Process Safety and Environmental Protection 9 3 ( 2 0 1 5 ) 173–181

Contents lists available at ScienceDirect

Process Safety and Environmental Protection

journal homepage: www.elsevier.com/locate/psep

Assessment of air dispersion characteristic in

underground mine ventilation: Field measurement
and numerical evaluation

Arif Widiatmojo a,∗ , Kyuro Sasaki a , Yuichi Sugai a , Yoshiaki Suzuki b ,

Hiroyuki Tanaka b , Kagemi Uchida b , Hiroyuki Matsumoto b
a Department of Earth Resources Engineering, Faculty of Engineering, Kyushu University, 744 Motooka, Nishi-Ku,
Fukuoka-Shi 819-0395, Japan
b Kushiro Coal Mine Co., Ltd., 5-2-23, Okotsu, Kushiro-Shi, Hokkaido 085-0811, Japan

a b s t r a c t

The environmental safety of an underground mine depends strongly on its ventilation system. An efficient venti-
lation system provides fresh air, removes hazardous gases and dust, and maintains the temperature and humidity
at appropriate levels. One of the most important factors in removing hazardous gases and dust is the dispersion
behaviour in the mine network. This factor determines the longitudinal spreading and the average air residence
time of gases or particulate matter throughout the mine. This paper describes tracer gas measurement in an under-
ground mine and the utilisation and analysis of the dispersion characteristics using numerical simulations. The
concentration–time curve obtained from the measurement is simulated to evaluate the effective diffusion coeffi-
cient that reflects the general dispersion characteristic of an entire mine. The evaluated values of effective diffusion
coefficient are then compared to other data from several studies. The diffusivities obtained in this study were higher
than other analytical and empirical results. More research is still required to identify the main factors causing such
higher diffusivities. However, the results from the present work can be an important standpoint for future work.
Numerical simulation conducted in this research was confirmed to be effective in detecting several leakage paths
occurring in the mine ventilation network.
© 2014 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: Tracer gas; Underground mine ventilation; Dispersion; Numerical simulation

1. Introduction or leakages that lead to power losses and inefficiency. This

Underground mines are traversed by numerous access lev-
els directly off the main ventilation shafts, which in turn are
intersected by raises, cross cuts or tunnels in a very complex
network. Ventilation has the vital aim of ensuring the safety
and health in underground working environments. One of the
roles of mine ventilation is to dilute and exhaust hazardous
gases, airborne dust and other particulate matter to comply
with the mine safety regulations. However, as mining progress
advances, at certain levels, the ventilation becomes non-
productive because some airflow routes have short-circuits
dynamic progress makes the mine ventilation network more
complicated and the possibility of recirculation or leakage
arises. Consequently, the prediction of the concentration of
contaminants in the ventilation network is practically diffi-
cult. Moreover, mine ventilation also addresses comfort and
economic constraints. For instance, excessive air motion will
sometimes cause discomfort because the total ventilation cost
must be as low as possible to keep the whole process econom-
ically feasible.
Conventional ventilation surveys using anemometers or
smoke tubes in the mine are sometimes inaccurate and

