 � 10 (38,500–132,000)

Table 2. Permeability measurements over the gob for deep cover (Khademian Z. et al., 2022).
Borehole ID Location First Panel Range m2 (mD)

Pre-mining- 295-m depth and 98 m away from the tailgate gateroad 2:42 � 10 13 2:81 � 10 13

Borehole A (245-285)
Post Mining- 265-m depth at the Uniontown coal horizon and at the 6:61 � 10 14 1:21 � 10 13

Borehole B center of the mined panel (67- 123)

VEP-S 370 m depth at the Sewickley coal horizon (328 m) and 4:64 � 10 16 1:75 � 10 14

over the abutment pillar (0.47-17.73)

VEP-U 370 m depth at the Uniontown coal horizon (291 m) and 1:84 � 10 16 5:33 � 10 15

over the abutment pillar (0.19-5.4)

2.2 Geomechanical model for aperture prediction

The permeability of rock mass surrounding the gas well is required to quantify the flow of
shale gas from a hypothetical breach location to the mining area. Khademian et al. (2022)
developed a geomechanical modeling methodology for estimating fracture aperture changes
due to longwall mining. This approach was based on the DFN technique for explicitly model­
ing fractures in rock using 3-Dimensional Distinct Element Method (3DEC) software and cal­
culating their apertures affected by longwall stress and deformations. A shallow, 145-m-cover
mine in the Pittsburgh coal seam was used to calibrate the mechanical and hydraulic proper­
ties of fractures using pre-mining permeability measurements, post-mining pillar stress, and
surface subsidence. The same calibration parameters were used in a deep-cover (341-m) long­
wall mine in southwestern Pennsylvania in the Pittsburgh coal bed, and the modeled perme­
ability results were shown to agree reasonably with the field permeability measurements. The
calibrated properties of fractures were fracture friction coefficients being 60% of the strata
internal friction, initial fracture apertures of 0.3–0.5 mm, fracture density (total fracture area
in unit volume of rock) of 0.15 in weak coal horizon, fracture density of 0.2 for the shallow
weathered zone, and fracture density of 0.3 for the interconnected fracture zone (with a height
of about 23 times the mining height).

509
Figure 2. Summary of the research approach.

2.3 Development of multizone DFN model and numerical model for permeability
As described in Section 1.0 (Introduction), it is impractical to exactly represent the fractures
in the overburden. Therefore, a stochastic DFN model is developed in Fracture Flow Code
(FFC) to predict mining-induced permeability. In a previous study conducted with FFC, the
overburden was classified into three zones: Tailgate Abutment pillar area, Gob area, and
Headgate chain pillar area. However, this approach has its limitations because a mining-
induced aperture could change within a short interval. Therefore, FFC is updated such that it
discretizes the overburden. The region over the abutment pillar is classified into 5-m zones
toward the edge of the panel, and the section over the gob is classified as 10-m zones. It is
assumed that these zones completely map the exact values of the aperture value from the geo­
mechanical model in 3DEC. Within each zone, the center location of the vertical fractures is
generated using uniform distribution, the length of the fractures is modeled with lognormal
distribution, the fracture density is imported from the 3DEC model, the aperture is modeled
with lognormal distribution using the peak value of each zone from 3DEC, and the fracture
orientation is modeled with Von Mises Fischer’s distribution. The model generates 100 DFN
realizations to replicate the potential variation in fracture geometries, and the average perme­
ability is presented in this study. Figure 3 shows one of the DFN realizations of the multizone
model generated in FFC for the shallow-cover site.
For each zone in Figure 3, the vertical and horizontal permeabilities are determined using the
method presented in a previous study (Zhang et al., 1996). Cubic law is used to model flow through
the fractures and Darcy’s equation is used to predict the permeabilities (Zhang et al., 1996):

In Equation 1, q indicates the direction of flow and velocity (m/s), and it is calculated by divid­
ing the sum of boundary flow from cubic law (Q, m3/s,) by the boundary area (A, m2),
rHx and rHy are the pressure head gradients in the x and y directions, respectively, and K is
the permeability coefficient or conductivity (m/s). The principal (qyy and qxx) and cross flow (qxy
and qyx) in the y and x directions, respectively, are obtained by applying boundary pressure inde­
pendently in both directions. The absolute permeability tensor (k in m2) is obtained from:

ρ is the density of fluid kg/m3, g is the acceleration due to gravity (m/s2), μ is the dynamic
viscosity of the fluid (Ns/m2). The principal absolute permeabilities in the x (kxx) and y (kyy)
directions are the horizontal and vertical permeabilities.

510
Figure 3. Stochastic DFN model generated for FFC.

3 RESULTS AND DISCUSSION

This section presents the permeability results from FFC for shallow and deep cover from the
center of the abutment pillar to 350 m (1,148 ft). Even though the width of the panel is
457.2 m (1,500 ft), the choice of 350 m (1,148 ft) is to focus on the area of interest closer to the
gas well in the abutment pillar and save on the computational time required for the full
model. Figures 4 and 5 show the average permeability from 100 DFN realizations for each of
the zones in the overburden for shallow and deep cover, respectively. These permeability maps
are developed using the same range for ease of comparison.

Figure 4. Permeability distribution for shallow cover.

It is observed that from the permeability map in Figure 4 that the values over the abutment
pillar (within 0–15 m) are about an order of magnitude lower than the edge of the panel. For
most of the strata, the permeability at the edge of the panel is highest and gradually decreases
towards the gob. Under this shallow-cover condition, the impact of mining-induced

511
deformation is significant within 100 m from the center of the panel and gradually decreases
towards the gob except for the strata within 40 m from the mine roof and about 10 m from
the surface. The strata directly closest to the mine roof (within 40 m) seems to have an
extended higher permeability region into the gob. The lowest permeabilities are toward the
center of the gob for strata between 10–100 m. Even though these observations are unique for
this site, the results show that mining-induced deformation could significantly impact the over­
burden under shallow cover. By comparing Figures 4 and 5, the mining-induced permeability
for most of the zones under deep cover in Figure 5 is significantly lower than in shallow cover
(Figure 4).

Figure 5. Permeability distribution for deep cover.

The highest permeability under the deep cover is near the edge of panel towards the surface
and could be due to the impact of subsidence. Like the shallow-cover site, the strata within
30–40 m from the mine roof have higher permeability compared with the strata above. Apart
from these locations, there are multiple extremely low permeability regions/zones in the over­
burden close to the range of in-situ or undisturbed rock mass. This is a clear distinction
between the mining-induced permeability for deep and shallow cover. Based on the difference
in induced permeability, it is predicted that the impact of a shale gas casing breach under shal­
low cover could create more concerns for the mine compared to deep cover. For the shallow
and deep cover, Figure 6 shows the exact plot of the permeabilities at about 100 m from the
center of the pillar for the stratum directly at the mine roof. The results show the permeability
at the edge of the panel (between 15–20 m) is about one to two orders of magnitude greater
than the permeability over the abutment pillar (0–15 m). The underlying reasons for perme­
ability differences are the abutment loading and caving process. At the center of the pillar, the
loads from retreated panels compact the fractures and reduce their aperture. Approaching the
panel edge, bedding plane separation and caved blocks significantly increase the fracture aper­
tures and thus their permeability.

512
Figure 6. Comparison of permeability for shallow and deep cover.

4 CONCLUSIONS

This study presents the comparison of mining-induced permeability for a deep and shallow
site in southwestern Pennsylvania. A discrete fracture network model is developed in FFC
with aperture values imported from the geomechanical model in 3DEC. This stochastic DFN
model generates 100 realizations to account for potential variation in fracture geometry and
the average for each zone is compared. The findings from this study shows that:
i. For both shallow and deep cover, there are notable permeability changes for the strata
within 30–40 m of the mine roof, and a shale gas casing breach within this horizon could
provide pathways for gas flow toward the mine,
ii. Under shallow cover, the high permeability zone for each stratum at the edge of the panel
could extend into the gob to about 100 m and,
iii. Under deep cover, there are multiple locations over the gob with very low permeability
values such as in-situ or undisturbed rock mass.
These findings provide insights into the potential impact of a casing breach under different
mining depth and could help toward development of best practices for the safe operations of
both coal mining and shale gas extraction.

DISCLAIMER

The findings and conclusions in this paper are those of the authors and do not necessarily
represent the official position of the National Institute for Occupational Safety and Health,
Centers for Disease Control and Prevention. Mention of any company or product does not
constitute endorsement by NIOSH.

REFERENCES

Ajayi, K., Khademian, Z., Schatzel, S., Watkins, E. & Gangrade, V. (2021) Numerical modeling of
longwall-induced permeability under shallow cover. Mine Ventilation. CRC Press.
Ajayi, K., Khademian, Z., Schatzel, S., WATKINS, E. & Gangrade, V. (2022) A Discrete Fracture Net­
work Model for Prediction of Longwall-Induced Permeability. Mining, Metallurgy & Exploration, 1–8.
Commonwealth of pennsylvania (1957) Joint Coal and Gas Committee, Gas Well Pillar Study. Harris­
burg, Pennsylvania:, Department of Mines and Mineral Industries Oil and Gas Division

513
Khademian Z., Ajayi, K. M., SU, D. W. H., Schatzel, S. J., Kim, B. H. & Esterhuizen, G. (2022) Rock­
mass permeability induced by longwall mining under deep cover: potential gas inflow from a sheared
gas well. SME Annual Conference & Expo. Utah.
Mine safety and health administration (2018) CFR § 75.323, Title 30, Electronic Code of Federal Regula­
tions. Mineral Resources, Department of Labor.
Pennsylvania department of environmental protection (2022) Longwall Mine Data. https://www.dep.
state.pa.us/dep/deputate/minres/LongwallForestlandStudy/Table1.pdf.
Su, D. W., Zhang, P., Van Dyke, M. & Minoski, T. (2018) Effect of Cover Depth on Longwall-Induced
Subsurface Deformations and Shale Gas Well Casing Stability. 52nd US Rock Mechanics/Geomecha­
nics Symposium. American Rock Mechanics Association.
Su, D. W., Zhang, P., Van Dyke, M. & Minoski, T. (2019) Effect of longwall-induced subsurface deform­
ations on shale gas well casing stability under deep covers. International Journal of Mining Science and
Technology, 29, 3–8.
Watkins E, G. V., Schatzel S, Hollerich C, Addis J (2020) Permeability Determination for Potential Inter­
action between Shale Gas Wells and the Coal Mine Environment due to Longwall-induced Deform­
ations. 2020 SME Annual Meeting and Exhibit. Phoenix, AZ.
Watkins, E., Karacan, C. Ö., Gangrade, V. & Schatzel, S. (2021) Assessing Gas Leakage Potential into
Coal Mines from Shale Gas Well Failures: Inference from Field Determination of Strata Permeability
Responses to Longwall-Induced Deformations. Natural Resources Research, 1–14.
Zhang, P., Dougherty, H., su, D. & Trackemas, J. (2019a) Influence of Longwall Mining on the Stability of
Gas Wells in Chain Pillars. Proceedings of the 38th International Conference on Ground Control in Mining.
Zhang, P., su, D. & Lu, J. (2019b) Influence of Longwall Mining on the Stability of Shale Gas Wells in
Barrier Pillars. American Rock Mechanics Association. New York.
Zhang, X., Sanderson, D., Harkness, R. & LAST, N. (1996) Evaluation of the 2-D permeability tensor
for fractured rock masses. International journal of rock mechanics and mining sciences & geomechanics
abstracts. Elsevier.

514
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

New applications of jet fans in underground mines for haulage

ramps and block cave ventilation control

C. Stewart
Minware, Brisbane, Australia

ABSTRACT: Jet fans are extensively used in civil road tunnels, large enclosures (such as under­
ground car parks) and other applications requiring the movement of air in large low-pressure
voids but are rarely systematically used in underground mining. Despite a relatively low cost and
compact size compared to typical mine axial fans, jet fans have found little practical use in most
mines, apart from occasional applications in coal and room and pillar mines. Jet fans have limited
pressure capabilities and are inefficient in moving airflow through higher-resistance airways.
This research explores the technical merits of jet fans for new applications in mining envir­
onments and highlights areas where the fans may be highly beneficial compared to conven­
tional fans or ventilation controls such as louvres or doors. Jet fans can excel in applications
that benefit from producing targeted ventilation flows while allowing unrestricted vehicle
movement that would otherwise be hindered by doors or ventilation controls.
An overview of the theory and application of jet fans in mining environments is provided
and examples and theoretical effectiveness of potential uses in ramp flow control and block
cave extraction ventilation is researched using modelling software. Further detailed design
and practical test work for specific applications is intended for future research.

1 INTRODUCTION

1.1 Background
Jet fans, sometimes known as impulse or induction fans are derived from a specialised vari­
ation of axial fans and came to prominence in the early 1960s in US road tunnels (ErLeong &
Mashrae 2022). The technology was later enhanced by Japanese fan manufacturers in the
1970s and then in South Korea in 1986.
While jet fans can be based on either axial or centrifugal design, the most common type is axial
due to compact sizing and the useful horizontal in-line thrust vector produced for tunnelling.
Axial jet fans consist of a cylindrical fan casing designed to generate a high-velocity jet of air
from the discharge of the fan. The accelerated airflow through the fan generates an impulse or
‘thrust’ which is imparted and spread to the surrounding airflow via the jet stream. The jet can be
directed into large chambers to create air movement (Drenda et al., 2018) or for improved fan
ventilation (Konduri et al., 1998), however, the primary use of jet fans is for tunnel ventilation by
creating a net positive pressure and increasing (or decreasing) overall air movement to greater
amounts than the discharge air from the fan alone (ErLeong & Mashrae 2022).
The overall static pressure induced for large tunnels is relatively low, rarely exceeding a few
tens of pascals (10 – 30 Pa) however in low resistance applications this may be sufficient to
generate significant changes in airflow, particularly when multiple fans are used in parallel or
spaced along the tunnel in series. Jet fan thrust produced is not significantly influenced by sur­
rounding tunnel air velocity, working well in contraflow, counterflow or even stagnant tunnel
conditions.

DOI: 10.1201/9781003429241-53

515
1.2 Jet fan design
Jet fans are a relatively simple and robust design consisting of an inline impellor, motor, motor
shroud and straightener arrangements within a cylindrical casing, and are often supplied with
inlet and exhaust silencers to reduce noise. Unlike higher-pressure axial fans, the blade size to
hub ratio tends to be high to optimise low pressure movement of the high air volumes and the
casing design from inlet to exhaust is usually straight with no expanding discharge diffuser.
Fan capabilities are normally specified by thrust (N) produced in a static air environment,
and specifications may also include volume flow, discharge velocity, fan power, and relevant
dimensions data such as diameter and width. Fans can also be designed in variable speed and
reversible configurations to provide optimum control of thrust and direction. Fully reversible
or bidirectional fans can produce similar thrust in both directions. Partially bidirectional fans
can produce highly efficient thrust in one direction but lower efficiency in reverse, while uni­
directional is only designed to produce thrust in one direction. Theoretical fan efficiency is
based on the motor shaft power versus the velocity pressure and air volume flow produced
but specifications often rated efficiency on the thrust (N) to power (kW) ratio, with high
ratios indicating better efficiency.

Figure 1. Example of road tunnel jet fans (FläktWoods, 2020).

The most important data for performance calculations is the theoretical thrust produced,
which is a product of the kinetic energy imparted to the air mass through the fan into the dis­
charging jet stream. The momentum of the discharged air mass is diffused and transferred
into the surrounding air mass creating a net positive pressure. Thrust in Newtons can be calcu­
lated as follows.

Where ρ = air density (kg/m3), Q= airflow volume (m3/s) and A = Fan outlet area (m2)
The theoretical thrust however is eroded based on installed conditions, and the actual or
installed thrust depends on the prevailing air velocity into the fan, the placement of the fan in
relation to the centreline of the drift, and the angle of the airflow relative to the surrounding
wall. The installed thrust can be estimated by the following equation.

Where tunnel velocity correction, wall proximity correction factor and jet inclination correc­
tion factor, ranging from 1.0 (horizontal) to 1.1 (5° - 7° tilt to centre).
To convert thrust into a pressure rise across a tunnel, the following equation can be used.

516
where A = Area of Tunnel (m2)

Figure 2. Calculation of k1 and k2 factors (compliments (FläktWoods, 2020).

where DF = Jet Fan Diameter (m), DT = tunnel diameter (m), Z = distance of jet fan axis to
tunnel wall or ceiling (m)

1.2.1 An example
A fan manufacturer offers an 8.5kW jet fan at a specified 278N of thrust, an outlet velocity of
30.5 m/s, a diameter of 560mm, and is installed in a 5m x 5m tunnel with airflow of 2.0m/s. The
jet fan is installed horizontally in line with the tunnel axis, with the fan centre axis 1.0m below
the ceiling. The applied fan thrust and pressure rise due to the fan can be calculated as follows.
Separation Factor =(2 × 1.0 – 0.56/(5.0 - 0.56) = 032
k1 = 0.93, k2 = 0.94 (from Figure 2), k3 = 1.0 (horizontal alignment)
Installed Thrust (Ti) = 278 × 0.93 × 0.94 × 1.0 = 243N
Pressure Rise Δ P=243/25 =9.7 Pa

The example above shows the induced pressure of a jet fan can be relatively small, and of
little use for conventional air movement through the shafts and tunnels of a large mine. In
comparison, the applied pressure of a typical mine axial fan through a pressure seal such as
a door or bulkhead can be several magnitudes greater than the 9.7 Pa shown in the above
example.
The advantage of a jet fan however is that it can help distribute or correct airflow in low-
pressure, open environments, without the need for a sealing door, brattice or bulkhead that
would otherwise restrict traffic flow or add to ventilation infrastructure cost.

2 JET FAN APPLICATIONS IN UNDERGROUND MINES

Mines have a safety and legal responsibility to ensure adequate airflow is delivered to all
working areas of the mine to meet or exceed minimum velocities, satisfy diesel equipment
requirements, and achieve dilution standards. In a large dynamic mine with many work areas,

517
this requirement can be challenging and may require complex and expensive ventilation circuit
arrangements, frequent adjustment of controls, or overventilation of areas to ensure minimum
standards are always met.
In most cases, the level of control required comes at the cost of substantial resources and
mining infrastructure, expensive and restrictive (to traffic) controls such as doors and louvres,
energy for ventilation fans, and of course manpower to manage and ensure everything works
as required.
In some applications however, jet fans may alleviate much of this burden by removing the
need for some controls and allowing optimisation of airflow without wastage.
In road tunnels for example, as illustrated in Figure 3, the airspeed is often regulated
depending on the volume of traffic, gas concentrations and the time of day. A control system
using air velocity sensors will activate jet fan operations, adjusting the fan speeds and some­
times the direction to ensure the targeted air velocity is achieved. Vehicle movement piston
effects are often enough to drive much of the flow, but during traffic slowdowns or jam condi­
tions, the jet fans will ensure that targeted flow is achieved.

Figure 3. Jet fan components & considerations for tunnel air velocity control (not to scale).

The figure above shows an example of methods of controlling air velocity depending on
traffic volume and speed at the time. The jet fan operation is controlled via a linked PID (pro­
portional, integral, derivative) control system that operates jet fan numbers, speed, and direc­
tion. During times of rapid, high-volume vehicle movement, piston effects may be sufficient to
move the majority of required air volume through the tunnel, and minimal jet fan use may be
required. During low-volume vehicle times or when traffic slowdowns occur, more jet fans are
run (and at higher speeds) to achieve safe air volumes. Reversible jet fans can even be used to
retard ventilation flow, so excessive ventilation is not consumed at the expense of adjoining
tunnels, or to prevent tunnel fumes from exiting the portal.

2.1 Performance and efficiency comparison of high-pressure auxiliary fans vs jet fans
Auxiliary fans are occasionally used in mines for jet fan applications and are free-hung with­
out ducts or walls to boost flow along stagnant drives or push air into dead-end headings or
large chambers. However, the efficiency and effectiveness of using conventional auxiliary fans
in jet fan applications should be considered. The fan total pressure efficiency of a high-
pressure auxiliary fan operating under no static pressure may be only 40% or lower.
A comparison of an auxiliary and a jet fan is provided below.
A 1259mm diameter and 55kW motor power (Howden AXN-1259) using Equation 1 sug­
gests a theoretical thrust of up to 1185N would be possible. A similar specialised jet fan design
of the same power and size (Howden APA1250/578) is catalogued (Howden Ventilatoren
GmbH, 2022) to produce almost double the thrust at 2025N. A smaller, lower power jet fan
(e.g., Howden APA 1000/403 1000mm fan with 30kW motor) will produce the same thrust as
the higher powered larger 55kW mine auxiliary fan.

518
 In summary, it is rarely an economical solution to employ high-pressure auxiliary fans in jet
fan applications unless specialised discharge reduction cones are used to improve fan jet per­
formance (Wolski, 1995) or the fan is employed in very short-term applications for convenience.

2.2 Jet fan influence on parallel airways

The control of airflow between parallel airway paths is one of the more difficult ventilation
problems for mines, requiring restrictive doors, balancing regulators, or other complex
arrangements. A parallel airway configuration may occur between connecting orebodies, mul­
tiple parallel haul ramps or in the case of block cave mines as a design requirement for mul­
tiple extraction or haulage levels crosscuts across the cave footprint.
A jet fan can have a huge influence on airways with similar static pressure on both sides as
shown below in Figure 4. In this example, 500m long crosscuts connect two similar pressured
perimeter drives. While there will be frictional losses and system pressure differences in the
perimeter drives, these can be offset by injection or rejection of airflow at strategic locations
along the perimeter drifts. The calculations for the estimated airflow influence of a jet fan in
a simple parallel drive arrangement are as follows.

Figure 4. Theoretical pressure balance equation for parallel drift flow control.

Kirchhoff’s law applied to ventilation and used in the Hardy Cross algorithm indicates that
the sum of the pressure losses (or gains) in any loop circuit must equal zero. The figure above
can be described by the following equation.

Where P1ab and P2ab is the frictional loss (Pa) of the perimeter drift between crosscuts
Pja and Pja is the combined crosscut frictional pressure loss and jet fan addition (Pa).
Pva and Pvb is the piston effect (Pa) of the vehicle movements in the drifts.
Assuming similar system pressures at P1ab and P2ab at both ends, Atkinson’s formula can be
used to estimate airflow influence from a jet fan. Assuming the piston effect of the vehicle is
negligible, a +/- 15 Pa jet fan pressure influence is available (the capability of a modest-sized
11kW jet fan). The airflow difference in a 500m long 20m2 drift can be calculated as shown
below.

Where k = Friction factor (kg/m3), C = circumference (m), L = length (m), A = area(m2)

Assuming a friction factor of 0.012 kg/m3 (ρ = 1.2 kg/m3) and a drift circumference of 18m.

519
 In summary, the application of jet fans in equal pressure crosscuts with a thrust influence of
+/-15 Pa can change the airflow in each by over +/- 30 m3/s. The movement of vehicles may
increase or decrease this amount due to piston effects, depending on direction, however, the
average overall increase remains the same. This offers a huge range of ventilation control poten­
tially suitable for most large diesel equipment. The above calculations of course can be modelled
more quickly by ventilation simulation software to help design and prove larger systems.

2.3 An application for jet fan airflow regulation in haulage ramps

Haulage ramps in underground mines can create unique ventilation problems associated with
the natural ventilation pressure of heated air, and from the piston pressure effects of large
moving machinery. The effect is often exacerbated in ventilation designs that have multiple par­
allel pathways for airflow (for example two ramps system joined to the same primary circuit).
The piston effect is a well-known phenomenon produced by both the aerodynamic drag of
moving machinery, and the displacement of air by a moving object in a limited-size void. The
movement of trains in subway systems is a familiar concept, with high-pressure zones ahead
of the moving train, and low-pressure zones behind the train creating air movement through
train tunnels.
Truck and other vehicle movements within a ramp system create similar effects, and while the
bi-directional movement of vehicles should in theory dampen the effect, higher downward speeds
and sporadic vehicle movements can create highly variable airflow during a working shift.
Another cause for ventilation variation is thermal heat effects on natural ventilation. While
normally small, the influence of both natural ventilation and piston effects can become very
noticeable in parallel pathways. The effect can (for example) cause ventilation to change or
even reverse as the intensity of vehicle traffic increases the heat in a region. The change is
often perplexing to ventilation professionals, who may measure good ventilation at the start
or end of a shift, only to find it stagnant or reversing at busy times.
An example was modelled in Ventsim. A dynamic (transient) simulation of six (6) trucks
travelling up and down a ramp. The truck movement piston effect and heat both at times
boost and retard the primary airflow. While the lower ramp airflow in this model is relatively
stable due to the influence of dedicated primary ventilation controls, the upper ramp shown in
the figure sits adjacent to competing ventilation regions and is more easily influenced. The
movement of trucks in this area adversely affects air movement, causing it to drop to hazard­
ously low levels. The results are shown in Figure 5.
A single theoretical 650mm 15kW jet fan with an airflow velocity-based sensor, control
system and speed controller is placed in the ramp to generate pressures to stabilize the airflow
to a nominal 25 m3/s (sufficient for a single truck passing through this decline section).
Figure 6 shows the theoretical effectiveness of the jet fan operation, with near-complete sta­
bilisation of the airflow based on frequent changes in the jet fan speed and direction to create
induced pressures of -10 Pa to +15 Pa in the ramp. In this example, despite a maximum power
of 15kW, the jet fan typically operates well under maximum power.
While the above example appears effective in theory, it should be noted that actual jet fan
control may not be as responsive as the simulated example and the PID controller would need
to be tuned and optimised to deliver the best results. In addition, airflow setpoints that require
pressure beyond the jet fan capabilities would only be partially achieved, unless larger or mul­
tiple jet fans in series or parallel configurations were used.

2.4 Conceptual application of jet fans in large block cave haulage and extraction
Block cave mines are usually designed with multiple parallel extraction and haulage drifts
accessed by perimeter or ‘rim’ drives and ventilated with intake air entering one or both sides

520
Figure 5. Variations to airflow (left) with no jet fan control at the model ramp location shown (Right).

Figure 6. Airflow controlled (left) with jet Fan variable pressure (right) to achieve 25 m3/s.

and exhausted via regulated raises in the mid-section or opposite side to a separate ventilation
level. Multiple parallel pathways create unique ventilation difficulties for control, with even
small variances in rim drive pressure, airways resistance or vehicle activities causing uneven
flow distribution. For effective jet fan control, the resistance or pressure differences must be
small, which restricts a ventilation design layout to one that has similar resistance along all
available pathways.
The concept is shown in Figure 7 where the left design has similar distances from primary
intake to exhaust regardless of the parallel path taken, while the right design has increasingly
long distances between intake and exhaust, the further away the parallel drift is from the pri­
mary circuit. The variable resistance caused by different distances and therefore the large vari­
ation in pressure losses means the right-hand design is likely unsuitable for low-pressure jet
fan control.
For efficiency and productivity, both sides of the cave footprint are often worked from the
same extraction drive, so a central exhaust raise and regulator must be used and adjusted to
suit activities at the time. Ideally, for two-sided access, the central exhaust would draw in
equal amounts of air from both sides however this is rarely the case because of unequal pres­
sure bias between each extraction drift side. The concept is shown in Figure 8 where an intake
air drift distributes fresh air via raises and regulators onto the extraction footprint perimeter
rim drive to equalise pressure. A central exhaust drift below the extraction level draws exhaust
air through connected raises and regulators.
Methods can be implemented to control the airflow variability, including:

521
Figure 7. Schematic of parallel drift showing equal path resistance design vs unequal path resistance.

• An alternative single-sided entry and production with an exhaust raise at the end.
• Doors centrally or on both sides to control flow.
• Regulating intake flow and pressure along the perimeter rim drives.
• Over-ventilating the area to ensure minimum flows are achieved.
None of these methods however are particularly efficient or cost-effective and may impede
traffic or result in poor control of airflow and dust.

Figure 8. Conceptual block cave extraction level layout with example jet fan locations.

2.4.1 A jet fan solution

Provided sufficient airflow can be delivered to both sides of the perimeter rim drive, the rela­
tive differentials of pressure between extraction drift sides can be relatively small. A compact
extraction drift size area of 20m2 and 200-400m in length, with airflow requirements of 20 m3/s,
needs very little pressure to influence flow as shown in the example in section 1.2.1.
Installing jet fans in each extraction drift – either a single bi-directional jet fan mid-drift or
uni-directional jet fans on each end near the rim drive, together with an ultrasonic airflow
sensor, provides much of the control needed to balance airflows. Provided the central exhaust
regulator is adjusted accordingly, the jet fans (through a control system) can help provide
equal flow on both sides (for two-sided operation), or occasionally confine airflow on one side
(for a single loader operation).
A jet fan solution was modelled for the example shown in Figure 8. Regulators were used to
distribute fresh air along the rim drive, and exhaust regulators were used to regulate net

522
airflow into each extraction drift. A model simulation of a Howden APR 800/330 (800 mm
diameter) reversible 22 kW jet fan with a static thrust of 710 N was used to provide an esti­
mated maximum 30 Pa pressure influence in a 20 m2 drift.
Network modelling of jet fan systems cannot simply use ‘fixed’ flows as this gives an unreal­
istic expectation of available jet fan pressures. Instead, an automated flow-regulated jet fan
effect was modelled by using a conventional fan curve (shown in Figure 9) to apply positive or
negative jet fan pressure bias of up to 30 Pa to target a set flow point of 20 m3/s for a loader
operation. Slightly higher bias pressures were allowed outside the target range to improve
simulation convergence and reflect the change in jet fan effects in counterflow conditions.

Figure 9. Modified Drift Pressure ‘Fan’ Curve for Jet Fan Operation.

Three types of extraction drive activities were modelled within the same footprint.
• dual production requiring full airflow on both sides (20 m3/s per side, 40 m3/s exhaust)
• single-sided production requiring full airflow on one side only (20 m3/s exhaust).
• temporarily no activity with minimised flow (assumed 2m3/s exhaust leakage)
An attempt to model equal flows in extraction drifts without jet fans is shown in Figure 10.
The solid line represents a target airflow, and the columns show modelled airflow achieved.

Figure 10. Airflow modelling in extraction drifts #1 to #26 without jet fans.

523
 Drifts 1-4 and 24-26 are open but not operational and require only minimal airflow, while
the remaining central drifts are in production and require 20m3/s each. Only one side of the
footprint results are shown in the figure. Airflow is well outside the target range in many
extraction drifts, with some having insufficient airflow, and others wasting too much airflow.
Further design work could be done to try to improve shortfalls by balancing regulators, how­
ever this was generally found to increase air wastage in unused drifts or drifts that already
have adequate ventilation.
When jet fans are applied along each drift as shown in Figure 11, an even flow distribution
at the targeted airflow is achieved in all drifts. The required jet fan pressures (shown in the
shaded curve) vary from -20 Pa (reversed) to 10 Pa (forward), with an average jet fan influence
of 8 Pa or around 25% of maximum jet fan power. The requirement for fine adjustments of
most regulators is removed with only broad intake system adjustments required to ensure
both perimeter rim drifts have sufficient airflow, and the jet fan pressures remain operating
within their working range.
Of particular benefit is the ability of jet fans to use counter-flow to reduce airflow in tem­
porarily inactive crosscuts provided some airflow remains for cooling the fans. While outside
the scope of this paper, a similar arrangement could also be applied to haulage level ventila­
tion controls where multiple parallel pathways to different haulage chutes or areas could be
influenced without the need for ventilation controls and doors which impede traffic.

Figure 11. Airflow modelling in extraction drifts #1 to #26 with jet fans.

2.5 Operational considerations

It is acknowledged that many challenges in operating jet fans in production areas need to be
overcome. The placement in production areas subjects the jet fans to vehicle clearance issues
and potential blast damage. The heat generated by jet fan operation may also contribute to
temperature buildup, particularly in stagnant areas where jet fans can recirculate. The power,
maintenance and repair cost of operating jet fans must also be considered and the reliability
issues of sensors and control automation in difficult areas are also acknowledged. Each point
is addressed briefly below.
Clearance: Jet fans are relatively small in diameter (smaller than the duct used to originally
ventilate the development headings) and clearance can be further improved by shallow
recesses in the back or drift sides if required.

524
 Blast Damage: Jet fan design is relatively simple and robust, and the fans are free hanging
and therefore not subject to stagnating blast pressures like regulators or doors. To increase
proximity distances from blasting and production on a two-side extraction footprint, instead of
using a mid-extraction drive located bi-directional fan, a single-direction jet fan could be located
at both ends of the extraction drive on the rim drive directing jet airflow into each side.
Heat and Recirculation: Generated heat from jet fans is expected to be relatively small com­
pared to mobile equipment present, with an expected wet bulb increase (for example) of 0.3°C
for a 30kW jet fan at full power inducing 1m/s velocity of air in a 25m2 drift. In most cases, jet
fans will run substantially less than full power. For lower velocity ‘retarding’ operation, recir­
culation and heat buildup are a concern, however, even at 0.25m/s velocity for the above
example, the temperature increase is only 1.2°C (unless operated in a recirculating restriction
mode to reduce airflow).
Cost: The electrical cost for a single 30kW jet fan operating at 25% duty at a mine electrical cost
of 15c/kWh will be $9800 per year, a modest cost to pay for the control provided, and the likely
significant saving compared to the additional airflow that would otherwise be required to ensure
minimum flow standards are met. As within auxiliary fans, occasional maintenance of blades and
motors through wear and tear will be required, however, the relatively low capital cost of jet fans
may relegate them to a disposable item after several years of use in dusty environments.
Communications: Digital communication networks and automation and a standard features
in most modern mines, and are robustly constructed to ensure any communication, automa­
tion or sensors can be installed in most areas. Wi-Fi communication can be an option in par­
ticularly hazardous areas where communication cables may be damaged.

3 CONCLUSION

The capabilities of jet fans for use in underground mining have likely been underappreciated,
and the application of jet fans may have significant ventilation control, cost, and safety bene­
fits. The use of jet fans and ventilation control systems for road tunnels is a well-proven tech­
nology but the methods have not generally been transferred to underground mines, perhaps
because of preconceived notions of the limited effectiveness of jet fans in high-pressure, high-
resistance environments.
The examples presented in this paper demonstrate the potential of jet fans in unique appli­
cations which would otherwise need to be solved with over-designed ventilation airflows to
counter wastage, or with numerous, expensive, and complex ventilation controls and infra­
structure. Other issues such as improved control of gases, dust and radon could also be ana­
lysed. While the flexibility, effectiveness, and robustness of jet fans in these applications have
only been demonstrated as a concept with computer modelling, with further test work, control
system development, and cost analysis, these examples and many more applications may bene­
fit. Jet fans have the potential to remove significant complexity and cost while improving effi­
ciency, productivity, and flexibility in many underground mine applications.

REFERENCES

Drenda, J., Domagała, L., Musioł, D., Pach, G. & Różański, Z. (2018) An analysis of selected jet fans
used in chambers of KGHM mines with respect to the air stream range. Tunnelling and Underground
Space Technology, 82, 303–314.
Erleong, C. W., Mashrae Mies (2022) Enhancing the design of ventilation systems. The Singapore Engin­
eer. The Singapore Engineer.
Fläktwoods (2020) Jetfoil and Large JM Aerofoil Fans for Tunnel Ventilation. FläktWoods.
Howden Ventilatoren Gmbh (2022) Air for Underground Traffic. Howden.
Konduri, I., Mcpherson, M. & Topaz, E. (1998) Model and field investigations of jet fans for face
ventilation.
Wolski, J. K. (1995) Performance analysis of the jet fan used as a booster fan. Society for Mining, Metal­
lurgy, and Exploration, Inc., Littleton, CO (United . . ..

525
Occupational health and safety in mine ventilation
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Ventilation research findings for enhanced worker safety when

mining near unconventional gas wells in longwall abutment pillars

S.J. Schatzel, K.M. Ajayi, Z. Khademian, R. Kimutis, M.L. Harris, M. Van Dyke &
J.D. Addis
CDC/NIOSH/PMRD/MSSB

H. Dougherty & E. Watkins

Formerly National Institute for Occupational Safety and Health, Pittsburgh, PA, USA

ABSTRACT: Unconventional gas wells continue to be drilled through current and future
coal reserves in Pennsylvania, West Virginia, and Ohio. The ability for both mining and gas-
producing activities to coexist safely is a continuing question for federal and state regulatory
entities as well as for industry representatives. A hypothetical, gas well casing failure result­
ing from mining-induced ground movements could produce unsafe conditions and an explo­
sion hazard in nearby operating mines. The National Institute for Occupational Safety and
Health (NIOSH) is conducting research to characterize a hypothetical breach from an
unconventional gas well near an operating longwall coal mine and any resulting mine safety
consequences for a range of mining conditions. Regional conditions include overburden
depths of under 152 m (500 ft), between 152 m (500 ft) and 274 m (900 ft), and overburden
depth over 274 m (900 ft). Geologic conditions also include typical southwestern Pennsylva­
nia topography and stream valley environments. Differing ventilation schemes commonly
used in Pittsburgh coal bed mines have been considered in assessing the distribution of gas
from a hypothetical breach in mine workings. Multiple technical approaches are utilized to
address the ventilation research questions. These methods include experimental techniques,
analytical methods and interpretations, and numerical and physical modeling tasks. The
NIOSH team is providing scientific input to regulatory and industry partners in the develop­
ment of new recommendations for shale gas wells influenced by longwall mining.
A summary of the current findings by the research team, across all tasks, is provided.
Although this summary represents years of research by the NIOSH team, refinements to the
technical interpretations may continue to be produced over the life of the project as more
research data become available.

1 INTRODUCTION

Longwall mining induces surface and subsurface subsidence producing stress changes in the
overburden. Gas wells near longwall mining are protected by abutment pillars or barrier pil­
lars. Interested parties have raised some concerns about the possibility that mining condi­
tions could produce excessive stresses that result in deformation and damage to gas well
casings and suggested that regulatory guidelines need to be updated (Commonwealth of
Pennsylvania, 1957, 2017; John T. Boyd Co., 2016; Mark & Rumbaugh, 2020). NIOSH is
providing scientific findings from mining research to improve worker and safety and health
(Su et al., 2020; Schatzel & Su, 2020; Zhang & Su 2021).

DOI: 10.1201/9781003429241-54

529
2 EXPERIMENTAL METHODS

This research endeavor is a combined ground control and ventilation research effort. To be
successful, the two overall research components broadly interact as the movement of ground
produces fracturing which is a critical component of the hypothetical inflow hazard. Mining-
induced fracturing related to the full extraction mining of the coal seam produces transport
pathways for the movement of gas away from the hypothetical breach towards the operating
mine resulting in a significant safety risk. This discussion is focused on the ventilation portion
of the research with our goal of providing scientific input for the industry and operations asso­
ciated with the longwall mining of coal near unconventional gas wells.
To conduct research on ventilation objectives, a series of technical approaches were devel­
oped Figure 1 shows the technical approaches that continue to progress in this research. They
can be broadly categorized as methods to estimate the amount of hypothetical inflow, tech­
niques to produce portrayals of movement through overburden and mine ventilation systems,
and a new methodology to distinguish gas sources. A brief summary of each of the technical
approaches is included in the subsections that follow.

Figure 1. Overview of technical approaches used in the ventilation research component.

The review of these approaches follows the path and chronological order of gas migrating
away from a damaged well in the event of a hypothetical casing breach. Gas is transported
away from the well through the fractured overburden for different sites, overburden depths
and stratigraphic units. The fractures in the overburden are characterized in terms of the frac­
ture aperture and permeability in the strata and then an estimated inflow quantity to the
nearby mine is given. Gas movement from the hypothetical breach is simulated moving
through the ventilation system and concentrations of added gas are quantified in key areas of
the mine. Findings from NIOSH’s proposed methodology for distinguishing gas sources as
contributors to the gas emissions underground are reviewed.

2.1 Field measurements of pressure for monitoring

Field measurements for permeability monitoring are done in experimental boreholes as part
of the ventilation research component. These boreholes are configured by NIOSH to monitor
changing permeabilities at specific stratigraphic horizons of interest (Watkins et al., 2021).
These units are chosen based on previous experiments or ground movement data as being sites
of likely high-magnitude, horizontal movement. It is these positions that are considered the
most probable locations for a hypothetical breach. Boreholes are configured with one section
open to the stratigraphic unit of interest. Communication between the inner casing annulus
and surrounding strata occurs through slotted casing matching the thickness of the

530
stratigraphic unit. Falling head slug tests are performed with the addition of incremental
amounts of water and head loss rates are measured with a downhole piezometer.

2.2 Discrete Fracture Network (DFN) 3DEC simulations

Simulations are made using commercial DFN 3DEC code which incorporates geomechani­
cal changes in the rockmass. It describes fractured rocks as a population of individual frac­
tures. A pseudo-2D, two-panel longwall model is constructed for analyzing the fracture
permeability induced by the mining deformation. All fluid transport is assumed to move
through the fracture system. Lithologic descriptions from the study site are used to describe
strata and fracture locations in longwall gobs. Rockmass permeability can be estimated
through calculating the fracture permeability, which is a function of the fracture hydraulic
aperture and fracture geometry. The model assumes a constant starting aperture width.
Changes in overburden rock stresses as the face advances modify the aperture width (Khade­
mian et al., 2022). Permeability data are used to produce inflow simulations using a second
DFN technology described below.

2.3 Discrete Fracture Network (DFN) Fracture Flow Code (FFC) simulations

The DFN FFC analysis simulates fluid transport through a stochastic representation of fracture
systems. This technology is used to map fracture apertures in longwall gobs following a log
normal distribution. This study develops a stochastic DFN model for the site using fracture
parameters obtained from the site core log data and geomechanical analysis. Flow from
a breached gas well casing is modeled through the strata’s fracture network using cubic law. Frac­
ture position data is produced from lithologic input based on information from field sites and
from DFN 3DEC results to generate 100 two-dimensional realizations (Ajayi & Schatzel, 2020).

