MODELLING OF THE BLAST FURNACE PROCESS WITH A VIEW TO OPTIMISE THE

STEEL PLANT ENERGY SYSTEM

Christer Ryman (1), Mikael Larsson (2), Tommi Niemi (1) and Carl-Erik Grip (2,3)

(1) MEFOS – Metallurgical Research Institute AB, Box 812, SE-971 25 Luleå, Sweden
(2) Luleå University of Technology, Div of Energy Engineering, SE-971 87 Luleå, Sweden
(3) SSAB Tunnplåt AB, SE-971 88 Luleå, Sweden

Abstract

A new process integration tool, based on mixed integer linear programming, has been
developed and used to analyse the energy and material system at the steel plant system SSAB
Tunnplåt AB in Luleå, Sweden (SSAB). The modelled steel plant system also includes an
adjacent CHP (Combined Heat and Power) plant. A one-dimensional static blast furnace
model consisting of a heat- and a mass-balance model with a user friendly web interface has
been developed as a part of this work.

This paper gives an overview of the process simulation model for the blast furnace and
describes in particular the interaction with the mixed integer linear programming based
process integration method. Calculated blast furnace operation practices, originating from BF
operational data from SSAB are used as input to the global analysis model for process
integration calculations. The analysis approach, using a BF process simulation model and a
mathematical site optimisation model is then exemplified with calculations for the
corresponding energy use and CO2 emission for the system. The described methods can be
useful tools for process optimisation and evaluation.

Key words: Blast furnace, Steel plant, Energy system, Model, Optimisation,
Process integration
1. INTRODUCTION

1.1. Background

In the ore based steel production, the blast furnace has a significant role both from energy and
environmental point of view. Due to the use of coal based reductants, the blast furnace counts
for 60-70% of the total energy conversion for the typical integrated steel production site. The
blast furnace process is also the largest source of CO2 emissions from the blast
furnace/oxygen furnace production route. Although large efforts have been paid to develop
alternative iron making processes during the last decades, the blast furnace stands as the
dominating reduction process and is still undergoing positive development towards less
energy use and lowered CO2 emissions.

Historically, and to some extent this is still valid, the possibility to confirm and control several
of the most important process parameters such as the position and distribution of the reaction
zone, flow of solids, gases and liquid phases, the extension of the coke column etc has been
limited. For this reason a number of different process models have been brought forward to
explain and analyse which phenomena’s that occur in the process. The heat and mass balances
for the Fe-O-C system was early explained in diagrammatic form by Reichardt, and later by
Rist [1-2]. These charts were later developed to what today is known as the C-DRR diagram
for description of the working point of the process in comparison to the theoretical limitations.
In the 1960-ies professor Muchi in Japan started research based on mathematical analysis of
the blast furnace process based on transport phenomena [3]. When computers became
available much more complex calculations could be solved and the simulation models were
further developed. The present model, consisting of conservation equations on momentum,
energy and chemical species for all phases considering viscosity, specific heat and more was
recently described by Yagi [4].

For on-line applications, process optimisation and operation comparison it is difficult to use
too complicated models. The models contain vast amounts of information, including hundreds
of more or less disordered parameter values. It is therefore time consuming to interpret and
evaluate the results. The use of models based on staged heat- and mass-balances for process
optimisation and case studies are therefore justified. The latter method was therefore selected
in this work.

1.2. Studied system

In industrial systems, such as the steel plant, several processes are connected together. A
change in one process unit may result in unpredicted changes in other parts of the system. The
steel plant system SSAB consist of a coke oven plant, one blast furnace (BF), two basic
oxygen furnaces (BOF), and two continuous casting systems (CC). The process gases from
coke ovens, BF and BOF´s are only partly utilised within the steel production processes and
therefore arise as a surplus in the system. An adjacent 320 MW CHP plant is operating with
the process gases as primary fuel. Since the year 2000 the SSAB plant is operating one blast
furnace, replacing the earlier two-furnace operation. The new BF No 3 has a heart diameter of
11.4 m, a working volume of 2540 m3 and the production during 2003 was close to 2.3 Mton
of hot metal. The BF is operated on a 100% pellet burden with pulverised coal injection. The
high iron of the burden materials gives low slag volumes, typically in the range 150-170
kg/thm, and reductant rates in the range 460-480 kg/thm. The final product from the steel
plant is slabs which are transported to rolling mills situated 800 km south of Luleå.