∗
Corresponding author. Tel.: +81 92 802 3329; fax: +81 92 802 3329.
E-mail addresses: arifw@mine.kyushu-u.ac.jp, widiatmojo arif@yahoo.com (A. Widiatmojo).
Available online 18 April 2014
http://dx.doi.org/10.1016/j.psep.2014.04.001
0957-5820/© 2014 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
174 Process Safety and Environmental Protection 9 3 ( 2 0 1 5 ) 173–181
inappropriate under certain conditions. In addition, the leak-
ages that may occur are difficult to determine and quantify.
The tracer gas method can be expected to overcome these
problems. This method can also be used to measure the vol-
umetric flow rate as long as the perfect mixing of tracer with
air flow is ascertained.
The dispersion phenomenon in underground ventilation
has an important role affecting particulates or gases spreading
from its source throughout the mine. Thus, certain measure-
ment and simulation methods are required to investigate their
effects on the mine environment and operation safety.
Dispersion or diffusion in an underground mine ventilation
network is very complex. The diffusivity is the total outcome of
turbulent eddy diffusion and turbulent mixing in the complex
interconnecting airways.
Airflows in most underground mines consist of many
branches that separate or re-join. Under certain circum-
stances, airflow cannot be studied by using the conventional
methods of ventilation survey such as anemometer or smoke
tube. For example, only leakage and recirculation rates in
mined areas and air exchange rates at ventilation headings or
a dead end can be estimated using these conventional meth-
ods.
Based on the purpose of tracer gas application in
underground mines, tracer gas measurements in a mining
environment may be classified into three main techniques;
dilution, decay and leakage. The principle of tracer dilution is
that the initial to final concentration of tracer gas is directly
proportional to the initial and final volume occupied.
The dilution technique can be broken down into static sys-
tems and dynamic systems. An example of a static system is
measuring the volume of a non-ventilated room in which a
fully mixed tracer gas shows a constant concentration. The
use of a tracer gas for determining the airflow rate is an
example of a dynamic system. Tracers may also be used to
determine the rate of dilution or exchange in a semi-enclosed
room by observing the tracer concentration that decreases
over time. In early applications, a tracer gas was used to eval-
uate the average air residence time in the stope. Currently,
tracer gases have various applications in underground mines,
such as leakage investigation, evaluation of air route, recir-
culation, dust control and methane control (Stokes et al.,
1987; Thimons and Kissell, 1974; Timko and Thimons, 1983;
Patterson and Luxbacher, 2012; Xu et al., 2013).
In this study, the tracer gas measurement was conducted by
transient release (pulse release). Principally, the measurement
was carried out by releasing tracer gas at an upstream position
instantaneously and recording the resulting concentration at
certain downstream positions. Fig. 1 shows the possibilities of
concentration–time curves that could be obtained from tracer
gas measurements with a pulse release method in a single
airway. A wider curve suggests higher diffusivity and is usually
followed by longer trailing behind due to very slow diffusion
Fig. 1 – Illustration of tracer gas diffusion within mine
airway by pulse release (Sasaki et al., 2009).
Fig. 2 – Illustration of tracer gas dispersion mechanism and
tracer gas measurement in underground mine ventilation
network.
of tracer gas. Theoretically, the less diffusivity, the closer the
concentration–time curve is to a perfect Gaussian distribution.
2. Methods
A tracer gas measurement using SF6 (sulphur hexafluoride,
non-toxic) gas was conducted at Kushiro Coal Mine Co., Ltd.,
in December 2011. Kushiro Coal Mine Co. Ltd. is located at
Kushiro City, Hokkaido (latitude: 42.959, longitude 144.422), in
northern Japan. The objective of the measurement was mainly
to obtain general longitudinal dispersion characteristics of the
entire ventilation network of the mine by releasing the tracer
gas from its intakes and measuring the arriving tracer concen-
tration at the main fan as illustrated in Fig. 2.
Currently, the ventilation system consists of an exhaust
shaft driven by 2 × 190 kW axial fans arranged in parallel,
delivering 66.67 m3 /s of fresh air from four intakes, namely
Harutori, Daini, Harutorishakou and Tuukishakou. From these
four intake shafts, only two (Harutori and Daini) provide
fresh air to the main working area. The other two intakes
(Harutorishakou and Tuukishakou) are used as access for the
conveyor belts and piping. Table 1 shows the ventilation air
balance from intakes and exhausts in 2011 while the distri-
bution of fresh air from each intake is presented in Table 2.
Compared to its size in 1997, when a similar tracer gas mea-
surement was conducted, the mine size and total flow rate
sustained a major reduction due to an operational change that
altered the flow network. Figs. 3 and 4 show the mine venti-
lation network in 2011 and 1997, respectively. In a previous
study with the same procedure as the 2011 measurement,
SF6 was released at the Okotsu intake and at the arriving
concentration at the Shireto main fan was monitored (Sasaki
et al., 2009). The coloured lines in Fig. 3 represent the air
routes from the Harutori and Daini intakes. The air from
both intakes was mixed long before it reached the working
area.
Tracer gas concentration measurements were carried out
using an SF6 gas monitor (Bruel & Kjaer Model 1302) with
a high measurement resolution (10 ppb) at approximately
50–60 s sampling intervals. The measurement principle is the
Photoacoustic Spectrometry (PAS) method that can be used
to measure SF6 and hydrocarbon gas concentrations. Com-
pared to traditional spectrometry, the photoacoustic method
has several advantages as described by Einfeld (1998).
 Process Safety and Environmental Protection 9 3 ( 2 0 1 5 ) 173–181 175

Table 1 – Comparison of ventilation network outlines of Kushiro Coal Mine in 2011 and 1997.
Measurement Tot. airway length (m) Tot. airflow (m3 /s) Number of main fans

2011 17,171 66.67 1

1997 34,210 320 2
Ratio 1:2 1:4.8 –

Table 2 – Measured airflows from intakes to an exhaust during measurements in 2011.

