EU

The OFFGAS project has been focused on investigating the potential for process

KI-NA-25048-EN-N
optimisation and efficiency increases in the electric steelmaking process regarding
oxygen injection and energy transfer to scrap and melt at the EAF. A special empha-
sis has been put onto the application of permanent off-gas analysis systems installed
at the furnaces investigated. The investigations were carried out by a consortium of
research institutes (RWTH, BFI, CSM, CRM) and industrial partners (DEWG, TKN,
MH, ORI, Tenova) representing in total five electric steelmaking plants with a wide
range of produced steel grades and EAF technology.

The off-gas analysis systems in combination with process and energy models allow
for a continuous online monitoring and control of the EAF melting process and provide
an important tool for the steel plant workers in order to optimise the EAF process with
respect to optimal efficiency of oxygen lancing, carbon injection and addition, energy
transfer either from the arc and from the gas phase to the melted pool (post-combustion,
scrap preheating). The technical objective, to significantly increase the availability and
applicability of the off-gas signals for online assessment and control of post-combustion

Improved EAF process control using online offgas analysis — OFFGAS

and for comprehensive EAF energy monitoring, has been achieved. Additionally, EAF
process models that were up to now mainly used for off-line process assessment of
various aspects of EAF steelmaking, have been tested, developed and applied for online
EAF process and energy control.

Improved EAF process

control using online offgas
analysis — OFFGAS

EUR 25048

doi:10.2777/18439
Interested
Interestedin
inEuropean
Europeanresearch?
research?

RTD
RTD info
info isis our
our quarterly
quarterly magazine
magazine keeping keeping you you inin touch
touch with
with main
main developments
developments (results,
(results, HOW TO OBTAIN EU PUBLICATIONS
programmes,
programmes, events,
events, etc.).
etc.). ItIt isis available
available inin English,
English, French
French and
and German.
German. AA free
free sample
sample copy
copy Free publications:
or
orfree
freesubscription
subscriptioncan canbebeobtained
obtainedfrom: from:
• via EU Bookshop (http://bookshop.europa.eu);
Directorate-General
Directorate-GeneralforforResearch
Researchand
andInnovation
Innovation • at the European Union’s representations or delegations. You can obtain their
Information
Informationand
andCommunication
CommunicationUnit
Unit contact details on the Internet (http://ec.europa.eu) or by sending a fax
European
EuropeanCommission
Commission to +352 2929-42758.
1049
1049Bruxelles/Brussel
Bruxelles/Brussel
BELGIQUE/BELGIË
Priced publications:
BELGIQUE/BELGIË
Fax
Fax+32
+32229-58220
229-58220 • via EU Bookshop (http://bookshop.europa.eu).
E-mail:
E-mail:research@ec.europa.eu
research@ec.europa.eu
Internet: Priced subscriptions (e.g. annual series of the Official Journal of the
Internet:http://ec.europa.eu/research/rtdinfo.html
http://ec.europa.eu/research/rtdinfo.html
European Union and reports of cases before the Court of Justice
of the European Union):
• via one of the sales agents of the Publications Office of the European Union
(http://publications.europa.eu/others/agents/index_en.htm).

EUROPEAN
EUROPEANCOMMISSION
COMMISSION
Directorate-General
Directorate-Generalfor
forResearch
Researchand
andInnovation
Innovation
Research
Research Fund for Coal and SteelUnit
Fund for Coal and Steel Unit

Contact:
Contact:RFCS
RFCSpublications
publications
Address:
Address: EuropeanCommission,
European Commission,CDMA
CDMA0/178,
0/178,1049
1049Bruxelles/Brussel,
Bruxelles/Brussel,BELGIQUE/BELGIË
BELGIQUE/BELGIË

Fax
Fax+32
+32229-65987;
229-65987;e-mail:
e-mail:rtd-steel-coal@ec.europa.eu
rtd-steel-coal@ec.europa.eu
 European Commission

Research Fund for Coal and Steel

Improved EAF Process Control Using On-Line Offgas
Analysis – OFFGAS
H. Pfeifer, T. Echterhof, V.Y. Risonarta, L. Voj
RWTH Aachen University
Templergraben, 55, 52056 Aachen, GERMANY
H.-P. Jung, S. Lenz, C. Beiler
Deutsche Edelstahlwerke GmbH
Obere Kaiserstrasse, 57078 Siegen, GERMANY
H.-H. Ballewski, H. Mees
ThyssenKrupp Nirosta GmbH
Oberschlesienstrasse, 16, 47807 Krefeld, GERMANY
B. Kleimt, R. Pierre
VDEh – Betriebsforschungsinstitut GmbH
Sohnstrasse, 65, 40237 Düsseldorf, GERMANY
H.-J. Krassnig
Stahl- und Walzwerk Marienhütte GmbH
Südbahnstrasse, 11, 8021 Graz, AUSTRIA
F. Cirilli
Centro Sviluppo Materiali S.p.A
Via di Castel Romano, 100/102, 00128 Roma, ITALY
U. De Miranda
ORI Martin S.p.A.
Corso Garibaldi, 49, 20100 Milano, ITALY
M. Pustorino
Tenova S.p.A.
Via Monte Rosa, 93, 20149 Milano, ITALY
P. Nyssen, D. Borenstein, C. Ojeda, E. Abreu, P. Simon, B.Vanderheyden
Centre de Recherches Metallurgiques
Avenue du Bois Saint-Jean, 21, 4000 Liège, BELGIUM

Contract No RFSR-CT-2006-00004
1 July 2006 to 30 June 2009

Final report
Directorate-General for Research and innovation

2011 EUR 25048 EN

LEGAL NOTICE

Neither the European Commission nor any person acting on behalf of the Commission
is responsible for the use which might be made of the following information.

Europe Direct is a service to help you find answers

to your questions about the European Union

Freephone number (*):

00 800 6 7 8 9 10 11
(*) Certain mobile telephone operators do not allow access to 00 800 numbers or these calls may be billed.

A great deal of additional information on the European Union is available on the Internet.
It can be accessed through the Europa server (http://europa.eu).

Cataloguing data can be found at the end of this publication.

Luxembourg: Publications Office of the European Union, 2011

ISBN 978-92-79-22160-6

doi:10.2777/18439

ISSN 1831-9424

© European Union, 2011

Reproduction is authorised provided the source is acknowledged.

Printed in Luxembourg

Printed on white chlorine-free paper

Table of contents
1  Summary .................................................................................................................................... 5 
2  Scientific and technical description of the results .................................................................... 11 
2.1  Objectives of the project .......................................................................................................... 11 
2.2  Description of activities and discussion ................................................................................... 12 
2.2.1  Commissioning of off-gas analysis at the EAF, evaluation of the current state of EAF
operation (WP 1) ...................................................................................................................... 12 
2.2.1.1  Adaptation and commissioning of existing off-gas analysis system at DEWG Siegen (T 1.1) 12 
2.2.1.2  Installation and commissioning of ABB off-gas analysis system at TKN Bochum (T 1.2) .... 14 
2.2.1.3  Installation and commissioning of LINDARC off-gas analysis system at MH (T 1.3) ........... 17 
2.2.1.4  Installation and commissioning of EFSOP off-gas analysis system at ORI (T 1.4) ................ 24 
2.2.1.5  Installation and commissioning of off-gas analysis system for long term analysis at PA (T 1.5)
................................................................................................................................................. 25 
2.2.2  Off-gas measurements for control of decarburization (WP 2) ................................................. 28 
2.2.2.1  Off-gas measurements and correlation with nominal carbon input, assessment of carbon mass
balance (T 2.1) ......................................................................................................................... 28 
2.2.2.2  Correlation of measured critical points of decarburization with carbon mass balance (T 2.2) 30 
2.2.2.3  Adaptation of calculated amount of oxygen to measured carbon mass flow in off-gas (T 2.3)
................................................................................................................................................. 31 
2.2.2.4  Correlation with chromium contents in slag from slag analysis (T 2.4) .................................. 32 
2.2.2.5  Development of empirical model of optimum decarburization (T 2.5) ................................... 33 
2.2.2.6  Test of model and adaptation of model parameters (T 2.6) ..................................................... 34 
2.2.2.7  Visualisation of calculated chromium oxidation from thermo-chemical modelling (T 2.7).... 35 
2.2.3  Development of on-line energy monitoring system in stainless steel production (WP 3) ....... 36 
2.2.3.1  Data acquisition of the off-gas analysis (gas composition, volume flow rate and temperature)
to the EAF process control (T 3.1) ........................................................................................... 36 
2.2.3.2  Acquisition of temperature and volume flow rate of the hot water system to EAF process
control (T 3.2) .......................................................................................................................... 38 
2.2.3.3  Development of energy monitoring system and dynamic energy process model (T 3.3) ........ 39 
2.2.3.4  Efficient oxygen use and optimum process rules regarding electric energy demand (T 3.4-3.6)
................................................................................................................................................. 42 
2.2.4  Development of dynamic control of oxygen supply for post-combustion by off-gas laser
absorption spectrometry at Marienhütte (WP 4) ...................................................................... 45 
2.2.4.1  Plant trials with additional off-gas measurement by portable VIS/IR detector systems, to
provide additional off-gas data (e.g. flow rate, H2 content) for modelling purposes and
comparison of gas analysis signals (T 4.1) .............................................................................. 45 
2.2.4.2  Adaptation of BFI dynamic mass and energy balance model to the conditions of the
Marienhütte EAF (T 4.2) ......................................................................................................... 52 
2.2.4.3  Performance of simulation trials to study the effect of post-combustion oxygen on the energy
balance of the furnace (T 4.3) .................................................................................................. 64 
2.2.4.4  Development of dynamic control strategies for post-combustion oxygen supply on the basis of
real-time laser-based off-gas measurement and model-based calculation of further off-gas data
(T 4.4) ...................................................................................................................................... 65 
2.2.4.5  Test and optimisation of control strategies with the help of the BFI simulation model (T 4.5)
................................................................................................................................................. 70 
2.2.4.6  Implementation and test of dynamic control of post-combustion oxygen on the basis of
measured off-gas values at the Marienhütte furnace (T 4.6) ................................................... 72 
2.2.4.7  Evaluation of achieved energy savings (T 4.7) ........................................................................ 76 

3
2.2.5  Development and commissioning of continuous feeding EAF process dynamic control model
based on off-gas analysis (WP 5) ............................................................................................. 76 
2.2.5.1  Development of a long term data collecting and remote monitoring system to control the
reliability of the off-gas measurement system (T 5.1) ............................................................. 76 
2.2.5.2  Development of a new model to establish the distribution of the carbon between bath slag and
free board (T 5.2) ..................................................................................................................... 78 
2.2.5.3  Validation of model developed by experimental measurement on the plant (carbon
concentration in steel, carbon content in slag and in dust sample inside EAF) (T 5.3) ........... 84 
2.2.5.4  Development of a system for continuous measurement of off-gas flow rate model (T 5.4).... 88 
2.2.5.5  Development of new on-line model for the dynamic control of carbon and oxygen injection in
the EAF and dynamically control of the post combustion between the EAF and the Consteel
tunnel (T 5.5) ........................................................................................................................... 93 
2.2.5.6  Application and trial of the new on-line model (T 5.6) ........................................................... 95 
2.2.6  Development of CRM process control at PA with special emphasis on off-gas analysis (WP 6)
................................................................................................................................................. 99 
2.2.6.1  Modification of CRM dynamic metallurgical model in order to include off-gas information as
input (T 6.1) ............................................................................................................................. 99 
2.2.6.2  Performance of simulation trials with modified injection pattern of post-combustion oxygen
lances, the off-gas measurements being active (T 6.2) .......................................................... 103 
2.2.6.3  Optimisation of the injection pattern in order to decrease energy consumption and iron losses
with the help of the dynamic metallurgical model (T 6.3) ..................................................... 111 
2.2.6.4  Performance of simulation trials with the optimised injection pattern of post-combustion
oxygen lances, the off-gas measurements being active (T 6.4) ............................................. 115 
2.2.6.5  Evaluation of achieved energy savings (T 6.5) ...................................................................... 116 
2.2.7  Assessment of plant trials and development of EAF energy process model (WP 7) ............. 116 
2.2.7.1  Correlation of off-gas energies for stainless, special and carbon steel heats with electric energy
demand (T 7.1) ....................................................................................................................... 116 
2.2.7.2  Influence of oxygen lancing and injection strategy to off-gas energy and electric energy
demand (T 7.2) ....................................................................................................................... 120 
2.2.7.3  Influence of burner to off-gas energy and electric energy demand (T 7.3) ........................... 124 
2.2.7.4  Influence of dedusting system to electric energy demand (T 7.4) ......................................... 127 
2.2.7.5  Development of additional process adaptation in order to minimize electric energy demand
and oxygen consumption (T 7.5) ........................................................................................... 129 
2.2.7.6  Development of an energy/mass model based on off-gas measurements for Consteel process
control (T 7.6) ........................................................................................................................ 132 
2.2.8  Final plant trials with adapted process models and assessment (WP 8) ................................ 137 
2.2.8.1  Plant trials and assessment at DEWG (T 8.1) ........................................................................ 137 
2.2.8.2  Plant trials and assessment at TKN (T 8.2) ............................................................................ 140 
2.2.8.3  Plant trials and assessment at MH (T 8.3) ............................................................................. 140 
2.2.8.4  Plant trials and assessment at ORI (T 8.4) ............................................................................. 147 
2.2.8.5  Plant trials and assessment at PA (T 8.5) ............................................................................... 148 
2.3  Conclusions ............................................................................................................................ 151 
2.4  Exploitation and impact of the research results ..................................................................... 154 
3  List of Figures ........................................................................................................................ 155 
4  List of Tables ......................................................................................................................... 162 
5  List of References .................................................................................................................. 163 
6  Abbreviations ......................................................................................................................... 164 

4
1 Summary
The objective of the project was to increase the efficiency of EAF oxygen injection and energy transfer
to the scrap and melt. In expansion to short-term investigations using off-gas analysis to monitor energy
flow rate of the off-gas, permanent on-line monitoring and control of the EAF process on the basis of
off-gas analysis has been implemented. The continuous availability of an on-line monitoring system of
the EAF melting process has shown its effectiveness to provide an important tool for the steel plant
workers in order to optimize the EAF process with respect to optimal efficiency of oxygen lancing,
carbon injection and addition, energy transfer either from the arc and from the gas phase to the melted
pool (post-combustion, scrap preheating). The technical objective, to significantly increase the
availability and applicability of the off-gas signals for on-line assessment and control of post-
combustion and for comprehensive EAF energy monitoring, has been achieved. Additionally, EAF
process models that were up to now mainly used for off-line process assessment of various aspects of
EAF steel making, have been tested, developed and applied for on-line EAF process and energy control.
The work programme of the OFFGAS project partners was organised by division into 8 work packages
that were subdivided into various tasks. After installation and commissioning of the off-gas analysis
systems at the industrial partners (WP 1), off-gas data were assessed together with additional process
data in order to define the current state of EAF operation and to develop process models: optimal
decarburization (WP 2), energy monitoring system and optimal energy efficiency (WP 3), optimal post-
combustion (WP 4), optimal Consteel process (WP 5), advanced CRM process model at ArcelorMittal
Differdange (WP 6). Results from plant trials and process modelling have been used to compare and
adapt the different process models and rules from each industrial partner (WP 7). Final plant trials with
the process models developed and/or adapted have been conducted as last work package (WP 8).
WP 1: Commissioning of off-gas analysis, evaluation of current state of EAF process
In this work package RWTH activity has been to support the industrial partners DEWG and TKN and
also CRM at AM Differdange in adaption and installation as well as commissioning of off-gas analysis
systems in the primary dedusting system of industrial EAFs. During measurements the off-gas sampling
and analysis systems have been adjusted and reworked by the partners with regard to long-term
availability and efficient maintenance. An initial set of off-gas measurement data has been established
by industrial measurements in order to determine the present state of EAF operation.
A tuneable diode laser (TDL) system (LINDARC) for in-situ measurement of CO and O2 concentration
as well as temperature of the off-gas was installed directly at the hot gas duct of the primary EAF
dedusting system of the Marienhütte (MH) AC EAF. High reliability and low maintenance effort of this
system was evident. An improvement of the protection of sensitive electronic parts (e.g. plugs,
cables ...) was performed, to adapt the system to the harsh environment in the off-gas stream of the Arc
Furnace.
The energetic performance of the MH AC furnace at the beginning of the project has been assessed by a
statistical energy balance model of BFI. This model calculates the electrical energy demand of EAF
from the most relevant consumption figures of a furnace. The difference of the calculated demand to the
actual electrical energy consumption is a measure for the energetic performance of the furnace. This
model calculation has been used throughout the project, to follow the improvements achieved by the
dynamic control of the post-combustion oxygen developed for the MH furnace on the basis of the
LINDARC off-gas analysis.
In co-operation with CSM and Tenova at the ORI Martin Consteel plant an EFSOP off-gas analyzing
system giving real time measurement of the off gas composition (CO, CO2, H2, O2) continuously
through the entire heat without any interruption has been commissioned and installed. Additionally a
thermocamera to measure the off-gas temperature exiting the furnace at the “4th hole” and a
downstream off-gas analyzer at the end of the Consteel tunnel that gives real time measurement of CO,
CO2, O2 have been installed at the Consteel plant.
WP 2: Off-gas measurements for control of decarburization
The objective of this work package was to use the off-gas analysis system to determine the critical point
of decarburization of highly alloyed steel heats and to develop a process model predicting the amount of
carbon removed from the melt to adapt the amount of oxygen injected as well as a model estimating the
chromium content of the slag based thermo-chemical calculations.

5
All these objectives have been reached by DEWG and RWTH. The developed models have been tested
off-line and adapted as needed within this task. They showed good agreement between modelled values
and data acquired by measurements and slag and steel analysis and indicated a potential for oxygen
savings by adaption of the procedure for oxygen injection.
WP 3: Development of on-line energy monitoring system in stainless steel production
Within this work package TKN with the assistance of RWTH did link the data acquired from the off-
gas system (gas composition at two measuring points, gas temperature) and the hot water cooling
system to EAF process control and put a newly developed dynamic energy process modelling and
monitoring system into operation. The energy model comprises the main energy sinks and thus expands
the process data already available to the furnace operator.
Additionally the current process rules regarding the oxygen use and its influence on the electric energy
input have been investigated and adapted rules have been proposed in order to reduce electric energy
demand.
WP 4: Development of dynamic control of oxygen supply for post-combustion by off-gas laser
absorption spectrometry at Marienhütte
The structure of an existing dynamic EAF process model of BFI was adapted to the configuration of the
MH furnace. This model is based on coupled mass and energy balance calculations and was used to
observe the current EAF process state. The measured values of the off-gas measurement system
LINDARC were incorporated as additional model input data. With the help of this model, control
strategies for post-combustion oxygen were developed to improve its energetic efficiency.
The validation of the model was started on the basis of a number of heats from a trial campaign at the
MH furnace performed with an IR/VIS off-gas analysis system of RWTH Aachen. The data from this
measurement campaign were further evaluated to acquire information on off-gas flow rate and off-gas
components like H2 and CH4, which cannot be measured on-line by the LINDARC system, but are
important for a dynamic energy balance calculation and also for a comprehensive on-line control of the
post-combustion oxygen. A further trial campaign with a thermocouple probe for the off-gas
temperature provided by RWTH was performed, to derive a soft sensor for calculation of a more
accurate value for the off-gas temperature on the basis of the cooling cycle data of the laser probe. On
the basis of this calculated off-gas temperature, more reliable information on the sensible energy losses
via the off-gas was provided for model calculations.
The comparison of the different analysis techniques IR/VIS vs. Laser confirmed a correlation between
CO content and H2 as well as CH4 content. A dynamic control strategy of the post combustion system
was developed, where the oxygen flow rate is set proportional to the CO/O2 ratio in the off-gas, which
can be measured continuously by the LINDARC system. In case this ratio is lower than a preset value,
the post combustion oxygen injectors are switched off.
The trials performed at the MH furnace with this closed loop control of the post-combustion oxygen
were evaluated regarding the energetic behaviour and the achieved energy savings with the help of the
BFI statistical and dynamic models for the EAF. Due to the dynamic control depending on the evolved
CO the energetic efficiency of the post-combustion oxygen was increased significantly.
WP 5: Development and commissioning of continuous feeding EAF process dynamic control model
based on off-gas analysis
CSM activity for this project has been focused on the adaptation of an already available BOF dynamic
model, suitable to describe in real time steel and slag composition and temperature, to EAF (the
dynamic model of the EAF has been named as dyCoSMelt®).
The model simulates the chemical reactions occurring inside the steel bath and slag and the melting
kinetic of the scrap. The model runs integrated with the ‘FreeBoard’ model (based on EFSOPTM off gas
analyser), developed by Tenova, which simulates the gas phase inside the EAF. The two models have
been used to build the online control system for EAF continuous charging configuration (Consteel) that
has been tested at the ORI Martin plant.
An additional tool needed for the mass/energy balance for the process has been developed by CSM,
computing in real time the off gas flow rate; the algorithm is based on static pressure and temperature
measurement at the furnace connecting car and has been implemented into the Tenova system.
Calibration of the software sensor is performed by Pitot measurement in the fume duct and carbon
balance between this point and the measurement performed by EFSOP at the exit of the furnace.

6
CFD calculations have been also performed to evaluate the flow field at the connecting car, that collects
gas from EAF and where the static pressure sensor is located; the aim was to study the suitability of
such a measurement in providing indication for proper furnace off gas MFR calculation. A relationship
for gas MFR calculation has been tested and the obtained results indicated that the developed virtual
sensor can be used for the purposes of the project.
Off-gas temperature long term measurement has been implemented by Tenova thermo-camera (installed
at fourth hole). As parallel activity for cross check evaluation, CSM designed and tested with the
support of ORI a specially designed suction pyrometer. The pyrometer configuration was aimed at
avoiding in particular clogging from dust (to guarantee at least one complete heat measurement), error
due to furnace radiation and to decrease as low as possible the time of response of the thermocouple.
In order to be able to get a proper running of the model the following side activities have been also
performed:
 collection and examination of data from industrial heats,
 calibration campaign of the model, also by collection of steel and slag samples.
After calibration activity and sensitivity analysis, comparison with experimental data showed that the
CSM bath/slag dynamic model (dyCoSMelt®) is able to simulate in a good manner the EAF Consteel
process and can be used for the online control system.
ORI Martin contribution to the project has been mainly related to support
 Tenova in installing and maintaining the proper running of the EFSOP system (off-gas analysis
and temperature)
 CSM and Tenova for the monitoring modules calibration (measuring and heat tests campaigns)
and for the controlling modules development and application (also Level 2 software upgrading
for the aims of the project).
In particular ORI Martin contributed by:
 Mechanic, electric and automation plant equipment modification for EFSOPTM installation,
 follow up and EFSOPTM maintenance over the project period,
 commissioning and supporting external company for downstream gas speed, composition and
temperature for modules calibration,
 performing steel and slag sampling and analysis campaigns for modules development and test
activities,
 updating DB structures for managing properties of materials suitable for real time modules
computation,
 providing level 2 real time data for furnace additions and off-gas mass flow rate calculation.
Steel and slag sample have been taken and analyzed in order to provide data for dynamic mass and
energy balance models calibration and then general validation in describing process evolution.
Also some activities for pyrometer support devices assembling and assistance during measurements
have been performed.
ORI Martin collaborated also to the integration in their automation system the general structure of the
control tools, installed with the currently running control software.
The objective of Tenova activities was related to the development of a control system able to elaborate
the off-gas analysis and to control oxygen and coal addition to the steel bath in order to achieve the
target value of carbon concentration, reducing (or optimizing ) the energy consumption.
To achieve this goal, the off-gas analyzer system EFSOPTM has been installed at ORI Martin Consteel®
plant.
A freeboard model, based on the off-gas analysis and temperature has been developed and coupled with
CSM dynamic model, in order to have a complete representation of the phenomena which occur in the
EAF, (both in liquid and gas phase).
Over the installation and commissioning, an off-gas analysis and plant/process database have been
implemented. This database is currently used to calibrate and refine the Tenova and CSM dynamic Bath
& Slag and Freeboard models.

7
The development of a complete EAF model is the key to develop a control system. The control logic
routines (agreed among the partners) have then been written, installed and tested at Ori Martin.
Tenova developed also the complete HMI and automation system on which all the models and the rule
code routines are currently running and controlling the Ori Martin’s EAF and Consteel®.
WP 6: Development of CRM process control at PROFILARBED with special emphasis on off-gas
analysis
Off-gas analysis measurement has been performed in AM Differdange. The probe has been installed in
different locations (furnace elbow, furnace gap and post-combustion chamber). The off-gas flow-rate
exiting the furnace and the air ingress have been assessed and compared to the model simulation.
Different operating patterns were simulated with the CRM model and tested industrially. An optimum
operating pattern has been deduced and applied during three weeks. A gain estimated to 0.5 euro per ton
of steel produced has been obtained during this period.
WP 7: Assessment of plant trials and development of EAF energy process model
At the MH AC furnace, plant trials for investigation of different operating patterns for oxygen lancing,
oxygen injection and gas burners were performed. The influence of these different operating patterns on
the thermal and chemical off-gas losses and on the electrical energy demand was investigated with the
help of the statistical and the dynamic energy balance models of BFI. The energetic benefit and the
influences on the off-gas losses of each trial were presented. Furthermore the off-gas losses were
investigated in detail for the phases of melt down of each scrap basket and the refining period. High off-
gas losses during the refining phase lead to a significant decrease of the energetic performance of the
furnace. The influence of the blowing carbon and its effect on the energetic performance has been
investigated too.
The trials performed at DEWG and TKN were investigated by RWTH regarding the impact of process
parameters on electric energy demand and oxygen consumption. The off-gas losses were investigated in
relation to the electric energy demand for carbon and stainless steel at DEWG and TKN and in both
cases a correlation between high energy consumption and high off-gas losses could be validated. The
investigation of the oxygen injection in relation to off-gas losses and energy consumption lead to a
general correlation of rising off-gas losses with increasing oxygen injection. The energy input is slightly
decreasing for TKN and increasing for DEWG with an increase of oxygen injection.
The investigation of influence of the dedusting system on the energy demand performed by RWTH
showed that a reduction of the off-gas mass flow for example by increased air tightness of the furnace
of a dynamically controlled off-gas flow rate reduces the off-gas losses as well as the electric energy
demand.
Based on the result gained in WP 5 the process description models have been improved by Tenova and
dedicated equipment has been installed (i.e. the downstream off-gas analyzer). The significance to cross
correlate the off-gas composition up and down stream has been taken in account. In particular a new
feature on the off-gas flow rate model has been developed. The off-gas flow rate can be calculated
based on the dynamic real time measure of both off-gas analyzers.
WP 8: Final plant trials with adapted process models and assessment
During final trials at DEWG performed in cooperation with RWTH the decarburization model
developed in WP2 was put into online operation. The results of the trials indicate that the average
consumption of oxygen can be decreased by the application of the model. The amount of chromium in
the slag is also reduced by optimized oxygen injection procedure.
The dynamic mass and energy balance model has been implemented for on-line observation of the
energetic status of the Marienhütte furnace. Besides the calculated actual melt temperature, the off-gas
analysis and the most important operating parameters are visualized. This tool supports the operator to
run the furnace in an optimal way. The accuracy of the model was improved continuously during the
project. For the finally collected data of about 1000 heats, the modelling error standard deviation for the
melt temperature before tapping was below 20 K.
Trials at ORI Martin have confirmed accuracy, reliability and repeatability of the energy/mass model
applied to the Consteel process control system.

8
As a general positive consequences deriving from the application of the control system, the following
results can be mentioned.
 a reduction in oxidizing oxygen and an increasing in energetic yield (this affects positively the
slag quantity and composition. The FeO the CaO and the amount of slag have been reduced;
 process more stable. This involves an easier and better process/production management for all
the plant;
 the post combustion reaction increase the off-gas volume so, at the same fume system working
condition, more high is the post combustion less is the false air sucked from the slag door and
the gap. The post combustion effect along the entire conveyor increase the thermal exchange
and so the scrap preheating efficiency. The post combustion effect is visible also in the off-gas
chemical analysis measured by EFSOP®.
In addition to the on line feedback implemented, some indication messages have been implemented to
advice operator:
 Warning about bath stratification and possible uncontrolled reaction (three level of probability)
 Group (increase/decrease) for managing the bath oxidation (higher or lower the window) by
oxygen lancing or coal addition
 Group (increase/decrease) for managing the scrap melting and bath temperature by conveyor
speed (scrap feeding rate)
 Warning about FeO exceeding (coal injection addition for FeO reduction as suggestion)
These information help the plant operator, not only to better perform in terms of consumptions saving,
but also to learn and understand the process improving their own knowledge. ORI Martin is currently
using all the equipment installed.
The new generation of EAF control system, which elaborate simultaneously the data from SW and HW
sensors has been used by Tenova to develop a commercial system named iEAF.

9
10
2 Scientific and technical description of the results

2.1 OBJECTIVES OF THE PROJECT

Mass and energy flow rates at the EAF are not steady-state but time-dependent during a heat, e.g.
electric power input, injection of oxygen and coal, gas burners, post-combustion injectors, CO
emission, off-gas enthalpy, heat load of the cooling system. Some of the energy flow rates are precisely
known for process control: electric energy input, heat transfer to the cooling system. For some mass
flow rates the efficiency of mass transfer is known to a lesser extent: injection of oxygen and coal.
However, other important EAF mass and energy flow rates are poorly known due to technical difficulty
to achieve a reliable signal during long-term measurements: off-gas enthalpy, carbon mass flow rate in
off-gas. Currently, EAF process optimization is based on means of characteristic values per heat:
electric energy demand, metallic yield, tapping temperature, tap-to-tap time etc. In the case that a set of
reliable and more complete data about the important time-dependent EAF mass and energy flow rates
are available the set-up of new and more precise process optimization will be possible. Dynamic
process control will be developed. Since years off-gas analysis systems based on infrared absorption are
available for measurements at industrial furnaces, e.g. from ABB, Emerson Process Management,
Goodfellow EFSOP™ system. Recently, new detector systems have been developed based on tuneable
diode laser absorption techniques in-situ at the hot gas duct, e.g. the LINDARC™ system. Application
of these systems at industrial EAFs is still rare.
Time-dependent process control is still restricted to control of electric power input. Control of other
mass and energy flow rates as oxygen injection for decarburization and post-combustion, gas burners,
coal injection, heat transfer to cooling system requires reliable and comprehensive on-line process
monitoring. In this project, continuous analysis of EAF off-gas is used to develop comprehensive
energy monitoring and to improve energy efficiency by enhanced control of the EAF process. In
particular, the following objectives shall be achieved in order to reduce energy demand and thus to
increase EAF productivity:
i. implementation of on-line energy monitoring to increase energy efficiency
ii. monitoring of decarburization of the melt to increase oxygen efficiency
iii. decrease of heat load of the dedusting system by optimal post-combustion and heat transfer to
the scrap in the furnace vessel
iv. decrease of oxygen lancing and injection time and, consequently, tap-to-tap time.

11
2.2 DESCRIPTION OF ACTIVITIES AND DISCUSSION
2.2.1 Commissioning of off-gas analysis at the EAF, evaluation of the current state of EAF
operation (WP 1)
An overview on the production characteristics of the EAFs of all industrial partners is given in Table 1.
Table 1: Production characteristics of EAFs at industrial partners
TKN DEWG MH ORI AM
Differdange
Tapping weight [t] 150 140 / 120 35 75 155
Transformer Power 135 105 35 35 2 * 70
[MVA]
Power supply AC AC AC AC DC
Stainless steel X X - - -
Special steel - X - - -
Construction steel - - X X X
Carbon content of alloys 5.5 - 7.8 6.5 - 7.3 - - -
Silicon content of alloys 2.8 - 4.5 0.1 - 3.9 - - -
Carbon content at tapping 1.5 0.07 / 0.60 n.a. n.a. 0.06
Mean oxygen input [m3/t] 5.5 22.6 / 8.8 44 17.5 42.5
Gas burners - - X - X
Consteel technology - - - X -
2.2.1.1 Adaptation and commissioning of existing off-gas analysis system at DEWG Siegen (T 1.1)
The EAF at DEWG with a tapping capacity of 140 tons is used for production of carbon steel (140 t)
and stainless steel (120 t) (Table 1). The AC EAF has a water-cooled top cover, water-cooled wall
panels and a lance manipulator equipped with two lances. One lance is used for oxygen injection and
the second one is used for dust injection. The dust is injected by compressed air. The average tap-to-tap
time is 75 min. In order to decrease the electric energy input there is oxygen injected into the melt.

A H
B

J
F

Figure 1: EAF layout at DEWG Siegen

A: measurement point A, B: measurement point B, H: water injector, F: oxygen/dust injector,
J: dust silo, 3: gap between EAF vessel and primary dedusting system
Figure 1 shows a schematic of the EAF and dedusting system layout at DEWG in Siegen. The
sampling of the off-gas was realized at point A and point B by using water-cooled probes. The
installation position directly behind the gap between the cover elbow and the primary dedusting system
(Figure 2) was chosen to minimize the clogging of the probe and to maximize the protection against
mechanical problems. Also the influence of air, which ingresses into the off-gas stream at the gap, to

12
the measured concentration values is minimized. The water-cooled probe has been installed at the
centre of the elbow. Therefore the off-gas sampling is carried out directly from the EAF off-gas stream.

Figure 2: Water-cooled probe installed at the elbow near the EAF (point A) at DEWG Siegen
During the course of the project the measurement point was moved downstream from point A to point
B. Here the concentration of the different off-gas species is more evenly distributed in radial direction
and it is therefore easier to get a representative off-gas sample sufficient for the control of
decarburization (cf. section 2.2.2).
The opening of the water-cooled probe used for the off-gas sampling was enlarged during the project.
At the beginning the diameter of the opening was 12 mm and has been enlarged to 15 mm. With this
measure the resistance against clogging was successfully increased and within the project the sampling
of off-gas worked without any problems.
To prevent clogging of the sampling probe and tubing to the analysis system, the probe and tubing are
flushed every 15 minutes (carbon steel) with pressurised nitrogen. For stainless steel the flushing
sequence is set to 25 min. The two flushing sequences have been chosen because of the adapted dust
injection while producing carbon steel. Due to the flushing sequence there is a periodic zero value of all
concentrations.

Figure 3: Existing off-gas analysis system (left) and layout of the ABB analysis system installed (right)
Figure 3 shows the layout of the ABB analysis system installed. The analysis system is located near the
EAF vessel (about 2 m). So the time delay caused by the time the off-gas sample needs to reach the
inlet of the analyse system, is minimized. Testing gases are used to calibrate the analyser cells regularly.
The off-gas sample taken is piped to the off-gas inlet and then to a set of filters, e.g. a fine filter ensures

13
that there will be no dust or dirt taken to the analyser cells. The cooling trap will separate the water
vapour content from the off-gas sample, because the analysis system installed is very sensitive to water.
Eventually there is also the pump, which is delivering the off-gas sample to the ABB analyser cells. The
analyser consists of three analyzer units (O2,CO2,CO). The CO and CO2 concentrations are measured
using an infrared detector cell. The O2 concentration is measured using a paramagnetism detector cell.
The data is captured by the PLC (programmable logic controller) of DEWG Siegen.
Figure 4 shows the concentration curves of the off-gas sampled at point A. With increasing air
tightness of the EAF vessel, the CO concentration decreases while the CO2 concentration reaches about
30 % (1st basket).

Figure 4: Measured off-gas composition at point A (DEWG; carbon steel)

2.2.1.2 Installation and commissioning of ABB off-gas analysis system at TKN Bochum (T 1.2)
The EAF at TKN with a tapping capacity of 150 tons is used for production of stainless steel (Table 1).
The AC EAF has a water-cooled top cover, water-cooled wall panels and a lance manipulator equipped
with one lance for oxygen injection.
 

Point B

Point A

Figure 5: EAF layout at TKN Bochum

To determine the CO, CO2 and O2 content in the off-gas of this EAF two measuring points with off-gas
sampling probes have been permanently installed in the off-gas system. In Figure 5 a schematic of the

14
EAF at TKN Bochum is shown with the measuring point A and point B realized for the sampling of off-
gas. Point A is located directly after the EAF elbow. The off-gas dedusting system at TKN is equipped
with a sliding muffle. This sliding muffle is installed to regulate the amount of air intake through the
gap. A water-cooled probe has been installed onto the sliding muffle between EAF elbow and the
stationary primary off-gas duct sampling the off-gas. Figure 6 shows the layout and Figure 7 a picture
of the measurement point A. The water-cooled probe has been installed at the centre of the elbow, so
that the off-gas sampling is carried out direct from the EAF off-gas stream. The chosen installation
position leads to small influence of the air intake.

Figure 6: Layout of the off-gas measurement installed at point A at TKN Bochum

Figure 7: Installation of the water-cooled sampling probe in front of the primary off-gas duct
Measurement point B is located after the hot-water cooled heat recovery boiler and directly before the
DEC (Direct Exhaust Control) flap. This flap is regulating the mixing ratio of primary off-gas to off-gas
from the canopy dedusting of the housed EAF. The sampling probe installed at point B is not water
cooled because maximum off-gas temperatures at this point are at 570°C.
Volumetric flow and temperature of the off-gas at both positions have also been measured within this
project using thermocouples and Pitot tubes measuring the differential pressure in the off-gas flow.
At both measurement points the off-gas is sampled by the probes and then lead in a heated sample gas
line to the attached off-gas analysis system. Before reaching the analyzers the gas is filtered and
condensate is precipitated. Gas processing and analyzers are housed in air-conditioned cabinets within
the dog-house and above the secondary filter chamber, respectively. ABB off-gas analysis systems,
shown in Figure 8, have been installed on both measurement points. These systems are measuring the
CO, CO2 concentration of the sampled off-gas by infrared spectrometry (URAS 14) and the O2
concentration of the sampled off-gas is measured the paramagnetic property of oxygen (MAGNOS 16).
They are similar to the ABB analysis system at the DEWG Siegen EAF. Since no H2 has been measured

15
at the EAF at TKN during the initial off-gas measurement campaign, no H2 analyzers have been
adapted to the ABB off-gas analysis systems.

1 3

7 2

6 4 5 8

Figure 8: The ABB off-gas analyzer

1 - off-gas sample inlet; 2 - calibration gas inlet; 3 - filter unit; 4 - water trap; 5 - flow meter;
6 - filter; 7 - overpressure protection; 8 - analyzer module
Because of the dust load of the primary off-gas of ca. 8 kg/tsteel the off-gas probes are equipped with a
filter unit for primary cleaning as well as an automatic flushing box. Probe, sample gas line, filter unit
and flushing box have the cleaned regularly to prevent a clogging by dust. At point A the probe is
therefore flushed with pressurized nitrogen regularly within charging breaks for 3 minutes and if the
sample volume flow measured at the analyzer falls below a certain threshold, respectively.
Clogging of the probe at point B is in opposition to point A no problem. A cleaning of the filter unit
once a day is sufficient. The probe at point B is currently in continuous operation since 12 month.
In the course of the project there have been made some changes to the design of the water-cooled probe.
In 2008 the filter unit for primary cleaning has been installed directly at the water-cooled probe. But
this lead to a clogging of the filter and probe usually within 3 heats. Therefore this modification was
rapidly discarded. The positioning of the filter unit directly before the flushing box lead to a durability
of 3 to 5 days of production. The design of the flushing box has been changed too. In 2008 also the
position of the off-gas sampling at point A was shifted in relation to the previous position for about 0.7
m downstream in order to be save from mechanical damage to the water cooled probe. Figure 9 shows
the positions of off-gas sampling near the EAF. Shifting the position of off-gas sampling is for sure
cause of an increase of O2 content measured in the off-gas, but this is compensated by an increase of the
volume flow rate of the off-gas sampled so that the retention time could be decreased from 42-60 s to 4-
6 s.