2. METHOD

2.1. Development of a new BF simulation tool

A one-dimensional static blast furnace model, consisting of a mass balance and a heat balance
model has been developed. The main principle is to divide the furnace in upper and a lower
zone, which enables to determine the heat losses separately in each zone. In the upper zone
only gas phase reactions and reaction between gaseous and solid phases occur. No solid
carbon is consumed, and it is assumed that all top charged iron oxides are reduced to FeO1.05.
In the lower zone all iron oxides from the upper zone and possible iron oxides injected in the
tuyere level are reduced to liquid iron. The interface between the two zones is defined by the
thermal reserve zone, for a general description see Figure 1. Initially this calculation method
was used in a spreadsheet model developed at SSAB. The calculation method is based on
mass-/energy balances and carbon-/oxygen balances which are solved iteratively [5]. The
calculation algorithms then have been transformed to object oriented code in Java. This was
made to get a clearer method for computation. The new model then could be furnished with
user friendly interfaces for in- and output of process data. The graphic user interface was
suited for web-access and has been demonstrated as an on-line calculation tool on internet [6].

Iron ore
Reductant (Coke)
Slag former
Gas

Heat loss (q)

UPPER
MODEL
ZONE

RESERVE Coke
Gas
ZONE FeO1.05
Heat loss (q)

LOWER Injected
MODEL reductant
ZONE Hot blast

Liquid iron
Slag

Figure 1. General description of a two-stage BF model

 2.2. BF validation experiments

Three experimental campaigns and one reference campaign were carried out at BF no 3 using
the operational modes briefly described below.

All coke operation (Case 1)

Operation with coke charged from the top to cover the complete carbon need for
the process, and no reductants injected at tuyere level.

Injection of BF flue dusts (Case 2)

Since the blast furnace is operated on pellet burden there is no sinter plant in
operation to recirculate dusts and sludges generated in the process chain. At the
SSAB Luleå site, such materials have instead been recirculated to the BF,
essentially in form of cold bonded briquettes. An alternative practice, applicable
to flue dusts, is to inject them in the tuyere level.

Injection of alternate coal qualities (Case 3)

Trials have been made to evaluate injection coals with different volatility. The
normal practice at the SSAB Luleå site is to use an injection coal with high
volatility, appr. 40%, but evaluation trials have been made with an alternate
quality with considerably lower volatility, appr. 20%. The energy content of the
two coal types were reported to be the same.

Reference operation practice

Operational practice which is representative for the BF involved in this study.
This is operational data close to the average operational practice for the SSAB
blast furnace during 2003.

The operational data was used to verify the BF model, and to perform a further system
evaluation. These experiments were co-ordinated with current process development activities
at SSAB. Operational data are reported in the result section of this paper.

2.3. Total optimisation model, Process integration tool

The common aim of all process integration (PI) methods is to minimise the energy use, the
approach to achieve this varies between different methods. Generally three different types of
PI methods are used, the pinch analysis method, the exergy method and the mathematical
programming method. In the pinch analysis method the heat transfer possibilities within the
system is analysed aiming to minimise the external heating and cooling demand for the
system. With the exergy analysis method, the energy quantity for various streams are analysed
and the losses in the system usually quantified accordingly. With the mathematical
programming method, mathematical description of the system is used in the analysis, often
together with an optimisation routine. A further description how the different PI methods have
been used in Swedish steel industry can be found in [5, 7].