2.4 Computational Fluid Dynamics (CFD) simulations-Ansys Fluent

CFD is a type of fluid mechanics simulation that uses numerical analysis to analyze and solve
problems that involve fluid flows. A finite volume model was created, using Ansys Fluent, to
simulate gas transport from the hypothetical breach through fractured rock. into the mine.
and through the ventilation network. This technique provides a numerical analysis of fluid
flow showing interactions of fluids (air-methane species). This Ansys Fluent model incorpor­
ates field site permeability and mine ventilation data to estimate methane inflow in a worst-
case scenario of a ruptured shale gas wall drilled through a longwall abutment pillar. The
caved and, fractured and zones that represent the overburden of the CFD model are described
as three-dimensional porous zones, each with a prescribed permeability value that is constant
and isotropic. The volumes of flow are based on DFN results and defined by boundary condi­
tions (Ren et al., 2011; Watkins & Gangrade, 2022). The resulting simulation is a full-scale,
two panel longwall model.

2.5 Network Simulations, VentSim

Simulations of gas from hypothetical breaches into a longwall mine’s ventilation system
were done using VentSim commercial software. Ventilation air transport rates in Pitts­
burgh seam mines were used in this and all ventilation-based tasks. This technology is
a circuit-based approach with airflow through branches. The mine model data were used
as input to the VentSim software and a section of gob was added. The gob consisted of
an approximately 10 m by 10 m grid. The gob in the model consists of 4 different resist­
ances similar to the permeability zones in the LIAM model. Quantified inflows from the
DFN simulations were included to add gases at nodes in the network with guidance
from CFD simulations on the transport pathways (Dougherty et al., 2022). The VentSim
simulations provide a strong visual representation of gases moving through the ventila­
tion network.

531
2.6 Longwall Instrumented Aerodynamic Model (LIAM)
A 1:30 scale physical model of a longwall panel that replicates a cooperating Pittsburgh coal
bed mine was built. The ventilation and reservoir properties have a high degree of similitude
with full-scale mines operating in the study area. To ensure good geometric scaling, the LIAM
is built with a very high attention to detail, including shields, cribs, 3D printed shearer, regu­
lators, standing supports, gob vent boreholes, and curtains. To simulate the reservoir proper­
ties of the gob for the Pittsburgh seam, five types of materials having of different
permeabilities are used. The model was adapted to incorporate known quantities of dilute
sulfur hexafluoride tracer gas as a proxy for hypothetical breached gas mixed in mine air.
Flow paths can be defined, and the increase in methane (CH4) concentrations can be quanti­
fied in the LIAM using this technology (Gangrade et al. 2019, 2022).

2.7 Gas compositional analysis, distinguishing gas sources

Other sources of gas, primarily hydrocarbons not produced from coal seams, are present near
longwall mines in southwestern PA and potentially other mines in U.S. locations and inter­
national sites. Additions of gas from these sources could produce hazards for operating long­
wall mines, particularly when combined with conventional coalbed gas emissions. There is
a possibility that underground storage facilities or shale gas wells could be contributors in
hypothetical, unplanned mine emissions. A methodology to distinguish potential gas contribu­
tions from these sources was developed based on interpretations of gas chromatography (GC)
results (Schatzel & Su, 2020; Schatzel et al., 2022). The overall composition of these gases is
primarily related to the type of organic matter that produced them and to the thermal matur­
ity, or rank, of the organic material generating the gas, forming a theoretical basis for distin­
guishing these sources.

3 RESULTS AND DISCUSSION

In NIOSH’s efforts to protect miners throughout the range of scenarios that may present
themselves during longwall mining near unconventional gas wells, several key, impactful vari­
ables were identified as most influential. The described technical approaches were utilized to
provide scientific input for safety enhancements in the event of a hypothetical well breach.
These variables and the current technical findings are discussed below.

3.1 Modification of overburden permeability due to mining-induced fracturing

Overburden depth is very impactful in the creation of mining-induced fracturing permeability.
For first panel mining, there is a significant increase in permeability in going from deep
(>274 m, 900 ft) to shallow cover overburden (<152 m, 500 ft), measuring two orders of mag­
nitude greater for the maximum value in the shallow case (Figure 2 and Figure 3). These are
the regional depth categories used in the study region. Fracture permeability, produced by
longwall mining, substantially increases the overburden fracture permeability over in-situ
values and changes with longwall face position (Figure 3). Second panel mining under shallow
cover (<152 m, 500 ft) increases maximum permeability by about 1.5 to 3 times the maximum
from the first panel mine-by according to the experimental data set. Regionally the Sewickley
and Uniontown coal beds can be very important for methane control and ground control in
Pittsburgh coal bed longwall mines.

3.2 Impact of topography on fracture network behavior

Another parameter found to be very important in the magnitude and character of mining-
induced fracture permeability is surface topography. Sites showing more typical topographic
features for southwestern PA such as rolling hills and plateaus, show different permeability

532
Figure 2. Panel 1 mining under deep cover, Uniontown horizon. Mine-by shown by vertical, dotted red
line. Maximum permeability during monitoring reached about 5.4 mD, purple line shortly after the face
passed the monitoring site. The undisturbed rock permeabilities are shown by the initial readings. Max­
imum permeabilities are two orders lower at the deep cover site compared to the shallow cover site.

Figure 3. Panel 1 mining under shallow cover, Uniontown horizon. Mine-by shown by vertical, dotted
red line. Maximum permeability of about 246 mD, purple line. Permeabilities increased as the face
approached but reached a maximum almost two months after the face passed, possibly effected by the
stream valley location of the monitoring site.

behavior than sites near stream valleys. For more typical topography away from stream val­
leys, the first panels mine-by showed permeability dropping to about 10% to 20% of the max­
imum value achieved for about 610 m (2000 ft) after mine-by (Figure 4). For sites near stream
valleys, first panel mine-bys show the permeability remains in the range of 90% of the max­
imum value achieved for about 610 m (2000 ft) after mine-by (Figure 5). Note there is also
a significant difference in overburden depth between the two sites which may contribute to
these findings.

533
Figure 4. Panel 1, typical regional topography, Sewickley horizon. Mine-by of longwall face shown by ver­
tical, dotted, red line. Green, horizontal line shows 20% of maximum permeability value, which is close to the
post mining values during this time interval. Maximum permeability was achieved shortly after face passed.

3.3 Importance of the location of maximum horizontal movement of strata in gas transport
There is a high degree of importance regarding the location of maximum horizontal move­
ment of strata in gas transport and the amount of predicted inflow quantities associated with
a hypothetical well breach. The zone of maximum horizontal movement or unconventional
subsidence was assumed to be the most likely position for the hypothetical well breach. From
our ground control research partners, it is understood that zones of maximum unconventional
subsidence are positioned where large contrasts in bending strength occur in adjacent strata.
Multiple stratigraphic zones of these high contrasts in bending strength were identified at the
study sites by monitoring ground movement, and through ground movement simulations
using numerical modeling software FLAC 3D. This software can be used to solve complex
geotechnical problems for three-dimensional analyses of rock.
These factors contributing to overburden deformation are incorporated in the overburden
fracture characterizations completed by DFN modeling. The two simulation methods por­
tray the changes to vertical fracture in response to ground movement (3DEC) and to fluid
transport through fractured rock, or FFC. The combination of these DFN simulations pro­
duce a simulation of inflow to an active mine in the event of a hypothetical casing breach.
The summation of these results to date has shown that, for flow simulations under inter­
mediate cover (between 152 m, 500 ft and 274 m, 900 feet) and deep cover (>274 m, 900 ft)
where the zone of maximum movement was above the mining-induced fracture zone (more
than about 300 ft above the Pittsburgh seam), no hypothetical breached gas reached the
mine (Table 1).
When the horizon of maximum horizontal displacement occurs above the zone of mining-
induced fracturing (typically more than 91 m, 300 ft above the Pittsburgh seam in the region),
the likelihood of hypothetical breached gas reaching the mine is low. Research has shown the
dominant means of gas transport is through the induced fracture network formed by longwall
mining. When the induced fracture network is not in contact with the location of maximum
horizontal movement, the opportunity for appreciable gas migration from a hypothetical well
breach is limited. For simulations under intermediate cover (between 152 m, 500 ft) and deep
cover (>274 m, 900 feet) where the zone of maximum movement was above the mining-
induced fracture zone, no gas from a hypothetical well breach reached the mine (Table 2).

534
Table 1. A short summary of inflow simulations for intermediate depth (between 152 m, 500 ft and
274 m, 900 ft) at 20,700 kPa (3000 psi).
Sites Hypothetical breach location 1 borehole 10 boreholes

Tunnel Ridge 51.2 m (168 ft) below surface 0 m3/s (0 cfm) 0 m3/s (0 cfm)
NV35 120 m, 395 ft below surface <5-4 m3/s, <1 cfm Not determined

Figure 5. Sewickley horizon monitoring borehole at the, shallow, stream valley monitoring site.
Mine-by of longwall face shown by solid, red line. Dashed purple line shows 90% of maximum perme­
ability value, which is close to the post mining permeability values.

An additional consideration for these unconventional well pads is the number of wells
drilled from a site. For the locations adjacent to mining, the well pads can contain different
numbers of wells. As part of this study, hypothetical well breach inflows were assessed for dif­
ferent numbers of wells on a pad, both at flowing and shut-in pressures. Increasing from 1 to
10 boreholes typically increases inflow prediction quantity about 2 to 3 times the single well
rate at 20,700 kPa (3000 psi).

3.4 Summary of hypothetical inflow findings

A series of simulations of the transport of hypothetical breached gas were made based on the
inflow estimates for Pittsburgh coal bed mine ventilation systems. These simulations for mines
under shallow cover (<152 m, 500 ft) with a hypothetical breach at the Sewickley horizon
showed relatively high inflow rates at shut-in pressures of 20,700 kPa (3000 psi) (Table 2). If
the zone of maximum movement is very shallow—above the mining-induced fracture zone
more than 91 m (300 ft) above the Pittsburgh seam—the site would experience a much-
reduced likelihood of inflow or reduced inflow quantities.
Simulations of the transport from the hypothetical breached gas for mines under intermedi­
ate cover show no hypothetical inflow to the mine when maximum ground movement is above
the mining-induced fracture zone for the regions Pittsburgh coal bed mines under 20,700 kPa
(3000 psi) of pressure. If a hypothetical breach occurs under intermediate cover within the
mining-induced fracturing and at 20,700 kPa (3000 psi) shut-in pressures, some limited gas
inflow to the mine may occur (Table 2).

535
Table 2. Summary of hypothetical inflow findings for differing depths.
20,700 kPa 20,700 kPa
(3000 psi) (3000 psi)
Overburden depth 1 casing 10 casing Comments

Shallow cover sites (under 152 m, 500 ft, 0.347 m3/s, 0.672 m3/s, Maximum movement
depth of overburden cover at well site) (736 cfm)1 (1420 cfm)1 within fractured zone2
Intermediate cover sites (between 152m, 500 ft 0 m3/s, (0 0-0.00656 m3/s, Maximum movement
and 274 m, 900 ft, depth of overburden cover cfm) inflow (0–13.9 cfm) above fractured zone
at well site) inflow
Deep cover sites (over 274 m, 900 ft depth of 0–0.035 m3/ 0–0.0854 m3/s, Maximum movement
overburden cover at well site) s, (74 cfm) (181 cfm) above and within frac­
inflow inflow tured zone2
1
Inflow data for shallow cover site in stream valley at 2410 kPa (350 psi)
2
Inflow data with hypothetical breach above fractured zone expected to produce different behavior

Under deep cover, quantities of inflow from a hypothetical breach are anticipated to be low
under shut-in pressures—20,700 kPa (3000 psi). No inflow quantities were predicted from
a hypothetical breach under deep cover above the mining-induced fractured zone under shut-
in pressures.

3.5 Behavior of gas transport in underground mine ventilation systems and possible input for
monitoring or mitigation
Different inflow scenarios were simulated using the model gas migration methods shown in
Figure 1. Two fundamentally different ventilation schemes were used in these simulations and
are described in Dougherty et al., 2022. If estimated inflow quantities from a hypothetical well
breach do not exceed about 0.19 m3/s (400 cfm), mitigation by the mine’s standard ventilation
system may be possible. Ventilation systems using conventional Pittsburgh coal mine strat­
egies showed good potential for controlling this level of additional inflow.
In the Pittsburgh ventilation systems studied, gas from a hypothetical well breach of
0.19 m3/s (400 cfm) was primarily concentrated at the bleeder evaluation points (BEPs) and
bleeders but below statutory limits, with some limited amounts of the hypothetical breached
gas in the gob and no gas on the active longwall face (Table 3). The lower simulated inflow
quantity of 0.16 m3/s (340 cfm) showed identical distribution patterns regarding of where the
hypothetical breached gas accumulated, was absent.
A second, higher level of hypothetical breach inflow was considered. These simulations
showed 0.99 m3/s (2100 cfm) of gas entering the mine through roof fractures. Ventilation air­
flow simulations for a conventional Pittsburgh coal mine and a hypothetical breach of
0.99 m3/s (2100 cfm) showed methane concentrations at elevated levels, above or near statu­
tory limits in the BEPs and the bleeders. Hypothetical inflows of this magnitude or greater
will likely require mitigation to re-establish safe conditions underground.

3.6 Potential mitigation methods and monitoring guidance

As NIOSH research has progressed, the focus of the interested parties has changed in the event
of any potential for a hypothetical well breach, however unlikely towards monitoring and detec­
tion and away from mitigation or rehabilitation activities. Consequently, the simulations for
any gas from a hypothetical well breach can be used as characterizations of transport within the
underground ventilation system as a guide to detect this unplanned gas underground.
The locations for accumulations and higher concentrations of methane shown in Table 3
have been consistent across different inflow quantities, simulation methods and the two gen­
eral ventilation schemes. The ventilation simulations include CFD, the LIAM and VentSim
network modeling (Figure 1). The plan for monitoring gas underground could benefit from
this guidance on where hypothetical breached gas will accumulate. The simulations can show

536
Table 3. The Table shows results from the LIAM physical model for ventilation testing using tracer gas
mixture as a proxy for introduced gas from a hypothetical well breach through roof fractures. Results are
in close agreement with other simulation methods for transport in the ventilation network. Multiple tail­
gate, bleeder, and gob vent borehole/gab gas venthole locations are standardized test locations in entries
or the gob.
0.19 m3/s, 400 cfm hypo­
Monitoring location 0.16 m3/s, 340 cfm hypothetical inflow thetical inflow
Increased CH4 concentration due to hypothetical well breach, % CH4

Tailgate Bleeder Inby BEP 0.131 0.301

Evaluation Point
Tailgate Bleeder outby BEP 1.080 1.328
Evaluation Point
Tailgate #1 0.000 0.000
Tailgate #2 0.033 0.049
Tailgate #3 0.047 0.064
Longwall Face 0.000 0.000
Back Bleeder #5 0.000 0.000
Back Bleeder #6 0.016 0.019
GVB #1 0.002 0.004
GVB #2 0.000 0.000

how the gas moves through the mine and the magnitude of concentration change due to the
added gas from the hypothetical breach.

3.7 Methodology for distinguishing gas sources near well sites or within mines
A GC-based methodology was developed to discriminate coalbed gas, unconventional
(shale) gas, and underground storage field gas. The anticipated need to retrieve and ana­
lyze samples with a relatively short turnaround time makes GC methods attractive for
this application. Gas analysis for stable isotopes of carbon and hydrogen have been used
to discriminate gas sources (Coleman, 1991). These analyses can have limitations for rou­
tine, frequent sampling and analysis in this application where short turnaround times for
results and lower analytical costs for multiple samples are desirable. Bivariate plots using
the hydrocarbon index and carbon dioxide (CO2) concentration show coalbed gas plot­
ting above 2% CO2 with shale gas and storage field gas below this value and below 80
on the hydrocarbon index (Figure 6).
Plots of the hydrocarbon index and CH4/CO2 ratio show coalbed gas below 80 for the CH4
/CO2 ratio with shale gas and storage field gas over 110 for the CH4/CO2 ratio and below 80
on the hydrocarbon index. A strategy for implementing this technique uses small diameter
boreholes to collect samples that can be analyzed by GC to identify incursions of non-coal gas
(Schatzel et al., 2022).
The limitations of this methodology have not been completely established. In southwestern
PA, differences in both organic matter type and rank for coal and non-coal sources create
a good theoretical basis for discrimination of the produced gases. The accumulation of more
data is in progress and will better define performance parameters for the methodology.

4 CONCLUSIONS AND SUMMARY

Overburden depth is very impactful in the creation of mining-induced fracturing permeability.

For first panel mining, there is a significant increase in permeability in going from deep
(>274 m, 900 ft) to shallow cover overburden (<152 m, 500 feet), measuring two orders of
magnitude greater for the maximum value in the shallow case.

537
Figure 6. Graphical methods of discriminating gas sources using gas chromatography and graphical,
interpretive methods. Plots of the hydrocarbon index and CO2 concentrations (a) and CH4/CO2 ratio (b)
with potential gas sources occupying different portions of the plot.

Differences in topography at the monitoring sites were also very influential for mining-
induced fracture permeability. Where topography near the monitoring site included a stream
valley, the site retained much of the increase in permeability after the face passed, up to 90%
of the maximum value. In more typical western PA topography, away from stream channels,
induced permeability dropped to 10 to 20% of the maximum value in the time frame of moni­
toring presented.
The position of maximum horizontal movement has a great effect on the amount of hypo­
thetical breached gas that can reach the mine workings. Enhanced inflows were estimated at
the shallow cover, stream valley site, within the fractured zone. The intermediate and deep
cover sites showed essentially no inflows when the hypothetical breach location is above the
fracture zone.

538
 Gas transport in the ventilation system for a hypothetical 0.19 m3/s (400 cfm) gas breach
showed a good possibility for managing gas within statutory limits. If a hypothetical breach
of 0.99 m3/s (2100 cfm) were to occur, this quantity would be excessive and require mitigation
to comply with statutory limits for the mine.
A methodology for distinguishing gas sources near well sites or within mines was reviewed.
A GC-based method was developed to distinguish coalbed methane, shale gas, and storage
field gas sources. The proposed method successfully distinguished the gas sources in the data
set. The goal of the ventilation research is to allow for the intersection of these two industries
to co-exist with science-based recommendations forming a framework for worker safety.

REFERENCES

Ajayi, K. M. & Schatzel S. J., 2020. Transport model for shale gas well leakage through the surrounding
fractured zones of a longwall mine. International Journal of Mining Science and Technology, DOI:
10.1016/j.ijmst.2020.05.0, published online May 29, 2020.
Coleman, D. D., 1991. The Use of Geochemical Fingerprinting to Identify Migrated Gas. 67th Annual Tech­
nical Conference and Exhibition of the Society of Petroleum Engineers, Washington, DC, October 4–7,
1992.
Commonwealth of Pennsylvania, Department of Mines and Mineral Industries, Oil and Gas Division.
1957. Joint Coal and Gas Committee, Gas Well Pillar Study. Harrisburg, Pennsylvania, 28 pp.
Commonwealth of Pennsylvania, Department of Environmental Protection Oil and Gas Management,
2017. Guidelines for Chain Pillar Development and Longwall Mining Adjacent to Unconventional
Wells, Technical Guidance Document 800-0810-004, Harrisburg, Pennsylvania, 27 pp.
Dawson KJ, Istok JD, 1991. Aquifer testing: design and analysis of pumping and slug tests. Chelsea, MI:
Lewis Publishers, Inc.
Dougherty, H., Watkins, E., Kimutis, R., 2022. A network model analysis of an unconventional gas well
breach above an underground coal mine. MineXchange: 2022 SME Annual Conference and Expo, Feb­
ruary 27-March 2, 2022, Salt Lake City, Utah, preprint 22–008. Englewood, CO: Society for Mining,
Metallurgy & Exploration, 2022 Feb, 33–36.
Gangrade., V., Schatzel, S. J., Harteis, S. P., Addis, J. D. 2019. Investigating the Impact of Caving on
Longwall Mine Ventilation Using Scaled Physical Modeling. Springer, Mining, Metallurgy & Explor­
ation, 36(4),729–740.
Gangrade, V., Kimutis, R. A., Watkins, E. A., Schatzel, S. J., Addis, J. D., Hollerich, C. A. 2022. Simu­
lating the Impact of a Shale Gas Well Breach on Longwall Mine Ventilation Utilizing a Scaled Phys­
ical Model. Society of Mining, Metallurgy and Exploration Mine Expo 2022, Salt Lake City, UT,
February 27 to March 2, 2022.
Khademian, Z., Ajayi, K. M., Su, D. W. H., Schatzel, S. J., Kim, B. H., Esterhuizen, G. S. 2022. Rock­
mass permeability induced by longwall mining under deep cover: potential gas in-flow from a sheared
gas well. Society of Mining, Metallurgy and Exploration Mine Expo 2022, Salt Lake City, UT, Febru­
ary 27 to March 2, 2022.
John T. Boyd Co., 2016. The Pennsylvania Department of Environmental Protection’s Review of the Gas
Well Pillar Study Update. PO 4300311202 and 4300400813. September 6, 22 pp.
Mark, C., & Rumbaugh, G. M., 2020. Assessing risks from mining-induced ground movements near gas
wells. International Journal of Mining Science and Technology, 30. 11–16.
Ren, T., Balusu, R., Claassen, C., 2011. Computational Fluid Dynamics Modelling of Gas Flow Dynam­
ics in Large Longwall Goaf Areas. In: 35th APCOM Symposium, Application of Computers and Oper­
ations Research in the Minerals Industry, 603–613.
Schatzel, S. J. & Su, D. H. W., 2020. The origin of gas in coal and shale. In Pramod Thakur, Steven
J. Schatzel, Kashy Aminian, Gary Rodvelt, Morgan H. Mosser, Joseph S. D’Amico (eds) Coal Bed
Methane: Theory and Application, Amsterdam, The Netherlands, Elsevier, 25–34.
Schatzel, S. J., Ajayi, K., Watkins, E. A., Gangrade, V., 2022. Gas source discrimination methods to
identify the occurrence of a hypothetical, unconventional gas well breach into a nearby longwall mine.
Society of Mining, Metallurgy and Exploration Mine Expo 2022, Salt Lake City, UT, February 27 to
March 2, 2022.
Su, W. H., Zhang, P., Dougherty, H., Van Dyke, M. A., 2020. Shale gas production under the influence
of underground longwall coal production. In: Proceedings of the IV International Conference on Fossil
& Renewable Energy, Houston, TX, Feb. 17–19.

539
Watkins, E. A., Karacan, C. O., Gangrade, V., Schatzel, S. J., 2021. Assessing Gas Leakage Potential
into Coal Mines from Shale Gas Well Failures: Inference from Field Determination of Strata Perme­
ability Responses to Longwall-induced Deformations. Natural Resources Research. 10.1007/s11053-
021-09859-9
Watkins, E. A., Gangrade, V., 2022. Optimization of Auxiliary Fan Placement for Large-opening Under­
ground Stone Mines. Society of Mining, Metallurgy and Exploration Mine Expo 2022, Salt Lake City,
UT, February 27 to March 2, 2022.
Zhang, P., & Su, W. H., 2021. FLAC3D Modeling Evaluation of the Comprehensive NV-35 Gas Well
Pillar Study. In: Proceedings of the 55th US Rock Mechanics/Geomechanics Symposium, June 18–25,
ARMA-21-A–1193

540
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Optimization for fire evacuation applying maximum flow cost

algorithm to improve the time-response in underground coal mines

S. Lotero, H. Khaniani, V. Androulakis, M. Hassanalian, S. Shao & P. Roghanchi

New Mexico Tech, Socorro, NM, USA

ABSTRACT: During an underground mine fire, the presence of smoke and toxic gasses,
low visibility, and changes in the ventilation system will make it extremely difficult to identify
evacuation measures and the optimum path to safety. This paper presents an algorithm for
solving the maximum cost flow network problem (MCFNP) for fire evacuations considering
the distribution of toxic gasses inside the mine in real-time. The proposed algorithm will iden­
tify the optimum evacuation paths requiring minimum decision-making time. The algorithm
holds data in nodes and arcs accumulating values at each iteration and updating the network
depending on the mine conditions. To achieve this, a fire simulation was performed in
VentSimTM software to extract the air velocity, gas concentrations, visibility, and changes in
the ventilation system data throughout the incident zone. The presented algorithm finds the
evacuation paths improving time-response and comparing the decision-making of miners in
a successful evacuation. The prediction of the concentration of toxic gasses, recommends
a safety path and avoids exposure to the danger zone despite the presence of the shortest-path
and road capacity toward the surface.

1 INTRODUCTION

Underground mine workers may encounter situations such as explosions, fires, floods, roof
falls, and other emergencies that put their lives at risk (Ronald S. Conti 2005; Mahdevari,
Shahriar & Esfahanipour 2014). One of the most dangerous dynamic situations in under­
ground coal mines is fire. It is the second most frequent cause of death throughout history
with more than 734 fatalities from 1900-2006. Despite all the technological advancements, this
number is continually increasing due to fire accidents underground (Centers for Disease Con­
trol and Prevention 2009; An et al. 2021). Fire incidents also generate large economic losses in
the mining industry (Saleh & Cummings 2011). Some critical impacts of fire in the under­
ground environment are damage to the mine structure, potential explosion, generation of
smoke and toxic gases, visibility, and production loss (Hansen 2010). The most common cause
of death is monoxide carbon (CO) poisoning, (U.S Department of Labor 2007; Yuan, Zhou &
Smith 2016). Table 1 shows the quantity of CO, the allowable exposure time by the law, and
the symptoms produced by the gas(Centers for Disease Control (CDC) 1988; U.S Department
of Labor 2007).
Effective escape from the contaminated environment will significantly reduce the risk of
fatalities and injuries during a fire event (Adjiski et al. 2015). Mine evacuation refers to
a series of protocols and procedures that allows a mine to respond, establish control, and
escape as soon as possible from the danger (Kingery 1960). During a fire event, the mine ven­
tilation system plays an important role in delivering fresh air to the miners. However, the
mine ventilation system also transports contaminants to different parts of the mine causing
contaminations in workplaces (Kingery 1960; Conti 1993). The mine ventilation network and

DOI: 10.1201/9781003429241-55

541
Table 1. Health effects from exposure to CO levels in Underground mines(U.S Department of Labor
2007).
CO levels Time exposure
Exposure level [ppm] [h] Symptoms

Minimum 1 - 70 8 Food poisoning and flu symptoms

Medium 150 - 300 5 Dizziness, vomiting, tinnitus, breathing difficulties,
Impaired judgment, confusion, drowsiness, memory,
and walking problems.
Extreme 400 > 3 to 4 Convulsion, coma, collapse, unconsciousness, brain
damage, death.

spatial distribution of the mine depend on the method of coal extraction, room and pillar is
a traditional method for extracting coal (Kingery 1960).
This method consists of mining the rock forming “rooms” and leaving behind “pillars” as
a support of the overburdened rock weight. These procedures form “grids” as a tunnel net­
work pattern inside the mine (Khayreddine Rahab 2011). The ventilation network analysis is
concerned with the interactive behavior of airflows within connected branches of an integrated
tunnel network. Knowing how those branches are interconnected then we can predict, quanti­
tatively, the distribution of airflow for given locations and duties of fans(Mcpherson 1984).
During a fire event, the carried away smoke and toxic gasses through the tunnels force the
workers to evacuate the mine or find a safe place to refuge during the incident(Onifade 2021).
Miner’s decision-making capabilities determine the success of survival and control of the evacu­
ation flow (Ray & Singh 2007). However, identifying the highest concentration of the toxic gasses
is of vital importance to optimize the escape path reducing miner exposure to contaminations.
Ford-Fulkerson is a maximum flow cost optimization algorithm to find the “cheapest” possible
way of sending a certain amount of flow through a network [17]– [19]. Maximum Cost Flow
Problems are composed of a series of nodes (points) and arcs (arrows), with a beginning and an
end corresponding to the uses of graph theory (Harary & Norman 2013; Lavrov 2020).

2 LITERATURE REVIEW

Readers may refer to Ford-Fulkerson (1953) for an interpretation of evacuation as a dynamic

problem to withdraw a group of people from a danger zone to safety place (Neumann 1984;
Chavez 1993). Cui, An, and Chao (2014) proposed a novel generalized maximum cost flow
model to optimize the distribution pattern of two types of flow introduced by the conflict
between them for vehicle accidents (Wang et al. 2016).
According to MSHA (2018), fresh air must be distributed to the production areas where
mine personnel are located, and the incoming air might need to be heated or cooled (Mine
Safety and Health Administration 2007). According to the U.S department of labor (2006)
“workers may not be exposed to CO over 50 ppm averaged over 8-hour workday; the excur­
sion (or short-term limit) for CO in coal mines is 75 ppm for 15 minutes without proper
approved respiratory protection” (U.S Department of Labor 2007).
Brake (2018) explained that underground fire modeling software has had limited availability
and has been technically challenging and time-consuming to use especially for routine mine
investigations (Brake 2013). VentSimTM software is used to design the ventilation system of
an underground mine. VentFIRETM, a module of VentSimTM, permits to recreate of a fire
event by providing CO concentration data(Howden 2020). Multivariable models described
how changes in the velocity of fans affect both the airflow and the pressure over fans (Yang,
Yao & Wang 2022). Further, developing a fire simulation about the dynamic flow impulsed
by the ventilation system allows the identification of the movement of carbon concentrations
(Adjiski et al. 2015; Wang et al. 2016).

542
3 METHODOLOGY

The simulation of the multivariable model has been executed using VentFIRETM to execute fire
simulation. We used the software to collect the amount of carbon monoxide (CO) produced by
the fire. The data extracted was used as input in a code developed in Python (Yuan et al. 2016).
For this paper, only the concentration of carbon monoxide produced by the fire inside the mine
was considered to identify the critical points where miners should not take during the evacuation,
and how this prediction can contribute to the decision-making. To execute the simulation, the fol­
lowing assumptions were carried out to develop a realistic fire environment for accurate results:
A. Fire event: Mobile diesel equipment fire accident; mine truck MT2010 (Epiroc 2022).
B. The burning substance is diesel combined with coal; it is assumed that the mine truck was
loading around 50% of its capacity (10000 kg), and the tank was full of diesel (380 liters).
C. The fire occurred due to a vehicle rollover; the diesel spilled 50% of the tank on the surface
(An et al. 2021). The fire cannot be extinguished via an onboard suppression system or
portable extinguishers.
D. Ten miners are 1 km from the surface. The speed at which they walk is 1.1 m/s (Howden
2020).
E. Wireless Sensor Network Type: Carbon monoxide sensors. The sensors were distributed
every 100 m starting from the workplace (Conti 1993). In total, 15 sensors were put in the
VentFIRETM simulation. Refer to Figure 3.
F. The duration of the fire is 1 hour with a growth period of 30 min, the delay fire start is 5
minutes, and the growth and delay periods are 30 minutes each.
The ventilation system corresponds to an underground coal mine shown in Figure 2. The
direction of the gasses flowing south to north is for the intake air (dark blue). Thus, fresh air
is delivered through the central airways, and the contaminated air is exhausted along the side
of the coal panel (red). The conveyor is in a center airway with a brattice curtain regulating
the airflow, where the fire accident occurs (brown).

Figure 1. Underground coal mine ventsimTM model: Airway type of the mine ventilation.

Part of the mine is extracted from the whole model to have a clearer visualization of the
simulation (Figure 2). The data collected is only from the zone that covers the workplace, the
fire accident, and the exit to the surface, where the 15 sensors are located. However, the entire
ventilation system is working as a whole mine model to conserve logical results.
The ventilation system is crucial to understand how the gasses are being transported
through the fire model and predicting the hazards once the miners are notified about the fire
event. The model and the evacuation are dependent parameters. However, the velocity of the
airflow is relevant to see how fast the gas is being transported and it’s considered for the ana­
lysis of the distribution of the gas.

543
Figure 2. The ventilation model was extracted to perform the fire simulation.

In this section, the multiple emergency flow route (MEFR) model is carried out. The first
flow path is the distribution of the parameters created by the fire (source 1) and how the venti­
lation delivered them to the miners’ location (sink 1). The second flow path is the evacuation
of the mine from the workplace (source 2) to the surface (sink 2). The objective of MEFR,
using a Ford-Fulkerson algorithm, is to minimize the time evacuation and the risk generated
by the CO. Knowing which the maximum-cost flow is delivering through the mine, we deter­
mined where is the highest concentrations of CO. To develop the mathematical model,
a generic maximum-flow algorithm is modified, to present the solution to the evacuation
problem. Given a directed graph with quantities, G = ((N, A)) be a directed graph with node
set N and edges set A, the graphic G = ((N, A)) representing the maximum capacities of the
tunnels. The capacity c(i,j) is a positive real value for each arc (i,j) ϵ A, this capacity repre­
sents the quantity of the hazardous parameters. delivered into the mine or the distribution of
the fresh air. The new graph G = (N, Af) for a pre-flow fp content the new edge with the
residual values Af. The algorithm is presented in (Ford 1955; Ahuja 1987; Goldberg & Tarjan
1988; Kyi & Naing 2018; Lavrov 2020).

Figure 3. A General max-flow cost algorithm for evacuation problems (Goldberg & Tarjan 1988).

The maximum permissible limit for CO concentration in coal mines is 75 ppm for 15 min­
utes without proper approved respiratory protection (U.S Department of Labor 2007). In this
study, the worst-case scenario where the miners do not have proper respiratory protection is

544
considered. Figure 4 shows the CO concentration map of the mine after the end of the fire
simulation, while Table 2 contains the data collected by the CO concentration sensors during
the simulation. The least affected part of the mine is the ‘intake airway’ part that lies north of
the fire (see sensors no. 1 and 3), as the direction of the intake air forces the fire fumes towards
the interior of the mine. At the same time, the CO concentration at the ‘return airway’ part of
the mine that lies north of the working area exceeds the permissible limit (of 75 ppm) only at
the last 10 minutes of the simulation (see sensors no. 2, 4, 8, 9, and 11). This means that the
CO concentration would allow the miners to safely evacuate moving along the return air path
towards the main exit up to 1 hr. after the fire occurs without protection.
Naturally, the area where the CO accumulates mostly and the fastest is at the site of the fire
and south of the fire: at the site of the fire the CO concentration increases from 0 to 216 ppm
within the first 10 minutes from the beginning of the fire (see sensor no. 5), while the part
exactly south of the fire exceed the safety limit within the next 5 minutes (15 minutes from the
fire – see sensors no. 6 and 7). As time passes, the CO fumes move further inside the mine and
the CO accumulation at the west part of the active coal panel (refuge chamber location) has
exceeded the safety limit between 25 and 30 minutes from the starting of the fire (see sensors
no. 10, 12, 13, and 15). This means that the workers would have approximately 20 minutes
available to reach the vicinity of the refuge chamber and approximately 35 minutes to safely
enter the chamber (without protective equipment).

Table 2. Carbon monoxide concentration results in ppm as read by the 15 sensors during 1 hr of
simulation.
Time [s] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
300 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
600 0 0 0 0 216 57 19 0 0 0 0 0 0 0 0
900 0 0 0 0 871 287 170 0 0 4 0 0 42 0 0
1200 0 0 0 0 1960 680 476 0 0 67 0 4 228 0 43
1500 0 0 0 0 3482 1246 948 0 0 237 0 65 570 0 217
1800 0 1 0 1 5435 1955 1564 0 0 545 0 235 1071 0 541
2100 0 2 0 2 7830 2789 2307 0 0 983 1 537 1711 2 1024
2400 0 5 0 6 7845 3009 2763 4 7 1543 17 973 2483 24 1645
2700 0 22 0 26 7852 3010 2788 33 45 2089 77 1540 2723 97 2383
3000 0 74 0 85 7881 3010 2790 117 144 2238 213 2076 2737 251 2637
3300 0 181 0 202 7875 3011 2791 282 329 2266 444 2258 2740 506 2684
3600 0 357 0 392 7884 3012 2793 547 617 2273 786 2305 2741 873 2699

Figure 4. VentFireTM model for the CO concentrations in the mine.

545
 Table 3 shows the CO concentrations from the fire simulation and are compared with the
permissible exposure limit time according to MSHA(U.S Department of Labor 2007)

Table 3. CO concentrations [ppm] with and without protection.

Color
Gas CO [ppm] 0 50 100 200 300 400 400>

With protection 28800 18000 10800 900

Time exposure [sec] Without protection 900 300 100

The data collected from VentSimTM is the flow used for each arch of the model as the cap­
acity to deliver the gas. The miners’ decision-making depends on the points with the highest
concentrations of CO. As the simulation was run, the fire incident was reported after 10 min­
utes (i.e., the capacity and flow data extracted from the model were taken from minute 10
onwards). After obtaining the data, we used the code to find the safest path considering the
highest concentrations of CO in the mine. In addition, based on the distance between each
airway that the miners must travel, the algorithm selects the longest routes and discards them
until the shortest one is obtained, optimizing the desired escape route.
Figure 5 shows the capacity and flow that is delivered in each tunnel A = arc (i,jn).
The sequence of graphs also shows the update of the data according to the algorithm
proposed. In addition, the intake shaft has a direction from west to east and from north
to south. To make the graph simpler and the values more visible, every two real mine
junctions V = nodes (i,jm) represents one in the proposed scheme, and the values calcu­
lated between each arc will be interpolated so that the results do not change (Goldberg
& Tarjan 1988; Lavrov 2020).

Figure 5. Distribution of CO based on the Ford-Fulkerson algorithm after 20 minutes of the fire event.

4 RESULTS

The number of iterations generated by the algorithm depends on the possible number of
routes that can reach a point or sector. However, the algorithm selects all the possible paths
and accumulates the flow, and calculates the maximum flow from the destination sink.
Table 4 shows the results of the first flow path due to CO concentrations in the mine section
10 minutes after the fire alarm. It shows the dangerous points through which miners cannot

546
cross. Each color (green, blue, and yellow arrows) represents the arbitrary routes selected by
the algorithm in Figure 5 and represents the column 1,4, and 7 in Table 4, respectively. The
source 1 is the fire and the sink 1 is the workplace.

Table 4. Distribution of the CO concentrations results using the algorithm Ford- Fulkerson.
The first iteration of the flow path 1.
CO Flow CO Flow CO Flow rc Max-
1st Route [ppm] 1 2nd Route [ppm] 2 3rd Route [ppm] 3 (i,j) flow

Fire → 2 1960 37 Fire → 1 1760 31 (Fire) → 1 1729 36 29 133

2→3 1456 37 1→4 810 31 1→3 1200 36 29 133
3→9 562 37 4→5 730 31 3→5 644 36 29 133
9→ 8 476 37 5→8 587 31 5→8 520 36 29 133
8 → 11 459 37 8 → 11 459 31 8 → 11 355 36 29 133
11 → 16 222 37 11 → 12 238 31 11 → 12 202 36 29 133
16 → 21 66 37 12 → 17 212 31 12 → 13 192 36 29 133
21 → Sink 1 37 37 17 → Sink 1 31 31 13 → 18 184 36 29 133
18 → 23 9 36 29 65
23 → Sink 1 0 36 29 65

Figure 6 shows the safest and shortest Evacuation Route (SSER) based on the information
analyzed previously. The values in green are the results of the flow after the code has per­
formed in Python. The results are scaled by 1000 to understand the maximum values.

Figure 6. Route of escape performing the Ford-Fulkerson algorithm.

Table 5 shows the results of the safest path as a function of the distance from each branch
and the walking speed at which the miners evacuate. The time taken to exit depends on the
constant speed at which miners walk and the distance from each arch. The code selects the
minimum value of time it takes for the miners to travel from one node to the other until they
reach the exit.

547
 Table 5. SSER using Ford-Fulkerson algorithm form the workplace to surface.
Accumulative time [s]
Paths Flow Time [s] Bifurcation 1 Bifurcation 2
Workplace ->1 12.2 44.6 45 45
1→2 14.7 56.4 101 101
1→3 30.39 45.5 90 90
2→4 14.7 45.5 146 146
3→4 30.39 46.3 136 136
3→5 0 45.5 136 136
4→6 9.39 28.9 165 175
4→7 35.7 57.8 194 204
5→6 0 45.5 181 181
6→9 9.38 89.1 254 270
7→9 35.7 47.2 241 241
7→8 0 46.8 241 241
8 → 10 0 30.8 272 272
9 → 10 0 46.7 288 301
9 → 11 22.5 106.3 348 361
10 → 12 0 107.6 379 396
11 → 12 7.38 45.6 393 406
11 → 13 14.9 45.9 394 407
12 → 14 7.38 46.0 425 439
13 → 14 14.9 45.9 440 440
13 → 15 0 45.3 439 452
14 → 16 7.38 29.3 455 469
15 → 16 0 45.8 485 485
16 → surface 7.38 133.4 588 618
SSER Workplace→1→3→4→6→9→11→12→14→16→Surface

5 CONCLUSIONS

This study used Ford-Fulkerson algorithm to identify the safe path to safety considering the
concentrations of CO during a fire event in an underground coal mine. We used VentSimTM
software to capture various ventilation data, including airflow and airflow direction, visibility,
oxygen level, and CO concentrations. The algorithm helps miners to make a more assertive
decision and mitigate the risks involved in the prolonged inhalation of carbon monoxide.
Improving the code can positively impact the mining industry by decreasing fatality rates and
the costs of lost miners. Similar studies can be conducted using the proposed algorithm to
identify the path to safety in various scenarios, such as roof fall, explosion, and other mine
disasters. This will allow the mine operators to utilize a data-driven approach for mine evacu­
ation planning. The incorporation of the code in an interface that miners can control will help
rescue team to establish better communication with the miners involved in the fire emergency.
It should be noted that several parameters play an important role to choose the optimum
evacuation route.