Figure 9: Off-gas sampling positions at point A in 2007 (left) and 2008 (right)
Since the service lifetime of the water-cooled probe of 3-5 days was still not satisfactory in 2009
another change has been made to the probe. The probe is usually clogging because of water condensing

16
within the sampling channel of the probe and a subsequent setting of the dust with this condensed
water. To prevent a condensation of water within the probe an air insulation layer has been built into the
lance between water cooling and sampling channel. A mechanical drawing of the changed probe is
shown in Figure 10 and in Figure 11 a cross-section of the probe is shown. In a first trial this changed
design lead to a complete prevention of water condensation within sampling channel of the probe and
therefore to a prevention of clogging. The modified probe has been running within the trial for 4 weeks
without any incident of clogging.

Figure 10: Mechanical drawing of water-cooled gas sampling probe with air insulation layer

Cooling water inlet

Cooling water outlet

Isolation

Figure 11: Cross-section of the modified water-cooled probe

2.2.1.3 Installation and commissioning of LINDARC off-gas analysis system at MH (T 1.3)
At the 35 t AC EAF of Marienhütte, a laser-based system for off-gas analysis has been installed. The
laser gas monitor is an optical instrument, that transmits infrared light from a temperature stabilised
single mode diode laser to a detector, diametrically opposite to the transmitter (Figure 12).

17
Figure 12: Principle of Installation

Figure 13: Principle of Absorption

The diode lasers operate at approximately room temperature. The laser gas monitor is based on a
measuring technique known as single-line spectroscopy in the near infrared (NIR) spectral range. The
single-line spectroscopy measuring technique is based on the selection of one single absorption line in
the near infrared spectral range for the specific gas. The spectral width of the diode laser is considerably
narrower than the spectral width of the absorption line for the gas chosen. By varying the diode laser
current, the diode laser wavelength is scanned across the absorption line (Figure 13). The light detected
in the receiver unit varies in intensity as a function of wavelength during the scan of the laser, due to the
absorption of light from the specific gas molecules in the optical path between diode laser and the
receiver.
The detected shape and size of this single absorption line is used to calculate the amount of gas between
the transmitter unit and the receiver unit. Absorption lines from other gases are not present at this
specific wavelength, and therefore, will not interfere with the single absorption line or the resulting gas
concentration [5].
At the AC EAF of Marienhütte, secondary energy is provided by a door burner (TB4) and the so called
KT-system which is a combination of oxygen, natural gas and carbon injectors. The KT-Injectors (KT1-
3) are located at the cold spots of the shell (Figure 14). Each flow is regulated by a multiple point
chemical package controlling system against the progress of each heat. The system can be operated in a
burner mode during melt down and a lancing mode with a focused oxygen jet to refine the heat and to
build up the foamy slag by carbon injection. Additional oxygen to prepare the door area for sampling is
supplied by an Oxygen-Lance (LM). For the post combustion of evolved CO and H2 four PC-Injectors
with a soft oxygen jet are placed tangentially at the top furnace (PC1-4).

18
Figure 14: Top view of the AC EAF at Marienhütte
The laser system, which has been installed at the MH EAF, consists basically of three parts: (a) The
transmitter unit; (b) The receiver unit; (c) The electronic evaluation unit with visualisation. The laser
units are protected by a water cooled housing. The laser system is assembled just behind the gap after
the elbow in order to deliver off-gas concentration close to the place of emergence in the furnace
(Figure 15). Due to the typical harsh environment of an electric arc furnace the sensitive transmitter
and receiver unit are mounted outside the off-gas system to prevent from high heat load and even
mechanical damage. These parts are additionally cooled by water. To evade undesirable effects of
entered false air from the gap the path length is reduced by two water cooled pipes (Figure 16).
A

1
3
4
5

2
Figure 15: Layout of the EAF at MH
A: Measurement point A (LINDarc), 1: gap, 2: doorburner, 3: KT oxygen/gas/coal-
Injectors, 4: PC-injectors, 5: oxygen-lance

19
Figure 16: Water cooled lances and housing
The whole system is purged by nitrogen to avoid dust in the pipes and on the optical equipment. For the
measurement of CO and O2 two lasers are installed: one small band laser for CO and another small
band laser for the detection of O2. Two lasers have to be used because the frequencies of CO and O2 are
too far apart to be covered by one laser. These small band lasers scan across the defined frequency band
continuously. A visualisation screen (Figure 17) in the furnace control station shows the current and the
historical values of CO, O2 and temperature. The historical data of the heats are stored such that the
furnace performance can be evaluated and enhanced. The outputs can be used to develop a dynamical
process regulation in order to increase the post combustion degree in the furnace as well as in further
consequence the chemical and electrical energy efficiency. In view of security it is possible to take
measures in time to avoid bag house damages due the fast response time of the laser system. A high
water and/or dust content does not cause a problem with this type of system. The measurement works
without interruption even under worst conditions (dirty scrap, carbon injection).

Figure 17: Visualisation screen

During the project, several measures were undertaken regarding the LINDARC off-gas analysis system
at the MH furnace, to improve its reliability and to allow an easier maintenance. First of all the
alignment of the water cooled lance for housing the laser senders and transmitters was improved. An
easy alignment procedure is required to ensure an optimal transmission of the laser beam in long term
operation. As shown in Figure 18, three adjusting screws were installed at the housing of the laser.
Thus the alignment can easily be checked and optimised during maintenance stops. Furthermore, a total
revamping of the electric units of the system, with assembly of a switch board in an air-conditioned area
was performed, see Figure 19. This ensures the protection of the electronic against high temperatures,
dust, electromagnetic fields etc., and allows an easy maintenance.

20
 3 x adjusting screws (120°)

Figure 18: Three adjusting screws for the housing of the laser

Figure 19: Electric units with assembly of a switch board in an air-conditioned area
Figure 20 shows the laser measurement signals of the oxygen and carbon monoxide laser during one
heat. Due to an increase of the cooling water flow rate of the water cooled lances from 100 to 120 l/min
a durability for the lances of at least 20.000 heats was achieved. The oxygen measurements are very
reliable without any maintenance.

Figure 20: Off-gas laser measurements with carbon monoxide laser drop out

21
However, since the revamping of the electronic switchboard the carbon monoxide signal is not stable
any more. The measurement drops out for an undefined time and begins again without any effort. The
reason for that behaviour was not determined exactly up to now. A malfunction of an electrical circuitry
was assumed. Thus electronic parts were replaced (Figure 21) step by step to get rid of this problem.

Figure 21: Electronic board encased by a water cooled spiral

To assess the state of operation and the energetic performance of the Marienhütte AC furnace at the
beginning of the project, to judge the efficiency of the chemical energy input, as well as the
improvements which were achieved with the developments performed within the OFFGAS project, a
statistical model for calculating the electrical energy demand of arc furnaces has been used. This model
has been developed by BFI by statistical analysis of average values from more than 50 furnaces [10,
11]. Within a recently finished ECSC project [12], a detailed evaluation of single heat values of more
than 5000 heats from 5 EAF‘s of different type (AC, DC, shaft, twin shell) was performed. This
evaluation led to a modification of the parameters of the model, in order to allow an accurate
calculation of the electrical energy demand both for (daily or monthly) average as well as for single heat
consumption figures. With this statistical model, the electrical energy consumption of a furnace can be
assessed in comparison to other furnaces, and variations of electrical energy consumption at the same
furnace can be analysed. The formula for calculating the electrical energy demand of arc furnaces is
given with the parameter values determined in [12] in Table 2.
Table 2: BFI formula for electrical energy demand of arc furnaces

It takes into account the specific consumption of total and several individual ferrous materials, slag
formers, burner gas, oxygen for blowing by lances and injectors as well as for post-combustion,
temperature before tapping and tap-to-tap time, divided in power-on and power-off time. All
consumption values - also the actual electrical energy consumption WE for comparison with the
calculated demand WR - are related to the tap weight, i.e. the weight of steel in the ladle per heat,
including ferroalloys added at tapping. For measured energy losses by vessel cooling and gas
exhaustion, the deviation for an individual heat from the mean value over all heats is taken into account
by the factor NV, which has to be determined specifically for each furnace. As the mean value of the
electrical energy demand is thus not affected by the energy losses, furnaces with and without
measurement of these losses can be compared in evaluating their electrical energy consumption. For an

22
individual furnace however, the variation in energy losses from heat to heat may explain a considerable
part of the variation in electrical energy consumption [12].
For first calculations with the statistical BFI model for the electrical energy demand, process data and
consumption figures of 136 heats produced at the AC EAF of Marienhütte in the period from 15. –
19.09.2006 were collected. From the data supplied by MH, the specific electrical energy demand WR
was calculated and compared to the actual values WE. In Table 3 the average input values used for this
evaluation are listed for the first trial campaign.
Table 3: Average input values of MH heats for BFI statistical model for the electrical energy demand
Variable WR WE GA gE gScrap gZ TA tS tN MG MO2G ML MLJ1 MLJ2 MLJ3 MN MCoke
Unit kWh/t kWh/t t kg/t kg/t kg/t °C min min m3/t m3/t 3
m /t
3
m /t
3
m /t
3
m /t
3
m /t kg/t
Campaign 15.-19.9.06
351.7 378.2 36.1 1151 1141 36.2 1631 28.4 13.3 6.7 13.4 6.8 16.3 6.4 9.0
136 heats

Regarding the input values for the MH furnace, the following explanations have to be given:
gScrap is the amount of metallic charge materials related to the tap weight. There is no distinction
between different scrap types. Only the charging of skulls is recorded separately. In gE additionally the
amount of alloy materials added at tapping is included.
gZ is the specific amount of slag former additions (only lime).
MG is the specific amount of burner gas for the door burner and the three KT injectors in burner mode
respectively shrouding gas in oxygen injector mode.
MO2G is the specific amount of burner oxygen for the door burner and the three KT injectors in burner
mode resp. for shrouding in oxygen injector mode. A stoechiometric consumption of oxygen is assumed
(MO2G = 2 * MG), as no separate recording of oxygen consumption figures of the KT injectors for
shrouding is performed.
ML is the specific oxygen consumption of the door lance
MI1-3 is the total specific oxygen consumption of the three KT injectors, minus the stoechiometric
consumption for combustion of the natural gas for shrouding.
MN is the specific oxygen consumption of the 4 post-combustion injectors.
MCoke is the specific consumption of carbon input (sum of batch charging and injection).
500
Mean deviation: -27 kWh/t
calculated electrical energy demand in kWh/t

Standard deviation: 22 kWh/t

450

400

350

300

250
250 300 350 400 450 500
actual electrical energy consumption in kWh/t

Figure 22: Calculated versus actual electrical energy demand for MH single heats
In Figure 22, the electrical energy demand calculated according to the BFI statistical model is plotted
against the actual electrical energy consumption for the evaluated 136 single heats of the MH furnace.
The mean deviation of the model calculations, that means the average difference between calculated

23
and actual electrical energy consumption, is -27 kWh/t. It is a measure for the energetic performance of
the furnace, as with the statistical model it is compared to a standard EAF with an effective utilisation
of electrical and chemical energy input. The fact that the mean error is negative means that the energetic
performance of the MH furnace is not as good as could be expected from a standard performance.
Possible reasons result from a more detailed evaluation of the average consumption figures of the
furnace: (a) The energetic efficiency of the blown oxygen is not as high as assumed within the statistical
model. Different to the expected value of 4.3 kWh/m³, the efficiency is about 4.1 kWh/m³ for the
oxygen input from the door lance, and only about 3.4 kWh/m³ for the KT injectors; (b) The ratio
between total oxygen input (door lance + KT injectors + Post combustion injectors = 29.5 m³/t) and
carbon addition (batch charging + injection = 9 kg/t) is very high. Thus possibly not enough
carbonaceous material to be combusted by the injected oxygen is available. This might also explain the
high oxygen content in the EAF off-gas, as can be found from the off-gas analysis.
2.2.1.4 Installation and commissioning of EFSOP off-gas analysis system at ORI (T 1.4)
Background
Electric arc furnaces (EAF) are generally operated according to a fixed working pattern. To improve
EAF operations a dynamic control system is required that allows to reduce the energy demand (and
specific CO2 emissions) and to enhance EAF productivity. This aspect constitutes the main goal for
electrical steelmaking, but as important as it is also to improve repeatability and flexibility of the
process.
In this direction, the off-gas analysis system coupled with a dynamic process simulator plays an
important role; can therefore be better understood the oxidation process thus allowing an on-line
process control.
Installation of off-gas analysis system
This task is devoted to the installation of the off-gas analysis system on the Consteel® Plant of Ori
Martin. The EFSOP system is a patented water cooled gas sampling probe coupled with sensors for
measurement of O2, CO, CO2, H2 with proven industrial technology (IR for CO, CO2, thermal
conductivity for H2, electrochemical analysis for O2).
The off-gas analysis is necessary to enhance the Consteel® process control. The final scope is to
maximize the energy input controlling, at the same time, the composition of the steel and the slag
through:
 Dynamic control of O2 and C injection to enhance the efficiency of the post combustion;
 Dynamic control of O2 and C injection in order to optimize the melting and refining process
control.

Room with control panels and off-gas analyzer

EFSOP probe
cabinet
Figure 23: EFSOP probe installed and the box with control panels of gas probe and gas sensors
Figure 23 shows the EFSOPTM probe installed on the first Consteel® hood and the room with the
EFSOPTM control cabinet and gas sensors.
The off-gas sample is drawn to the analyzer cabinet using a high volume pump. In order to minimize
the condensation the off-gas passes through a heated sample line to minimize condensation of water

24
vapour in the sample. At the analyzer, a rugged conditioning system further cleans the sampled gas with
more filters and condenses the water vapour out of the sample. The clean, dry sample is then analyzed
for carbon monoxide (CO), carbon dioxide (CO2), hydrogen (H2) and oxygen (O2).
The off-gas measurements are converted to 4 to 20 mA signals and sent to the PLC. The sampling
system also uses an automatic compress air purging system, to purge the analyzer filters and the heated
sample line. The EFSOP acquires and stores automatically gas analysis data, including gas temperature.
Details about gas temperature measurements are reported in section 2.2.5.4. Figure 24 shows the user
friendly graphical interface of EFSOPTM system.

Figure 24: EFSOP cabinet HMI - Off-gas analysis system detail

2.2.1.5 Installation and commissioning of off-gas analysis system for long term analysis at PA (T 1.5)
Initially the plant of Esch-Belval was selected but due the interest of Differda nge, the trials were
performed there. The plant managers of Differdange were readily interested in off-gas analysis
information due to the high improvement potential, the Differdange furnace being characterised by long
burners and post-combustion preheating phases.

Gap
Water supply

Probe

PC Roof
chamber
Probe

Slag
door

Figure 25: Details of the probe installation in AM Differdange

The Differdange furnace is a twin shell DC furnace, the tap weight is 155 tons/heat. The carbon steel
production reached 1.25 kton in 2006. From the beginning of the project, process and maintenance
engineers showed a high involvement. As an example, a direct connection on high pressure cooling
water of the elbow was authorised. Existing side holes were available in the elbow and an existing
footbridge facilitates the installation of the measurement point at the roof elbow. Figure 25 shows the
details of the installation at ArcelorMittal Differdange. The water supply is given in the left side picture.

25
The right side show a general view of the furnace and its connection to the post-combustion chamber;
the gap is closed, the furnace having been stopped for maintenance operations. Figure 26 detailed the
RWTH probe.
Gas sampling tube

Water Water
inlet outlet

Figure 26: Details of the RWTH water cooled probe for off-gas sampling
It consists of a double envelope cooling tube which provides a sealed connection to the gas analysers.
This simple and robust design allows long term lastingness in the harsh environment of the EAF.
Figure 27 illustrates the location of the probe sampling the exhaust gas directly in the furnace elbow.
The behaviour of the probe was satisfying, the measurement being reliably achieved during a complete
week without any intervention on the measurement equipment. This location before the gap provides
the off-gas information directly at the furnace exit, avoiding the possible influence of the mixing zone
of the gap. Anyway, the design of the gap which enters the post-combustion chamber in both plants
does not allow to locate the sampling probe after the gap as done for the other plants in this project. The
RWTH analysis system allows quantifying the following gas species: O2, CO, CO2, H2, and CH4. The
nitrogen content is assumed to balance these compounds. Due to the location of the gas analysers and
the used pump, the response time of the measurement is around 3 minutes; it should be possible to
reduce it to a value lower than 1 minute by increasing the gas sampling flow rate or one may decrease
the length of the off-gas sampling tube. Two measurement campaigns in the furnace elbow have been
performed in Differdange. The first one allowed defining all the technical installation aspects (location
of the probe, path for the gas sampling pipe, installation of the analysers in a safe technical room close
to the furnace, availability of the purging air and electrical supply). Some improvements have been
brought for the second campaign: (a) synchronisation of the cleaning sequence with the tapping
operation thanks to a plant information provided by the PLC hardware; (b) the availability of all
injectors prior to the recording (one PC lance was missing during the first campaign), steady working
operations are required for the measurement in order to guarantee the reliability of the conclusion which
will be deduced; (c) the analysers information has been transferred to the plant computers in order to
allow operators following the gas analysis during the whole melting process. (d) a second measurement
point has been added in order to appraise the exhaust gas flow rate.

Figure 27: Location of the sampling probe in the elbow at the AM Differdange furnace
View at start-up (left) and after three working days (right)
For the second measurement campaign (2007), a second sampling point was located in the exhaust duct
as illustrated in Figure 28. Figure 29 shows the point A at the EAF. The probe was installed at the
midpoint of the EAF elbow. So off-gas has been sampled without any influence of the air intake at the
gap. The second measurement point B has been added in order to appraise the off-gas volume flow rate.
The principle is to use the Venturi flow rate measurement existing in the main duct to the filter
baghouse and chimney, deducing the primary flow from a carbon mass balance. The second point B

26
measurement has been carried out with portable equipment; the recording period was limited to two
heats.

Secondary

Air Air
Point A

F2 F1 Venturi

PC
chambers

Water
Filter

Point B Fan
Quench

Figure 28: Location of the sampling point A & point B in the exhaust gas system at AM Differdange

Water cooled probe at POINT A

Figure 29: Location of the probe at point A

During the summer of 2008 a major revamping of the AM Differdange furnace has been carried out.
The upper and lower shells have been changed as well as the roof, replacing spray cooling technology
to classical tube panels. The volume of the furnace vessel has been increased, allowing to charge more
easily light scrap. As a consequence, the sampling flange in the elbow and the access footbridge have
been removed.
For the next measurement campaign (2008-2009), duct as illustrated in Figure 30, the sampling point A
was moved to the point C in the post-combustion chamber, assuming an incomplete burning of the
exhaust gas at the entrance of the post-combustion chamber.

27
 Secondary

Air Air
Point A

Point C Venturi
F2 F1

PC
chambers
Water Filter

Point B Fan
Quench

Figure 30: Location of the sampling points in the exhaust gas system
Figure 31 shows the measuring at point C which doesn’t allow anymore the appraisal of the exhaust
gas flow rate. It nevertheless provides the advantage of an easy maintenance.

Probe

Probe

Figure 31: Sampling probe installed in the inlet of the PC chamber (ArcelorMittal Differdange)
2.2.2 Off-gas measurements for control of decarburization (WP 2)
2.2.2.1 Off-gas measurements and correlation with nominal carbon input, assessment of carbon mass
balance (T 2.1)
At the investigated EAF of DEWG, the off-gas analysis at point A was performed 0.5 m behind the gap
between EAF elbow and hot-gas duct. Here, the gradients of off-gas concentration, temperature, and
off-gas velocity in the radial direction are considerable high. The installation of measurement lances at
point A experiences many problems, i.e. the lance blockage by the dust, the lance melt-down due to the
high off-gas temperature, and difficulty with temperature and off-gas concentration due to the
inhomogeneous off-gas profile in the radial direction [9]. For this reason, the measured off-gas
composition and off-gas volume flow rate at point A cannot be integrated at a whole duct area without
any correction. If the off-gas analysis is performed at the gap between EAF elbow and hot-gas duct then
the result of off-gas analysis may be integrated throughout the diameter of the EAF elbow with several
remarks and the integration to the diameter of the hot-gas duct is not reasonable. For this reason, the
integration over the bended hot-gas duct with considerable a higher diameter than the EAF elbow is not
reasonable. Therefore, further analysis developed in this work will utilize the result of the off-gas
analysis at point B.

28
Figure 32: Off-gas concentration, temperature and integrated carbon mass in the off-gas at point B (an
exemplary stainless steel heat)
The result of the off-gas analysis at point B shows a very small CO concentration which is found to be
lower than 2000 ppm (Figure 32). The contribution of CO concentration on the complete carbon mass
balance is only up to 0.19 kg/t per heat which is only equal to about 1.9 % of the total carbon mass in
the off-gas. Thus, the CO concentration at point B can be neglected for the determination of carbon
concentration in the off-gas. For this reason, the installation and maintenance cost of the off-gas
analysing system to control the elements oxidation can be reduced since the installation of a CO sensor
at point B is less important. Generally, the carbon input for the steelmaking results from the scrap, the
alloys, the NG burner, the electrode oxidation, the slag additive, and the coal injection (eq. (1)).
mC,scrap + mC,slag additive + mC,alloy + mC,NG burner + mC,electrode + mC,coal injection =
(1)
mC,steel + mC,off-gas + mC,slag + mC,dust
where a conservation law is obeyed, mC,in = mC,out.
20
. C,in
m
mC,out
m
C Coff-gas
mC,in, mC,out, mC,off-gas [kg/t]

,off-gas
15

10
mC,tap (+ mC,slag)

5
mC,off-gas

0
0 5 10 15 20
mC,in [kg/t]
Figure 33: Carbon mass balance of stainless steel heats
The total carbon concentration at point B is mostly contributed by CO2 gas since the CO concentration
at point B is below 2000 ppm. There is no carbon input from NG burner and coal injection for stainless
steel grades at the investigated EAF. The carbon concentration in the slag is very low, i.e. below

29
0.05 wt% (see Table 4), therefore its contribution on the complete carbon mass balance can be
neglected in this case. A non-consistent carbon mass balance still exists since the determination of the
carbon mass balance is based only on the result of the off-gas analysis at the primary dedusting system
and neglects the off-gas outflow through the secondary dedusting system. The difference between
carbon output and carbon input varies in the range of -9 to 18 % (average value 2.7 %) (Figure 33).
Some part of carbon concentration in the off-gas which goes out through the secondary dedusting
system will contribute on the inconsistent carbon mass balance. Another factor for the inconsistent
carbon mass balance is possibly due to the inconsistent data of carbon input mainly from the scrap and
the dust.
2.2.2.2 Correlation of measured critical points of decarburization with carbon mass balance (T 2.2)
As previously discussed, the CO concentration in the off-gas at point B is in the order of ppm and thus
it can be neglected for the determination of the carbon mass from the liquid steel during the refining
period. On the other hand, the carbon mass in the off-gas during the refining period in the investigated
EAF comes only from the oxidized carbon in the liquid steel since there is no other carbon input from
the NG burner, the electrode oxidation, and the carbon injection during the refining period. For these
reasons, the CO2 analysis can be used to control decarburization and to determine the end of oxygen
injection at the investigated EAF. For both austenitic and ferritic stainless steel grades, the result of the
off-gas analysis shows the existence of three consecutive stages of the oxygen injection, i.e. the early
stage, the middle stage, and the end stage. In Figure 34 and Figure 35 the measure CO2 content in the
off-gas is shown over the amount of oxygen injected, which is because of fixed volume flow rates also
an equivalent of time elapsed during the decarburization process. In the early stage when mainly silicon
is oxidized by the oxygen, the amount of CO2 in the off-gas is rising. The middle stage is the phase of
intense decarburization with a steady high amount of CO2 in the off-gas. In the last phase, the end stage,
most of the carbon is oxidized and the oxygen is slagging chromium and iron from the melt. The
amount of CO2 in the off-gas is decreasing in this phase. See also section 2.2.2.7 and Figure 42 for
more details.
15
Austenitic stainless steel grade

10
xCO 2 ,off-gas [%]

1st critical point 2nd critical point

(3.4 kg O2/t) (10.5 kg O2/t)

5
Early stage Middle stage End stage

0
0 5 10 15
Oxygen injection [kg O2/t]
Figure 34: Three consecutive decarburization stages during the refining period (austenitic steel grade)

30
 15
Ferritic stainless steel grade

1st critical point 2nd critical point

(8.6 kg O2/t) (17.8 kg O2/t)
xCO 2,off-gas [%] 10

Early stage Middle stage End stage

5

0
0 5 10 15 20 25
Oxygen injection [kg O2/t]

Figure 35: Three consecutive decarburization stages during the refining period (ferritic steel grade)
2.2.2.3 Adaptation of calculated amount of oxygen to measured carbon mass flow in off-gas (T 2.3)
Figure 36 shows that carbon mass in the off-gas increases with the increase of oxygen injection. It is
also shown that a certain amount of oxygen injection, i.e. 3.6 kg/t, is required to start the carbon
oxidation. However, this value should be interpreted carefully because it is based on the empirical data
of many charges, not based on charge wise. For example, CID 200881 shows not good decarburization
process because it lies far below the regression line.

10
mC,off-gas at refining period [kg/t]

8
CID 200855

2
3.6 kgO2/t
CID 200881
R2 = 0.7821
0
0 5 10 15 20

Oxygen injection [kg/t]

Figure 36: Oxygen injection vs. measured carbon content in off-gas

31
2.2.2.4 Correlation with chromium contents in slag from slag analysis (T 2.4)

Figure 37: Steel sampling (left) and slag sampling (right)

During the measurement campaign at DEWG, analysis of slag and tapped steel were performed for 18
charges in order investigate the behaviour of elements oxidation (Figure 37). The slag analysis shows
that the highest chromium content in slag is 11.40% for CID 200855 while the lowest one is 5.47% for
CID 200857 (Table 4). The average chromium oxide content in slag is 8.07%.
Table 4: Slag analysis of stainless steel heats during measurement campaign
Elements content in slag [%wt] Slag mass
CID
Al2O3 CaO FeO MgO MnO SiO2 C S TiO2 Cr2O3 [kg]

200853 5.17 35.09 2.24 11.52 3.18 26.05 0.02 0.20 0.78 11.04 18716
200855 3.78 35.91 3.62 7.46 4.20 28.55 0.05 0.18 1.00 11.40 12181
200857 5.02 40.04 0.74 8.44 3.25 31.28 0.02 0.18 1.01 5.47 17399
200861 5.69 34.72 1.41 7.69 3.19 30.05 0.02 0.79 0.76 6.88 18310
200863 4.78 36.94 3.42 7.66 3.68 29.03 0.03 0.62 0.79 8.45 17315
200865 5.17 40.48 1.61 7.35 3.05 30.92 0.03 0.72 0.88 5.79 16059
200867 4.60 36.65 2.65 7.42 3.77 32.39 0.02 0.19 0.96 7.87 17114
200869 4.32 37.23 1.42 8.17 3.73 32.20 0.02 0.21 1.06 7.96 16473
200871 2.92 36.26 3.58 10.69 4.25 26.23 0.04 0.19 0.43 10.68 13202
200873 4.47 36.28 2.69 8.55 3.51 37.78 0.04 0.21 0.91 8.17 16970
200875 4.36 37.50 4.23 7.75 3.34 29.65 0.04 0.24 1.12 7.20 16601
200877 4.83 36.17 2.30 8.68 5.37 29.00 0.05 0.17 0.98 7.50 16771
200879 5.33 36.21 1.86 11.17 4.59 29.00 0.03 0.17 0.98 6.57 16695
200881 5.18 34.90 3.78 8.75 4.56 28.26 0.06 0.14 1.01 9.40 17501
200883 4.69 37.64 2.24 10.4 3.86 29.57 0.03 0.14 1.03 6.11 18939
200887 3.95 36.66 4.27 7.57 3.67 28.04 0.03 0.18 1.03 10.26 16909
200889 3.75 37.37 6.24 7.92 3.25 26.28 0.15 0.13 0.94 9.21 16686
200899 3.30 33.19 10.91 8.53 1.62 26.14 0.03 0.18 0.81 8.55 22587
Mean 4.52 36.35 3.26 8.82 3.60 29.31 0.04 0.30 0.91 8.07 17024
The oxygen prefers to oxidize chromium and iron, instead of carbon in the end stage of the
decarburization. Figure 38 shows the amount of FeO and Cr2O3 in the slag in relation to the amount of
excess oxygen in the end stage. The excess oxygen in the end stage is calculated as the amount of
oxygen added after reaching the second critical point. However, the correlation between the excess

32
oxygen in the end stage and the Cr2O3 concentration in the slag is not as strong as the correlation
between the excess oxygen and the FeO concentration. On the other hand, one heat with the highest
excess oxygen indicates a higher FeO concentration in the slag compared than Cr2O3 concentration. A
weak correlation between the excess oxygen and the Cr2O3 concentration in the slag may occur since
there are distinct groups of the chromium oxidation. Although there is no clear explanation about the
existence of these distinct groups, the chromium activity is surprisingly found to decrease with a higher
chromium concentration in the liquid steel at a low carbon concentration and a high chromium
concentration. For this reason, the chromium oxidation may be retarded and the oxygen will thus
oxidize other elements instead of chromium, e.g. iron.

Figure 38: Influence of excess oxygen injection on chromium and iron oxidation (without
decarburization control)
2.2.2.5 Development of empirical model of optimum decarburization (T 2.5)
It is very difficult to mathematically estimate the 2nd critical point of oxygen injection since many
parameters should be taken into consideration, e.g. the thermodynamic, the mass and heat transfer and
the reaction kinetics. In many cases, these influencing parameters can be contrarian to each other. For
instance, the discrepancy of the thermodynamic and mass transfer influence about the sulphur presence
on the decarburization rate. Moreover, the validation of the models with a well controlled input mass of
elements from pilot plants may result in some difficulty to scale the empirical model to industrial
production process. The empirical models for the critical points of oxygen injection have been
developed based on the results of off-gas analysis and operational data. The empirical models for the 1st
and 2nd critical point of decarburization are functions of the elements and electrical energy input before
the refining period starts (eq. (2) ).

mO , A· a ·m a ·e B (2)

mi Mass of dust, iron, carbon, silicon, manganese, aluminium, phosphor, sulphur, chromium,
molybdenum, nickel, copper, oxygen, CaO, MgO, SiO2, Al2O3 input before refining
e specific electrical energy input before refining
A, B, ai Coefficients derived from and fitted by empirical modelling
Figure 39 shows the 2nd critical point of oxygen injection calculated by model in relation to the value
derived from off-gas analysis data for 18 heats. The 2nd critical point varies between 6 and 14 kgO2/t for
these heats and a good agreement between model and real value can be observed.

33
Figure 39: The empirical model of the 2nd critical point of oxygen injection
2.2.2.6 Test of model and adaptation of model parameters (T 2.6)
Before on-line plant trials, off-line plant trials were performed in order to identify possible unexpected
problems. For this reason, empirical model of chromium concentration, which takes account the amount
of oxygen injection, was developed (Figure 40). For these off-line plant trials, carbon concentration,
nickel concentration, molybdenum concentration and mass of tapped steel were also developed.
30
105 stainless steel heats (Cr-Ni)
xCr,steel [wt%] - Model

20

10

R² = 0.95
0
0 10 20 30
xCr,steel [wt%] - Real

Figure 40: Real values vs. model values of chromium concentration at tapping (stainless steel heats)
In the first step of this off-line test, the oxygen consumption based on the empirical model developed in
section 2.2.2.5 is calculated (cf. Figure 39). After that, chromium, carbon, nickel, and molybdenum
concentration as well as mass of tapping steel are determined (Figure 41). The off-line plant trial show
a good result. As expected, the oxygen consumption can be decreased from 12.91 kg O2/t down to 9.27
kg O2/t. The nickel and molybdenum concentration decreases due to an increase of carbon and
chromium concentration due to less oxygen injection (cf. Table 5).

34
 Start

Data of scrap, alloy, and slag

additive input

Determination of oxygen
consumption (2nd critical point)

Determination of elements concentration

and steel mass at tapping

End

Figure 41: Flow chart of off-line test strategy

Table 5: Result of off-line test

Without With
Parameter
decarburization control decarburization control
[kg 12.91 9.27
Oxygen consumption O2/t]
Chromium concentration [wt%] 17.83 18.15
Carbon concentration [wt%] 0.59 0.78
Nickel concentration [wt%] 8.79 8.48
Molybdenum [wt%] 0.81 0.80
concentration
Mass of tapped steel [t] 110.08 110.48

2.2.2.7 Visualisation of calculated chromium oxidation from thermo-chemical modelling (T 2.7)

The input data for the simulations presented in this section includes the type and mass of elements input
in the liquid steel and the slag before the refining period as well as the initial and end temperatures of
the oxygen injection. After the simulation result is generated, the amount of the oxidized silicon,
chromium, and iron in the slag phase as well as carbon in the gas phase are then calculated as sum of
their oxides (eqs. (3) to (6)).
28 28
mSi,slag  mSiO2 ,slag  mSiO,slag (3)
60 44
12 12
m C,off gas  m CO,off  gas  m CO2 ,off  gas (4)
28 44
104 52
m Cr,slag  m Cr2O3 ,slag  m CrO,slag (5)
152 68
112 56
m Fe,slag  m Fe2O3 ,slag  m FeO,slag (6)
160 72
From the result of thermochemical simulations, two critical points of the oxygen injection referring to
the decarburization process can be identified, i.e. the 1st and the 2nd critical point (Figure 42). For this
reason, the simulation result shows the existence of three consecutive stages of the oxygen injection, i.e.

35
the early stage, the middle stage, and the end stage. The 1st critical point correlates with the critical
silicon concentration, mSi,critical, while the 2nd critical point correlates with the critical carbon
concentration, mC,critical. The 1st critical point ends the early stage. The middle stage is between the 1st
and the 2nd critical point while the end stage starts after the 2nd critical point appears. An intensive
silicon oxidation by the injected oxygen is shown to take place in the beginning of the refining period
before the 1st critical point is reached, i.e. at 4.8 kg O2/t for this particular heat. No carbon oxidation is
found in this stage. Referring to the Ellingham-Jefferson Diagram, the change of standard Gibbs
energies for the magnesium, aluminium, titanium, and silicon oxidation are lower than the carbon and
chromium oxidation when the oxygen injection starts at about 1500 oC. Before the oxygen starts to
intensively oxidize chromium and carbon, the intensive silicon oxidation takes place until the relative
silicon-lean condition in the liquid steel is achieved.
The result of the thermodynamic simulation shows that the intensive carbon oxidation takes place after
the intensive silicon oxidation finishes, i.e. in the middle stage, while the chromium oxidation takes
place in the middle and the end stage. After the 1st critical point is reached, the generated carbon oxides
in the off-gas increases with the increase of oxygen injection. On the other hand, the chromium
oxidation seems to follow an exponential path at the end of oxygen injection. After the 2nd critical point
appears, the carbon oxidation decreases, the intensive chromium oxidation continues further, and the
intensive iron oxidation starts. This simulation result shows that the chromium oxidation is more
intensive than the iron oxidation since the intensive chromium oxidation already takes place since the
middle stage.
15
Temperature 1500 oC
CID SS-XP30284
Mass of oxidized elements [kg/t]

Early stage Middle stage End stage

10
C

1st critical point 2nd critical point

(4.8 kg O2/t) (18.3 kg O2/t)
5
Si
Cr

Fe
0
0 5 10 15 20 25
Oxygen injection [kg O2/t]

Figure 42: The result of the thermodynamic simulation of elements oxidation

2.2.3 Development of on-line energy monitoring system in stainless steel production (WP 3)
2.2.3.1 Data acquisition of the off-gas analysis (gas composition, volume flow rate and temperature)
to the EAF process control (T 3.1)
Gas composition
As already described in section 2.2.1.2 off-gas sampling probes have been installed at point A and point
B in the primary dedusting system at TKN Bochum in combination with ABB off-gas analysis systems.
The water-cooled probe at point A has been installed at the centre of the elbow. An ABB off-gas
analysis system for point A has been installed in the doghouse. The signals of CO, CO2, O2 gas
concentration have been made available to the EAF process control system and are displayed on a
monitor in the control room (Figure 43). The concentration of the off-gas sampled at point B is also
measured and provided to the EAF process control system.

36
 3

Figure 43: Monitor display of ABB analyzer unit at the dispatcher

1 - off-gas composition; 2 - On/Off; 3 - display of calibration gas inlet
Volume flow rate
As already shown in the schematic in Figure 5 (section 2.2.1.2) the EAF at TKN Bochum is housed.
The amount of primary to secondary off-gas volume flow is regulated by the DEC (Direct Exhaust
Control) flap. For typical EAF operation the DEC flap is opened to maximum value while melting and
opened to 50 % while charging and tapping in favour of the bypass duct of the secondary dedusting
system. With the closure of the DEC flap the capacity of the canopy hood dedusting is therefore
increased in order to provide maximum collection of secondary off-gas in the doghouse. Within this
doghouse secondary off-gas is coming from gaps while tapping and charging. The concentration of off-
gas from the doghouse in the dedusting system is very low because of the high secondary volume flow
rates. The sum of primary and secondary volume flow is more or less constant during the heat. The
primary off-gas volume flow before mixing at the bypass amounts to about 100,000 m3(Vn)/h. The
secondary off-gas volume flow amounts to about 450,000 m3(Vn)/h.
The off-gas dedusting system at TKN is also equipped with a sliding muffle. This sliding muffle is
installed to regulate the amount of air intake through the gap. During the melting process the sliding
muffle moves forward to close the gap, so that the amount of infiltrated air is minimized. When the
oxygen is injected into the melt the sliding muffle moves backwards and opens the gap. Within this
muffle position there will be air intake to burn the CO content of the off-gas stream coming from the
EAF vessel.
As already described in section 2.2.1.2 for the initial measurements a pitot tube measurement was
installed at point B. Figure 44 shows the Pitot tube installed and the data captured and derived.
700 — Temperature [°C]
— Differential pressure [Pa]
— V [m (V )/s]
dp [Pa]; V [m (Vn)/s]; Temperature [°C]

3
n
600
— V [m (V )/s]
3
B

500

400

300
3

200

100

0
0 20 40 60 80 100 120
Time [min]

Figure 44: Pitot tube for measuring the differential pressure at point B (left); Data acquisition: Off-gas
temperature, differential pressure and volume flow rate (right)

37
The plotted off-gas volume flow rate is not constant. The off-gas volume flow rate is changing due to
the DEC flap position and the sliding muffle. In case the DEC flap is been completely opened the
volume flow rate is 80 m3(Vn)/s (150 m3(Vb)/s). In case the DEC flap is been almost closed (20%) the
volume flow rate is 25 m3(Vn)/s (80 m3(Vb)/s). Equation (7) shows the derived relation between off-gas
volume flow rate and DEC flap position (fDEC [%]), which will be used for the calculation of volume
flow rates within the energy monitoring system and dynamic energy process model.

off  gas [m N / h]  0.6875  f DEC  11.25
3
V (7)

Temperature
Additionally the off-gas temperature has been measured at point B before the DEC flap. This off-gas
temperature is also continuously measured and recorded by a pen writer.
2.2.3.2 Acquisition of temperature and volume flow rate of the hot water system to EAF process
control (T 3.2)
The hot water cooling system at the Bochum EAF consists of two circulations. The first one is the wall
cooling system of the furnace with a forced circulation, and the second circulation is the hot gas line
cooling system which is buoyancy driven under operating conditions. The wall cooling system of the
furnace consists out of 14 single panels which are built as pipe-by-pipe constructions each with a
cooling water inlet and outlet. The panels are connected in parallel to the cooling water system and have
a total cooling surface of 37.5 m². The hot gas line cooling system is also a pipe-by-pipe construction
with a total cooling surface of 196 m². In this system the hot primary off-gas of the EAF is post-
combusted and cooled down. The cooling water in both circulations has an inlet temperature of Tinlet =
220°C at a pressure of pB = 20 bar. The heat transferred into the hot cooling water is not leading to an
increase of temperature but to a partial evaporation. The mixture of cooling water and saturated steam
produced is fed into a steam drum, where the saturated steam is separated from the cooling water and
subsequently dispensed into the plants steam network. Figure 45 shows a schematic of the hot water
cooling system described.