A process integration tool has been developed for the system based on mixed integer linear
programming, and used to analyse the energy and material system at the steel plant system
SSAB. In this case the modelled steel plant system also includes the adjacent CHP plant. The
main product from the system is the cast slabs.
The different processes have been modelled separately and connected together by each
primary product and by-product interactions. The model sophistication varies between simple
linear process models, empirically derived process relations based on engineering practice,
and models based on mass and energy balances for the different processes. The driving force
for the model is the production of the final product from the system, i.e. first-rate steel slabs.
Each sub-process, identified in Figure 2, is linked to the next processing step by the primary
product from each process i.e. coke, hot metal (HM), and liquid steel (LS). The steel demand
from the slab-casting units will determine the production rate in the BOF, which in turn will
determine the hot metal rate for the BF and so forth. The external material use is based on the
process requirements for each sub-process. The consumption and excess of by-products are
also determined from each sub-process model. The arrows indicate the main product flow
within the model. The efficiencies and specific process-related data are based on actual
production data from the steel plant and the CHP plant. The process integration model has
been further described elsewhere [8].

Figure 2. Schematic model description of steel plant system

2.4. Integration of BF simulation results into total analysis model

The BF operation practices calculated from the BF simulation model is converted and
included into the total analysis model as different BF operation cases. The calculated ingoing
resources and outgoing hot metal production are coupled together for each case. The produced
rest products (BF slag and flue dust) and recovered energy in the top gas are also coupled to
the hot metal production. The feasible region in which the BF can be operated is limited
within these cases. Totally 10 different cases are possible to include in the optimisation, from
which the optimal operation practice is chosen either as one of the cases or as a mixture of the
different cases. This implies that the region between the different cases are analysed as well.
The use of different BF cases enables the possibility to study the effect of different specific
changes in the BF on the total steel production system.

2.5. Objective function

The process integration tool uses an optimisation routine to analyse different aspects on the
system. The objective function defines what to be minimised in the system, e.g. production
cost, emissions or energy use. The choice of calculation philosophy and system boundary will
affect how the objective function is defined. Principally three different calculation
philosophies can be used in the analysis; a) Inventory analysis, b) Effect-oriented analysis,
and, c) Life-cycle assessment (LCA) analysis. Which of these calculation principles to be
used is depending on how the results from the analysis are going to be used. The different
principles differ mainly in respect to how the “costs” (monetary, emissions or energy use) are
included. Direct “cost” from a system are those physically originating from the system.
Indirect “costs” are affected by the operation but are accounted elsewhere (e.g. other
companies, systems) outside the analysis border.

Effect oriented analyses have been used to analyse the global effect of the changes in the steel
plant. Indirect emissions from raw material preparation and rest material use outside the steel
plant are included. The system boundary includes the total system Steel plant/Power plant.
Two objectives have been defined for in the analysis, Energy and CO2 (emission), see Table
1.

Table 1. Coefficients in the objective function

Energy CO2 Energy CO2
[GJ] [tonne] [GJ] [tonne]
Included debits Included credits
Coking coal [t/h] [27.0 - 35.9] [ 2.0 – 3.2] Coke breeze [t/h] -40.90 -3.62
External coke [t/h] 40.9 3.7 By-product: Tar [t/h] - -3.3
PCI coal [t/h] [28.2 – 29.3] [2.8 – 3.0] Sulphur [t/h] - -
Transport coal/coke [t/h] - 0.051 Benzene [t/h] - -3.3
Pellet [t/h] 0.3 0.1 BF slag [t/h] - -
External Scrap [t/h] - 0.019 BOF slag [t/h] - -
Internal scrap [t/h] - 0.019 Power [MWh] -3.6 -0.6
Pig Iron [t/h] - - Heat [MWh th] - -
Mn ore [t/h] - -
Limestone [t/h] - 0.4
Lime [t/h] - 0.4
Dolomite [t/h] - 0.4
Quartz [t/h] - -
Briquettes [t/h] 3.9 0.4
Power [MWh] 3.6 0.6
Oil [t/h] 46.2 3.1

3. RESULT AND DISCUSSION

3.1. Translation from spreadsheet to sequential programme code

The existing spreadsheet model consisted of heat- and mass-balances and the equations were
solved using the built-in iteration facility in the spreadsheet software. Translation of the heat-
and mass-balances into Java code were comparatively straight forward, the iterations on the
other hand, was a more difficult matter. The Java program needs exact instructions on each
iteration and also in which order they are to be performed. This turned out to be a difficult
task. The spreadsheet program itself chooses the road of iteration and it is almost impossible
to follow. The formulation of these routines used a large effort and became the largest
obstacle in the time schedule of the project.
 3.2. BF model validation

The BF simulation model has been validated against four defined operational practices
summarised in Table 2. The parameters in the input rows have been measured and the output
rows are both measured and calculated using the model. The agreement between measured
and calculated parameters has been verified to be consistent.