Remarks Name Airflow (m3 /s) Total (m3 /s)

Intake Harutori intake (#1) 40 63.64

Daini intake (#28) 10
Harutorishakou intake (#25) 6.67
Tuukishakou intake (#27) 6.67
Exhaust Harutorikou exhaust (#17) 66.67 66.7

Unbalance = 5%.

Fig. 3 – Schematic figure showing the Kushiro Coal Mine’s ventilation network in 2011 (plan view).
176 Process Safety and Environmental Protection 9 3 ( 2 0 1 5 ) 173–181

(4) Measure balloon diameter in three different traverse

directions and calculate the average diameter by
dave = (d1 + d2 + d3 )/3.

Considering the buoyancy effect imposed by the surround-

ing air to its surface, the true mass of gas in a balloon, wt (kg),
is calculated by:

wt = (wf − w0 ) + air Vb (1)

where air (kg/m3 ) is air density and Vb is the balloon volume

calculated by

Fig. 4 – Schematic figure showing the ventilation network 4

Vb = ␲d3 (2)
in 1997 (Sasaki et al., 2009). 24 ave

The SF6 volume inside a balloon, V (m3 ), is calculated by

wt
V= (3)
SF6

where the density of SF6 , SF6 (kg/m3 ), is calculated based on

the ratio between the molecular mass of SF6 , MRSF6 (kg/mol)
and air, MRair (kg/mol).
By taking air = 1.2 kg/m3 , MRSF6 = 0.146 kg/mol and
MRair = 0.029 kg/mol, the SF6 density, SF6 , is:

MRSF6
SF6 = air = 6.041 kg/m3 (4)
MRair

The SF6 was released at two intakes: the Harutori intake

Fig. 5 – Releasing SF6 at ventilation intake by breaking
balloons.
Figs. 5 and 6 show the photographs taken during release
and data acquisition from gas monitor, respectively.
Prior to the measurement, the preparation was conducted
as follows:
(1) Measure the mass of balloons, w0 (kg);
(2) Fill balloons with SF6 gas from the gas cylinder;
(3) Measure balloons mass after filling with SF6 , wf (kg);
and the Daini intake. To confirm the measurement consis-
tency, the releases were repeated twice at each intake. The
sampling point was located at the horizontal shaft close to
the main fan by using 40-m flexible tubing connected to a
long steel pitot tube. The pitot tube was inserted into the
shaft while the other side of the pipe was connected to the
gas monitor. The tracer gas was released when the detected
concentration of tracer gas from the previous release was at a
negligible level. This sampling technique requires 24 h of con-
tinuous operation of the gas monitor. For protection against
snowfall, the gas monitor was kept in a small warehouse (see
Fig. 6).
3. Numerical simulation

The governing advection–diffusion equation in cylindrical

coordinate is expressed as:

   
∂C ∂c ∂2 C ∂2 C 1 ∂C
+ Um = DL + Dr + (5)
∂t ∂x ∂x2 ∂r2 r ∂r

where DL and Dr are diffusion coefficients in the longitudinal

Fig. 6 – Monitoring of SF6 gas concentration.
and radial directions, respectively, Um is the mean velocity, C
is the tracer gas concentration, and x and r are distances in
the longitudinal and radial directions.
In a fully developed turbulent flow of an airway, the concen-
tration gradient in the radial direction is assumed to be much
smaller than the diffusive and convective transport in longi-
tudinal direction. This assumption may be valid only when
the longitudinal distance is much longer than the diameter of
the tunnel. Moreover, for the large size mine to be modelled,
it is impractical to consider two spatial coordinates due to a
computational constraint (Colella et al., 2009).
 Process Safety and Environmental Protection 9 3 ( 2 0 1 5 ) 173–181 177

i Eij j one particular junction. For example, if there are five junctions
that are connected to junction #15, we denote it as J(15, 1),
Lij J(15,2),. . ., J(15, 5). The actual junction numbers connected to
Umij junction #15 can be any number between 1 − NT , (except #15),
where NT is total number of junctions in the network. Thus,
the properties of the subsequent airways are following this
Δτ
system. Supposing that the tracer gas concentration arriving
at junction i is Ci (t) based on Kirchoff’s first law of nodal flow,
Ci Cj then the upstream concentration for the next calculation is
defined by:
κ=1 κ=1

time time t (i, k) Um (i, k) A (i, k)