ACKNOWLEDGMENTS

The project is funded by the National Institute for Occupational Safety and Health
(U60OH012351).

REFERENCES

Adjiski, V., Mirakovski, D., Despodov, Z. & Mijalkovski, S., 2015, ‘Simulation and optimization of
evacuation routes in case of fire in underground mines’, Journal of Sustainable Mining, 14(3), 133–143.

548
An, W., Li, Q., Liang, K., Yin, X. & Wang, Z., 2021, ‘Study on Combustion Characteristics and Evacu­
ation during Intake Airway Fire in Coal Face under Different Ventilation Conditions’, Combustion
Science and Technology, 193(8), 1378–1399.
Brake, R., 2013, Fire modelling in underground mines using Ventsim Visual VentFIRE software, THE
AUSTRALIAN MINE VENTILATION CONFERENCE, Adelaide.
Centers for Disease Control and Prevention, 2009, ‘Underground Coal Mining Disasters and Fatalities—
United States, 1900–2006ʹ, MMWR, 57, 1–7.
Centers for Disease Control (CDC), 1988, ‘NIOSH recommendations for occupational safety and health
standards 1988.’, MMWR supplements, 37(7), 1–29.
Chavez, J.D. and H.L.H., 1993, ‘Duality Theorems for continuos analog of Ford-Fukerson Flows in
Networks’, Advances in applied mathematics, 14, 369–388.
Conti, R.S.L.C.D., 1993, Fire sensors and the detection of underground mine fires, Longwall USA confer­
ence, Pittsburgh, PA.
Epiroc, 2022, Minetruck MT2010 Underground articulated truck with 20-tonne load capacity.
Ford, L.R. and F.D.R., 1955, ‘Maximal Flow through a network’, 1st edn.
Goldberg, A. V & Tarjan, R.E., 1988, ‘A New Approach to the Maximum-Flow Problem’, Journal of the
Association for Computing Machinery, 35(4), 921–940.
Hansen, R., 2010, Smoke spread calculations for fires in underground mines.
Harary, F. & Norman, R.Z., 2013, Graph Theory as a Mathematical Model in Social Science *.
Howden, B., 2020, VentSim DESIGNTM User Guide VentSim DesignTM Software, vol. Version 5.4.
Khayreddine Rahab, 2011, SURFACE AND UNDERGROUNDMINING TECHNOLOGY, Wrocław.
Kingery, D.S., 1960, Introduction to mine ventilating principles and practices., Bulleting 589, vol. 1,
Bureau of mines, Washington.
Kyi, M.T. & Naing, L.L., 2018, ‘Application of Ford-Fulkerson Algorithm to Maximum Flow in Water Dis­
tribution Pipeline Network’, International Journal of Scientific and Research Publications (IJSRP), 8(12).
Lavrov, M., 2020, When augmenting paths fail Proving the residual graph theorem Max-flow algorithms.
Mahdevari, S., Shahriar, K. & Esfahanipour, A., 2014, ‘Human health and safety risks management in
underground coal mines using fuzzy TOPSIS’, Science of the Total Environment, 488–489(1), 85–99.
Mcpherson, M.J., 1984, ‘Mine ventilation planning in the 1980s’, International Journal of Mining Engin­
eering, 2(3), 185–227.
Mine Safety and Health Administration, 2007, ‘Accident, illness and injury, and employment self-
extracting files (part 50 data): 2006ʹ, US Department of Labor, Mine Safety and Health Administration.
Neumann, M.M., 1984, A Ford-Fulkerson type theorem concerning vector-valued flows in infinite networks,
vol. 34.
Onifade, M., 2021, Towards an emergency preparedness for self-rescue from underground coal mines,
Process Safety and Environmental Protection, 149, 946–957.
Ray, S.K. & Singh, R.P., 2007, ‘Recent developments and practices to control fire in underground coal
mines’, Fire Technology, 43(4), 285–300.
Ahuja, R.K.& O.J.B., 1987, ‘A fast and simple algorithm for the maximum flow problem’, Massachusetts
Institute of technology, 87, 10–48.
Ronald S. Conti, L.L.C.W.J.W. and C.P.L., 2005, ‘Fire Response Preparedness for Underground
Mines’, CDC Workpalce Safety and Health, DEPARTMENT OF HEALTH AND HUMAN SER­
VICES, Pittsburgh, PA.
Saleh, J.H. & Cummings, A.M., 2011, Safety in the mining industry and the unfinished legacy of mining acci­
dents: Safety levers and defense-in-depth for addressing mining hazards, Safety Science, 49(6), 764–777.
U.S Department of Labor, 2007, Carbon Monoxide and Mining Health Hazard Information Card HH-28.
Wang, K., Jiang, S., Ma, X., Wu, Z., Shao, H., Zhang, W. & Cui, C., 2016, ‘Numerical simulation and
application study on a remote emergency rescue system during a belt fire in coal mines’, Natural Haz­
ards, 84(2), 1463–1485.
Yang, B., Yao, H. & Wang, F., 2022, A Review of Ventilation and Environmental Control of Underground
Spaces, Energies, 15(2).
Yuan, L., Zhou, L. & Smith, A.C., 2016, ‘Modeling carbon monoxide spread in underground mine fires’,
Applied Thermal Engineering, 100, 1319–1326.

549
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Heat mitigation for underground coal mine refuge alternatives

D.S. Yantek & L. Yan

NIOSH Pittsburgh Mining Research Division

ABSTRACT: Federal regulations require mines to place refuge alternatives (RAs) in under­
ground coal mines in the United States. Because miners trapped in an RA could experience
heat exhaustion or heat stroke, RA regulations mandate an apparent temperature (AT) limit
of 35°C (95°F) in RAs. The National Institute for Occupational Safety and Health (NIOSH)
conducted in-mine AT tests to evaluate three RA heat mitigation devices: a borehole air
supply (BAS), a battery-powered air conditioner (BPAC), and a cryogenic air supply (CAS).
The BAS limited the AT in a 60-person built-in-place (BIP) RA to 25°C (77°F) when the out­
side air temperature was 10°C–23.9°C (50°F–75°F) with relative humidity from 50%–95%.
During tests on a portable tent-type RA with a 29.4°C (85°F) mine air temperature, the
BPAC and CAS kept the AT below 35°C (95°F). AT tests showed the CAS allowed for 12
miners in a 29.4°C (85°F) 30-person BIP RA. RA manufacturers and mines can use this infor­
mation to ensure RAs can meet the AT limit.

1 INTRODUCTION

Heat buildup inside a refuge alternative (RA) is a serious concern. Heat sources for an RA
include the RA occupants and the RA’s carbon dioxide (CO2) scrubber system. RAs have
a limited ability to dissipate heat due to the low thermal conductivity of mine strata materials.
If conditions become severe enough, RA occupants could experience heat exhaustion or heat
stroke (Bernard, 2011, OSHA, 2003). To guard against severe heat stress, federal RA regula­
tions for underground coal mines specify a maximum apparent temperature (AT) of 35°C (95°
F). The AT is a heat index that can be used to determine how severe a thermal environment
appears (Steadman, 1979).
To ensure that the AT limit is not reached inside an RA, heat mitigation devices or occu­
pancy derating could be necessary for warm or hot mines. The National Institute for Occupa­
tional Safety and Health (NIOSH) examined the cooling capabilities of an air-conditioned
borehole air supply (BAS), a battery-powered air conditioner (BPAC), and a cryogenic air
supply (CAS) through RA AT testing in its Experimental Mine in Bruceton, PA. This paper
discusses the results of heat mitigation device AT testing.

2 HEAT MITIGATION DEVICE DESCRIPTION

2.1 BAS – Borehole Air Supply

The BAS (see Figure 1) consists of a rotary blower powered by a 460-V electric motor with
a power rating of 18.6 kW (25 hp). The BAS incorporates a 45-kW (154,000-BTU/hr) heater
and a 39.3-kW (134,000-BTU/hr) air conditioner. The BAS can provide a maximum airflow
of 2039 m3/hr (1200 CFM) at a pressure of 20.7 kPa (3 psi). At the minimum required fresh
airflow rate of 21.2 m3/hr (12.5 CFM) per person (CFR, 2018, CFR, 2019, CFR, 2022), the

DOI: 10.1201/9781003429241-56

550
Figure 1. Pictures of a BAS for RAs.

BAS can supply fresh air for 96 miners. To use the BAS, a flexible duct is connected from its
outlet to the top of a borehole.

2.2 BPAC – Battery-powered Air Conditioner

The BPAC (see Figure 2) was developed as part of Broad Agency Announcement (BAA) Con­
tract 200-2016-91901 with HDT Global (DRS, 2016). The BPAC was designed to keep the
thermal environment of a 30-person portable RA under the AT limit in a 29.4°C (85°F) mine.
The BPAC consists of a vapor compression refrigeration system attached to a movable,
rugged frame. To match the required cooling load, the BPAC uses a variable-speed compres­
sor with two variable-speed condenser fans. The BPAC is powered by a bank of 48-V lead
acid batteries contained within a separate battery tray. The BPAC and its battery tray would
be stored near an RA until needed post-disaster.

Figure 2. Picture of a BPAC connected to a portable tent-type RA using flexible duct.

2.3 CAS – Cryogenic Air Supply

The CAS (see Figure 3) was developed as part of BAA contract 200-2016-91194 (Cryo Life Sup­
port Systems, 2016). The CAS was also designed to keep a 30-person RA under the AT limit.
The CAS maintains air as a liquid at cryogenic temperatures until gaseous air would be needed
in the event of a mine emergency. The primary components of the CAS are a cryocooler and
a 2000-L horizontal dewar. The CAS uses a water-cooled, 480-V cryocooler with a power con­
sumption of 7.5–7.8 kW (Yan et al., 2017, 2020, 2021). Liquid air is created by filling the dewar
with the necessary quantities of liquid nitrogen and liquid oxygen to generate liquid air with the
desired oxygen concentration. A pressure switch that senses the dewar pressure cycles the cryo­
cooler on and off as needed to maintain the air inside as a liquid.

551
 In the event of a mine emergency requiring breathable air from the CAS, the following pro­
cedure would be followed. First, power to the CAS would be turned off. Next a vacuum-
jacketed hose would be connected from the CAS to a heat exchanger positioned within an
RA. A pressure buildup valve on the CAS would be activated to allow vapor pressure from
the dewar to force liquid air to flow to the heat exchanger. As the liquid air passes through the
heat exchanger, it would absorb heat from the surroundings and become gaseous air at
roughly 5°C–15°C (41°F–59°F), thus providing cooling to the RA.

Figure 3. Picture of a cryogenic air supply for refuge alternatives.

3 APPARENT TEMPERATURE (AT) TESTS

NIOSH researchers developed an AT test method to evaluate the thermal environment of RAs
(Yantek, 2014, 2019). This test method uses devices called simulated miners (SMs) to provide
the representative metabolic heat from actual miners and heated water tanks to provide the rep­
resentative heat input from an RA’s CO2 scrubber system. The SMs and heated water tanks use
electrical resistance heaters as heat sources. The SMs also served to provide moisture input to
the RAs via vapor or simulated sweat using water from the heated water tanks. The heat input
for the SMs was set at 117 W per person, and the heat input of the heated water tanks was set
to 27.5 W per SM (Yantek, 2014, 2019). During the AT tests, the total heat input was controlled
using programmable variable autotransformers (PVAs). The heat input from the heated water
tanks was controlled via manual variable autotransformers (MVAs).
Each of the heat mitigation devices were tested using either a built-in-place (BIP) RA, port­
able tent-type RA, or both. The BAS was tested on a 60-person BIP RA constructed inside
the NIOSH Experimental Mine in Bruceton, PA. The BPAC was tested on a 20-person port­
able tent-type RA that was installed in a crosscut in the Experimental Mine. The CAS was
tested on the 20-person portable tent-type RA and a 30-person BIP RA.

3.1 Setup for BAS testing in 60-person BIP RA

The 60-person BIP RA was built by constructing two stoppings using two layers of solid
20.3-cm-thick (8-in-thick) concrete blocks at either end of an entry. The interior dimensions of
the BIP RA are 13.7-m (45-ft) long by approximately 6.1-m (20-ft) wide and roughly
2.0-m (6.5-ft) high. The BIP RA floor is a nominally 20.3-cm-thick (8-in-thick) layer of con­
crete, and the roof and ribs are covered with a roughly 25.4-mm-thick (1-in-thick) layer of
shotcrete. The BIP RA is approximately 30.5 m (100 ft) below the surface.
The BAS delivered air to the BIP RA via a borehole located near the left rear corner (when
viewed from the entry door). A 2.4-m-long (8-ft-long), 20.3-cm-diameter (8-in-diameter) PVC
pipe was connected to the borehole to provide a long straight section to allow for accurate
measurement of the delivered airflow. The outlet of the PVC pipe delivered the air near the
right rear corner (when viewed from the entry door). The BAS delivered air at a nominal

552
volumetric flow rate of 1317 m3/hr (775 SCFM), which is slightly higher than the required
21.2 m3/hr (12.5 SCFM) of fresh air per person for 60 people. The BAS air conditioner was
set to deliver air at a dew point of 12.8°C (55°F).
The air temperatures and relative humidity (RH) levels inside the RA were measured using
Vaisala HMP155 (Vaisala, 2022b) and HMP110 Humidity and Temperature Probes (Vaisala,
2022a). For ease of reference, the RA was split into three equal-sized hypothetical sections:
Section 1, Section 2, and Section 3. One Vaisala HMP155 was placed in the center of Section 2
at mid-height, one Vaisala HMP110 was placed at the center of Section 1 at mid-height, and
one Vaisala HMP110 was placed at the center of Section 3 at mid-height. Averaging resistance
temperature detectors (RTDs) with a length of 1.2 m (4 ft) were used to monitor the air tem­
perature at additional locations.
The nominal total heat input was set to 8670 W to provide the representative metabolic
heat input of 60 occupants. The SMs were evenly distributed using 6 SMs across the width of
the BIP RA with 10 SMs along the length of the BIP RA. Because borehole air supplies pro­
vide enough air to dilute miner-generated CO2, heated water tanks were not needed. The
water tanks were used only to provide moisture via the SMs.

3.2 Setup for BPAC and CAS testing in 20-person portable tent-type RA
The 20-person portable tent-type RA was deployed in an entry in the Experimental Mine. Two
insulated walls were built to create an isolated test area. One wall was about 3.0 m (10 ft) from
the metal box of the RA, and the other wall was about 3.0 (10 ft) from the tent end of the RA.
The insulated-wall frames were built using 2x4 wooden studs spaced 40.6 cm (16 in) from center
to center with a layer of 50.8-cm-thick (2-in-thick) polystyrene foam attached to each side of the
wall frame. The cavities of each wall were filled with fiberglass insulation batts.
The air temperatures and RH levels inside the RA were measured using Vaisala HMP155
(Vaisala, 2022b) and HMP110 Humidity and Temperature Probes (Vaisala, 2022a). For ease
of reference, the RA was split into three equal-sized hypothetical sections: Section 1, Section 2,
and Section 3. One Vaisala HMP155 was placed in the center of Section 2 at mid-height, one
Vaisala HMP110 was placed at the center of Section 1 at mid-height, and one Vaisala
HMP110 was placed at the center of Section 3 at mid-height.
The mine air temperatures around the RA were measured using Lascar Electronics EL-
WiFi-T+ High Accuracy WiFi Temperature Data Loggers (Lascar Electronics, 2022). Four
data loggers were positioned over Sections 1, 2, and 3 of the tent and over the RA’s metal box
midway between the RA and the mine roof. Four data loggers were positioned around the
sides of the RA, and one data logger was positioned at either end of the RA. These six data
loggers were positioned at mid-height and 30.5 cm (12 in) from the RA.
The heat input was set to the nominal value corresponding to 30 occupants, 4335 W. This
value was selected because each device was designed to provide cooling for an RA with 30 occu­
pants. The SMs provided a nominal 3510 W, and the heated water tanks provided a nominal
825 W. Be-cause the space and volume inside the 20-person RA is less than would be required
per RA regula-tions, these conditions are more severe than those of an actual 30-person RA.
To represent a hot mine, the mine air temperature around the RA for these tests was ele­
vated to 29.4°C (85°F) using portable thermostatically controlled space heaters that were posi­
tioned around the RA. The heaters were on for the entire duration of the tests. The test was
begun as soon as the desired mine air temperature was reached.

3.3 Setup for CAS testing in 30-person BIP RA.

The 30-person BIP RA was created by partitioning the 60-person BIP RA with an insulated
wall. The resulting BIP RA was 6.86-m (22.5-ft) long by approximately 6.10-m (20-ft) wide
and roughly 1.98-m (6.5-ft) high. The insulated-wall frame was built using 2x4 wooden studs
spaced 40.6 cm (16 in) from center to center. A layer of 50.8-mm-thick (2-in-thick) polystyrene
foam was attached to each side of the wall frame. The cavities between the studs were filled
with fiberglass insulation batts.

553
 The air temperatures and RH values were measured using Vaisala HMP155 and HMP110
Humidity and Temperature Probes located at the centers of three equal-sized hypothetical sec­
tions: Section 1, Section 2, and Section 3. One Vaisala HMP155 was placed in the center of
Section 2 at mid-height, one Vaisala HMP110 was placed at the center of Section 1 at mid-
height, and one Vaisala HMP110 was placed at the center of Section 3 at mid-height. The
mine strata surface temperatures and temperatures at depth were measured using patch-style
Class A RTDs and PVC rods with RTDs affixed.
In order to evaluate the CAS in hot-mine conditions, the interior of the BIP RA was pre­
heated to a strata surface temperature of 29.4°C (85°F) using a 30-kW heater that was posi­
tioned near the center of the BIP RA and six MVA-controlled baseboard heaters that were
distributed within the BIP RA. The 30-kW heater and the baseboard heaters were operated
together to rapidly increase the BIP RA air and strata surface temperatures. After the initial
preheat period, the 30-kW heater was turned off and the MVAs were iteratively adjusted so
that the mine strata temperature stabilized at about 29.4°C (85°F) for 8 hours prior to begin­
ning the test.
The number of occupants that could be housed in the BIP RA with the CAS operating was
determined through a set of iterative tests. To provide cooling, liquid air was delivered to the
BIP RA from the CAS to a heat exchanger located inside the BIP RA. Because it was assumed
that CO2 scrubbers would be necessary for a CAS, the total heat input for each test was the
number of SMs used times 144.5 W, which includes 117 W to represent metabolic heat input
and 27.5 W to represent CO2 scrubber heat input. The liquid airflow rate from the CAS was
kept roughly constant. Small adjustments to the flow rate were made in an attempt to keep
the AT below 35°C (95°F) and to ensure that the liquid air would last for 96 hours.

3.4 AT calculation
To develop an equation to calculate the AT, or heat index, using the dry-bulb air temperature
and relative humidity, Rothfusz carried out a multiregression analysis on the tabular data
from Steadman (Rothfusz, 1990). The National Weather Service refined this approach to cal­
culate the AT. For dry-bulb air temperatures above 26.7°C (80°F) and an RH above 13%, the
AT was calculated using (National Weather Service, 2023):

where T is the dry-bulb air temperature in °F amd RH is the percent relative humidity. If
Equation 1 resulted in an AT below 26.7°C (80°F), the following simplified formula was used
to calculate AT (National Weather Service, 2023)

4 RESULTS

4.1 BAS tested in 60-person BIP RA

The BAS was able to keep the AT below the AT limit. The highest AT was calculated to be
24.2°C (75.5°F) and this occurred in Section 1, which was furthest from the location of the air
delivered from the BAS. The air temperature in Section 1 varied from 20.2°C–24.0°C (68.4°F–
75.2°F) with an average of 22.6°C (72.7°F), and the RH ranged from 54.4–80.7% with an aver­
age of 62.9%. The AT in Section 1 varied from 20.2°C–24.2°C (68.3°F to 75.5°F) with an aver­
age of 22.6°C (72.7°F). For these tests, the outside air temperature ranged from 9.2°C–25.8°C

554
(48.6°F–77.5°F) with an average of 17.7°C (63.8°F), and the outside RH ranged from 36.8%–
95.2% with an average of 72.0%.

4.2 BPAC tested in 20-person portable tent-type RA

The BPAC unit kept the AT under 35°C (95°F) until battery issues occurred during testing, and
the test was ended prior to 96 hours (refer to Figure 4). The average mine air temperature near
the RA varied from 28.3°C–29.4°C (83°F–88°F). The test was expected to last for 96 hours, but
higher than expected power consumption caused the BPAC’s main breaker to trip at about 38
hours. With the breaker tripped, the AT increased from about 29.4°C (88°F) to about 37.8°C
(100°F) within about 2 hours. After the breaker was reset, the BPAC was able to bring the AT
below 35°C (95°F). However, the battery ran out of power around 60 hours, so the test was
terminated. Just prior to the batteries running out of power, the AT was 33.3°C (92°F).

Figure 4. (a) Mine air average temperature and (b) RA interior temperature and %RH for BPAC AT
test.

4.3 CAS tested in 20-person portable tent-type RA

The mine air temperature around the RA was elevated to a nominal 29.4°C (85°F). After the
first 18 hours of the test, the actual average mine air temperature around the RA varied from
about 28.3°C–29.4°C (83°F–85°F) as shown in Figure 5. For the first 20 hours of the test, the
CAS was able to keep the AT below 35°C (95°F). However, due to ice build up on the heat
exchanger inside the tent, the AT exceeded 35°C (95°F) after about 18 hours. After 48 hours,
researchers temporarily halted the test to enter the tent and remove the builtup ice from the
heat exchanger. When the test was resumed, the CAS kept the AT below 35°C (95°F) from 60
hours until about 90 hours, when the CAS ran out of liquid air. Without cooling from the
CAS, the AT climbed to 60°C (140°F) in less than 8 hours.

4.4 CAS tested in 30-person BIP RA

The BIP RA was preheated to a mine strata surface temperature of 29.4°C (85°F). The corres­
ponding mine strata temperatures at 0.15-m (6-in), 0.61-m (24-in), and 1.22-m (48-in) deep
were 29.4°C (85°F), 24.4°C (76°F), and 14.4°C (58°F), respectively. With the CAS operating,
tests were conducted with the heat input of 24, 16, and 12 miners. The cryogenic airflow rate
was adjusted to keep the AT below 35°C (95°F) and to have enough liquid air to last 96
hours. An 8-hour-long test with the heat input of 12 miners was conducted without the CAS
for comparison.
When the heat input was set to that of 24 miners, the AT was 37.4°C (99.4°F) when the
CASran out of liquid air at 91 hours (refer to Table 1). With the heat input of 16 miners, the
CAS lasted the entire duration, but the AT reached 35.2°C (95.3°F), just over the AT limit.

555
Figure 5. (a) Average mine air temperature and (b) average RA interior temperature, %RH, and AT
for CAS AT test.

Table 1. Summary of CAS AT test results for a 29.4°C (85°F) 30-person BIP RA.
Parameter Test 1 Test 2 Test 3 Test 4

Number of SMs 24 16 12 12
Test Duration 96 hrs 96 hrs 96 hrs 8 hrs
CAS used Yes Yes Yes No
Time when CAS empty 91 hrs 96 hrs 91 hrs –
AT when CAS empty 37.4°C (99.4°F) – 30.3°C (86.6° F) –
48.1°C 35.2°C 37.9°C 40.7°C
AT at end of test (118.6°F) (95.3°F) (100.3°F) (105.2°F)

The CAS kept the AT below 35°C (95°F) with the heat input corresponding to 12 miners, but
at 91 hours, the CAS ran out of liquid air. At this time, the AT was 30.3°C (86.6°F). For the
test without the CAS with the heat input of 12 miners, the AT reached 40.7°C (105.2°F) after
just 8 hours.

5 DISCUSSSION

The BAS was able to keep the AT well below the AT limit. In addition, the BAS had no perform­
ance issues during testing. However, the outside atmospheric conditions were not as severe as
desired, as the dry-bulb air temperature was a maximum of only 25.8°C (77.5°F). Ideally, the
BAS would be evaluated with an outside air temperature matching that of southern mines, such
as the ones in Alabama, which are hotter than other United States underground coal mines. In
Brookwood, AL, July is the hottest month with an average high temperature of 33.1°C (91.6°F)
and an average RH of 73% (Weather U.S., 2022). Furthermore, to ensure adequate cooling in
Alabama-mine-like conditions, the BIP RA should be preheated to 29.4°C (85°F) as was done
with the evaluations of the BPAC and the CAS.
The BPAC was able to keep the AT below the limit, but higher than expected power con­
sumption caused issues with the unit’s power system. Because the device uses a compressor
that adjusts to the cooling demand, and the cooling demand was greater than expected, the
power consumption exceeded the capacity of the batteries and of the circuit breaker. In order
to ensure the device could last for an entire 96 hours, the battery capacity would have to be
increased. Under the test conditions here, the unit operated for about 60 hours. Therefore, the
battery capacity would need to be increased by about 60% to ensure the BPAC could operate
for 96 hours. The BPAC would need to satisfy federal requirements for permissibility in under­
ground coal mines. As tested, the BPAC was not evaluated subject to these requirements.

556
 The CAS was able to provide enough cooling to keep the 20-person portable RA below the
AT limit, but it had issues with ice buildup on the heat exchanger, and it ran out of liquid air
prior to the end of the 96-hour test. The heat exchanger design must be capable of having ice
buildup removed easily without risking damage. For the 20-person portable RA test, the CAS
2000-liter dewar had about 1850 liters of liquid air inside at the beginning of the test. During the
test, the CAS was set to deliver liquid air to the heat exchanger inside the tent at a nominal rate
of 220 liters/min. However, the CAS had limited flow rate control capability, so the actual flow
rate exceeded the target amount. Better flow control capability would be needed for a final
product.
The CAS tests on the 30-person 29.4°C (85°F) BIP RA showed promising results. With the
heat input of 24 occupants, the unit was able to keep the AT below 37.8°C (100°F) for 91 hours.
A small increase in liquid airflow rate and a larger volume of liquid air would have allowed the
unit to provide enough cooling to bring the AT below 35°C (95°F) and to last the full 96 hours
with 24 occupants. The tests with the heat input of 16 occupants showed the CAS would have
kept the AT below 35°C (95°F) with a slight increase in liquid airflow rate. The tests highlighted
the importance of having good flow control.
One of the challenges with the CAS is that this is a novel technology for mining applica­
tions. Because gas suppliers do not currently offer liquid air, mines would have to mix their
own liquid air, or a supplier would need to be convinced to provide this commodity. Another
challenge is the maintenance of a CAS in a mining environment. This would require the devel­
opment of a maintenance schedule and instructions. It is likely that a vendor would need to
provide maintenance services.

6 CONCLUSIONS

All three devices were able to reduce heat buildup within the RAs. The BAS would need to be
tested in more severe thermal conditions like those in southern mines to ensure the device would
be capable of keeping the apparent temperature (AT) below 35°C (95°F). Both the BPAC and
CAS were able to keep the AT under 35°C (95°F), but both devices had issues. The BPAC needs
improved battery capacity. In addition, the BPAC would need to pass permissibility testing. The
CAS would benefit from a heat exchanger design that is less prone to ice buildup and has
improved flow control.
To ensure hot mines can be kept under the AT limit without occupancy derating, cooling
devices would need to be used with RAs. The devices tested here could all serve this need.
The BAS is closest to being a usable product. The BPAC and CAS need some additional
improvements before they are ready to be used to provide cooling to RAs in productions
mines.

7 LIMITATIONS

The heat mitigation devices were evaluated only in the NIOSH Experimental Mine. If an
actual mine has less conductive strata, the resulting ATs would be higher. The BAS evaluation
was not conducted with hot outside ambient conditions or a preheated mine, both of which
would be more severe than the tested conditions. All of the devices described in this publica­
tion would require additional development prior to implementing them in commercial mines.

8 DISCLAIMER

The findings and conclusions in this report are those of the author(s) and do not necessarily
represent the official position of the National Institute for Occupational Safety and Health,
lefts for Disease Control and Prevention. Mention of any company or product does not con­
stitute endorsement by NIOSH.

557
REFERENCES

Bernard, T. E. 2011. Physiological analysis of human generated heat in a refuge alternative. NIOSH
Contract Report 254-2011-M-40932. Tampa, FL, University of South Florida.
CFR. 2018. Code of Federal Regulations, 30 CFR 7.506, Breathable air components. Washington, DC,
U.S. Government Publishing Office. https://www.govinfo.gov/content/pkg/CFR-2018-title30-vol1/pdf/
CFR-2018-title30-vol1-sec7-506.pdf
CFR. 2019. Code of Federal Regulations, 30 CFR 75.1506, Refuge alternatives. Washington, DC, U.S.
Government Publishing Office. https://www.govinfo.gov/content/pkg/CFR-2019-title30-vol1/pdf/
CFR-2019-title30-vol1-sec75-1506.pdf
CFR. 2021. Code of Federal Regulations, 30 CFR 7.508, Harmful gas removal components. https://
www.govinfo.gov/content/pkg/CFR-2021-title30-vol1/pdf/CFR-2021-title30-vol1-sec7-508.pdf.
Cryo Life Support Systems. 2016. Refuge alternative heat mitigation system utilizing advanced liquid air
technologies. Broad Agency Announcement (BAA) Contract 200-2016-91194. Pittsburgh, PA, USA,
the National Institute for Occupational Safety & Health. https://www.cdc.gov/niosh/mining/research
program/contracts/contract_200-2016-91194.html.
DRS. 2016. Development of Heat Mitigation Systems for Refuge Alternatives (ras) used in Underground
Coal Mines. Broad Agency Announcement (BAA) Contract 200-2016-91901. Pittsburgh, PA, USA, the
National Institute for Occupational Safety & Health. https://www.cdc.gov/niosh/mining/researchpro
gram/contracts/contract_200-2016-91901.html.
Lascar Electronics. 2022. Lascar Electronics EL-WiFi-T+ High Accuracy WiFi Temperature Data
Loggers. Accessed on 1/13/2023. https://www.lascarelectronics.com/easylog-el-wifi-tplus.
National Weather Service. 2023. The Heat Index Equation. Weather Prediction Center. Accessed on 1/
13/2023. https://www.wpc.ncep.noaa.gov/html/heatindex_equation.shtml.
OSHA. 2003. 1910.146 - permit-required confined spaces. Occupational Safety and Health Administra­
tion (OSHA). https://www.osha.gov/laws-regs/regulations/standardnumber/1910/1910.146.
Rothfusz, L. P. 1990. The heat index equation (or, more than you ever wanted to know about heat
index). Fort Worth, Texas: National Oceanic and Atmospheric Administration, National Weather Ser­
vice, Office of Meteorology, SR 90–23. https://www.weather.gov/media/ffc/ta_htindx.PDF.
Steadman, R. G. 1979. The assessment of sultriness. Part I: A temperature-humidity index based on human
physiology and clothing science. Journal of Applied Meteorology and Climatology, 18, 861–873. https://
journals.ametsoc.org/view/journals/apme/18/7/1520-0450_1979_018_0861_taospi_2_0_co_2.xml.
Vaisala. 2023a. Humidity and Temperature Probe HMP110. Accessed on 1/13/2023. https://www.vaisala.
com/en/products/instruments-sensors-and-other-measurement-devices/instruments-industrial-measure
ments/hmp110.
Vaisala. 2022b. Humidity and Temperature Probe HMP155. Accessed on 1/13/2023. https://www.vaisala.
com/en/products/weather-environmental-sensors/humicap-humidity-temperature-probe-hmp155.
Weather U.S. 2022. July weather forecast Brookwood, AL. Accessed on 1/13/2023. https://www.weather-
us.com/en/alabama-usa/brookwood-weather-july#temperature.
Yan, L., Fernando, R., Yantek, D., Carr, J., Reyes, M., Degennaro, C., Yonkey, J. & Srednicki, J. 2021.
Storage time and venting characteristics for cryogenic air supplies on cryocooler shutdown. The 21st Inter­
national Cryocooler Conference. Boulder, CO. https://cryocooler.org/resources/Documents/C21/085.pdf.
Yan, L., Yantek, D., Lutz, T., Yonkey, J. & Srednicki, J. 2017. Heat mitigation for underground coal
mine refuge alternatives using cryogenic air or borehole air supplies. ASME International Mechanical
Engineering Congress and Exposition. American Society of Mechanical Engineers.
Yan, L., Yantek, D., Reyes, M., Whisner, B., Bickson, J., Srednicki, J., Damiano, N. & Bauer, E. 2020.
Cryogenic air supply for cooling built-in-place refuge alternatives in hot mine. Mining, Metallurgy &
Exploration, 37, 861–871.
Yantek, D. 2014. RI 9695 - Investigation of temperature rise in mobile refuge alternatives. Publication
No. 2014–117, US Department of Health and Human Services, Centers for Disease Control and Pre­
vention. National Institute for Occupational Safety and Health, Office of Mine Safety and Health
Research. https://www.cdc.gov/niosh/mining/UserFiles/works/pdfs/2014-117.pdf.
Yantek, D.S., Yan, L., Damiano,N.W.,Reyes, M.A., and Srednicki, J.R.. 2019. A test method for evalu­
ating the thermal environment of underground coal mine refuge alternatives. International Journal of
Mining Science and Technology 29(3): 343–355. https://www.ncbi.nlm.nih.gov/pmc/articles/
PMC9278555/.

558
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Occupancy derating for underground coal mine refuge alternatives

D.S. Yantek & L. Yan

NIOSH Pittsburgh Mining Research Division

ABSTRACT: Federal regulations require refuge alternatives (RAs) to be installed in under­

ground coal mines in the United States. The thermal environment of RAs is a concern because
trapped miners could experience heat exhaustion or heat stroke, which is deadly. To prevent
exposing RA occupants to heat stress, regulations for RAs in underground coal mines man­
date an apparent temperature (AT) limit of 35°C (95°F). To examine occupancy derating to
meet the AT limit, researchers at the National Institute for Occupational Safety and Health
(NIOSH) conducted RA thermal simulations with validated thermal simulation models using
actual mine strata and temperature data. NIOSH examined occupancy derating for a portable
23-person tent-type RA and a portable 6-person rigid RA across a range of initial temperat­
ures for the NIOSH Experimental Mine strata. Occupancy derating was necessary for mine
air dry-bulb temperatures as low as 15.6°C (60°F). NIOSH also examined occupancy derating
for each RA in five mines across the United States using the measured worst-case temperat­
ures and mine strata of each mine. The results showed occupancy derating would be necessary
for the mines with relatively warm initial temperatures. RA manufacturers and mines can use
this information to ensure that RAs do not exceed the AT limit.

1 INTRODUCTION

On December 31, 2008, the Mine Safety and Health Administration (MSHA) published its
final rule “Refuge Alternatives for Underground Coal Mines” (Federal Register, 2008), which
became effective on March 2, 2009. This rule mandates underground coal mines to install
refuge alternatives (RAs) so that all miners, contractors, and visitors have access to an RA in
case of an inescapable mine emergency. The RA regulations require RAs to provide
a breathable air environment, food and water, and supplies needed to sustain trapped miners
for up to 96 hours. The regulations also provide requirements for the breathable air environ­
ment, available space and volume, and thermal conditions.
Due to human metabolic heat, carbon dioxide (CO2) scrubber system heat, and a limited
ability to dissipate heat, the thermal environment of an occupied RA can become severe
enough to cause heat exhaustion or heat stroke, which can be deadly (OSHA, 2003, Bernard,
2011). RA regulations for U.S. underground coal mines specify a maximum apparent tem­
perature (AT) of 35°C (95°F) (Steadman, 1979). For indoor environments without solar load­
ing, the AT is calculated using dry-bulb temperature and percent relative humidity (%RH)
(Rothfusz, 1990). The regulations require RA manufacturers to conduct testing to determine
an RA’s maximum occupancy without exceeding the AT limit.
Many factors affect the resulting RA thermal conditions including the number of RA occu­
pants, the mine air and strata temperature, and the mine strata thermal properties (Yan et al.,
2020, Yantek et al., 2017). As the number of occupants increases, the heat input also increases,
and the resulting RA thermal conditions become more severe. Higher initial mine air and
strata temperatures also increase the severity of an occupied RA’s thermal conditions. Mine
strata materials have low thermal conductivity, which limits conductive heat transfer from an
RA to the mine.

DOI: 10.1201/9781003429241-57

559
 To ensure that the AT limit is not reached, heat mitigation devices or occupancy derating
could be necessary for warm or hot mines. The National Institute for Occupational Safety and
Health (NIOSH) used test-result-validated RA thermal simulation models with actual mine
temperatures, %RH values, and strata compositions to examine occupancy derating for two
portable RAs: a 23-person portable tent-type RA and a 6-person portable steel RA. The
stated occupancies are based on the space and volume requirements of the RA regulations for
underground coal mine RAs (CFR, 2019). This paper discusses the results of the occupancy
derating simulations.

2 OCCUPANCY DERATING ANALYSIS METHOD

The concept of occupancy derating of an RA is to limit its occupancy to less than full capacity
in order to ensure that the conditions within the RA remain below the 35°C (95°F) AT limit.
The resulting thermal conditions within an RA are a function of the number of occupants, the
RA design, the mine strata thermal properties, and the mine air and strata temperatures. For
a given RA, it would be impractical to conduct testing to represent the conditions in all mines.
In order to examine the need for occupancy derating, NIOSH used validated thermal simula­
tion models (refer to Figure 1) of a 23-person portable tent-type RA and a 6-person portable steel
RA that were tested in the NIOSH Experimental Mine (EM) (Yan et al., 2016a, Bissert et al.,
2016, Yan and Yantek, 2018, Yan et al., 2020, Klein et al., 2017, Klein and Hepokoski, 2017)
(Yan et al., 2016b) (Klein, 2017a) (Klein, 2017b). In these models, the mine strata were modeled
as twenty-four 7.62-cm-thick (3-in-thick) layers up to a modeling depth of 1.83 m (6 ft). The
models included transient heat conduction into the mine roof, rib, and floor. The mine strata ther­
mal properties for each mine were used for the layers in the models. Human thermal models
(HTMs) were used to input the heat of actual occupants (Bernard et al., 2018, Klein and Hepo­
koski, 2017). The HTMs model the complex nature of heat transfer from the human body and
account for all manners of body heat transfer including vasodilation, perfusion, and evaporation
of sweat. The CO2 scrubber system heat input was set to 27.5 W per person (Shumaker, 2013).
For each RA, the mine dry-bulb air temperature, %RH, and mine strata temperatures were used
as initial conditions. A 96-hour transient thermal analysis was conducted with the maximum RA
occupancy to determine the final RA air temperature and %RH. The AT limit was calculated
from the predicted RA air temperature and %RH (Rothfusz, 1990). If the AT limit was reached,
the number of HTMs was reduced and the analysis was repeated. The maximum number of occu­
pants was then found by iteratively increasing or decreasing the number of HTMs.

Figure 1. Thermal simulation models used for occupancy derating analysis (a) 23-person portable
tent-type RA and (b) 6-person portable steel RA.

Initially, NIOSH examined occupancy derating for both RAs based on the EM. This ana­
lysis was conducted using the thermal simulation models of the RAs with the mine strata ther­
mal properties and thermal conditions of the EM. The EM floor strata was modeled as

560
a 15.2-cm-thick (6-inch-thick) layer of concrete over 168 cm (66 in) of siltstone, the EM rib
strata was modeled as a 2.5-cm-thick (1-inch-thick) layer of shotcrete over bituminous coal,
and the EM roof strata was modeled as 31 cm (12 in) of slate, 61 cm (24 in) of bituminous
coal, and 91 cm (36 in) of shale. To simplify the model, the 2.5-cm-thick (1-in-thick) layer of
shotcrete on the EM roof was ignored in the analysis. Thermal simulations were conducted
across a range of initial mine air, mine strata, and RA temperatures from 12.8°C–23.9°C (55°
F–75°F) in 2.8°C (5°F) increments. The initial %RH for these simulations was 85% based on
measured data from the RA tests in the EM.
Next, NIOSH examined occupancy derating for the portable tent-type and metal RAs for
five U.S. mines located in the Southwestern, Midwestern, Southern, Western, and Eastern
regions of the U.S. NIOSH collaborated with these mines to determine the mine temperatures,
%RH values, and mine strata properties to be used in the simulations (Bissert et al., 2016,
Yantek et al., 2017). The mine air temperature, %RH, and strata temperatures near existing
RAs were measured by mine personnel using air-temperature-measurement resistance tem­
perature detector (RTD) probes, a %RH instrument, RTD-instrumented PVC rods (refer to
Figure 2), and RTD thermometers provided by NIOSH. The mine strata temperatures were
measured at the surface and at depths of 0.61 m (24 in) and 1.22 m (48 in). Mine personnel
documented the temperatures at least once per month for an entire year. The worst-case com­
bination of mine air temperature, %RH, and mine strata temperatures were used as initial
conditions in the thermal simulations (see Table 1). Through conversations with mine person­
nel and NIOSH ground control researchers, the mine strata composition and layer thicknesses
were assumed based on the geographic location of each mine (see Table 2). The strata proper­
ties used in the thermal simulations are shown in Table 3.

Figure 2. (a) RTD-instrumented PVC rod for measuring strata temperatures and (b) picture showing
RTD-instrumented rod installed into the mine rib.

3 RESULTS

3.1 Results for NIOSH Experimental Mine

The analysis of the 23-person portable tent-type RA with the EM strata properties showed
that occupancy derating would be necessary for an initial temperature of 21.1°C (70°F) or
greater (see Table 3). At an initial temperature of 21.1°C (70°F), the occupancy would have to
be decreased from 23 to 16, a 30% reduction. For an initial temperature of 23.9°C (75°F), the
occupancy would have to be reduced from 23 to 9, a 61% reduction.