Steam to plant network (p=20bar) Sound absorber

Water to plant network

p = 20 - 23 bar
Condensor
M Water from
Vessel plant network

Separator
Hot gasline cooling system
Buoyancy Circulation
Valve pB=20 bar
A=196 m2)

M Sleeve
Circulating
pumps
PW=144 kW
p=7 bar

Wall cooling system n=14

Recovered Energy Forced Circulation
45 to 56 kg Steam/tSteel pB=20 bar
30 to 37 kWh/tSteel A=37,5 m2

Figure 45: Pressurized hot water cooling system at the Bochum EAF
The usual way to measure the amount of heat transferred into a water cooling system, by measuring the
volume flow of cooling water and the inlet and outlet temperature, is not applicable in this case. This is
because the single-largest part of the heat transferred is absorbed by the evaporation of cooling water
and only a small part is going into an increase of temperature. Therefore a measuring technique for a
hot water cooling system would have to take into account the amount of steam produced inside the
system.

38
A possible solution for this task is described in a patent by Brod [3]. The patent describes a way to
prevent the overheating of hot water cooled wall elements of EAFs by measuring the change of density
of the hot cooling water in a wall element. This measurement can be accomplished by using a restrictor
in the inlet and outlet of each wall element to record the mass flow of the cooling water in known
condition (pressure, temperature, density) at the inlet and calculate the density of the cooling water in
the outlet out of the differential pressure reading of the restrictor in the outlet. Even though this
measuring setup would provide detailed data on the heat transferred to the hot water cooling system of
the furnace, this solution is not feasible because of the high constructional and financial effort needed.
The budgeted time and costs for this task do not allow an implementation of this measuring setup.
Another more general way to determine the amount of energy transferred to the cooling water system is
to use the measurement setup already in place. Two sets of measurement data are of interest here. First
the amount of steam dispensed into the plant network measured after the steam drum and secondly the
amount of steam produced in the wall cooling system, which is measured in the return line of the hot
cooling water from the furnace to the steam drum. The data of this two metering points are recorded on
pen writers. An example of this data can be seen in Figure 46. Out of the given conditions of the
saturated steam (p = 20 bar, TSteam = 220°C) an energy content of 2.38 GJ/tSteam can be assumed and the
energy flow from the furnace and gasline cooling system can be derived from the steam volume flows
measured.

Amount of steam produced 0-40 t/h

Figure 46: Chart of the steam produced in the hot water cooling system at the Bochum EAF
2.2.3.3 Development of energy monitoring system and dynamic energy process model (T 3.3)
The dynamic energy process modelling and monitoring system developed is primarily based on
measurement data acquired from the primary dedusting system. The measurement data used for the
system are the off-gas composition measured at point A and point B simultaneously and the off-gas
temperature measured at point B (cf. Figure 5). Since the setting of the DEC flap is usually not changed
during a heat, a constant standard volume at point B of 30 m³ (STP)/s is assumed regarding calculations
and modelling within the system.
Additionally used are the data acquired from the hot water cooling system of the EAF as described in
section 2.2.3.2, more precisely the total amount of steam produced in the cooling system and the
amount of steam produced in the wall cooling elements of the furnace. All other values presented in the
system are derived from these values by simple analytic equations or modelling.
The system is able to display to the furnace operator the off-gas composition and temperatures at point
A and B. It also reports the actual energy loss by water cooling as well as the thermal, chemical and
total energy loss by the off-gas at point A and B. It also computes and displays the time dependent
course of the integrated energy losses.
The heat transfer rates from the furnace system to the cooling water system (CWS) are calculated
according to equation (8) using the specific energy content of the steam of eSteam = 2.38 GJ/tSteam at the
given conditions of the steam (p = 20 bar, TSteam = 220°C) and the mass flow rates of steam measured.

39
This way the heat transfer rates for the total system as well as the furnace are calculated. The heat
transfer rate in the hot gas line cooling system is calculated as balance according to equation (9).

CWS  m Steam  eSteam
Q  (8)
  
Q CWS,hot gasline  QCWS,total  QCWS,furnace (9)

The losses through the off-gas at point A and point B can be divided into the sensible heat and the latent
heat of the off-gas. The sensible heat can be calculated according to equation (10) where the molar
enthalpy hoff-gas can be substituted by equation (11).
  n
H th off  gas  h off  gas (10)
T
h off gas   x i   cp,i  dT (11)
T0

di
c p,i  a i  bi T  ci T 2  (12)
T2
Here the molar heat capacity cp can be again substituted by equation (12). The coefficients ai, bi, ci and
di needed for the calculation of the molar heat capacity have been taken from literature. The molar flow
rate can be derived from the mass flow of off-gas and the molar mass of the off-gas according to its
composition (equations (13) and (14)).
 off  gas
m
n off gas  (13)
M off gas

M off gas   x i  Mi (14)

The off-gas mass flow again can be deduced from the volume flow (STP) of off-gas and the standard
state density off the off-gas according to equations (15) and (16). For all calculations values for the
molar mass Mi and density N,i have been taken from literature.

 off gas  V
m STP,off  gas STP,off gas (15)

STP,off gas   x i STP,i (16)

Regarding the off-gas composition nitrogen accounts for a great percentage of the off-gas. But since it
is hard to measure and because it is assumed from prior measurements that there are no other significant
species besides the measured O2, CO and CO2, the amount of N2 in the off-gas is determined by a
balance according to equation (17).
x N 2  1  x CO  x CO2  x O2 (17)
The latent heat of the off-gas is depending on the content of combustible gasses in the off-gas. At TKN
only CO has been measured therefore this is the only combustible gas taken into account in the
modelling system. The latent heat can then be calculated using the standard volume flow of off-gas as
well as the amount of CO in the off-gas xCO in combination with the net calorific value hCO, the
density CO and the molar mass MCO as shown in equation (18).

 V  CO  h CO  x CO
H ch off  gas  (18)
M CO
The total heat flow rate of the off-gas is then according to equation (19) equal to the sum of latent and
sensible heat.
  
H off  gas  H ch  H th (19)

40
At measurement point B all needed values for these calculations like off-gas composition, off-gas
temperature and standard volume flow are known from measurement or assumption. At measurement
point A only the off-gas composition is known. Figure 47 shows the measured off-gas composition at
points A and B for an exemplary heat at TKN. An aggravating factor is that between measurement point
A and point B through the gap between furnace elbow and primary dedusting system an unknown
amount of air is diluting the off-gas and then furthermore reacting with the CO in the off-gas in the post
combustion zone.
However assuming that the leak air is free of any carbon a carbon mass balance between point A and
point B can be used to determine the off-gas volume flow rate at point A. Equation (20) for the
calculation of the volume flow rate at point A is derived from the carbon mass balance.
MC MC
x CO,B CO   x CO2 ,B CO2 
M CO M CO2
 V
V   (20)
A B
M MC
x CO,A CO  C  x CO2 ,A CO2 
M CO M CO2
With the volume flow rate at point A and the CO content in the off-gas known from measurement the
latent heat of the off-gas can directly be calculated using equation (18).
42 — N2 [%] — CO2 [%] 120
— O2 [%] — CO [%]
35 100

28 80
O2 [%], CO2 [%]

N2 [%], CO [%]
21 60

14 40

7 20

0 0
0 10 20 30 40 50
Time [min]

2400 — N2 [%] — CO2 [%] 120

— O2 [%] — CO [ppm]
2000 100
CO2 [%]; O2 [%]; N2 [%]

1600 80
CO [ppm]

1200 60

800 40

400 20

0 0
0 10 20 30 40 50
Time [min]

Figure 47: Off-gas composition monitoring at point A (top) and point B (bottom)

41
To estimate the sensible heat and the total heat flow rate of the off-gas at point A an energy balance is
used. According to equation (21) it is assumed that the sum of the heat flow rate at point B and the heat
transfer rate to the water cooling of the hot gas line is equal to the total heat flow rate at point A.
  
H off  gas,A  H off  gas,B  QCWS,hot gasline (21)

The sensible heat at point A can then be derived from the total heat flow rate and the already calculated
latent heat (see equation (19)). Using the equations for the molar enthalpy (11) and the molar heat
capacity (12) it is possible to deduce an off-gas temperature from the sensible heat of the off-gas at
point A.

Figure 48: Operator view of the dynamic energy modelling and monitoring system
Figure 48 shows the operator view of the developed dynamic energy modelling and monitoring system
with exemplary data. The two monitors in the top row show the measured off-gas composition and the
modelled off-gas temperature at point A (left) as well as the measured off-gas composition and
temperature at point B. In the middle row the heat flow and mass flow rates for the off-gas at point A
(left) and at point B (centre) as well as the heat transfer rates into the cooling water system (right) are
shown. The bottom row is showing the time dependent course of the integrated energy losses for point
A (left), point B (centre) and the cooling water system (right).
2.2.3.4 Efficient oxygen use and optimum process rules regarding electric energy demand (T 3.4-3.6)
This section covers the tasks 3.4 “Assessment of efficient oxygen use and electric energy input”, 3.5
“Definition of optimum process rules regarding electric energy demand” and 3.6 “Plant trials,
adaptation of process model parameters and rule definitions”.

42
During oxygen injection, the volume of the hot reaction zone, VL, depends on the penetration angle, Ф,
the penetration depth, L, and the free distance between the oxygen lance and the surface of the liquid
steel, L0, (eq. (22)) [14].

 _
 tan 2   [L d  L 0 ) 3 L30 ]
2
  (22)
VL 
3

3
2

Figure 49: Reaction zones during oxygen injection into the liquid steel [14]. Hot reaction zone (1),
Liquid steel-slag zone (2), Slag zone (3), and Liquid steel zone (4)
There is abundant oxygen amount inside the hot reaction zone. Thermodynamically, the free oxygen,
mO2,free, preferably reacts with the less-noble elements, e.g. silicon, in the beginning of the refining
period. The required oxygen for silicon oxidation at a particular reaction site and at a particular time,
mO2,Si-ox, can be simply determined from the stoichiometric thermodynamic equilibrium (eqs. (23) and
(24)). However, due to the abundant oxygen amount in this zone, the oxygen amount is
stoichiometrically much higher than is required to react with silicon. This circumstance results in excess
oxygen, mO2,ex. The excess oxygen can be determined from the difference between the free oxygen and
the required oxygen for silicon oxidation (eq. (25)).
Si + O2 ↔ SiO2 (23)
32
m O 2 ,Si  ox  m Si (24)
28
32
m O 2 ,ex  m O 2 ,free  m Si (25)
28
This excess oxygen therefore reacts with other elements inside the hot reaction zone, which are nobler
compared than silicon, e.g. carbon and chromium. In case the amount of this excess oxygen is still
excessively high only for carbon and chromium oxidation at a particular time and a particular reaction
site, the rest excess oxygen also oxidizes the next nobler elements, e.g. iron.
Different than the melting strategy of stainless steel making at DEW Siegen whereby oxygen is injected
only after the liquid steel reaches its liquid phase, the oxygen injection of stainless steel making at TKN
Bochum is simultaneously performed with electrical energy input. Since not all charged scrap and alloy
reach their liquid phase, the oxygen only oxidizes small part of liquid steel. This causes extra large
amount of excess oxygen due to the limited oxygen capacity of silicon (eq. (24)). This excess oxygen
will oxidize the next nobler elements, e.g. manganese, chromium, iron, and nickel. Due to extra large
amount of excess oxygen, nickel oxidation can be even found during stainless steel making at TKN

43
Bochum. Moreover, Cr2O3 concentration in slag was found to be very high compared than DEW
although the amount of oxygen injection is much lower compared than DEW. Different than DEW, the
oxygen injection in EAF of TKN Bochum is subjected only to oxidize silicon, i.e. only until the 1st
critical point, while further decarburization is performed in AOD converter.
In order to avoid the oxidation of precious metals, e.g. iron and metal, the oxygen injection at TKN
Bochum is suggested not to be performed in the very beginning whereby only small part of charged
scrap and alloy reaches their liquid phase. However, if the oxygen injection is performed some minutes
later, the tap-to-tap time may increase. Meanwhile, oxygen injection at TKN Bochum did not strongly
correlate with the tap-to-tap time during plant trials (Figure 50). In contrast, the electrical energy input
correlated more strongly with the tap-to-tap time.
  Electrical energy input [kWh/t]
350 400 450 500 550 600
200
TKN Bochum
Tap-to-tap time [Minutes]

150

100

50
O2 injection
Electrical energy input
0
0 10 20 30
Specific oxygen injection [kg O2/t]
Figure 50: Oxygen injection and electrical energy input vs. the tap-to-tap time
Mainly the off-gas enthalpy at point A which can be determined from the off-gas measurement at point
B (see section 2.2.3.3) has been used for this investigation. The thermal off-gas enthalpy is strongly
correlated with the electrical energy input in the plant trials (Figure 51). In contrast, the chemical off-
gas enthalpy is weakly correlated with the electrical energy input. From this result, it is expected that
the electrical energy input experiences high energy loss to the off-gas.
100
Hthermal
. R² = 0.29
Off-gas enthalpy at point A [kWh/t]

Hchemical.
R² = 0.58
75

50

25

0
400 450 500 550 600
Electrical energy input [kWh/t]
Figure 51: Electrical energy input vs. thermal and chemical off-gas enthalpies at point A

44
According to the plant trials to decrease the energy loss to the off-gas, several measures can be
performed as follow:
 Sealing off all gaps in the EAF vessel.
 The slag door is closed during power on. However, the slag door at TKN Bochum is opened
during power on since oxygen injection is simultaneously performed with the electrical energy
input during stainless steel making at TKN Bochum. This is different that the melting strategy
of stainless steel at DEW.
 Besides closing the slag door in case the oxygen lances are not inserted into the EAF vessel, the
off-gas volume flow rate is suggested to be decreased.
 The thermal energy loss at point A increases with higher tap-to-tap time (Figure 52). For this
reason, a decrease of tap-to-tap time, e.g. by decreasing the power off-time, is suggested to
decrease the energy loss through the off-gas. The homogeneous convection coefficient between
liquid steel and the off-gas is estimated at 2.4 W/(m2 K) [6] and the EAF inner diameter is 7 m.
Thus, the energy loss from the liquid steel to the off-gas is estimated at about 8.7 MW when the
liquid steel temperature is at 1600 °C.
100
Hthermal
.
Hchemical.
Off-gas enthalpy at point A [kWh/t]

75

50

R² = 0.52
25

0
0 50 100 150 200
Tap-to-tap time [minutes]
Figure 52: Tap-to-tap time vs. thermal and chemical off-gas enthalpies at point A
The process model parameters have been validated by the plant trials. There were only minor
adjustments of model parameters.
2.2.4 Development of dynamic control of oxygen supply for post-combustion by off-gas laser
absorption spectrometry at Marienhütte (WP 4)
2.2.4.1 Plant trials with additional off-gas measurement by portable VIS/IR detector systems, to
provide additional off-gas data (e.g. flow rate, H2 content) for modelling purposes and
comparison of gas analysis signals (T 4.1)
On-line off-gas analysis from the laser-based system described under Task 1.3 was used to study the
post combustion process at the MH EAF. With the help of simulation calculations and plant trials,
optimal conditions for post combustion oxygen injection were acquired.
To validate the laser-based off-gas analysis, an additional off-gas measurement was realized at point A
and a second point B downstream in the primary dedusting system. The portable off-gas analysis
system (RWTH) has been installed for collecting off-gas analysis data of further components, especially
of H2 and CO2, which are important to monitor the post combustion process. Additionally, the accuracy
of the laser-based off-gas analysis system has been checked. Therefore two portable analysis systems
(RWTH) have been installed at MH EAF. Both measurement locations are indicated in Figure 53.

45
Figure 53: Off-gas measurement points during RWTH trial campaign at MH furnace
Measurement point A was at the same position as the LINDarc water-cooled probes (Figure 54),
directly behind the gap after the elbow. Measurement point B was about 50 meters further downstream
in the off-gas system, without any significant further leakage air ingress between measurement point A
and B, besides the one through the gap.
The chosen position of the RWTH water-cooled probe is at the gap between EAF vessel and primary
dedusting system (Figure 55). At point A the off-gas analysis was performed simultaneous to that at
point B. The analysis equipment consists of water-cooled probes, filters and signal converters for
measuring the off-gas composition (CO,CO2,O2,H2,CH4) (Figure 56). At point B the off-gas
temperature is measured by a coated thermocouple. The measured analysis data for differential pressure
of the off-gas flow, off-gas temperature, and off-gas composition have been recorded continuously.

Figure 54: Installation of LINDarc and RWTH (portable) water-cooled off-gas probes at the EAF
elbow

46
 POINT A
Post combustion zone

Dust settling
Air intake

Figure 55: Air intake at the gap between EAF vessel and primary dedusting system
Filter
Off-gas sampling
Sample
preparation Analyzer

Gas heater
Cooling-water 4..20mA
Compressed air

PC

Temperature Pt/Rh30-Pt6/Rh 4..20mA

Ni/Cr-Ni Signal converter

Pressluft

Volume flow rate

4..20mA

Signal converter
Cooling-water

Figure 56: Layout of the RWTH (portable) off-gas analysis system

In Figure 57 the results of the simultaneously measured concentration curves at point A are presented.
The laser-based analysis system is detecting O2 and CO to a limited amount up to 60%. The recorded
analysis data of the two systems were corrected on the time axis. The correlation of the detected
concentrations versus the entire time is good. In the case of the CO values of the two systems correlate
in a very good way. However, the O2 values of the LINDarc system are lower in relation to the portable
system (RWTH) for about 3 % to 5 %.
Figure 58 and Figure 59 show the simultaneous measured off-gas composition at point A (O2, CO2,
CO, H2, CH4) and point B (O2, CO2, CO). Additionally shown is the power on, off-gas temperature, and
differential pressure. The off-gas volume flow rate is to be calculated from the measured differential
pressure (point B) of the gas flow. Measured off-gas mass flow rates (e.g. carbon mass flow) at point B
will be checked with the complete mass carbon balance. The carbon mass flow rate at point B will give
the value of the carbon-mass flow rate at point A. The measured H2 content at point A indicates an
additional post combustion potential of the primary off-gas. The high amount of H2 at point A is
strongly correlated to the high amount of CO. The CO post combustion is then higher when there is a
low amount of H2 at point A. However, the measured H2 content at point A gives additional information
on the post combustion potential during the melt down period. The CO2 content reaches 15 % while the
final period at point A and 12 % at point B (Figure 60). The CO2 content measured at point A indicates
the post combustion inside the EAF vessel.

47
Figure 57: Measured CO and O2 concentration curves (point A): (RWTH (portable) vs. LINDarc)
25 — O [%] — H [%] — N [%]
2 2 2 95
— CO [%] — CH [%]
O2 [%], CO2 [%], CO [%], H2 [%], CH4 [%]

2 4

— CO [%] — Power on
20 89

15 83

N2 [%]
10 77

5 71

0 65
0 5 10 15 20 25 30 35 40
Time [min]
Figure 58: Measured concentration H2 and CH4 curves (RWTH (portable system)) (point A)
25 — O [%] — Temperature [°C]
2 3000
— CO [ppm] — dP [100 mbar]
CO [ppm], Temperature [°C], dP [100 mbar]

— CO [%] 2

20 2400
O2 [%], CO2 [%]

15 1800

10 1200

5 600

0 0
0 10 20 30 40
Time [min]
Figure 59: Measured concentration curves and differential pressure, and off-gas temperature (RWTH
(portable system)) (point B)

48
 2 — oxygen — pc air
— coal — power on
— CH 4

On
1

Off
0
0 5 10 15 20 25 30 35 40
Time [min]

Figure 60: Operating data

The measurements of more than 100 heats confirmed the occurrence of high O2 and CO values at the
same time. This thermodynamic disequilibrium can be explained by the small furnace volume and a
high off-gas velocity (approx. 20 m/s) that preclude sufficient reaction time.
As the measurements indicate, the absolute O2 and CO values of the laser are lower than those of the
extractive method, especially in the melting phase. Possible reasons for deviation are: Dry gas
(extractive) – wet gas (laser). If the gas sample is dried the water molecules are removed and the ratio
of all other components increases [5] (a); Measurement point (extractive) – line (laser). The laser
technique averages the concentration of the molecules against the measurement path and doesn’t fall
victim to possible different off-gas strata’s (b); Insufficient quenching of the extractive sample at high
temperatures [13]. Probe sampling of CO is increasingly inaccurate at temperatures greater than 1000 K
(c); Inaccurate laser measurement path due to nitrogen purging. If the nitrogen flow rate is too high a
nitrogen pillow builds up and reduces the measurement path. On the other hand off-gas (and inherent
dust loading) will penetrate the lances if the flow rate is too low (d). The comparison of the off-gas
measuring data from 20 heats shows good correspondence of the CO/O2 ratio between the different
measuring techniques. In Figure 61 the average ratios of 20 heats at the measurement point A are
diagrammed. That good agreement assured the idea of using the CO/O2 ratio from the laser instrument
as basis for the dynamic control of the post combustion system (PC).
0.6

0.5

0.4
CO/O2 Laser

0.3

0.2

0.1

0
0.0 0.1 0.2 0.3 0.4 0.5 0.6
CO/O 2 VIS A

Figure 61: Correlation of CO/O2 ratio from laser and IR/VIS

49
To confirm the accuracy of the off-gas temperature values provided by the LINDARC laser-based
analysis system, a trial campaign with an additional temperature probe based on a thermocouple
provided by RWTH was performed at the MH furnace. The thermocouple probe was installed close to
the location of the LINDARC water-cooled lance (Figure 62). In addition indirect information about
the off-gas temperature, derived from the cooling water temperature at the outlet of the water-cooled
lance of the LINDARC system was evaluated.

Figure 62: Installation of the thermocouple probe close to the LINDARC water-cooled lance

70 900

water outlet Lindarc

power on
800
65 carbon
thermo couple

700
60

600
55
water outlet lance [°C]

thermo couple [°C]

500

50

400

45
300

40
200

35
100

30 0
0 20 40 60 80 100 120
minutes

Figure 63: Correlation between the off-gas temperature measured by the thermocouple and the outlet
cooling water temperature from the LINDARC water-cooled lance
As can be seen in Figure 63, a very good correlation between the off-gas temperature measured by the
thermocouple and the outlet cooling water temperature exists. The derived equation

Tcooling _ water  34.66C (26)

Toffgas 
0.03
specifies the relation between the measured off-gas temperature and the cooling water temperature of
the LINDARC lance. Figure 64 shows the measured off-gas temperature of the thermocouple and the

50
off-gas temperature calculated from the cooling water temperature using equation (26). The
correspondence between these two values is very good. Furthermore a comparison of this calculated
off-gas temperature to the one measured by the LINDARC system was performed. Figure 65 shows
that the LINDARC value increases rapidly after charging a new basket and decreases to the end of the
melt down process of each basket. The evolution of the calculated temperature from the cooling water is
oppositional to the measured one from the LINDARC system. Herein the maximum temperature is
reached at the end of the melt down of each basket, which is more reasonable. The time delay between
the calculated and the measured off-gas temperature is with a value below 20 seconds negligible. Thus
the off-gas temperature calculated from the cooling water seems to be more reliable and was used as
input for the BFI dynamic energy balance model to calculate the sensible heat of the off-gas, see under
Task 4.2.
1000

thermo couple
900 calculated temperature

800

700

600
temperature [°C]

500

400

300

200

100

0
0 20 40 60 80 100 120
minutes

Figure 64: Measured off-gas temperature by the thermocouple probe and calculated off-gas
temperature with equation (26)

heat 149996

oxygen carbon monoxide power on

30 1200
temperature laser calc. temperature

18 seconds
25 1000

20 800
[Vol.-%]

[°C]

15 600

10 400

5 200

0 0
0 5 10 15 20 25 30 35
minutes

Figure 65: Calculated off-gas temperature and measured temperature from the LINDARC system

51
2.2.4.2 Adaptation of BFI dynamic mass and energy balance model to the conditions of the
Marienhütte EAF (T 4.2)
The BFI dynamic process model for observation of the current EAF process state is based on coupled
mass and energy balance calculations. It is installed under the block-oriented simulation software
Matlab / Simulink for dynamic process simulation. The dynamic process model has been applied for the
DC EAF of Georgsmarienhütte, Germany, within one ECSC [16] and one RFCS [17] project. For the
purpose of the OFFGAS project, its structure was adapted to the configuration of the MH AC furnace
and the input data which are available at this furnace. Figure 66 shows the structure of the process
model with the main function blocks and the most important input values which are available at the MH
AC-EAF.

Metallic material input via basket Weight of

charge materials
Slag former additions Meltdown energy
requirement

Hot heel (amount and temperature)

Electrical energy
Burner natural gas
Energy
Lance and KT injector oxygen
input
Post-combustion oxygen

Off-gas analysis ( CO, O2 ) Actual Melt

Post-combustion
energy content temperature

Radiation and convection

Off-gas temperature Energy
Cooling water flow rates and temperatures losses

Figure 66: Structure of the dynamic process model with input data available at the MH AC-Electric
Arc Furnace
The block ‘weight of charge materials and meltdown energy requirement’ calculates within a mass
balance the total weight (steel and slag) of the melt from the metallic materials (distinction of only one
scrap type / scrap mix and charging of skull) charged with up to three scrap baskets per heat, and slag
former additions (only lime). From the specific meltdown energy of each material, the totally required
energy is calculated. Additionally the amount of the hot heel in the furnace has to be considered in this
calculation.
The ‘energy input’ comprises on the one hand the electrical energy input, and on the other hand the
chemical energy input. The latter one consists of the energy input by burner gas (one door burner), by
oxygen input through a door lance and three KT shell injectors, and by the injection of post-combustion
oxygen. The ‘post-combustion’ is calculated from the off-gas analysis, which is measured by the
permanently installed LINDARC system (only CO and O2), and for a trial period also by the RWTH
Aachen off-gas measurement system.
The ‘energy losses’ consist of losses by off-gas and water cooling and overall radiation losses. In
calculation of the off-gas losses, the off-gas temperature to determine the sensible heat of the off-gas,
and the CO content to calculate the chemical energy content is considered. The thermal losses by the
water cooling circuits (roof and wall panels) are calculated from cooling water flow rates and the
difference between inlet and outlet cooling water temperatures. Radiation losses are taken into account
by model parameters for the different process phases like charging, meltdown, refining etc.
The difference between energy input and energy losses gives the ‘Actual energy content’ of the total
charge, consisting of molten steel and slag. Within this balance, again the hot heel with its energy

52
content has to be considered. The actual energy content is related to the meltdown energy requirement,
which is defined for a reference temperature of e.g. 1600 °C. Thus the current charge temperature is
calculated, assuming that molten steel and slag have the same temperature.
For a first validation of the BFI dynamic process model, the cyclic and acyclic process data of trial
heats performed during the measurement campaign with the RWTH Aachen off-gas measurement
system in September 2006 were provided by MH. The data were reasonable and consistent and thus
allowed to follow continuously the energetic behaviour of the EAF during treatment. From these data,
corresponding input data files for simulation with the dynamic process model under Matlab / Simulink
were generated.
The input data and the simulation results of the dynamic process model are shown for one example heat
in Figure 67 to Figure 70.

Figure 67: Electrical energy input, KT burner and oxygen injection rates of a MH example heat

53
In the upper part of Figure 67, data of the electrical energy input as active power and accumulated
energy input are shown. The break in the power input after about 15 minutes indicates the period of
charging of the second basket. This is also shown by a binary charging signal, and also the tapping
period is indicated by a binary signal.
The middle part of the figure shows natural gas and oxygen flow rates of the door burner and the three
KT injectors. The KT injectors can be operated in a burner mode, or in an oxygen injection mode,
where natural gas and oxygen are used as shrouding gas to focus the oxygen jet. In this part of the
figure, the natural gas and the oxygen flow rates for the KT injectors in burner mode resp. for shrouding
of the oxygen jet are shown.
In the lower part of the figure the oxygen flow rates of the three KT injectors in injection mode and of
the door lance are displayed. Furthermore oxygen is injected via four post combustion injectors which
are operated directly after charging of each scrap basket. The lower part of the figure also includes the
total carbon injection via the three KT injectors and the door lance during the slag forming period.
Figure 68 shows inlet and outlet temperatures and the flow rates of the complete water cooling system,
which is divided into five circuits.
The first cooling circuit covers the furnace roof, the second the furnace wall (both in the upper part of
the figure), the third and fourth circuit the off-gas elbow and the post-combustion chamber (both in the
middle part of the figure), and the fifth one the suction channel (lower part of the figure). In the lower
part additionally the off-gas temperature measured in the suction channel is shown.
Figure 69 shows three diagrams with off-gas analysis results of the LINDARC and the RWTH Aachen
off-gas measurement system. The measurement location is approximately the same for both systems
and is described in detail under task 4.1.
In the upper part of the figure the off-gas analysis (O2 and CO concentration) and the off-gas
temperature of the LINDARC measurement system are displayed.
The middle part of the figure shows the off-gas concentrations of CO, CO2, CH4, H2, O2 of the RWTH
Aachen off-gas measurement system. Additionally a sound pressure signal is recorded for
synchronisation of the off-gas analysis data with the process data provided by the MH data acquisition
system.
The third diagram shows a comparison of the off-gas concentrations of CO and O2 of the two different
measurement systems. The correspondence between the different measurement systems is obvious for
this heat. A detailed comparison of the two off-gas analysis systems is described under task 4.1.
The result of the dynamic model calculations is shown in Figure 70. In the upper part of the figure, the
energy inputs by electrical energy, burner gas, lance and KT injector oxygen as well as post-combustion
oxygen are displayed as power versus time.
In the middle part of the figure, the calculated power losses are displayed. The losses via the off-gas
consist of the sensible heat of the off-gas, and the chemical energy of the not completely burned CO.
For the off-gas flow rate, which is not measured at the MH furnace, in a first approach a constant value
was assumed. Radiation losses are calculated with a power loss rate which is different for the meltdown
and the refining period. Furthermore a smaller loss rate is assumed for power-off times, which is
increased when the furnace roof is open. The cooling water energy losses from the wall and roof
cooling circuits are determined from the water flow rates and the difference between outlet and inlet
water temperature.
The lower part of the figure shows the calculated temperature of the melt as result of the energy
balance, i.e. the difference of energy inputs and losses related to the reference energy content. The times
when the two baskets are charged are clearly indicated by the temperature drop. The simulated steel
melt temperature is compared to five temperature measurements performed before tapping. It can be
seen that the simulated temperature matches most of the temperature measurements relatively good.

54
Figure 68: Cooling water flow rates and cooling water temperatures of a MH example heat

55
Figure 69: Off-gas measurement data by LINDARC and RWTH Aachen measurement systems of a
MH example heat

56
Figure 70: Energy input and loss rates as well as melt temperature evolution calculated for a MH
example heat
The results of the IR/VIS off-gas measurement campaign performed by RWTH at the MH furnace were
evaluated at BFI to calculate values for the off-gas flow rate and the H2 content of the off-gas, which
can both not be measured with the permanently installed laser-based system.

57
Off-gas measurements were performed by RWTH at two points in the off-gas system of the MH
furnace. Measurement point A was at the same position as the laser-based sensor, directly behind the
gap after the elbow. Measurement point B was about 50 meters further downstream in the off-gas
system, without any significant further leakage air ingress between measurement point A and B, besides
the one through the gap. Both measurement locations have been indicated in Figure 53.
In the Figure 71 and Figure 72 the measurement results from measurement point A and B are shown
for one example heat. From off-gas differential pressure p, off-gas temperature TB and off-gas
analysis, the off-gas flow rate under standard pressure and temperature conditions can be calculated.
The gas velocity vmax at measurement point B is calculated from the measured differential pressure p:

2  p
v max  (27)

The gas density  under operating conditions can be determined from the off-gas temperature TB and
the density under STP conditions N, which can be derived from the off-gas composition.
273,15
  N  (28)
273,15  TB
The off-gas flow rate under operating conditions QB is given from the calculated off-gas velocity, the
cross sectional area of the off-gas duct A, and a correction factor of 0.83 for v/vmax which considers the
not ideally homogeneous flow of the off-gas at the measurement point B.

Q B  0.83  A v max  3600 (29)

The off-gas flow rate under STP conditions QNB is finally given from the flow rate under operating
conditions QB and the measured off-gas temperature TB.
273,15
Q NB  Q B  (30)
273,15  TB

Figure 71: Off-gas analysis (CO, CO2, CH4, H2 and O2 concentration) at measurement point A

58
Figure 72: Off-gas analysis (CO, CO2, and O2 concentration), off-gas temperature and differential
pressure at measurement point B
The input data for this calculation are shown together with their results in Figure 73 for the example
heat. According to the calculation, the off-gas flow rate QNB under STP conditions at measurement
point B is on the average around 100.000 m³ (STP)/h. From this the off-gas flow rate at point A under
STP conditions QNA, i.e. at the exit of the furnace directly behind the gap, can be calculated from an off-
gas carbon balance. Under the assumption that no ingress of CO or CO2 occurs between measurement
point A and B, a carbon balance between these two points can be set up:

Q NA   COA  CO2A   Q NB   COB  CO2B  (31)

Q N A  Q NB 
 CO B  CO 2B 
(32)
 COA  CO2A 

Figure 73: Measured off-gas data and calculated off-gas flow rates at measurement point A and B

59
The result of this carbon balance is also shown in Figure 73. At measurement point A, the flow rate has
a value of around 50.000 m³ (STP) with larger deviations between the meltdown and the refining phase.
This value seemed to be quite high for the relatively small MH EAF with a capacity of only 45 tons.
Thus it was checked by a carbon balance for the investigated example heat, as listed below:
Average off-gas flow rate STP at measurement point B 99.987 m³ (STP)/h

CO2 volume in off-gas at measurement point B 5308 m³ (STP)

CO2 by combustion of burner and KT injector natural gas - 249 m³ (STP)

CO2 by combustion of carbonaceous materials 5059 m³ (STP)

Carbon removal observed from off-gas measurement 2710 kg

Carbon addition with baskets - 131 kg

Lance carbon injection - 371 kg

Electrode consumption (assumed 2kg/t) - 78 kg

Remaining carbon in liquid steel at tapping (0.14 % in 38 t) - 53.2 kg

Carbon input by materials charged with the baskets (43.2 t) 2185 kg

Average Carbon content of charge materials 4.97 %

This carbon content of the charge materials is unreasonable high. It can be assumed that the scrap has
an average carbon content of about 0.3 %, together with organic components charged with the scrap a
carbon content of 0.6 % can be estimated. This would lead to a correction factor of about 0.36 for the
off-gas flow rate under STP conditions at measurement point B. This high correction factor might be
explainable by the location of the point B of measurement for the differential pressure in the off-gas
system, which seems not to be ideal for determining the off-gas flow rate. Furthermore the utilisation of
a pitot tube might have caused inaccuracies in determining the differential pressure.
From the off-gas flow rate under STP conditions at measurement point A the leakage air ingress at the
gap of the elbow can be determined. This is possible via a nitrogen balance, under the assumption that
the off-gas analysis at the measurement point A is taken inside the flame of the furnace off-gas, so that
it is not influenced by leakage air ingress QNL at the gap. Under this assumption the nitrogen balance is:
100   COA  CO2A  O2A  CH 4A  H 2A  79 100   COB  CO2B  O2B 
Q NA   Q NL   Q NB  (33)
100 100 100
The resulting leakage air flow rate is shown in Figure 74. For a check of this balance, the sum of QNA
and QNL is compared to QNB, showing a good correspondence for almost the whole treatment, with the
exception of the phases where the furnace roof is opened for scrap basket charging. The leakage air
ingress flow rate through the gap is during most of the melting time on the same level as the furnace
off-gas flow rate. However this is only true for STP conditions. Under operating conditions the furnace
off-gas flow rate is much higher, as its temperature is higher than the one of the leakage air, which can
be assumed to have approx. room temperature when entering the gap. This relationship can be used to
calculate the off-gas temperature at the measurement point A, using the assumption that under operating
conditions the total flow rate has to be the same at point A and B:
273  TA
Q B  Q A  Q L  Q NA   Q NL (34)
273
QB is known from the differential pressure measurement, QNA from the carbon balance and QNL from
the nitrogen balance. Thus the off-gas temperature at measurement point A can be calculated to:

60
 273   Q B  Q NL 
TA   273 (35)
Q NA
This calculated value for the off-gas temperature TA is plotted in Figure 74 together with the off-gas
flow rates. During meltdown of the first scrap basket values of about 1400 °C are reached, during
refining the temperature drops to a level of about 800 °C, which is a reasonable result. Thus the RWTH
off-gas measurement campaign at the MH furnace provided useful information on the process
behaviour and the off-gas conditions. The next step was to derive correlations between these results and
the laser-based off-gas measurement values which are acquired at the MH furnace continuously, and to
calculate H2 and CH4 content of the off-gas which can not be measured by the laser-based analysis
system, see under Task 4.3.
 

Figure 74: Calculated off-gas flow rates at measurement point A and B together with leakage air
ingress at the gap

Figure 75: CO and H2 content of off-gas at measurement point A together with CO/H2 ratio and
difference

61
To express a sufficient control strategy for the PC-injectors the information of the amount of additional
combustible gases like H2 and CH4 is required. For that examination, the IR/VIS measurement data
described under Task 4.1 have been used. Figure 75 shows a typical evolution of the off-gas H2 and CO
content at the measurement point A for the same example heat which was already evaluated under Task
4.2.
During scrap meltdown an almost constant ratio between the CO and the H2 off-gas content can be
found. During refining the H2 content is lowered to almost zero, whereas the CO content stays on a
level of about 3 %. In Figure 76 and Figure 77 this relationship of CO between H2 and CH4
respectively has been confirmed to exist. The correlation of CO and H2 from 20 heats can be seen in
Figure 76. A distinction between the melting and refining phase is readily identifiable.

Figure 76: Correlation of CO and H2 (IR/VIS) Figure 77: Correlation of CO and CH4 (IR/VIS)
Typical high hydrogen amounts occur during melt down due to several H2 sources like natural gas from
the burners, organic impurities of the scrap etc. Much less H2 is recorded during refining when the
burners are off and the scrap is already melted down. A possible source of evolved hydrogen is the
water coming from the spray rings to cool the electrodes. The best utilisation of Post Combustion
generated heat is achieved during melt down. Therefore the correlation of CO, H2 and CH4 during the
melt down of each basket has been evaluated separately. In Figure 78 and Figure 79 the average
concentrations of CO, H2 and CH4 respectively from each first basket of 20 heats are correlated
exemplified.

Figure 78: Correlation of CO and H2 (IR/VIS) of Figure 79: Correlation of CO and CH4 of the first
and CH4 (IR/VIS) of the first basket basket
The linear equations of the correlation lines from all three baskets of 20 heats are below-mentioned
(equation (36) and (37)):

g(H 2 )  0.853  CO  0.808 (36)

g(CH 4 )  0.126  CO  0.116 (37)

The dynamic model of BFI has been adapted to the results of the trials with the off-gas measurement
system from the RWTH Aachen. As described above, the values for the concentration of H2 and CH4,
which are not provided by the LINDarc system, can be calculated by linear equations, which depend on

62
the concentration of carbon monoxide. According to these results, the equations (36) and (37) are used
in the dynamic model to calculate the off-gas H2 and CH4 contents during scrap melt down, see Figure
80.