Ci→J(i,k) (t) = Ci (t)
 n(i)
Fig. 7 – Illustration of numerical simulation using Eq. (8). (i, k) Um (i, k) A (i, k)
k=1
By using this assumption, Eq. (5) can be reduced to the 
following form: (i, k) = 1; Um (i, k) ≥0
× (9)
  (i, k) = 0; Um (i, k) < 0
∂C ∂C ∂2 C
+ Um = DL (6)
∂t ∂x ∂x2
The flow is positive (Um (i,k) ≥ 0) if the air is flowing from
Longitudinal diffusion in the mine ventilation network may node i to J(i,k) and vice versa. To reduce the calculation time,
vary, depending on the flow condition in each airway. In the concentration distribution at the separating node is stored and
present simulation, the longitudinal diffusion is regarded as used as a starting concentration for the other routes so that it
an effective diffusion coefficient, E (m2 /s), representing the is not necessary to recalculate from the release point.
average value for the whole network system. The concentra- Other assumptions used in the simulation are:
tion distribution within a single airway with length, L (m), can
be written as a solution of Eq. (6):
• The flow is steady incompressible Newtonian fluid;
 2
 • Mixing between air and SF6 is perfectly inert, so that the
AC(x, t) 1 {L − Um t}
= √ exp − (7) density difference between air and tracer gas are negligible;
V 2 Et 4Et
• There is no chemical reaction between tracer gas and ven-
tilation air or other properties;
where A (m2 ) is the cross sectional area of the airway at a
• The amount of SF6 in the flow is sufficiently small that all
measurement position.
physical properties of ventilation air are homogeneous;
A discretised form of Eq. (7) for each airway is:
• The mixing and spreading of tracer gas in ventilation air is
controlled mostly by a turbulent diffusion mechanism at a

t/
Ci→j () Qij high Reynolds number flow, therefore the difference from
Cj (t) =  1/2
2Aij Eij {t − } actual air density is negligible;
=1
 • The airflow quantity at each airway is constant and exter-
2

Lij − Umij {t − } nal disturbances such as mining equipment that sometimes
× exp −  (8) move in the airway are negligible.
4Eij {t − }

where Cj is the gas concentration at a downstream node, Ci (() The effect of different effective diffusion coefficients
is the gas concentration at an upstream node, Lij is airway (E = 2 m2 /s, E = 4 m2 /s, E = 8 m2 /s) at a straight-single airway,
length, Qij is airflow quantity, t (s) is elapsed time since releas- L = 100 m, Um = 1 m/s, V = 0. 05 m3 and d = 5.05 m, are shown in
ing and  (s) is calculation time step (see Fig. 7). The value of Fig. 9.
 must be considered small enough to minimise the error.
A scheme to treat the calculation of flow separation at the
junction was also considered in this numerical study. Let NT be 1.2
the total number of junctions connecting the airways, and the
E = 2 m2/s
junction number is denoted as i = 1 − NT , while the definition of 1
E = 4 m2/s
the kth junction, which is connected to node i is j = J(i,k) for k = 1 E = 8 m2/s
Concentration (ppm)

0.8
to n(i) and n(i), is the total number of junctions connected to
junction i. Fig. 8 shows the number of junctions connected to 0.6

0.4
2
0.2
1 n(j)

i j 0
0 0.5 1 1.5 2 2.5 3 3.5 4
Elapsed Time (minutes)
1
n(i) 2
Fig. 9 – Effect of axial effective diffusion coefficient for flow
Fig. 8 – Schematic definition of airways connectivities. condition: Um = 1 m/s, d = 5.05 m, L = 100 m, V = 0.05 m3 .
178 Process Safety and Environmental Protection 9 3 ( 2 0 1 5 ) 173–181

Table 3 – Comparison of SF6 volume between release and monitored concentration profiles.
Release position Measurement Released SF6 (m3 ) Calculated SF6 (m3 ) % Disagreement