561
Table 1. Mine dimensions and worst-case combinations of %RH, mine air temperature, and mine
strata temperatures for five mines across the U.S.
Strata Temperatures
Mine
Mine Air Air % 0.61 m (24 1.22 m (48
Mine Width Height Temp. RH Surface in) Depth in) Depth

Southwestern 5.94 m 2.90 m 18.9°C 69 18.3°C 18.9°C 18.9°C

(19.5 ft) (9.5 ft) (65.6°F) (65.2°F) (65.6°F) (66.0°F)
Midwestern 5.49 m 1.68 m 20.6°C 92 19.4°C 18.3°C 17.8°C
(18 ft) (5.5 ft) (69.4°F) (67.2°F) (65.2°F) (64.1°F)
Southern 6.40 m 2.93 m 27.2°C 92 28.3°C 26.7°C 26.1°C
(21 ft) (9.6 ft) (81.3°F) (82.8°F) (80.1°F) (78.7°F)
Western 4.88 m 2.59 m 17.2°C 59 15.0°C 15.0°C 13.9°C
(16 ft) (8.5 ft) (62.7°F) (58.7°F) (58.5°F) (57.1°F)
Eastern 5.79 m 2.21 18.9°C 96 17.2°C 17.2°C 17.2°C
(19 ft) (7.25 ft) (66.4°F) (63.4°F) (62.9°F) (62.5°F)

Table 2. Mine strata composition of five mines used in occupancy derating study. Note: If only one mater­
ial is listed, the entire 1.8-m-thick (6-ft-thick) strata layer is assumed to consist of the listed material.
Strata Composition
Mine Floor Ribs Roof

Southwestern mudstone bituminous coal 72 cm (2.5 ft) bituminous coal, 49 cm

(1.6 ft) shale, 18 cm (0.6 ft) bitumin­
ous coal, 9.1 cm (0.3 ft) mudstone
Midwestern 122 cm (4 ft) bituminous coal shale
claystone, 61 cm
(2 ft) sandstone
Southern shale 91 cm (3 ft) bituminous coal, shale
91 cm (3 ft) shale
Western bituminous coal 5.1 cm (2 in) shotcrete, 89 cm 30 cm (1 ft) bituminous coal, 122 cm
(35 in) bituminous coal, 89 cm (4 ft) sandstone, 30 cm (1 ft) bitumin­
(35 in) mudstone ous coal
Eastern shale bituminous coal/shale 91 cm (3 ft) bituminous coal, 91 cm (3
composite ft) shale

Table 3. Mine strata thermal properties used in the occupancy derating studies.
Thermal Conductivity Specific Heat
Material Density (kg/m3) (W/m-K) (J/kg-K)

shale, mudstone, claystone 2600 1.0 1000

bituminous coal (non-EM) 1346 0.33 1380
Sandstone 2500 1.7 920
Siltstone 2600 2.7 1000
concrete, shotcrete (measured) 2039 1.3 1088
EM bituminous coal (measured) 1500 0.58 1297

Note: For measured values, core samples collected by NIOSH were tested by Decagon Devices, Inc., 2365 NE
Hopkins Court, Pullman, WA 99163.

The analysis of the 6-person portable steel RA with the EM strata properties also showed
that occupancy derating would be necessary for an initial temperature of 21.1°C (70°F) or
greater (see Table 4). At an initial temperature of 21.1°C (70°F), the occupancy would have to

562
be decreased from 6 to 4, a 33% reduction. For an initial temperature of 23.9°C (75°F), the
occupancy would have to be reduced from 6 to 2, a 67% reduction.

Table 4. Derating results for 23-person portable tent-type RA from simulations of the EM. Red indi­
cates the AT exceeded 35°C (95°F). Note: The initial %RH for these simulations was 85 %RH.
Number of Final RA Air­
Initial Temperature Occupants Temperature Final RA %RH Final AT

12.8°C (55°F) 23 22.8°C (73.0°F) 85.2 22.2°C (72.0°F)

15.6°C (60°F) 23 25.1°C (77.1°F) 85.7 25.9°C (78.6°F)
18.3C (65°F) 23 28.1°C (82.5°F) 90.2 33.9°C (93.1°F)
21.1°C (70°F) 23 30.8°C (87.5°F) 90.2 44.1°C (111.4°F)
17 28.5°C (83.3°F) 92.8 36.0°C (96.8°F)
16 27.8°C (82.1°F) 93.9 33.9°C (93.0°F)
23.9°C (75°F) 17 31.2°C (88.2°F) 92.4 46.6°C (115.8°F)
11 28.8°C (83.9°F) 94.9 37.7°C (99.9°F)
9 27.9°C (82.2°F) 97.0 34.6°C (94.3°F)

Table 5. Derating results for 6-person portable steel RA from simulations of the EM. Red indicates the
AT exceeded 35°C (95°F). Note: The initial %RH for these simulations was 85 %RH.
Number of Final RA Air­
Initial Temperature Occupants Temperature Final RA % RH Final AT

12.8°C (55°F) 6 22.9°C(73.2°F) 83.9 22.5°C (72.5°F)

15.6°C (60°F) 6 25.3°C (77.6°F) 84.0 26.5°C (79.7°F)
18.3C (65°F) 6 27.4°C (81.3°F) 90.0 31.9°C (89.4°F)
21.1°C (70°F) 5 28.5°C (83.3°F) 94.5 36.3°C (97.4°F)
4 27.4°C (81.3°F) 93.3 32.2°C (90.0°F)
23.9°C (75°F) 3 28.7°C (83.6°F) 96.1 37.4°C (99.3°F)
2 27.3°C (81.2°F) 95.6 32.5°C (90.5°F)

3.2 Results for five mines across the U.S.

The analysis of the 23-person portable tent-type RA for the five mines across the U.S. showed
that some mines would not require derating, while others would require derating (see Table 5).
The Midwestern mine and the Southern mine would require derating. The Midwestern mine
would require the occupancy to be reduced from 23 to 21, a 9% reduction. The Southern mine
would require the occupancy to be reduced from 23 to 1, a 96% reduction. The Southwestern,
Western, and Eastern mines would not require derating. It should be noted that, if the initial tem­
perature was a few degrees higher, the Southwestern mine would also require derating as the pre­
dicted final air temperature would likely exceed the AT limit.
The analysis of the 6-person portable steel RA for the five mines across the U.S. showed that
only the Southern mine exceeded the AT limit (see Table 6). For the Southern mine, the AT limit
would be exceeded even with only a single occupant. Therefore, the Southern mine would need to
implement cooling to use the 6-person portable steel RA. At full occupancy, the Midwestern
mine was just below the AT limit. If the initial temperature of the Midwestern and Southwestern
mines were a degree or two higher, the number of occupants for the 6-person portable metal RA
would need to be reduced for each mine.

563
Table 6. Derating results for 23-person portable tent-type RA from simulations of five mines across the
U.S. Red indicates the AT exceeded 35°C (95°F).
Number
Initial AirTem­ Initial of Occu­ Final AirTem­ Final
Mine perature % RH pants perature % RH Final AT

Southwestern 18.7°C (65.6°F) 69 23 28.1°C (82.5°F) 92.5 34.3°C (93.7°F)

Midwestern 20.8°C (69.4°F) 92 23 28.6°C (83.4°F) 93.3 36.3°C (97.4°F)
22 28.2°C (82.8°F) 92.6 35.1°C (95.1°F)
21 27.9°C (82.2°F) 91.9 33.8°C (92.8°F)
Southern 27.4°C (81.3°F) 92 9 30.7°C (87.2°F) 99.4 47.1°C (116.8°F)
4 29.0°C (84.2°F) 98.7 39.4°C (103.0°F)
3 28.6°C (83.5°F) 98.8 37.9°C (100.2°F)
2 28.2°C (82.8°F) 99.0 36.3°C (97.3°F)
1 27.8°C (82.1°F) 99.3 34.7°C (94.5°F)
Western 17.1°C (62.7°F) 59 23 26.1°C (78.9°F) 84.9 27.9°C (82.3°F)
Eastern 19.1°C (66.4°F) 96 23 31.4°C (80.6°F) 88.6 30.6°C (87.1°F)

Table 7. Derating results for 6-person portable metal RA from simulations of five mines across the U.S.
Red indicates the AT exceeded 35°C (95°F).
Number
Initial AirTem­ Initial % of Occu­ Final RA Final
Mine perature RH pants AirTemperature % RH Final AT

Southwestern 18.7°C (65.6°F) 69 6 27.7°C (81.8°F) 90.0 32.7°C (90.8°F)

Midwestern 20.8°C (69.4°F) 92 6 28.2°C (82.7°F) 91.6 34.6°C (94.2°F)
Southern 27.4°C (81.3°F) 92 2 30.0°C (86.0°F) 98.6 43.6°C (110.4°F)
1 28.6°C (83.5°F) 98.3 37.7°C (99.9°F)
Western 17.1°C (62.7°F) 59 6 24.3°C (75.8°F) 83.3 24.7°C (76.4°F)
Eastern 19.1°C (66.4°F) 96 6 26.6°C (79.8°F) 85.7 29.1°C (84.4°F)

4 DISCUSSION

The derating simulation results show that the occupancy derating percentages depend on
the initial mine temperature and mine strata composition based on geographical location.
For the portable tent-type RA, the predicted AT was just under the AT limit when the
mine initial temperature was 18.3°C (65°F) for the EM and for the Southwestern mine. If
the mine strata thermal conductivity was slightly lower, the AT limit would have been
reached because the final RA air temperature would have been higher. At an initial mine
temperature of 21.1°C (70°F), the portable tent-type RA needed to be derated by 30% and
the portable metal RA needed to be derated by 33% for the EM. In addition, for the Mid­
western mine with a mine initial temperature of just under 21.1°C (70°F), the portable
tent-type RA needed to be derated by 10% and the portable metal RA was almost at the
AT limit. For the EM, at an initial mine temperature of 23.9°C (75°F), the portable tent-
type RA and portable metal RA would have to be derated by 60% and 67%, respectively.
For the EM at an initial mine temperature of 26.7°C (80°F), using the dry-bulb tempera­
ture rise per occupant and relative humidity for the 23.9°C (75°F) initial mine temperature,
the occupancy of the portable tent-type RA would have to be derated by 90% and the port­
able steel RA would not be able to be used without cooling. For the Southern mine, which
had an initial mine temperature of about 26.7°C (80°F), the derating analysis showed that
the portable tent-type RA would have to be derated by 96% and the portable metal RA
would require cooling.

564
 Based on the occupancy derating analysis results for the EM and the actual mines, a table
of expected percent occupancy derating needed for compliance was developed (refer to
Table 7). In the development of this table, we considered the results from more than 30 in-
mine tests of the portable 23-person tent-type and 6-person metal RAs. In addition, we con­
sidered simulation results that examined the resulting thermal conditions for mine strata
compositions having “low” to “high” thermal conductivity and a range of mine entry sizes
(Bissert et al., 2016). Furthermore, we considered simulation results that examined how ini­
tial mine air temperatures, initial mine strata surface temperature, and initial mine strata
temperature at depth affects the resulting thermal conditions (Yantek et al., 2017). To
account for variations in mine strata composition, entry size, and RA thermal properties,
the lowest percent occupancy reductions from the analysis were decreased slightly, and the
highest percent occupancy reductions from the analysis were increased slightly. This table
should not be interpreted as a rigid set of percent occupancy reductions as it is based on
thermal simulations and has not been validated by testing. Furthermore, the table is based
on results from only one portable tent-type RA and one portable metal RA and a few mines
with assumed mine strata composition based solely on geographical location. However,
Table 7 can be used as a guide to provide an expected possible range of percent occupancy
reductions. It should be noted that above 26.7°C (80°F), it is likely that cooling or dehumidi­
fication would be required because the buildup of moisture inside an RA would likely result
in a relative humidity of greater than 90%. At this relative humidity, the 35°C (95°F) AT
limit would be reached at an RA air temperature of 28.3C (83°F). For initial mine temperat­
ures of 29.4°C (85°F), the AT limit would be reached for a relative humidity of 75%. Based
on all NIOSH testing and thermal simulations, the relative humidity would exceed 90%.
Therefore, cooling or dehumidification would be necessary for mine air temperatures above
29.4°C (85°F).

Table 8. Approximate percent occupancy derating needed for a 23-person portable tent-type
RA and a 6-person portable steel RA.
Range of Initial Mine Temperatures % Occupancy Derating Needed for Compliance

< 15.6°C (60°F) Derating unnecessary

16.6°C–18.3°C (60°F–65°F) 0%–10%
18.3°C–21.1°C (65°F–70°F) 0%–40%
21.1°C–23.9°C (70°F–75°F) 25%–75%
23.9°C–26.7°C (75°F–80°F) 50%–90%
26.7°C–29.4°C (80°F–85°F) 90%–100%
>29.4°C (85°F) Cooling/dehumidification required

5 CONCLUSIONS

Although it would be impractical to conduct testing for all possible mine temperatures and
mine strata compositions, our analysis results can be used as a rough guide for planning pur­
poses. In order to ensure compliance with the 35°C (95°F) AT limit, testing or simulations
using validated thermal simulation models should be used to estimate the final AT, which pri­
marily depends on initial mine temperature and mine strata composition.
In the absence of cooling or dehumidification, our analysis indicates that occupancy derat­
ing of RAs may be required for some mines. For mines below 18.3°C (65°F) with very low
mine strata thermal conductivity, the occupancy may need to be slightly reduced. For mines
between 18.3°C–21.1°C (65°F–70°F), occupancy derating would be necessary for mines with
strata having low thermal conductivity. From 21.1°C–26.7°C (70°F–80°F), the results indicate
that occupancy derating would be necessary. Above (80°F), occupancy derating would
approach 100%, and cooling or dehumidification would be needed to ensure that the thermal
conditions inside an RA stay below the 35°C (95°F) AT limit.

565
6 LIMITATIONS

Because the analysis presented here was performed on only two RAs with conditions from only
five mines across the U.S. and one nonproduction mine, the results should not be assumed to
directly apply to mine-specific RA installations. Additional testing or analysis would be neces­
sary to determine occupancy derating percentages for a specific RA installation.

REFERENCES

Bernard, T., Yantek, D. & Thimons, E. 2018. Estimation of metabolic heat input for refuge alternative
thermal testing and simulation. Mining Engineering, 70, 50.
Bernard, T. E. 2011. Physiological analysis of human generated heat in a refuge alternative. NIOSH Con­
tract Report 254-2011-M-40932. Tampa, FL, University of South Florida.
Bissert, P., Yantek, D., Klein, M. & Yan, L. 2016. Analysis of heat loss mechanisms for mobile tent-type
refuge alternatives. Transactions of Society for Mining, Metallurgy, and Exploration, Inc, 340, 70.
CFR. 2019. Code of Federal Regulations, 30 CFR 75.1506, Refuge alternatives. Washington, DC, U.S.
Government Publishing Office. https://www.govinfo.gov/content/pkg/CFR-2019-title30-vol1/pdf/
CFR-2019-title30-vol1-sec75-1506.pdf accessed on 2/24/2023.
Federal Register. 2008. Refuge Alternatives for Underground Coal Mines; Final Rule, Department of
Labor, Mine Safety and Health Administration, 30 CFR Parts 7 and 75. Washington, DC, US Gov­
ernment Printing Office.
Klein, M. 2017a. Mine shelter thermal analysis - 23-person inflatable RA TAITherm model validation and
analysis. Report 3018–001, Revision 4.1. ThermoAnalytics, Inc., 23440 Airpark Blvd., Calumet, MI.
Klein, M. 2017b. Mine shelter thermal analysis - steel RA TAITherm model validation and analysis.
Report 3090–001, Revi-sion 4.1. ThermoAnalytics, Inc., 23440 Airpark Blvd., Calumet, MI.
Klein, M. & Hepokoski, M. 2017. Human thermoregulation model for analyzing the performance of
mine refuge alternatives. Mining Engineering, 69.
Klein, M., Yantek, D., Hepokoski, M. & Yan, L. 2017. Prediction of human core temperature rise and
moisture loss in refuge alternatives for underground coal mines. Transactions of Society for Mining,
Metallurgy, and Exploration, Inc, 342, 29.
OSHA. 2003. Permit-required confined spaces. Department of Labor, Occupational Safety and Health
Administration, https://www.osha.gov/laws-regs/regulations/standardnumber/1910/1910.146 accessed
on 2/24/2023.
Rothfusz, L. P. 1990. The heat index equation (or, more than you ever wanted to know about heat
index). Fort Worth, Texas: National Oceanic and Atmospheric Administration, National Weather Ser­
vice, Office of Meteorology, SR 90–23. https://www.weather.gov/media/ffc/ta_htindx.PDF accessed on
2/24/2023
Shumaker, W. A. 2013. Information relayed from Wesley Shumaker, Mechanical Engineer, MSHA
Approval and Certification Center, to NIOSH Pittsburgh Mining Research Division regarding the
heat used to represent a refuge alternative CO2 scrubber system during heat and humidity testing. U.S.
Department of Labor, Mine Safety and Health Administration, Approval and Certification Center.
Triadelphia, WV.
Steadman, R. G. 1979. The assessment of sultriness. Part I: A temperature-humidity index based on
human physiology and clothing science. Journal of Applied Meteorology and Climatology, 18, 861–873.
Yan, L. & Yantek, D. 2018. Portable refuge alternatives temperature and humidity tests. Mining engin­
eering, 70, 43.
Yan, L., Yantek, D., Klein, M. & Bissert, P. 2016a. Interior thermal environment of a 6-person metal-
type refuge alternative (RA). ASME International Mechanical Engineering Congress and Exposition.
Phoenix, Arizona, USA, American Society of Mechanical Engineers.
Yan, L., Yantek, D., Klein, M., Bissert, P. & Matetic, R. 2016b. Validation of temperature and humidity
thermal model of 23-person tent-type refuge alternative. Mining engineering, 68, 97.
Yan, L., Yantek, D. & Reyes, M. 2020. Underground mine air and strata temperature change due to the
use of refuge alternatives. Mining, Metallurgy & Exploration, 37, 773–781.
Yantek, D., Yan, L., Bissert, P. & Klein, M. 2017. Effects of mine strata thermal behavior and mine ini­
tial temperatures on mobile refuge alternative temperature. Mining engineering, 69, 41.

566
Renewable / alternative energy in mine ventilation
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Incorporating droplet dynamics to improve the reduced-order

model of spray freezing for mine heating applications

M. Mohit
McGill University, Montreal, Quebec, Canada

S. Akhtar
Polytechnique Montreal, Montreal, Quebec, Canada

M. Xu & A.P. Sasmito

McGill University, Montreal, Quebec, Canada

ABSTRACT: Spray freezing technology has been shown its exceptional efficiency, safety, and
sustainability for underground mine heating. Designing spray freezing systems requires
a mathematical model with the rigorous formulation and fast computation methods, capable of
predicting performance indicators. Existing models for spray freezing often take the droplet
motion and velocity distribution as a priori, thus making it less feasible in practice. In the present
work, a novel reduced-order model is developed to calculate the distributions of droplet velocity
and residence time for various spray configurations (namely, flat fan, hollow cone, and full cone)
and droplet diameters. The velocity distributions are then incorporated into a robust heat transfer
model for mine heating to improve the prediction of the droplets freezing time and overall heat
transfer rate (HTR). Consequently, the residence time distribution is used along with droplets
diameter distribution to calculate the ice packing factor (IPF) and cooling capacity of the system.

1 INTRODUCTION

In recent years, studying different clean technologies for providing the heating demand of
underground mines has attracted researchers’ attention due to the importance of employing
eco-friendly systems in mining industries. Spray freezing technology as a renewable and effi­
cient method for air heating in under-ground mines of sub-arctic regions is one of these tech­
nologies. In this method, the sub-arctic air stream is heated by using the water latent heat of
solidification. An example of spray freezing application for underground mine heating has
been introduced by Stachulak (1991). Through spraying water into the cold air stream, water
droplets freeze, and the air temperature increases. The schematic of the spray freezing setup is
illustrated in Figure 1 (a). To efficiently evaluate the performance of spray freezing setups, it
is necessary to develop mathematical models with high accuracy and low computational cost.
The two important aspects of spray freezing modeling are the droplet freezing and droplet
dynamics. Droplet freezing process has been the topic of study by many scholars due to its
importance from both the theoretical and practical point of view. The solid-liquid phase
change phenomenon of water droplets is categorized under the subject of Stefan problem by
many scholars. Various numerical, analytical, asymptotic, and hybrid methods have been
developed in the literature to solve the Stefan problem in different conditions and coordinates
(Rubinšteĭn, 2000; Meirmanov, 2011). One powerful method to investigate Stefan problems
subject to various boundary conditions is the perturbation series analysis which has been used
in many works (Caldwell and Kwan, 2003; Akhtar, Xu and Sasmito, 2021b). It should be

DOI: 10.1201/9781003429241-58

569
noted that most results on droplet solidification modeling do not take nucleation and den­
dritic growth stages into account. This shortcoming has been addressed by Akhtar, Xu and
Sasmito (2021a) by developing a reduced-order semi-analytical model for droplet solidifica­
tion that can capture all freezing stages. On the other hand, several researchers have studied
droplets motion in different conditions. Analyzing droplets motion and dynamics is essential
to calculate spray characteristics in various fields like spray cooling, spray freezing, and pesti­
cide spraying. The equation of motion for droplets, bubbles, and solid particles has been dis­
cussed by Michaelides (1997). This equation of motion can be numerically solved for droplets
to find their trajectory and velocity (Deshpande, Gao and Trujillo, 2011; Zhang, Zhang and
Bai, 2022). However, droplets dynamic has not been considered in the literature of spray-
freezing modeling to the best of authors’ knowledge. Furthermore, to fully capture the drop­
lets dynamic, it is necessary to take the spray shape into account. This issue has not been
addressed in the literature as well.
Akhtar, Xu and Sasmito (2021) studied the thermal characteristics of the spray freezing setup
by developing a reduced-order theoretical-statistical model. However, the effects of droplets
motion are not considered in that study. In the present work, a reduced-order model for droplets
motion has been combined with the droplet freezing model introduced by Akhtar, Xu and Sas­
mito (2021a) to accurately capture system-scale characteristics of spray freezing setup. Different
spray shapes have been studied and the droplets velocity distribution and residence times have
been calculated by numerical approaches. After that, the results have been coupled with the drop­
let freezing model to calculate heat transfer coefficient (HTC), HTR, IPF, and cooling capacity.

2 MATHEMATICAL MODELING

The mathematical modeling of the spraying freezing system can be divided into three parts,
namely, a) Droplet freezing modeling; b) Droplet dynamics modeling; c) Spray modeling. The
first two parts are treated as particle-scale processes, while the third one can be seen as
a macro-scale process. The following assumptions are made in the present work:
i) Droplets are spherical with no deformation and there is no change in the droplet volume
during solidification;
ii) Droplets break-up and coalescence are neglected;
iii) The thermo-physical properties are assumed to be phase-dependent in the droplet freezing
model (Sections 2.1.1 and 2.1.2);
iv) When a droplet collides with the system boundaries, it does not contribute to the heat
transfer anymore.
In the following subsections, each modeling part will be discussed.

2.1 Particle-scale parts

2.1.1 Droplet freezing model

For droplets solidification process, the model introduced by Akhtar, Xu and Sasmito (2021a) is
used in the present work. As discussed by Akhtar, Xu and Sasmito (2021a), the droplet freezing
process consists of five different stages, namely, water super-cooling, nucleation, dendritic
growth, equilibrium freezing, and solid sub-cooling. Figure 1 (b) illustrates different stages of
droplet freezing. The mathematical representation of each stage has been given in the work of
Akhtar, Xu and Sasmito (2021a). A similar approach is employed in the present work to calcu­
late the droplet thermal characteristics and freezing time. The following correlation holds
between the dimensionless time and the interface position (Akhtar, Xu and Sasmito, 2021a):

where:

570
 The dimensionless variables are defined as:

where ri and rI,i are the droplet and interface radiuses, respectively, t is time, and αs is the ice
thermal diffusivity. Ste and Bi are the Stefan and Biot numbers, respectively, as defined below:

Equations (1)-(3) will be used to calculate the interface position of each droplet in the spray
freezing process. The schematic of the solid-liquid interface motion is shown in Figure 3 (c).

Figure 1. (a) Spray freezing setup (b) Droplet freezing curve (c) Solid-liquid interface motion for
a droplet.

2.1.2 Droplet dynamics model

The droplets motion is modeled numerically by solving the equation of motion on
a Lagrangian mesh. The position and velocity of a droplet are correlated as below:

where xd and ud are the droplet position and velocity vector, respectively. The drag force and
gravity are the two significant forces acting on a droplet and other terms in the equation of

571
motion can be neglected (Zhang, Zhang and Bai, 2022). Writing the second Newton’s law for
a single droplet gives:

which can be simplified as:

where:

In above equations, md, ρd, rd, and Dd are the droplet mass, density, radius, and diameter,
respectively, ug, ρg, μg are the gas velocity vector, density, and dynamic viscosity, respectively,
g is the gravitational acceleration vector, and Red is the Reynolds number according to the drop­
let-gas relative velocity. CD is the drag coefficient which can be calculated for different values of
Red as below (Zhang, Zhang and Bai, 2022):

From (5) to (8), the droplets position and velocity can be calculated over time using numer­
ical approaches. Furthermore, the residence time of each droplet, i.e., the total time that each
droplet remains within the system domain, can be calculated by comparing its position with
the system boundaries.

2.1.3 Effects of droplet motion on HTCs

While the models proposed in Sections 2.1.1 and 2.1.2 are solved separately, in the present
work to improve the computational efficiency, the effects of the droplet motion on the freezing
model is taken into account. In fact, droplets motion affects the calculation of the overall
HTC. As discussed by Akhtar, Xu and Sasmito (2021a), the overall HTC is calculated by con­
sidering the effects of forced convection, radiation, sublimation, and evaporation:

where indices eq, c, r, s, and e are referring to terms “equivalent”, “convection”, “radiation”,
“sublimation”, and “evaporation”, respectively. Coefficient hr is not affected by the droplets
velocity because it is not a function of Reynolds number (Red), but hc, hs, and he are func­
tions of Red. In previous works (Akhtar, Xu and Sasmito, 2021; Akhtar, Xu and Sasmito,
2021a), since droplets velocity is neglected, Red is calculated based on the air velocity:

In the present work, Reynolds number is calculated as given below:

This value will be used in the calculation of hc, hs, and he instead of Reold
d . It should be
noted that ud is obtained from the droplets dynamic model given in Section 2.1.2 and is

572
a function of droplet diameter, droplet initial direction, and spray shape. In the following
part, a novel approach for considering the spray shape and droplets initial velocity direction
in the droplet’s velocity calculation will be discussed.

2.2 System-scale part: Spray modeling

A spray consists of a mass of droplets with different diameters and velocities. Several distribu­
tions have been used in the literature to represent sprays diameter distribution like normal
and Rosin-Rammler distributions. In the present work, it is assumed that the flat spray shape
is produced by a fan spray and hollow cone and full cone shapes are produced by a pressure
swirl atomizer. The experimental correlations reported by Liu (1999) are used to calculate the
Sauter mean diameter (SMD) of these sprays and a Rosin-Rammler distribution is then util­
ized to find the droplet diameter distribution. The distribution is discretized by considering
n different values for droplet diameter where n is set to 20 in the present work. The probability
distribution of diameters for each spray shape is illustrated in Figure 2 (a) and (b).
To calculate droplets velocities over time, it is necessary to model their initial direction. Here,
it is assumed that all droplets have an equal initial velocity magnitude, but their initial directions
are different. The droplets initial velocity directions are different for each spray shape. In the
present work, the droplets different initial directions are modeled by assuming a finite number
of paths for droplets in each spray shape as depicted in Figure 3. In this figure, each black circle
indicates a droplet path which can be seen as a straight line that connects the spray position to
that circle. Some of the droplet paths are shown with dotted lines in Figure 3. The x, y, and
z-components of the initial droplets’ velocity can be calculated for each droplet path in each
spray shape using the spray angle θ which is a known variable based on spray characteristics,
and the number of paths. To incorporate droplet diameter distribution in this framework, it is
assumed that the distributions in all droplet paths are equal. The residence times and velocities
in different paths are then averaged to find a mean value for each droplet diameter.

2.3 Computational framework and other considerations

The presented mathematical model is computed by Python programming language using
a hybrid method (both analytical and numerical approaches). IPF and HTR are considered as
large-scale indicators of the setup performance. The IPF is defined as the ratio of the frozen
water over the overall mass of sprayed water:

where m_ s is the mass flow rate of the ice formed during the operation and m
_ t is the total mass
flow rate of sprayed water. To calculate IPF, the frozen portion of each droplet should be
calculated and then integrated for all droplets:

in which f(Di) is the probability of a droplet with diameter Di to exist, N_ is the particle injec­
tion rate of the spray, and ms,i is the mass of the frozen portion of a droplet with diameter Di.
N_ can be calculated as:

where mi is the mass of a droplet with diameter Di. For a droplet with diameter Di,
variable ms,i in (14) is calculated by finding the interface position based on the droplet resi­
dence time (τr,i) and freezing time (τf,i) using equations (1)-(3). It should be noted that the

573
volume change occurring during the water freezing process is neglected in the droplet freezing
model developed by Akhtar, Xu and Sasmito (2021a). This will result in the mass conservation
violation due to the phase-dependency of density which yields a considerable error in IPF cal­
culation. Hence, a correction is applied to IPF calculation by using the mass conservation
equation. The total mass of water is conserved during the process, hence:

where ml.i is the volume of unfrozen water of the droplet with diameter Di and Δm is the mass
difference due to neglecting the droplet diameter movement. Δm can be added to the numer­
ator of the IPF correlation (14) to compensate for the effect of droplet volume increase during
the freezing. Therefore:

The HTR is calculated as:

in which ei is the total heat emission of a droplet with diameter Di during solidification and
can be calculated as:

Figure 2. Probability distributions (a) Flat spray (Liu, 1999) (b) Cone sprays (Liu, 1999) (c) Experimen­
tal data from Santangelo (2010).

where hst,i is the HTC of each freezing stage, Ti is the temperature profile over time, T∞ is the
air temperature, and Ai is the droplet surface area.

3 MODEL VALIDATION

To validate the model, the experimental data from the frood-stobie mine reported by Stachulak
(1991) has been used. The droplet distribution in this part is adopted from the experimental
results proposed by Santangelo (2010) which suitably mimics the situation in frood-stobie mine.
This distribution is depicted in Figure 2 (c). As discussed by Stachulak (1991), the total amount
of ice produced during the winter operation of the frood-stobie mine is approximately mice,exp
=145.000 tons. Using the developed model in the present work with the distribution shown in

574
Figure 2 (c), the IPF will be equal to 92%. Considering 1,200 operating hours of two stopes
during the winter and an average water flow rate of 18.9 kg/s as reported by Stachulak (1991)
and Trapani and Chen (2017), the total amount of ice formed during winter can be calculated
using the model output as below:

The relative error is equal to:

Figure 3. Modeling droplet initial directions (a) Flat spray (b) Hollow cone spray (c) Full cone spray.

4 RESULTS AND DISCUSSION

The developed model has been run using the specifications given by Stachulak (1991) for the
frood-stobie mine. The spray shapes discussed in Section 2.2 along with their corresponding
diameter distributions have been considered in running the code one by one. For each spray,
different parameters have been calculated and averaged using the given distribution. The
results are summarized in Table 1. As can be seen from this table, the values of IPF and HTR
are greater for hollow cone and full cone sprays compared to flat spray. This difference is due
to the different diameter distributions or SMDs of the sprays as reflected in Figure 2. While
larger droplets require more time for freezing, their residence time is smaller compared to
finer droplets due to the larger gravity force imposed on them. Hence, the ratio of the frozen
volume over the droplet volume is smaller for droplets of larger diameter which in turn results
in a smaller IPF for sprays with larger SMDs.
Comparing the results for hollow cone and full cone spray shapes shows that although the
spray shapes and droplet trajectories are different, but the average values listed in Table 1 are
quite close. The reason is that the diameter distribution is the same for both cases as they are
produced by a specific type of nozzle.

575
Table 1. Average values for different spray shapes.
Spray shape
Parameter and unit Symbol Flat spray Hollow cone Full cone

Sauter mean diameter (μm) SMD 822.83 258.49 258.49

Average residence time (s) �τR 7.34 7.46 7.46
Average freezing time (s) �τF 19.19 4.81 4.80
Average nucleation time (s) �τN 8.79 2.17 2.17
Average relative velocity (m/s) �vrel 4.52 1.72 1.74
Average interface radius (μm) �rI 540.54 44.30 44.24
Heat transfer rate (MW) HTR 1.98 5.56 5.58
Ice packing factor (%) IPF 30.82 94.11 94.14
Cooling capacity (GJ) Cc 8,404.7 25,664.2 25,672.4

The ice produced during the cold days of winter can be stored and utilized in summer for
cooling purposes. The cooling capacity (Cc) of a spray-freezing setup is calculated by multiply­
ing the rate of ice production by the water latent heat of fusion and the total operating time:

where top is the total operating time. As can be inferred from (21), a larger IPF will result in
a greater cooling capacity, hence using cone-sprays will provide more cooling capacity during
the summer operations.

5 CONCLUSION

In the literature of spray freezing modeling, there is a research gap in bridging between the
models of small-scale droplet freezing phenomena and system-scale spray characteristics. The
present work aimed to address this shortcoming by developing a reduced-order model for cal­
culating droplets velocity and residence times and incorporating this model into the existing
droplet-scale models of freezing phenomena and experimental correlations of spray character­
istics. For modeling the droplets dynamic, different spray shapes, i.e., flat, hollow cone, and
full cone sprays have been modeled in a computationally efficient manner. Droplets diameter
distribution is then included in the model for each spray shape to determine the system-scale
characteristics of the setup like the HTR and IPF. The total amount of ice formed during the
winter operations has been used as a benchmark to validate the model. The model results indi­
cate that sprays with finer droplets (like cone sprays) will provide greater HTR and IPF. The
developed model can be extended to capture other contributing phenomena like droplet-
droplet and droplet-wall collisions.

ACKNOWLEDGEMENT

The authors wish to thank the Fonds de recherche du Québec - Nature et technologies (FRQ-
NT) (PR-300597) and the Natural Sciences and Engineering Research Council of Canada
(NSERC) (RGPIN-2021-02901). The first author would also like to show his gratitude to the
McGill Engineering Doctoral Award (MEDA) for supporting his doctoral research.

REFERENCES

Akhtar, Saad, Xu, M. and Sasmito, A. P. 2021a. Development and validation of a semi-analytical frame-
work for droplet freezing with heterogeneous nucleation and non-linear interface kinetics, Inter-
national Journal of Heat and Mass Transfer. Elsevier Ltd, 166, p. 120734.

576
Akhtar, Saad, Xu, M. and Sasmito, A. P. 2021b. Development and validation of an asymptotic solution
for a two-phase Stefan problem in a droplet subjected to convective boundary condition, International
Journal of Thermal Sciences. Elsevier Masson SAS, 164(December 2020), p. 106923.
Akhtar, S., Xu, M. and Sasmito, A. P. 2021. Spray freezing for mine heating a statistical perspective,
Mine Ventilation, (2008), pp. 357–365.
Caldwell, J. and Kwan, Y. Y. 2003. On the perturbation method for the Stefan problem with time-dependent
boundary conditions, International Journal of Heat and Mass Transfer. Elsevier, 46(8), pp. 1497–1501.
Deshpande, S., Gao, J. and Trujillo, M. F. 2011. Characteristics of hollow cone sprays in crossflow,
Atomization and Sprays. Begel House Inc., 21(4).
Liu, H. 1999. Empirical and Analytical Correlations of Droplet Properties, in Liu, H. B. T.-S. and E. of
D. (ed.). Norwich, NY: William Andrew Publishing, pp. 238–314.
Meirmanov, A. M. 2011. The stefan problem. Walter de Gruyter.
Michaelides, E. E. 1997. Review—the transient equation of iviotion for particles, bubbles, and droplets,
Journal of Fluids Engineering, Transactions of the ASME, 119(2), pp. 233–247.
Rubinšteĭn, L. I. 2000. The stefan problem. American Mathematical Soc.
Santangelo, P. E. 2010. Characterization of high-pressure water-mist sprays: Experimental analysis of drop­
let size and dispersion, Experimental Thermal and Fluid Science. Elsevier Inc., 34(8), pp. 1353–1366.
Stachulak, J.S., 1991. Ventilation strategy and unique air conditioning at Inco Limited. CIM Bul-letin
(Canadian Institute of Mining and Metallurgy); (Canada), 84(950).
Trapani, K. and Chen, Z. 2017. Computational fluid dynamic modelling of the Frood-Stobie ice stope
thermal storage for mine ventilation heating, Proceedings of the Eighth International Conference on
Deep and High Stress Mining, pp. 289–298.
Zhang, Haibin, Zhang, Hao and Bai, B. 2022. Numerical Study on Flow Structures of a Hollow Cone
Spray in Crossflow, Heat Transfer Engineering. Taylor & Francis, 43(8–10), pp. 737–753.

577
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Analysis of small-scale lithium-ion batteries under thermal abuse

A. Iqbal, G. Xu, R.I. Pushparaj & O.B. Salami

Department of Mining and Explosives Engineering, Missouri University of Science and Technology,
Rolla, MO, USA

ABSTRACT: Lithium-Ion Batteries (LIB) have dominated the energy market for several
decades due to their high energy density and long-life cycle. However, several fire accidents in
electric vehicles have raised questions about their safety concerns. Researchers have identified
thermal runaways to be the major reason for the fire susceptibility of LIBs. One of the major
risks in accidents involving LIB fires is toxic gaseous emissions. Hence, it is necessary to
understand those toxic emissions to properly counter them. In this study, small-scale battery
thermal abuse tests will be performed, and the toxic gases will be analyzed using different
tools such as Fourier Transform Infrared (FTIR) spectroscopy ultimately. 5 different chemis­
tries of roughly the same nominal capacity i.e., 3000 mAh, will be tested and a comparison of
the gaseous emissions will be presented in future studies, which has not been done before.
However, so far, we have performed preliminary experiments to study the surface temperature
of cells. The results of this study showed that the thermal runaway of the cell is triggered when
the surface temperature of the cell reaches 190°C, and it keeps on increasing for up to 208°C,
even without the use of the heating source, the duration of the experiment varies from 30min
to 45min. Based on the study results, suitable experiment scenarios are designed the study the
gaseous analysis of the LIB on the cell level.

Keywords: Thermal abuse, Small-scale LIB, Cell surface temperature

1 INTRODUCTION

Lithium-ion batteries (LIBs) are the most commonly used batteries in battery electric vehicles
(BEV) due to their high energy density and long lifetime (Larsson et al., 2016). However, sev­
eral BEV accidents involving battery fires have raised concerning alarms about the fire risks
of LIBs (Sun et al., 2020)(CMTeam, 2022). Since 2006, Federal Aviation Administration
(FAA) has recorded about 433 fire accidents involving LIBs in aviation (FAA, 2022). In 2012,
A Toyota Prius (hybrid) was recorded to have caught fire after being flooded by hurricane
Sandy (Labovick, 2021)(GARTHWAITE, 2012). In 2013, two Tesla Model Ss were reported
to have caught fire after being involved in two separate collisions(Musk, 2013)(George, 2013).
Further down the timeline, different car makers have experienced BEVs catching fire accidents
such as Volkswagen in 2017 (Traugott, 2021), BMW i8 (Bruce, 2019), and two separate events
involving Porsche Panamera in 2019 and a Taycan in 2020 happened in Portugal and USA
respectively (Labovick, 2021). Several other events have happened when BEV caught fire
while charging or parking idle (Wayland, Micheal. kolodny, 2021)(Yoney, 2021).
Existing research has identified thermal runaways to be the probable reason for most of the
fire accidents in BEVs (Sun et al., 2020)(Liu et al., 2021)(Liu et al., 2016). Thermal runaway
happens when the battery pack ignites spontaneously due to overheating, according to the lit­
erature. It is caused by either electrical abuse, mechanical abuse/collision, or thermal abuse (Sun
et al., 2020)(Liu et al., 2016)(Zhong et al., 2019)(Zhao et al., 2021). Once thermal runaway is

DOI: 10.1201/9781003429241-59

578
triggered, chemical reactions start to take place inside the cell/pack, causing the emission of
toxic fumes. Due to the presence of fuel and oxidizer inside the cell, these chemical reactions will
most likely continue spontaneously, giving rise to heat and toxic gases during the process
(Doughty and Roth, 2012). Several studies have analyzed the thermal characteristics of LIB
fires, for instance, Yangyang Fu et.al (Fu et al., 2015) concluded from their experimental study
that the stored electrical energy has a significant impact on the burning behaviors of LIBs. In
the contrast, Larsson et.al (Larsson et al., 2016) stated that the stored electrical energy has nom­
inal dependence on the thermal properties of the battery.
One of the most under-explored areas of BEV fires is the amount and types of gases emitted
during fires. These toxic gases pose an immense threat to human health (Peng et al., 2020), as
fatalities and injuries are mostly caused by the inhalation of toxic gases in fire accidents (Stec,
2017). Several researchers have performed experiments on the fire behavior and propagation
of the fire in LIB cells (Fu et al., 2015)(He et al., 2020). However, very little information is
present on the types and amount of toxic gases as suggested by a review performed by Sun
et al. (Sun et al., 2020). The presence of toxic gases such as CO, CO2, HCl, HF, SO2, and
POF3 has been confirmed by different studies(Larsson et al., 2017) (Ribière et al., 2012), How­
ever, the amount of the gases has still not been well understood. Larsson et al (Larsson et al.,
2017) indicated the presence of HF, POF3, and PF5, However, the cells used by Larsson et.al
(Larsson et al., 2017) were of different chemistries and the number of cells in each battery
sample varied. In a separate study (Lecocq et al., 2016), the fire behavior of Li-ion batteries
with LiPF6 chemistry was analyzed and the presence of toxic gases such as HF, and SO2 was
confirmed, however, they were also unable to quantify the amount of the gases and also the
experiment was done on only one chemistry of LIB. This research is also endorsed by
a different study in which LIB cells were dissembled and the heat value and gases analysis was
performed, confirming the presence of PF5, CO2, and HF (Kriston et al., 2019). However, the
quantification and complete analysis of the gases is unavailable. Besides, reignition is another
major problem related to LIB fires and several researchers have studied the effect of different
extinguishers on LIB fires, and so far, a specifically effective extinguisher against LIB fires is
non-existent (Maloney, 2014)(Luo et al., 2018)(Su et al., 2014).
The main objective of this study is to evaluate the toxic fumes emitted during fire incidents
involving Li-ion batteries. However, so far, only preliminary tests are performed, and the results
are used as a guideline for the future study of the project. Similar studies have been performed
by other researchers; however, the difference lies in the number of chemistries of LIB used, the
types of gases analyzed, and the SOC relative to the LIB chemistries tested. The objective of the
research will be achieved by first identifying and quantifying the toxic fumes by performing fire
abuse tests. Applying fire directly to the battery cells, the cells will be left to burn out completely
at different states of charge (SOC). The smoke of the batteries will be analyzed for the type and
amount of gases, with the help of a multi-gas detector and also spectroscopy using FTIR. This
analysis will be useful in identifying and quantifying the gases emitted during battery fires,
which can further be used in designing specialized protective equipment and fire extinguisher
designs, to elevate the health and safety standards of battery-powered vehicles.