Figure 80: Calculated concentrations of H2 and CH4 according to equation (36) and (37)
During refining the contents of H2 and CH4 in the off-gas are set to zero. A further modification of the
input data for the dynamic model became necessary concerning the temperature and the flow rate of the
off-gas.
The lack of a permanent measurement system for the off-gas flow rate made an assumption of a
constant rate necessary. The analysis of the off-gas measurements performed by the RWTH Aachen has
shown that the previously assumed constant level for the flow rate of 380 m³ (STP)/h was too high.
According to the evaluation of this campaign shown in Figure 70, the value has been adapted to 300 m³
(STP)/h.

Figure 81: Off-gas temperature measured by LINDARC system and calculated from cooling water
temperature of laser lance according to formula (26)
According to the trial campaign at MH with the thermocouple lance provided by RWTH Aachen
described under Task 4.1, the BFI dynamic mass and energy model has been modified concerning the
consideration of the off-gas temperature. The comparison between the measured temperature from the
LINDARC system and the temperature from the thermocouple has shown an incongruity. However, as

63
described under Task 4.1, there is a very good correlation between the off-gas temperature measured by
the thermocouple and the cooling water temperature of the LINDARC laser lance. Figure 81 shows a
comparison between the off-gas temperature of the LINDARC system and the off-gas temperature
calculated according to formula (26) from the cooling water temperature of the LINDARC lance. The
evolution of the calculated temperature is much smoother, and the continuous temperature increase
during the melt down of each basket seems to be much more reasonable than the evolution measured by
the LINDARC system. Thus the off-gas temperature calculated from (26) was used in the dynamic
mass and energy balance model of BFI to determine the sensible heat losses of the off-gas.
2.2.4.3 Performance of simulation trials to study the effect of post-combustion oxygen on the energy
balance of the furnace (T 4.3)
If CO and H2 are present in the furnace gases, their oxidation releases more energy per unit oxygen than
burning injected carbon or natural gas. The theoretical limit for post-combustion of CO at bath
temperatures (1600 °C) is 5.8 kWh/m³ (STP) oxygen [15]. Jones [8] published results of different post-
combustion practices at several EAFs.
Energy savings between 1.2 and 5.04 kWh/m³ (STP) oxygen were obtained. The PC oxygen efficiency
assumed in the BFI energy model is 2.8 kWh/m³ (STP) [11]. Adams [1] expects a thermal efficiency of
the reaction between hot CO and cold oxygen of 50 % at the most, i.e. 3.1 kWh/m³ (STP) or less.
However, the reported PC efficiencies vary in a wide range. The conclusion that PC efficiency depends
highly on the several PC techniques (injectors, stoichiometry, lancing ...) can be drawn. The knowledge
of the PC oxygen efficiency for further evaluations regarding the energetic performance of the EAF is
of particular importance. Thus from October to December 2007 at the MH furnace two trial campaigns
with different PC oxygen flow rates were performed. To minimise the influence of different scrap
qualities, the trials were performed alternating between two different flow rate values.
Table 6: Input data for BFI statistical model of campaign 4 with 400 vs. 300 m³ (STP)/h PC oxygen per
injector
Variable WR WE GA gE gScrap gZ TA tS tN MG MO2G ML MLJ2 MN MCoke
3 3 3 3 3
Unit kWh/t kWh/t t kg/t kg/t kg/t °C min min m /t m /t m /t m /t m /t kg/t
Campaign 10.-18.10.07
125 heats all
370.6 365.9 34.5 1204 1191 37.6 1628 26.7 12.9 7.2 14.3 6.9 14.4 8.1 10.2

Campaign 10.-18.10.07
62 heats with PC O2 = 300
372.7 369.5 34.5 1208 1195 37.3 1627 27.0 12.9 7.2 14.4 7.0 14.5 7.4 9.7

Campaign 10.-18.10.07
63 heats with PC O2 = 400
368.5 362.2 34.6 1200 1187 37.8 1630 26.5 13.0 7.1 14.3 6.8 14.4 8.8 10.8

Difference between two

campaings
-7.3 0.1 -8.0 -8.0 0.5 3.0 -0.5 0.1 -0.1 -0.1 -0.2 -0.1 1.4 1.1

Difference in electrical
energy demand in kWh/t
-3.2 0.5 0.9 -0.5 0.1 0.8 0.9 0.4 -

Corrected difference in
electrical energy consump.
-7.2 5.1

This method ensured comparable conditions regarding the scrap input, because of similar scrap qualities
taken out from the wagons sequentially. The results of calculations with the statistical BFI model for the
first campaign with flow rates of 300 and 400 m³ (STP)/h post-combustion oxygen are shown in Table
6. The difference in electrical energy consumption between these two campaigns is 7.3 kWh/t.
However, for comparison of the total effective energy input, it is necessary to correct this value,
because the other relevant average input parameters of the two trial variations differ. According to the
equation of the statistical model, the difference in electrical energy demand caused by the different
input parameters is shown in row five of the table. The corrected difference in electrical energy
consumption is 7.2 kWh/t. Dividing this effective saving by the additional PC oxygen consumption of
1.4 m3/t results in an energetic efficiency of post-combustion oxygen of 5.1 kWh/m3. Similar results of
the statistical model calculations for the second campaign with PC oxygen flow rates of 200 and 400 m³
(STP)/h are shown in Table 7. In this case the electrical energy consumption was effectively reduced by
11.9 kWh/t, and the efficiency of the additional post-combustion oxygen is 4.1 kWh/m3. The result of
both trial campaigns shows that the energetic efficiency of PC-oxygen is significantly higher than
assumed with 2.8 kWh/m³ (STP) in the statistical BFI model.

64
Table 7: Input data for BFI statistical model of campaign 6 with 400 vs. 200 m³ (STP)/h PC oxygen per
injector
Variable WR WE GA gE gScrap gZ TA tS tN MG MO2G ML MLJ2 MN MCoke
3 3 3 3 3
Unit kWh/t kWh/t t kg/t kg/t kg/t °C min min m /t m /t m /t m /t m /t kg/t
Campaign 26.11.-06.12.07
195 heats all 365.4 394.7 34.7 1223 1213 37.4 1631 29.0 11.6 8.2 16.4 8.6 14.8 7.1 11.3

Campaign 26.11.-06.12.07
99 heats with PC O2 = 200
366.5 397.7 34.8 1221 1210 37.4 1631 29.3 11.6 8.3 16.6 8.6 15.2 5.7 11.5

Campaign 26.11.-06.12.07
96 heats with PC O2 = 400 364.2 391.6 34.7 1226 1215 37.5 1630 28.6 11.7 8.1 16.2 8.6 14.5 8.6 11.0
Difference between two
campaings
-6.1 -0.1 5.0 5.0 0.1 -1.0 -0.7 0.1 -0.2 -0.4 0.0 -0.7 2.9 -0.5
Difference in electrical
energy demand in kWh/t 2.0 0.1 -0.3 -0.7 0.1 1.6 0.0 3.0 -
Corrected difference in
electrical energy consump. -11.9 4.1

2.2.4.4 Development of dynamic control strategies for post-combustion oxygen supply on the basis of
real-time laser-based off-gas measurement and model-based calculation of further off-gas
data (T 4.4)
The Equation (36) and (37) describe the correlation between H2 resp. CH4 content in the off-gas and the
CO content measured by the laser-based system at MH can be used to calculate the ideal CO/O2 ratio
according to equation (38) in order to ensure a stoichiometric combustion.

nCO  g(H 2 )  g(CH 4 )  mO 2  nCO 2  xH 2 O (38)

The calculated criteria (equation (39)) for the dynamic control of the PC-injectors contains all
coherences of the combustible gases CO, H2 and CH4 and negates the need for additional H2 and CH4
analyses.
CO 0.458
  0.848 (39)
O 2 CO[%]  0.54
The line shape function is shown in Figure 82. The PC-Injectors have to be switched on if the CO/O2
ratio measured with the laser instrument exceeds the minimum ratio needed for stoichiometric
combustion as formulated with the criteria (equation (39)) above. Otherwise if the criterion is not
fulfilled and the ratio is lower there is no need to add additional oxygen by the injectors.

Figure 82: Line shape function of the post combustion criteria

65
At the MH furnace trials with a dynamic control of the PC-injectors have been performed. Due to
technical reasons related to the control system, in a first step the factor 0.75 instead of the function (39)
described above was used approximately for the CO/O2 criteria. Figure 83 shows a typical off-gas
profile of a heat with three baskets when the dynamic control is working. The CO/O2 ratio related to the
off-gas profile in Figure 83 is diagrammed in Figure 84 as well as the operation modes of the PC-
injectors, burners, Power On, carbon and oxygen lance.

Figure 83: Off-gas profile

Figure 84: Operation mode and CO/O2 ratio

The PC-Injectors are switched on at the beginning of the melt down of each basket. The four PC-lances
keep running 60 seconds with a flow rate of 300 m³ (STP)/h oxygen each. The duration of one minute is
fixed because of the typical strong flame evolution. After the first minute the CO/O2 criteria of 0.75 is
the constraint for the PC-Injectors. Is the ratio below 0.75 the injectors are switched off, otherwise they
keep switched on for one further minute. In Figure 84 the CO/O2 ratio increases after seven minutes
steadily. By exceeding a ratio of 0.75 the injectors start for at least one minute. At runtime the ratio
decreases continuously. After one minute running period the ratio is below 0.75 and the PC-injectors
are switched off. The duration of the validity of the PC-control is only during melt down. After starting
the lancing mode or during the use of the oxygen lance the post combustion system is switched off
using air to keep the nozzles free respectively. Bag house explosions: To counter the danger of bag

66
house explosions due to too high carbon monoxide amounts in the off-gas an important action is
implemented.
If the CO/O2 ratio exceeds 4.0 the moveable sleeve moves back immediately whereby the gap after the
elbow increases allowing a higher amount of false air coming into the fourth hole duct. It can be seen in
Figure 84 after 35 minutes where carbon is injected to build up the foamy slag. The ratio rises
suddenly. By exceeding 4.0 the sleeve moves backwards and the ratio drops down. Examination of the
energetic behaviour of the EAF using dynamic PC-control: The goal at that point of the project was to
increase the efficiency of the oxygen injection for post combustion without negative consequences
related to the electrical energy behaviour: (a) when is oxygen required for post combustion during the
progress of the heat? (b) how much oxygen is when required during the progress of the heat? By the
completion of the first goal a better electrical performance of the EAF shouldn’t be expected. However,
the oxygen efficiency should be increased.
0.6

0.5

CO/O2
0.4

0.3

0.2

0.1

0
0 50 100 150 200 250 300 350
PC [Nm 3 ]

Figure 85: Correlation of CO/O2 and post combustion oxygen

Figure 86: Results of alternating trial with 160 heats

67
In Figure 85 the relation between the PC-oxygen consumption and the average CO/O2 ratio per heat is
shown. Higher ratios result in higher oxygen consumption because the PC-injectors are operating for a
longer time. The dynamic control of the PC-injectors depending on the CO/O2 ratios leads to highly
varying oxygen consumption of the different heats. To evaluate the influence of these varying
conditions field trials over 160 heats have been performed. To erase the influence of different scrap
qualities the trial has been progressed alternating. That means one heat was carried out without dynamic
control (automatic OFF) the following heat with dynamic control of the PC-system (automatic ON).
During the automatic OFF mode the injectors are operating about five minutes at the beginning of the
melt down of each basket. The alternating trials ensured comparable conditions because of similar scrap
qualities taken out from the wagons successively. Figure 86 shows the result of the performance from
80 heats operated with the automatic ON mode compared to 80 heats operated with the OFF mode.
There was no significant difference in the electric energy consumption or in the power on time
noticeable whereas the oxygen consumption for post combustion decreased in the automatic ON mode
dramatically from 6 to 3 m³ (STP)/tscrap.

Figure 87: Electrical energy, power on time and tap to tap time against post combustion oxygen

Figure 88: Correlation of CO/O2 ratio and offgas temperature at measurement point K3
In Figure 87 the daily averages of the electrical energy [kWh/tgb], the tap to tap time and the power on
time [min] of 112 production days equal to 3739 heats are correlated to the PC-oxygen consumption
[m³ (STP)/tgb]. The distribution of the PC-oxygen consumption in the automatic ON mode is between 2
and 9 m³ (STP)/tgb. For comparison only the range of the automatic OFF mode is also identified in the

68
diagram. All heats have been centred in that area between 6.5 and 8.5 m³ (STP)/tgb before the PC-
control system was implemented. With the automatic ON mode the average PC-oxygen consumption
spread out resulting in a shift of the average consumption to lower values. There was no significant
influence neither to the electric energy consumption nor to the power on and tap to tap time noticeable.
Consequently the electric energy input seems to be independent of the oxygen input for post
combustion. But high PC-oxygen consumption due to a high CO/O2 ratio (Figure 85) corresponds to
bad scrap qualities. A large amount of impurities or in other words a scrap quality with bad melting
behaviour leads to an insufficient efficiency of the burners and the electrodes resulting in high off-gas
temperatures. Therefore much more oxygen is required to compensate that loss. This is demonstrated by
Figure 88, where the off-gas temperature at a measurement point K3, which is 35 m downstream in the
hot gas duct, is correlated to the CO/O2 ratio. The results of the first campaigns with the above
described control of the PC oxygen were also evaluated with the BFI statistical model for the electrical
energy consumption. They were compared with previously evaluated consumption figures which were
acquired during the RWTH off-gas measurement campaign in September 2006.
Regarding PC oxygen control, the following data acquisition campaigns were evaluated with the BFI
model: (a) 1st campaign (15.09.2006 to 19.09.2006), 136 heats (Standard PC oxygen control via fixed
patterns); (b) 2nd campaign (20.02.2007 to 24.02.2007), 153 heats (Automatic PC oxygen control); (c)
3rd campaign (20.03.2007 to 24.03.2007), 164 heats (Trials with alternate Auto / Standard PC oxygen
control).
Table 8: Input data for BFI statistical model of different campaigns at the MH furnace
Variable WR WE GA gE gScrap gZ TA tS tN MG MO2G ML MLJ1 MLJ2 MLJ3 MN MCoke
Unit kWh/t kWh/t t kg/t kg/t kg/t °C min min m3/t m3/t m3/t m3/t m3/t m3/t m3/t kg/t
Campaign 15.-20.9.06
351.6 378.1 36.1 1151 1140 36.2 1631 28.4 13.3 6.7 13.4 6.8 16.3 6.4 9.0
135 heats
Campaign 20.-24.2.07
357.7 388.5 35.6 1184 1171 35.9 1633 29.0 12.2 7.0 13.9 8.7 16.3 5.2 10.2
153 heats
Campaign 20.-24.3.07
362.5 399.4 35.5 1193 1182 37.7 1621 29.6 12.1 6.8 13.7 7.3 16.8 6.0 11.2
164 heats all
Campaign 20.-24.3.07
357.9 399.3 35.4 1192 1180 38.7 1618 29.4 11.7 6.8 13.6 7.2 16.5 8.1 11.2
83 heats Auto = 0
Campaign 20.-24.3.07
367.3 399.4 35.7 1195 1183 36.7 1624 29.9 12.5 6.9 13.8 7.4 17.1 3.9 11.2
81 heats Auto = 1

500
March 2007 March 2007
Auto = 1 Auto = 0
calculated electrical energy demand in kWh/t .

Mean deviation: -32.1 kWh/t -39.7 kWh/t

Standard deviation: 22.0 kWh/t 25.9 kWh/t
450

400

350

300

250
250 300 350 400 450 500
actual electrical energy consumption in kWh/t

Figure 89: Calculated vs. actual electrical energy consumption for MH heats of campaign March 2007
with alternating PC oxygen control mode

69
The average input data of these data acquisition campaigns as used for the BFI statistical model for the
electrical energy demand are shown in Table 8. The comparison of the campaigns from September
2006 and February / March 2007 shows that the electrical energy consumption was higher in the second
period, and the iron yield lower with automatic control of PC oxygen input, although an improvement
in both energy consumption and yield would be expected. One reason is that in the winter months the
energy consumption is generally higher, on the other hand the influence of the varying scrap quality
overrules all other effects. This was the reason to set up the campaign with alternating PC control
strategy, performing one heat under automatic control and the next heat with standard patterns.
Electrical energy consumption, yield and also oxygen consumption via door lance and KT injectors are
nearly equal in both cases.
The significant difference is that the PC oxygen consumption is in the automatic control mode by 4.2
m³ (STP)/t liquid lower than in the standard mode, leading to the conclusion that in automatic mode the
energetic efficiency of the PC oxygen is higher. In Figure 89 the expected electrical energy demand is
plotted against the actual consumption for both cases of PC control mode for the heats of the campaign
in March 2007. The difference between calculated electrical energy demand and actual consumption,
which is a measure for the energetic performance of the furnace, is by about 8 kWh/t lower for the heats
with automatic PC oxygen control. This amount can be considered as net energy saving due to the
control of PC oxygen based on the CO/O2 ratio of the laser-based off-gas measurement.
2.2.4.5 Test and optimisation of control strategies with the help of the BFI simulation model (T 4.5)
The trial campaign of MH with different strategies of PC oxygen control was also evaluated with the
dynamic BFI model. The injectors were set to a fixed flow rate of oxygen. In automatic control mode
they were switched on or off against a certain CO/O2-ratio, without automatic control they were
switched on for a certain time after charging of each basket. In the following figures the simulation
results of one example heat for each control mode are shown.
Without automatic control With automatic control

Figure 90: Oxygen injection rates for example heats without and with automatic PC control
In Figure 90 the oxygen and carbon injection regime for the two control modes is shown. The
difference between these two heats is mainly the dynamic control of post-combusting oxygen. Without
the automatic control, see the left hand side figure, the injectors are switched on after charging of each
basket for a fixed time. This means that more or less always the same amount of post-combustion
oxygen is applied. In the automatic mode on the right hand side, the PC oxygen is switched off much
earlier, resulting in a lower consumption and a more effective use of PC oxygen.
In the upper part of Figure 91, the measured data from the LINDarc system are shown. After charging a
basket the concentration of CO rises immediately before it decreases slowly during the melting process.
While the furnace roof is open to fill in new scrap, the oxygen rises to maximum and the carbon
monoxide becomes nearly zero, because of the air streaming into the system. At the beginning of the
carbon injection the concentration of CO rises again to even higher levels than during scrap melting.
The middle part of Figure 91 shows the ratio between CO and O2 and its mean values during the
melting of each basket and the carbon injection phase. At the beginning of each melt down phase the
ratio rises very fast to slope down during the further meltdown process. During carbon injection the
CO/O2 ratio is also very high due to the high amount of produced CO. In the same figure the differential
furnace pressure is shown. It is obviously a measure for the produced amount of combustion gases in
the furnace, as it runs almost in parallel to the CO/O2 ratio.

70
 Without automatic control With automatic control

Figure 91: Measured LINDarc off-gas values, CO/O2 ratios and furnace pressure, and calculated
concentrations of H2 and CH4 for example heats without and with automatic PC control
Without automatic control With automatic control

Figure 92: Energy input and loss rates and melt temperature evolution calculated with the dynamic
energy balance model for example heats without and with automatic PC control

71
The diagrams at the bottom of Figure 91 display the calculated concentrations of H2 and CH4. The
equations to calculate these values were derived from the results of the campaign with the off-gas
measurement system from the RWTH Aachen, as described under Task 4.3.
The results of the calculations with the dynamic energy balance model are shown in Figure 92. The
upper part shows the power inputs, the middle part the power losses. The calculated melt temperature
resulting from the balance between inputs and losses is for both heats close to the measured values.
2.2.4.6 Implementation and test of dynamic control of post-combustion oxygen on the basis of
measured off-gas values at the Marienhütte furnace (T 4.6)
The dynamic PC-control described in Task 4.4 is only a simple version of a closed loop control. The
injectors were set to a fixed flow rate and switched on or off depending on a certain CO/O2-ratio. This
simple control already resulted in an increased PC-oxygen efficiency. The next step of development
aimed at controlling not only the timing but also the flow rate of oxygen input depending on the off-gas
composition. Thus it was required to vary the oxygen flow rate of the PC-injectors continuously,
adjusting it to the CO and O2 profile of the off-gas. The oxygen flow rate of each injector is limited
between m1= 200 and m2= 400 m³ (STP)/h. The range of the CO/O2-ratio is between n1= 0.40 and n2=
0.75. In that range the flow rate is adjusted linear proportional according to equation (40). If the CO/O2-
ratio exceeds 0.75 the flow rate remains at 400 m³ (STP)/h. If the ratio is below 0.40 the injectors are
switched off. This control regime is shown in Figure 93. For any trials and improvements the
boundaries m1, m2, n1 and n2 can be adapted easily.
n
 Nm³  m 2  m1  CO 
2
  m 2  m1  
Q       m  n    (40)
 h  n 2  n1  O 2  n1 
1 1
 n 2  n1  

Figure 93: Linear proportional PC-control strategy

A large number of heats were performed with the settings described above. In Figure 94 the calculated
oxygen flow rate according to equation (40) and the actual flow rate of the PC-injector 1 are shown for
an example heat with a three basket charging sequence. The operation mode of KT-burners, oxygen-
lance, PC-system, carbon injection and power-on is also figured against the progress of the heat. In the
beginning the PC-injector flow rate is set to its maximum of 400 m³ (STP)/h, as described in the
previous dynamic PC-control. After a lapse of time the flow rate is determined according to equation
(40). The minimum operating time of the PC-injectors is set to 30 seconds to protect the valves against
mechanical wear. For example, if the CO/O2-ratio goes below n1, the flow rate will retain at the value
m1 for at least 30 seconds. The deviation between calculated and actual flow rate can be explained by
the attenuation factor of the regulation controller. Figure 95 shows the total oxygen and natural gas
input for the three basket sequence heat, the off-gas profile is plotted in Figure 96. Post-combustion
oxygen is injected only if required due to high levels of CO in the off-gas. Due to the dynamic control
depending on the evolved CO, the PC oxygen efficiency was increased drastically. A reduction of up to
40 % PC-oxygen consumption was obtained.

72
Figure 94: O2 flow rate of PC-injector 1 with dynamic control

Figure 95: Total oxygen and natural gas

Figure 96: CO and O2 profile

73
Figure 97 shows a typical evolution of the dynamically controlled PC oxygen flow rate for one of the 4
PC injectors.

30 O2 [%] CO [%] 500

power-on door lance
criteria [Nm3/h] PC-Injector [Nm3/h]

25
400

oxygen post-combustion PC1 [Nm /h]

3
20
300
O2, CO [%]

15

200

10

100
5

0 0
0 5 10 15 20 25 30
minutes

Figure 97: Dynamically controlled PC oxygen flow rate together with off-gas analysis values for an
example heat
Further trial heats with this dynamic PC control were performed at the furnace of Marienhütte, which
were evaluated with the dynamic BFI model. In automatic dynamic control mode the flow rate of the
PC-injectors is calculated from the CO/O2 ratio in the off-gas. Below a fixed level of CO/O2 the PC-
injectors are switched off. This strategy guarantees an optimal use of the post-combustion oxygen. In
the following figures the simulation results of one example heat with automatic dynamic control and in
comparison to one heat without automatic control are shown.

Figure 98: Natural gas and oxygen input for the two example heats
The upper part of Figure 98 shows the gas and oxygen flow rates of the three KT burners and the door
burner. In the lower part of this figure, the oxygen and carbon injection regime for these two example
heats is shown. In the right hand side diagram the variation of the flow rate of PC oxygen and the

74
different operation time of the PC-injectors is quite perceptible. In comparison to this the flow rate of
PC oxygen is constant for the heat without automatic control. In the upper part of Figure 99 the values
from the LINDARC measurement system are shown. Although the analysed concentration of carbon
monoxide is quite low, the ratio to the O2 content was reasonable enough to be used as input for the
dynamic PC control. In the lower part of this figure the measured off-gas temperature from the
LINDARC system and the off-gas temperature calculated from the temperature of the cooling water
from the laser lance are shown. After charging a new basket the level of the calculated off-gas
temperature is increasing continuously.

Figure 99: LINDARC off-gas values and a comparison between the measured LINDARC off-gas
temperature with the calculated from the cooling water of the LINDARC laser lance

Figure 100: Energy input and loss rates for example heats

75
The upper part of Figure 100 shows the power inputs for the two example heats, the lower part figures
the corresponding power losses. The results of the calculations with the dynamic energy balance model
are shown in Figure 101. For both heats the calculated melt temperatures are close to the measured
values.

Figure 101: Melt temperature evolution calculated with the dynamic energy balance model
2.2.4.7 Evaluation of achieved energy savings (T 4.7)
After the implementation of the on-line dynamic process model of BFI at Marienhütte (see under task
8.3), the LINDARC off-gas measurement system was out of order for some month. Collected
consumption data from this time period were used for comparison purpose with data after reinstallation
of the LINDARC system. Table 9 shows the results of this evaluation. With dynamic PC control, the
average electrical energy consumption decreased by almost 2.8 kWh/t. The consumption of PC-oxygen
(+1.5 m³ (STP)/t) and injected/lance oxygen (+1.3 m³ (STP)/t) has been increased. The most significant
difference between the two time periods is the yield of the scrap. The scrap quality became worse,
which caused a higher energy demand of about 20 kWh/t. Taking all changes of consumption and
operating data into account, the savings in specific electrical energy input are about 15 kWh/t. At least
parts of these savings are due to the dynamic control, which increased the input of PC-oxygen because
of a higher CO-content in the off-gas caused by worse scrap quality.
Table 9: Comparison between two time periods with and without dynamic PC-control
Variable WR WE GA gE gScrap gZ TA tS tN MG MO2G ML MLJ2 MN MCoke
Unit kWh/t kWh/t t kg/t kg/t kg/t °C min min m3/t m3/t m3/t m3/t m3/t kg/t

Dynamic PC-control Off 352.0 352.0 39.5 1126 1115 35.4 1631 29.4 10.5 6.9 16.1 9.2 9.3 6.0 8.3

Dynamic PC-control On 363.7 349.2 41.1 1177 1167 34.1 1636 29.9 13.0 6.9 16.8 9.3 10.5 7.5 7.4

Difference between two

campaings
-2.8 1.6 51 52 -1.3 5 0.5 2.5 0.0 0.7 0.1 1.2 1.5 -0.9

Difference in electrical
energy demand in kWh/t
12.0 20 -1.3 2 0.5 2.5 0.3 -2.0 -0.6 -5.1 -4.3 -

Corrected difference in
electrical energy consump.
-14.8

2.2.5 Development and commissioning of continuous feeding EAF process dynamic control model
based on off-gas analysis (WP 5)
2.2.5.1 Development of a long term data collecting and remote monitoring system to control the
reliability of the off-gas measurement system (T 5.1)
In order to prove and control the EFSOPTM system reliability is necessary to ensure a network
communication between it and the plant. The task scope is to implement a remote data historical system
to evaluate and prove the reliable off-gas chemistry measurement.
The measured data (gas composition by EFSOPTM system, EAF, Consteel® process and fume system
data) are continuously stored in a database. A Supervisor Control And Data Acquisition (SCADA)
system has been appositely implemented to collect and store in a historical file chemical measurement
and plant data. Moreover, automatic e-mailing system every night sends all the historical files and the

76
alarms signal to a remote server. This permits a fast and accurate remote troubleshooting both from the
furnace pulpit and from a remote office.
Figure 102 shows the EFSOPTM Human-Machine Interface (HMI), which displays the status of the
EFSOPTM system, furnace off-gas trends and other relevant process information. The reliability for this
new EFSOPTM prototype system, designed and customized for Ori Martin Consteel®, has been
measured splitting the analysis in two parts:
 analyzer reliability
 sampling line reliability

Figure 102: ORI Martin - EFSOPTM HMI (Human-Machine Interface)

1- Analyzer reliability
The possible analyzer damages have been identified and analyzed:
(1) sample leak (CO, CO2, H2, O2). The sampled off-gas are loose in the analyzer cabinet;
(2) vent system plugging: the discharge system is plug;
(3) general mechanicals damage;
(4) PLC - Analyzer communication blackout. Absence of PLC communication with SCADA
system;
(5) power blackout and electrical fault.
The above mentioned possible damages have been monitored using the historical data stored. The
SCADA database gathers all the diagnostic signals coming from the EFSOPTM analyzer PLC. No
deviation has been recorded - from the critic signal value reference – in the observance period.
I.e. in the analyzer cabinet is installed a CO monitor (safety reason). The CO ppm concentration in the
analyzer cabinet is monitored. In case of off-gas leak inside the analyzer (first possible damage
mentioned in the list) a deviation might occur and an alarm would be sent immediately in the EAF
pulpit and to the remote server.
The analyzer has been completely reliable (100%); in fact none of the above mentioned damages ever
happen.

77
2- Sampling system reliability
The off-gas sucked from the furnace is carried to the analyzer through a sampling line composed by
cooling probes, flexible tubes and hard tubes. The links between components are made by fitting
connections and welding.
It is possible to check the reliability of the system looking for the damage that causes sampling signal
interruption. Two main causes of damages have been individuated:
1- sample gas watering by air leak: this happens when the line pressure is too negative, and air is
sucked from the connections. As a consequence the sample gas is mixed with air and high
oxygen concentration is revealed and signal intensity is low.
2- sample gas line break (with obvious sample lose): in this case the sucking system works
properly but there is a break (or a leak) in the sampling line. As a consequence atmosphere
concentration oxygen (20% about) is measured.
Some general rules have been developed for a fast check of sampling line reliability:
 oxygen should be less than 18%
 the extractive line negative pressure is within standard limits
 the sample gas flow rate is above the standard minimum value
 the chemistry composition is considered valid by the EFSOPTM internal control algorithms that
detect air in-leakage, low concentration, calibration errors, etc..
All of the above mentioned conditions have to be simultaneously satisfied in order to consider a
measurement point as “good chemistry”. With “good chemistry” it means a measure of reasonable
values without the occurrence of any fault. To identify the “good chemistry” have been applied on the
data some filters based on the concepts just shown. A chemistry measurement is considered to be
“good” once: (a) it shows less than 18% Oxygen (the extractive line is not broken or damaged); (b) the
extractive line negative pressure is within standard limits (The line is not plugged); (c) the sample gas
flow rate is above the standard minimum value; (d) the chemistry is considered valid by the EFSOPTM
internal control algorithms that detect air in-leakage, low concentration, calibration errors, etc.. In this
case the reliability of the system has been calculated as the ratio between the time of “good chemistry”
during power on and the total power on time. Three months testing of the system has been carried out.
The 95.41 % of the Power On time the analyzer and the system have worked reliably given good
measurements of the off-gas composition.
2.2.5.2 Development of a new model to establish the distribution of the carbon between bath slag and
free board (T 5.2)
To improve EAF operations it is necessary a dynamic control system that allows to reduce the energy
and material demand (and therefore the specific CO2 emissions) without affecting EAF productivity.
EAF dynamic model has been developed in order to reproduce the EAF cycle of the ORI Martin
Consteel®. In what follows terminology showed below will be used:
 Freeboard is the EAF gas phase
 Bath is the system constituted by steel and slag
Figure 103 reports the scheme of the developed model; the physical system is schematized as the
joining of freeboard (the zone of the furnace above the bath) and bath. Freeboard sub-set modeling has
been developed by Tenova with the main goal of optimizing the post-combustion in EAF and provide
net power and oxygen to the bath model CSM that can be considered as the second sub-set to complete
the dynamic description of process during the heat.

78
 EFSOP
Free Board Model

O2 injection, Electric
O2 for C, Power, Water Cooling
Fe, Mn, Si Net Power
oxidation
Charging
Charging--Injections
Injections
(plant
(plantdata)
data)

CaO, FeO, SiO2, …, Temp

Scrap, Fluxes,
Carbon Sources, …
C, Mn, Si, Temp

Composition
CompositionandandTemperature
Temperatureevolution
evolution
ofofthe
themetal-slag
metal-slagsystem
system

Figure 103: General scheme of EAF model

Bath/slag model
The bath/slag dynamic model simulates the chemical reactions inside the liquid steel and the slag. The
model uses the following input data:
 C lump feeding rate
 C powders feeding rate
 Scrap feeding rate
 Pig iron feeding rate
 Fluxes feeding rate
 Available oxygen for the melt oxidation (by Freeboard)
 Available power for heating up the melt (by Freeboard)
For calculating dynamically bath composition the following chemical reactions are considered:
Direct oxidation with oxygen
Fe+OFeO
C+OCO
Si+2OSiO2
Mn+OMnO
FeO reduction
C+FeOCO +Fe
Si+2FeOSiO2 + 2Fe
Mn+FeOMnO +Fe
The following kinetic expressions match the above reactions:
Direct oxidation with oxygen
C+OCO dCO/dt = k1CO·CC
Mn+OMnO dMnO/dt = k1MnO·CMn
Si+2OSiO2 dSiO2/dt = k1SiO2 CSi
Fe+OFeO dFeO/dt = k1FeO·O
FeO reduction
C+FeOCO +Fe dCO/dt = k2FeOCFeO CC
Si+2FeOSiO2 + 2Fe dSiO2/dt = k2SiO2·CFeO CSi

79
Mn+FeOMnO + Fe dMnO/dt = k2MnO·CFeO CMn
All the parameters k1X and k2X are the kinetic constants of the equations and require model tuning with
experimental data. Cx indicates concentration in the steel or slag phase. The underline means the specie
is considered to be a free element.
In the zone 1 (see Figure 103) when solid material (scrap, pig iron) are charged into the furnace, a
melting law, based on thermal conductivity of the material is considered [2]:

M 1
d ln   K ther  K area   Tsteel  Tmat  (41)
dt  Fe  cp  Tsteel  Tmat 

where
M: mass of solid material
CP: specific heat of the solid material
Tsteel: temperature of liquid steel
Tmat: temperature of solid material entering into the steel
Kther: coefficient of heat transfer
Karea: coefficient proportional to the area of the material, and varies with material size
Fe : heat of fusion of iron
For lump coal the following law of dissolution is considered:

dC
k (42)
dt
k is a constant which varies with coal lump size
Software implementation of the model has been also carried out. Following Figure 104 reports an
example of input and output window of the model running as a standalone simulator (Matlab
environment during development).

Input window Output window

Figure 104: Graphical interfaces of the software for model input and output as standalone application
Tenova freeboard model structure
Initial work towards a dynamic control system for the EAF-ConsteelTM has focused on building a
mathematical model for calculating a dynamic mass and energy balance applied to the gas-phase of the
EAF. The idea of calculating a mass/energy balance of the gas-phase of the EAF is not new. What
differentiates the current approach from that of others is that here the balance is calculated dynamically
using actual measurements of off-gas composition and temperature.
The freeboard model is illustrated in the Figure 105. Given measured composition, temperature and
static pressure of the furnace off gas; as well as the rate of injection of oxygen, (methane if burners will

80
be added) and carbon the model calculates a mass and energy balance of the freeboard based on
elemental and enthalpy equations.
Specifically the model calculates:
 The rate of water entering the gas phase from a hydrogen balance
 The rate of carbon (as CO) entering the gas phase from a carbon balance
 The rate of oxygen (as O2) leaving the gas phase from an oxygen balance
 The rate of air entering the gas phase from a nitrogen balance
 The rate of heat exchange from the gas phase from the enthalpy balance.

Internal Energy Balance

(Bournoulli Principle)
to estimate off-gas rate  y CO 
Rate of water in-
leakage is calculated
y 
from hydrogen balance  CO 2 
 yH 
 2 
Pstatic , T offgas , 
 y H 2O 
WATER
INLEAKAGE
y 
 O2 
 y N 2 

OXIDATION
Rate of air in-leakage is
calculated from nitrogen
CO balance
Carbon

Heat loss from gas-phase

Rate of CO from decarb is calculated from
& combustion is Rate of oxidation enthalpy balance
calculated from carbon (C, Fe, Si, etc.) is calculated from
balance oxygen balance

Figure 105: Freeboard model

GIVEN:
 Off-gas composition and temperature
 Static Pressure
 Oxygen, fuel and carbon rates
CALCULATE:
 Off-gas flow rate
 Water-in leakage
 Air-in leakage
 Rate of Oxidation
 Rate of Decarburization
 Off-gas Energy Balance
Note that the EFSOP system measures the off gas composition on a “dry basis”. The water in the
sampled off gas is removed by cooling and condensation before the gases are analyzed. The amount of
water removed from the sample is not measured. The fraction of water in the off gas is significant and

81
must be accounted for end up the balance. To do this, the gases are assumed to be at equilibrium at the
measured temperature and static pressure. Then, a routine based on the minimization of Gibb’s Energy
is used to calculate the fraction of water that is in thermodynamic equilibrium is calculated. The dry
chemistry is then scaled according to this equilibrium water content.
The elemental and enthalpy balance equations are detailed in the table below. The equations are written
as ordinary differential equations. In practice, however, the steady-state approximation is used (i.e. set
the derivative term to zero).
OXYGEN BALANCE
  yCO og
   yCOog
 
    
d   2  y og
    2  y og
  (43)
n   og   2  zOB  zCO  
CO2 CO2
dC
 2  y air
 z  z InLeak
 2  z Oxid
 z 
dt   y H 2O    y Hog2O  
2 O2 air H 2O O2 og

  


 2  yOog 

  2  y og
O2   2 
HYDROGEN BALANCE

d   2  y H 2    2  y Hog2  
og

  4  zCH 4  2  z H 2O  zog    (44)

n  B InLeak

dt  2  y H O    
 2  y H 2O 
og og
  2 
NITROGEN BALANCE
d
dt
 
n  y Nog2  y Nair2  z air  zog  y Nog2
(45)

CARBON BALANCE where:

y: concentration (percentage)
d   yCO    yCO 
og og
(46)
n  og   zCH 4  zCO  zog  og 
B dC  z: molar fraction (moles)
dt   yCO2   yCO 
 2  n: numbers of moles calculated in off
 
gas flow rate
ENTHALPY BALANCE og: off-gas
 Hˆ air  z air   B: bath
 InLeak InLeak 
 Hˆ H 2O  z H 2O  
 deCarb deCarb   Hˆ O2  zO2  
 
B B
d (47)
n  Hˆ og   H CO  zCO    
ˆ   Q
dt  ˆ Oxid Oxid   Hˆ CH 4  zCH 4  
B B

 H O2  zO2  
 ˆ 
 H og  zog 
Thermo camera installation
To have a reliable off gas flow rate model is a critical aspect necessary to close the mass balance and to
tune the freeboard model.
At the beginning, measurement of furnace off-gas temperature has been attempted with traditional
sheathed thermo-couples. Although thermocouples have been able to provide temperature
measurements for brief periods of time, they have been found to be un-reliable long-term in the harsh
EAF environment. Tenova has identified, and installed, an infra-red (IR) gas-phase pyrometric
technology based on the emissivity of carbon dioxide and is working towards adopting the technology
to the EAF-Consteel environment.
The pyrometer is working reliably since the installation. The installation position can be seen in Figure
106. In Figure 107 it is possible to see (green line) the temperature trend over a heat.