Harutori inlet 1st 0.0195 0.0201 3.08

2nd 0.0180 0.0186 3.33
Daini inlet 1st 0.0192 0.0201 4.69
2nd 0.0200 0.0209 4.50

4. Results measurements are clearly wider in comparison to the 2011

The initial volume of tracer gas released from Daini and
Harutori inlets is presented in Table 3 in comparison with
the total calculated SF6 volume, based on the measured
concentration–time curve:
measurements, indicating longer air residence time and a
higher degree of diffusivity. In general, all of these concen-
tration profiles show a trailing effect at the end.
All of the results from the 2011 measurement show two
main peaks. The first peak was mostly leakage airflow of
 15.8 m3 /s in total (31.8% of total supplied fresh air from Daini
∞
V=Q Cdt (10) 8E-05
0
Measurement #1
Measurement #2
where Q (m3 /s) corresponds to the airflow rate at the 6E-05 Simulation
measurement position. The measured volume of SF6 was
calculated using the common Riemann method by dividing CA
Leakages
the concentration–time curve into small rectangular ele- V-1 4E-05 Routes Main Routes
ments, with the time interval similar to the sampling interval (m )
recorded by the gas monitor. All of the calculated SF6 volumes
2E-05
show slightly larger values compared to the initial measure-
ments. The reasons for these disagreements are due to the
error in the airflow measurement and truncation error as a 0
result of the fairly long sampling interval (50–60 s). However, 0 50 100 150 200 250
Elapsed time (min)
the error is considerably lower (less than 5%). In 1997, Sasaki
released 0.659 m3 of SF6 and calculated the SF6 volume as Fig. 10 – Tracer gas measurement results for release at
0.635 m3 (3.6% disagreement) (Sasaki et al., 2009). Daini inlet and numerical simulation using 1-D
The mine ventilation network connectivity and airway data advection–diffusion equation.
(e.g., cross sectional areas, average velocity, airway length)
8E-05
used for simulations are the latest data at the time of the Measurement #1
measurements and were supplied by the ventilation divi- Measurement #2
Simulation
sion of the mining company as presented in Table 4. Based 6E-05
on the evaluation of the raw ventilation network data, there
Leakages
are a total of 9 and 18 routes for airflow coming from Daini CA routes
4E-05
and Harutori intakes, respectively (including possible leakage V-1
(m )
routes). The value of the effective diffusion coefficient, E, was
adjusted accordingly to obtain the best matching curve with 2E-05 Main
routes
the measurement results. The best matching curves for both
releases (Daini and Harutori intakes) were obtained by giving 0
E = 45 m2 /s. For the case of the 1997 measurements carried out 0 50 100 150 200

by Sasaki et al. (2009), the matching curve was obtained by Elapsed time (min)
giving E = 200 m2 /s.
Fig. 11 – Tracer gas measurement results for release at
Harutori inlet and numerical simulation using 1-D
5. Discussion advection–diffusion equation.
4E-06
During the measurements, it was difficult to maintain the Simulation
same amount of SF6 to be released at each intake. To compare
Measurement
the results, the concentration is presented as the normalised 3E-06
value, CA/V (m−1 ). Figs. 10 and 11 are the simulation results CA
presented with the measurement results. The measurement V-1 2E-06
results for both Daini and Harutori are consistent, possibly (m )
implying that there were no significant factors such as the 1E-06
change of flow rate in the airways, or no mechanical distur-
bances affected the tracer gas route.
0
Comparing the Daini and Harutori releases, the shapes of 0 100 200 300 400
the concentration profiles are almost similar, but the differ- Elapsed time (min)
ence of arrival time is noticeable because these routes are
nearly the same except for the first long shaft connected to Fig. 12 – Measurement results from measurement in 1997
both intakes. The measurement and simulation results for and numerical simulation using 1-D advection–diffusion
the 1997 measurements are presented in Fig. 12. The 1997 equation.
 Process Safety and Environmental Protection 9 3 ( 2 0 1 5 ) 173–181 179

Table 4 – Network connectivity, airways and flow properties for measurement in 2011.
Node #from Node #to Flow rate Q(m3 /s) Area A (m2 ) Length (m)