2 METHODOLOGY

2.1 Lithium-ion battery incinerator

To conduct the experiments, we designed and built the Lithium-ion battery Incinerator. The
dimensions of the incinerator are 20×20×60 inches. It is made up of 26-gauge meta sheets and
¾” metal cube rods screwed together, and it is equipped with a 20×20×24” chamber for fire,
and other divisions for the data logger and DC power supply. We fixed caster wheels to the
bottom of the structure to easily move it in and out of the lab to protect ourselves from toxic
exposures. The top of the incinerator is transitioned into a 4” duct with the help of an alumi­
num transit. Which is extended with a flexible plastic duct for about 5 feet. At the end of the
duct, a 4” inline duct fan is fixed to pull the smoke out of the chamber Figure 1.

579
Figure 1. Lithium-ion battery incinerator.

2.2 Cell selection

LIBs are characterized by medium to high energy density, with a variety of chemistries. These
cells have layers of Anode (copper foil coated with a specialty carbon) and Cathode (aluminum
foil coated with a lithiated metal oxide or phosphate) separated by a microporous polyolefin
separator. The electrode is an organic solvent and a dissolved lithium salt provides a medium
for lithium-ion transport (Williams and Back, 2014). After a careful review of the literature, we
found out that the emission of toxic gases is majorly influenced by the state of charge (SOC)
and the cathode chemistry of the LIB. Hence, we decided on the five chemistries that are most
common among electric vehicle manufacturers i.e. Lithium Iron Phosphate (LFP), Lithium
Nickle Manganese Cobalt (NMC), Lithium Nickle Cobalt Aluminum (NCA), Lithium Titanate
(LTO), and Lithium Manganese Oxide (LMO) at five different SOC i.e. 0%, 25%, 50%, 75%,
and 100%. The detailed information on the cells can be seen in. So far, NiCad cells at 100%
SOC were tested as preliminary experiments.

Table 1. Detailed specifications of the cells.

Chemistry Voltage (V) Capacity (Ah) Packaging No of cells Manufacturer

LFP 3.3 2.5 26650 Cylindrical 10 Lithiumwerks

NMC 3.6 2.5 18650 Cylindrical 20 Samsung
NCA 3.6 2.25 18650 Cylindrical 10 Panasonic
LTO 3.3 3.0 18650 Cylindrical 10 Titanate batteries
LMO 3.75 2.8 18650 Cylindrical 10 Samsung

2.3 Abuse initiation

Different studies have used different fire sources including an electric heater (Fu et al., 2015),
a thermal oven (Larsson and Mellander, 2014), and a propane burner (Larsson et al., 2016).
In this study, small-scale battery fire tests are conducted, and we used 18 Gauge Kenthal A-1
Annealed round, Nickle Chromium (Ni-Chrome) resistance wire, which is powered by NICE
POWER DC power supply to heat the cells. The wore is a wire wound around the cell as

580
shown in Figure 2 and is gradually heated up at a rate of 4°C per min until the cell starts
smoking off. A high-resolution thermal camera (FILR E4) Figure 7a, is used to capture the
burning process of the battery based on the flame profile and temperature data, through the
transparent window of the chamber.

Figure 2. a: the Nickel Chromium resistance wire, b: application of heat using Ni-Chrome resistance wire.

2.4 Emissions analysis

For thermal management of the cell, we made use of the cell surface temperature which was
recorded using J-thermocouples attached to the surface of the sample with the help of heat-
resistant aluminum adhesive tape. While the data was recorded using the Anabi AT44532 tem­
perature data logger
Figure 3 Thermocouples (TC) location on the cell is shown in Figure 4, TCs are designed to
be attached to the cell at 1/4th of the length from each end.

Figure 3. a: J thermocouples, b: the AT4532 temperature data logger.

For the analysis of the smoke, we made used Fourier Transform InfraRed (FTIR) spectros­
copy. Thermo Scientific Antaris IGS analyzer Figure 5, which is also reported to have been
used by F Larsson (Larsson et al., 2017). The sample is burned in the chamber, and the smoke
is pulled out by the inline duct fan at an average speed of 40 L/sec, however before the smoke
is released into the atmosphere, the sample is sucked into the FTIR gas chamber using an
external vacuum pump at 0.05 L/sec. The gas cell conditions are set at 180°C and 650 mmHG,
and these were the conditions at which the FTIR Fire Science Calibration Library will be used
for gaseous analysis. The experiment lasted for 45 min and the results are based on real-time
analysis of the smoke.

581
Figure 4. Location of thermocouples on the cell.

Figure 5. Thermo scientific antaris IGS analyzer.

Fire will be directly applied to a battery cell in the chamber, once it starts to smoke, it will
be collected and analyzed in by the multi-gas detectors and the sample of the gas will also be
collected for FTIR analysis for the POF5, PF3, and HF. The experimental setup was bor­
rowed and modified from a study performed by F. Larsson (Larsson et al., 2017) in 2017.
Once the gas sample is analyzed, and based on those analyses, PPEs designs, and Fire suppres­
sion techniques will be recommended.
Figure 6 shows the illustration of the experimental setup used for the analysis of Toxic emis­
sions of LIB cells. Starting from Step 1 includes lithium-ion battery cells indicated by A1,
which are first charged to the desired state of charge with a MIBOXER Cell charger indicated
by A2 after that in step 2, J thermocouples (B1) are attached to the cell surface which is con­
nected to Anabi AT44532 temperature data logger (B2). In the next step, the 18 Gauge
Kenthal A-1 Annealed round Nickle Cobalt resistance wire (C1) is wound into a coil and
powered with a DC power supply (C2), and the system is placed in the incinerator. The cell is
then heated till it starts smoking and the smoke is directed toward the FTIR (D1) in step 4,
and the gases are analyzed with OMNIC software using a personal computer (PC). The data
from the temperature data logger is also stored in the PC using AT4500 software, as shown in
Figure 6, Step 4 and step 5 are not a part of this report.
Compared to the similar setup used by other researchers, the advantage is that this setup is
easily and commercially available in the market, this setup is much easier and economical to
build and use for battery experiments. Other researchers have used a similar setup, however,
they have some equipment such as SBI apparatus, which is much more lab specific and cost too

582
Figure 6. Experimental setup used for toxic gas analysis of the LIB emissions.

much. This setup is a modified version of that used by Larsson et.al (Larsson et al., 2017) How­
ever, the difference is that the designed experiment can collect the temperature data using both
thermocouples and a thermal imaging camera Figure 7(a), which will be further processed for
fire profiling analysis. This will help us to better understand the fire profile and devise effective
suppression solutions and help in the development of emergency PPEs for LIB fires.
Another aspect of the novelty of this research is that so far, no researcher has identified and
analyzed all the gases at once while studying five different LIB chemistries with the same
power and charge capacities. For instance, F Larsson et.al (Larsson et al., 2017) studied three
chemistries but with variable capacities, P. Ribière et.al (Ribière et al., 2012) and Yangyang
Fu et.al (Fu et al., 2015) studied only one chemistry. Hence studying different chemistries at
the same capacity level will give us an insight into how different chemistries behave under the
same circumstances, and it will be a realistic comparison.

3 PRELIMINARY RESULTS

Initially, trial experiments were performed to find a suitable method and apparatus for the
experiments. The experiments were repeated 6 times, and most of the results are the same as
described below. Different types of abuse methods were applied including propane burner,
heating strip, and Ni-Chrome resistance wire. Tests indicated problems with controlling the
heat applied to the sample using the propane burner. However, the heating strip and the Ni-
Chrome resistance wire techniques performed better. Ni-Chrome resistance wire with a DC
voltage controller, as it is economical compared to the heating strip, is recommended, and will
be used further in this research.
The surface temperature of the cell was measured with the help of thermocouples, TC1 is
attached to the top (valve) end while TC2 is connected to the bottom end Figure 4, data were
recorded continuously every 10 sec. In these preliminary experiments, it was observed that the
cell starts smoking when the surface temperature of the cell exceeds 190°C. The electrolyte is
ejected from the cell. A mild hissing sound is evident that intensifies over the remainder of the
experiment until all the gases are emitted from the cell. The burning process was also captured
with a thermal camera to analyze the flame profile and surface temperature of the cell
Figure 7.

583
Figure 7. A: FLIR E4 Thermal Imaging Camera, b: Thermal image of the burning process of the
cell.

The power voltage of the DC power supply was increased after each 10 min. which can be
seen in Figure 8. The power of the supply was kept constant at 20W for the first 20 min. It can
be seen on the surface of the temperature profile of the cell. Here, the rate of temperature
increase doubles in the third section of the experiment. When the temperature of the cell
reached 220°C, it started smoking vigorously, and the power supply was turned off. The cell
kept smoking for 10 more mins and the temperature of the cell decreased thereafter. Figure 8
shows the data captured by only two thermocouples because only two of them were attached
to the cell.

Figure 8. Data Logger results of the surface temperature of the cell trial 1; TC1 (red), TC2 (yellow).

In another trial was repeated with the same procedure and different results were achieved
this time for the surface temperature of the battery. The experiment lasted for more than
one hour and the liquid was still confined within the cell. The maximum surface temperature
was seen to be up to 280 °C. At the time, t=34.0 min, the power source was turned off. The
surface temperature was seen to rise to 190 °C. It further rose to 208 °C without any external
source. It can, therefore, be inferred that thermal runaway occurs when the temperature of the
cell reaches 190 ° Figure 9.

584
Figure 9. Data logger results of the surface temperature of the cell.

4 CONCLUSIONS

The preliminary tests run on the small-scale LIB yield some scientific results. They are
described below:
• It takes 45 mins for the cell to burn out with the 50W DC power supply.
• The cell starts to smoke when the temperature reaches 190 °C. This also could be assumed
to be the temperature at which the thermal runaway begins.

5 FUTURE STUDIES

For future recommendations, this study provides valuable input on cell behavior under ther­
mal abuse. The authors will continue to work on capturing the smoke coming out of the cell
and analyze the nature of the smoke for the type and amount of different toxic gases with the
help of FTIR, further in the project.

NOTE

This is a work in progress, and the study is still not complete yet.

REFERENCES

Bruce, C., 2019. Smoking BMW i8 Dumped In Water By Firefighters [WWW Document]. URL https://
www.motor1.com/news/315476/bmw-i8-fire-reponse-netherlands/
CMTeam, 2022. How Many Electric Cars Catch Fire Every Year: Electric Car Fire Statistics (2022 Find­
ings) [WWW Document]. URL https://www.carsmetric.com/electric-car-fire-statistics/#:~:text=Key
Electric Car Fire Statistics 1 Electric-powered vehicles,vehicles and 3.4%25 for hybrid vehicles. More
items.
Doughty, D.H., Roth, E.P., 2012. A general discussion of Li ion battery safety. Electrochem. Soc. Inter­
face 21, 37.
FAA, 2022. Lithium Battery Incidents [WWW Document]. URL https://www.faa.gov/hazmat/resources/
lithium_batteries/incidents
Fu, Y., Lu, S., Li, K., Liu, C., Cheng, X., Zhang, H., 2015. An experimental study on burning behaviors
of 18650 lithium ion batteries using a cone calorimeter. J. Power Sources 273, 216–222.
Garthwaite, J., 2012. Mystery at Port Newark: Why Did 17 Plug-In Cars Burn? NewYork Times.

585
George, P., 2013. Another Tesla Model S Caught Fire After A Crash In Mexico [WWW Document].
URL https://jalopnik.com/another-tesla-model-s-caught-fire-after-a-crash-in-mexi-1453376349.
He, X., Restuccia, F., Zhang, Y., Hu, Z., Huang, X., Fang, J., Rein, G., 2020. Experimental study of
self-heating ignition of lithium-ion batteries during storage: effect of the number of cells. Fire Technol.
56, 2649–2669.
Kriston, A., Adanouj, I., Ruiz, V., Pfrang, A., 2019. Quantification and simulation of thermal decompos­
ition reactions of Li-ion battery materials by simultaneous thermal analysis coupled with gas analysis.
J. Power Sources 435, 226774.
Labovick, 2021. ELECTRIC VEHICLE FIRE INCIDENTS AND STATISTICS [WWW Document].
URL https://www.labovick.com/blog/electric-vehicle-fire-incidents-and-stats/
Larsson, F., Andersson, P., Blomqvist, P., Mellander, B.-E., 2017. Toxic fluoride gas emissions from
lithium-ion battery fires. Sci. Rep. 7, 1–13.
Larsson, F., Andersson, P., Mellander, B.-E., 2016. Lithium-ion battery aspects on fires in electrified
vehicles on the basis of experimental abuse tests. Batteries 2, 9.
Larsson, F., Mellander, B.-E., 2014. Abuse by external heating, overcharge and short circuiting of com­
mercial lithium-ion battery cells. J. Electrochem. Soc. 161, A1611.
Lecocq, A., Eshetu, G.G., Grugeon, S., Martin, N., Laruelle, S., Marlair, G., 2016. Scenario-based pre­
diction of Li-ion batteries fire-induced toxicity. J. Power Sources 316, 197–206.
Liu, X., Wu, Z., Stoliarov, S.I., Denlinger, M., Masias, A., Snyder, K., 2016. Heat release during
thermally-induced failure of a lithium ion battery: Impact of cathode composition. Fire Saf. J. 85,
10–22.
Liu, Y., Sun, P., Niu, H., Huang, X., Rein, G., 2021. Propensity to self-heating ignition of open-circuit
pouch lithium-ion battery pile on a hot boundary. Fire Saf. J. 120, 103081. https://doi.org/10.1016/J.
FIRESAF.2020.103081
Luo, W., Zhu, S., Gong, J., Zhou, Z., 2018. Research and development of fire extinguishing technology
for power lithium batteries. Procedia Eng. 211, 531–537.
Maloney, T., 2014. Extinguishment of lithium-ion and lithium-metal battery fires. US Fed. Aviat. Adm.
Washington, DC, USA 46–51.
Musk, E., 2013. Model S Fire [WWW Document]. URL https://www.tesla.com/blog/model-s-fire
Peng, Y., Yang, L., Ju, X., Liao, B., Ye, K., Li, L., Cao, B., Ni, Y., 2020. A comprehensive investigation
on the thermal and toxic hazards of large format lithium-ion batteries with LiFePO4 cathode.
J. Hazard. Mater. 381, 120916.
Ribière, P., Grugeon, S., Morcrette, M., Boyanov, S., Laruelle, S., Marlair, G., 2012. Investigation on
the fire-induced hazards of Li-ion battery cells by fire calorimetry. Energy Environ. Sci. 5, 5271–5280.
Stec, A.A., 2017. Fire toxicity–The elephant in the room? Fire Saf. J. 91, 79–90.
Su, C.-H., Chen, C.-C., Liaw, H.-J., Wang, S.-C., 2014. The assessment of fire suppression capability for
the ammonium dihydrogen phosphate dry powder of commercial fire extinguishers. Procedia Eng. 84,
485–490.
Sun, P., Bisschop, R., Niu, H., Huang, X., 2020. A review of battery fires in electric vehicles. Fire Tech­
nol. 56, 1361–1410.
Traugott, J., 2021. Two-Day-Old VW Golf Hybrid Explodes While Driving [WWW Document]. URL
https://carbuzz.com/news/two-day-old-vw-golf-hybrid-explodes-while-driving.
Wayland, Micheal. Kolodny, L., 2021. Fires, probes, recalls: The shift to electric vehicles is costing auto­
makers billions [WWW Document]. URL https://www.cnbc.com/2021/08/19/fires-probes-recalls-auto
makers-spend-billions-in-shift-to-evs.html
Williams, F.W., Back, G.G., 2014. Lithium battery fire tests and mitigation. NAVAL RESEARCH LAB
WASHINGTON DC CHEMISTRY DIV.
Yoney, D., 2021. Hyundai Kona Electric Fires In Norway And Korea Cause Concern [WWW Docu­
ment]. URL https://insideevs.com/news/515983/kona-electric-fire-norway-korea/.
Zhao, J., Xue, F., Fu, Y., Cheng, Y., Yang, H., Lu, S., 2021. A comparative study on the thermal runaway
inhibition of 18650 lithium-ion batteries by different fire extinguishing agents. Iscience 24, 102854.
Zhong, G., Mao, B., Wang, C., Jiang, L., Xu, K., Sun, J., Wang, Q., 2019. Thermal runaway and fire
behavior investigation of lithium ion batteries using modified cone calorimeter. J. Therm. Anal.
Calorim. 135, 2879–2889.

586
Ventilation monitoring and measurement
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Gauge and tube surveys: What is their future and that of

underground measurements generally as mines transition towards
greater use of Big Data and Artificial Intelligence systems?

D.J. Brake
Mine Ventilation Australia

ABSTRACT: Gauge and Tube (also known as Trailing Hose or P-Q) Surveys have been
widely used in underground mine ventilation for frictional pressure loss surveys and ventilation
model calibrations for a very long time. They are one of the most basic types of underground
ventilation measurement. However, there are many factors that have been contributing to and
accelerating the decline of this type of survey and underground measurements more generally. In
most cases, barometric pressure surveys are now the preferred method for validating ventilation
circuits and models and can be completed more quickly, more cheaply, with fewer personnel and
with less disruption to mine operations. In addition, the growing use of real-time instrumentation
of the ventilation system combined with Big Data analysis techniques is reducing the amount of
manual measurements and other data collection required. Artificial Intelligence is also likely to
assist in the future with providing statistically reliable advice regarding the location, type and fre­
quency of underground measurements, as well as answers to many other ventilation questions
including validating ventilation models and providing site-specific advice on management of
upset conditions such as fan or ventilation control failures or outages, fires or explosions. This
paper discusses the factors behind these trends and identifies some of the advantages and disad­
vantages of both gauge and tube and barometric pressure surveys, as well as the broader poten­
tial use of Big Data and Artificial Intelligence to assist with decisions regarding underground
measurements generally and assisting with other ventilation-related advice.

1 INTRODUCTION

Gauge and tube ventilation surveys have been the mainstream ventilation survey technique
for measuring underground flow directions and volumes for a very long time. However, this
type of survey started falling from favour in Australian hardrock mines from the early 1980s
and was almost entirely abandoned by the early 1990s. Gauge and tube surveys persevered
somewhat longer in coal mines but have also now been abandoned for many years.
The coup de grâce for the decline of this type of survey occurred during the transitional
period when underground mines went from 5-day operations to 7-day operations. This major
structural change in the industry was usually associated with moving from 8-hour to 12-hour
shifts and in many cases in Australia, also associated with moving from residential to fly-in,
fly-out (FIFO) operations.
The nature of gauge and tube surveys means they need to be undertaken when there is the
minimum possible activity in the mine resulting in the minimum disruption to the ventilation cir­
cuits. This was manageable when gauge and tube surveys could be undertaken when the mine
was mostly idle over a weekend and ventilation staff lived in a nearby town and could complete
the survey on “overtime”. When mines moved to 7-day operations, gauge and tube surveys
became increasingly problematic.

DOI: 10.1201/9781003429241-60

589
 However, the reasons for their declining use were not only related to the loss of available
non-production time underground to conduct the survey but also to the availability of suitable
personnel and their time and the rise of alternative measurement technologies and processes.
In addition, the rapid advances in the use of real-time instrumentation, Big Data analysis
techniques and Artificial Intelligence are likely to become a significant factor in the science
and practice of underground measurements and ventilation model validation, fault-finding
and decision-making in the medium term.
This paper summarises the author’s opinions about Guage and Tube Surveys in particular
and extends these reflections to the case of underground measurements more generally.

2 VENTILATION SURVEYS

In most cases, the purposes of a P-Q survey in a mine are to:

• Check airflow directions for compliance with the required safe and effective circuit config­
uration, fresh air bases and egress routes
• Measure windspeeds and hence volumetric airflow rates to ensure statutory or good prac­
tice compliance with minimum or maximum wind speeds, diesel requirements, gas dilution,
recirculation, etc.
• Measure pressure differentials for management of unsafe pressures on ventilation controls,
avoidance of excessive leakage or spontaneous combustion and the like
• Measure frictional pressure gradients in a mine to establish high-resistance or high-cost
paths for debottlenecking by appropriate circuit changes or installation of circuit fans
• Provide a baseline (trending) for fault-finding in the ventilation circuits
• Validate the as-built ventilation model to ensure it is a reliable tool for future mine plan­
ning, e.g. modelling major extensions to ventilation circuits or assessing fan duties for the
purchase or relocation of primary fans, etc. This last objective has really only arisen since
computer modelling of mine ventilation networks became practical in the 1970s.
From the mine’s point of view, the method used to measure the parameters required to
achieve these purposes is not important, providing the objectives themselves are achieved. In
particular, it is not important to obtain a direct measurement of the resistance of individual
airways or even the entire mine for the mine ventilation engineer. The critical underground
measurements for validating a ventilation model are airflow directions and volume flows,
pressure differentials across key parts of the circuit and checking that fans are operating on
their curves (Brake, 2015; Rowland, 2011).
Even more broadly than this, it could be argued that the future of mine ventilation measure­
ments will be to use real-time instrumentation combined with fewer, but smarter, manual
measurements and integrate these very targeted manual measurements with the use of increas­
ingly sophisticated Big Data and mine ventilation Artificial Intelligence technologies to signifi­
cantly improve the timeliness and reliability of mine ventilation decisions and advice.

3 PROBLEMS WITH GAUGE AND TUBE SURVEYS

The theory and techniques for gauge and tube (G&T) surveys are well described by McPher­
son (2007). However, there are a number of drawbacks of using gauge and tube surveys in
modern mines, despite their historical popularity.
• In a gauge and tube survey, the resistance of every section of airway (calculated from every
flow and pressure measurement of the gauge and tube as it is advanced) has to be added to
the resistance of the previous gauge and tube measurement. Any errors that creep in even­
tually get carried through the system (at least unless a second closed loop can be formed).
• Airway resistances often change over time for a variety of reasons, so that just surveying
the incremental new parts of the ventilation circuit since the last survey can and will

590
 eventually result in problems in the model as older airways are altered and these changes
are not necessarily recognised.
• There is significant manpower and time required and hence significant labour cost.
• There is the opportunity cost to these time-consuming surveys. While the ventilation team
is undertaking a gauge and tube survey, they are not doing other, often critical, ventilation
work and this lost time cannot be made up
• There is the disruption to the operations with restrictions on equipment travel and/or venti­
lation control operation during the period of the survey to obtain accurate results
• There are potential safety issues for the survey team if surveying in high activity areas, e.g.
on main ramps or travelways or in exhaust airways during production periods.
• It is not possible or very difficult to undertake a gauge and tube survey in many vertical
raises or other airways in which travel through is not possible.
Expanding the remit of this discussion from gauge and tube surveys to the role of underground
ventilation measurements more generally, it could easily be argued that most of the underground
measurements taken by the ventilation professional are never used again and contribute little to
the effective running of the mine, or the validation of the ventilation model in the mine. In some
cases, the collection of underground measurements actually reduces the ability of the ventilation
professional to perform other more value-adding activities and therefore decreases the overall
effectiveness of the ventilation system and negatively impacts on the ventilation outcomes on site.

4 CHANGING TECHNOLOGIES

There are four trends over the past few decades impacting on the methodology for conducting
ventilation measurements surveys underground.
Firstly, the development of very comprehensive and accurate ventilation modelling pro­
grams. Compared to the pre-computer era or even the period of early ventilation modelling
software, modern mine ventilation modelling programs can comprehensively include all the
factors that drive flow and pressure in underground ventilation networks including variable
density, compressible airflow, shock losses and full thermodynamic modelling as required. To
some extent they can also not only model steady-state conditions but even transient condi­
tions. Any errors in current ventilation models are due to errors in the physical inputs of the
airways and leakage resistances etc., rather than due to assumptions or approximations in the
software itself, as was previously the case.
Secondly, the ready availability of inexpensive high accuracy barometers. In many respects,
the traditional method of undertaking ventilation surveys in a mine is analogous to the trad­
itional method of undertaking a topographical survey of surface land: in historical land survey­
ing practices, surveyors would use levels or plane tables and theodolites to comprehensively
measure the three dimensions of many surface points and create contour maps from these
measurements. However, with the advent of, firstly, accurate aerial mapping and more recently
differential GPS, these older methods of land surveying have been mostly superseded. It is very
possible today to use aerial photographs to develop high resolution contour maps with only
the occasional spot heights (or spot elevations) surveyed by humans, if at all. In a similar way,
high accuracy barometers have made similar advances possible with ventilation surveys as will
be discussed in the next section.
Thirdly, the increasing trend to instrumenting the ventilation system in a mine, often part of
a more comprehensive Ventilation-on-Demand system. The use of fibre-optic communication
backbones has also made real-time Big Data type of analyses now possible. Instrumentation,
VOD and Big Data will continue to decrease the number of manual measurements needing to be
taken, assuming suitable software becomes available to process this data into useful information.
Fourthly, the development of so-called Artificial Intelligence is beginning to have a major
impact on assisting with solving many complex problems in society and industry ranging from
medical diagnosis to setting up spreadsheet formulas. AI is likely to have an equally profound
impact on the profession of mine ventilation over the decade to come.

591
5 BAROMETRIC PRESSURE SURVEYS

The theory and techniques for barometric (or altimeter) pressure (BP) surveys are well
described by McPherson (2007). BP surveys are not new and date back to around the 1920s
(McElroy & Kingery, 1957; Harris et al, 1973). However, it is the ready available and low cost
of high accuracy barometers (Derrington, 2015) that has made the technique a far more
attractive option than previously.
A barometric pressure survey can be used in at least two ways:
• For airways which have a significant frictional pressure loss along their length, a BP survey
can be used to directly measure frictional pressure loss, resistance and friction factor for
that airway
• For validation of a ventilation model, effectively checking the integrity of the model against
measured “spot heights” (in the pressure sense).
In the latter case, a number of nodes are selected throughout the mine and the barometric
pressure (and wet and dry bulb temperatures) are accurately measured at those locations. If
the airflows in the mine are known (and these can be measured easily and accurately), and the
elevation of the BP positions is accurately known (to within about 0.5 m, which can now usu­
ally just be done simply and easily using the survey floor strings in the as-built survey data­
base), then the pressure loss between surface and those node points can be accurately
measured and compared to the predicted values from the ventilation model. In addition to
correlating the model by comparing predicted and actual BP values, it is good practice to also
measure the differential pressure (by manometer) across key ventilation controls, which can
also then be checked against the modelled values of differential pressure (DP) at these loca­
tions. Ventilation controls are also inspected to estimate leakage (m3/s) so that this can be
accurately modelled. Measuring DP and flow across ventilation controls (effectively a gauge
and tube survey of a single “airway”) remains very useful in terms of measuring the leakage
resistances of different types and constructions of ventilation controls (e.g. doors, air cross­
ings). Any discrepancies between measured and modelled values of BP, DP and leakage can
be investigated. Once any differences are within a reasonable tolerance, the model can be con­
sidered to be validated and fit for purpose.
The main advantages of barometric pressure surveys are:
• They are quick and inexpensive (compared to gauge and tube)
• There are no cumulative errors as each measurement stands alone as a fixed reference point
for validation in the model
• Any changes in the resistance of old airways or significant new sources of leakage are
detected because each BP survey is, in effect, a full new mine survey.
• Very importantly, if the as built ventilation model has failed to include some new piece of
development somewhere, or some important development has not been removed from the
model (i.e. these changes have inadvertently not been included in the ventilation model)
then the BP survey will detect this, but an incremental gauge and tube survey of only the
new development in a mine will not.
• A BP survey can be conducted by one person, assuming the base (or control) station tech­
nique is used, with the surface barometer at a suitable location in data-recording mode
• A BP survey can be conducted with the mine operating as normal with generally little loss
of accuracy.
• A BP survey results in negligible disruption to mine operations.
Barometric pressure surveys are particularly useful for very large operations or those with
many airways that are difficult or impossible to traverse. For example, the three integrated
Mount Isa underground mines circulated 6000 m3/s of air through over 600 km of ventilated
airways over a strike length of 5 km, from surface to 1900 metres below surface and with
levels every 40 m (every 20 m in the upper areas).

592
6 THE IMPACT OF VENTILATION SYSTEM INSTRUMENTATION,
VENTILATION-ON-DEMAND AND BIG DATA

The increasing range of responsibilities of the mine ventilation engineer has already put great
pressure on his or her ability to collect manual measurements from underground. Even where
these are collected, sometimes for statutory purposes, they are often not used for any useful
subsequent analysis. Most new mines are now being set up with a fibre optic communications
backbone and larger mines are implementing an increasing degree of real-time instrumenta­
tion and monitoring, and in some cases, adopting a form of remote and semi-autonomous
control of the ventilation system (VOD). The ability to collect and usefully manage large sets
of data (Big Data) will be an essential precursor to the development and refinement of Artifi­
cial Intelligence systems in mine ventilation engineering.

7 THE IMPACT OF ARTIFICIAL INTELLIGENCE ON MINE VENTILATION

ANALYSIS INCLUDING VENTILATION MODEL VALIDATION

Artificial Intelligence (AI) has been defined (Copeland 2023) as “the ability of a digital computer
or computer-controlled robot to perform tasks commonly associated with intelligent beings. . .
developing systems endowed with the intellectual processes characteristic of humans, such as the
ability to reason, discover meaning, generalize, or learn from past experience”. Examples would
include some robotics, self-driving cars, chatbots, virtual travel agents and many others.
So how could AI assist with ventilation decision-making and model validation in the future?
With respect to ventilation model validation, an interesting paper by Griffith and Stewart
(2021) effectively described using a limited number of gauge and tube survey data to improve the
calibration of a mine ventilation computer model by setting up a sensitivity matrix of actual
versus measured values to modify the resistance of the most probable airways that need resist­
ance adjustment. It is likely that this approach could be modified to do one of the following:
• Use the same limited (or nil) P-Q data, along with accurate barometric pressure data and
differential pressure measurements, and with the option for the user to set or relax friction
factors and airway sizes and other parameters, to develop the most reliable true estimate of
the ventilation model of the mine, or
• Assess the model and provide a list of ventilation controls or fan locations and the type of
measurements needing to be undertaken at each location (e.g. pressure difference across
a ventilation control, pressure and flow through a fan, barometric pressure difference to
surface), to provide a statistically-valid confidence level for the model. For example, the AI
system could determine that if certain specific measurements are taken, it would then be in
a position to use these measurements in the model to determine how much uncertainty is
left in the reliability of the overall model, or perhaps even provide advice about the uncer­
tainty regarding a particular question from the ventilation engineer such as this: “Given the
existing ventilation model and the results of the last flow and pressure survey and the accur­
acy of these measurements, what additional measurements would need to taken to be 95%
confident that a predicted fan pressure and flow at (for a new fan duty at some location)
would be within 3% on both flow and pressure”.
• In fact, all five questions posed in the dot points in section 2 of this paper could be candi­
dates to put to an AI system in the future.
Artificial Intelligence will certainly impact on the role of measurements and data collection
and analysis more generally for the mine ventilation professional in the future. However, this
author’s view (Brake, 2022) is that the impacts will mostly be positive. AI could provide statis­
tically sound advice answering many of the questions that come up frequently (or less fre­
quently) for the ventilation engineer allowing for faster, more reliable and more timely
decisions. In addition, AI could remove the need to take many of the often-tedious and time-
consuming measurements that perhaps add no real value to the mine ventilation monitoring

593
program whilst simultaneously improving the reliability of the analysis, freeing up the ventila­
tion professional for more important parts of the job.

8 CONCLUSION

Given the increasing demands on the ventilation engineer’s time, the shrinking size of the ven­
tilation department on most mine sites, and the 24/7 nature of mine operations, barometric
pressure surveys, when combined with volume surveys, are increasingly the most practical
method of validating ventilation models. This very useful survey technique should be part of
the repertoire of every practicing mine ventilation engineer and its specific requirements (e.g.
timing) should be included in the mine’s Ventilation Management (Control) Plan.
With regard to ventilation measurements more generally, it is likely that Big Data analysis
and Artificial Intelligence will play an increasingly useful role in many technical related mat­
ters, including those in mine ventilation, over the next decade. AI is likely to develop the point
where it can provide independent and statistically sound advice regarding the types and loca­
tions of measurements to be taken underground to achieve a statistically reliable estimate of
the overall validity of a ventilation model or to rapidly answer specific questions regarding
fault-finding or diagnosis of the ventilation system including under upset conditions such as
power failures, fan failures, fires or explosions.

REFERENCES

Brake D J, 2015. Quality Assurance Standards for Mine Ventilation Models and Ventilation Planning,
Proc 3rd Aust Mine Vent conf (Belle B (ed)). AusIMM. pp. 221–228.
Brake D J, 2022. Developing and maintaining mine site ventilation capability and new trends in mine
ventilation career paths. The 6th Australian Mine Ventilation Conference (Belle, B (ed). AusIMM.
Gold Coast, 10-12 Oct 2022. pp 181–204.
Copeland, R J, 2023. Artificial Intelligence. Artificial intelligence | Definition, Examples, Types, Applica­
tions, Companies, & Facts | Britannica. Encyclopaedia Britanica.
Derrington, A.S. 2015. Development of a Low-cost Instrument for Barometric Pressure Surveys. The
Australian Mine Ventilation Conference. AusIMM. Chalmers, D (ed). Sydney, NSW. AusIMM. 31
Aug-2 Sep 2015.
Griffith M D & Stewart C M, 2021. Ventilation model calibration with limited survey data. Proc 18th
NAMVS, Tukkaraja, P (ed). 12–17 Jun 2021. Online. Taylor & Francis. pp 509–517.
Harris, E.J., Dalzell, R.W, Kline, R.J & Miller, E.J. 1973. A Method for Calculating Mine Ventilation
Pressure Losses Using Computers and Desk-Top Calculators, A Supplement to Information Circular
7809. Information Circular 8594. US Bureau of Mines.
McElroy, G.E & Kingery, D.S. 1957. Making ventilation-pressure surveys with altimeters, Information
Circular 7809. US Bureau of Mines.
McPherson, M.J. 2007. Subsurface Ventilation Engineering. http://www.mvsengineering.com/ 835 pp.
Rowland, J.R. 2011. Ventilation Surveys and Model Building–A Consultant’s Opinion. Australian Mine
Ventilation Conference/Sydney, NSW, 5-6 September 2011.

594
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Improving the accuracy of field airflow measurements for tunnel

ventilation fans

R.E. Ray, Jr. & E. Fuster

WSP USA Inc., New York, NY, USA

ABSTRACT: Accurately measuring the airflow quantity delivered by tunnel ventilation fans
during site acceptance testing can be difficult due to the fan configuration and constrained
sizes of adjoining plenums. Most axial flow tunnel ventilation fans are reversible with the
drive motor attached directly to the fan impeller. Obstructions created by motor supports pre­
clude performing pitot tube traverses in the fan section and the presence of transition duct­
work and sound attenuators on each side of the fan dictate that airflow measurements be
taken at the end of the fan equipment train. Point air velocity measurements taken with an
anemometer at the fan room and plenum interface are averaged to calculate the fan airflow
quantity. Limited floor space in ventilation buildings may result in narrow plenums adjacent
to fan rooms that produce non-uniform flow with high velocities concentrated on one side of
the fan ductwork, making the accuracy of anemometer point traverses problematic. This
paper presents the results of a study of alternative grid spacing of point anemometer traverses
to yield more accurate field measurements using CFD modeling of a sample fan room and
plenum arrangement.

1 INTRODUCTION

The NFPA (National Fire Protection Association) 130 Standard for Fixed Guideway Transit
and Passenger Rail Systems (2020) requires passenger rail and transit (subway) tunnels and
stations to be equipped with emergency ventilation systems that provide tenable egress condi­
tions for passengers evacuating from a train fire incident. To meet the standard, a series of
shafts are typically spaced along the length of the tunnels to facilitate a “push-pull” ventilation
system utilizing reversible axial flow fans. For a fire emergency in the tunnel, passengers
would walk towards the shaft supplying air, while smoke is exhausted from a different shaft in
the opposite direction. Passenger rail stations encompassing multiple tracks with much larger
enclosed space than tunnels may be equipped with an exhaust emergency ventilation system to
maintain smoke layers above the walking height of passengers. Station platform ventilation
fans are also usually reversible to allow them to work in tandem with adjoining tunnel ventila­
tion shafts if required.
As noted by Ray and Gamble (2010), since ventilation systems serve a critical life safety
function for passenger rail and transit tunnels used by the public, factory acceptance testing of
tunnel ventilation fans is required for most projects to confirm that the fans deliver the design
duty airflow. Factory testing is conducted in accordance with procedures described in AMCA
210-07, “Laboratory Methods of Testing Fans for Certified Aerodynamic Performance
Rating,” with pitot tube traverses performed across a measurement plane located a minimum
of 8.5 test duct diameters away from the fan with essentially a uniform air velocity profile
across the duct.
In addition to factory acceptance tests, site acceptance testing is also performed to verify
that tunnel ventilation fans deliver the design duty airflow quantity in the field. As shown in

DOI: 10.1201/9781003429241-61

595
Figure 1, an elevation view of a typical reversible axial fan equipment train, round-to-
rectangular transition ducts are attached to each side of the fan with sound attenuators placed
downstream of the transitions. Measuring velocity and total pressures with a pitot tube in the
fan section is not feasible since the drive motors are coupled directly to the fan impeller and
motor support structures create disturbances to airflow and obstruct insertion of the tube.
The increase in cross-sectional area through the transition ducts creates angular airflow that
also precludes using a pitot tube since the bent tip section of the pitot tube needs to be aligned
with the flow. As a result, measurement of fan total pressure in the field is not feasible. Since
parallel full height baffles in the sound attenuators create channeled airflow with no access for
measurement instrumentation, suitable locations for measuring fan airflow quantity are gener­
ally limited to the ends of the fan equipment train.

Figure 1. Typical reversible tunnel ventilation fan configuration.

2 MEASURING FAN AIRFLOW QUANTITY AT THE PLENUM INTERFACE

Air velocity traverses taken across the end of the fan equipment train in an adjoining air plenum
can be performed by direct measurement of velocity using vane anemometers or by measuring
velocity pressure with a pitot tube and calculating the velocity. Due to the large size of the duct­
work, pitot tube traverses are cumbersome to perform, requiring custom tubes long enough to
reach the center of the ducts and sufficiently stiff to avoid deflection with the fan operating.
Digital anemometers with remote readouts and attached to extension rods are easier to use and
can allow testing personnel to remain to the sides of the duct for all measurements.
Field testing specifications typically require anemometer traverses to be performed twice
across the measurement plane and averaged. Conducting continuous traverses at the plenum
interface without personnel standing in front of a portion of the duct would be difficult, and
repeatability of successive traverses would be problematic since testing contractors do not
have experience with this method. Instead, point air velocity measurements taken at the end
of the fan equipment train are averaged and multiplied by the duct cross-sectional area to esti­
mate the fan airflow quantity. As noted by Jorgensen (1999), the most frequently used method
to locate traverse stations is to divide the duct into equal areas and measure at the centroid of
the subdivided areas. In determining the number of point measurements comprising the tra­
verse grid, practical considerations need to be weighed against selecting a sufficient number of
locations to establish the airflow distribution exiting (or entering) the fan ductwork. A test
grid with too few velocity measurements will not accurately reflect the airflow quantity
delivered by fan, while a grid with a large number of measurement points may take an exces­
sive amount of time to complete. For a facility with multiple fans operating in parallel, the
time required to complete two traverses for each fan in both exhaust and supply operation can
be substantial. McElroy (1935) recommended not less than 16 traverse points for square cross

596
sections and not less than 24 for rectangular cross sections where one dimension is twice the
other for “precise traversing methods” in major underground air currents or tests of primary
fan performance. He further noted that a minimum spacing of 0.304 m (1 ft) between grid
points to be satisfactory in mine airways, and that if closer spacing appears to be required, “it
is very probable that the conditions of turbulence would prevent a high degree of accuracy by
any method of measurement”.

Figure 2. Air velocity traverse being performed with digital anemometer.