82
 Looking through the gap

Figure 106: Installation position of the pyrometer

 

Temperature (°C) 
Composition (%) 

temperature 

Time (min) 
Figure 107: Example of pyrometer measurement data (green line) in one heat
Freeboard model simulates the chemical reactions occurring in EAF gas phase and the energy transfer
from gas phase to the bath. The coupling of the freeboard model with the bath/slag model permits to
obtain a complete simulation of the EAF process.
Summarizing the freeboard model needs the following input data:
 Gas composition (CO, CO2, O2, H2, N2,), provided by EFSOP sensor;
 Gas temperature and pressure
 Off-gas flow rate
 Injected oxygen
 Injected carbon
The model calculates the chemical reactions occurring in gas phase and makes mass and energy
balances. The calculation gives the following output:
 Air leakage in furnace
 Water leakage in furnace
 Oxygen entering bath
 Carbon exiting bath
 Off-gas energy losses

83
  Panels energy losses
 Energy transfer to bath
In the mass balance are considered the following terms:
 O2 in the bath
 CO from the bath
 Air leakage in the furnace
 Water leakage in the furnace
 Steam in the off-gas
The energy balance is performed considering the following contributions:
Energy input = Efreeboard reactions + Efreeboard sensible energy +E (CO, CO2 from bath)
Energy output = E off-gas losses + E losses from panels + E transfer to bath
Assuming the arc energy goes completely to the bath and knowing the temperature of the bath the
freeboard model will calculate:
 Energy loss from off-gas
 Energy loss from panels and roof
 Energy transferred to the bath or from the bath
2.2.5.3 Validation of model developed by experimental measurement on the plant (carbon
concentration in steel, carbon content in slag and in dust sample inside EAF) (T 5.3)
Sensitivity analysis of the bath/slag dynamic model has been performed during the module development
in order to verify the model capability to respond to the variable process situations. Following table
reports an example of model calculations in standard and modified Consteel® operations.
Table 10: Model calculation for sensitivity analysis

Case Calculated (%FeO) in slag Calculated [%C] in the steel

standard 32.5 0.075

without C powder 37.9 0.043
double C powder (360 kg) 28.0 0.130
+630 m³ (STP) O2 33.4 0.059
Performed calculations indicated that the model was able to respond to input variations: respect a
defined standard condition, pulverised carbon and oxygen were modified. If coal injection is
suppressed, FeO concentration into the slag increases and C concentration in the steel decreases. If the
injection of coal is increased, a reverse trend is obtained.
In order to perform model calibration steel and slag samples have been collected. Figure 108 refers to
input of metallic materials (scrap and pig iron; the values on Y axis are percentage of total charged per
minute). Figure 109 refers to C and O2 injection.
The following assumptions were made for the calculations:
 scrap flow rate described by conveyor speed (homogeneity of scrap height and density) up to
total charged; concerning scrap composition %C has been set at 0.1 and Si/Mn content has been
tuned to reach a good description of elements and their oxides in slag
 whole carbon added (lump and powder) is assumed at 100% C
 whole O2 injected is assumed into the bath for reaction
 a curve for air leakage has been set (the same for two test heats)
 PC energy contribution has been neglected
 starting slag weight has been estimated by operator

84
 0.7 4.5
4.0
0.6
3.5
total charged per minute (%)

total charged per minute (%)

0.5
3.0
0.4 2.5

0.3 2.0
1.5
0.2
1.0
0.1 0.5
0 0.0
0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000

Time (s) Time (s)

Scrap charge Pig Iron charge
Figure 108: Input of metallic materials (scrap and hot metal; the values on Y axis are percentage of
total charged per minute
0.40 0.6

0.35
0.5
total charged per minute (%)

total charged per minute (%)

0.30
0.4
0.25
AC 5923
0.20 0.3
AC 6960
0.15
0.2
0.10

0.05 0.1

0.00 0.0
0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000
time (s) time (s)

Figure 109: Coal lump and oxygen injected (on Y axis % of total charged per minute)
Figure 110 and Figure 111 report the results of the optimisation of the kinetic parameters, showing that
a good agreement has been obtained for steel carbon content and bath temperature.

0.30

0.25

0.20

AC 5923
%C

0.15
AC 6960

0.10

0.05

0.00
0 500 1000 1500 2000 2500 3000

Figure 110: Comparison of calculated carbon content and measured

85
 1900

1800

1700

<T> metal/slag, °C
1600

1500 AC 5923
1400 AC 6960

1300

1200

1100

1000
0 500 1000 1500 2000 2500 3000
time, s

Figure 111: Average temperature of the slag/metal system as calculated by the code and
measurements in the liquid pool close to the slag door
Several campaigns with multiple sampling (steel and slag) and temperature measurements have been
carried out by ORI Martin to check the dynamic and the fitness between measured and calculated data,
at first in the calibration mode the second as the validation mode.
Figure 112 shows an example (as post processing) of multiple temperature measurement (that are
generally carried out in the second half of the heat) compared with the computation (temperature of the
computation is the average into the bath).

Figure 112: Comparison and computed (graph) and measured (table) bath temperature

86
Figure 113 and Figure 114 show further examples of the evolution of the steel and slag composition
over 50 minutes heat; the input was built with typical pattern of the Consteel® process in ORI Martin.
The obtained results (sensitivity analysis and calibration) indicate that the model was able to respond to
variation of input materials.
Concentration (%)

Time (s)

Figure 113: Steel composition evolution - example

Concentration (%)

Time (s)

Figure 114: Slag composition evolution - example

Partition of carbon provided by coals (lump and injected) is managed by coefficients (named RBS,
“Ripartitori Bagno Scoria” that means partition coefficient between metallic bath and slag) that are set
by technicians into the interface (properly modified by ORI Martin) in which data related to each
material are stored.
Figure 115 and Figure 116 show two different print screen of dataset for a scrap type and the coal
lump respectively; meaning of parameters are showed with labels.

87
 Scrap type:
steel sheet

Material group code for

computation (scrap,
coals, fluxes, etc)

Melting
coefficient
Figure 115: Print screen of datasheet for model input of scrap type

lump coal

Partition coefficient
(slag / steel)

Fluxes composition or
steriles (in scrap)
Figure 116: Print screen of datasheet for model input of coal and fluxes
2.2.5.4 Development of a system for continuous measurement of off-gas flow rate model (T 5.4)
Continuous off-gas flow rate measurement is necessary both for modeling purposes (to calculate the
amount of C and O transferred to the bath and calibrate the model) and for the online control system (to
elaborate the gas analysis from EFSOPTM probe and apply furnace control logics).
For this reason a virtual sensor, able to elaborate the information coming from already existing sensors
and to calculate the gas flow rate has been developed and tested.

88
For model calibration a standard measurement based on gas flow rate measurement with a Pitot tube
and simultaneous gas analysis downstream was performed.
Gas flow rate measurements by Pitot Tube

Figure 117: Scheme of sampling points for gas flow rate continuous measurement
The flow rate continuous measurement has been performed by means of a Pitot tube, installed in the
duct far from the EAF, coupled with gas sensors (Figure 117). The following data have been recorded:
Off-gas analysis from EFSOPTM at fourth hole
 Process parameters (lance oxygen, scrap, pig iron, injected coal, electrical power)
 Gas velocity and composition measured in the fume duct
With this data it was possible to calculate the off-gas flow rate from the EAF.
The off-gas flow rate QEAF is determined by the following mass balance:
Q EAF   Cox
EFSOP
 Q out   Cox
out
(48)

where:
Qout: is the gas flow rate measured downstream
xCOxEFSPO is the gas composition measured by EFSOP
xoutCOx is the gas composition measured downstream
As an example results of the mentioned calculation procedure is reported for heat AC2081

Heat AC 2081
25000 20
18
Q EAF
20000 16
EAF / Leak Air MFR ratio

Q EAF / Q LA
EAF MFR (Nm3/h)

14
15000 12
10
10000 8
6
5000 4
2
0 0
0 500 1000 1500 2000 2500
progr time (s)

Figure 118: Results of mass flow rate calculations at fourth hole

89
Development of Mass Flow rate virtual sensor
The development of a virtual sensor, able to elaborate the signal from already existing sensors and to
calculate the off-gas mass followed two approaches:
Approach 1: by static pressure and temperature downstream (fume duct) measurements and algorithm,
total (furnace + leakage) mass flow rate of the gas is calculated – the on time EAF off-gas MFR
determination needs, in this case, a CO2 concentration analyser, at the same point to perform carbon
balance with the EFSOP at the 4-th hole.
Approach 2: by static pressure measurement at the connecting car and managing of the EAF / total
MFR (calculated at that point) versus process parameters.
The idea was to use the static pressure (standard measure) just downstream the connecting car inlet to
evaluate mass flow rate and correlate it to several process parameters detected in the furnace to
determine the amount coming out through the fourth hole; it needs calibration, provided by the
measurement campaigns carried out for carbon balance in the fume duct.
The Approach 2 as been selected as the most suitable for a continuous online application. Figure 119
summarise the logic scheme of the mass flow rate virtual sensor.
Q_SP: is the gas flow rate calculated from static pressure measurement
Q_EAF: is the off-gas flow rate
Q_AL: is the flow rate of air leakage

Q_AL
Static Pressure, Temp
Calculated
Q_EAF
Q_SP

Q_EAF / Q_SP

From carbon balance (downstream

measurements)
- average value (approx)
- dynamic evaluation through furnace parameters
relationship

Figure 119: Logic scheme of the virtual sensor for on line gas flow rate measurements (approach 2)
CFD calculations have been also performed to characterize the flow inside the pre-heater and evaluate
the suitability of static pressure measurement points through it in determining off-gas MFR. The flow
field calculated is reported in Figure 120. The results obtained with CFD calculations indicated that the
position of the sensor of static pressure is far enough the vortical flow which is present at the beginning
of tunnel.

Velocity contours (m/s)

x
x1 x2 x3 x4

x1= -2 m x2= -1 m x3= +1 m x4= +5 m

Figure 120: CFD calculations of flow field

90
According to this approach the off-gas mass flow rate is continuously computed by a mathematical
relationship that takes into account the pressure (dP) at the exit of the furnace (so called ‘connecting
car’), the temperature of the gas (T), the off-gas composition (%N2, the nitrogen content)
MFR  f  dP   k  T, %N 2  (49)
The parameters of the algorithm are calibrated through double point measurements campaigns (carbon
balance).
The relationship (approximated to linearity in the graph) of variation of flow rate with static pressure
has been verified and reported in Figure 121.
Figure 122 reports the comparison, for a test heat of calculation of Q_EAF from static pressure
measurement and QEAF measured by double point measurement and carbon balance. The agreement is
good.

110000

100000

90000

80000
Q_SP, Nm3/h

70000

60000

50000

40000

30000

20000
0 1 2 3 4 5 6 7
static pressure * (-1), mmH2O

Figure 121: Static pressure measurement and gas flow rate Q_SP calculated from mass balances and
developed algorithm
25000

Q_EAF_algo
20000
Q_EAF_CB
Q_EAF, Nm3/h

15000

10000

5000

0
0 500 1000 1500 2000 2500 3000 3500
time, s

Figure 122: Comparison of calculation of Q_EAF from static pressure measurement and Q_EAF
measured by double point measurement and carbon balance
Off-gas temperature measurement
A continuous off-gas temperature measurement is necessary for the mass flow rate calculation. A
Tenova dedicated sensor has been installed and is currently providing the measurement used by the
FreeBoard model to calculate the efficiency of energy transferred to the bath.

91
A second temperature probe has been installed closer as much as possible to the location of the pressure
sensor, for gaining a calibration and a cross check in measurement of the thermo camera.
To reach this goal a suction pyrometer has been appositely designed. This probe has been studied in
order to avoid two typical problems that are encountered in the EAF fumes environment:
 Clogging by dust in the off-gas
 Slow response of thermocouple
The clogging of dust has been solved with a new design of the apparatus, while the slow response of the
thermocouple has been solved with a new design of the protective tube (Figure 123). Usually an
alumina protective tube is used, which slow down the thermal exchange with off-gas. The developed
tube is made with steel, with higher thermal diffusivity respect to alumina for furnace lab test results
with different materials – blue line relates to the steel casing, magenta to the alumina). On the inner side
of the steel tube a very thin alumina protective layer was deposited, to avoid thermocouple
contamination as a consequence with contact with steel.
Protective
Metallic tube
allumina layer

thermocouple

Figure 123: Improvement of thermocouple protective tube to decrease the time of response.
Figure 124 shows the water cooled probe and the position at the EAF fourth hole outlet.

Figure 124: Water cooled probe and the position at the EAF fourth hole outlet
Figure 125 reports the temperature measurement during one heat with the new developed suction
pyrometer, confirming the reliability of this apparatus.

92
 1400

1200

Temperature, °C
1000

800

600

400

200

0
18.00 18.10 18.20 18.30 18.40 18.50 19.00 19.10
time, hh.mm

Figure 125: Temperature measurement during one heat with suction pyrometer
2.2.5.5 Development of new on-line model for the dynamic control of carbon and oxygen injection in
the EAF and dynamically control of the post combustion between the EAF and the Consteel
tunnel (T 5.5)
The scope of this task was to build a dynamic model able to control the operative process based on the
EFSOP® off-gas chemistry measurement. The final issue was to find out the optimum oxygen and coal
managing as well as post combustion ratio between furnace and Consteel® pre-heater in order to
optimize the energy consumption over the heat.
In order to evaluate the ORI Martin operative practice methodology it has been made an accurate
process data analysis. The idea has been to “make a picture” of the plant status with the purpose of
being able to evaluate coherently the research progress. Having a so defined “baseline” it is possible to
appreciate the benefits deriving from the application of the off-gas analysis system coupled with the
control module.
The needed to have a reference appear immediately looking for the ORI Martin production. ORI Martin
produces special steel using selected scrap. To assure the high quality steel the process is oriented to
obtain an optimum slag condition in terms of foaminess and chemical composition. All the process data
analyzed are filtered to reject the heats not significant respect to the standard process and the acquisition
errors. Nevertheless all the parameters have been related to the balance consumptions to the plant; thus,
it has been guarantee a perfect match with the real operative practices.
It is important to account all the factors relevant to the process in order to correctly evaluate the benefits
of any practice. This is especially true for the ORI Martin Plant whose process has important
constraints:
ORI Martin Plant produces many different high quality steel grades. Each grade has its own operating
practice and materials and energy consumption;
 because of the high steel quality requirements optimum slag composition and foaming
conditions become a constraint too;
 environmental constraints play a fundamental role. Ori martin is particularly close to an area
densely populated.
To consider all the different kind of steel produced (about 150) the steel grades production has been
divided into families of steels with common final requirements and practices. The steel grade has
grouped in two “family” depending on the slag target composition. The slag target S007 is used to
produce high carbon steel grade, the S004 for low carbon steel grade. For each family has been
calculated a “baseline”. The final baseline is a weighted average of the single baselines accordingly
with the plant production. Similar calculations have been carried out analyzing the pig iron in charge
(Figure 126), the heats executed with basket additions (7.5%), the tap temperature (Figure 127), the
composition and the scrap size.

93
Figure 126: ORI Martin pig iron in charge. Progressive n° of heats vs. Pig iron in charge

Figure 127: ORI Martin tap temperature. Progressive n° heats curve Vs Tap temperature
Looking out all this parameters it has been possible to assure an excellent process reproducibility. In
addition, many other parameters were analyzed in order to reach a better and more comprehensive
understanding of the benefits/effects of the new practice:
 chemical and physical characteristics of carbon and lime;
 slag chemical composition;
 composition and weight of fume system dust;
 productivity;
 electrode consumption.
Post combustion oxygen controlled by feedback (CLC)
ORI Martin installed one PC injector on the left side of the slag door, between the slag door and the
conveyor. The intent is to increase and focus post combustion reactions in the coldest area of the
furnace, right above the location where the scrap falls into the bath. Post combustion is then completely
achieved in the conveyor.
Thus there has been the possibility to drive the process acting on the oxygen injection by a post
combustion burner installed between the slag door and the connecting car. The position and the
direction of such burner (height and angle-shot) have been modified more times, in order to better
control and increase the post combustion efficiency. During the end of June 2007 Ori Martin has
completed the necessary PLC modifications to power on the dynamic control (post combustion oxygen
injection), controlled by EFSOP® system; so Tenova and ORI have been implemented the dynamic post
combustion oxygen control (CLC; close loop control). The first step in the Consteel® optimization is

94
based on the off-gas redox reaction. The target is to keep as stable as possible a fixed value in terms of
CO+H2. The PC logic is based on two values to set up. A minimum and a maximum CO+H2 value;
under the minimum the burner works in “low post combustion”, above the maximum it works in “max
post combustion”. Between those set points the burner works in “middle post combustion”.
In Figure 128 is schematically shown the control logic. All the parameters are easy to set by the user.
For all control limits it is possible open or close the valves link with the correspondent post combustion
oxygen line. It has already implemented the possibility to set the oxygen flow rate numerically (in case
of proportional valves installation).
Figure 129 shows a portion shot in the EFSOP® HMI screen concerning the post combustion control. It
has possible to see the tow, definable from the user, thresholds necessary to characterize the chemical
atmosphere on the furnace (reducing or oxidant) and to set the post combustion control limits.
It is possible to act one by one on the valve or to set up the flow rate singly.

Figure 128: Oxygen injector post combustion control logic

Figure 129: EFSOP® HMI screen shot; dynamic control post combustion set point
The carbon feeding and injection has been adjusted basing on the CO concentration. The carbon feeding
has been slightly increased (Table 12). The idea has been to approach the refining phase having a pretty
high carbon concentration in order to reduce the iron oxidation. The strong CO development sustains
the slag foaminess even if the slag temperature increases. That’s allowed to reduce to overall carbon
injected (Table 12).
The oxygen and the carbon feeding/injection pattern has been studied and optimized thanks to the ORI
Martin’s furnace expertises.
Further aspects of the online control system are reported in section 2.2.7.6, while the screen shot of the
bath status with the continuous variation of carbon concentration is reported in Figure 191.
2.2.5.6 Application and trial of the new on-line model (T 5.6)
After put on line the control logic model efficiency some trials on field have been carried out by
Tenova. The goal has to evaluate the post combustion effect on the Consteel® process flat bath and find
out its best ratio between furnace and Consteel® conveyor. The carried out trials demonstrated a great
flexibility in the dynamic PC control. The results gained in terms of consumption and energy savings
represent big economical advantages.

95
In the trials period the produced steel quality has been, compared with the baseline period, almost
constant (Table 11). As already described (in the baseline paragraph) the different kind of steel
produced may be split basing on the slag target (S004 and S007). The value appears rather stable; so it
is possible a coherent results evaluation.
Table 11: Production expressed in terms of target slag during the baseline and performance period
Steel produced (slag target)
Period S004 S007 Total
Baseline 7.57% 92.43% 100.00%
Trials Period 6.63% 93.37% 100.00%
The results, shown in Table 12 in detail, demonstrate the usefulness in terms of process control of the
information available with EFSOP® system and the CLC logic efficiency. The results analysis and the
furnace observation help to understand the benefits given by controlled injection oxygen.
The oxygen blown reacts immediately, around the beginning of the conveyor, in the reducing furnace
atmosphere.
The energy obtained from CO combustion causes:
 decreasing in thermal loss from the bath (The heat acts like a “cover”. It protects the slag and
the bath from an excessive irradiation. It has paid attention to avoid a slag super heat to
guarantee a good foaminess. Moreover the slag has good also in terms of chemical
composition);
 increase the thermal exchange and the scrap temperature in proximity to the furnace (The heat
increases the scrap temperature and speeding up its melting. Generically the post combustion
warms the cold furnace zone);
 increase the thermal exchange in the end of conveyor (The higher off-gas temperature, give by
the high post combustion ratio, increase the thermal exchange in the end of conveyor end
before, through the connecting car, the scrap fall down into the furnace. It doesn’t mean that all
the post combustion is carried out in the furnace; however to obtain the maximum energetic
yield exists an optimal ratio between the post combustion developed in the conveyor and in the
furnace).
Table 12: EFSOP® PC CLC results (Best month achievement over the trials period)
Parameter Deviation
Electric Energy kWh/t -2.92%
Oxygen m³/t +1.76%
Carbon in Charge kg/t +0.98%
Carbon injected kg/t -4.35%
Dolo Lime kg/t +0.74%
Lime kg/t -25.46%
EAF slag kg/t -22.49%
Fe in slag kg/t -29.91%
It has been others improvement to achieve the benefits shown. Since September 07 the burner has
moved and directed to work only as post combustor. It has seen:
 a reduction in oxidizing oxygen and an increasing in energetic yield (This change affects
positively the slag quantity and composition. The FeO the CaO and the amount of slag have
decreased. In October 07 has been recorded the lower mass slag produced since the beginning
of the year);

96
  process more stable (The process is in the last months has been more stable with consumption
close in a small range. This involves a better process and production manages for all the plant);
 higher inlet off-gas temperature measured at the quenching tower (The information about the
tower inlet temperature cannot to be used to prove he post combustion efficiency and the heat
performance. However they are indicative of the effects already explained and may help a
better understanding of the physical events. The post combustion reaction increase the off-gas
volume so, at the same fume system working condition, more high is the post combustion less
is the false air sucked from the slag door and the gap. The post combustion effect along the
entire conveyor increase the thermal exchange and so the scrap preheating efficiency). The post
combustion effect is visible also in the off-gas chemical analysis measured by EFSOP®.
The post combustion oxygen blowing decreases the sensible energy in the furnace increasing the
chemical one. It is possible to see it in Figure 130; CO (red) and H2 (yellow) drop and CO2 (purple) rise
quickly.
Even if the results gained by the logic implemented were very good there was not clear understanding
about the post combustion ratio between the EAF and the Consteel® pre-heater.
EFSOP CLC OFF

EFSOP CLC ON

Figure 130: Comparison with chemical analysis between post combustion ON vs. OFF
where CO2 is magenta, CO is red, H2 is yellow, O2 is blue

97
The combustion distribution has a very deep impact on the control of the process. The best optimization
with a dynamic control is reached when balancing the post combustion distribution between the furnace
and within the tunnel.
Thus has been chosen to install a downstream analyzer at the end of the Consteel® tunnel. A system has
been designed, deriving from the EFSOP® technology, and installed downstream at the exit of the
preheating tunnel in order to extract and analyze the O2%, CO% and CO2% content of the off-gas.
Lots of effort has been done to make the analyzer work reliably (Figure 131). The main installation
purpose has been to have a continuous downstream chemistry measured to be compared with the
EFSOP® upstream one.

Figure 131: Example of measured downstream chemistry

Figure 132: HMI main page updated – the upstream and the downstream chemistry can be seen
together

98
The main HMI page has been modified in order to see the chemical analysis together (Figure 132).
That allowed to cross correlate the post combustion distribution helping the furnace operator to carry
out the heat.
Last but not least the downstream chemistry helps to tune the freeboard and bath and slag model.
Looking at the oxygen excess it can be understood if the nitrogen balance (air in leakage) and then the
off gas flow rate model are working correctly.
2.2.6 Development of CRM process control at PA with special emphasis on off-gas analysis
(WP 6)
2.2.6.1 Modification of CRM dynamic metallurgical model in order to include off-gas information as
input (T 6.1)
The CRM model has been applied and tuned for the Differdange furnace. Several phases of data are
necessary to obtain the required format to enter the CRM model.
This work required the extraction of cyclic data from the furnace data acquisition system. Some data
formats being inadequate and some information being not available in a cyclic form (for example,
temperature sampling and scrap charging sequence), a pre-processing of the input data has been carried
out prior the running of the model. Figure 133 gives an example of results of simulation with the EAF
model, comparing the exhaust gas composition to the on-line measurement.
70
(MWh), (%)
Heat 7405
60

50
CO

H2O
40
H2
30

O2
20
CO2
10

0
11 h 30 11 h 40 11 h 50 12 h 00 12 h 10 12 h 20 12 h 30
Energy CH4 CO CO2 H2 O2
PC lances CO model CO2 model O2 model H2O model

Figure 133: Comparison between off-gas measurement at AM Differdange with model simulation
Several remarks come from the examination of these evolutions: (a) The decomposition of water into
hydrogen is not foreseen in the CRM model, high H2O values are observed during the preheating phases
coming from the combustion of the natural gas; (b) The response time of the oxygen analyser induces a
slow decrease of the oxygen level compare to the model calculation; (c) Except for hydrogen, the
calculated and measured levels are in the same order of magnitude; further considerations will be
required to complete this comparison. A first way to compare the model gas to the exhaust gas
measurement could be to assume a complete reaction of water with the carbon monoxide according to
the water gas shift reaction:

CO  H 2 O  CO 2  H 2 (50)

Figure 134 shows the resulting evolution:

99
 70

(MWh), (%)
60
Heat 7405

50
CO
40 H2O
H2

30

O2
20
CO2
10

0
11 h 30 11 h 40 11 h 50 12 h 00 12 h 10 12 h 20 12 h 30
Energy CH4 CO CO2 H2 O2
PC lances CO model * CO2 model * H2 model * O2 model H2O model *

Figure 134: Comparison between exhaust gas measurement with model simulation, assuming a
complete decomposition of water in the fume (ArcelorMittal Differdange)
This assumption adversely affects the carbon dioxide level. Moreover, a significant amount of water is
still present during the preheating of the first and second basket. A possible explanation of these
behaviours could be the effect of the off-gas volume flow rate which could significantly differs from the
calculation, the model hypothesis being a constant air ingress flow entering the vessel and a fixed post-
combustion ratio. The second analyser located in point B allows the assessment of the gas flow rate at
the furnace exit according to the following relationship (“carbometry” principle):

Q Furnace    CO A   CO 2 A   Q Venturi
Exhaust   CO 2 B (51)

An example of this calculation is given in Figure 135 for the same heat.
300 000 70

Nm³/h Heat 7405 (MWh), (%)

60

Picks 50
200 000 due to the
analysers
reponse time Flow rate (model) 40

Flow rate 30

100 000
(offgas information)
20
CO2 point A

10

CO2 point B
0 0
11 h 30 11 h 40 11 h 50 12 h 00 12 h 10 12 h 20 12 h 30

Q furnace Q model Energy CO2 PC lances CO2 point B

Figure 135: Comparison between off-gas volume flow rate deduced for the off-gas measurements with
model simulation
The exhaust gas flow rate deduced from the gas analysis on point B is in the order of magnitude of the
model calculation. Abnormal picks are observed at the melting start of each basket, they are due to the
analysers response time, the steady state levels for CO and CO2 being reached after several minutes.
The on-line use of the off-gas measurements to appraise the furnace exhaust gas flow rate is
questionable and therefore should not enhance the accuracy of the model. During the last measurement
campaign, only one furnace vessel was operating, hopefully the instrumented one. In this particular
circumstance, the second dedusting system was closed and didn’t influence the flow rate assessment. In

100
normal operation, these calculations would be difficult to conduct for a double shell furnace. As already
mentioned, the Differdange furnace is equipped with a Venturi flow measurement in the exhaust duct.
In order to check the validity of this information, this flow rate has been compared to the value given by
the quench unit (thermal balance between the cooling water flow and the temperature delta in the gas).
Figure 136 gives this comparison. Both flow rates are in the same order of magnitude, despite the
calculation at quench is no more available when the water spraying is stopped, this gives a scattered
signal.
600 000 70

Nm³/h Flow rate (MWh), (%)

Heat 7405
500 000
at quench 60

Venturi flow rate

50
400 000

40

300 000

30

200 000
20

CO2
100 000
10

0 0
11 h 30 11 h 40 11 h 50 12 h 00 12 h 10 12 h 20 12 h 30

Q quench Q Venturi Energy CO2 PC lances

Figure 136: Comparison between Venturi off-gas volume flow rate and the value deduced from the
quench unit heat balance
On the basis of the main events occurring during the process, it is possible to deduce the measurement
time delay. As illustrated in Figure 137 and Figure 138, a value of 2’40” has been assessed. The main
time delay comes from the length of the sampling tube to the analysers (the drop down of the CO, CO2
and H2 curves are immediate at the opening of the roof for charging the scrap basket).
The second analyser located in point B allows the assessment of the gas flow rate at the furnace exit
according to the following relationship (“carbometry” principle):

Q Point A    CO A   CO 2 A   Q Exhaust
Venturi
  CO 2 B (52)

70 PC lances
Coal
(MWh), (%) Burners
60
Energy
With
correction
50
CO
Original
40

2'40" delay
H2
30

20

CO2
10

O2
0
14 h 45 14 h 55 15 h 05 15 h 15 15 h 25 15 h 35 15 h 45

Figure 137: Assessment of the exhaust gas measurement delay

101
 70 PC lances
Coal
(MWh), (%) Burners
60
Energy

50 CO

40

2'40" delay
30
H2
20
CO2

10

O2
0
16 h 45 16 h 55 17 h 05 17 h 15 17 h 25 17 h 35

Figure 138: Assessment of the exhaust gas measurement delay (continuing)

The synchronisation between the analysis values in point A and B is mandatory to calculate the correct
flow rate at the furnace outlet. The new result is given in Figure 139.
Moreover, a nitrogen balance on the furnace vessel allows assessing the air ingress according to the
following equation:

Q Point A  100   CO A   CO 2 A   O 2 A   H 2 A   Q Furnace

Air ingress
 79 (53)

200 000 70

Q point A
60
Q air ingress
C
Q process gas
150 000
Q model 50
PC lances
Energy
Gas flow (Nm3/h)

Energy (MW)
40

100 000

A 30

B 20
50 000

10

0 0
11:30 11:40 11:50 12:00 12:10 12:20 12:30

Figure 139: Assessment of the exhaust gas flow rate and air ingress
Air ingress is significant during the preheating phase of the first scrap basket. As illustrated in Figure
140 (picture A), the furnace atmosphere is under negative pressure and promotes air ingress.
When the melting of the second basket is progressing (time B in Figure 139 and Figure 140), flames
are generated in the roof gap due to the high process gas flow.
The refining phase (time C in Figure 139 and Figure 140) is characterised by huge gas flow and air
ingress, mainly due to the opening of the slag door for de-slagging.

102
Figure 140: View of different process phases; Time A (at left): Burner preheating of the first basket;
Time B (at center): Melting of the second basket; Time C (at right): Refining phase.
Comparing the model calculation to the exhaust gas flow rate deduced from the gas analysis on point B,
they appear to be in the same order of magnitude. The air ingress is under-estimated at the start of the
preheating phases and during the refining period.
2.2.6.2 Performance of simulation trials with modified injection pattern of post-combustion oxygen
lances, the off-gas measurements being active (T 6.2)
The aim of the first campaign carried out in the ArcelorMittal Differdange was to appraise the potential
gain in the exhaust gas in regards of post-combustion reaction inside the furnace. Considering the
evolution of the different gas species, the origin of high hydrogen level in the measured value, up to
30%, must be first clarified before implementing the available exhaust gas information as input in the
EAF CRM model. As mentioned in the previous report, Figure 141 gives the evolution of the species
present in the exhaust gas as a function of temperature, assuming an equilibrium state. It is based on the
water gas shift reaction (equation (54)).
CO  H 2 O  CO 2  H 2 (54)
This simple equilibrium is applicable during the preheating phase, when burner combustion produces
high volume of water gas. On flat bath, the water source could be the electrode cooling. In this case, the
cooling water could also react directly the electrode carbon according to the following reaction
(equation (55)).
Celectrode  H 2 O  CO  H 2 (55)

[X] (%)
60

50
H2O
40

30 CO
C
H2
20

CO2
10 N2

O2
0
0 500 1000 1500 T (°C) 2000

Figure 141: Thermodynamic equilibrium of the exhaust gas

If we assume that this reaction is the predominant mechanism of hydrogen formation, we can calculate
the electrode erosion as a function of time. To achieve this, we will deduce the CO and CO2 volume
flow from the combustion of the burner gases and the coal injection. Figure 142 shows this evolution,
the exhaust gas flow rate is assessed on the basis of the CO and CO2 content measured at the furnace
elbow. Considering this electrode wear mechanism the only source of hydrogen, the electrode erosion

103
can be easily calculated on the basis of the exhaust gas volume and hydrogen content. This evolution is
given in Figure 143.
Integrating this electrode erosion curve, a total amount of 850 kg of carbon should have reacted with
water to form the hydrogen present in the fumes. This result doesn’t correspond to the theoretical
erosion value, which is function of the electrode current and close to 130 kg per heat. Moreover, the
electrode wear on flat bath is overestimated. Flat bath phase corresponds also to the coal injection
period; this observation tends to anticipate the reaction of the fine particles of the injected coal with the
water present in the furnace chamber atmosphere.
70 Heat 6624 Coal injection 80000
(%)
(MWh) Oxygen blowing (Nm³/h)
PC lances operating
Electrical energy 70000
60
consumption (MWh)
Gas flow rate
60000
50

50000
Cleaning of
40 the probe
Second CO Tapping
basket 40000

30 First basket CO+CO2 flow

H2 rate 30000

20
O2
20000
CO2
10
10000
CH4
0 0
2 h 30 2 h 45 3 h 00 3 h 15 3 h 30 3 h 45

Figure 142: Exhaust gas volume assessment through mass carbon balance

70 Heat 6624 Coal injection 6000

(%)
(MWh) Oxygen blowing (kg, kg/h)
PC lances operating
Electrical energy
60
consumption (MWh) 5000
Electrode
erosion (kg/h)
50
Tapping 4000
Cleaning of
40 the probe
Second Theretical
H2 basket Electrode
3000

30 First basket erosion (kg)

2000
20
Electrode
erosion (kg) 1000
10

0 0
2 h 30 2 h 45 3 h 00 3 h 15 3 h 30 3 h 45

Figure 143: Appraisal of the electrode erosion assuming cooling water dissociation

104
 70 Heat 6624 Coal injection 80000
(%)
(MWh) PC lances operating
Oxygen blowing (Nm³/h)
Electrical energy 70000
60
consumption (MWh)
Coal injection Gas flow rate
60000
50 yield = 54 %

50000
Cleaning of
40 the probe
Second CO Tapping
basket 40000

30 First basket CO+CO2 flow

H2 rate 30000

20
O2
20000
CO2
10
10000
CH4
0 0
2 h 30 2 h 45 3 h 00 3 h 15 3 h 30 3 h 45

Figure 144: Exhaust gas volume assessment through mass carbon balance with introduction of a coal
injection yield
70 Heat 6624 Coal injection 6000
(%)
Oxygen blowing
(MWh) PC lances operating (kg, kg/h)
Electrical energy
60
consumption (MWh) 5000
Electrode
Coal injection
yield = 54 %
erosion (kg/h)
50
Tapping 4000
Cleaning of
40 the probe
Second Theretical
H2 basket Electrode
3000

30 First basket erosion (kg)

2000
20
Electrode
erosion (kg) 1000
10

0 0
2 h 30 2 h 45 3 h 00 3 h 15 3 h 30 3 h 45

Figure 145: Appraisal of the electrode erosion assuming cooling water dissociation with introduction
of a coal injection yield
Assuming a coal injection yield of 54%, the calculated gas volume is reduced as illustrated in Figure
144. On the basis on this additional hypothesis, the new resulting electrode erosion can be deduced. As
shown in Figure 145, the new electrode erosion value copes with the theoretical value. All this
assumptions will have to be corroborated by the next measurement campaigns to be carried out in
ArcelorMittal Differdange, a technique to appraise the exhaust fumes flow rate is mandatory (set-up of
a second measurement point in the main duct). The remaining amount of methane detected in the fumes
should not come directly from the burners but from the reverse reactions with the hydrogen (equation
(56) and equation (57)).
CO  3 H 2  CH 4  H 2O (56)
CO2  4 H2  CH 4  2 H 2O (57)
During the second campaign of measuring the off-gas there were two sampling points. Point A was
located at the elbow near the EAF and point B was located at the end of the water-cooled hot gas duct.
At point A the post combustion is finally ended. The off-gas temperature was not to be measured while
this campaign. Figure 146 shows the operating pattern profile. While 1st Power on period (feeding the
1st basket into the EAF) the O2 injection is on. Simultaneously the Natural gas burner is switched on
too. The corresponding off-gas profile (point A) is shown in Figure 147. While 1st Power on period the

105
CO content reaches 30% while H2 is increasing up to 35%. The H2 and Co profiles/courses are very
similar. While 2nd Power on period O2 injection is switched on while Natural gas is on too. When the
Natural gas is off the Coal injection is switched on. While Natural gas is switched on the H2 and CO
profile are similar too. They begin to differ when the Coal injection is switched on. This is due to
foaming slag period. When the slag starts foaming the effective volume left in the EAF vessel is
decreasing. The post combustion zone inside the EAF vessel is changing its position. Consequently the
off-gas profile is much different. The H2 and CO profile are very similar at 1st Power on period as well
as in the beginning of the 2nd Power on period. One may say there is a technical problem measuring
high contents of H2 and CO simultaneously. In case of similar H2 and CO profile at point B the used
technique for off-gas measuring at point B is proven. Figure 148 shows the off-gas profile (point B):
(a) O2 and CO2 off-gas profile (top); (b) CO and H2 off-gas profile (bottom). The H2 and CO off-gas
profile are very similar too. The off-gas profile (point A) is shown in Figure 148. At point B there is
still a small CO content measured. This is due to a non total post combustion process. The H2 off-gas
content (point B) is 3 times lower than CO. Because of the same value of CO and H2 at point A while
1st Power on period the H2 post combustion is preferred. If one is looking for 100% post combustion of
CO he has to take the H2 content in account too.

— Natural gas — O2 Injection

— Coal Injection — Power on

0 10 20 30 40 50 60
Time [min]

Figure 146: Second campaign of exhaust gas measurement: operating pattern profile– Reference heat
(point A)
100 80
— — O [%] — CH [%]
H2 [%], CO [%], CO2 [%], O2 [%], CH4 [%]

H2 [%] 2 4
90
— CO [%] — CO [%] — Electric Energy input
2 70
80
60
Electricalinput (MWh)

70 Flushing
60 50
1st Bucket st
2 Bucket
50 40
40 30
30
20
20
10 10

0 0
0 10 20 30 40 50 60
Time [min]

Figure 147: Second campaign of exhaust gas measurement: off-gas profile– Reference heat (point A)

106
 30
—O [%]
2 — Natural gas — O2 Injection

25
—CO [%] 2 — Coal Injection

CO2 [%], O2 [%] 20

15

10

1500 500
— H2 [ppm] — Natural gas — O2 Injection
— CO [ppm] — Coal Injection
1200 400

900 300
CO [ppm]

H2 [ppm]
600 200

300 100

0 0
0 10 20 30 40 50 60
Time [min]

Figure 148: Second campaign of exhaust gas measurement: off-gas profile (CO2,O2) (top); off-gas
profile (CO,H2) (bottom) – Reference heat (point B)
During the second campaign performed in ArcelorMittal Differdange, the manager of the furnace
allows stopping the electrode cooling during the melting process. Figure 149 gives, as reference, the
evolution of the gas measurement at the end of the previous heat. The purge of the probe was set to a
cyclic mode, inducing some disturbance in the gas measurement due to a high oxygen residual level.
For this heat, the hydrogen grew up to 25%. Figure 150 illustrates this action of stopping the water
cooling of the electrode for several periods of time. The water flow rate supplied to the electrode is
measured by an existing classical water flow meter and is given as a function of time. A low hydrogen
value is observed, down to 5%, in the period when water cooling is stopped. This leads to corroborate
the conclusions previously given. A more precise quantification of the electrode cooling effect will be
possible by analysing the information brought by the second measurement point on the exhaust gas flow
rate.

107
 70 10

(MWh), (%) Electrode water flow Heat 7374 (m³ water / h)

60 0

50 -10

High H2 level !
40 -20

30 -30

Probe purge Probe purge

20 -40

10 -50

0 -60
8 h 25 8 h 30 8 h 35 8 h 40 8 h 45 8 h 50

Energy CH4 CO CO2 H2 O2 Electrode cooling

Figure 149: Second campaign of exhaust gas measurement – Reference heat

70 10

(MWh), (%) Electrode water flow (m³ water / h)

60 0

50
Heat 7375 -10

Stop of electrode water cooling

40 -20

Low Hydrogen level !