1 2 40.0 15 525
2 3 36.9 15 1020
3 4 50.0 11.2 150
4 5 23.3 11.6 80
5 6 5.0 3.2 880
6 8 5.8 10 880
8 9 9.2 10.8 130
9 16 20.0 10.5 805
5 8 7.5 11.2 150
5 11 10.8 11.6 150
11 12 5.0 3.2 900
12 13 5.0 10 900
13 9 10.8 12 130
11 13 5.8 11 100
28 29 10.0 17 645
4 23 26.7 9.9 600
23 14 7.5 10.2 750
14 15 7.5 10.2 65
15 16 28.3 14.6 215
23 18 19.2 3.2 150
18 24 2.5 3.2 65
8 24 3.3 10 175
18 19 16.7 14.3 953.8
19 21 16.7 7 225
21 22 16.7 14.3 1176
24 22 5.8 10.5 250
25 26 5.8 20 1460
27 26 7.5 5 300
26 16 13.3 12.8 570
16 17 63.3 13.4 825
22 15 22.5 11.3 340
29 3 13.1 14 1255
2 29 3.1 13 30

and Harutori intakes), while the second peak was airflow function of mean velocity and the friction factor, f(−) of the
through the working face, development headings and other flow. The turbulent friction factor can be obtained from:
areas of the mine. Further investigation on ventilation data,
coupled with simulation and measurement results, revealed f
u∗ = Um (12)
that the leakages most likely occurred at the ventilation ducts 8
used to supply the coal face (leakage #18–#24) and the devel-
opment heading (leakage #11–#13 and leakage #5–#8). This Eq. (11) is valid only when the laminar sub-layer and tran-
leaked air flows towards the exhaust shaft due to a strong pres- sition buffer layer are smaller in thickness (Re > 20,000).
sure imposed by the main fan. The two dashed curves in both The result of evaluated effective diffusion coefficients
Figs. 10 and 11 are the breakdown of the simulation to show based on 1997 and 2011 measurements is shown in Fig. 13.
the leakage path and the main routes, whereas the solid line These results are compared against the Reynolds number to
is the superimposed curves with these two. accommodate different types of fluids used in experiments.
One of the pioneers of diffusion research in turbulent pipe The results show higher values than empirical Eq. (13) by Sittel
flow was Taylor (1954), who combined the effect of radial dif-
fusion with the longitudinal diffusion. He used the Reynolds
analogy that states that radial diffusivity is analogous with
heat and mass transfer in turbulent flow, as well as with
transfer of fluid momentum. He neglected the contribution
of molecular diffusion in both the radial and the axial direc-
tions, which are negligibly small in comparison with turbulent
eddy mixing diffusion in high Reynolds numbers. The univer-
sal velocity profile is used to analyse axial mixing. He proposed
the following relationship to express the turbulent diffusion
coefficient, DL (m2 /s), in the longitudinal direction:

DL = 5.05du∗ (11)