The model accuracy depends upon the establishment of a correct mass balance. This calls for
input data that are as correct as possible (analysis, temperature, heat capacity etc). Especially
the quantity of the elements O, N, H, and C are of importance for the oxygen and carbon mass
balance. The model is in principle thermodynamically correct; it uses established
thermodynamic data, but as in many models there are also a few empirically established
relations.

Table 2. Validation of blast furnace model

Reference All coke BF flue dust Low volatile
operation operation, injection, coal injection,
practice,
Parameter Unit Ref. Case 1 Case 2 Case 3
Input:
Coke kg/thm 327 478 325 326
PCI kg/thm 145 - 141 143
Flue dust in PCI % - - 10 -
O2 enrichment % 4.50 - 4.04 4.70
Moisture g/Nm3 11.70 29.60 5.90 11.50
Blast temp. °C 1101 1049 1099 1089
Output: Model Model Model Model
Blast flow Nm3/thm 938 941 1147 1198 941 935 903 928
Top gas (BFG) flow Nm3/thm 1465 1470 1642 1646 1455 1461 1412 1417
Top gas (BFG) heat MJ/Nm3 3.06 3.13 2.66 2.65 2.97 3.04 3.00 3.03
Top gas (BFG) energy GJ/thm 4.49 4.60 4.37 4.35 4.32 4.44 4.24 4.30
Top gas (BFG)-ηCO % 55.20 55.00 54.90 54.80 55.60 55.50 55.70 55.50
Slag volume kg/thm 166 166 146 145 171 172 178 178
LKAB pellets kg/thm 1365 1367 1372 1372 1368 1369 1354 1354
Briquettes kg/thm 58 58 37 37 60 60 72* 72*
*) Divergent composition of recycled material (higher iron content)

A conclusion from the validation of the model is that it represents the SSAB BF No 3 with
such a standard that it can be used to calculate different operating practices. The model can be
used to compute the hot metal and slag and to assess the heat balance for compositions for a
given burden and reductant mix. This can be useful for process optimisation by the possibility
to calculate effects of modified operation practice. It can also be used in forecasting
experimental work, such as BF trial planning. Another application of the model is to use it as
an input source for total analysis of the production site by means of process integration, as
developed further in this study.

The new operational practices described by case 2 and 3 proved to be very interesting. An
objective with injecting of flue dusts is to get a more efficient use of the oxides and coal by
injecting them in the high temperature zone of the furnace. The consumption of reducing
agents was unchanged or slightly decreased compared to normal operation, but equipment
problems occurred. This practice was evaluated and reported by SSAB recently [9].
The injection of low volatile coal also indicated a consumption of reducing agents at
unchanged or slightly decreased levels compared to normal practice. This can be an indication
of better replacement ratio for the tested coal compared to the normal quality.

The calculated values from the validation study has been further analysed in the total
optimisation system to analyse the effect in case the total system is considered.

3.3. Effect of BF operation on total system

The operation practice of the BF will affect the overall mass- and energy balance for the
system. The three different operation practices have been analysed in three simulation cases
(1-2-3) and four optimisation cases (a-b-c-d). As basis for comparison a reference case (Ref.),
simulating the normal operation practice for the unit’s coke oven, BF and BOF, has been
included. The different BF operation practices are calculated as described in section 3.2. In
the optimised cases, a-d, the system is optimised based on energy and a CO2 emission
objective function. In case:
a) The different BF cases (ref, 1, 2, 3) is optised based on a CO2 emission minimisation
objective function. Only changes in the BF operation practice are allowed.
b) The total system including the other processes (coke oven and BOF) are optimised
based on a CO2 emission minimisation objective function.
c) Minimisation of the energy use, else as a).
d) Minimisation of the energy use, else as b).