3 SAMPLE FIELD TEST DATA

Fan airflow quantity and power measurements taken during site acceptance testing recently
completed for a passenger rail station are summarized in Table 1. Point anemometer traverses
were performed at the end of the fan equipment trains for both the exhaust and supply airflow
directions for eight reversible axial fans—four, 2.59-m (102-in.)-dia., 184 kW (250 hp) fans
and four, 2.16-m (85-in.)-dia., 110 kW (150 hp) fans. Velocity measurements were taken with
a digital vane anemometer equipped with a remote readout at a total of 36 points at the end
of the 3.66-m (12-ft) by 3.66-m (12 ft) fan ductwork for the 2.59-m (102-in.)-dia. fans and at
the end of the 2.44-m (8-ft)-wide by 2.74-m (9-ft)-high ductwork for the 2.16-m (85-in.)-dia.
fans. The measurement points were located at the center of 36, 0.61-m (24-in.) by 0.61-m (24-
in.) subdivided square areas in the larger ducts. This spacing placed measurement points
0.31 m (12 in.) from the duct walls and 0.61 m (24 in.) apart in between. For the smaller ducts,
measurement points were located at the center of six-wide by six-high, 0.41-m (16-in.) by 0.46-
m (18-in.) rectangles. This grid spacing placed measurement points 0.31 m (8 in.) from the
duct side walls and 0.41 cm (16 in.) apart across the duct, and 0.23 m (9 in.) from the top and

597
bottom duct walls and 0.46 cm (18 in.) vertically apart. The grid measurement locations were
marked on the screens installed across the ends of the ductwork with electrician’s tape. The
anemometer was secured to an adjustable painter’s extension pole to allow testing personnel
to remain to the side of the ductwork while taking measurements, as shown in the photograph
provided in Figure 2. A ladder was needed to help reach some measurement stations in the
top center of the larger duct. At each measurement station, the anemometer was set up to
read continuously over a 10-second interval and the high and low velocities were averaged.
A three-phase power quality analyzer wired directly to the fan motor terminals was used to
directly measure current, voltage and power factor and calculate power.
In addition to listing the field test airflow quantities and power for each fan in both direc­
tions of flow, Table 1 also displays the fan airflow quantity derived from plotting field-
measured power (corrected to the project air density) on the factory test airflow quantity vs.
bhp curves. This is considered the most accurate method of determining fan airflow in the
field for tunnel ventilation projects. The field-measured airflow quantities ranged from
approximately 112% to 132% of the airflow quantities determined from the bhp factory curves
for the 2.59-m (102-in.)-dia. fans in exhaust and 114% to 131% in supply. The field-measured
airflow quantities for the 2.16-m (85-in.)-dia. fans varied from approximately 99% to 109% of
the airflow quantities determined from the bhp factory curves in exhaust and 99% to 115% in
supply. It should be noted that the repeatability of the two traverses for each fan test was
excellent, with the differences in average air velocity between successive traverses falling in
a range of 0.1% to 3.1%. This indicates that the variations seen between the field-measured
airflow quantities and the airflow quantities determined from the factory airflow quantity vs.
bhp curves was not due to inconsistent handling of the anemometer and extension pole setup
or failing to measure the same grid point locations in successive traverses. It is also assumed
that the installed fan arrangement does not degrade the factory performance curves since the
field airflow measurements exceed those predicted from the factory curves and AMCA 201-02
(R2011) addresses “system effects” causing non-uniform flow conditions at the fan inlet or
outlet by increasing the fan system resistance, not de-rating the fan.
An additional velocity traverse was subsequently performed for SVF-4 operating in
exhaust with an increase from 36 to 64 measurement points to evaluate whether the accuracy

Table 1. Site acceptance testing airflow measurements for station ventilation fans.
EXHAUST TVF-1 TVF-2 TVF-3 TVF-4

kW (bhp) 138 (185) 144 (193) 131 (175) 156 (209)

Q measured: m3/sec (cfm) 145.7 (310,000) 148.1 (315,000) 143.4 (305,000) 152.8 (325,000)
Q bhp curve: m3/sec (cfm) 124.1 (264,000) 120.3 (256,000) 127.8 (272,000) 115.6 (246,000)
Measured Q/bhp Q 117.4% 123.0% 112.1% 132.1%
SUPPLY TVF-1 TVF-2 TVF-3 TVF-4
kW (bhp) 138 (185) 134 (180) 129 (173) 145 (194)
Q measured: m3/sec (cfm) 136.8 (291,000) 154.6 (329,000) 146.6 (312,000) 154.2 (328,000)
Q bhp curve: m3/sec (cfm) 119.9 (255,000) 122.2 (260,000) 127.8 (272,000) 117.5 (250,000)
Measured Q/bhp Q 114.1% 126.5% 114.7% 131.2%
EXHAUST TVF-5 TVF-6 TVF-7 TVF-8
kW (bhp) 62 (83) 70 (94) 69 (92) 66 (88)
Q measured: m3/sec (cfm) 70.0 (149,000) 72.9 (155,000) 72.4 (154,000) 70.5 (150,000)
Q bhp curve: m3/sec (cfm) 71.0 (151,000) 66.7 (142,000) 67.7 (144,000) 70.0 (149,000)
Measured Q/bhp Q 98.7% 109.2% 106.9% 100.7%
SUPPLY TVF-5 TVF-6 TVF-7 TVF-8
kW (bhp) 66 (89) 61 (82) 64 (86) 68 (91)
Q measured: m3/sec (cfm) 64.4 (137,000) 76.1 (162,000) 76.6 (163,000) 70.0 (149,000)
Q bhp curve: m3/sec (cfm) 64.9 (138,000) 67.7 (144,000) 66.7 (142,000) 64.4 (137,000)
Measured Q/bhp Q 99.3% 112.5% 114.8% 108.8%

598
of a 36-point traverse was significantly impacted by the spacing of velocity measurement
points. For the revised test, measurement points were placed in the center of 64, 0.41-m by
0.41-m (16-in. by 16-in.) subdivided duct sections. These points were within 0.20 m (8 in.) of
the duct walls and 0.41 m (16 in.) apart, compared to 0.31 m (12 in.) from the duct walls and
0.61 m (24 in.) apart for the original test traverses. Despite a 78% increase in velocity meas­
urement locations, the airflow quantity measured in exhaust was 151.8 m3/s 323,000 cfm,
only 1% less than the original measurement and still 131% of the airflow quantity predicted
from power measurements.
Velocity contours across the end of the fan equipment train developed from the field test
traverse measurements are provided in Figure 3 for TVF-4 operating in supply. Field airflow
quantity measurements for TVF-4 varied the most from those predicted from power measure­
ments (132.1% exhaust/131.2% supply) while field airflow quantity measurements for TVF-5
were closest to those predicted from power measurements (98.7% exhaust/99.3% supply). The
simplest explanation for these differences is that measurement points were spaced closer
together and closer to the duct walls for the smaller ductwork connected to TVF-5 versus the
larger ductwork connected to TVF-4. However, the velocity contours displayed in Figure 4
for TVF-4 supply operation indicate that the airflow is more concentrated in the upper center
to lower right side of the duct with velocities exceeding 13.2 m/s (2,600 fpm) in the upper
center. The overall average measured duct velocity was 8.8 m/s (1,736 fpm). The uneven flow
distribution may make it more difficult to accurately measure the average velocity for SVF-4
in supply.

Figure 3. Air velocity contours in fpm for SVF-4 supply traverse.

4 CFD AIRFLOW MODELING OF SAMPLE FAN ROOM

Based on the results of the 64-point traverse for TVF-4 operating in exhaust, field airflow
measurements were not repeated for TVF-4 or any other fans with additional grid points.
Instead, it was decided to study alternative velocity traverse grid spacings using computational
fluid dynamics (CFD) analysis to evaluate the number of points that would provide the closest
measurement to the actual airflow rate of a highly skewed flow distribution. The primary goal

599
of this analysis was to use CFD as a tool to easily compute airflow quantity from skewed vel­
ocity contours using different measurement patterns, not to simulate the internal airflow
through the fan equipment train with a high degree of accuracy. A three-dimensional model of
the TVF-4 fan room and adjoining plenums originally developed during construction using
AutoCAD was updated to include more detail of the fan equipment train. As illustrated in
Figure 4, representations of the fan dampers, the full height baffles installed in the sound attenu­
ators located on each side of the fan and the fan drive motor and foot-mounted motor pedestal
were added to the fan equipment train portion of the CFD model. The only missing features in
the geometry compared to the physical installation were the fan impeller and wire mesh screens.
Transient simulations of TVF-4 operating in supply were performed using ANSYS Fluent soft­
ware. The cases utilized the realizable k-ϵ turbulence model and the grid size was approximately
10 million polyhedral cells. The TVF-4 supply airflow quantity was set to approximately
117.5 m3/sec (250,000 cfm) to match the quantity determined from the factory performance
curves using the field-measured power. The upstream inlet and downstream outlet were modeled
as pressure boundaries, and to achieve the desired non-uniform flow at the plane of measure­
ment the Fluent “fan” internal boundary condition utilized was set to swirl the flow to produce
a similar velocity pattern to that depicted in Figure 3. An acceptably skewed pattern was devel­
oped after iterating the fan boundary settings. To determine if the solution was grid independ­
ent, two sensitivity cases were run, and the results were nearly identical.
Air velocities were measured for four different traverse grids across the end of the ductwork
adjacent to the exhaust plenum in the CFD model: (1) 36 points, (2) 49 points, (3) 81 points
and (4) 144 points. The 36-point grid was identical to that used in the field testing, while the
81-point and 144-point grids placed the air velocity measurement points at the center of 9 by 9
and 12 by 12 equal subdivided areas of the overall 3.66-m (12-ft) by 3.66-m (12-ft) duct,
respectively. Traverse points for the 81-point grid were within 0.20 m (8 in.) of the duct walls
and 0.41 m (16 in.) apart, compared to 0.15 m (6 in.) from the duct walls and 0.31 m (12 in.)
apart for the 144-point grid. The 49-point traverse was based on the log-Tchebycheff method
that utilizes a logarithmic distribution adjacent to the duct walls and a polynomial distribution
of velocity for the interior points. Spacing of the 49-point traverse grid points is depicted in
Figure 5. As noted by Jorgensen (1999), all points are also equally weighted in this method.
The velocity traverse points were represented as 6.99-cm (2.75-in.)-dia. circles in the CFD
model that mimic the size of the digital anemometer. Area-weighted averages of velocity for
each circle were averaged across the duct cross section to determine the airflow quantity meas­
ured by each traverse grid configuration.

Figure 4. Three-dimensional CFD model of TVF-4 fan equipment train and plenums.

600
Figure 5. Log-Tchebycheff traverse grid for 3.66-m (12-ft) by 3.66-m (12 ft) cross section.

5 ALTERNATE TRAVERSE GRID SPACING RESULTS

Velocity contours at the measurement plane at the end of the fan equipment train adjacent to
the exhaust plenum for the CFD simulation of TVF-4 operating in supply with the 36-point
traverse grid are shown below in Figure 6. As noted previously, the Fluent “fan” internal
boundary condition utilized in the CFD model was set to swirl the flow to attempt to skew the
higher air velocities in the 11.2 m/s (2,200 fpm) to 14.2 m/s (2,800 fpm) range to the upper
center and lower right portions of the cross section as was measured in the field. Table 2 pro­
vides a summary comparison of the actual airflow quantity set in the CFD model (as close as
possible to the quantity derived from the power measurement) and the airflow quantity com­
puted from the traverse of the CFD simulation cross section at the end of the fan equipment
train for each of the four alternative grid spacings. An examination of the airflow data shown
in Table 2 confirms that the CFD simulations predict that the supply airflow quantity
delivered by TVF-4 would still be overestimated by using any of the four alternate grid spa­
cings. The simulation of the “base case” of 36 points was somewhat closer to the assumed air­
flow quantity vs. the 36 points measured in the field (124.1% of assumed flow vs. 131.2%).
Using the centroid of subdivided equal areas, increasing the number of measurement points
from 36 to 81 would reduce the “overmeasurement” of the TVF-4 supply airflow from 124.1%
to 117.2% of the assumed airflow. Increasing to 144 measurement points would further reduce
the “overmeasurement” of airflow to 114.9% of the assumed airflow. Using the log-
Techebycheff grid spacing of 49 points results in a predicted TVF-4 supply airflow quantity of
120.3% of the assumed airflow. It would appear that using the log-Techebycheff method to
space measurement locations does not provide advantages over using centroids of subdivided
equal areas since the simulation data shows that its overestimated airflow of 120.3% of the
assumed airflow for 49 measurement points falls in between the 36 and 81-point equal area
centroid spacing overmeasurements. A comparison of the data contained in Tables 1 and 2
indicates that the field measurements for the smaller 2.16-m (85-in.)-dia. fans (TVFs-5, 6, 7
and 8) are all closer to the assumed airflows (98.7% to 114.8%) than the simulated measure­
ments for similar or even closer traverse point spacing for the 2.59-m (102-in.)-dia. fans.

601
Figure 6. Air velocity contours at inlet end of ductwork for CFD simulation of TVF-4 operating in
supply.

Table 2. TVF-4 actual CFD model supply airflow quantity vs. traverse CFD airflow quantity.
Actual CFD Q Traverse CFD Q
Traverse Grid m3/sec (cfm) m3/sec (cfm) Traverse CFD Q/ actual CFD Q

36 Points 119.2 (253,000) 147.9 (314,000) 124.1%

81 Points 120.6 (256,000) 141.3 (300,000) 117.2%
144 Points 120.1 (255,000) 138.0 (293,000) 114.9%
49 Points (Log-Tchebycheff) 118.2 (251,000) 142.2 (302,000) 120.3%

6 CONCLUSIONS

Obtaining accurate measurements of the airflow quantity delivered by tunnel ventilation fans
during site acceptance testing is challenging for a number of reasons, including skewed airflow
distribution through the fan caused by narrow adjoining plenums and the difficulty locating
measurement cross sections sufficiently distant from obstructions and transitions in the duct­
work. Recent field test data presented for a station with eight reversible tunnel ventilation
fans showed airflow quantities measured with anemometer traverses overestimating the fan
airflow quantity determined from field power measurements applied to factory test curves by
12 to 32 percent for 250,000 cfm fans and as much as 15 percent for 125,000 cfm fans. CFD
modeling of the fan with the highest overmeasurement in the field showed a reduction in over­
measurement from 24% to 15% by increasing the number of measurement points from 36 to

602
144 across 3.66-m (12 ft) by 3.66-m (12 ft) fan ductwork. Two complete traverses over the
cross section being measured are required for field testing of tunnel ventilation systems, how­
ever, and performing two, 144-point traverses for each direction of flow for a large number of
fans would be time prohibitive. In the case of the 117.5 m3/sec (250,000 cfm) fans, increasing
the number of velocity measurement points from 36 to 81 points would be more practical and
result in more accurate results, although still 17 percent high based on the results of the CFD
analysis described above.
Since nearly all of the fan airflow quantities measured in the field for the sample station
exceeded the derived values using power measurements, it is clear that the lower velocities adja­
cent to the duct walls are underrepresented in the traverses. Based on the CFD analysis, this
would also apply to the traverse with the largest number of measurement points (144), with
points 0.15 m (6 in.) from the duct walls. Because the CFD simulations also predict overmeasure­
ment of average air velocities for the unequal airflow distribution entering the fan equipment
train, it is unlikely that anemometer error is contributing to the overmeasurement of the fan
airflow.

REFERENCES

AMCA. 2011. Publication 201-02 (R2011), Fans and Systems. Arlington Heights, IL: Air Movement and
Control Association International, Inc.
ANSI/AMCA. 2008. Standard 2010-07 (ANSI/ASHRAE Standard 51-07), Laboratory Methods of Test­
ing Fans for Certified Aerodynamic Performance Rating. Arlington Heights, IL: Air Movement and
Control Association International, Inc.
Jorgensen, R. (ed). 1999. Fluid flow. In Fan Engineering, Chapter 2: 2–90. Kingsport, TN: Quebecor
World Book Services.
Jorgensen, R. (ed). 1999. Fluid flow. In Fan Engineering, Chapter 2: 2–94. Kingsport, TN: Quebecor
World Book Services.
McElroy, G.E. 1935. Engineering Factors in the Ventilation of Metal Mines: 12–13. Washington, D.C.;
U.S. Department of the Interior, Bureau of Mines.
National Fire Protection Association. 202. NFPA 130: Standard for Fixed Guideway Transit and Passen­
ger Rail Systems: 130-22-130-24. Quincy, MA.
Ray, R. E. & Gamble, G.A. 2010. The benefits of conducting factory performance tests for main mine
fans. In S. Hardcastle & D.L. McKinnon (eds.), Proceedings of the 13th U.S./North American Mine
Ventilation Symposium: 529–539. Sudbury, Ontario: MIRARCO – Mining Innovation.

603
Ventilation network analysis and optimization
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Case study on the abnormal airflow diagnosis method using

atmospheric monitoring data

L. Zhou, D. Bahrami & R.A. Thomas

Pittsburgh Mining Research Division
National Institute for Occupational Safety and Health (NIOSH), Pittsburgh, USA

ABSTRACT: A stable and well-maintained mine ventilation system is the key to ensuring
a safe and healthy working environment for miners. A sudden, unplanned, and significant
change in airflow termed as abnormal airflow is frequently observed in mine ventilation.
Some abnormal airflows can return to normal without manual intervention; however, some
abnormal airflows may cause catastrophic accidents if left unattended. In addition, abnormal
airflow may be a consequence of an accident such as a blocked airflow route due to a roof
fall. Promptly diagnosing and locating the cause of abnormal airflow can help prevent acci­
dents. Researchers at the National Institute for Occupational Safety and Health (NIOSH)
have developed a method to diagnose the cause of abnormal airflow for underground mine
ventilation systems. The purpose of this paper is to verify the developed method using experi­
mental tests conducted at NIOSH’s experimental mine. The airflows were monitored by
a real-time atmospheric monitoring system installed in the experimental mine during the tests.
As demonstrated in this case study, the developed abnormal airflow diagnosing method,
based on the resistance sensitivity and matching method, has been proven to be reliable.

Keywords: Mine ventilation, Network modeling, Atmospheric monitoring system, Abnormal

airflow

1 INTRODUCTION

Ventilation is the lifeline of underground mines to provide the underground workings with fresh
air in sufficient quantity and quality, dilute and remove dust and noxious gases, and regulate tem­
perature and other conditions. A stable and well-maintained mine ventilation system is the key to
ensuring a safe and healthy working environment for miners. The efficiency of mine ventilation
highly relies on airflow distributions and airflow quantities in the mine ventilation system. The
airflow distributions are closely monitored and controlled during the life of an active mine.
Changes in airflow quantity or direction are not rare in the routine management of ventilation.
The airflow changes can be categorized as: 1) planned changes for the purpose of maintaining a
safe and healthy working environment; 2) periodic changes associated with atmospheric condi­
tions such as barometric pressure, temperature, humidity, and so on; 3) expected temporary
changes from the interference of moving equipment; 4) sudden, unplanned, and significant
changes due to some hidden hazardous conditions or malfunction of ventilation fans. The fourth
category of airflow changes can be termed abnormal airflow if the changes in quantity exceed
a certain level. The abnormal airflow can be caused by some kinds of failures in ventilation. For
instance, unattended opening/closing of a ventilation door, airpath blockage due to roof collapse,
fan stoppage, or any other unforeseen hazardous condition. The abnormal airflows may cause
catastrophic accidents if left unattended. For example, insufficient air supplies resulting from
abnormal airflow may induce elevated levels of hazardous gas accumulation. Therefore, promptly

DOI: 10.1201/9781003429241-62

607
diagnosing and locating the cause of abnormal airflow can help prevent accidents. Researchers at
the National Institute for Occupational Safety and Health (NIOSH) have developed a method,
named as abnormal airflow diagnosis method, to assist the diagnosis of the cause of abnormal
airflows for underground mine ventilation systems (Bahrami & Zhou, 2022). Bahrami & Zhou
(2022) also verified the abnormal airflow diagnosis method using the cases generated with ventila­
tion network modeling. The method was found to be reasonably accurate in locating an unknown
source airway caused by a flow resistance change in a ventilation network. However, the verifica­
tion was based on model-based numerical data which tends to provide “perfect” cases without
involving any interferences from the real world. There is a need to test the proposed method using
real-world scenarios. For this reason, we have conducted a ventilation test at NIOSH’s experimen­
tal mine by changing airflow distribution while monitoring of the Atmospheric Monitoring
System (AMS) and testing the abnormal airflow diagnosis method using the experimental data. In
this paper, while we refer to our earlier work (Bahrami & Zhou, 2022), the focus is to experimen­
tally verify and demonstrate how to diagnose an abnormal airflow using the proposed method
with the monitored AMS data.

2 METHODOLOGY

The abnormal airflow diagnosis method referred to in this paper is based on the availability of
the sensitivity matrix for a given mine ventilation network and the matching method which
compares the monitored airflow changes with their sensitivities from the sensitivity matrix.
For a better understanding of the diagnosis procedure, a brief introduction to the sensitivity
matrix and the matching method is necessary. These have been described comprehensively by
the authors previously (Bahrami & Zhou, 2022; Zhou & Bahrami, 2022).

2.1 Sensitivity matrix

The airflow quantity distributed to each airway of a mine ventilation system depends on the
schematic connecting diagram of airways, the resistance of airways, and air pressures pro­
duced by mechanical fans or/and natural ventilation. For a given mine ventilation network,
a resistance change in an airway will affect the airflow quantities in other airways more or
less. The resistance sensitivity is frequently used to quantify airflow change in a certain airway
due to resistance change in another airway (Li et al., 2011; Dziurzynski et al., 2017; Griffith &
Stewart, 2019; Jia et al., 2020; Zhou & Bahrami, 2022). Mathematically, the resistance sensi­
tivity can be expressed as (Zhou et al., 2007; Jia et al., 2020):

Where Sij is the sensitivity of airway i to the change of resistance, ΔR, in airway j; Rj is the
resistance in airway j, and Qi is the volumetric airflow rate in the airway i. Given a ventilation
network with N airways, each airway can have N sensitivities to the rest of the airways and
itself. Therefore, an N � N sensitivity matrix S can be formed for the ventilation network
(Zhou & Bahrami, 2022):

2.2 Matching method

The sensitivity resistance matrix of a ventilation network serves as a benchmark to indicate
the interdependence of airflow and resistance of any two airways. If any unexpected airflows

608
are detected in a ventilation system, the airflow changes can be compared to their sensitivity
columns in the sensitivity matrix for each airway to find out which airway can cause a similar
airflow change pattern and has the closest airflow change quantities. Bahrami & Zhou (2022)
employed Root Mean Square Error (RMSE) to measure how well the changed airflows match
the sensitivity of each airway. For each airway in the network, a corresponding scale factor is
determined by minimizing the corresponding RMSE. It is assumed the airway with the min­
imum RMSE is the airway that has caused the abnormal airflow due to its resistance change.
For detailed information about the matching method, refer to Bahrami & Zhou (2022).

3 VENTILATION TESTS AT THE SAFETY RESEARCH COAL MINE

The Safety Research Coal Mine (SRCM) is one of the two experimental mines located at the
Bruceton Campus of NIOSH in Pittsburgh, PA. Despite the smaller size of entry (6.5 ft high
by 14 ft wide) compared to the operating mines nowadays, this room-and-pillar mine has
a complete ventilation network and has been frequently used for research regarding mine ven­
tilation, mine fires, and other studies (Zhou et al., 2017, 2018, 2019, 2020). The main fan,
installed at the surface above the return shaft, exhausts air from the mine. Stoppings, doors,
regulators, and brattices are used in the mine to direct the airflow to the desired routes. The
quantity of main airflow getting into the mine is controlled by the main fan and a door at the
main return entry. The layout of the SRCM is shown in Figure 1.
The SRCM is equipped with a mine-wide AMS to continuously monitor the atmospheric
condition of the mine. Nine air-velocity sensors are arranged throughout the mine to monitor
the real-time air velocity (as shown in Figure 1) for the ventilation tests conducted in this
study. The sensors are all mounted close to the roof in the middle of the entry. The airflows
are continuously monitored at these locations with a server on the surface to store air-velocity
readings every minute.

Figure 1. Air-velocity sensor locations within the SRCM.

609
 Within the SRCM, a door in the entry connecting B Butt and A Butt remains closed for
normal ventilation (as shown in Figure 1). To create an abnormal airflow scenario, we intention­
ally open this door and maintain the open status for approximately 17 hours for our testing. The
real-time air velocities at the nine AMS monitoring locations were recorded using air-velocity
sensors every minute. Figure 2 shows the monitored air velocities from Sensors S2, S3, S4, and
S5 during the time the door was open There was a large air-velocity drop at Sensor S2 from
around 500 ft/min to 250 ft/min when the door was opened. The air velocity returned to 500 ft/
min after the door was closed. Upon opening the door, a slight air-velocity increase can be seen
at Sensors S3 and S4. Sensors S1, S6, S7, S8, and S9 do not see significant air-velocity changes
upon opening and closing the door. Therefore, they are not plotted in the figure.

Figure 2. Monitored air velocities at Sensor S2, S3, S4, and S5.

4 SRCM VENTILATION NETWORK MODEL CONSTRUCTION AND

CALIBRATION

As it was mentioned above, the abnormal airflow diagnosis method proposed in this paper con­
sists of two major parts: one is the resistance sensitivity matrix for a ventilation network, and
the other is the matching method using RMSE. The generation of the resistance sensitivity
matrix of a ventilation network is based on one-time ventilation simulation results. Therefore,
the collaboration of the network model and the mine is critical to ensure a good application of
the abnormal airflow diagnostic method. Continuous efforts have been made to establish and
maintain a well-calibrated ventilation network model for the SRCM (Iannacchione et al., 2015;
Zhou et al., 2022). The schematic outline drawing of the SRCM ventilation network including
166 airways is displayed in Figure 3. The source airway of the abnormal airflow, where the
door is maintained open for the test, is labeled as Airway #40 in the network drawing.
Table 1 displays the comparison of the measured airflows from all nine air-velocity sensors
and simulated airflows using MFIRE, which is an open-sourced mine fire simulation program
having ventilation network simulation capability. The calculation of resistance sensitivity is

610
Figure 3. Schematic drawing of the SRCM ventilation network with airway numbers.

implemented in MFIRE as well (Zhou & Bahrami, 2022). Due to the fluctuation in the air-
velocity sensor readings, a week-long average air velocity at each sensor was used as the value
of the measured air velocity. With the known cross-sectional area for each sensor location, the
airflow rate can be obtained by multiplying the average sensor reading by the cross-sectional
area at each sensor location. However, the calculated airflow rate is based on the point reading
of the air-velocity sensor instead of the average airflow rate through the whole cross-sectional
area as the air-velocity sensor measures the air velocity where the sensor is mounted. The air-
velocity sensors used in the SRCM are typically installed at the upper section of the entry,
close to the roof or ribs, to avoid interference from the moving equipment and miners.
A correction factor for each sensor location was introduced and calculated to convert the
point airflow rate to the average airflow rate following the same procedure illustrated by Zhou
et al. (2017) and by Zhou & Bahrami (2022). The adjusted airflow rates using these correction
factors are considered the final airflow rates. The comparisons were made between the adjusted
airflow rates and the simulated airflow rates using MFIRE. The average difference for all the
sensor locations was 4.63% with the largest difference occurring at Sensor 4.

Table 1. Comparison of measured and simulated airflow at monitoring locations.

Sensor Measured Vel­ Area Airflow Correction Adjusted Air­ Simulated Air­
ID ocity (ft/min) (ft3) (kCFM) Factor flow (kCFM) flow (kCFM) Difference

1 436.69 65.83 28.75 1.01 28.46 29.56 3.71%

2 449.13 58.00 26.05 1.01 25.79 23.23 –11.03%
3 43.8 73.37 3.21 0.85 3.78 3.29 –14.91%
4 95.44 76.38 7.29 0.81 9.00 7.29 –23.45%
5 186.11 69.30 12.90 0.94 13.72 11.96 –14.73%
6 235.63 97.35 22.94 0.90 25.49 23.64 –7.82%
7 182.57 62.25 11.36 0.95 11.96 12.39 3.45%
8 191.58 62.25 11.93 1.10 10.84 9.43 –14.97%
9 479.2 70.71 33.88 1.13 29.98 30.88 2.90%
Average 4.63%

611
5 RESULTS AND DISCUSSION

By applying the derivative method implemented in MFIRE (Zhou & Bahrami, 2022), a 166-
by-166 resistance sensitivity matrix was obtained for the SRCM ventilation network. In the
sensitivity matrix, if we say the rows represent the source airways where resistance changes are
made and the columns represent the target airways whose airflow changed according to the
resistance change at each airway, an element Sij in the ith row and jth column is the rate of
airflow change in airway j responding to the resistance change in row i.
As shown in Figure 2, airflow changes occurred at Sensors S2, S3, S4, and S5 while opening
the door at Airway #40.
Table 2 summarizes the monitored airflow changes at all nine sensors and the correspond­
ing airway number (#) of each sensor in the SRCM ventilation network.

Table 2. Airflow changes at all monitoring locations.

Sensor S1 S2 S3 S4 S5 S6 S7 S8 S9

Airflow change (CFM) 0 13,167 3,413 2,173 –902 0 0 0 0

Airway # of sensor in network 2 155 81 88 77 117 118 157 19

In this case, the airflow change at Sensor S2 located in Airway #155 is obviously abnormal
due to the significant airflow reduction of 13,167 cfm. To find out the possible source airway­
(s), whose resistance change has caused similar airflow changes in all sensor airways including
Airway #155, all the monitored airflow changes, including zero changes, at all nine sensor air­
ways are set as target airflow changes. Then the columns, representing the corresponding
sensor airways, in the resistance sensitivity matrix are extracted out to be used to match the
input airflow changes. Before applying the abnormal airflow diagnosis method, each row is
screened for the sorted sensitivities to eliminate all-zero rows. An all-zero row means that the
resistance change in the airway represented by this row has no influence on all monitored air­
ways, and the calculated RMSE of the all-zero row airway will be zero as well. Figure 4 is
showing the RMSE for each non-zero sensitivity row. It can be seen that Airway #40 has the
least RMSE which indicates Airway #40 is the best matching airway to cause the airflow
changes in the monitored airways, as well as the abnormal airflow change in Airway #155. In
addition to Airway #40, Airways #145, #155, and a few others can also be the source airways
causing the airflow changes in the monitored airways.

Figure 4. RMSE of non-zero airways for the case.

612
6 CONCLUSIONS

The abnormal airflow in underground mine ventilation could be an indication of a malfunction

of the ventilation system. If left unattended, it may lead to catastrophic accidents. The cause of
the abnormal airflow needs to be diagnosed, located, and mitigated to restore a safe working
environment. An abnormal airflow diagnostic method has been developed previously by
NIOSH researchers. In this work, the diagnostic method is verified experimentally using
a ventilation test conducted at NIOSH’s experimental mine. This paper has clearly shown that
the abnormal airflow diagnostic method can diagnose the cause of abnormal airflow. In the ven­
tilation test, air-velocity sensors from an AMS were used to monitor and collect air-velocity
data. It has been demonstrated in this research that air-velocity sensors are a practical tool to
detect abnormal airflow. From the research that has been carried out, it is possible to conclude
that the abnormal airflow diagnostic method can be potentially used in practice in the future.
To ensure a successful application of the abnormal airflow diagnostic method in a mine
ventilation system, the constructed mine ventilation network model must be as close as pos­
sible to the real operating condition. The number and the selection of the monitored airflow
change inputs also plays an important role in the outcome of the proposed method. The min­
imum number of input airflow changes is three. Further research has been planned to study
the impact of the number and the selection of the input airflow changes regarding the confi­
dence level of the results from the proposed method.

DISCLAIMER

The findings and conclusions in this report are those of the authors and do not necessarily
represent the official position of the National Institute for Occupational Safety and Health,
Centers for Disease Control and Prevention. Mention of any company or product does not
constitute endorsement by NIOSH.

ACKNOWLEDGEMENT
The authors wish to thank John Soles of the Pittsburgh Mining Research Division for con­
ducting the full-scale AMS test.

REFERENCES

Bahrami, D. Zhou, L. 2022. A novel methodology to locate an abnormal airflow in underground mine
ventilation networks. Proceedings of the 2022 SME Annual Conference & Expo, Preprint 22-018, Salt
Lake City, February 27-March 2, 2022.
Dziurzynski, W. Krach, A. Palka, T. 2017. Airflow sensitivity assessment based on underground mine
ventiltion systems modeling. Energies, 10(10), 1451.
Griffith, M. & Stewart, C. 2019. Sensitivity analysis of ventilation models – where not to trust your simu­
lation. Proceedings of the 17th North American Mine Ventilation Symposium. Montreal, Canada:
The Canadian Institute of Mining, Metallurgy and Petroleum, pp. 654–660
Iannacchione A. Joseb A. Hornc, T. Iannacchione, G. Iannacchione, S. 2015. Modeling the ventilation
network at NIOSH’s Safety Research Coal Mine. Proceedings of the 15th North American Mine Ven­
tilation Symposium, June 20-24, 2015, Blacksburg, VA.
Jia, J. Jia, P. and Li, Z. 2020. Theoretical study on stability of mine ventilation network based on sensi­
tivity analysis. Energy Science & Engineering, 00, 1–8. doi:10.1002/ese3.699.
Li, G. Kocsis, C. Hardcastle, S. 2011. Sensitivity anlysis on parameter changes in underground mine ven­
tilation systems. Journal of Coal Science & Engineering, 17(3), 251–255.
Zhou, L. Luo, Y. Wu, F. 2007. Identificataion of abnormal airflow quantity in underground coal mines.
International Journal of Mineral Resources Engineering, 12(4), 257–265.
Zhou, L. Yuan, L. Thomas, R.A. Iannacchione, A. 2017. Determination of velocity correction factors
for real-time air velocity monitoring in underground mines. International Journal of Coal Science and
Technology, (2017) 4: 322–332.

613
Zhou, L. Yuan, L. Bahrami, D. Thomas, R.A. Rowland, J.H. 2018. Numerical and experimental investi­
gation of carbon monoxide spread in underground mine fires. Journal of Fire Sciences, 36(5): 406–418.
Zhou, L. Yuan, L. Bahrami, D. Thomas, R.A. Cole, G. 2019. Study on integration of real-time atmos­
pheric monitoring system monitoring data and MFIRE simulation. Proceedings of the 16th North
American Mine Ventilation Symposium, Montreal, Canada, April 28-May 1, pp. 794–803.
Zhou, L. Yuan, L. Thomas, R.A. Bahrami, D. Rowland, J.H. 2020. Determination of a mine fire inten­
sity using atmospheric monitoring system in a ventilation network. Proceedings of the 2020 SME
Annual Conference & Expo, Preprint 20-005, Phoenix, AZ, February 23–26.
Zhou, L. Thomas, R.A. Yuan, L. 2022. Experimental study of improving a mine ventilation network
model using continuously monitored airflow. Mining, Metallurgy & Exploration. 39, 887–895. https://
doi.org/10.1007/s42461-022-00574-4.
Zhou, L. & Bahrami, D. 2022. A derivative method to calculate resistance sensitivity for mine ventilation
networks. Mining, Metallurgy & Exploration. 39, 1833–1899 https://doi.org/10.1007/s42461-022-00630-z.

614
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Ventilation system upgrades at the Waste Isolation Pilot Plant

K.G. Wallace, Jr.

SRK Consulting, (U.S.), Inc., Clovis, CA, USA

I. Peña
Nuclear Waste Partnership LLC, Carlsbad, NM, USA

ABSTRACT: The Waste Isolation Pilot Plant (WIPP) facility, operated by the U.S. Depart­
ment of Energy (DOE), is the only transuranic waste repository in the United States. The
facility is designed for the permanent disposal of transuranic radioactive waste generated
through nuclear weapons production and research. On February 14th, 2014 a continuous air
monitor (CAM) alarmed indicating a radioactive contamination event had occurred in the
active emplacement panel underground. The ventilation system automatically switched to
a filtration mode where all exhaust air is sent through a bank of filters including High Effi­
ciency Particulate Air (HEPA) filters. The original filtered ventilation system had limited cap­
acity. Interim projects have included adding additional fans and filtration units to the surface
and an underground fan to enhance airflow to mining areas. The DOE has decided to main­
tain the WIPP facility for TRU waste disposal and to this end has authorized several signifi­
cant capital projects. These projects, which are currently under construction, includes a dust
and water extraction system on surface combined with one of the world’s largest HEPA filtra­
tion systems. The new Safety Significant Confinement Ventilation System (SSCVS) will be
capable of filtering up to 540,000 cfm. The dust extraction system will contain six 100,000 cfm
capable filtration units to remove salt dust and water before the air reaches the HEPA filters.
The second project is a new 28 ft diameter intake shaft 2,150 ft in depth capable of passing
500,000 cfm. New surface intake fans are designed for this shaft. An intake shaft at the site is
being reconfigured as an exhaust for the new intake fans. Both capital projects are significant
with expenditures expected to exceed $400 million. This paper describes the analyses per­
formed to size the new fans, filtration units and shaft to support work activities and to main­
tain proper flow alignment in the underground. The status of the capital projects will also be
described.

1 INTRODUCTION

In 1979, the United States Congress authorized the Waste Isolation Pilot Plant (WIPP) as
a research and development facility to demonstrate the safe disposal of transuranic radioactive
(TRU) waste from defense activities. TRU waste results chiefly from the production of nuclear
weapons from plutonium and enriched uranium. The term transuranic indicates that the waste
contains radionuclides with atomic numbers greater than 92, that is, greater than that of uranium.
TRU waste consists of a wide variety of contaminated materials from laboratory and production
operations, including discarded protective clothing, laboratory test equipment and reagents,
machine components, and solidified sludge. This waste has accumulated over the past 75 years
because of weapons development and production at U.S. defense facilities. The goal of WIPP is
to dispose legacy TRU waste from 22 generator sites nationwide.
The facility is managed by the Department of Energy (DOE) with Nuclear Waste Partnership,
LLC being the operating contractor. WIPP is the first geological repository in the nation for
which an application for permanent geological isolation has been approved. WIPP is located
approximately 35 miles southwest of Carlsbad, New Mexico (see Figure 1). The repository

DOI: 10.1201/9781003429241-63

615
horizon is located 655 m (2150 ft) below surface in the Salado salt horizon. The salt horizon was
formed about 250 million years ago and is 610 m (2,000 ft) thick. This salt horizon was selected
based on its stable geology, lack of water, high thermal conductivity and because of its “creep”
characteristics which will eventually encapsulate the waste in the emplacement rooms. The first
shipment of TRU waste was received in 1999.

Figure 1. Location of WIPP facility.

2 ORIGINAL WIPP VENTILATION SYSTEM DESCRIPTION

WIPP currently consists of four shafts and eight panels for the disposal of TRU waste. The
original ventilation system at the WIPP facility utilizes three shafts for intake air, the salt-
handling shaft (SHS), waste handling shaft (WHS) and air-intake shaft (AIS). A single
exhaust shaft (ES) is utilized for the return airway at the facility. During normal operations,
fresh air is supplied primarily by the AIS with secondary intake from the SHS. Additionally,
the SHS is used for personnel and material access as well as removal of the mined salt.
Nuclear waste is lowered through the waste handling shaft which is equipped with an enclosed
headframe. The 6.1 m (20 ft) diameter waste handling shaft provides access for personnel and
equipment to the repository with limited intake downcasting the shaft. The air from the
waste-handling shaft is directly routed to the exhaust shaft after ventilating the waste-
handling shaft station. Figure 2 gives a cross-section of the WIPP facility and the original
infrastructure.
Figure 3 illustrates the original surface fan configuration at WIPP. Primary ventilation
through the facility was achieved by running of one or two of the 450-kW (600 hp) centrifugal
main fans (labeled 700A, B and C). During concurrent mining and waste-handling operations,
two of the three fans operate in parallel to provide 230 m³/s (490,000 cfm). The surface fan
arrangement also included a filtration mode. In this mode the airflow decreases to 28 m³/s
(60,000 cfm). The shift from full airflow to filtration mode is accomplished by turning off the
main fans and starting one of the three 175 kW (235 hp) 860 centrifugal standby filtration
fans. A series of isolation dampers diverts the air through the filtration system where the air is
routed through a series of high-efficiency particulate air (HEPA) filters.
The underground ventilation system is divided into four circuits. These are the North area,
where the primary equipment shop is located along with a salt disposal investigation (SDI)
experimental area, the construction circuit, TRU waste disposal circuit and the waste shaft
station circuit. Regulators strategically located in each circuit ensure airflow courses from an
area where there is no TRU waste towards the disposal circuit. Figure 4 illustrates the four
underground circuits at the beginning of 2014.

616
Figure 2. Geology at WIPP with original shaft and underground design.

Figure 3. Repository configuration with air splits shown (in 2014).

3 CONTAMINATION EVENT

On February 14, 2014 a radiation detection sensor, called a continuous air monitor (CAM),
alarmed in Panel 7. The CAM measures airborne radioactivity within the exhaust of the panel
where active waste emplacement operations and final storage occurs. The CAM alarmed in the
evening when no personnel were underground at the time of the incident and triggered an

617
Figure 4. Original fan system at WIPP.

automatic ventilation system switch from one primary unfiltered fan (700C) to filtration mode
(860C) where all air exhausting the underground passes through high efficiency filter banks. The
cause of the release was due to a single drum that contained incompatible materials resulting in
a spontaneous combustion event in Room 7 of Panel 7. Figure 5 shows the damage to the TRU
container. The release contaminated certain areas in the underground, namely downstream of
where the release occurred. After the release, DOE and NWP have implemented numerous tem­
porary fan systems to increase the available airflow to the underground. The interim ventilation
system (IVS) consists of two additional surface HEPA filtration units tied to dedicated exhaust
fans. In 2016 the IVS was placed into service, increasing filtered exhaust by an additional 54,000
cfm raising the total filtered exhaust to 114,000 cfm. The Supplemental Ventilation System
(SVS) fan was installed in 2017. This fan is located at the base of the AIS and pushes air into
the construction circuit. This increased the ventilation system capacity for the construction and
the north area of the repository by 46,000 cfm for a total intake volume of 180,000 cfm. With
SVS operating, airflow is routed in series from construction to the disposal circuit. The north
circuit is ventilated by using the SHS as an exhaust. The IVS and SVS were installed as tempor­
ary measures until a permanent ventilation system could be engineered and installed.