Fig. 13 – Relationship between normalised values of

where d (m) is the airway diameter of the pipe and u* (m/s) longitudinal diffusion coefficient with the Reynolds
is the turbulent friction velocity in an arbitrary sub-layer as a number.
180 Process Safety and Environmental Protection 9 3 ( 2 0 1 5 ) 173–181
Fig. 14 – Evaluated effective diffusion coefficient against
(Um d) from 1997 and 2011 measurement compared with
Taylor’s equation (11).
et al. (1968), empirical Eq. (14) by Wen and Fan (1975) and Tay-
lor’s equation (11) for f = 0.015–0.05.
DL = 1.13 × 10−6 Re0.875 (13)
DL 3.0 × 107 1.35
= + 0.125 (14)
Um d Re2.1 Re
The results of Widodo (2008) for laboratory scale measure-
ment in straight pipe of 30 m length and 0.025 m diameter
show relatively good agreement with Taylor’s analytical equa-
tion (11). Thus, Taylor’s theory for the smooth-straight pipe is
supported. According to the results by Davidson et al. (1955),
the longitudinal dispersion of hydrogen (H2 ) injected into a
refinery gas pipeline is relatively high. Despite being unable
to determine the exact effects of 20 elbows with 90◦ and 8
elbows with 45◦ bends, they stated that it has a significant
contribution to the enhanced longitudinal diffusivity by inten-
sified mixing at these bends. Moreover, they also accounted
for the effect of imperfect mixing during injection. The results
presented by Hull and Kent (1952) from the radioactive tracer
injected into oil pipelines clearly show the increasing diffu-
sion coefficients as the distance from the releasing position
increases. They also noted the intensive mixing due to the
bends and elevation differences.
As described in Fig. 14, the effective diffusion coefficients
are plotted against Um d. The linear coefficients of Taylor’s
equation (11) are 0.253 and 0.399 taking f = 0.02 and f = 0.05,
respectively, as representative value for a relatively rough air-
way (McPherson, 1993). The results of the present study have a
linear gradient of 16.6 or 41.5–65.5 times larger than the results
obtained from Taylor’s analytical equation.
By observing the network properties supplied by the mine
in 1997 and 2011, the comparison of flow properties between
these two measurements are presented in Table 5. The results
of the diffusion coefficient evaluated from the 1997 and 2011
measurements show that the reduction of mine size, average
Table 5 – Comparison of average travelled distance, mean velocity and airway diameter and evaluated value of effective
diffusion coefficient, E, for 1997 measurement and 2011 measurement.
L total (m) Um average (m/s)
1997 Measurement 18,187 4.06
2011 Measurement 4,545 2.89
Ratio 4:1 1.4:1
(2011–1997)% 25 71.4
flow rate and average diameter have a relationship with the
reduction of the effective diffusion coefficient. It may be pos-
sible to perform a rough correlation of the decrease in the
averaged air velocity and the diameter of the airway to the
effective diffusion coefficient using Eqs. (11) and (12) by replac-
ing DL with E, resulting in a 0.55 times factor between the
1997 and 2011 measurements, while from this study, the effec-
tive diffusion coefficient of the 2011 measurement (45 m2 /s)
is 0.225 times smaller compared to the 1997 measurement
(200 m2 /s).
This observation, combined with the results presented
in Fig. 13, bring out a piece of evidence that the evaluated
diffusion coefficients for mine ventilation are much larger
compared to the results of Davidson et al. (1955). Both cases
have bends and consist of a network. The important ques-
tion that arises is what could be the possible reasons for this
very large diffusion coefficient. There must be some effect that
enhanced the diffusion coefficients. In the mine airways, there
are many small wall cavities due to blasting or special purpose
rooms and dead ends with or without any auxiliary ventila-
tion. There is a possibility that some portion of the tracer gas
is retained for several moment in these small cavities, recir-
culates and then re-joins the main flow, causing wider spread
and some delay time on the concentration–time curve. On a
small extent of the cavities, it may be possible to consider
this effect as wall roughness in which the average roughness
of the walls increases the friction factor as expressed by the
following relationship proposed by Colebrook (1939):
 