3.4. Outcome of optimisation calculation

The result from the different calculations are shown in Figure 3. The specific energy use and
CO2 emission from integration of the different BF cases are shown in cases 1-2-3. The
reference case is for comparison. For these cases boundaries have been introduced to govern
the system, resulting in a simulated energy and CO2 emission. The optimisation routine is not
enabled. The cases a-b-c-d includes optimisation in the calculations.
[10 GJ/ton slab; 10 kg CO2/ton slab]

2.0 Specific energy use

Specific CO2 emission
1.9

1.8
3

1.7

1.6

1.5
1

1.4
a) c)
Case 1

Case 3

b) d)
Case 2
Ref.

Figure 3. Specific energy use and CO2 emission including the emission from the CHP
The different BF operation practices, cases 1-2-3, are compared with the reference case, Ref.
It can be seen that integration of the all coke operation practice, case 1, significantly increases
both the specific energy use and the CO2 emission. By integrating the BF flue dust injection
operation practice, case 2, the both the specific energy use and CO2 emission can be
decreased. By operating the BF with low volatile coal injection, no significant change can be
seen. The lowest energy use and CO2 emission are found from the optimised case d) and b).
For these cases the BF is operated with BF flue dust injection, the coke oven is operated with
a decreased volatile matter (VM) in the coking coal mix, the ratio HM/scrap in the BOF
converter is decreased. The amount of pig iron is decreased. The specific changes proposed in
the different units are shown in Table 3.

Table 3. Summarised calculation results – System analysis (fixed ton/h production rate)
← →
Simulation Optimisation
Coke ovens Unit Ref. Case 1 Case 2 Case 3 a) b) c) d)
VM (in the coking coal mix) wt% 24.8 24.8 24.8 24.8 24.8 22.2 24.8 22.0
Ash (in the coking coal mix) wt% 8.4 8.4 8.4 8.4 8.4 9.8 8.4 9.3
Coke yield factor (dry) - 1.278 1.278 1.278 1.278 1.278 1.245 1.278 1.243
Coke oven gas (external COG) knm3/h 20.0 20.0 20.0 20.0 20.0 19.4 20.0 18.9
Heat value MJ/nm3 18.0 18.0 18.0 18.0 18.0 18.0 18.0 18.0
BF
BF cases used
Ref. % 100 0 0 0 0 0 0 0
Case 1 % 0 100 0 0 10 0 0 0
Case 2 % 0 0 100 0 90 100 100 100
Case 3 % 0 0 0 100 0 0 0 0
Pellet kg/thm 1367 1372 1369 1354 1370 1369 1369 1369
BOF slag kg/thm 42 31 39 42 38 40 40 40
Briquette kg/thm 58 37 60 72 58 60 60 60
Coke kg/thm 327 478 325 326 340 325 325 325
PCI kg/thm 145 0 141 143 127 141 141 141
BF flue dust injected % in PCI 0 0 10 0 10 10 10 10
Blast knm3/thm 0.942 1.199 0.936 0.928 0.962 0.936 0.936 0.936
O2 % 4.5 0.0 4.0 4.7 3.5 4.0 4.0 4.0
3
BF gas (external BFG) knm /h 256.2 165.1 250.9 241.8 242.5 231.0 250.9 231.0
Heat value MJ/nm3 3.13 2.64 3.03 3.03 3.01 3.03 3.03 3.03
BOF
HM kg/tls 907 907 907 907 907 861 907 861
Scrap incl. pig iron kg/tls 159 159 159 159 159 210 159 210
Pellet kg/tls 16 16 16 16 16 0 16 0
3
BOF gas (external BOFG) knm /h 28.7 28.7 28.7 28.7 28.7 27.5 28.7 27.5
Heat Value MJ/nm3 7.0 7.0 7.0 7.0 7.0 7.0 7.0 7.0
CHP
BF gas % 70 51 66 64 63 62 66 62
BOF gas % 16 21 16 16 16 15 16 15
Coke oven gas % 15 28 18 21 21 23 18 22
Heat MW th. 80.8 80.8 80.8 80.8 80.8 80.8 80.8 80.8
Power MW el. 90.1 66.1 90.1 90.1 90.1 88.9 90.1 88.1
Condense/ back pressure - 1.54 0.86 1.54 1.54 1.54 1.50 1.54 1.48
← →
Simulation Optimisation
For the optimised cases b) and d), the total coke use for the system is decreased which results
in a decreased need for external coke. A decreased volatile mix in the coking coal mix results
in a slightly decreased coke oven gas production and an increased internal coke production,
which also decreases the demand for external coke. The BF is operated with the BF flue dust
injection operation practice. In the BOF the scrap rate is increased by decreasing the use of
pellet as cooling agent. The ration HM/scrap is decreased. In the CHP system only minor
changes regarding the distribution of the fuel gases resulting in changed electricity production
are shown.