30 -30

Probe purge Probe purge

20 -40

10 -50

0 -60
10 h 15 10 h 20 10 h 25 10 h 30 10 h 35 10 h 40

Energy CH4 CO CO2 H2 O2 Electrode cooling

Figure 150: Second campaign of exhaust gas measurement – Stop of the electrode cooling
During this second campaign, the operating pattern of the post-combustion oxygen injectors has been
modified in order to increase the oxygen input in the vessel. Figure 151 compares the modified
operations to the normal situation. The oxygen flow rate is given as a function of the electrical energy
value, which is currently, used an indicator of the melting progress. The same operating pattern has
been applied for each scrap basket. The changes consist in speeding up the starting of the lance,
increasing the flow rate in the steady phase and pursuing the oxygen blowing at a reduced flow further
in the melting. This action has been conduct for 29 heats. No significant difference has been observed
between the normal and modified practice.

108
 PC oxygen lance flow rate (Nm³/h)
900
800
700
600
500 Normal PC
400
300
200
100
0

900
800
700
600
500 Improved PC
400
300
200
100
0
0 5 10 15 20 25
Electrical energy (MWh)

Figure 151: Modification of the operating pattern of the PC oxygen injectors

The CRM model has been applied to this data set as illustrated in Figure 152. Figure 153 and Figure
154 compare the exhaust gas composition for normal and modified practice. Both evolutions are
similar, the correspondence is even better in the case of the modified practice. For the next tests in
Differdange, the question was to know further improvements are possible by increase the oxygen input.
Figure 155 compares the yearly average values of electrical energy consumption with various European
furnaces. The selected parameter is the ratio of the supplied oxygen to the required one to fully react
with fossil input (coal and natural gas) for producing carbon dioxide. Single heats from normal and
modified practices have been added on the same graph. It appears that some energy savings could come
from an increased use of the oxygen supply. These first tests were focused on the increase of the oxygen
flow rate of the PC lance. Because of their poor efficiency and because the lance moved for an
impingement blow increasing the flow, the improvement was limited. The next measurements
campaigns were devoted to the increase of the oxygen supply by various operating patterns.
T model (°C)
1750

1725

1700

1675

1650

1625

1600 Normal Practice

Enhanced Practice
1575 False T° 1/1 line

measurements
1550
1550 1575 1600 1625 1650 1675 1700 1725 1750 T dip (°C)

Figure 152: Results of CRM model simulations for normal and modified practices

109
 70

(MWh), (%)
Normal practice
60

50
CO
H2O
40

30

H2 O2
20

10 CO2

0
16 h 45 16 h 55 17 h 05 17 h 15 17 h 25 17 h 35 17 h 45
Energy CH4 CO CO2 H2 O2
PC lances CO model CO2 model O2 model H2O model

Figure 153: Comparison between exhaust gas measurement with model simulation for normal practice
70

(MWh), (%) Modified practice

60

50

40
H2O CO

30

H2 O2
20

10
CO2
0
14 h 50 15 h 00 15 h 10 15 h 20 15 h 30 15 h 40 15 h 50
Energy CO CO2 H2 O2 PC lances
CO model CO2 model O2 model H2O model

Figure 154: Comparison between exhaust gas measurement with model simulation for modified
practice
550 Yearly averaged data and single heats
Electrical energy
consumption
(kWh/ton)
500

450

400

350

European furnaces
300 Differdange - Average
Heats Differdange - Normal practice
Heats Differdange - Modified practice Stoechiometric ratio
250
oxygen to carbon
and natural gas

200
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Figure 155: Comparison of the electrical energy consumption to a theoretical stoechiometric ratio for
ArcelorMittal Differdange and European furnaces [4]

110
2.2.6.3 Optimisation of the injection pattern in order to decrease energy consumption and iron losses
with the help of the dynamic metallurgical model (T 6.3)
After the revamping of the furnace a new preheating pattern has been implemented in AM Differdange.
Figure 156 shows the new preheating practice. In order to implement a preheating in the operation
pattern; a waiting time of 8 minutes had to be added to allow the load of the second basket, between the
melting of the first and the second basket. The use of preheating allows using less electrical energy that
is partly “replaced” by fossil energy which is less expensive than electrical energy (40%
approximately).

Power On 

Figure 156: New preheating practice in ArcelorMittal Differdange. B indicates when the burners are
working, PC when there is a post-combustion, RF a refining, F1 is the first furnace and F2
is the second one
With this new preheating mode, further modifications in the operation pattern of the electric arc furnace
number 2 where gas analysis was active as illustrated in Table 13.
Table 13: Modifications of the operation pattern made by the EAF engineer
Operating pattern Number of heats Modifications
1 33 Beginning of the tests with increasing of O2 supply by RCB lances

2 Limiting oxygen during preheating with the aim of decreasing the oxidation
2
14 Decrease of fossil energy at the first basket

3 33 Increase of post combustion at the first basket

Decrease of burners energy at the first basket

2
4 Decrease of refining oxygen at the second basket
16 Little alteration of refining oxygen at the second basket

Modification of refining oxygen. They tried to decrease the oxygen injected

5 57
6 99
7 33
by RCB lances because the steel is too much oxydated at this stage
Decrease of post combustion for the RCB burners
Increase of post combustion for the burners 2 and 5 (PC lances) (1)
Increase of post combustion at the first basket
Remove of the post combustion by RCB burners (2)
Reincrease of the refining oxygen: there is no more oxydation problems

(1) Increasing of post combustion for the PC lances 2 and 5 do not compensate the decrease at the RCB
burners. Then, it is needed to increase post combustion at the first basket.
(2) Post combustion by the RCB burners is not efficient and causes problems for the ladle metallurgy coming
later.

111
As explained in the Table 13, enhanced post combustion was tested with lances of the RCB burners and
PC lances.
RCB burners have an inclination of 50° while PC lances 2 and 5 have an inclination of 28°.
Furthermore, the RCB burners are closer of steel bath than PC lances 2 and 5. This could explain the
fact that post combustion by lances of the RCB causes over-oxidation.
On Figure 157, we can see the oxygen injected by the lances of the RCB burners (for post combustion
and refining) and the oxygen of the PC lances. The oxygen injected by the lances of RCB burners
decreases strongly between the operation pattern 3 and 4. Over-oxidation was avoided by decreasing the
oxygen injected by lances of RCB burners. Between operation patterns 4 and 6, the oxygen injected for
post-combustion was progressively transferred of the lances of the RCB burners to the PC lances.

O2 injected for post‐combustion and refining
4500
4000
3500
3000 O2 of PC lances
2500
Nm3

O2 refining and PC by 
2000 lances of RCB burners
1500 O2 burners
1000 CH4 burners
500
0
0 1 2 3 4 5 6 7 8
Operation patterns
Figure 157: O2 injected in the EAF with the lances of the RCB burners and the PC lances
As a reminder, the aim of preheating was to use less electric energy and to “replace” it by fossil energy
that is less expensive. As most of the CO is formed during the preheating, less post combustion oxygen
is necessary at the first basket and the electric energy necessary for melting is lower.
A preliminary study has been made in two parts. First part, analyzing preheating effects and the second
part studying the relative influence of the different operating patterns (As a reminder, the operation
patterns are described on the Table 13).

Figure 158: Cost of O2, CH4 and electrical energy for furnace 1 and 2 in function of preheating
Figure 158 show the cost evolution by ton of scrap according to the fossil energy injected during the
preheating (preheating = preheat phase of the first basket without electrical energy). The cost of the
total oxygen injected and the cost of the methane injected (burners) increase similarly and in the same

112
proportions. On the other hand, it is clear that when the fossil energy injected increases the cost of
electrical energy decreases. In this study, standard costs were used, the costs must be taken in relative
and should not be compared to other European plants.
It is clear that when the preheating energy increases, the electrical energy necessary for the melting
decreases. Nevertheless, when we look at the total energy cost (Figure 159), there is no significant
effect of the preheating. It is even observe an increase of the energy cost with the preheating energy for
the first furnace. This is not the case for second furnace (possible effect of the post combustion
efficiency?). However, by the lower electrode consumption, some additional benefits could be gained.

Figure 159: Energy cost for furnace 1 and 2 in function of preheating

Regarding to the effect of the operation patterns on averages of the energy cost for each operation
pattern, the oxygen injected in the furnace 2 had a direct effect on the energy cost. Figure 160 shows
the averages of the energy cost and of the total oxygen injected by lances of RCB burners (for post-
combustion and refining) for each operation pattern. These results indicate:
 For operation pattern 3, there is a minimum of the energy cost and a maximum for volume of
oxygen injected.
 For operation pattern 7, there is a minimum of oxygen injected and a maximum of the costs.

Figure 160: Averages of energy cost and oxygen injected by lances of RCB burners (for post-
combustion and refining) for each operation pattern for furnace 2
The relationship between both parameters is clear. We have lower energy cost for higher ratio of
oxygen injected by the lances of RCB burners. The potential benefits can turn around 1 Euros per Ton
of scrap.
Figure 161 shows the averages of energy cost and oxygen injected by PC lances for each operation
pattern. It is clear that when oxygen injected by PC lances increases, the energy cost increases
proportionally. This result is discussed further with the analysis of post-combustion.

113
Figure 161: Averages of energy cost and oxygen injected by PC lances (for post-combustion) for each
operation pattern for furnace 2
Simulations were performed with the EAF model of the CRM by reproducing the different operating
patterns performed on the plant. The post combustion and the scrap composition were supposed
constant and a typical heat was selected to perform simulations. Note that the inputs of oxygen
correlated to operating patterns have been multiplied by a factor 1.5 in order to clearly see the effect of
the oxygen input variations on the energetic cost. By adapting the input of electrical energy to get the
same tap temperature for each operating pattern an evaluation of the energetic cost was done. The
energetic costs are valid in relative and are not comparable with the real results of Figure 160 and
Figure 161. Figure 162 shows the total energetic cost per ton of steel produced, obtained with the
model. It is clear that the energetic cost decrease with oxygen injection by RCB burners and increase
with oxygen injection by PC lances as noted with the previous results.

Energetic cost obtained for each operation 
pattern with the EAF model
22
Energetic Cost (Euros/Tons (of 

21
scrap)*yield

20

19

18
0 1 2 3 4 5 6 7 8
Operating Pattern

Figure 162: Energetic cost per ton of steel produced obtained with the model for each operating
pattern
The post-combustion ratio has been calculated with the following formula:
CO 2
PC  . (58)
CO  CO 2
Figure 163 shows trends for post-combustion after avoiding heats with oxygen measured average
bigger than 15. It is clear that post-combustion increase between operation patterns 1 and 3. Post-
combustion for operation pattern 4 decreases because of oxidation problems as said before. For

114
operation patterns 4 to 6, we can observe a global increase of post-combustion due to the transfer of
oxygen injected from RCB burners to PC lances.

Data For O2 measured Average < 15 
(%w/s) Adjusted
1
PC ratio  adjusted (%) 0,8 4 2 4 4
0,6 20 5 1
0,4
0,2
0
0 2 4 6 8
Operation patterns

Figure 163: Trends of the PC ratios for each pattern when oxygen measured average (calculated with
electrical energy inferior to 35 MWh) is inferior to 15 (%w/s)
Various operating patterns have been tested while the gas analysis facilities were operating on furnace
number 2. The main conclusions are:
1) The preheating does not have any effect on the energy cost calculated (O2, CH4 and electrical
energy). Preheating decrease significantly the electrical energy consumption. The electrode
consumption and refractory lining wear should also be taken in account.
2) Additional Injection of oxygen with the lances of RCB burners decreases the energy cost while
injection of oxygen with PC lances increases the energy cost.
3) The post-combustion ratios calculated for each operation pattern are strongly related to the
additional injection of oxygen what make credible the ratios calculated. However, the post-
combustion ratios are not directly related to the energy cost. We explain these phenomena by
the fact that, following the way we inject oxygen in the furnace, the energetic value of post-
combustion will not be the same. It means that the additional oxygen injected by the lances of
RCB burners have a better energetic yield than additional oxygen injected by PC lances.
These conclusions lead us to studied the possibility to increase the stoechimetric ratio of the burners
(RCB and classical) to get a smaller rate of oxygen and then avoid over oxidation. This is studied in the
following points.
2.2.6.4 Performance of simulation trials with the optimised injection pattern of post-combustion
oxygen lances, the off-gas measurements being active (T 6.4)
Simulations were performed with the EAF model of the CRM by reproducing the different operating
patterns performed on the plant (as explained in the precedent point). Two operating patterns were
added:
1) Operating pattern 0: taken during a period after the operating pattern 7 (52 heats) with a
standard practice.
2) Operating pattern 8: by projecting an operating pattern running with an over-
stoichiometry practice for the burners, also called modified practice.
By adapting the input of electrical energy to get the same tap temperature for each operating pattern an
evaluation of the energetic cost was done. Figure 164 shows that operating pattern 8 (over-stoichimetry
practice) has the smallest energetic cost. Note that energetic cost of all operating pattern are smaller
than the reference operating pattern (operating pattern 0).
The next campaign in AM Differdange was focused on the application of this new operating pattern
(number 8) in order to avoid the oxidation problems observed in the operating pattern 3.

115
Figure 164: Energetic cost per ton of steel obtained for each operating pattern with the EAF model of
the CRM
2.2.6.5 Evaluation of achieved energy savings (T 6.5)
Figure 165 shows the average energetic cost for each operating pattern. Note that the operation pattern
2 contains only 10 heats what is not representative. The other operation patterns contain at least 20
heats. It is clear that these results are highly correlated to energetic cost calculated with the model
(Figure 164). We can clearly see that savings cost of 1 euro’s can be obtained for the operating pattern
3 (over-stoichiometry) relating to operating pattern 0 (reference).

Average energetic cost for each operating 
pattern
27
Energetic  cost (euros) / Ton of 

26
25
24
scrap*yieled

23
22
21
20
19
18
0 1 2 3 4 5 6 7 8
Operation Scheme

Figure 165: Average energetic cost per ton of steel obtained for each operating pattern
2.2.7 Assessment of plant trials and development of EAF energy process model (WP 7)
2.2.7.1 Correlation of off-gas energies for stainless, special and carbon steel heats with electric
energy demand (T 7.1)
The dynamic process model for mass and energy balance of BFI as described in detail under task 4.2 is
a suitable tool to examine the off-gas energies by using the input data from the different measurement
systems at the Marienhütte furnace. The measured values and the cyclic input data (from the PLC
controllers) allow the description of the energetic state of the heat during the whole time of the melt
down process. Due to the fact, that all relevant process values are calculated at any time, the EAF
process can be divided into different phases, which can be examined separately. For example the melt
down phases after charging of a basket, or the phase of refining can be investigated separately and
compared for different trial campaigns and evaluation purposes.

116
Based on the CO concentration measured by the LINDARC system and the off-gas temperature derived
from cooling water data at the elbow of the off-gas channel, the off-gas losses of a heat can be recorded
at the Marienhütte furnace for each heat. An evaluation was performed to investigate the effect of the
off-gas losses on the energetic performance of the furnace, as well as operating parameters influencing
the off-gas losses. This was done for about 300 heats performed from September to October 2009. The
data were collected by the on-line model, which was implemented at Marienhütte in March 2009, see
under task 8.3.
In the first step of this investigation it was clarified that the specific off-gas energy losses depend on the
specific electrical energy consumption. Figure 166 shows the increase of thermal and chemical energy
losses with the consumption of electrical energy.
160
Thermal off-gas losses
140 Chemical off-gas losses
120
Off-gas losses in kWh/t

100

80

60

40

20

0
280 300 320 340 360 380 400 420 440
Electrical energy consumption in kWh/t

Figure 166: Thermal and chemical off-gas losses depending on specific electrical energy consumption
80
Thermal off-gas losses
60 Chemical off-gas losses
Model deviation in kWh/t

40

20

-20

-40

-60

-80
0 20 40 60 80 100 120 140 160
Off-gas losses in kWh/t

Figure 167: Deviation of model for electrical energy demand depending on off-gas losses for MH
heats
In Figure 167 the model deviation of the BFI statistical energy balance model described under task 1.1,
which is a measure for the energetic performance of the furnace, is plotted against the off-gas losses. It
can be clearly seen that the energetic performance of the furnace is deteriorated with increasing
chemical and thermal off-gas losses.

117
  
80

60

Model deviation in kWh/t 40

20

Mean values for share

-20
thermal off-gas losses

-40
26.0 % 1. basket
21.8 % 2. basket
28.9 %
-60 3. basket
23.3 %
Refining
-80
0 10 20 30 40 50 60
Relative thermal off-gas losses in %

Figure 168: Model deviation against thermal off-gas losses for different phases of MH heats
 
80

1. basket
60
2. basket
3. basket
40
Refining
Model deviation in kWh/t

20

-20 Mean values for share

chemical off-gas losses

-40 27.3 %
22.0 %
-60 31.7 %
18.9 %
-80
0 10 20 30 40 50 60 70
Relative chemical off-gas losses in %

Figure 169: Model deviation against chemical off-gas losses for different phases of MH heats
For a more detailed investigation, the evaluated heats were split into 4 phases, 3 phases of melt down of
a charged basket and the refining phase. In Figure 168 the model deviation of the whole heat is plotted
against the share of thermal off-gas losses for the 3 phases of melt down of each basket and for the
phase of refining. The significant negative correlation between the off-gas losses and the performance
of the furnace during the refining phase indicates the importance of the refining phase.
The mean share of the thermal off-gas losses during the refining phase is 23.3 % for the investigated
heats. For the heats with the best performance this share is below 20 %. This means that more than
80 % of the thermal off-gas losses are generated during the melt down process of the 3 baskets and the
largest benefit can be reached when most of these losses accrue during the melt down phase of basket 3.
To improve the energetic performance of the furnace it is thus favourable to shorten the refining phase
by adjusting the aim tapping temperature and analysis without any significant delay and to charge the
baskets as soon as possible. A helpful tool to achieve this is the on-line model of BFI, which helps the
operator to meet the aim tapping temperature more exactly without delays e.g. by additional
temperature measurements and to decide, when the next basket can be charged.

118
In Figure 169 the model deviation is plotted against the share of chemical off-gas losses for the 4
phases. The significant negative influence of the off-gas losses on the performance of the furnace is
similar to the one of the thermal off-gas losses.
150 300
Off-gas enthalpy
Tap-to-tap time
Off-gas enthalpy at Point B [kWh/t]

Tap-to-tap time [min]

100 200

50 100

0 0
350 400 450 500 550 600
Electrical energy input [kWh/t]
Figure 170: Electrical energy input vs. off-gas enthalpy at Point B and tap-to-tap time (TKN –
stainless steel)
150 150
Off-gas enthalpy
Tap-to-tap time
Off-gas enthalpy at Point B [kWh/t]

Tap-to-tap time [min]

100 100

50 50

0 0
350 400 450 500 550 600
Electrical energy input [kWh/t]
Figure 171: Electrical energy input vs. off-gas enthalpy at Point B and tap-to-tap time (DEW – carbon
steel)
During stainless steel making at TKN Bochum, the electrical energy input correlates with the off-gas
enthalpy at point B and tap-to-tap time (Figure 170). In contrast, the electrical energy input during
stainless steel making at DEW does not strongly correlate with the tap-to-tap time, although the
electrical energy input shows a strong correlation with the off-gas enthalpy (Figure 171). Meanwhile,
higher energy loss through off-gas occurs if tap-to-tap time increases (Figure 172). If melting time
increases while the other energy sinks, e.g. cooling water and off-gas, are constant per unit time, the
energy losses both for carbon steel heats and stainless steel heats also therefore increase with higher
tap-to-tap time. To achieve the appropriate oxygen amount, tap-to-tap time will be longer due to limited
volume flow rate of oxygen lance. However, a method to decrease tap-to-tap time by simultaneous
oxygen injection with electrical energy input should be paid attention since it may cause high metal loss
(see WP 3.4). Use of multiple numbers of oxygen lances at various positions is suggested to decrease
tap-to-tap time since it will decrease the ratio of hot reaction zone and cold steel zone (see WP 3.4).

119
 120
Carbon steel (DEW)

Off-gas enthalpy at Point B [kWh/t]

100 Stainless steel (DEW)

80

60

40

20

0
0 50 100 150
Tap-to-tap time [minutes]
Figure 172: Tap-to-tap time vs. off-gas enthalpy at Point B
2.2.7.2 Influence of oxygen lancing and injection strategy to off-gas energy and electric energy
demand (T 7.2)
At the Marienhütte furnace a trial campaign with different door lance oxygen operating modes was
performed in January 2008. To minimise the influence of different scrap qualities, the trials were
performed alternating between the standard and an extended use of the door lance.
Table 14: Input data for BFI statistical model of campaign with different operating modes for the door
oxygen lance
Variable WR WE GA gE g Scrap gZ TA tS tN MG MO2G ML MLJ2 MN MCoke
Unit kWh/t kWh/t t kg/t kg/t kg/t °C min min m3/t m3/t 3
m /t
3
m /t m /t
3
kg/t
Campaign 15.- 18.01.08
95 heats 363.3 385.1 34.1 1226 1213 37.4 1624 28.0 10.6 7.9 15.8 8.0 14.9 8.4 9.8
Standard / Ext. Lance O2
Campaign 15.- 18.01.08
47 heats with Stand. Lance
368.6 391.9 34.1 1231 1219 37.4 1627 28.4 10.8 8.0 16 7.0 15.4 8.3 9.8

Campaign 15.- 18.01.08

48 heats with Ext. Lance
358.0 378.4 34.2 1220 1208 37.4 1621 27.6 10.4 7.8 15.6 9.0 14.5 8.5 9.7
Difference between two
campaings
-13.5 0.1 -11.0 -11.0 0.0 -6.0 -0.8 -0.4 -0.2 -0.4 2.0 -0.9 0.2 -0.1
Difference in electrical
energy demand in kW h/t
-4.4 0.0 -1.8 -0.8 -0.4 1.6 3.9 -0.6 -
Corrected difference in
electrical energy consump.
-11.0 5.5

The extended use means, that the oxygen input was increased by nearly 30 %. The results of evaluations
with the statistical BFI model for this campaign are shown in Table 14. The difference in electrical
energy demand between these two operating modes is 13.5 kWh/t. However, for comparison of the total
effective energy input by the additional lance oxygen, it is necessary to correct this value, because the
other relevant average input parameters of the two trial variations differ. According to the equation of
the statistical model, the difference in electrical energy demand caused by the different input parameters
is shown in row five of the table. The corrected difference in electrical energy consumption is
11.0 kWh/t. Dividing this effective saving by the additional door lance oxygen consumption of 2.0 m3/t
results in an energetic efficiency of door lance oxygen of 5.5 kWh/m3. This value is higher than
expected within the energy demand formula, where it is set to 4.3 kWh/m3. In Figure 173, the electrical
energy demand calculated according to the BFI statistical model is plotted against the electrical energy
consumption for this campaign.

120
 500
Extended Standard
door lance door lance

calculated electrical energy demand in kWh/t

Mean deviation: -20.3 kWh/t -23.3 kWh/t
Standard dev.: 33.0 kWh/t 31.3 kWh/t
450

400

350

300

250
250 300 350 400 450 500
actual electrical energy consumption in kWh/t

Figure 173: Calculated vs. actual electrical energy consumption with extended and standard door
lance operation
The examination of the influence of oxygen lancing and injection on the off-gas energy was performed
with the data collected from September to October 2009. In Figure 174 the correlation between the
thermal and chemical off-gas losses are plotted against the specific oxygen injection by the 3 KT-
injectors. Especially the thermal off-gas losses clearly increase with higher oxygen injection. A similar
behaviour can be found for the oxygen lancing by the door lance.

160
Thermal off-gas losses
140 Chemical off-gas losses
Off-gas losses in kWh/t

120

100

80

60

40

20

0
0 5 10 15 20
3
Oxygen injection in Nm /t

Figure 174: Thermal and chemical off-gas losses according to specific oxygen injection

121
  
1500
Total TKN Bochum
Electrical
Chemical

Energy input [kWh/t]

1000

500

0
0 5 10 15
Oxygen injection [kg O2/t]
Figure 175: Specific oxygen injection vs. chemical, electrical, and total energy inputs for TKN
Bochum
1200
Carbon steel DEW Siegen
Stainless steel
1000
Specific total energy input [kWh/t]

800

600

400

200

0
0 10 20 30 40 50
Oxygen injection [kg O2/t]

Figure 176: Specific oxygen injection vs. specific total energy input of carbon and stainless steels for
DEW Siegen
The off-gas enthalpy at point B at TKN Bochum does not strongly correlate with oxygen injection
(Figure 175). This might be due to air infiltration and cooling-water spraying between point A and
point B at TKN Bochum which changes the off-gas enthalpy at Point B. On the other hand, both carbon
and stainless steel heats at DEW Siegen show positive correlations between oxygen injection and off-
gas enthalpy at point B (Figure 176). At DEW Siegen, there is no cooling-water spraying between
point A and point B.
For DEW Siegen, both carbon and stainless steel heats show positive correlations between oxygen
injection and off-gas enthalpy at point B (Figure 177). Meanwhile, the off-gas enthalpy at point B at
TKN Bochum does not strongly correlate with oxygen injection (Figure 178). This might again be due
to air infiltration, and cooling-water spraying between point A and point B at TKN Bochum which
changes the off-gas enthalpy at Point B. At TKN Bochum, the specific off-gas enthalpy significantly
decreased when a lime of instead of limestone and when lower amount of oxygen injection are applied.
Limestone decomposition is a significant endothermic process which decreases the temperature of
liquid steel. Meanwhile, the composition of tapped steel at TKN Bochum did not significantly change
with lower amount of oxygen injection (Table 15).

122
Table 15: EAF characteristics when using lime instead of limestone
Electrical energy O2 injection Coal input Element concentration at tapping [wt%]
Case
[kWh/t] [kg/t] [kg/t] C Si Mn Cr
Limestone 470 12.7 2.6 1.51 0.13 0.33 16.10
Lime 460 7.5 1.1 1.52 0.23 0.53 17.36

120
Carbon steel DEW Siegen
Off-gas enthalpy at Point B [kWh/t]

Stainless steel
100

80

60

40

20

0
0 10 20 30 40 50
Oxygen injection [kg O2/t]

Figure 177: Specific oxygen injection vs. off-gas enthalpy at point B of carbon and stainless steels for
DEW Siegen
120
TKN Bochum
Specific off-gas energy at Point B [kWh/t]

90

60

30

Use limestone
Use lime
0
0 10 20 30
Specific oxygen injection [kg O2/t]
Figure 178: Specific oxygen injection and off-gas enthalpy at point B of carbon and stainless steels for
TKN Bochum
In parallel, results of trial campaigns performed at the Twin shell AC furnace of AM Differdange,
which was also described under Task 6.2, were evaluated with the help of the BFI model to calculate
the electrical energy demand. At this furnace the standard mode for post-combustion oxygen input was
varied, increasing the oxygen flow rate and extending the duration of PC oxygen input. In Table 16 the
average input data for the BFI model are shown for the heats operated under the two different PC
modes.

123
Table 16: Input data for BFI statistical model of campaigns at AM Differdange furnace with different
operating modes for post-combustion oxygen

As can be seen the specific consumption of PC oxygen, here assigned with the variable MN is slightly
increased for the modified PC control mode. However, at the same time the specific oxygen input by
injectors MI was increased by 1.6 Nm³/t, and the burner gas input MG by 1.1 Nm³/t, whereas the other
influence parameters stayed more or less constant. Due to these measures the electrical energy
consumption WE slightly decreased, but not as much as would have been expected due to the increased
chemical energy input. Thus the difference between calculated electrical energy demand WR and actual
consumption WE has a higher negative value for the modified operating practice. This means that the
energetic performance of the furnace was deteriorated by this modification.
2.2.7.3 Influence of burner to off-gas energy and electric energy demand (T 7.3)
In November 2007 a trial campaign with different operating modes of the door burner was performed at
the MH furnace. The trials were performed alternating between the standard and an extended use of the
door burner.
500
Extended Standard
Door burner Door burner
calculated electrical energy demand in kWh/t

Mean error: -22.1 kWh/t -16.9 kWh/t

Standard dev.: 24.6 kWh/t 26.0 kWh/t
450

400

350

300

250
250 300 350 400 450 500
actual electrical energy consumption in kWh/t
Figure 179: Calculated vs. actual electrical energy consumption with extended and standard use of the
door burner
The extended use of the door burner means, that the input of natural gas was increased by nearly 20 %.
The results of calculations with the statistical BFI model for the electrical energy demand are shown in
Table 17. The electrical energy demand for the extended mode is about 4.2 kWh/t smaller than for the
standard mode. As described in Task 7.2 it is necessary to correct the difference because of the different
input parameters. The effective difference in electrical energy consumption is 5.3 kWh/t. Dividing it by

124
the additional door burner natural gas consumption of 1.3 m3/t, this results in an effective efficiency of
door burner natural gas of 4.1 kWh/m3. This value is significantly lower than expected within the
energy demand formula where it is set to 8 kWh/m3. This means that an extended use of the door burner
decreases the electrical energy consumption, but the energetic efficiency of the burner gas is lower than
in the main melt down period. In Figure 179, the electrical energy demand calculated according to the
BFI statistical model is plotted against the electrical energy consumption for this campaign.
Table 17: Input data for BFI statistical model of campaign with different operating modes for the gas
burners
Variable WR WE GA gE g Scrap gZ TA tS tN MG MO2G ML MLJ2 MN MCoke
Unit kWh/t kWh/t t kg/t kg/t kg/t °C min min m3/t m3/t 3
m /t
3
m /t
3
m /t kg/t
Campaign 12.-18.11.07
173 heats all
363.7 383.1 35.1 1198 1187 35.9 1632 28.5 12.2 7.2 14.3 7.6 14.3 9.0 11.9
Campaign 12.-18.11.07
93 heats Standard Door 368.0 385.0 35.2 1196 1185 35.9 1631 28.9 12.6 6.6 13.1 7.7 14.3 8.9 11.8
Burner
Campaign 12.-18.11.07
80 heats Extended Door 358.7 380.8 35.0 1201 1190 35.9 1632 28.0 11.6 7.9 15.8 7.4 14.3 9.1 12.1
Burner
Difference between two
campaings
-4.2 -0.2 5.0 5.0 0.0 1.0 -0.9 -1.0 1.3 2.7 -0.3 0.0 0.2 0.3

Difference in electrical
energy demand in kW h/t
2.0 0.0 0.3 -0.9 -1.0 1.3 0.0 -0.6 -

Corrected difference in
electrical energy consump.
-5.3 4.1

In Figure 180 the correlation between the thermal and chemical off-gas losses are plotted against the
specific natural gas input by the 3 KT-burners. The off-gas losses increase with higher natural gas
inputs.

160
Thermal off-gas losses
140 Chemical off-gas losses
120
Off-gas losses in kWh/t

100

80

60

40

20

0
5 5.5 6 6.5 7 7.5 8 8.5 9 9.5
3
Gas burner in Nm /t

Figure 180: Thermal and chemical off-gas losses according to specific natural gas input

125
 80
1. basket
60 2. basket
3. basket
Model deviation in kWh/t
40
Refining
20

-20

-40

-60

-80
0 0.5 1 1.5 2 2.5 3 3.5 4
3
Gas burner in Nm /t

Figure 181: Model deviation against specific natural gas input for different phases of MH heats
In Figure 181 the model deviation is plotted against the specific use of the gas burners for the 4
different phases of the melt down process. During the melt down of the 3 baskets the use of the gas
burners is more extensive than in the refining phase. The negative correlation during the refining phase
indicates the low efficiency of the burners when the scrap is already molten down. However, in this
phase most of the gas consumption is explained by the stand by operation mode of the burners.

550
Previous new preheating
practice
calculated electrical energy demand in kWh/t

Mean Error: -56.8 4.4 kWh/t

Standard dev..: 28.8 32.7 kWh/t
500

450

400

350

New preheating practice

Previous practice
300
300 350 400 450 500 550
actual electrical energy consumption in kWh/t

Figure 182: Comparison between calculated electrical energy demand and actual consumption for
previous and new preheating practice at the AM Differdange furnace
As described in Task 6.3, the Twin shell AC furnace of AM Differdange moved to a new preheating
operating practice. The two situations were evaluated with the help of the BFI model to calculate the
electrical energy demand. As illustrated in Figure 182, the difference between calculated electrical
energy demand WR and actual consumption WE has a positive value for the new preheating practice
compare to the previous one. This means that the energetic performance of the furnace was significantly
improved by this modification.

126
2.2.7.4 Influence of dedusting system to electric energy demand (T 7.4)

Figure 183: Schematic display of the off-gas flow

The volume flow rate of primary dedusting system affects directly the EAF energy balance by
influencing the mean off-gas retention time inside the EAF vessel ∆t (eq. (59)) and the volume flow rate
of the infiltrated air  V FL . The mean off-gas retention time is a function of the free-board volume (VEAF),
the normalized off-gas volume flow rate in the point A  V N 21  , the gap width between elbow and the off-
gas primary dedusting system (∆Sp), the diameter of the primary dedusting system (D), and the off-gas
temperature leaving the EAF vessel (T20). A corrective factor (C) needs to be applied. The different
fluid flows inside furnace and primary dedusting system can be seen in Figure 183.

VEAF 298.15 K  1  Sp T20 

t  . 1   (59)
V N 21 T20  C D TSp 
 
Meanwhile, minimization of the infiltrated air into the EAF can be achieved by:
 increasing the tightness of the EAF vessel (e.g. by closing the slag door and by sealing off all
EAF gaps),
 dynamically adjusting the gap width between the EAF elbow and the primary dedusting system
which depends on the EAF period,
 adjusting the off-gas volume flow rate of the primary dedusting system through a dynamic or a
static DEC control.
On the other hand, the off-gas volume flow rate of the secondary dedusting system should be increased
in order to compensate the decreasing off-gas volume flow rate of the primary dedusting system.
In 5 measurement campaigns at TKN heats were measured and evaluated (Table 18). In the campaign
1, the EAF was in actual operation without any improvement methods. Thus, the EAF characteristics in
campaign 1 serve as the reference values for the following four measurement campaigns. The strategies
to increase the EAF energy efficiency, e.g. a decrease of air infiltration and a decrease of off-gas
volume flow rate, were performed in the measurement campaign 2 until 5. Due to the installation of
measurement lances, the minimum gap width between the EAF elbow and the primary dedusting
system could not be changed during all measurement campaigns and remained constant at 1.5 m.
The static DEC control of the primary dedusting system was applied in the measurement campaign 1 to
4 and the DEC was dynamically controlled only in the measurement campaign 5. The evaluation of 5
measurement campaigns shows a clear influence of the different operation modes of the primary off-gas
dedusting system on the off-gas mass, the off-gas enthalpy and the electrical energy consumption
(Table 19 and Figure 184). Compared to the initial off-gas mass in the measurement campaign 1
(average value 939 kg/t), the measurement campaign 2 and 3 demonstrated a decrease of the off-gas
mass between 105 and 143 kg/t which is equal up to a 15 % decrease. The off-gas mass decreased even
more significantly by 44 % in the measurement campaign 5 when a dynamic DEC control was applied.

127
Table 18: Various EAF operation modes in each measurement campaign
Measurement Method to decrease air Method to increase
DEC control
campaign infiltration chemical energy efficiency
Campaign 1 Static DEC, By- Nothing Nothing
(initial pass valve 50 %
campaign) open in the melt-
down period
Campaign 2 Similar with Sealing of the EAF Nothing
campaign 1 leakage:
between vessel and upper
shell
between elbow and upper
shell
Campaign 3 Similar with Similar with campaign 2 Nothing
campaign 1 plus slag door closed in
melt-down period
Campaign 4 Static DEC, By- Similar with campaign 3 Reduction of oxygen injection and
pass valve 100 coal input
% open in the Using lime instead of limestone
melt-down
Decreasing the off-gas volume
period
flow rate at primary dedusting
system (compensated with higher
off-gas volume flow rate at the
secondary dedusting system)
Campaign 5 Dynamic DEC Similar with campaign 3 Nothing
Table 19: EAF and off-gas properties in each measurement campaign
O2 Electrical Off-gas properties (primary dedusting system)
Measurement
injection energy enthalpy
campaign temperature [°C] mass [kg/t]
[kg/t] [kWh/t] [kWh/t]
Campaign 1 7.9 494 (σ±21) 298 939 (σ±292) 82 (σ±20)
Campaign 2 8.3 477 (σ±24) 313 796 (σ±127) 69 (σ±8)
Campaign 3 8.5 474 (σ±31) 293 834 (σ±290) 67 (σ±16)
Campaign 4 5.5 459 (σ±24) 216 619 (σ±76) 36 (σ±6)
Campaign 5 8.9 459 (σ±27) 308 529 (σ±130) 48 (σ±13)
The lower off-gas volume rate at the primary dedusting system decreased directly the off-gas enthalpy
at the primary dedusting system. In the measurement campaign 4, the oxygen injection and the coal
addition were decreased which thus decreased the influence of the chemical energy on the off-gas
enthalpy. Although coke addition is necessary to form the foaming slag, the foaming slag cannot be
developed during stainless steel making. For this reason, coke addition was not further performed at
TKN. Since limestone decomposition is an endothermic reaction, replacing limestone by lime in the
measurement campaign 4 contributed also to a lower total energy demand. This could decrease the
electrical energy consumption. In the campaign 4, the amount of coal injection was decreased which
contributes also on the decreasing off-gas enthalpy. Besides that, a 100% opening of by-pass valve
contributes also to a significant decrease of off-gas mass and enthalpy. Compared to campaign 3, the
off-gas enthalpy in campaign 4 decreased significantly by 46% from 67 kWh/t down to 36 kWh/t. An
application of dynamic DEC control in campaign 5 delivered a promising result. The off-gas electrical
energy consumption in campaign 5 could be significantly decreased although there were no special

128
measures to increase the chemical energy efficiency. Compared to the campaign 1, the off-gas enthalpy
in the campaign 5 decreased by 41% from 82 kWh/t down to 48 kWh/t and the specific electrical
energy consumption fell by 7% from 494 kWh/t down to 459 kWh/t.
500 650
Off-gas enthalpy

Electrical energy consumption [kWh/t]

Electrical energy
400 550
Off-gas enthalpy [kWh/t]

494
477 474
459 459
300 450

Static DEC, 50 % by pass valve open

in melt-down period Dynamic DEC
200 350
Static DEC,100 % by pass valve
open in melt-down period

100 250
82 69 67
48
36

0 150
Campaign 1 Campaign 2 Campaign 3 Campaign 4 Campaign 5

Figure 184: The influence of different operating modes on the off-gas enthalpy and the electrical
energy consumption
2.2.7.5 Development of additional process adaptation in order to minimize electric energy demand
and oxygen consumption (T 7.5)
In June 2009 the tap weight at the Marienhütte furnace has been increased from about 35 t up to more
than 40 t. In Table 20 the consumption data are figured for a time period before, during and 3 months
after the increase of the tap weight. The average consumption of electrical energy decreased by about
13 kWh/t.
Table 20: Input data for BFI statistical model for three time periods in 2009
Variable WR WE GA gE gScrap gZ TA tS tN MG MO2G ML MLJ2 MN MCoke
Unit kWh/t kWh/t t kg/t kg/t kg/t °C min min m3/t m3/t 3
m /t 3
m /t m /t 3
kg/t
April 2009
159 heats
361.5 362.3 35.9 1144 1133 36.2 1631 27.4 10.1 7.3 16.5 8.1 9.2 5.8 8.9

May - June 2009

244 heats
350.8 355.0 37.9 1131 1120 35.9 1628 28.3 10.9 7.1 16.5 8.9 9.8 5.9 8.2

Sept. - Oct. 2009

314 heats
363.7 349.2 41.1 1177 1167 34.1 1636 29.9 13.0 6.9 16.8 9.3 10.5 7.5 7.4

A further obvious difference is the decrease of the consumption of blowing carbon. The correlation
between the deviation of the statistical model and the injected blowing carbon is plotted in Figure 185.
The model deviation increases with the amount of injected blowing carbon, which proves the
assumption of a larger electrical energy demand for this kind of carbon addition.
Up to now, there was no term to consider the effect of the charged or injected carbon in the statistical
energy balance model of BFI. However the consumption of blowing carbon also correlates with the
electrical energy input and the power on time, as more carbon is blown for heats where a longer
electrical energy input is needed. This fact makes it difficult to integrate the term of blowing carbon
into the statistical model in a consistent way without changing the other parameters of the model.