1 ε 2.51
 = −2 log +  (15)
f 3.7da Re f
where ε (m) is the average roughness height of the mine air-
way, and da (m) is the diameter of the airway. Thus, based
on Eqs. (11) and (12), the diffusion coefficient will increase as
well. However, for a larger volume of cavities, the effect of
the recirculation of the cavity should be treated in different
ways because the effect is much larger than wall roughness.
To consider the cavity flow and its effect on the delay time,
one should consider at least 2-D or (better) 3-D flow simula-
tion in which the cavity dimensions are taken into account.
On the contrary, in reality it is impractical to quantify the pres-
ence of cavities. Additionally, the mine network is too large to
be simulated practically by 2-D or 3-D numerical flow simula-
tion. Alternative numerical simulations are required to tackle
this difficulty. One of the expected numerical approaches
is the Lagrangian random walk method. This type of parti-
cle tracking method allows us to obviate the requirement of
a calculation grid that is mandatory in numerical methods
involving partial differential equations (PDEs), for instance,
the CFD method (Widiatmojo et al., 2013).
d Average (m) E (m2 /s)
5.05 200
3.86 45
1.3:1 4.44:1
76.9 22.5
 Process Safety and Environmental Protection 9 3 ( 2 0 1 5 ) 173–181
6. Conclusions
Tracer gas measurements using SF6 have been conducted
successfully in the Kushiro Coal Mine, Hokkaido, Japan. The
numerical simulation has been performed by considering a
one-dimensional advection–diffusion equation and employ-
ing several assumptions. The total amount of 15.8 m3 /s (31.8%
of total fresh air from Daini and Harutori intakes) leakage was
observed. The locations of the leakages also have been iden-
tified. The 2011 measurements showed that the total airflow
and total airway length of the mine has shrunk compared to
the 1997 conditions. Further investigations on network prop-
erties of each route obtained from the 2011 measurements
showed that the average total travelled distance of the tracer
gas, the average gas velocity, and the average diameter have
been reduced to 25%, 71.4% and 76.9% of its conditions in 1997,
respectively. The evaluated effective diffusion coefficient for
2011 measurement was 45 m2 /s or 0.225 times smaller than its
value in 1997 (200 m2 /s). Considering Taylor’s analytical equa-
tion (11) indicating that the effective diffusion coefficient is
linearly proportional to Um d, the linear coefficient between
the 1997 and 2011 measurements is 16.58 or approximately
41.5∼65.5 times larger than the linear coefficient used in Tay-
lor’s equation for evaluating the friction factor, f = 0.02–0.05.
This phenomenon indicates an intensified mixing of air that
enhances the diffusivities more than the average velocity and
diameter difference. The recirculation flows at cavities are
expected to be one major cause of this phenomenon. Further
numerical simulation is required to take the cavity flow effects
into consideration. In addition, new numerical methods must
be proposed because the present numerical simulation (i.e.,
CFD) has limitations in modelling the full scale of large mine
ventilation system.
Acknowledgements
The authors would like to thank the Ministry of Education,
Culture, Sport, Science and Technology of Japan (MEXT No.
090069) and Kyushu University Global COE Novel Carbon
Resources Science (NAHG 23J001) for the financial support. The
authors are also thankful to Dr. Amin Yousefi-Sahzabi from
Kyushu University for his valuable discussions and advice.
References
Colebrook, C.F., 1939. Turbulent flow in pipes with particular
reference to the transition between the smooth and rough
pipe laws. J. Inst. Civ. Eng. Lond. 11, 133–156.
181
Colella, F., Rein, G., Borchiellini, R., Carvel, R., Torero, J.L., Verda,
V., 2009. Calculation and design of tunnel ventilation systems
using a two-scale modelling approach. Build. Environ. 44,
2357–2367.
Davidson, J.F., Farquharson, D.C., Picken, J.Q., Taylor, D.C., 1955.
Gas mixing in long pipelines. Chem. Eng. Sci. 4,
201–205.
Einfeld, W., 1998. Environmental Technology Verification Report:
Photoacoustic Spectrophotometer, Innova AirTech
Instruments Type 1312 Multi-gas Monitor. US Environmental
Protection Agency.
Hull, D.E., Kent, J.W., 1952. Radioactive tracers to mark interfaces
and measure intermixing in pipelines. Ind. Eng. Chem. 44,
2745–2750.
McPherson, M.J., 1993. Subsurface Ventilation and Environmental
Engineering. Springer, London.
Patterson, R., Luxbacher, K., 2012. Tracer gas applications in
mining and implications for improved ventilation
characterisation. Int. J. Min. Reclam. Environ. 26,
337–350.
Sasaki, K., Widiatmojo, A., Arpa, G., Sugai, Y., 2009. Airflow
measurements and evaluation of effective diffusion
coefficient in large scale of mine ventilation network using
with tracer gas method. MMIJ. J. 125, 614–
620.
Sittel, C.N., Threadgill, W.D., Schnelle, K.B., 1968. Longitudinal
dispersion for turbulent flow in pipes. Ind. Eng. Chem. Fund.
7, 39–43.
Stokes, A.W., Kennedy, D.J., Hardcastle, S.G., 1987. A real-time
tracer gas analyzer – an investigational tool for mine
ventilation studies. Min. Sci. Technol. 5, 187–196.
Taylor, G., 1954. The dispersion of matter in turbulent flow
through a pipe. Proc. R. Soc. Lond. A 223, 446–
468.
Thimons, E.D., Kissell, F.N., 1974. Tracer gas as an aid in mine
ventilation analysis. US Bureau of Mines.
Timko, R.J., Thimons, E.D., 1983. New techniques for reducing
stopping leakage. U.S. Dept. of the Interior, Bureau of
Mines.
Wen, C.Y., Fan, L.T., 1975. Models for flow systems and chemical
reactors. Marcel Dekker, Inc., New York.
Widiatmojo, A., Sasaki, K., Widodo, N.P., Sugai, Y., 2013. Discrete
tracer point method to evaluate turbulent diffusion in circular
pipe flow. J. Flow Control. Meas. Vis. 1, 57–68.
Widodo, N.P., (Ph.D. thesis) 2008. Study on tracer gas method for
mine ventilation measurement and evaluation of gas
diffusion coefficient. Kyushu University, Fukuoka,
Japan.
Xu, G., Luxbacher, K.D., Ragab, S., Schafrik, S., 2013. Development
of a remote analysis method for underground ventilation
systems using tracer gas and CFD in a simplified laboratory
apparatus. Tunn. Undergr. Sp. Technol. 33,
1–11.