3.5. Applicability for other systems

The total analysis model has been made to fit the existing steel plant system at SSAB, and
have therefore been tested and validated in various studies before. By using the already
developed models, the analysis including the calculation with the BF simulation model and
integration calculations for the system can usually be conducted within two working days.
The time effort is of course depending on required changes in the model and the experience
from the users.

If the same model is going to be used for other systems, time is needed to adapt and validate
the model and model results. The work load is depending on the modelling work needed for
inclusion of further processes (e.g. rolling mills). However, the costs for this effort can be
considered to be less than the cost and risk of alternative work on experiments and tests.

4. CONCLUSIONS

A model for blast furnace simulation has been developed. The model is based on a two-stage
heat- and mass-balance with a web-based user interface and can be used within a short
learning time. The model has been validated by full-scale blast furnace trials at SSAB with
good agreement. This type of model can be used for optimisation of the blast furnace process
e.g. by comparison of different operational modes and also for planning of experimental
work.

The model has been used as an input source for total analysis of the production site by means
of process integration. The integration of the blast furnace model into a total optimisation
model results in suggestion of new operational practices not achieved always achieved by
straight forward simulation alone. The performance of the total system is always highly
depending on the interaction between the different unit processes. By use of optimisation
models a production system can be optimised with a limited effort and time.

5. ACKNOWLEDGMENT

This study has been supported by the Swedish National Energy Administration, SSAB
Tunnplåt AB and LKAB.
6. REFERENCES

1 P. REICHARD, Arch. Eisenhüttenwesen, 1(1927), pp. 27-101.

2 A. RIST, N. MEYSSON, N. J. Met. 1967, pp. 50-59.
3 I. MUCHI, Transactions ISIJ, 1967, pp 711-728.
4 J. YAGI, M. CHU, H. NOGAMI, Numerical investigation of blast furnace performance under innovative
operations, Scanmet II – 2nd International Conference on Process Development in Iron and Steelmaking,
Luleå, Sweden, Vol. 1, pp 93-102.
5 C.E. GRIP, M. LARSSON, J. DAHL: Energy Optimization by Means of Process Integration in an Integrated
Steel Plant with Surrounding Community, 84th Steelmaking Conference Proceedings, ISS, Warrendale, PA,
(2001), pp. 543.
6 Kobolde & Partners AB, www.raceway.nu, 2003 (in Swedish).
7 C.E. GRIP, A. THORSELL, Swedish national research programme for energy saving by means of process
integration, Scanmet II – 2nd International Conference on Process Development in Iron and Steelmaking,
Luleå, Sweden, Vol. 1, pp 407-416.
8 M. LARSSON, J. DAHL, Reduction of the specific energy use in an integrated steel plant – The effect of an
optimisation model, ISIJ International, Vol.43, No.10, 1664, (2003).
9 B. JANSSON, L. SUNDQVIST ÖKVIST: Injection of BF flue dust into the BF a full-scale test at BF No 3
in Luleå, Scanmet II – 2nd International Conference on Process Development in Iron and Steelmaking, Luleå,
Sweden, Vol. 2, pp 265-274.