129
 80

60
Model deviation in kWh/t
40

20

-20

-40

-60

-80
0 2 4 6 8 10 12 14 16
Specific amount of blowing carbon in kg/t

Figure 185: Correlation between model deviation and the amount of blowing carbon
Closing of the Slag Door
The infiltrated air during the melt-down and finishing periods increases the energy loss, mainly through
nitrogen heating (Figure 186). In order to prevent a higher energy loss through the off-gas, the amount
of the infiltrated air into the EAF vessel must be minimized, e.g. by closing the slag door. At DEW, the
off-gas temperature at point A is up to 1 820 °C (average 234 °C). The total infiltrated air through the
opened slag and tapping door at DEW is between 2 and 5 m³ (STP)/s. Thus, the average energy loss due
to the nitrogen heating during stainless steel making at the investigated EAF is estimated between 4.7 to
11.8 kWh/t.
0.5
N2
O2.
0.4
Heating enthalpy [kWh/kg ]

0.3

0.2

0.1

0.0
0 500 1000 1500
Temperature [oC]

Figure 186: Heating enthalpy of nitrogen and oxygen (p = 1 atm)

Maintaining the Initial Temperature of the Refining Period
To increase the decarburization reaction with a minimum chromium oxidation, many works suggested
achieving a high temperature of liquid steel before the start of oxygen injection. Based on the empirical
model developed in this work, the liquid steel temperature before the refining period in the investigated
EAF is proposed to be kept in a minimum value since a higher temperature leads to a lower chromium

130
concentration in the liquid steel. Energy loss due to the convection between liquid steel with off-gas and
with cooling-water system may increase with a higher tap-to-tap time and a higher temperature
gradient. Energy loss of the EAF is defined through the heat loss via water-cooling system, Eloss,water-
cooling, off-gas, Eloss,off-gas, and radiation, Eloss,radiation, (eq. (60)).

E loss   E loss,cooling  water dt   E loss,off _ gas dt   E loss,radiation dt (60)

t t t

The homogeneous convection coefficient above the liquid steel is assumed at 2.4 W/(m2·K) [6] and the
EAF inner diameter of DEW is 6.98 m. Thus, an increase of the energy loss due to the convection heat
transfer between the liquid steel and the infiltrated air at DEW is at 917 W for each 10 °C increase of
the liquid steel temperature. Moreover, the heat loss through radiation increases with a high temperature
of liquid steel. By assuming the emissivity of the liquid steel is approximately 0.28 [7], an increase of
the liquid steel temperature by 10 °C will increase the heat loss through radiation by 3.8 kW.
Heat loss through water-cooling system increases with a higher liquid steel temperature as well. Energy
losses through water-cooling system, off-gas, and radiation are increasing as the integral of energy loss
over time is increasing. For this reason, a higher temperature of the liquid steel by increasing the
electrical energy input at DEW should be performed in the finishing period, not in the melting down
period. In contradiction, some works suggested maintaining a high temperature of liquid steel before the
refining process is necessary in order to increase decarburization and decrease chromium oxidation.
Transfer of a part of the PC oxygen to burners
As described in Task 6.4, in AM Differdange, an optimal operating pattern has been deduced from
several tentative modifications: additional oxygen volume has been input for post-combustion purposes
thanks to the burners working in over-stoichiometric mode. The two situations were evaluated with the
help of the BFI model to calculate the electrical energy demand. As illustrated in Figure 187, the
difference between calculated electrical energy demand WR and actual consumption WE has a slightly
higher positive value for the over-stoichiometric practice compare to the other tested operating patterns.
This means that the energetic performance of the furnace was further improved by this modification.

500
Model deviation operating pattern 1 to 8
Mean Error: 4.4 6.5 kWh/t
calculated electrical energy demand in kWh/t

Standard dev..: 32.7 29.3 kWh/t

450

400

350

1 to 7
8
300
300 350 400 450 500
actual electrical energy consumption in kWh/t

Figure 187: Comparison between calculated electrical energy demand and actual consumption for
each operating pattern at the AM Differdange furnace (cf. Table 24)

131
2.2.7.6 Development of an energy/mass model based on off-gas measurements for Consteel process
control (T 7.6)
To solve the EAF Gas-Phase mass and energy balance are used 4 different models (Figure 188):
 Chemistry Calibration Model [CCM]
 Wet Chemistry Model [WCM]
 Off-Gas flow rate Model [FRM]
 FreeBoard Model [FM]
The final objective of the four models is to calculate the energy, carbon and oxygen transferred to the
bath. These data constitute the input of the bath/slag model which simulates the process.

Figure 188: Scheme of the relationships among the models in order to simulate the whole EAF
Consteel® process and to build the control system.
The drawing shows a simplified scheme of the models interaction between them (including also the
bath and slag phase model) and with the plant.
All the models have already been described, here only the main concepts are reminded.
The Chemistry Calibration Model corrects the chemistry measured by the EFSOP (if needed because
of non perfect calibration or measurement not available due to the purging of the analysis system).

132
The Wet Chemistry Model calculates the water concentration in the off-gas and subsequently the
WET chemistry (CO, CO2, H2, O2, N2 and H2O) and takes as input the Off-gas chemistry from EFSOP,
Off-Gas Temperature and Pressure.
The Off-Gas Flow Rate Model calculates the Off-Gas flow rate. This model has been described
previously.
The Freeboard Model: calculates the Mass and Energy Balance of the Gas Phase and takes as input
the wet concentration, the Off-Gas temperature and Flow rate, O2 injection, Electrical Power and Panel
Losses.
The Bath and Slag: calculates the mass/energy balance in the molted phase. This model has been
described previously.
The system is coded to warn the operator both for safety and for process control.
The first level of warning is given by the probability of having stratification into the furnace. That
might to be due to an unbalanced charging operation - that could lead to an uncontrolled reaction into
the furnace – or the amount of not molten scrap. The alarm duration and the carbon content are warning
the operator of possible unsafe production condition.
A second level of warning is deal with the:
 yield of the iron - due to the charge oxidation,
 extra power supplied by post combustion in off-gases
Using proper thresholds, dependent by the operative production (steel grade, slag target, etc..), the
system warns the operator about excessive steel oxidation and the possibility of using post combustion
oxygen to improve the available power.
The scheme shows the general link between computed status of the furnace, thresholds and process
parameters.

Figure 189: Scheme of the logic of control system

The post combustion control has been based on the same concept described on the Task 5.5. The
innovative step done has been to relate the flow rate calculated whit the gas concentration measured.
Then a post combustion quantitative control has been implemented; dedicated thresholds have been
determined for different production operative mode. The good results gather about the global mass and
energy furnace balance secure a good flow rate estimation, then the post combustion control has been
based on robust data.

133
The real time simulation of the bath/slag status is the base for determining critical conditions occurrence
respect to:
 probability of charge/liquid pool stratification and possible uncontrolled reaction into the
furnace,
 deviation from target condition for carbon content of the melt,
 deviation from target condition for melt temperature,
 excessive oxidation of the melt (yield control)
Post combustion in the freeboard is also controlled to optimise heat transferring to the scrap or to the
bath.
Concerning the liquid phase carbon content, iron oxide amount and temperature (also its gradient) status
is continuously monitored and their values compared to thresholds (acceptability windows) to provide a
warning and/or to suggest a corrective action to the operator.
A safety level of warning is provided for avoiding bath stratification and uncontrolled reaction into the
furnace looking at the temperature status and at the carbon content.
In particular three level of probability for uncontrolled reaction occurrence have been implemented
(low, medium, high) following an ‘and’ logic for different conditions verified:
 if the cooling gradient of the bath exceed a defined threshold
 if the bath temperature calculated is outside a defined window
 if the bath carbon content calculated is outside of a defined window
While a condition is verified the alarm is “low”; while two conditions are verified at the same time is
“medium”; while all of them are verified it is “max”.
A second level of control, i.e. the process control, has been implemented coupling warning to
suggestion for managing devices/parameters as conveyor speed (charge feeding), oxygen mass flow
rate, coal addition and carbon injection.
Figure 190 resumes criteria for the carbon content and temperature control during the melting phase
changing (increasing or decreasing) the oxygen lancing and conveyor speed respectively.

Figure 190: Criteria followed by control system for carbon content and temperature, during charging,
acting on oxygen lancing
The refining phase (overheating) is led comparing the bath and slag condition respect to the end point
target of the heat (tap aim carbon and temperature). Figure 191 shows an example of an on line running
(HMI print screen – page implemented in english) with messages of C/T reached condition.

134
Figure 191: Print screen of the developed interface to follow the bath status
Figure 192 shows the criterion for the iron oxide control, i.e the yield control, used in particular in the
case of high quality scrap charging. A warning is provided if FeO amount (mass in kg or %FeO) target
is exceeded, where the latter is defined (continuously) as a function of the carbon content.

Figure 192: Criterion for iron oxide control

The technologist through a suitable parameters table, as showed in Figure 193, sets parameters and
thresholds in the HMI (page implemented in Italian kindness to the operator).
Following the table form top to bottom rows allow setting of
 time frequency for evaluation,
 cooling rate threshold for the bath,
 target temperature,
 target carbon content,
 upper and lower tolerance for temperature and carbon (four lines),
 kg of FeO not to be exceeded (if equals to zero the %FeO criteria is taken in account),
 coefficients for linking %FeO threshold respect to carbon content of the bath (two lines).

135
Figure 193: Parameters table to set the following parameters: time frequency for evaluation, cooling
rate threshold for the bath, target temperature, target carbon content, upper and lower
tolerance for temperature and carbon (four lines), kg of FeO not to be exceeded
Several warning messages have been implemented to advice operator, which appearance has priority
over all windows of the HMI. Figure 194 shows all messages available (see the labels)
Warning about bath stratification and possible uncontrolled reaction
 Group (increase/decrease) for managing carbon content (higher or lower the window) by
oxygen lancing or coal addition
 Group (increase/decrease) for managing temperature of the bath by conveyor speed (scrap
feeding rate)
 Warning about FeO exceeding (coal addition for reduction as suggestion)
Because of the big advantages demonstrated by the post combustion managing, explained in the task
5.5, the PC dynamic control has been left as implemented at the first stage.

136
 1

4
2
3

Figure 194: Print screen of the warning messages for the operator
2.2.8 Final plant trials with adapted process models and assessment (WP 8)
2.2.8.1 Plant trials and assessment at DEWG (T 8.1)

Figure 195: Monitor display of the off-gas compositions used during the plant trial
A permanent off-gas analysing system was adopted during the plant trials. The lances for O2, CO, and
CO2 measurement were installed about 5 m from the EAF elbow. Such installation was chosen since the
available off-gas analyser had been already installed near the EAF vessel. The result of the off-gas
analysis indicates an almost complete CO post combustion at this measurement point which is shown
by a very low CO concentration in the off-gas, i.e. less than 1 %. The measurement of the off-gas

137
volume flow rate is unnecessary for the investigated EAF since there is no additional carbon input
during the refining period. The end of the efficient oxygen injection is thus merely determined when the
efficient decarburization is finished. After the off-gas evaluation in the off-gas analyser, the off-gas
composition is shown in a monitor display which is located in the operator room (Figure 195) so that
the EAF operator can precisely determine when the oxygen injection must be ended.
In order to quantify the change of the oxygen injection, the actual amount of the oxygen injection is
compared with the initial set value which had been already determined before the refining period started
(eq. (61) and Figure 196).
m O2 ,actual  mO2 ,set
 O2 injection  100% (61)
mO2 ,set

6
15 stainless steel heats

5
Number of heats

4 Equal to a decrease of tap-

to-tap time by about 8
3 minutes

0
No decrease 0 - 10 10 - 20 20 - 30
Decrease of the oxygen injection [%]

Figure 196: Decrease of the oxygen injection during the plant trial phase 1 and phase 2
The plant trials delivered a successful result. The amount of oxygen injection can decrease from the
initial set value. On the other hand, the carbon concentration in the liquid steel can be maintained in a
low value as well. Since the oxygen injection during plant trials was finished at the 2nd critical point, the
critical carbon concentration in the liquid steel may exist between 0.47 and 0.8 wt%C, even though it
may not represent the precise values due to the change of steel composition after the slag reduction by
ferro-silicon and due to alloy addition in the finishing period.
During the plant trial phase 1, the mass of ferro-silicon input was intensely decreased since the amount
of the chromium and iron oxides in the slags were expected to be lower. This was carried out in order to
prevent a very high silicon concentration in the tapped steel, i.e. above 0.15 wt%, due to the unreacted
excess silicon with the chromium and iron oxides. A very high silicon concentration in the tapped steel
is harmful for the next process after the EAF, i.e. the VOD operation, since the silica as the product of
the silicon oxidation will accelerate the erosion of VOD refractory. As previously expected, the
decrease of the oxygen injection results in a very low silicon concentration in the tapped steel since the
silicon oxidation is expected to take place in early stage of oxygen injection. For this reason, the amount
of ferro-silicon was given as a nominal value, i.e. in the similar amount with the set value, during the
plant trial phase 2. The oxygen consumption in the plant trial phase 2 decreased up to 24 % (average
10 % decrease) compared with the set value which corresponds to the decrease of the oxygen injection
time and tap-to-tap time up to 8 minutes.
As previously expected, the oxygen injection until the 2nd critical point was confirmed to increase the
iron and chromium yield (Table 21, Table 22 and Table 23). Thermodynamically, the intensive
chromium oxidation takes place after the intensive silicon oxidation finishes. For this reason, the
chromium concentration in the slag is still high even though the oxygen injection was finished at the 2nd

138
critical point. When the amount of ferro-silicon was decreased during the plant trial phase 1, only the
amount of the iron oxide could be significantly decreased but the decrease of chromium oxide was less
significant. The more negative Gibbs energy of iron reduction by silicon contributes to a more effective
decrease of FeO concentration compared than the Cr2O3 concentration. A more significant decrease of
Cr2O3 mass in slag was achieved when the amount of ferro-silicon was given as its nominal value
during the plant trial phase 2. The decarburization control by using the off-gas analysing system
decreases the slag mass by 7.2 % from 159.3 kg/t down to 147.8 kg/t. This benefits the stainless steel
making process in the EAF. There is an additional economic saving since less slag needs to be treated
and deposited. Less CaO input can be achieved due to lower Cr2O3 and FeO content in the slag. In the
end, an additional energy saving can be obtained due to less energy loss through slag. Nevertheless,
lower CaO concentrations in the slag are necessary in order to maintain high Cr2O3 activity in the slag
for achieving effective Cr2O3 reduction by ferro-silicon.
Table 21: Result of the slag analysis after the steel and slag tapping (with the decarburization control,
plant trial phase 1)
The elements concentration in the slag [wt%] Basicity
CID
Al2O3 CaO FeO MgO MnO SiO2 TiO2 Cr2O3 (CaO/SiO2)
213095 3.39 44.75 1.69 7.52 5.00 24.58 0.81 11.38 1.82
213097 3.63 42.33 0.82 6.53 5.24 30.25 0.98 7.42 1.40
213099 4.35 42.81 0.47 7.37 4.03 29.84 1.04 7.07 1.43
213103 16.16 37.97 0.91 7.66 1.66 30.65 1.31 2.47 1.24
213123 3.50 44.28 0.95 7.97 1.84 31.04 1.03 6.52 1.43
Mean 6.21 42.43 0.97 7.41 3.55 29.27 1.03 6.97 1.46
Table 22: Result of the slag analysis after the steel and slag tapping (with the decarburization control,
plant trial phase 2)
The element concentration in the slag [wt%] Basicity
CID
Al2O3 CaO FeO MgO MnO SiO2 TiO2 Cr2O3 (CaO/SiO2)
214091 6.48 46.94 0.79 6.89 1.13 33.63 1.05 2.00 1.40
214093 5.38 47.26 0.47 7.33 2.03 32.92 1.28 3.12 1.44
214095 5.12 47.31 0.49 7.82 0.90 33.02 1.29 3.60 1.43
214097 4.67 41.56 0.71 7.93 2.02 34.73 1.33 6.63 1.20
214099 7.09 40.21 1.71 7.93 2.02 34.73 0.74 8.35 1.16
214107 5.18 48.45 1.06 8.56 1.38 29.16 0.96 4.80 1.66
214109 10.56 49.83 0.90 7.40 0.89 25.37 0.82 2.89 1.96
214111 7.22 45.84 1.40 7.03 2.12 28.62 0.97 5.48 1.60
214113 6.16 46.20 0.73 7.84 2.11 31.15 1.08 4.01 1.48
214115 7.48 48.89 0.73 8.32 0.88 30.23 1.15 1.37 1.62
214117 5.78 50.15 0.73 7.17 0.85 30.64 1.36 2.47 1.64
214123 3.25 45.77 0.89 5.73 1.13 35.22 1.12 6.72 1.30
Mean 6.20 46.53 0.88 7.50 1.46 31.62 1.10 4.29 1.49

139
Table 23: The average value of CaO, FeO and Cr2O3 concentration in the slag without and with the
decarburization control
mslag mFeO,slag mCr2O3,slag
Case
[kg/t]
Without decarburization control 159.3 5.4 12.6
With decarburization control (the plant trial phase 1) 138.2 1.3 9.5
With decarburization control (the plant trial phase 2) 147.8 1.3 6.2

2.2.8.2 Plant trials and assessment at TKN (T 8.2)

The economic crisis in 2008/2009 did intensely influence the production volume, operating plans and
development activities for the TKN steel plant. These circumstances made it difficult to perform the
final plant trials like initially planned.
Due to the reduced production the number of trials performed during this task was too small to deduce
unambiguous and statistically certain quantitative results regarding the influence of the advanced
process control on productivity, economic and environmental implications.
Nevertheless it was possible to perform several final trials to evaluate and validate the online use of the
dynamic energy process modelling and monitoring system developed and described in section 2.2.3.3.
The system already proved to be useful on a qualitative level in giving the furnace operator additional
information on the process regarding off-gas and cooling water energy losses, unknown prior to the
implementation of the system.
2.2.8.3 Plant trials and assessment at MH (T 8.3)
The dynamic control strategy of the oxygen flow rate for post-combustion has been described in detail
under WP 4. The required flow rate Q [m3/h] is calculated with formula (62) which is based on the post-
combustion criteria CO/O2. The criteria consider the correlation between CO and H2 as well as CH4
from IR/VIS measurements. The parameters m1 and m2 define the chosen minimum and maximum
oxygen flow rate. By the parameters n1 and n2 the range of the criteria CO/O2 is defined.
n
 Nm3  m 2  m1  CO 
2
  m 2  m1  
Q        m1  n1    (62)
 h  n 2  n1  O 2  n   n 2  n1  
1

Figure 197 shows the calculated and actual oxygen flow rate of one injector during the progress of one
heat with dynamic post-combustion control.

30 O2 [%] CO [%] 500

power-on door lance
criteria [Nm3/h] PC-Injector [Nm3/h]

25
400
oxygen post-combustion PC1 [Nm /h]
3

20
300
O2, CO [%]

15

200

10

100
5

0 0
0 5 10 15 20 25 30
minutes

Figure 197: Calculated and actual oxygen flow rate of PC-Injector 1 based on the off-gas composition

140
 Parameters:
m1 = 200 n1 = 0.5
m2 = 400 n2 = 0.9
In order to find the best operation practice of the dynamic post-combustion control with an efficient
input of oxygen several trials with different parameter settings were performed. Figure 198 shows the
specific electrical energy and post-combustion oxygen consumption per tonne liquid steel of approx. 70
heats.
400

350

300
[kWh/t]

250

200

150

100
0 2 4 6 8 10 12 14
3
PC [Nm /t]

Figure 198: Electrical energy and post-combustion oxygen consumption per tonne liquid steel of 70
heats
The energy consumption increases with higher post-combustion oxygen consumption. The reason for
this contradictory result can be explained by the chosen parameters m1=200, m2=400, n1=0.5 and
n2=0.9. By using these values the post-combustion system operates also at low CO/O2-ratios for a long
time with low oxygen flow rates. Below a certain CO/O2-value no additional post-combustion energy
can be generated by oxygen injection. Therefore post-combustion oxygen is not injected efficiently. It
will be more efficient to inject oxygen at higher CO/O2-ratios with higher flow rates.
In Figure 199 the taken action is shown:
Increase: n1 from 0.5 to 0.7
m1 from 200 to 300

m2
PC-Injector [Nm3/h]

m1

n1 n2

CO/O2-Laser
Figure 199: Linear-proportional control curve with changed parameters m1, m2, n1 and n2

141
Figure 200 shows the result of a trial with the corrected parameters as described in Figure 199. Post-
combustion oxygen is now injected at higher CO/O2-ratios with higher flow rate. With the dynamic
control the electrical energy consumption is equal at different oxygen consumption but the average
oxygen consumption decreased from 8 to 5 m3/t compared to the original automatic off mode.
450

400

350

300
[kWh/t]

250

200

150

100
1 2 3 4 5 6 7 8 9 10

PC [Nm3/t]

Figure 200: Electrical energy and post-combustion oxygen consumption with improved parameters
Several trials with different m and n values showed that the PC-system is most efficient at the
maximum flow rate of 400 m3/h oxygen (m1=m2). It would be advantageous for electrical energy
consumption to increase m2 but that is not possible due to the installed facilities. Higher flow rates can
also run the risk of destroying the water cooled panels if scrap covers the nozzles.
This operation mode is equal to the first automatic PC-mode whereby the injectors are only switched
ON or OFF with a constant flow rate depending on the CO/O2-ratio resulting in a reduction of post-
combustion oxygen consumption by 40 %.
In case of a laser measurement breakdown the post-combustion system operates in the automatic off
mode five minutes at each basket with constant flow rates. The following trial aims at optimisation of
the automatic off mode by reducing the operation time from five to four minutes. The trial with 70 heats
was performed alternating between the automatic ON and OFF mode. In Figure 201 the operation of
the post-combustion in the automatic ON mode is shown.
Figure 202 shows the result of the different operation modes. In the automatic mode the PC-oxygen
consumption is only 1 m3/t lower compared to the automatic OFF mode with shortened operation time.
No significant differences in the electrical energy consumption or yield were observed.
As the trial showed the automatic OFF mode with reduced operation time is a suitable practice in case
of laser measurement malfunction.
The laser measurement system does not only provide information about the off-gas composition and
temperature. There is also information about the transmission of the laser light between sender and
receiver available. The received light intensity depends on the dust load in the measurement path
between the water cooled lances.
In the following the transmission signal through the heat is examined more detailed in order to get
information about the dust load in the off-gas.
The relationship between dust load and transmission is explained in formula (63):
 transmission, temperature)
dust load[g / Bm3 ]  f (V, (63)

142
Figure 201: Off-gas measurements and post-combustion with automatic ON mode

350
dynamic PC-Automatic
ON / OFF
300 67 heats

250

200

150

100

50

0
scrap [t] power-on [min] PC O2 [Nm³/t] [kWh/t] ttt [min] power-off [min] yield [%]
OFF 41,49 29,6 5,6 339,28 39,7 10,03 0,83
ON 41,52 29,7 4,6 339,31 39,2 9,48 0,83

Figure 202: Results of the trial automatic ON vs. automatic OFF with shortened operation time
In Figure 203 the transmission, the calculated dust load, the calculated off-gas temperature, the power-
on time and the lancing time is shown.

143
 100 1200

90 Zn 907 °C
ZnO 1300 °C Pb 1750 °C 1000
80 PbO 1470 °C

Temperature [°C], total dust [kg]

70
800
Transmission [%],
dust load [g/Bm ]
3

60

50 600

40
400
30

20
200

10

0 0
0 5 10 15 20 25 30 35
time
power-on transmission carbon door-lance
dust load temperature kum. dust

Figure 203: Calculated dust load and transmission during one heat
At the beginning the transmission is very high at 80 %. The furnace is switched off and no dust is
present in the offgas. When the melt down starts the transmission decreases and the calculated dust load
rises to approx. 10 g/m3. Due to an increase of the off-gas temperature the off-gas density decreases
leading to lower dust concentrations. At the end of the melt down of each basket the off-gas
temperature reaches the highest values. The furnace door is opened to operate with the oxygen
manipulator. A high amount of false air and additional oxygen from the manipulator pass through the
furnace resulting in a high dust load of more than 100 g/m3. Because the boiling temperature of zinc is
907 °C it is assumed that zinc is leaving the furnace at this process period. Zinc is the main component
of the dust (approx. 38 %). In the refining period there is flat bath condition. Due to the CO formation
in the melt and the bursting of the bubbles slag and melt droplets are carried out with the off-gas. The
dust load in the refining period is between 20 and 30 g/m3. The total cumulated dust amount is about
550 kg or 15 – 16 kg/t which is a realistic value.
The dynamic process model for mass and energy balance of BFI as described in detail under WP4 has
been implemented on-line at the Marienhütte furnace. The calculated heat status from the model
supports the operator during the melting process.
A separate OPC-Server was installed to retrieve process data required for the on-line EAF-model from
BFI. The server makes it possible to access all required process data from different sources
independently from the existing system. Figure 204 shows the interfaces exemplified.
In Figure 205, the screenshot from the main screen mask of the on-line process model is shown for one
example heat. This mask displays the actual melt temperature calculated by the BFI model (white line)
and the measured temperatures (red squares) from the CELOX system. After a reasonable measurement
the model is adapted to this value, which can be seen by a jump of the displayed melt temperature. The
horizontal lines at 1530 °C and 1630 °C are used for reference of the target tapping temperature region.
The calculated temperature helps the operator to decide, when the next basket can be charged and when
the aim melt temperature is reached and thus the tapping process can be started. Furthermore this mask
shows the signals of charging (green) and tapping (blue).

144
 OPC-
Server

Figure 204: OPC-Server data processing

Figure 205: Chart of the calculated heat temperature with the measured temperatures and the signals
charging and tapping
An enlargement of the temperature range between 1500 °C and 1800 °C can be selected by the tabs at
the top to improve the monitoring during the refining phase. The charts of the different energy inputs
concerning electrical and chemical energy and the charts of the energy losses, comprising the losses by
off-gas, cooling water as well as by overall radiation and convection can be selected by the tabs too.
In Figure 206, the measured off-gas analysis from the LINDARC system (O2: red line, CO: green line)
and the off-gas temperature (white line) derived from the cooling cycle of the post-combustion channel

145
are shown. The concentration of CO is below 5 % and O2 can be measured during the whole melting
process for this example heat. The flow rates of oxygen injected by the door lance, the three KT
injectors and the four post-combustion injectors together with the amount of injected carbon can be
displayed too.

Figure 206: Chart of the off-gas temperature and concentrations of CO and O2

The on-line model collects all relevant cyclic and acyclic input data during the melting process and
stores these data for each heat. These collected data can be used for off-line evaluations to optimize the
model and to perform analyses for the processes at the Marienhütte furnace, e.g. to assess the
performance of the control system for post-combustion oxygen.
The optimisation of the dynamic process model was performed with the data of about 1050 heats which
were collected after the on-line implementation in March 2009. In Figure 207, the calculated melt
temperature is plotted against the first plausible measured temperature. A modelling error standard
deviation of about 33 K, respectively converted into an error of the energy balance of 7.6 kWh/t, was
achieved. With respect to the total energy input of around 690 kWh/t, this is a relative error of the
energy balance of below 1.1 %, which is already a good accuracy. Figure 208 displays the calculated
melt temperatures adapted to the previously reasonable measured ones against the further plausible
measured temperatures. The accuracy can be improved with this adaption, the standard deviation has
been decreased to a value of about 20 K.
The utilisation of the on-line model has reduced the average number of temperature measurements from
3 to 2.5 per heat. The share of plausible temperature measurements has been increased from 85 % to
90 %, as the model indicates when the meltdown process has been completed and thus a temperature
measurement makes sense. In general the performance of the furnace has been improved by decreasing
the number of interruptions caused by the measurement process.

146
 1750
Mean error: -3.3 °C
Standard deviation: 33.4 °C

1700

Calculated melt temperature in °C

1650

1600

1550

1500
1500 1550 1600 1650 1700 1750
Measured melt temperature in °C

Figure 207: Calculated melt temperature vs. first plausible measured melt temperature
1750
Mean error: 1.5 °C
Standard deviation: 19.7 °C

1700
Calculated melt temperature in °C

1650

1600

1550

1500
1500 1550 1600 1650 1700 1750
Measured melt temperature in °C

Figure 208: Calculated melt temperature adapted to previous reasonable measured ones vs. further
plausible measured melt temperature
2.2.8.4 Plant trials and assessment at ORI (T 8.4)
Market crisis has lowered development activity due to the overnight operation only and at the time of
project ending a long term tests for evaluating (on a statistical base benefits) the implemented system
have been scheduled, to be carried out as soon as the production rate is enough stabilized and normal
production restored to be compared with the historical condition.
Nonetheless in the mean time ORI Martin plant made further major innovation (i.e. the fume system
revamping, modification in the chemical energy equipment and then furnace layout). Then more time

147
might be needed to recalculate a new baseline accordingly with the new furnace configuration and to
collect enough data to be statistically meaningful.
In any case at the time of writing all computation modules, interfaces and control modules have been
installed on line and their functionality and reliability have been fully tested. As further improvement,
an innovative device for direct measuring of the off gas flow rate could enhance the system overcoming
the need of calibrating software algorithms, a system therefore able to manage even “non standard”
conditions.
2.2.8.5 Plant trials and assessment at PA (T 8.5)
In order to evaluate the better efficiency of a new operating pattern with over-stoechiometry practice, a
campaign was performed (operating pattern 8 on Table 24) and a reference operating pattern (operating
pattern 0 on table 2) was added to previous operating patterns (Table 13).
Table 24: Modifications of the operation pattern made by the EAF engineer
Operating pattern Number of heats Modifications
0 52 Period of reference taked between the operating pattern 7 and 8

1 33 Beginning of the tests with increasing of O2 supply by RCB lances

2 Limiting oxygen during preheating with the aim of decreasing the oxidation
2
14 Decrease of fossil energy at the first basket

3 33 Increase of post combustion at the first basket

Decrease of burners energy at the first basket

2
4 Decrease of refining oxygen at the second basket

16 Little alteration of refining oxygen at the second basket

Modification of refining oxygen. They tried to decrease the oxygen injected

5 57
by RCB lances because the steel is too much oxydated at this stage

Decrease of post combustion for the RCB burners

6 99
7 33
Increase of post combustion for the burners 2 and 5 (PC lances) (1)
Increase of post combustion at the first basket
Remove of the post combustion by RCB burners (2)
Reincrease of the refining oxygen: there is no more oxydation problems

Increasing of the stoechiometric ratio of the RCB burners frome to 2 to 2.1 in

order to improve the post-combustion ratio. Oxygne which do not react with
8 231
methane is usefull for post combustion because his speed is lesser than
injection with RCB lances

(1) Increasing of post combustion for PC lances 2 and 5 do not compensate the decrease at the RCB
burners. Then, it is needed to increase post combustion at the first basket.
(2) Post combustion by the RCB burners is not efficient and causes problems for the ladle metallurgy
coming later.
The operating pattern 8 (cf. Table 13) has been tested during one month in the plant of ArcelorMittal
Differdange.
Figure 209 shows the correlation of the cost per ton of steel of furnace 1 and furnace 2. The campaign
was conducted during one month. Note that this correlation should be useful to quantify the savings for
an operating pattern by comparing a furnace with a modified practice and a furnace without
modification in the operating pattern.
Figure 210 the cost per ton of steel produced for all operating patterns. The operating pattern 8 (circle
in red) shows clearly saving of 0.75 euro’s per ton of steel produced related to operating pattern 0
(standard practice).

148
 Cost for per ton of steel produced for each heat for  
operation pattern 8 (for furnace 1 and 2)
29

Euros/Ton of Steel Produced
27
25 Furnace 1 (reference)
23
Furnace 2 (modified 
21 practice)
19 10 per. Mov. Avg. (Furnace 1 
(reference))
17 10 per. Mov. Avg. (Furnace 2 
15 (modified practice))
4000 4100 4200 4300 4400 4500 4600
Heat

Figure 209: Energetic cost per ton of steel produced for each heat for furnace 1 and 2. The moving
averages of the cost are plotted on this figure
Average energetic cost for each operating 
pattern
27
Energetic  cost (euros) / Ton of 

26
25
24
scrap*yieled

23
22
21
20
19
18
0 1 2 3 4 5 6 7 8 9
Operation Scheme

Figure 210: Energetic cost per ton of steel produced for each operating pattern
Figure 211 shows the post-combustion ratio for each operating pattern. The operating pattern 8 has a
post-combustion ratio almost equal to this of operating pattern 3. This is logical because operating
pattern 3 and 8 have gave better savings. But, for operating pattern 3, there were oxidation problems,
and for operating pattern 8, it is not the case.
The savings of operation mode 8 is suitable in comparison with the savings costs evaluated in the point
3.3.6.4. It is important to emphasise that there is no more oxidation problems for operating pattern 8 as
it was the case for operation pattern 3. Furthermore, the predictions of the model (cf. Figure 164) are
very good, in relative, for the operating pattern 8. These considerations prove that operating pattern 8 is
better than the others operating patterns because there are no more oxidation problems and savings are
around 0.75 euro’s.
Figure 212 shows the scrap yield for each operating pattern expressed in kg of scrap per ton of liquid
steel produced.
The observed yields seem consistent with the previous considerations:
 The operating patterns 1 to 3 show the worse scrap yield which can be explained by the
observed oxidation problems,
 Except for the number 4 (very few heats), the operating patterns 4 to 7 lead to the best scrap
yield, number 7. But as said before, the energetic cost is degraded,

149
  The over stoechiometry practices (number 8) doesn’t deteriorate the scrap yield and appears to
be the most valuable operating pattern.
Data for O2 measured averages < 15 (%w/s) 
adjusted
0.8
0.7

PC ratio adjusted (%)
4
0.6 4
22
12
43 16

20
0.5 5

0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7 8 9
Operation patterns

Figure 211: PC ratio averages (with oxygen measured averages inferior to 15 %w/s), adjusted with the
parameter alpha, function of operation patterns. Note that the PC ratio for the reference
period is not represented because we don’t have an off-gas analysis of this period

Figure 212: Scrap yield for each operating pattern

150
2.3 CONCLUSIONS
In close cooperation with RWTH at the DEWG furnace in Siegen an empirical model for the optimum
decarburization has been developed. The adapted off-gas analysis system has been utilized to monitor
and control the decarburization of high alloyed steel heats by oxygen injection. The application of
model and off-gas analysis system lead to a decrease of oxygen injection of over 10 % for 6 out of 15
heats in the final plant trials and therefore led to a decrease of tap-to-tap time. The heats still met all
demands in regard of slag and steel analysis and additionally the mean amount of Cr2O3 in the slag
could be considerably reduced which also is economically and environmentally beneficial.
At TKN the data acquired from the newly installed off-gas measurement systems have been combined
with data from the hot water cooling system of the plant in the EAF process control system. A dynamic
energy process modelling and monitoring system has been developed by RWTH and has been put into
operation at TKN. The system presents data for the main energy sinks and thus expands the process
data already available to the furnace operator. Additionally the current process rules regarding the
oxygen use and its influence on the electric energy input have been investigated and adapted rules have
been proposed in order to reduce electric energy demand.
At the MH furnace, a dynamic closed-loop control strategy for the input of post-combustion oxygen
depending on the off-gas analysis was developed and implemented in the on-line control system. It
guarantees an optimal use and an increased energetic efficiency of the post-combustion oxygen.
Intensive plant trials with different control strategies for post-combustion injection and different
configurations of other process parameters were performed at the MH furnace. The results of these
trials were evaluated with the statistical and dynamic energy balance models of BFI with regard to
losses via the off-gas and overall energetic performance of the furnace.
The dynamic process model of BFI has been implemented on-line at Marienhütte for observation of the
actual energetic status of the furnace, represented by the calculated melt temperature. This tool supports
the operator by providing further information of the actual heat, especially on the energetic status, the
achieved degree of melt down and the current melt temperature. The latter one can be calculated with
an accuracy of less than 20 K error standard deviation.
The utilisation of the on-line model has reduced the average number of temperature measurements from
3 to 2.5 per heat. The share of plausible temperature measurements has been increased from 85 to 90 %,
as the model indicates when the meltdown process has been completed and thus a temperature
measurement makes sense. In general the performance of the furnace has been improved by decreasing
the number of interruptions caused by the measurement process, and has thus lead to an on the average
slightly reduced tap-to-tap time.
The trials performed at DEWG and TKN were investigated by RWTH regarding the impact of process
parameters on electric energy demand and oxygen consumption. The off-gas losses were investigated in
relation to the electric energy demand for carbon and stainless steel at DEWG and TKN and in both
cases a correlation between high energy consumption and high off-gas losses could be validated. The
investigation of the oxygen injection in relation to off-gas losses and energy consumption lead to a
general correlation of rising off-gas losses with increasing oxygen injection. The energy input is slightly
decreasing for TKN and increasing for DEWG with an increase of oxygen injection.
The investigation of influence of the dedusting system on the energy demand performed by RWTH
showed that a reduction of the off-gas mass flow for example by increased air tightness of the furnace
or a dynamically controlled off-gas flow rate reduces the off-gas losses as well as the electric energy
demand.
The application of an off-gas measurement in AM Differdange has allowed for defining an optimum
operating pattern leading to savings in energetic cost of 0.5 euro per ton of steel produced. The CRM
EAF model has been used for simulating several possible operating patterns. The post-combustion ratio
has been slightly increased. Further gain could be obtained thanks to the installation of supplementary
burners/injectors in this furnace which was out of scope of the present project. The improvement will be
pursued in this way by the plant.
In close cooperation with CSM Tenova developed, installed and improved a new process model based
on off-gas analyzing systems and plant data at the ORI Martin Consteel plant. The new comprehensive
Consteel process model includes dynamic models for chemistry calibration, wet chemistry, off-gas flow
rate, a freeboard model as well as a bath and slag model. The output of these models and the newly

151
available measurements (i.e. off-gas composition) combined with purposely developed control code
constitute the new applied control system at ORI Martin. Due to the new online models the control
system is now able to feed the oxygen and carbon injector with dynamic set points chosen according to
the current process conditions.
Using e.g. the EAF post combustion model a decrease in thermal loss from the bath protecting the slag
and the bath from an excessive radiation could be achieved. Additionally an increase in the thermal
exchange and the scrap temperature in proximity to the furnace (the available energy increases the scrap
temperature speeding up its melting. Generically the post combustion warms the cold furnace zone) as
well as an increase in the post combustion efficiency along the conveyor could be achieved. The higher
off-gas temperature, given by the high post combustion ratio, increases the thermal exchange in the end
of conveyor end. This results in a higher scrap preheating.
In general the application of the new control system at the Consteel plant of ORI Martin lead to a
reduction in oxidizing oxygen and an increasing in energetic yield (this affects positively the slag
quantity and composition. The FeO, the CaO and the amount of slag have been reduced). The process
also is more stable. This involves an easier and better process/production management for all the plant.
Additionally the post combustion reaction increases the off-gas volume so at the same fume system
working condition the higher the post combustion the fewer false air is sucked from the slag door and
the gap. The post combustion effect along the entire conveyor increase the thermal exchange and so the
scrap preheating efficiency. The post combustion effect is visible also in the off-gas chemical analysis
measured by EFSOP.
Comparison of off-gas analysis systems
A comparison of the utilized off-gas analysing systems is given in Table 25. The systems used at DEW,
TKN, ORI and AM are very similar in principle. They all rely on water cooled sampling probes in
combination with some kind of sample conditioning (filter, water-condensation, etc.). The cleaned off-
gas sample is then analysed by conventional IR gas sensors (CO, CO2, CH4), electrochemical cells or
paramagnetic sensors (O2) and thermal conductivity sensors (H2). The long-time reliability of these
sensors is proven and tested. This setup also allows for the installation of the sensitive and valuable
analyser in a protected area of the plant. Nevertheless in all installation special measures have to be
taken to prevent a clogging of sampling probes and sample line, usually a purging with pressurized air
is implemented. The response time of these systems based conventional analysers is, apart from the
time needed for sample conditioning, depending on the length of the sample line (distance probe –
analyser) as well as the volume flow rate of the sample taken.
The system used at MH on the contrary is a laser-based spectroscopic analyser measuring directly in the
primary off-gas system. Therefore the response time of the system is almost instantaneous and no
sampling or sample conditioning equipment is needed. A disadvantage of the system is the proximity of
the delicate optical and electrical systems to the furnace which leads to a potentially higher risk of
damages to the analyser system. During the project period, several measures were taken successfully to
achieve low maintenance and high operational availability of the laser-based off-gas analysis system.
Direct laser-based off-gas measurement is a rather new technology by comparison and right now not for
all desirable gas species a dedicated laser is available. So some interesting species like CO2 cannot be
measured with this system up to now. Nevertheless from Marienhütte point of view, the laser off-gas
analysis is believed to be a key tool to drive the tuning of the EAF process, thanks to its real-time
response. Besides the possibility of a closed-loop control for CO post-combustion in the EAF as well as
the optimization of the burners O2/CH4 ratio, there is a particular interest in using the laser off-gas
analysis as safety device. New developments of laser diodes enable the reading of the water content
directly in order to detect dangerous water leaks inside the EAF. Additionally, the range of chemical
species in the off-gas to be analysed expands steadily.