Figure 5. Waste container that self-combusted.

618
4 FUTURE MINING AND VENTILATION PLANS AT WIPP

The original operational life of WIPP, at the time of the radiologic release, was to finish emplacing
waste in Panels 7 and 8 and in the main north/south airways between Panels 1 and 4. At that time
this would have ended the operational phase for WIPP. However, after the radiologic event, DOE
re-evaluated the amount of TRU waste currently emplaced at WIPP and concluded that the facil­
ity needed to continue operation under the current waste operating permit. This resulted in
a major study to determine where new potential panels could be mined, a new surface fan config­
uration and an option for a new shaft to access the facility. DOE developed two primary criteria
as a design basis, the first was that for lifecycle planning the facility was to operate to the year
2050. The second criterion was that all exhaust air from the disposal circuit will be filtered with
HEPA systems.
A series of tradeoff studies were performed to determine where additional panels could readily
be developed. The Land Withdraw Act of 1992 set aside 16 sections of land for WIPP. The cur­
rent WIPP underground is in the approximate center of this area. The selected alternative was to
mine additional panels to the west of the existing repository. The primary reason for this selection
was ease of access to the west for mining, faster development to the next panel for waste handling
after Panel 8 is filled, straight forward ventilation system design, and sufficient area to the west for
panel development to achieve the design goals. The current design is to include two additional
panels (11 and 12) to the west with the ability to add additional panels.
Once the mining direction was determined, the engineering team began assessing geotechnical
considerations for the new design. The cross sectional area of new airways was slightly adjusted as
well as the distance between cross-cuts and distance between parallel airways. The panel design
was also adjusted with a greater length between the first room in the panel and the main access
west drives.
The original mains at WIPP are oriented north/south and consist of four entries. South of the
SHS, the second main from the west is a common intake to a regulator that controls air to the
disposal circuit. Air not entering this regulator continues in this drive to ventilate the construc­
tion circuit where it returns in the furthest west main entry. The third entry from the west is the
disposal intake and the fourth is the disposal exhaust. In the new design a second disposal
intake is included. The reason for this is for ground control considerations. With a single dis­
posal intake, any ground control function interrupted waste handling operations. With two
intakes this challenge is virtually eliminated.
The mining of salt at WIPP is primarily by full face continuous mining machines. No dust
suppression is applied during mining. Because of this, exhaust airflow from the construction
circuit can have significant airborne salt dust. In the original ventilation system design, this
dust reported to the surface and was exhausted directly to atmosphere. The new design is to
have the air pass through HEPA filters prior to discharge. In addition to dust, the exhaust
airflow can have significant moisture. This moisture is introduced during certain times of
the year. The facility is in the high desert of New Mexico and monsoonal moisture, typically
coming from the south in the summer, is very common. As the air upcasts the exhaust shaft,
this moisture condenses resulting in significant water discharging through the fans.
Water and salt dust could cause plugging of the pre-filters in the HEPA filtration system. This
would result in constant filter change outs. To minimize the impact of salt and moisture, a surface
Salt Reduction Building (SRB) is being constructed. The purpose of this system is to remove mois­
ture through a demister and salt dust through filtration before it reaches the HEPA filter system.

5 MINE VENTILATION DESIGN

The first ventilation analyses assumed a new surface salt and water reduction system upstream
of a HEPA filter system. These structures will replace the system shown on Figure 3. The cur­
rent design included Panels 11 and 12 as shown on Figure 6. Modelling was performed incorp­
orating natural ventilation pressure (NVP) for both cold (winter) and hot (summer)
conditions. Air expansion/contraction was accounted for in each intake and exhaust shaft.
Airflow requirements were based on diluting diesel exhaust, namely from the small mine

619
Figure 6. Ventilation system design for future WIPP panel development.

trucks, and the equipment used to transport the TRU waste from the Waste Shaft to the
active emplacement panel room. In addition, there are ventilation requirements for each waste
handling room of 20 m3/s (42,000 cfm). It is assumed three rooms are actively ventilated in
a normal operation. The construction area requires 35.4 m3/s (75,000 cfm) in the active mining
zones. Ventilation for the main shop and SDI is assumed at 28.3 m3/s (60,000 cfm) and the
waste shaft station at 11.8 m3/s (25,000 cfm). Bulkhead resistance values were based on his­
toric values at WIPP. Airway resistances were based on measured friction factors and pres­
sure/quantity measurements. Regulators were incorporated with fixed quantity branches and
adjusted until the airflow design values given above were achieved.
The base model for WIPP determined that the maximum exhaust airflow required is
245 m3/s (520,000 cfm) at a static pressure of 2.5 kPa (10.0 in. w.g.) (as measured at the collar
of the exhaust shaft. All system losses downstream of the collar were calculated by HVAC
engineers assigned to specify the primary fans).

6 SAFETY SIGNIFICANT CONFINEMENT VENTILATION SYSTEM

Once the total flow and required delivered pressures were determined, design engineers began to
develop concepts for a salt reduction system and parallel HEPA filtration system. The salt reduc­
tion system consists of six independent moisture and salt reduction units. Each unit is comprised
of a CFT GmbH demister upstream of a CFT GmbH Model HTKK 1/1500-2S dry scrubber.
A Spendrup 152-091-1800-C-1 600 hp fan is connected to each unit. The fan motors are on vari­
able speed drives for airflow control. Each unit is capable of handling a throughput of 120,000
cfm. The fans are fitted with an outlet silencer. Figure 7 shows a SRU during factory acceptance
testing. The salt reduction units are in a separate building called the Salt Reduction Building
(SRB). The SRB has one primary intake duct tied to a manifold to each SRU. A common exhaust
plenum takes the air back to the intake to the Safety Significant Confinement Ventilation System
(SSCVS) in the New Filter Building (NFB). Photos of an installed SRU is shown on Figure 8.

620
Figure 7. Demister, filter and fan system for a single SRU (at the time of factory acceptance testing).

Figure 8. Photos of installed filter and fans.

The SSCVS was designed based on each HEPA filter unit being capable of 12.7 m3/s (27,000
cfm). To achieve a design flow of 255 m3/s (540,000 cfm) this resulted in a minimum of 20 filter
units being available. The system was designed with 22 units. A filter unit is comprised of four
in line filters. The first filter is a MOD filter followed by a HIGH filter then two HEPA filters.
With this train contamination can be removed from the airstream. The basic design of the SRB
and SSCVS system is shown on Figure 9. This figure also shows a typical HEPA filter unit.
Mechanical engineers calculated the pressure losses through the HEPA filters, ducts, dampers
and elbows and computed a maximum possible pressure for the SSCVS fans. A total of six fans
were selected for the design, each fan capable of 70.8 m3/s (150,000 cfm) at a design pressure of
7.0 kPa (28.1 in. w.g.) (at an air density of 0.96 kg/m3 [0.06 lbm/ft3]). At maximum flow four fans
would be in operation each at approximately 63.7 m3/s (135,000 cfm). A Clarage 86.625 SW5731
CHS Fan was selected for the project. A 746 kW (1000 hp) motor on a variable speed drive is
included in the fan design. Because the fans are internal to the SRB, the fan motors will be water
cooled.
Because the resistance in each HEPA filter unit will vary, depending on the amount of load­
ing, the airflow through each unit will be controlled by a flow damper connected to an airflow
sensor on the intake to the HEPA filter unit. The damper will not modulate continuously,
rather be set and compared by personnel operating the system. Differential pressures will also
be monitored across each filter bank within a HEPA unit. Increases in pressure for a constant
flow will indicate when filters should be changed out.

621
Figure 9. Isometric view of SRB and NFB system.

The SRB system is designed to have the internal fans simply overcome the losses of the SRU
and duct work. Flow through the system will be balanced with the SSCVS total flow. In the
event of a radiological release, the SRB will be isolated from the system with high efficiency
dampers and all airflow from the underground will be directed through the SRB bypass duct.
The surface SRB and SSCVS are currently under construction. Completion is expected by
late 2024. The existing fan system connects the north side of the exhaust shaft. The SSCVS con­
necting duct will be to the east. Figure 3 shows the direction of the SSCVS in relation to the
existing fan system. Prior to connecting the SRB and SSCVS to the Exhaust Shaft, a series of
tests will be performed on the system. The 4.3 m (14 ft) diameter horizontal duct from the
Exhaust Shaft location will be fitted with a regulator. The regulator will be partially closed to
simulate the resistance required to support the repository. Testing will include various fan oper­
ations, including full flow, filter unit configurations combined with fan operation, duct integrity,
damper integrity, filter flow adjustment other operational considerations. It is expected that by
conducting a comprehensive testing program the operation once the system is connected to the
Exhaust Shaft will be simplified.

7 NEW SHAFT DESIGN

Historically at WIPP extracting salt is a challenge as the SHS is only 3.04 m (10 ft) in diameter
and consists of a single 7.25 tonne (8 ton) skip. The shaft and its internal components were
constructed in the early 1980s. To expect this infrastructure to last the life-cycle time to 2050
would be a challenge. After the SSCVS study was commenced, DOE embarked on a second
study to evaluate the addition of a new shaft. A series of alternative studies were performed to
evaluate the location of a new shaft, the size and the long term use of the shaft.
The results of these studies concluded that a new 8.54 m (28 ft) diameter shaft would be
constructed due west of the AIS and SHS. The location is shown on Figure 6. This location
was selected since the new mining panels were being constructed to the west and the location
was readily accessible on surface. The shaft diameter was selected to accommodate a possible
new hoisting system in the future. This new hoist would replace the SHS for primary salt
removal and provide personnel and equipment access to the repository. A large conveyance
for personnel and equipment with two salt skips are envisioned for the shaft.
The shaft will be an intake to the repository providing air to the construction and disposal
circuits. A surface intake fan is to be installed on the shaft. The design has two parallel fans
installed on the surface connected to a plenum (see Figure 10). Only one fan will operate at

622
a time with the second being a fully redundant fan. The fans are designed to intake over
235 m3/s (500,000 cfm). With a forcing fan on the collar of the shaft, the airflow from the
construction panel will course to the AIS where it will exhaust on surface. This flow path sig­
nificantly reduces the salt dust loading passing to the Exhaust Shaft.
To facilitate airflow exhausting the AIS a new horizontal duct with stack was constructed
on the collar of the AIS. This design will allow the dust to settle away from adjacent buildings
and parking areas. Figure 11 shows the exhaust duct and stack on the AIS. The shaft project
includes underground excavations around the shaft station and two airways connecting to the
primary north/south drifts. The shaft project completion date is expected to be midyear 2024.

Figure 10. Intake fan with plenum to new Figure 11. Exhaust stack at the AIS.
shaft.

8 SUMMARY

This paper describes two significant ventilation upgrades at the WIPP facility. The SSCVS con­
sists of a 255 m3/s (540,000 cfm) exhaust fan system. The entire exhaust system will be capable
of being filtered for any radiological release. This filter system includes HEPA filters. The result­
ing design is the largest HEPA filtration system in the world. A total of six 746 kW (1000 hp)
fans are included in the design with a maximum of four fans operating in parallel. A water and
salt dust removal system are being installed upstream of the HEPA filtration units to reduce
exhaust contaminants from reaching the filters. This system incorporates six demister and dry
dust filters each tied to a vane axial fan. It is designed to filter the entire maximum flow through
the system. In the event of a radiological release underground, the water and salt dust filtration
system will be turned off and the air diverted directly to the SSCVS.
In addition to the primary exhaust system upgrade, WIPP is installing a new 8.54 m (28 ft)
diameter intake shaft with surface intake fans. This system will provide upwards of 236 m3/s
(500,000 cfm) of intake air to the underground. A new exhaust path will be established to take
salt laden air from the construction system and route it to the air intake shaft which will be an
exhaust. On the surface of this shaft is an exhaust duct and stack to minimize salt dust over adja­
cent buildings and the parking lot. The system can be adjusted to operate without the intake fans
on the new shaft since the SSCVS can handle the maximum expected airflow at the facility.
With this system it is expected that WIPP can maintain operations for construction of Panels
11 and 12 and beyond should DOE elect to do so. In addition, the system will be robust and
can be managed to adapt to any normal or off-normal operating condition. A future project to
install a new hoisting system in the new shaft will further enhance WIPPs operational flexibility.

REFERENCES

CFT GmbH, Beisenstraße 39-41, 42964 Gladbeck, Germany, www.cft-bmbh.de.

623
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Ventilation network optimization: Realizing energy savings while

promoting worker health and safety

A.K. Ngcibi & M. Mochubele

University of the Witwatersrand

F.S. Bergh
Howden Africa

ABSTRACT: The COVID-19 pandemic has certainly shocked the world and made everyone
to rethink, rework and reimagine the workplace of tomorrow. The 21st-century engineer will
have to work smarter to manage the changing environment and employ technological tools to
aid them in the quest for zero harm. This paper outlines the concept of remotely and autono­
mously supplying ventilation to underground workings, commonly referred to as Ventilation-On
-Demand (VOD). For ease of reference, the presentation refers to VOD. However, the proposed
system should rather be viewed as a Ventilation Optimization Solution. Through remote moni­
toring of working conditions, these systems can be used extensively to improve workplace condi­
tions, whilst offer financial benefits. This is made possible by employing sensors for real-time
monitoring integrated with hardware to address sub-standards in real time.

1 INTRODUCTION

Globally, the mining industry is grappling with a number of issues which include shortage of
electricity supply and associated rising costs. A casing point is a South African metalliferous
mining industry which is heavily reliant on electricity and highly sensitive to energy cost. This
begs the question: how will mines be able to supply qualitative and quantitative air in the
underground working in a cost-effective manner?
Historically, South African mines would afford to adopt the ‘brute-force’ ventilation strat­
egy which means ventilating all the underground workings at full capacity 24/7 a day regard­
less of activity or the underground environmental conditions. The year 2007 can be marked as
a defining point of paradigm shift as South Africa started experiencing unprecedented electri­
city tariff increase and load-shedding. This is one of the major reasons why the industry is
looking at reducing the electricity cost by focusing research on energy efficient projects such
as Ventilation On Demand (VOD). The guiding principle of energy cost saving using VOD is
based on the cubic relationship between power and air quantity (P α Q3). This means any
reduction in quantity will result in significant energy savings. This research will also go a long
way in addressing the economic sustainability and indirectly reducing carbon footprint by
using energy derived from fossil fuel efficiently.
The scope of this paper is on reducing the cost of ventilation using VOD system in under­
ground metalliferous mines and also improving the health and safety.
This is a literature review paper, and the layout is made up of research background, litera­
ture review, health and safety, conclusion and recommendations. The literature review pro­
vides views/conclusions from different literature illustrating how real time monitoring, in
conjunction with the remote control of ventilation devices can support mines on their journey
to zero harm. In addition, the literature review provides the feasibility study which provide
the cost of VOD and the payback period.

DOI: 10.1201/9781003429241-64

624
2 RESEARCH BACKGROUND AND CONTEXT

The dynamic underground environment, and the dependence on human intervention, are major
challenges in supplying adequate fresh air to areas where employees are required to travel and/
or work. This can result in a deterioration of workplace conditions and is a constant threat to
the health and safety of employees. Covid-19 forced skilled and experienced personnel to work
from home, which increased the risk of mismanagement of the ventilation systems.
Inadequate supply of fresh air has detrimental effects on the health and safety of the
employees. Mine ventilation professionals must empower themselves with the available tools
to ensure workplaces remain safe and risk-free to the health of employees.
Mines need to become more innovative and utilise the available infrastructure to distribute
the air in a more optimised and autonomous fashion by adopting optimisation systems. This
approach will minimise the dependency on human intervention and make it possible for
skilled and experienced personnel to work from home.

3 LITERATURE REVIEW

Historically, South Africa (SA) had very competitive electricity tariffs. The abundance and
rates did not warrant an energy efficient approach towards mine ventilation. This notion is
supported by Johnson (2012) in his study which concluded that South African mines are pro­
duction driven where energy efficiency is not part of the mind set and culture.
Resultantly, the forbearers of mine ventilation in SA could afford to design ventilation sys­
tems based on the worst-case scenario as highlighted by Kocsis, et al (2003). The approach
was to “set and forget” as the brute-force ventilation system would always ensure more than
the required air supply to the underground workings (Mochubele, 2014).
As early as 2003, Kocsis et al recognised the potential of VOD systems for energy savings,
with Johnson in 2012 and Mochubele (2013) all reaffirming the same notion. The health and
safety improvements of the systems are neglected in the literature, with only claims to ‘main­
tain’ health and safety being made. The literature review focused on VOD implementation
and the associated health and safety improvements and energy savings.

3.1 Ventilation on demand implementation

The widely accepted VOD definition is that it is a process of adjusting mine airflow to supply
qualitative and quantitative airflow in the required area, in real-time, to maintain environmental
standards. VOD is a system of sensors and remotely controlled hardware capable of real time
adjustments to mine ventilation equipment, such as fans and regulators, to meet underground
atmospheric quality requirements for personnel and equipment based on their status and location.
VOD systems set out to improve health and safety, reduce energy usage and increase gen­
eral productivity. Even so, not all mines are lining up for the technology. This is due to the
high cost associated with installing monitoring and communication systems and the subse­
quent maintenance required to keep the system functional and effective (Wallace, Prosser, &
Stinnette, 2014). This cost of implementation is rarely ever stated in literature reports, as most
always quantify the savings but not the capital nor operation cost of the systems.

3.1.1 VOD levels of control

The high capital and maintenance costs along with the steep learning curve associated with
VOD systems has led to mines considering implementing VOD strategies usually take
a phased approach. The approach is based on the levels of control of the system shown in
Figure 1. Mines can phase-in VOD systems by starting with the basic Level 1 and 2 control,
and then scaling up the system over years to the more sophisticated Level 5 control, which
yield best results in terms of energy cost saving.
Level 1 offers the most basic and simple VOD system, allowing for manual remote control
of the ventilation equipment. This is widely used in mines for control over primary fans.

625
Figure 1. VOD levels of control (Howden, 2018).

Energy savings can be achieved through the system when used with variable speed drives
(VSDs) on the fan motors instead of guide vanes which reduce quantity by reducing the fan
rotational speed. For health and safety improvements in the mine, it is worthy to invest in
level 3 of controls which require extensive monitoring of the environmental conditions and
level 4 which track the locations of personnel and machines. Level 5 refers to completely
autonomous dynamic ventilation-on-demand systems which incorporates time-based, quality-
based and event-based ventilation optimization.

3.1.2 Remotely controlling VOD systems

Nick Holland, former CEO of Goldfields spoke of the possibility of mining at 5km (Mining
Weekly, 2020), citing that it is only possible through remote or autonomous mining due to the
potential dangers and risks. Mochubele (2014) reported on the potential benefits of automa­
tion, with electrical cost reduction being top of the list. The added and most significant benefit
of remote mining is the removal or reduction of personnel from the high-risk underground
environment.
VOD implementation is a preliminary step towards the goal of ‘ZERO HARM’, which in
reality can only be achieved through the complete removal of personnel from the hazardous
underground environment. VOD systems allow for tracking of the positions of personnel and
machines and real-time monitoring of the airflow properties and gas remotely. Events are
scheduled and setpoints are created for desired airflow quantities and velocities, temperatures,
percentage gases in the general atmosphere etc.
Computer programmes such as Ventsim Control compute the air demand and determine
the necessary fan and regulator set points required to achieve these environmental standards.
Hardware such as Programmable Logic Controllers (PLC’s) or remote Inputs and Outputs (I/
O’s) enable the remote control of equipment such as fan starters, Variable Frequency Drives
(VFD’s) or actuators.

3.2 Health and safety implications

The Mine Health and Safety Act (MHSA) of 1996 (No. 29) was the first stand-alone health
and safety legislation for mines in South Africa. Under the MHSA, the government, mining
industry (employer) and workers are all required to promote health and safety in the work­
place (Mohapi & Zarske, 2018). This means that mining companies are under pressure to pro­
actively protect employees from all present and future risks. Implementation of new
technologies such as VOD in underground mines is one way that companies can keep ahead
of the health and safety curve.

3.2.1 Early entry examinations

Efficient underground ventilation and cooling forms a critical part of sustainable underground
mining. Ventilation provides oxygen while removing dust and post-blasting fumes and is used
in conjunction with cooling for temperature control (Wallace, Prosser, & Stinnette, 2014).

626
 There is a need for re-entry examinations in underground mines after blasting, to ensure the
environmental conditions are within acceptable limits before re-entry into working areas.
These are usually done manually with hand-held devices, increasing the risk of exposure to the
observer, to not just noxious fumes, but seismicity risks as well. Through VOD, examinations
can be done remotely by checking the real-time gas readings in the system. This reduces the
risk of exposure and hence improve the health and safety position of the mine.
A hypothetical mine has a re-entry standard of 45 minutes, with early-entry examinations
conducted manually by the re-entry team. The re-entry team involves the shift supervisor
(miner); the crew leader; and two safety representatives. Hand-held gas detection instruments
are used for this and the following hazards are identified for this process:
i) Short term overexposure to CO
ii) Short term overexposure to NO2
iii) Flammable gas accumulation and fire risk
The time-weighted average occupational exposure limits (TWA.OEL) represent the minimum
acceptable exposure over a longer period i.e. 8 hours. The Short Term Exposure Limit (STEL)
refer to acceptable exposure over a short period, usually 15 minutes, as long as the time-
weighted average is not exceeded. Please see the TWA and STEL values in Table 1 below.

Table 1. Occupational Exposure limits

(MHSA, 1996).
Gas TWA.OEL STEL.OEL

CO 30 ppm 100 ppm

The acute effects of CO exposure include but are not limited to mental confusion, vomiting,
loss of muscular coordination, loss of consciousness and death.
High exposures, above 70ppm, leads to victims rapidly becoming mentally confused and
can lose muscle control and consciousness without experiencing any mild symptoms
CO takes longer than all the gases to dissipate to the OEL.TWA levels. This occurs after more
than 45 minutes from blasting according to the figure below. With a re-entry time of 45 minutes
in place at the mine, the probability of acute exposures of the re-entry team is quite high.

Figure 2. Blast fume dissipation trend (Brake, 2015).

627
 Potential accumulations in muck piles and lingering accumulations caused by ineffective and
damaged ventilation infrastructure poses a risk to the re-entry team. The likelihood for this
event to occur is probable (35% - 65%) Figure 2 illustrates the dissipation trend of blast fumes.

3.2.2 Real time monitoring

VOD systems incorporate real-time gas and airflow sensors which allow for monitoring of
environmental conditions in real-time. Whilst reducing the risk of exposure, it also increases
the monitoring capacity of underground environmental conditions. The real-time monitoring
and remote/ autonomous control means changes in environmental conditions can be dealt
with timeously by either directing more airflow to the affected areas or evacuating personnel
from the areas in time. This improves the health and safety of underground mines. This goes
a step further than simply “maintaining” health and safety standards, but suggest that these
solutions can improve workplace conditions.
A practical example is the VOD solution deployed at Eleonore mine in Canada. The solu­
tion automates the operation of all ventilation equipment including main fans, auxiliary fans
and airflow regulators (Howden, 2021).
The system includes 30 ventilation monitoring stations (VMS) which determine the quantity
and quality of airflow at various positions in the mine. Each VMS is equipped with airflow,
carbon monoxide (CO), nitrous dioxide (NO2) and propane (C3H8) sensors. The system
incorporates a mine-wide tracking system for locating personnel and vehicles. Both personnel
and vehicles are issued with radio frequency identification (RFID) tags that connect to any
one of the 250 zone-based WIFI access points.
The RFID location data provides enhanced safety and management of the movement of
personnel and vehicles, while the VMS system improves the monitoring of environmental con­
ditions in the production zones. The combined data allows for ventilation requirements for
each ventilation zone to be calculated by the VOD logic software, based on personnel and
vehicle location and environmental conditions.

3.2.3 Remote control and management of a ventilation system

In a post-Covid-19 world, more and more skilled employees work from home, and mines will
have to follow suit to attract skilled employees to the sector. The danger is that mines will
always need to remain ventilated and environmental conditions maintained to ensure safe,
healthy and sustainable production.
Furthermore, as mines operate on multiple shifts there’s a risk that conditions can deterior­
ate during periods where ventilation experts are not present e.g. exposure to pollutants and/ or
fire or influx of gas during a night shift that could endanger the lives of hundreds of persons.
Managing a ventilation system remotely will empower mine ventilation engineers to carry
out their legal duties even if being on-site is not possible. A control system will assist by log­
ging all changes to the ventilation system, complete with alerts based on varying levels of
severity. This function will also fulfil the legal requirement of an early warning system. In add­
ition to this, a control system allows the mitigation of sub-standard conditions in real-time.
Naturally, the system and associated hardware needs to be maintained to ensure it operates
satisfactorily. The hardware kits installed have a proven record of operating in the harsh
underground environment, but care and maintenance are critical in ensuring sustainable and
efficient operation thereof.

3.3 Energy cost savings

Airflow in underground mines is driven and controlled through pressure differentials created
by the use of fans. The associated cost of ventilating comes from the energy dissipated by
these fans to create a pressure differential. The absorbed power is based on the relationship
between mine resistance, fan pressure and airflow. Costs can be reduced either by using smal­
ler fans where possible or reducing the speed of the fan. VOD allows for the manoeuvrability
of fans operation point (speed) based on demand to improve energy efficiency, whilst main­
taining quality of mine environmental conditions.

628
 The implementation of a Ventilation Optimisation Solution can:
• Schedule events for reduced flows during shift-change and breaks;
• Reduce ventilation in zones where no production is occurring
• Reduce ventilation rates in zones where ventilation exceeds the requirements
• Monitor and control air quality within regulatory limits to maintain an adequate level of
gases, airborne pollutants and temperature
• Post-blast ventilation management to minimise blast clearing times.
Through these controls, the mine can achieve cost savings and improve health and safety
through increased monitoring and control over the airflow to respond to conditions in real-
time.
A case study at Newmont’s Eleonore Gold Mine in Canada, shows the mine has achieved
• a 43% reduction in mine heating costs
• a 56% reduction in underground ventilation electricity costs; and
• a decline in surface ventilation electricity costs of 76% through VOD implementation
(Gleeson, 2019).

3.4 Ventilation optimisation study at a South African mine

A feasibility study performed at a South African operation yielded the following results. The
operation could not establish additional vertical shafts to cater for their life of mine commit­
ment due to surface topography. The scope of work was to revise the ventilation design and
strategy to ensure the ventilation commitments comply with the mine standard.
The future mine will use seven (7) working ventilation districts for production. Of these 7
sections:
• 3 Sections are production sections with an average intake velocity in the last through road
required to ≥ 0.75m/s (2 roadways).
• The 4 remaining sections are non-production sections with an average intake velocity in the
last through road required to ≥ 0.15m/s (2 roadways).

3.4.1 Optimisation of primary surface fans

The study proposed fitting VSDs to the Primary Fans. This will enable the primary fans to be
operated at different speeds, and in doing so will enable the mine to capitalise on periods
where the ventilation demand is reduced.
In 2023, the two primary fans need to run at 100% speed during production (P), however,
savings can be realised during non-production (NP) hours. From 2024 onwards, due to the
underground activity, primary fans need only run at 85% of their capacity during production
time to meet the demand. Tables 2 and 3 below illustrate this:

Table 2. Characteristics of primary fans to be integrated 2023.

Power Starter Flow Fan Speed Percentage: Production time (P) and
Quantity (kW) Type (m³/s) Non- production time (NP)

2 700 VSD 176 P:100%

NP: 1 x fan Off; 1x fan 57%

Table 3. Characteristics of primary fans to be integrated 2031.

Power Starter Flow Fan Speed Percentage: Production time (P) and
Quantity (kW) Type (m³/s) Non- production time (NP)

2 700 VSD 176 P:85%

NP: 1 x fan Off; 1x fan 57%

629
3.4.2 Ventilation optimisation of underground ventilation network and booster fans
The proposed design, with air crossings placed strategically within the underground work­
ings, eliminated the need for underground booster fans. Ventilation control devices comprise
of air crossings and automated regulators in the return airway of each section. The running
cost of the booster fans add to annual savings for the mine. Figure 3 below depicts an auto­
mated louvered regulator used to automatically regulate airflow flowing into the ventilation
districts.

Figure 3. Automated Louvered regulator.

3.4.3 Ventilation optimization of auxiliary fans

The following VOD strategy concerning auxiliary fans was proposed. All auxiliary fans are
switched off for the two (2) hours of non-production time i.e. travelling time, vehicle inspec­
tions etc. The auxiliary fans in the non-productive sections remain switched-off for the dur­
ation of the shift. Auxiliary fans in the non –production sections will be switched on
automatically when personnel and/ or machinery enter sections, or when real time monitors
indicate sub-standard conditions.

3.5 Summary price: VOD solution

The system has a capital cost of USD 1.8 million which translates to just over ZAR
28 million. This is based on the primary fans being retrofitted with VSDs and the auxiliary
fans being retrofitted with auxiliary links. The inability to retrofit would impact the capital
costs of the system as new fans would be required for the system to work optimally.

3.6 Financial analysis

Figure 4 graphically represents the financial analysis conducted for the VOD system outlined
in this case study:
From Figure 4, the system payback period is shown to be just below 2 years. This is the
point when the net present value (NPV) is equal to zero, and it increases to just below USD
13 million in 2031. Concerns about the VOD system being capital intensive should be

630
Figure 4. Financial analysis USD.

discarded. VOD systems have short payback periods of about 2-3 years as they pay for them­
selves through the energy cost savings achievable through this dynamic control of airflow.
These systems have proven to be critical not just for energy savings but for improving the
health and safety in underground mines as well.

4 HEALTH AND SAFETY IMPLICATIONS

VOD systems allow mines to view environmental condition data in real-time, whilst logging
changes in the ventilation system and notifying officials of sub-standards for localized evacu­
ations. The system also allows the user to implement controls in real-time that will safeguard
worker health.
VOD systems help limit the time of exposure to airborne pollutants by preventing a build-
up of contaminants in production zones through re-directing noxious gases directly to return
airways. This will aid the operations existing controls in creating an environment that will go
a long way to achieving “ZERO HARM”.

5 CONCLUSION

Ventilation-on-demand consists of different levels of control i.e. manual-based, time-based

scheduling, quality-based (requires flow and gas readings), event-based (requires tracking and
tagging); and autonomous systems. Adoption of the systems is heavily swayed by the high
capital and learning curve costs associated with the higher levels of control. Case studies on
mines with VOD systems show that they have the potential to reduce energy costs while main­
taining, and improving, health and safety conditions in underground mines. A feasibility
study carried out in one of the South African mines demonstrated that although VOD system
has high capital cost of R28 million the payback period is less than two years.
Real-time monitoring of environmental parameters is crucial and something that is being
employed extensively within the mining industry. A VOD solution will enable mines to act
and mitigate exposure to airborne pollutants in real-time. Based on the information contained
in this report, it is evident that VOD solutions are advantageous. Mines need to consider util­
ising these technologies to improve workplace conditions and reduce energy costs.

631
6 RECOMMENDATIONS

Based on the conclusion and information contained in this technical paper, it is evident that
VOD or Ventilation Optimization solutions are advantageous to improving health and safety.
The removal of employees from dangerous work places or conditions, such as re-entry exam­
inations, shows to be the most effective way of improving health and safety. The current
system only removes the re-entry team from the hazardous area during re-entry examinations.
The team will still go and work in the very same areas for an entire shift where environmental
conditions can be readily reduced to a sub-standard. Mines should explore modernisation
techniques such as remote or automated mining solutions. VOD systems are an enabler to
such solutions according to Mochubele (2014).

REFERENCES

Brake, R., 2015. A Review of Good Practice Standards and re-Entry Procedures after Blasting and Gas
Detection Generally in Underground Hardrock Mines. Virginia, 15th North American Mine Ventila­
tion Symposium.
Dominguez, C. R., Martinez, I. V., Pena, P. M. & Ochoa, R. A., 2019. Analysis and evaluation of risks
in underground mining using the decision matrix risk-assessment (DMRA) technique, in Guanajuato,
Mexico. Journal of Sustainable Mining, 18(1), pp. 52–59.
Gleeson, D., 2019. International Mining. [Online] Available at: https://im-mining.com/2019/11/07/how
dens-eleonore-ventilation-on-demand-solution-wins-award/ [Accessed 25 August 2020].
Howden, 2021. VentSim Control: Eleonore. [Online] Available at: https://www.howden.com/en-us/case
studies/ventsim-control-eleonore#pdf [Accessed 16 November 2021].
Howden, V. S., 2018. Levels of control within a ventilation optimization system. Montreal: Howden
Ventsim Solutions.
Johnson D, F. C., 2012. An overview of energy efficiency in South African Hard Rock Mines, ??:
Research gate.
Kocsis, C. K., Hall, R. & Hardcastle, S. G., 2003. The integration of mine simulation and ventilation
simulation to develop a ‘Life Cycle’mine ventilation system. South African Institute of Mining and
Metallurgy, pp. 223–230.
MHSA, 1996. Mine Health and Safety Act 29 of 1996. [Online] Available at: Mine Health and Safety Act
[Accessed 21 August 2020].
Mining Weekly, 2020. Getting the best out of mining starts with policy certainty, says Gold Fields’ Hol­
land. [Online] Available at: https://www.miningweekly.com/article/getting-best-out-of-mining-starts-
with-policy-certainty-holland-2020-08-20 [Accessed 21 August 2020].
Mochubele, E. M., 2014. Effects of Increasing Rejection Temperatures on Electricity Demand for Venti­
lation and Cooling in Automated Metalliferous Underground Mines, Johannesburg: University of the
Witwatersrand.
Mohapi, G. & Zarske, R., 2018. Health and Safety in South African Mines: A Best Practice Report,
Frankfurt: Southern African-German Chamber of Commerce and Industry.
New Jersey Department of Health, 2000. Hazardous Substance Fact Sheet: Nitrogen Dioxide. New
Jersey, New Jersey Department of Health and Senior Services.
Oxford Reference, 2021. Quantitative Research. [Online] Available at: https://www.oxfordreference.com/
view/10.1093/oi/authority.20110803100357649 [Accessed 16 November 2021].
US Consumer Product Safety Commission, 2021. Carbo-monoxide Questions and Answers. [Online]
Available at: https://www.cpsc.gov/Safety-Education/Safety-Education-Centers/Carbon-Monoxide-
Information-Center/Carbon-Monoxide-Questions-and-Answers [Accessed 16 November 2021].
Wallace, K., Prosser, B. & Stinnette, J. D., 2014. The practice of mine ventilation engineering, California:
Mine Ventilation Services, Inc. Fresno.
Yahnke, K., 2021. Risk Assement Matrix. [Online] Available at: https://www.i-sight.com/resources/risk-
assessment-matrix/ [Accessed 16 November 2021].

632
Ventilation planning and design
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Design highlights for Agnico Eagle’s Macassa 4 Shaft primary

ventilation systems

K. Boyd, T. Mehedi & D. Witow

Hatch, Canada

M. Pinheiro-Harvey
Agnico Eagle, Canada

ABSTRACT: The 4 Shaft Project represents a step change for ventilation capacity at
Agnico Eagle’s Macassa Mine in Kirkland Lake, Ontario. This paper outlines the process
design, equipment selection, and experience in procuring new primary ventilation equipment
for this project. Equipment includes new 64 MMBtu/hr natural gas direct fired mine air heat­
ers with integrated silencers; twin 600 hp vane axial fresh air fans; and twin 3,000 hp centrifu­
gal primary exhaust fans with demisters and high-performance silencing measures. This paper
presents the description of obstacles tackled during the engineering design of the primary
mine ventilation equipment, which included space limitations on surface and noise concerns
due to proximity to local communities.

1 INTRODUCTION

In Northern Ontario, Canada, the town of Kirkland Lake has seen seven gold mining oper­
ations in the 20th century. After a successful exploration campaign in the early 2000’s, Kirk­
land Lake Gold re-established gold mining operations at the Macassa mine, which included
using the existing 3 Shaft as a service, skipping, and access shaft, and the 2 Shaft as secondary
egress. The ventilation circuit was driven by underground booster fans operating in parallel
within several of the historic tracked mining levels. The most recent configuration had fresh
air downcast in 3 Shaft and exhaust in 1 Shaft and 2 Shaft. After successful exploration at
depth, a new access shaft, called 4 Shaft, was approved, and began construction in 2018. To
accommodate the increased fresh air capacity, a pair of new exhaust raisebores to surface
were also approved for construction in parallel with the shaft.
Now merged with Agnico Eagle, the company’s 4 Shaft Project is now in the final stages of
construction in 2023. This shaft represents an opportunity to renew and improve mine access
in proximity to the deep SMC and Lower North orebodies including ore handling systems;
mine services and utilities; and ventilation. Additional fresh air will support mining activities
at depth which promises to improve safety and productivity.
To support production before the four-year shaft construction period was complete, the
SMC230 project was developed at Macassa in 2020 which included three new underground
fan stations and refit of three others which increased fresh airflow to operations at depth –
achieving the namesake of 230 kcfm airflow to the SMC orebody.
New ventilation infrastructure to leverage the new 4 Shaft and raisebore airways was div­
ided among two projects:
1. 4 Shaft Project, which would complete the 4 Shaft Heaters and 160L Booster Fans.
2. Raisebore Exhaust Fans Project, which would construct the new fans for the raisebores.

DOI: 10.1201/9781003429241-65

635
2 PROCESS DESIGN

The life-of-mine (LOM) ventilation schematic for Macassa can be seen in Figure 1. The venti­
lation circuit is negatively induced via the twin raisebore exhaust fans with the fresh air
coming from 3 Shaft and 4 Shaft. Multiple booster fans are installed mine-wide to direct the
airflow through various mining areas.
The design for the 4 Shaft Heaters, 160L Booster Fans, and Raisebore Exhaust Fans com­
menced with a studies phase to develop the basis of design including tradeoffs to confirm equip­
ment selection and configuration. New ventilation equipment was sized to maximize the capacity
of the new shaft and exhaust raisebores – to provide the highest degree of future flexibility.
Key process conditions forming the basis of the 4 Shaft Heater, 160L fans and the Raise­
bore Exhaust Fans are summarized in Tables 1 and 2.

Table 1. Summary of process parameters for 4 Shaft Heaters.

Parameter Units Value
3
Nominal Air Flow (Volumetric) m /s 307
Design Inlet Temperature (Min) °C db -44
Design Temperature Rise °C db 49
Thermal Capacity MMBtu/hr 60
Heating Fuel - Natural Gas
Heater Type - Direct Fired

Table 2. Summary of process parameters for 160L Fans.

Parameter Units Value
3
Nominal Air Flow (Volumetric)- combined m /s 307
Inlet Air Temperature °C db 5 to 36
Mine Static Pressure kPa 1.04
Fan type - Vane Axial – Arrangement 4
Airflow Isolation (Maintenance) - Louver Dampers (actuated)
Fan Configuration - Two Fans in Parallel

A schematic of the process design for the 4 Shaft Heater and 160L fans are shown in
Figure 2. The layout of various components was designed with the following key design aspects:
• 4 Shaft Heaters would include an acoustically insulated steel plenum above the Vent Shaft
with removable roof panels to allow future access for fan replacement.
• Relatively low-pressure application – with the 160L fans only having to overcome pressure
losses up to 4 Shaft Station on 160L.
• Fan flow slightly higher than 4 Shaft downcast demand to upcast 30 kcfm to 4 Shaft collar
which provides some make-up air for the headframe.
Prior to selection of equipment, a noise study was carried out to define the maximum allow­
able sound power level to be specified for the heater house intake that will meet community
noise limits at the nearest receptors and determine whether the underground fan discharge
levels can meet occupational noise exposure limits at 160L and at the Shaft collar work area.
Two fan operation conditions were evaluated:
• Both fans operating under typical conditions
• Maintenance scenario where one fan runs while the second is being maintained.

636
Figure 1. Macassa life of mine ventilation schematic in long section view process design.

Table 3. Summary of process parameters for Raisebore Exhaust Fans.

Parameter Units Value
3
Nominal Air Flow (Volumetric)- combined m /s 330
Inlet Air Temperature °C db 15
Mine Static Pressure kPa 7.69
Fan Type - Centrifugal
Fan Configuration - Two Fans in Parallel

The Raisebore Exhaust Fans are planned to be installed on surface, which is the preferred
location, however there are concerns regarding the noise impact on the adjacent community
due to limited mitigation opportunities with a surface installation. An underground installa­
tion to alleviate these concerns was considered, but ultimately the decision was made to go
forward with a surface installation due to technical and economic considerations.
This decision required modifications to the design of the fan system to include an independ­
ently supported acoustic enclosure surrounding the fan casing and all ductwork before the
existing silencer, and the addition of a second silencer of similar performance in series. Both
the considered underground and chosen surface layouts are shown in Figure 3.
The VFD cutouts for the 160L fans are located downstream of the fans (as shown in
Figure 4 Figure 4). To protect the equipment inside the cutout from high circulating/turbulent
air movement induced by the passing airstream, placement of deflection shields was suggested
based on CFD analysis. A sketch of predicted velocities and net forces are shown in Figure 5.
Figure 5 Finally, a standard operating procedure (SOP) was developed for accessing 160L to
promote safety given airflow velocities around the 160L shaft station.
Demisters were selected for the Raisebore Exhaust Fan application to avoid contamination of
local soils with dirty water droplets. The outlet side of the fan was selected for the demister loca­
tion, which reduced ductwork and footprint while also offering less velocity transition. In other
words, the exhaust air only requires one expansion transition to reach a low velocity required for
the demister, outlet airflow silencer, and stack. Turbulence from the fan discharge moving through
the evasé was a concern and CFD modeling was conducted to demonstrate to project teams and
demister vendor that the velocity profile was suitable for equipment integrity and performance.

637
Figure 2. Schematic showing 4 shaft heaters and 160L Booster fans in elevation view.

Figure 3. Section view of exhaust fan concept in underground chamber (a) and Selected configuration
showing elevation view of one fan on surface (b).

3 EQUIPMENT SELECTION AND PROCUREMENT

Equipment selection trade-off studies, which included workshops with Operations and vendor
consultations, were conducted by the project team to inform specification packages for competi­
tive bidding. After clarifications, the technical bid evaluation process identified compliant

638
Figure 4. Isolation view of the underground 160L Booster Fans with VFD location on right hand side.

Figure 5. CFD study results indicating high circulating flows in left view and shield walls with improved
conditions on the right.

offerings and a single vendor, Howden, a Chart Industries Company, was selected to provide all
the packages.
Figures 6 and 7 show the general arrangement of the 4 Shaft Heaters and 160L fans. To main­
tain access to the ventilation shaft for material movement, namely for 160L fan components, the
heater house is equipped with removable roof panels. Each burner will be connected to its individ­
ual valve train mounted on a steel rack inside a control room adjacent to the heater house. To
remove any sort of lifting hazards, the four burner valvetrains were divided into pairs and located
along opposite walls in the control room. The control room will house a control panel with
a burner management system (BMS) as required by the Technical Safety Standards Authority
(TSSA), and an integrated PLC system which will allow the Operators to monitor and control
the system either directly from a local operator station (panel mount touchscreen) or indirectly
from any HMI location elsewhere on the site.
Split baffle-type silencers were selected for the 160L fans so that a low-profile lifting trolley
can be installed within the drift excavation which will be used for initial installation and main­
tenance needs.

4 DESIGN LESSONS AND DISCUSSION

Highlights from the 4 Shaft Heaters include:

639
Figure 6. General arrangement for 4 Shaft Heater in Long Section across the Burner Room.

Figure 7. Elevation view arrangement showing one of the two of 160 Fan Systems.

• To fit the four 15-foot-long grid burners, a two-story heater building was used, making all
burners fully accessible without working at heights.
• Due to strict community noise limits, noise from combustion blowers and natural gas com­
bustion in the burners was a concern. A field study at an existing heater was conducted so
that a realistic sound power level for these components could be incorporated into the
noise control monitoring.
• Like the 3 Shaft Heaters, the burners and burner control systems utilized similar compo­
nents including flame rods for combustion monitoring, Maxon APX forced combustion air
grid burners, and Honeywell 7800 series flame safety relay.
• The plenum connecting the Vent Shaft and the heater was constructed as modular acoustic­
ally insulated panels to reduce site construction time.

Features of the 160L Fans are summarized:

• AMCA Arrangement 4 fans were used to minimize operating complexity.
• Low voltage (600 V) fan motors were selected to reduce both motor and VFD cost.
• Louvre-type rectangular isolation dampers were used for maintenance isolation – including
an open/close actuator.

640
• Motor brakes were included to allow any free-wheeling fan to be brought safely to a stop.
Disc-type brakes were used with rotors mounted to shaft extensions on the motor non-
drive end.
• Fan preventative maintenance was prioritized, and a Bently Nevada 1900/65 monitor was
selected to monitor a horizontal vibration probe on each fan bearing. With two vibration
probes per fan, a single four-channel monitor can monitor both fans.
• Temperatures of both motor bearings and windings are also monitored to improve main­
tenance planning.
• Variable speed drives were selected for fan modulation and are expected to have a long ser­
vice life in the relatively clean conditions on 160L.
• With the relatively unobstructed and short airway these fans are responsible for, stall sen­
sors were not required.
Features included on the Raisebore Exhaust Fan are as follows:
• Full condition monitoring instrumentation for 3,000 hp fan motors and fan bearing/shafts.
• Modulation with variable speed drive and option to also modulate with parallel-blade inlet
dampers. The inlet dampers also serve to isolate a fan system during maintenance or repair.
• Fitted wash bars were allocated for the stainless steel, blade-type demisters.
• Extensive noise engineering was applied to optimize noise control measures for the fans –
which includes a combination of baffle-type outlet silencers, acoustic lagging on exposed
ducting surfaces, and acoustic enclosure around the fan body.

641
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Upper Keel mine ventilation strategy at Eagle mine

K. Boyd, D. Witow & C. McGuire

Hatch, Canada

C. Gobbs
Lundin Mining Corporation, Eagle Mine, USA

ABSTRACT: Lundin Mining operates the Eagle Mine in the Upper Peninsula of Michigan
USA. The mine is preparing to construct the new Upper Keel zone – another high-grade
nickel orebody in proximity to the existing mine. This paper provides an overview on several
aspects of the ventilation design and strategy including the options for development and con­
struction, life-of-mine operations, and the innovative ability for on-shift blasting during the
construction phase while maintaining operations in the existing Eagle and Eagle East zones.

1 INTRODUCTION

Eagle Mine is owned by Lundin Mining Corporation, a global diversified base metal company
headquartered in Toronto. Eagle Mine started operations in 2014 with an expected life of
mine into 2027. The hard rock underground mine is located in the Upper Peninsula of Mich­
igan, USA. Nickel/copper ore is mined using both stoping and cut-and-fill mining methods
for the Eagle and Eagle East ore bodies. A satellite ore body known as the Upper Keel deposit
is located approximately 800m east of the existing portal and approximately 430m below
surface.

2 MINE-WIDE VENTILATION

The primary ventilation circuit for the mine uses two fresh air intakes: one through the mine
portal that is equipped with fresh air fans and direct-fired heaters and a Fresh Air Raise
(FAR) that is equipped with direct-fired heaters. Fresh air from the main decline and FAR is
distributed to Eagle by two 700hp exhaust fans located on the surface, at the collar of
a Return Air Raise (RAR). The Eagle East ventilation uses a branch circuit that is powered
by two 400hp fresh air fans to “push” fresh air from the main decline and FAR to the mining
zone. All exhaust air from both zones is collected on the 265L return air drive (RAD) via
exhaust raises and exhausted out the RAR. Additional details can be found in (Witow et al.,
2019).
The Upper Keel deposit is found adjacent to the main decline, at a shallower depth than the
other existing zones. During initial construction of this main decline, muck bays were con­
structed to allow the staging of development muck. Muck Bays 2 and 4 will be re-purposed to
establish development drives to access the top and bottom portions of Upper Keel, respect­
ively. The proposed ventilation plan will utilize fresh air from the upper portion of the main
decline drawn off at Muck Bay 2, circulated through this zone, down a series of FARs, and
returned to the Upper Keel ramp to the main decline at Muck Bay 4. A schematic of the over­
all proposed ventilation network is shown in Figure 1.

DOI: 10.1201/9781003429241-66

642
Figure 1. Eagle mine overall ventilation schematic outlining the ventilation of Eagle, Eagle East and
Upper Keel zones.

3 UPPER KEEL LIFE-OF-MINE VENTILATION:

A ventilation schematic showing the ventilation of the Upper Keel is shown in Figure 2. Air
for the Upper Keel will be drawn from the Main Decline at Muck Bay 2 by the new Upper
Keel Booster Fans. Fresh air will be delivered to the Upper Keel through a set of internal
fresh air ventilation raises. All the air will exit into the Upper Keel haulage ramp and exhaust
air will travel up the decline to rejoin the remaining air from the portal that has not been
directed to the Upper Keel.

Figure 2. Upper Keel Life-of-Mine ventilation schematic.

643
 The ramp section between Muck Bay 2 and 4 will have a reduced flow rate due to the
Upper Keel ventilation supply “bypassing” this area. Appropriate measures will ensure portal
air flow is greater than the Upper Keel demand resulting in s sufficient air flow to support
trucking in this ramp.
The proposed Upper Keel ventilation design draws and returns air to the same branch of
the primary ventilation circuit, creating an independent circuit resulting in:
1. No impact on fan flows or pressures within the Upper Keel.
2. Exhaust air from Muck Bay 4 will recirculate back up to Muck Bay 2 and through the
Upper Keel zone.
3. No increase or decrease in air flows or pressures in the Eagle primary circuit or fan stations.
To control the ventilation within the Upper Keel zone and to maintain a simple control
schematic, all the fresh air from the raises will be directed into the Upper Keel ramp on the
lowest mining level by constructing timber bulkheads which will block airflow from each of
the internal levels with the exception of the lowest one. Air from the lowest level will enter the
Upper Keel ramp which will upcast toward the main decline. Auxiliary fans will be placed in
the ramp to ventilate each active level.

4 UPPER KEEL BLAST MODE VENTILATION

As seen in the schematics above, exhaust air from Upper Keel rejoins the fresh air circuit for
Eagle and Eagle East. On-shift blasting could allow development crews to maximize develop­
ment rates in Upper Keel, where crews can blast at the face as soon as drilling and loading are
done and not wait until the end of the shift blasting window. Therefore, to enable on-shift
blasting in the Upper Keel, consideration must be made to isolate blasting gases to avoid
fumes re-entering the main decline and downstream workplaces. To achieve this, a single 54”
steel exhaust duct is proposed to pick up the exhaust air from the Upper Keel decline connec­
tion in Muck Bay 4 and direct it to the 265 RAD and through the airlock doors. Planning for
blast gas clearing was informed by previous testing (Carriere et al. 2017).
This configuration is chosen based on the historical development of both Eagle and Eagle
East declines that were completed with a similar single 54” steel duct. Thus, the headroom in
the ramp is sufficient so that the duct can be installed without impacting haulage activities.
The configuration of the fans at the 265 RAD airlock doors (Figure 3) was used during the
development of the Eagle East decline and should be familiar to site operations.

Figure 3. Eagle East development duct arrangement at 265L RAD Airlock.

644
 A ventilation schematic showing the ventilation flows during blast clearing and the general
airflow is shown in Figure 4. Following an on-shift blast, the Upper Keel Blast Fans located
at the 265L RAD will be turned on to draw all the air from the Upper Keel straight to
exhaust. To ensure that no blasting gases enter the main decline, the fresh air booster fans at
the top of the raise in Muck Bay 2 will be turned down and a door installed in Muck Bay 4
will be closed. This will completely isolate the Upper Keel and ensure that any fresh air drawn
through the portal and down the Eagle decline will be free of blasting gases, allowing produc­
tion activities to continue in either Eagle or Eagle East without delay.

Figure 4. Upper Keel Blast mode ventilation schematic.

The 80 kcfm airflow volume during blast mode is based on the practical amount of air that
can be induced at the 265L RAD by two 54”200 hp fans without the use of inline booster
fans. Having no in-line booster fans in the exhaust duct will keep all parts of this system under
negative pressure, ensuring that any leakage at duct joints or from mobile equipment damage
to a duct would not result in the escape of blast smoke.
Instrumentation for remote monitoring and control will be used for the Upper Keel Booster
Fans, Blast Door, and Upper Keel Blast Fans to ensure the safe and stable operation of the
blast mode system. In the event of any upset conditions, alarms will be programmed in the
mine’s control system, and appropriate system interlocks, including the shutdown of Upper
Keel Booster Fans, which can be programmed for additional layers of redundancy.

5 UPPER KEEL ACCESS VENTILATION STRATEGY

Initial development from Muck Bay 4 into the Upper Keel decline is proposed to occur under
a traditional forcing auxiliary ventilation arrangement until the steel exhaust duct would not
be damaged by concussion and fly rock. These initial development blasts will require end-of-
shift blasting allowing blast smoke to clear through the mine’s primary ventilation circuit,
without impacting Eagle and Eagle East operations.
The blast mode exhaust duct can be installed after development has commenced; at which
time the remaining development can utilize on-shift blasting. A fresh air fan will be required

645
to direct air to the working face and flush blast gases back up the ramp, where the exhaust
duct will collect all exhaust and direct it through the 265L RAD Airlock to the RAR. The
fresh air fan should be installed in an overlapping configuration with the intake of the exhaust
duct to prevent recirculation of blast smoke into the intake duct (Figure 5). The ducting over­
lap distance has not yet been calculated – and may be limited by available space.

Figure 5. Exhaust overlap ventilation in Muck Bay 4.

After completion of the first internal FAR, the Upper Keel Booster Fans will be installed,
and flow-through ventilation will be instituted. The construction of the Blast Door should be
scheduled to complete prior to the start-up of the Upper Keel Booster Fans.

6 MAJOR VENTILATION EQUIPMENT CONCEPTUAL DESIGN HIGHLIGHTS

The Upper Keel Booster Fan will be the driver of the ventilation circuit that begins at the top
of the FARs in Muck Bay 2. An example sketch of a potential installation for the fans is
shown in Figures 6 and 7. The fresh air booster fans will be controlled by a variable frequency
drive (VFD). The VFD will be networked to allow remote control of the fans from the
surface. This will be critical to ensure fan turndown is achieved during blast clearing.
In the blast clearing mode, the Upper Keel ventilation flow is matched to the capacity of
the exhaust duct at 80 kcfm. Based on this balance, all of the blasting fumes should be cap­
tured before entering the main ramp. As an additional measure of safety, a ventilation door
(blast door) is proposed for Muck Bay 4 which will prevent an imbalance of flows and prevent
any fugitive blasting gasses from reaching the main ramp. This door will normally be open
during production and will not affect truck haulage. In blast mode, the doors will close and all
air from the zone will be captured in the exhaust duct. The blast door will be required only
after the ramp and Upper Keel FAR are connected and a flow-through circuit is made.
An overhead steel sectional door, comparable to those currently used on 265 RAD, is pro­
posed for the blast door. The door system is shown in Figure 7. A higher-pressure door has
been selected to withstand upset conditions in the case the Upper Keel Booster Fans speed is

646
Figure 6. Design Sketch for upper keel booster fan station - Plan View.

Figure 7. Design sketch for upper keel booster fan station – Section view.

647
not correctly reduced or if there is an exhaust duct fan failure. This pressure would be too
high for a rubber roll-up door.

7 CONCLUSIONS

Having the Upper Keel zone located adjacent to the primary mine access presents an oppor­
tunity for a simple ventilation circuit both in construction and operations phases. Challenges
for minimizing blast clearing time can be overcome with a reduced-flow blasting mode in the
zone and an isolated exhaust duct connection to the return air raise. If Operations elect to
pursue on-shift blasting, the addition of a blast isolation door can provide a failsafe
configuration.

REFERENCES

Witow, D., Gobbs, C., Lam, J. & Harris, W. (2019) Lundin Eagle East Project – Mine Ventilation
Update. 17th North American Mine Ventilation Symposium. Canada.
Carriere, R., McGuire, C., McLaren, E., Witow, D. (2017) Studying Operational Improvements in Blast
Gas Clearing Using Ventilation Control. 16th North American Mine Ventilation Symposium. Colorado,
USA.

648
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Empirical and numerical investigation on the optimal length of

eddy airflow in dead-end tunnel

R. Morla & J. Chen

Norton Gold Fields, Kalgoorlie, WA, Australia

S. Karekal & A. Godbole

School of Civil, Mining, and Environmental Engineering, University of Wollongong, Australia

P. Tukkaraja
Department of Mining Engineering and Management, South Dakota School of Mines, Rapid City, SD, USA

P. Chang
WA School of Mines: Minerals, Energy and Chemical Engineering, Curtin University, Australia

ABSTRACT: This paper uses vortex flow modelling to find the optimal length of eddy air­
flow in dead-end gallery using air velocity, pressure and diesel particulate matter (DPM) simu­
lations and field investigations. Computational fluid dynamics (CFD) modelling conducted
for four different dead-end crosscut lengths (10 m, 15 m, 20 m, and 25 m), three different
crosscut angles (45°, 90° and 135°) and different air velocities in adjacent galleries revealed
that a distinct vortex flow develops in the dead-end crosscut. Results indicated that an eddy
airflow revolved in a curved form, while the air velocity and pressure decreased towards the
centre of the vortex and DPM concentration increased towards the centre of the vortex. The
eddy airflow influence distance in a dead-end crosscut depends on the crosscut angle and air
velocity in the adjacent gallery. If the air velocity in the adjacent gallery is one m/s and the
crosscut angle is 90°, eddy airflow ventilates up to 20 m from the entrance. Though the air
velocity in the adjacent gallery is 4 m/s, eddy airflow is not ventilating the crosscut after
30 m from the crosscut entrance. The eddy flow distance is lower in obtuse-angled crosscuts
than the acute-angled crosscuts.

1 INTRODUCTION

Dead-end workings are common in underground areas during development and production
mining (García-Díaz et al., 2019). Dead-end crosscuts adjacent to the main airways, like
decline, incline, tunnel etc., are commonly used in the subsurface to install substations,
pumps, stockpiles, etc. (Morla et al., 2021, Toraño Álvarez et al., 2002). Figure 1 shows a part
of a mine near portal. From the figure, it can be observed that the substation, stockpile, mono
pump, fuel bay, underground (u/g) toilet, sump, u/g store, etc. are in dead-end tunnel/drive/
cuddy/crosscuts/gallery. These areas are located adjacent to the main air gallery and most of
them are ventilated by eddy airflow. If the dead-end cuddy length is too long, eddy air flow
may not reach the endpoint of the gallery, causing accumulation of higher temperatures or
dust or DPM concentrations.
Diesel-powered vehicles are sometimes operated in unventilated areas like parking cuddies,
stockpiles, etc. (Morla et al., 2020a). As the lower airflow in dead-end workings, it takes longer
time to disperse exhaust fumes, including DPM. The DPM particles are made of carbon,

DOI: 10.1201/9781003429241-67

649
elemental (EC) and organic (OC), ash, sulphates and nitrates (Chang et al., 2019), particles are
small in size (nm) and their densities are low (0.3 gm/cm3 to 1.2 gm/cm3) (Bugarski et al., 2004),
they do not tend to settle quickly under their own weight (Morla et al., 2020a, Morla et al., 2019).
Prolonged exposure to DPM causes adverse health effects (Ristovski et al., 2012, AIOH, 2013,
Morla and Karekal, 2017, Morla et al., 2018, Chang and Xu, 2017).

Figure 1. A part of a mine with details.

As per Australian mine regulations (AIOH, 2013, MDG, 2008, WHSR, 2022) to minimize
the adverse health effects, the 8-hour time-weighted average exposure of mine personnel to
EC is limited to 0.1 mg/m3, total carbon (TC), 0.16 mg/m3 and diesel particulate (DP), 0.2 mg/
m3. As per the Western Australian mines regulations, to dilute diesel emissions, each location
where a diesel engine operates must be ventilated with a minimum airflow of at least 0.05 m3/
s/kW of the engine capacity (WHSR, 2022). Western Australian mine regulations also indicate
(WHSR, 2022), an average air velocity of at least 0.3 m/s is to be supplied at all working areas
of the mine where persons work.
The design of eddy airflow ventilated dead-end tunnels is important to maintain regu­
latory ventilation requirements. Mapping airflow patterns in dead-end crosscuts will
better understand air distribution and develop the optimum design of eddy flow venti­
lated tunnels. In this paper, studies were conducted using air velocity, pressure and
DPM field and modelling investigations to find the optimum length of eddy airflow in
a dead-end crosscut.

2 MATERIALS AND METHODS

2.1 Details of the field experiment

Field experiments were conducted in an underground mine to study the airflow and DPM
distribution in dead-end crosscuts. The airflow of the intake airway was controlled by
a regulator located at the return side of the airway. A calibrated ‘Airtec’ real-time DPM
monitor was used for this field study (Janisko and Noll, 2008, Khan, 2017). The flow rate of
the DPM monitoring instrument was adjusted to 2.83 × 10-5 m3/s (1.7 L/min). During the
investigation, the air velocity in the adjacent main gallery was maintained at 3 m/s. Air vel­
ocity was measured with an anemometer. The experimental site was a 100 m long tunnel
with a rectangular cross-section (width 6 m, height 2.7 m), with a dead-end crosscut
10 m long and at 90° with the main gallery. For the DPM experiment, a loader was operated
in dead-end crosscut up to DPM concentration reached to 820 µg/m3. Though the regulatory
limit of EC is 100 µg/m3, the experiment was conducted at a high concentration to precisely
monitor the DPM dispersion with time.

650
2.2 CFD modelling
CFD studies have been used in the mining industry to solve various critical ventilation issues
(Morla, 2013, Morla et al., 2013, Morla et al., 2015, Chang et al., 2020a, Chang et al., 2020b,
Chang et al., 2019a, Chang et al., 2019b). The commercially available CFD package ANSYS
Fluent was used for this DPM modelling (Tanguturi et al., 2013). The CFD simulations were
carried out in the sequence of steps outlined in the following paragraphs.
Figure 2 shows a CAD model representing the experimental gallery and the dead-end cross­
cut. The sampling points are at 5 m from the dead-end face. Figure 2 shows the mesh gener­
ated for the volume of the experimental gallery. The finer mesh cells were also used adjacent
to gallery walls, with seven layers of cells accommodated in the boundary layers. The compu­
tational domain and mesh consist of almost half a million tetrahedron-shaped computational
cells. Convergence and mesh-independent studies were conducted in this modelling. The
residual RMS error value is 10-4, and the domain has imbalances of less than 0.01. The min­
imum wall Y+ value for the model is 1.
The boundary conditions of the model were considered as having an intake air velocity of 3 m/s
at 300 K of dry bulb temperature and the initial DPM concentration in the dead-end crosscut was
considered as 820 µg/m3. The transient state modelling was used to model DPM concentration dis­
persion in a dead-end gallery. For this modelling, the diameters of the DPM particles are considered
between 1e-9 m to 1e-7 m with a mean diameter of 1e-8 m. DPM particles were treated as inert mater­
ials and the Rosin-Rammler diameter distribution was used. For physical models, the spherical drag
law was used as a drag parameter. As the air velocity is low (below 4 m/s), the standard k-ε turbulent
model was used. For stochastic tracking, a discrete random walk model with 10 tries and a 0.15-time
scale was used. The intake air and DPM are considered as two different phases. The Eulerian-
Lagrangian approach is used whereby the gas phase (air) was solved using the Eulerian approach,
and the particle-phase (DPM) was tracked using the Lagrangian approach.

Figure 2. Computational domain and sampling points locations.

The airflow in the tunnel was treated as a turbulent flow. The standard k-ε model, which is
one of the most commonly used turbulent models, is applied to simulate the airflow. The model
transport equation for k was derived from the exact equation, while the model transport equa­
tion for ε was obtained using physical reasoning and bears little resemblance to its mathematic­
ally exact counterpart. In the derivation of the k-ε model, the assumption is that the flow is fully
turbulent, and the effect of molecular viscosity is negligible. As the mine air is considered as
fully turbulent flow, the k-ε model is valid for mine air. The turbulent kinetic energy, k, and its
rate of dissipation, ε, are obtained from the following governing equations (ANSYS, 2013):

651
Figure 3. Meshed model of dead- end gallery.

where Gb is the generation of turbulent kinetic energy due to buoyancy, and Gk is the produc­
tion of turbulent kinetic energy due to the mean velocity gradient.
The particle flow is modelled using the Euler-Lagrange approach, where particle properties
are studied along the particle flow path (Chang et al., 2019a). These models define particle
flow by considering the various forces that act on the particle (Thiruvengadam et al., 2016).
The forces commonly encountered are the drag force between the fluid and the particle, the
lift force, the virtual mass force, etc.
The fluid phase is treated as a continuum by solving the Navier-Strokes equations, while
the dispersed phase is solved by tracking a large number of particles, bubbles, or droplets dis­
persed through the calculated flow field. The dispersed phase can exchange momentum, mass,
and energy with the fluid phase. DPM particles are tracked using the Lagrangian method in
the discrete phase, and the particle or droplet trajectories are computed individually at speci­
fied intervals during the fluid phase calculation. The force balance equation relates the particle
inertia with the forces acting on the particle and can be written as

�! �
Where ~ F is an additional acceleration (force/unit particle mass), FD ~
u up is the drag
force per unit particle mass, and

Here ~ u is the fluid phase velocity, �!up is the particle velocity, µ is the dynamic viscosity of
the fluid, ρ is the fluid density, ρp is the density of the particle material, and dp is the particle
diameter. Re is the relative Reynolds number, which is defined as:

The additional forces induced on the particle due to the fluid surrounding the particle due
to growth in the boundary layer is called the virtual mass force and is given by

652
Where Cvm is the virtual mass factor with a default value of 0.5, the fluid and the particle are
coupled together mathematically in the form of slip velocity.

3 RESULTS AND DISCUSSIONS

3.1 Results of field experiment and model validation for air velocity
Figure 4 shows the results of the airflow distribution at a cross-sectional area in the crosscut.
Lower air velocity is at the middle of the gallery and air velocity towards the gallery sides is
increasing. Table 1 compares the results of field experiments and modelled air velocities at vari­
ous locations shown in Figure 4 for validation purposes. To measure air velocity in the mine,
a rotating vane anemometer was used. This anemometer can measure the minimum air velocity
of 0.15 m/s; therefore, the field measurements did not consider air velocity below 0.15 m/s. It
can be observed from Table 1 that the base case simulated results were in good agreement with
the measured data in most cases. Some discrepancies between the simulated and measured
results can be due to the unevenness in the gallery wall surfaces that was not considered while
modelling. Overall, the difference varies from – 10% to + 8.6%, which are acceptable. From the
modelling studies, it is also observed that center of the vortex is 5 m from the entrance.

Figure 4. Results of 10 m crosscut velocity contours.

Table 1. Comparison of simulated DPM results with experimental results.

Sample Measured air Modelled air Difference Sample Measured air Modelled air Difference
point velocity (m/s) velocity (m/s) (%) point velocity (m/s) velocity (m/s) (%)

a 0.58 0.63 8.6 f 0.51 0.50 -2

b Error 0.03 N/A g 0.40 0.36 -10
c 0.54 0.53 -1.8 h Error 0.06 N/A
d 0.42 0.43 2.3 i 0.5 0.48 -4
e Error 0.05 N/A

Note: Difference % is the difference between simulation results and test results and is calculated as (Simulated
value– Experimental value)/ experimental value) × 100.

653
3.2 Vortex flow modelling of air velocity and pressure with different crosscut lengths and angles
After the initial dead-end crosscut modelling studies were conducted with a base case model,
detailed parametric investigations were carried out on the airflow patterns in unventilated
dead-end crosscuts of different geometries (crosscut lengths, orientation with respect to the
main airflow gallery and air velocities).
Figure 5 shows the CFD modelling results of air velocity distribution for dead-end crosscuts
10 m, 15 m, 20 m, and 25 m deep and two different orientations (45° and 135°) were con­
sidered for the 20 m long dead-end crosscut. For this simulation, intake air velocity was
assumed to be 3 m/s. The vortex flow patterns are plotted on a plane 1.2 m above the ground.
The air velocity in the crosscut is relatively low and the velocities are lower than 0.1 m/s in
some areas, as shown in Figure 5. The lower air velocities are at the centre of the vortex and
air velocity increases towards the periphery of the vortex. The locations of the centre of the
vortex for the 10 m, 15 m, 20 m and 25 m long crosscuts at 90° to the main gallery were found
to be at 4.5 m, 8.5 m, 10 m and 10 m from the crosscut entrance. The centres of the vortices in
the 20 m long crosscut oriented at 45° and 135° were found to be at 10 m and 6.5 m, respect­
ively, from the crosscut entrance.
Figure 6 shows the pressure distribution in a dead-end crosscut. For this modelling con­
sidered dead-end crosscuts of 10 m, 15 m, 20 m (90°), 25 m, 45° (20 m) and 135° (20 m), and
an air velocity in the main gallery of 3 m/s. The results show that air pressure is reducing
towards the centre of the vortex. There is higher air pressure in the main air gallery; the pres­
sure difference between the main air gallery and the dead-end crosscut (ΔP) is very low and is
below 0.4 Pa. A higher air pressure exists in the main air gallery.

Figure 5. Air velocity streamline patterns in dead-end crosscut.

3.3 Results of field experiment and model validation for DPM

Figure 7 shows the results of DPM distributions at a cross-sectional area in the crosscut.
Higher DPM concentration is at the middle of the gallery and DPM concentration towards
the gallery sides is decreasing. Table 2 compares the results of field experiments and modelled
air velocities at various locations shown in Figure 7(b) for validation purposes. To measure
DPM concentration in the field, a real-time DPM monitor was used. It can be observed from
Table 2 that the base case simulated results were in good agreement with the measured data in
most cases. Some discrepancies between the simulated and measured results can be due to the
unevenness in the gallery wall surfaces that was not considered while modelling. Overall, the

654
Figure 6. Pressure streamlines pattern in dead-end crosscut.

difference varies from – 2.5% to + 2.3%, which are acceptable. From the modelling studies, it
is also observed that center of the vortex is 5 m from the entrance Figure 7(a).

Figure 7. Results of 10 m crosscut DPM concentration contours.

Table 2. Comparison of simulated DPM results with experimental results.

Measured DPM Modelled DPM
Sample point concentration (µg/m3) concentration (µg/m3) Difference (%)

a 551 564 2.3

b 650 648 0
c 554 540 -2.5

Note: Difference % is the difference between simulation results and test results and is
calculated as (Simulated value– Experimental value)/ experimental value) × 100.

3.4 Vortex flow modelling of DPM concentration with different crosscut lengths and angles
Figure 8 shows the DPM distribution in different dead-end crosscuts. For this modelling,
dead-end crosscuts of 10 m, 15 m, 20 m (90°), 25 m, 45° (20 m) and 135° (20 m), an air velocity
of 3 m/s in the main gallery, and an initial DPM concentration in the dead-end crosscut of

655
820 µg/m3 were considered (Morla et al., 2020b). In all cases, transient flow modelling studies
were used for 180 sec for this modelling. A higher DPM concentration is observed at the
vortex centre and DPM concentration reduces towards the ends of the vortex. The DPM con­
centration is higher in obtuse-angled crosscuts than the acute-angled crosscuts.

Figure 8. DPM concentration streamline patterns in dead-end crosscut after 180 sec.

Figure 9 shows the changes in DPM concentration concerning air velocity inside the dead-
end crosscut. DPM concentration increases with decreasing air velocity. A high DPM concen­
tration is at the centre of the vortex. Figures 6 and 8 show changes in DPM concentration
concerning pressure differences between the main air gallery and dead-end gallery (ΔP). DPM
concentration is increases with decreasing air pressure. A high DPM concentration is at the
vortex centre, and the lower pressure drop is at the centre of the vortex.

Figure 9. Changes of DPM concentration with air velocity for 10 m, 15 m, 20 m, 25 m, 45° and 135°
dead-end crosscut.

656
3.5 Vortex flow modelling with different air velocities in adjacent gallery
Investigations have been conducted with different air velocities in dead-end crosscut’s adjacent gal­
lery. For this modelling considers a 90° dead-end crosscut of 50 m long, and air velocities in the
adjoining gallery are 0.5 m/s, 1 m/s, 2 m/s, 3 m/s and 4 m/s. Figure 10 shows the results of simula­
tions studies; if the air velocity in the adjacent gallery is 0.5 m/s, vortex-shaped airflow is spread up
to 15 m in the dead-end gallery. If the air velocity in the adjacent gallery is 1 m/s, 2 m/s, 3 m/s and
4 m/s, eddy airflow distance in the dead-end gallery is 20 m, 23 m, 27 m and 30 m, respectively.

Figure 10. Air velocity contours in 50 m long dead-end crosscut with different air velocities in adjacent
gallery.

4 CONCLUSIONS

Investigations concluded that eddy airflow was formed near the dead-end crosscut entrance and
revolved in a curved form, while the air velocity and pressure decreased towards the centre of the
vortex. DPM concentration increased towards the centre of the vortex. The eddy airflow distance
and velocity in a dead-end crosscut depend on the crosscut angle and air velocity in the adjacent
gallery. If the air velocity in the adjacent gallery is 1 m/s and the crosscut angle is 90°, the eddy
flow is up to 20 m from the entrance. The eddy airflow distance is lower in obtuse-angled crosscuts
than the acute-angled crosscuts. Though the air velocity in the adjacent gallery is 4 m/s, eddy flow
is not ventilating the crosscut after 30 m from the crosscut entrance. This research provided clear
information about ventilation distribution and DPM particle flow patterns in dead-end crosscuts.

REFERENCES

AIOH 2013. Diesel particulate matter Occupational Health Issues.: (The Australian Institute of Occupa­
tional Hygienists, Inc).
Bugarski, A., Schnakenberg, G., Noll, J., Mischler, S., Patts, L., Hummer, J., Vanderslice, S., Crum, M.
& Anderson, R. The Effectiveness of Selected Technologies in Controlling Diesel Emissions in an
Underground Mine - Isolated Zone Study at Stillwater Mining Company’s Nye Mine. 2004. 1–86.
Chang, P. & Xu, G. 2017. A review of the health effects and exposure-responsible relationship of diesel
particulate matter for underground mines. International Journal of Mining Science and Technology, 27,
831–838.
Chang, P., Xu, G. & Huang, J. 2020a. Numerical study on DPM dispersion and distribution in an under­
ground development face based on dynamic mesh. International Journal of Mining Science and Tech­
nology, 30, 471–475.
Chang, P., Xu, G., Mullins, B., Abishek, S. & Sharifzadeh, M. 2020b. Numerical investigation of diesel
particulate matter dispersion in an underground development face during key mining activities.
Advanced Powder Technology, 31, 3882–3896.

657
Chang, P., Xu, G., Zhou, F., Mullins, B. & Abishek, S. 2019a. Comparison of underground mine DPM
simulation using discrete phase and continuous phase models. Process Safety and Environmental Pro­
tection, 127, 45–55.
Chang, P., Xu, G., Zhou, F., Mullins, B., Abishek, S. & Chalmers, D. 2019b. Minimizing DPM pollution
in an underground mine by optimizing auxiliary ventilation systems using CFD. Tunnelling and Under­
ground Space Technology, 87, 112–121.
MDG 2008. Guideline for the management of diesel engine pollutants in underground environments.
MDG-29.
Morla, R. 2013. Optimum Inertisation Techniques for Blasting Gallery Method. 3rd International Work­
shop on Mine Hazard Prevention and Control,. Brisbane: CSIRO.
Morla, R., Balusu, R., Tanguturi, K. & Khanal, M. 2013. Prediction and control of spontaneous com­
bustion in thick coal seams. Coal Operators conference. Wollongong: University of Wollongong.
Morla, R., Balusu, R., Tanguturi, K. & Ting, R. 2015. Inertisation options for BG method and optimisa­
tion using CFD modelling. International Journal of Mining Science and Technology, 25, 401–405.
Morla, R., Godbole, A., Karekal, S., Bhattacharjee, R. M. & Balasubrahmanyam, N. 2018. Fundamen­
tal understanding of diesel-operated man riding vehicle DPM dispersion–a case study. Journal of Sus­
tainable Mining, 17, 105–110.
Morla, R. & Karekal, S. 2017. Diesel particulate matter investigations in underground coal mines.
Morla, R., Karekal, S. & Godbole, A. 2020a. CFD simulations of DPM flow patterns generated by
vehicles in underground mines for different air flow and exhaust pipe directions. International Journal
of Mining and Mineral Engineering, 11, 51–65.
Morla, R., Karekal, S. & Godbole, A. 2020b. Investigation of DPM dispersion in unventilated dead-ends
using transient flow modelling. International Journal of Mining and Mineral Engineering, 11, 121–133.
Morla, R., Karekal, S., Godbole, A., Bhattacharjee, R. M., Nasina, B. & Inumula, S. 2019. Effect of
ventilation air velocities on diesel particulate matter dispersion in underground coal mines. Inter­
national journal of mining and Geo-Engineering, 53, 117–121.
Ristovski, Z. D., Miljevic, B., Surawski, N. C., Morawska, L., Fong, K. M., Goh, F. & Yang, I. A. 2012.
Respiratory health effects of diesel particulate matter. Respirology, 17, 201–212.
Thiruvengadam, M., Zheng, Y. & Tien, J. C. 2016. DPM simulation in an underground entry: Compari­
son between particle and species models. International Journal of Mining Science and Technology, 26,
487–494.
WHSR 2022. Work Health and Safety (mies) Regulations. In: AUSTRALIA, W. (ed.).

658
 Underground Ventilation – Tukkaraja (Ed)
© 2023 The Author(s), ISBN 978-1-032-55146-3

Author Index

Abil, A. 115 Demirkan, D.C. 377 Khademian, Z. 507, 529

Addis, J.D. 499, 507, 529 Dias, T. 411 Khaniani, H. 541
Adhikari, A. 451 Diaz, J.C. 423 Kimutis, R. 529
Afrouz, S. 230 Dougherty, H. 499, 529 Klima, S. 247
Agioutantis, Z. 423 Durieux, D.W. 34, 174, 183 Klose, F.K.R. 469
Ajayi, K.M. 499, 507, 529 Duzgun, H.S. 377 Kocsis, C. 256, 411
Akhtar, S. 569 Kumar, A.R. 285, 399
Allen, C. 60, 83, 163 Falk, L.K. 60, 83
Amoah, N.A. 274, 285 Fernandez, J. 341 Lee, T. 104
Androulakis, V. 541 Finn, J. 322 Li, Z.B. 439
Animah, F. 221, 230 Fox, J.E. 208, 462 Lindgren, E. 411
Arnott, W. 256 Friend, S. 104 Liu, S. 3
Atchley, G. 411 Fuster, E. 595 Liu, Y. 313
Lotero, S. 541
Bahrami, D. 607 Gangrade, V. 69, 499 Lutz, E.A. 95
Balusu, R. 363 Gendrue, N. 3
Barone, T.L. 104 Ghosh, A. 313 Madureira, E. 305
Beck, T. 247 Gobbs, C. 642 Magauiya, N. 115
Belle, B. 363 Godbole, A. 649 Mardon, M. 135
Bergh, F.S. 624 Greth, A. 221, 230 Martikainen, A.L. 469
Bhattacharya, S. 3 McGuire, C. 34, 183, 642
Bingham, B. 256 Habibi, A.A. 25, 120, 135 McLean, B. 129
Black, D. 439 Harb, C. 305 Medina, A. 305
Bogin, G. 377 Hardcastle, S.G. 163 Mehedi, T. 183, 635
Bowling, J. 353 Harris, M.L. 499, 507, 529 Mochubele, M. 624
Boyd, K. 635, 642 Hassanalian, M. 541 Mohit, M. 569
Brake, D.J. 589 Hines, J. 330 Morla, R. 649
Brickey, A.J. 484 Homan, K.O. 120 Murphy, C. 256
Brown, M. 34 Hristopulos, D.T. 423
Brune, J. 264, 377 Hummer, J.A. 104 Nascimento, P. 256
Buaba, J.A. 484 Hurtado-Cruz, J.P. 45 Newman, A. 484
Bugarski, A.D. 104, 120 Ngcibi, A.K. 624
Burgess, J.L. 95 Iqbal, A. 393, 399, 578 Nunes, N. 411
Nurshaiykova, G. 115
Calizaya, F. 411 Jayaraman Sridharan, S. Nyqvist, J. 54
Carvajal-Meza, M.A. 45 192, 451
Chang, P. 313, 649 Jiang, H. 247 Osho, B. 256
Chen, J. 649 Johnson, J. 411
Chow, C. 330 Jones, T.H. 469 Pandey, A. 192
Clayton, A. 330 Juganda, A. 377 Peña, I. 615
Connot, J. 451 Juric, J. 439 Pilkington, E. 183
Pinheiro-Harvey, M. 635
Das, M.C. 296 Karekal, S. 649 Pinto, E. 135
Dasgupta, S. 54 Kaufman, M. 60 Potts, J.D. 239
De Souza, E. 15 Keles, C. 221, 230 Poulos, J. 411

659
Prabhu, E. 462 Sarver, E. 221, 230 Vanderslice, S. 104
Prabhu, M. 462 Sasmito, A.P. 569 Vanegas, A. 305
Prasojo, R. 25 Sastry, B.S. 192 Vaze, M. 54
Pushparaj, R.I. 393, 399, Schafrik, S. 423
578 Schatzel, S.J. 499, 507, 529 Wallace, K.G. 615
Schult, G. 353 Wang, X. 256
Qiao, M. 330, 439 Setiawan, I. 25 Watkins, E. 499, 529
Qureshi, A. 115 Shao, S. 541 Watson, K. 129
Shaw, J.K. 60, 163 Wijayanto, K. 135
Raj, K.V. 69 Sidrow, E. 264 Witow, D. 34, 183,
Rajapaksha, R. 305 Slouka, S. 264 635, 642
Rakhimov, D. 115 Soles, J. 384
Rawlins, C.A. 149 Stachulak, J. 129 Xu, G. 274, 285, 393, 399,
Rawson, T. 305 Stewart, C. 25, 515 578
Ray, R.E. 595 Xu, M. 569
Reed, R.J. 95 Tang, W. 384
Reed, W.R. 239 Taylor, S. 256 Yan, L. 550, 559
Ren, T. 330, 439 Thakur, P.C. 431 Yantek, D.S. 550, 559
Roberts, J. 330 Thomas, R.A. 384, 607 Young, D. 129
Roghanchi, P. 296, 305, Tom, K. 200 Yuan, L. 384
541 Torkmahalleh, M. 115
Rose, C. 135 Tsai, C. 264 Zaid, M.M. 274
Rubasinghege, G. 296 Tukkaraja, P. 451, 649 Zeinulla, A. 115
Rubeli, B. 129 Zhang, H. 83
Udofia, E. 484 Zhang, P. 499
Sabanov, S. 115 Uecker, L. 305 Zhao, Z. 313
Salami, O.B. 393, 399, 578 Zheng, Y. 239, 247
Salinas, V.P. 296 Van Diest, J. 353 Zhou, L. 607
Sandink, M. 256 Van Dyke, M. 499, 529 Zychowski, K. 296

660