152
Table 25: Comparison of off-gas analysis systems
Off-gas analyser Technology Advantages Disadvantages
ABB Infrared detectors for Proven and tested Response time depending
DEW, TKN CO and CO2 technology, long-time on distance between
Paramagnetic detector reliability of detectors, analyser and measuring
for O2 system can be set up in point, sample conditioning
protected area, detectors required, regular
available for almost all gas maintenance and
species calibration needed
LINDARC Laser-based NIR Measurement without time New technology, optical
MH spectroscopic delay, no sampling and and electrical device in
measurement of CO sample conditioning direct proximity of the
and O2 required, temperature furnace, only available for
measurement, no moving specific gas species, high
or consumable parts use of purging N2
EFSOP Infrared detectors for Proven and tested Response time depending
ORI Martin CO and CO2 technology, long-time on distance between
Electrochemical cell reliability of detectors, analyser and measuring
for O2 system can be set up in point, sample conditioning
Thermal conductivity protected area, detectors required, regular
detector for H2 available for almost all gas maintenance and
species calibration needed
RWTH Infrared detectors for Proven and tested Response time depending
AM Differdange CO, CO2 and CH4 technology, long-time on distance between
Paramagnetic detector reliability of detectors, analyser and measuring
for O2 system can be set up in point, sample conditioning
Thermal conductivity protected area, detectors required, regular
detector for H2 available for almost all gas maintenance and
species calibration needed

153
2.4 EXPLOITATION AND IMPACT OF THE RESEARCH RESULTS
The results of the research project have been used to improve the BFI and CRM models. They have also
been used in another running RFCS project named FLEXCHARGE. Tenova used the new generation of
EAF control system, which elaborate simultaneously the data from SW and HW sensors to develop a
commercial system named iEAF. The iEAF already proofed its potential in other Italian plants giving
good results. The iEAF impact is not only about energy saving but also in reducing CO2 emission (in
top charge furnaces). Starting from the system developed in ORI Martin a new RFCS proposal project
named SMARCON is now in evaluation.
Publications and conference presentations resulting from the project are given in Table 26.
Table 26: Publications and conference presentations regarding the OFFGAS project
Publication
Krassnig, H.-J.; Kleimt, B.; Voj, L.; Antrekowitsch, H.: EAF post-combustion control using
laser off-gas analysis, Stahl und Eisen 128 (2008)
Risonarta, V. Y.; Echterhof, T.; Voj, L.; Pfeifer, H.; Jung, H.-P.; Lenz, S.: Optimization of the
electric arc furnace process at Deutsche Edelstahlwerke, Stahl und Eisen 129 (2009), Nr. 11, S.
S55-S64
Conference presentation
Krassnig, H.-J.; Kleimt, B.; Voj, L.; Antrekowitsch, H.: EAF post-combustion control by on-line
laser-based off-gas measurements, 9th European Electric Steelmaking Conference EEC Krakau,
19. - 21.05.2008
Risonarta, V.; Voj, L.; Pfeifer, H.; Jung, H. P.; Lenz, S.: Optimization of electric arc furnace
process at Deutsche Edelstahlwerke, 9th European Electric Steelmaking Conference EEC
Krakau, 19. - 21.05.2008

154
3 List of Figures
Figure 1:  EAF layout at DEWG Siegen ................................................................................................ 12 
Figure 2:  Water-cooled probe installed at the elbow near the EAF (point A) at DEWG Siegen.......... 13 
Figure 3:  Existing off-gas analysis system (left) and layout of the ABB analysis system installed
(right) ................................................................................................................................ 13 
Figure 4:  Measured off-gas composition at point A (DEWG; carbon steel) ........................................ 14 
Figure 5:  EAF layout at TKN Bochum ................................................................................................. 14 
Figure 6:  Layout of the off-gas measurement installed at point A at TKN Bochum ............................ 15 
Figure 7:  Installation of the water-cooled sampling probe in front of the primary off-gas duct........... 15 
Figure 8:  The ABB off-gas analyzer ..................................................................................................... 16 
Figure 9:  Off-gas sampling positions at point A in 2007 (left) and 2008 (right) .................................. 16 
Figure 10: Mechanical drawing of water-cooled gas sampling probe with air insulation layer ............. 17 
Figure 11: Cross-section of the modified water-cooled probe ................................................................ 17 
Figure 12: Principle of Installation ......................................................................................................... 18 
Figure 13: Principle of Absorption ......................................................................................................... 18 
Figure 14: Top view of the AC EAF at Marienhütte .............................................................................. 19 
Figure 15: Layout of the EAF at MH...................................................................................................... 19 
Figure 16: Water cooled lances and housing .......................................................................................... 20 
Figure 17: Visualisation screen ............................................................................................................... 20 
Figure 18: Three adjusting screws for the housing of the laser .............................................................. 21 
Figure 19: Electric units with assembly of a switch board in an air-conditioned area ........................... 21 
Figure 20: Off-gas laser measurements with carbon monoxide laser drop out....................................... 21 
Figure 21: Electronic board encased by a water cooled spiral ................................................................ 22 
Figure 22: Calculated versus actual electrical energy demand for MH single heats .............................. 23 
Figure 23: EFSOP probe installed and the box with control panels of gas probe and gas sensors ......... 24 
Figure 24: EFSOP cabinet HMI - Off-gas analysis system detail .......................................................... 25 
Figure 25: Details of the probe installation in AM Differdange ............................................................. 25 
Figure 26: Details of the RWTH water cooled probe for off-gas sampling............................................ 26 
Figure 27: Location of the sampling probe in the elbow at the AM Differdange furnace ...................... 26 
Figure 28: Location of the sampling point A & point B in the exhaust gas system at AM Differdange 27 
Figure 29: Location of the probe at point A............................................................................................ 27 
Figure 30: Location of the sampling points in the exhaust gas system ................................................... 28 
Figure 31: Sampling probe installed in the inlet of the PC chamber (ArcelorMittal Differdange) ........ 28 
Figure 32: Off-gas concentration, temperature and integrated carbon mass in the off-gas at point B (an
exemplary stainless steel heat) .......................................................................................... 29 
Figure 33: Carbon mass balance of stainless steel heats ......................................................................... 29 
Figure 34: Three consecutive decarburization stages during the refining period (austenitic steel grade)
.......................................................................................................................................... 30 
Figure 35: Three consecutive decarburization stages during the refining period (ferritic steel grade)... 31 
Figure 36: Oxygen injection vs. measured carbon content in off-gas..................................................... 31 
Figure 37: Steel sampling (left) and slag sampling (right) ..................................................................... 32 
Figure 38: Influence of excess oxygen injection on chromium and iron oxidation (without
decarburization control) .................................................................................................... 33 
Figure 39: The empirical model of the 2nd critical point of oxygen injection ......................................... 34 
Figure 40: Real values vs. model values of chromium concentration at tapping (stainless steel heats) . 34 

155
Figure 41: Flow chart of off-line test strategy ........................................................................................ 35 
Figure 42: The result of the thermodynamic simulation of elements oxidation ..................................... 36 
Figure 43: Monitor display of ABB analyzer unit at the dispatcher ....................................................... 37 
Figure 44: Pitot tube for measuring the differential pressure at point B (left); Data acquisition: Off-gas
temperature, differential pressure and volume flow rate (right) ....................................... 37 
Figure 45: Pressurized hot water cooling system at the Bochum EAF ................................................... 38 
Figure 46: Chart of the steam produced in the hot water cooling system at the Bochum EAF .............. 39 
Figure 47: Off-gas composition monitoring at point A (top) and point B (bottom) ............................... 41 
Figure 48: Operator view of the dynamic energy modelling and monitoring system............................. 42 
Figure 49: Reaction zones during oxygen injection into the liquid steel [14]. Hot reaction zone (1),
Liquid steel-slag zone (2), Slag zone (3), and Liquid steel zone (4) ................................ 43 
Figure 50: Oxygen injection and electrical energy input vs. the tap-to-tap time .................................... 44 
Figure 51: Electrical energy input vs. thermal and chemical off-gas enthalpies at point A ................... 44 
Figure 52: Tap-to-tap time vs. thermal and chemical off-gas enthalpies at point A ............................... 45 
Figure 53: Off-gas measurement points during RWTH trial campaign at MH furnace ......................... 46 
Figure 54: Installation of LINDarc and RWTH (portable) water-cooled off-gas probes at the EAF
elbow................................................................................................................................. 46 
Figure 55: Air intake at the gap between EAF vessel and primary dedusting system ............................ 47 
Figure 56: Layout of the RWTH (portable) off-gas analysis system ...................................................... 47 
Figure 57: Measured CO and O2 concentration curves (point A): (RWTH (portable) vs. LINDarc) ..... 48 
Figure 58: Measured concentration H2 and CH4 curves (RWTH (portable system)) (point A) ............. 48 
Figure 59: Measured concentration curves and differential pressure, and off-gas temperature (RWTH
(portable system)) (point B) .............................................................................................. 48 
Figure 60: Operating data ....................................................................................................................... 49 
Figure 61: Correlation of CO/O2 ratio from laser and IR/VIS ................................................................ 49 
Figure 62: Installation of the thermocouple probe close to the LINDARC water-cooled lance ............. 50 
Figure 63: Correlation between the off-gas temperature measured by the thermocouple and the outlet
cooling water temperature from the LINDARC water-cooled lance ................................ 50 
Figure 64: Measured off-gas temperature by the thermocouple probe and calculated off-gas temperature
with equation (26) ............................................................................................................. 51 
Figure 65: Calculated off-gas temperature and measured temperature from the LINDARC system ..... 51 
Figure 66: Structure of the dynamic process model with input data available at the MH AC-Electric Arc
Furnace ............................................................................................................................. 52 
Figure 67: Electrical energy input, KT burner and oxygen injection rates of a MH example heat ........ 53 
Figure 68: Cooling water flow rates and cooling water temperatures of a MH example heat................ 55 
Figure 69: Off-gas measurement data by LINDARC and RWTH Aachen measurement systems of a
MH example heat .............................................................................................................. 56 
Figure 70: Energy input and loss rates as well as melt temperature evolution calculated for a MH
example heat ..................................................................................................................... 57 
Figure 71: Off-gas analysis (CO, CO2, CH4, H2 and O2 concentration) at measurement point A .......... 58 
Figure 72: Off-gas analysis (CO, CO2, and O2 concentration), off-gas temperature and differential
pressure at measurement point B ...................................................................................... 59 
Figure 73: Measured off-gas data and calculated off-gas flow rates at measurement point A and B ..... 59 
Figure 74: Calculated off-gas flow rates at measurement point A and B together with leakage air
ingress at the gap .............................................................................................................. 61 
Figure 75: CO and H2 content of off-gas at measurement point A together with CO/H2 ratio and
difference .......................................................................................................................... 61 
Figure 76: Correlation of CO and H2 (IR/VIS) ....................................................................................... 62 

156
Figure 77: Correlation of CO and CH4 (IR/VIS).................................................................................... 62 
Figure 78: Correlation of CO and H2 (IR/VIS) of and CH4 (IR/VIS) of the first basket ........................ 62 
Figure 79: Correlation of CO and CH4 of the first basket....................................................................... 62 
Figure 80: Calculated concentrations of H2 and CH4 according to equation (36) and (37) .................... 63 
Figure 81: Off-gas temperature measured by LINDARC system and calculated from cooling water
temperature of laser lance according to formula (26) ....................................................... 63 
Figure 82: Line shape function of the post combustion criteria.............................................................. 65 
Figure 83: Off-gas profile ....................................................................................................................... 66 
Figure 84: Operation mode and CO/O2 ratio .......................................................................................... 66 
Figure 85: Correlation of CO/O2 and post combustion oxygen .............................................................. 67 
Figure 86: Results of alternating trial with 160 heats ............................................................................. 67 
Figure 87: Electrical energy, power on time and tap to tap time against post combustion oxygen ........ 68 
Figure 88: Correlation of CO/O2 ratio and offgas temperature at measurement point K3 ..................... 68 
Figure 89: Calculated vs. actual electrical energy consumption for MH heats of campaign March 2007
with alternating PC oxygen control mode ........................................................................ 69 
Figure 90: Oxygen injection rates for example heats without and with automatic PC control............... 70 
Figure 91: Measured LINDarc off-gas values, CO/O2 ratios and furnace pressure, and calculated
concentrations of H2 and CH4 for example heats without and with automatic PC control 71 
Figure 92: Energy input and loss rates and melt temperature evolution calculated with the dynamic
energy balance model for example heats without and with automatic PC control ........... 71 
Figure 93: Linear proportional PC-control strategy ................................................................................ 72 
Figure 94: O2 flow rate of PC-injector 1 with dynamic control .............................................................. 73 
Figure 95: Total oxygen and natural gas................................................................................................. 73 
Figure 96: CO and O2 profile .................................................................................................................. 73 
Figure 97: Dynamically controlled PC oxygen flow rate together with off-gas analysis values for an
example heat ..................................................................................................................... 74 
Figure 98: Natural gas and oxygen input for the two example heats ...................................................... 74 
Figure 99: LINDARC off-gas values and a comparison between the measured LINDARC off-gas
temperature with the calculated from the cooling water of the LINDARC laser lance .... 75 
Figure 100:  Energy input and loss rates for example heats .................................................................. 75 
Figure 101:  Melt temperature evolution calculated with the dynamic energy balance model ............. 76 
Figure 102:  ORI Martin - EFSOPTM HMI (Human-Machine Interface) .............................................. 77 
Figure 103:  General scheme of EAF model ......................................................................................... 79 
Figure 104:  Graphical interfaces of the software for model input and output as standalone application
.......................................................................................................................................... 80 
Figure 105:  Freeboard model ............................................................................................................... 81 
Figure 106:  Installation position of the pyrometer ............................................................................... 83 
Figure 107:  Example of pyrometer measurement data (green line) in one heat................................... 83 
Figure 108:  Input of metallic materials (scrap and hot metal; the values on Y axis are percentage of
total charged per minute ................................................................................................... 85 
Figure 109:  Coal lump and oxygen injected (on Y axis % of total charged per minute) ..................... 85 
Figure 110:  Comparison of calculated carbon content and measured .................................................. 85 
Figure 111:  Average temperature of the slag/metal system as calculated by the code and
measurements in the liquid pool close to the slag door .................................................... 86 
Figure 112:  Comparison and computed (graph) and measured (table) bath temperature .................... 86 
Figure 113:  Steel composition evolution - example ............................................................................. 87 
Figure 114:  Slag composition evolution - example.............................................................................. 87 

157
Figure 115:  Print screen of datasheet for model input of scrap type .................................................... 88 
Figure 116:  Print screen of datasheet for model input of coal and fluxes ............................................ 88 
Figure 117:  Scheme of sampling points for gas flow rate continuous measurement ........................... 89 
Figure 118:  Results of mass flow rate calculations at fourth hole........................................................ 89 
Figure 119:  Logic scheme of the virtual sensor for on line gas flow rate measurements (approach 2) 90 
Figure 120:  CFD calculations of flow field.......................................................................................... 90 
Figure 121:  Static pressure measurement and gas flow rate Q_SP calculated from mass balances and
developed algorithm ......................................................................................................... 91 
Figure 122:  Comparison of calculation of Q_EAF from static pressure measurement and Q_EAF
measured by double point measurement and carbon balance ........................................... 91 
Figure 123:  Improvement of thermocouple protective tube to decrease the time of response. ............ 92 
Figure 124:  Water cooled probe and the position at the EAF fourth hole outlet ................................. 92 
Figure 125:  Temperature measurement during one heat with suction pyrometer ................................ 93 
Figure 126:  ORI Martin pig iron in charge. Progressive n° of heats vs. Pig iron in charge ................. 94 
Figure 127:  ORI Martin tap temperature. Progressive n° heats curve Vs Tap temperature ................. 94 
Figure 128:  Oxygen injector post combustion control logic ................................................................ 95 
Figure 129:  EFSOP® HMI screen shot; dynamic control post combustion set point .......................... 95 
Figure 130:  Comparison with chemical analysis between post combustion ON vs. OFF ................... 97 
Figure 131:  Example of measured downstream chemistry .................................................................. 98 
Figure 132:  HMI main page updated – the upstream and the downstream chemistry can be seen
together ............................................................................................................................. 98 
Figure 133:  Comparison between off-gas measurement at AM Differdange with model simulation.. 99 
Figure 134:  Comparison between exhaust gas measurement with model simulation, assuming a
complete decomposition of water in the fume (ArcelorMittal Differdange) .................. 100 
Figure 135:  Comparison between off-gas volume flow rate deduced for the off-gas measurements
with model simulation .................................................................................................... 100 
Figure 136:  Comparison between Venturi off-gas volume flow rate and the value deduced from the
quench unit heat balance ................................................................................................. 101 
Figure 137:  Assessment of the exhaust gas measurement delay ........................................................ 101 
Figure 138:  Assessment of the exhaust gas measurement delay (continuing) ................................... 102 
Figure 139:  Assessment of the exhaust gas flow rate and air ingress ................................................ 102 
Figure 140:  View of different process phases; Time A (at left): Burner preheating of the first basket;
Time B (at center): Melting of the second basket; Time C (at right): Refining phase. .. 103 
Figure 141:  Thermodynamic equilibrium of the exhaust gas ............................................................. 103 
Figure 142:  Exhaust gas volume assessment through mass carbon balance ...................................... 104 
Figure 143:  Appraisal of the electrode erosion assuming cooling water dissociation ....................... 104 
Figure 144:  Exhaust gas volume assessment through mass carbon balance with introduction of a coal
injection yield ................................................................................................................. 105 
Figure 145:  Appraisal of the electrode erosion assuming cooling water dissociation with introduction
of a coal injection yield ................................................................................................... 105 
Figure 146:  Second campaign of exhaust gas measurement: operating pattern profile– Reference heat
(point A).......................................................................................................................... 106 
Figure 147:  Second campaign of exhaust gas measurement: off-gas profile– Reference heat (point A)
........................................................................................................................................ 106 
Figure 148:  Second campaign of exhaust gas measurement: off-gas profile (CO2,O2) (top); off-gas
profile (CO,H2) (bottom) – Reference heat (point B) ..................................................... 107 
Figure 149:  Second campaign of exhaust gas measurement – Reference heat .................................. 108 
Figure 150:  Second campaign of exhaust gas measurement – Stop of the electrode cooling ............ 108 

158
Figure 151:  Modification of the operating pattern of the PC oxygen injectors .................................. 109 
Figure 152:  Results of CRM model simulations for normal and modified practices ......................... 109 
Figure 153:  Comparison between exhaust gas measurement with model simulation for normal
practice ............................................................................................................................ 110 
Figure 154:  Comparison between exhaust gas measurement with model simulation for modified
practice ............................................................................................................................ 110 
Figure 155:  Comparison of the electrical energy consumption to a theoretical stoechiometric ratio for
ArcelorMittal Differdange and European furnaces [4] ................................................... 110 
Figure 156:  New preheating practice in ArcelorMittal Differdange. B indicates when the burners are
working, PC when there is a post-combustion, RF a refining, F1 is the first furnace and
F2 is the second one ........................................................................................................ 111 
Figure 157:  O2 injected in the EAF with the lances of the RCB burners and the PC lances.............. 112 
Figure 158:  Cost of O2, CH4 and electrical energy for furnace 1 and 2 in function of preheating ..... 112 
Figure 159:  Energy cost for furnace 1 and 2 in function of preheating.............................................. 113 
Figure 160:  Averages of energy cost and oxygen injected by lances of RCB burners (for post-
combustion and refining) for each operation pattern for furnace 2 ................................ 113 
Figure 161:  Averages of energy cost and oxygen injected by PC lances (for post-combustion) for each
operation pattern for furnace 2........................................................................................ 114 
Figure 162:  Energetic cost per ton of steel produced obtained with the model for each operating
pattern ............................................................................................................................. 114 
Figure 163:  Trends of the PC ratios for each pattern when oxygen measured average (calculated with
electrical energy inferior to 35 MWh) is inferior to 15 (%w/s) ...................................... 115 
Figure 164:  Energetic cost per ton of steel obtained for each operating pattern with the EAF model of
the CRM.......................................................................................................................... 116 
Figure 165:  Average energetic cost per ton of steel obtained for each operating pattern .................. 116 
Figure 166:  Thermal and chemical off-gas losses depending on specific electrical energy consumption
........................................................................................................................................ 117 
Figure 167:  Deviation of model for electrical energy demand depending on off-gas losses for MH
heats ................................................................................................................................ 117 
Figure 168:  Model deviation against thermal off-gas losses for different phases of MH heats ......... 118 
Figure 169:  Model deviation against chemical off-gas losses for different phases of MH heats ....... 118 
Figure 170:  Electrical energy input vs. off-gas enthalpy at Point B and tap-to-tap time (TKN –
stainless steel) ................................................................................................................. 119 
Figure 171:  Electrical energy input vs. off-gas enthalpy at Point B and tap-to-tap time (DEW – carbon
steel) ................................................................................................................................ 119 
Figure 172:  Tap-to-tap time vs. off-gas enthalpy at Point B .............................................................. 120 
Figure 173:  Calculated vs. actual electrical energy consumption with extended and standard door
lance operation ................................................................................................................ 121 
Figure 174:  Thermal and chemical off-gas losses according to specific oxygen injection ................ 121 
Figure 175:  Specific oxygen injection vs. chemical, electrical, and total energy inputs for TKN
Bochum ........................................................................................................................... 122 
Figure 176:  Specific oxygen injection vs. specific total energy input of carbon and stainless steels for
DEW Siegen ................................................................................................................... 122 
Figure 177:  Specific oxygen injection vs. off-gas enthalpy at point B of carbon and stainless steels for
DEW Siegen ................................................................................................................... 123 
Figure 178:  Specific oxygen injection and off-gas enthalpy at point B of carbon and stainless steels
for TKN Bochum ............................................................................................................ 123 
Figure 179:  Calculated vs. actual electrical energy consumption with extended and standard use of the
door burner...................................................................................................................... 124 

159
Figure 180:  Thermal and chemical off-gas losses according to specific natural gas input ................ 125 
Figure 181:  Model deviation against specific natural gas input for different phases of MH heats .... 126 
Figure 182:  Comparison between calculated electrical energy demand and actual consumption for
previous and new preheating practice at the AM Differdange furnace .......................... 126 
Figure 183:  Schematic display of the off-gas flow ............................................................................ 127 
Figure 184:  The influence of different operating modes on the off-gas enthalpy and the electrical
energy consumption ........................................................................................................ 129 
Figure 185:  Correlation between model deviation and the amount of blowing carbon ..................... 130 
Figure 186:  Heating enthalpy of nitrogen and oxygen (p = 1 atm) .................................................... 130 
Figure 187:  Comparison between calculated electrical energy demand and actual consumption for
each operating pattern at the AM Differdange furnace (cf. Table 24) ........................... 131 
Figure 188:  Scheme of the relationships among the models in order to simulate the whole EAF
Consteel® process and to build the control system. ....................................................... 132 
Figure 189:  Scheme of the logic of control system ............................................................................ 133 
Figure 190:  Criteria followed by control system for carbon content and temperature, during charging,
acting on oxygen lancing ................................................................................................ 134 
Figure 191:  Print screen of the developed interface to follow the bath status.................................... 135 
Figure 192:  Criterion for iron oxide control ....................................................................................... 135 
Figure 193:  Parameters table to set the following parameters: time frequency for evaluation, cooling
rate threshold for the bath, target temperature, target carbon content, upper and lower
tolerance for temperature and carbon (four lines), kg of FeO not to be exceeded .......... 136 
Figure 194:  Print screen of the warning messages for the operator ................................................... 137 
Figure 195:  Monitor display of the off-gas compositions used during the plant trial ........................ 137 
Figure 196:  Decrease of the oxygen injection during the plant trial phase 1 and phase 2 ................. 138 
Figure 197:  Calculated and actual oxygen flow rate of PC-Injector 1 based on the off-gas composition
........................................................................................................................................ 140 
Figure 198:  Electrical energy and post-combustion oxygen consumption per tonne liquid steel of 70
heats ................................................................................................................................ 141 
Figure 199:  Linear-proportional control curve with changed parameters m1, m2, n1 and n2 .............. 141 
Figure 200:  Electrical energy and post-combustion oxygen consumption with improved parameters
........................................................................................................................................ 142 
Figure 201:  Off-gas measurements and post-combustion with automatic ON mode......................... 143 
Figure 202:  Results of the trial automatic ON vs. automatic OFF with shortened operation time .... 143 
Figure 203:  Calculated dust load and transmission during one heat .................................................. 144 
Figure 204:  OPC-Server data processing ........................................................................................... 145 
Figure 205:  Chart of the calculated heat temperature with the measured temperatures and the signals
charging and tapping....................................................................................................... 145 
Figure 206:  Chart of the off-gas temperature and concentrations of CO and O2 ............................... 146 
Figure 207:  Calculated melt temperature vs. first plausible measured melt temperature .................. 147 
Figure 208:  Calculated melt temperature adapted to previous reasonable measured ones vs. further
plausible measured melt temperature.............................................................................. 147 
Figure 209:  Energetic cost per ton of steel produced for each heat for furnace 1 and 2. The moving
averages of the cost are plotted on this figure................................................................. 149 
Figure 210:  Energetic cost per ton of steel produced for each operating pattern ............................... 149 
Figure 211:  PC ratio averages (with oxygen measured averages inferior to 15 %w/s), adjusted with
the parameter alpha, function of operation patterns. Note that the PC ratio for the
reference period is not represented because we don’t have an off-gas analysis of this
period .............................................................................................................................. 150 

160
Figure 212:  Scrap yield for each operating pattern ............................................................................ 150 

161
4 List of Tables
Table 1: Production characteristics of EAFs at industrial partners ........................................................ 12 
Table 2: BFI formula for electrical energy demand of arc furnaces....................................................... 22 
Table 3: Average input values of MH heats for BFI statistical model for the electrical energy demand23 
Table 4: Slag analysis of stainless steel heats during measurement campaign ...................................... 32 
Table 5: Result of off-line test ................................................................................................................ 35 
Table 6: Input data for BFI statistical model of campaign 4 with 400 vs. 300 m³ (STP)/h PC oxygen per
injector .............................................................................................................................. 64 
Table 7: Input data for BFI statistical model of campaign 6 with 400 vs. 200 m³ (STP)/h PC oxygen per
injector .............................................................................................................................. 65 
Table 8: Input data for BFI statistical model of different campaigns at the MH furnace ....................... 69 
Table 9: Comparison between two time periods with and without dynamic PC-control ....................... 76 
Table 10: Model calculation for sensitivity analysis .............................................................................. 84 
Table 11: Production expressed in terms of target slag during the baseline and performance period ... 96 
Table 12: EFSOP® PC CLC results (Best month achievement over the trials period) .......................... 96 
Table 13: Modifications of the operation pattern made by the EAF engineer ..................................... 111 
Table 14: Input data for BFI statistical model of campaign with different operating modes for the door
oxygen lance ................................................................................................................... 120 
Table 15: EAF characteristics when using lime instead of limestone .................................................. 123 
Table 16: Input data for BFI statistical model of campaigns at AM Differdange furnace with different
operating modes for post-combustion oxygen ................................................................ 124 
Table 17: Input data for BFI statistical model of campaign with different operating modes for the gas
burners ............................................................................................................................ 125 
Table 18: Various EAF operation modes in each measurement campaign .......................................... 128 
Table 19: EAF and off-gas properties in each measurement campaign ............................................... 128 
Table 20: Input data for BFI statistical model for three time periods in 2009 ..................................... 129 
Table 21: Result of the slag analysis after the steel and slag tapping (with the decarburization control,
plant trial phase 1)........................................................................................................... 139 
Table 22: Result of the slag analysis after the steel and slag tapping (with the decarburization control,
plant trial phase 2)........................................................................................................... 139 
Table 23: The average value of CaO, FeO and Cr2O3 concentration in the slag without and with the
decarburization control ................................................................................................... 140 
Table 24: Modifications of the operation pattern made by the EAF engineer ..................................... 148 
Table 25: Comparison of off-gas analysis systems .............................................................................. 153 
Table 26: Publications and conference presentations regarding the OFFGAS project ........................ 154 

162
5 List of References
[1] Adams, W.; Alameddine, S.; Bowman, B.; Lugo, N.; Paege, S.; Stafford, P.: Total energy
consumption in arc furnaces, MPT International 6/2002, p. 44-50
[2] Bekker, J. G.; Graig, I. K.; Pistorius, P. C.: Modelling and simulation of an electric arc
furnace process, ISIJ International, Vol. 39 (1999), No. 1, p. 23-32
[3] Brod, H.: Verfahren zur Unterdrückung von überhitzten Stellen in kühlmitteldurchflossenen
Wand- oder Deckenelementen eines Elektrolichtbogenofens, Patent DE 3103883 C2,
published 20.01.1983
[4] CRM: Enquiry concerning the European furnace performances, organized by CRM, not
published
[5] Dietrich, A.; Kaspersen, P.; Sommerauer, H.: Laser Analysis of CO and Oxygen in off-
gases, Proc. 7th Europ. Electric Steelmaking Conf., 26-29 May 2002, Venice, Italy.
[6] Guo, D.; Irons, G.: Modeling of radiation intensity in an EAF, Proc. of 3rd International
Conference on CFD in the Minerals and Process Industries, Melbourne, Australia, 10-12
December 2003, p. 223-228
[7] Guthrie, R.-I.-L.; Tavares, R.-P.: Mathematical and physical modeling of steel flow and
solidification in twin roll/horizontal belt thin strip casting machines, Proc. of International on
CFD in Mineral and Metal Processing and Power Generation, Melbourne, Australia, 3-4 July
1997, p. 41-54
[8] Jones, J.; Oliver, J.: A review of post-combustion in the EAF, 5th European Electric
Steelmaking Conference, June 19-23, 1995, Paris.
[9] Kirschen, M.; Pfeifer, H.; Tang, X.: CFD-simulation der CO-Nachverbrennung für
Lichtbogenöfen, 14. FOGI-Seminar, Essen, 04.-05.12.2002
[10] Köhle, S.: Improvements in EAF operating practices over the last decade. Proc. 57th Electric
Furnace Conference, Pittsburgh 1999, p. 3-14
[11] Köhle, S.: Recent improvements in modelling energy consumption of electric arc furnaces.
Proc. 7th European Electric Steelmaking Conf., Venice, 2002, p. 1.305-1.314
[12] Köhle, S.; Hoffmann, J.; Baumert, J. C.; Picco, M.; Nyssen, P.; Filippino, E.: Improving the
productivity of electric arc furnaces. Technical Steel Research, Final Report for ECSC
Project 7210-PR/132 (1999-2002), ECSC Report EUR 20803, 2003
[13] Nguyen, Q. V.; Edgar, B. L.; Dibble, R. W.: Experimental and Numerical Comparison of
Extractive and In Situ Laser Measurements of Non-Equilibrium Carbon Monoxide in Lean-
Premixed Natural Gas Combustion, Combustion and Flame Vol. 100 (1995), p. 395-406
[14] Oeters, F.: Metallurgy of steelmaking, Verlag Stahleisen mbH, Düsseldorf, 1994
[15] Thomson, M.J.; Kournetas, N. G.; Evenson, E.; Sommerville, I. D.; McLean, A.; Guerard, J.:
Effect of oxyfuel burner ratio changes on energy efficiency in electric arc furnace at Co-steel
Lasco, Iron and Steelmaking, Vol. 23 (2001), No. 3, p. 266-272
[16] Development of operating conditions to improve chemical energy yield and performance of
dedusting in Airtight EAF. Draft Final Report for ECSC Project 7210-PR/328 (2002-2005).
[17] Dynamic control of EAF burners and injectors for carbon and oxygen for improved and
reproducible furnace operation and slag foaming. Mid-Term Report for RFCS Project RFS-
CR-03031 (2003-2007), March 2005.

163
6 Abbreviations
AC/DC Alternating current / direct current
AM Arcelor Mittal
BFI VDEh - Betriebsforschungsinstitut
BOF Basic oxygen furnace
CFD Computational fluid dynamics
CID Charge ID
CLC Closed loop control
CRM Centre de Recherches Metallurgiques
CSM Centro Sviluppo Materiali
DB Database
DEC Direct exhaust control
DEWG Deutsche Edelstahlwerke
EAF Electric Arc Furnace
EFSOP Expert furnace system optimization process
HMI Human machine interface
ki Kinetic constant [mol/s]
mi specific mass of element i [kg/t]
MFR Mass flow rate
MH Marienhütte
NG Natural gas
(N)IR (Near) infrared
OPC Operations planning and control
ORI ORI Martin – Acciaieria E Ferreira di Brescia
PA Profilarbed
PC Post combustion
PLC Programmable logic controller
RWTH Rheinisch Westfälische Technische Hochschule
SCADA Supervisor control and data acquisition
STP Standard conditions for temperature and pressure (NIST)
T Task
TKN Thyssen Krupp Nirosta
VIS Visual
Vn Volume in STP [m³]
VOD Vacuum oxygen decarburization
WE Electrical energy consumption [kWh/t]

164
WR Calculated electrical energy demand [kWh/t]
WP Work Package

165
European Commission

EUR 25048 — Improved EAF process control using online offgas analysis — OFFGAS

H. Pfeifer, T. Echterhof, V.Y. Risonarta, L. Voj, H.-P. Jung, S. Lenz, C. Beiler, H.-H. Ballewski, H.
Mees, B. Kleimt, R. Pierre, H.-J. Krassnig, F. Cirilli, U. De Miranda, M. Pustorino, P. Nyssen, D.
Borenstein, C. Ojeda, E. Abreu, P. Simon, B.Vanderheyden

Luxembourg: Publications Office of the European Union

2011 — 165 pp. — 21 × 29.7 cm

Research Fund for Coal and Steel series

ISBN 978-92-79-22160-6

doi:10.2777/18439

ISSN 1831-9424
Interested
Interestedin
inEuropean
Europeanresearch?
research?

RTD
RTD info
info isis our
our quarterly
quarterly magazine
magazine keeping keeping you you inin touch
touch with
with main
main developments
developments (results,
(results, HOW TO OBTAIN EU PUBLICATIONS
programmes,
programmes, events,
events, etc.).
etc.). ItIt isis available
available inin English,
English, French
French and
and German.
German. AA free
free sample
sample copy
copy Free publications:
or
orfree
freesubscription
subscriptioncan canbebeobtained
obtainedfrom: from:
• via EU Bookshop (http://bookshop.europa.eu);
Directorate-General
Directorate-GeneralforforResearch
Researchand
andInnovation
Innovation • at the European Union’s representations or delegations. You can obtain their
Information
Informationand
andCommunication
CommunicationUnit
Unit contact details on the Internet (http://ec.europa.eu) or by sending a fax
European
EuropeanCommission
Commission to +352 2929-42758.
1049
1049Bruxelles/Brussel
Bruxelles/Brussel
BELGIQUE/BELGIË
Priced publications:
BELGIQUE/BELGIË
Fax
Fax+32
+32229-58220
229-58220 • via EU Bookshop (http://bookshop.europa.eu).
E-mail:
E-mail:research@ec.europa.eu
research@ec.europa.eu
Internet: Priced subscriptions (e.g. annual series of the Official Journal of the
Internet:http://ec.europa.eu/research/rtdinfo.html
http://ec.europa.eu/research/rtdinfo.html
European Union and reports of cases before the Court of Justice
of the European Union):
• via one of the sales agents of the Publications Office of the European Union
(http://publications.europa.eu/others/agents/index_en.htm).

EUROPEAN
EUROPEANCOMMISSION
COMMISSION
Directorate-General
Directorate-Generalfor
forResearch
Researchand
andInnovation
Innovation
Research
Research Fund for Coal and SteelUnit
Fund for Coal and Steel Unit

Contact:
Contact:RFCS
RFCSpublications
publications
Address:
Address: EuropeanCommission,
European Commission,CDMA
CDMA0/178,
0/178,1049
1049Bruxelles/Brussel,
Bruxelles/Brussel,BELGIQUE/BELGIË
BELGIQUE/BELGIË

Fax
Fax+32
+32229-65987;
229-65987;e-mail:
e-mail:rtd-steel-coal@ec.europa.eu
rtd-steel-coal@ec.europa.eu
 EU
The OFFGAS project has been focused on investigating the potential for process

KI-NA-25048-EN-N
optimisation and efficiency increases in the electric steelmaking process regarding
oxygen injection and energy transfer to scrap and melt at the EAF. A special empha-
sis has been put onto the application of permanent off-gas analysis systems installed
at the furnaces investigated. The investigations were carried out by a consortium of
research institutes (RWTH, BFI, CSM, CRM) and industrial partners (DEWG, TKN,
MH, ORI, Tenova) representing in total five electric steelmaking plants with a wide
range of produced steel grades and EAF technology.

The off-gas analysis systems in combination with process and energy models allow
for a continuous online monitoring and control of the EAF melting process and provide
an important tool for the steel plant workers in order to optimise the EAF process with
respect to optimal efficiency of oxygen lancing, carbon injection and addition, energy
transfer either from the arc and from the gas phase to the melted pool (post-combustion,
scrap preheating). The technical objective, to significantly increase the availability and
applicability of the off-gas signals for online assessment and control of post-combustion

Improved EAF process control using online offgas analysis — OFFGAS

and for comprehensive EAF energy monitoring, has been achieved. Additionally, EAF
process models that were up to now mainly used for off-line process assessment of
various aspects of EAF steelmaking, have been tested, developed and applied for online
EAF process and energy control.

Improved EAF process

control using online offgas
analysis — OFFGAS

EUR 25048

doi:10.2777/18439