See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/250159731

Modeling and Simulation of an Electric Arc Furnace Process

Article  in  ISIJ International · January 1999

DOI: 10.2355/isijinternational.39.23

CITATIONS READS

102 1,632

3 authors, including:

Ian K. Craig P. Christiaan Pistorius

University of Pretoria Carnegie Mellon University
132 PUBLICATIONS   1,376 CITATIONS    247 PUBLICATIONS   3,304 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Cooling and sizing behaviour of high titania slags View project

Titanium oxycarbide View project

All content following this page was uploaded by P. Christiaan Pistorius on 18 June 2014.

The user has requested enhancement of the downloaded file.

 ISIJ International. Vol. 39 (1999), No. 1, pp. 23-32

Modeling and Simulation of an Electric Arc Furnace Process

Johannes Gerhardt BEKKER.Ian Keith CRAIGand Petrus Christiaan PISTORIUS1)

Department of Electric.al and Electronic Engineering, University of Pretoria, O002, South Africa.
E-mail: jgbekker(ctJstarmail.com; icraigCc~ipostino,up.ac.za 1)
Department of Materials Science and Metallurgical Engi-
neering, University of Pretoria, Pretoria, O002, South Africa. E-mail: pistoriC(v)postino.up,ac,za

(Recelved on June 29. 1998, accepted in final form on September 16. 1998)

The model derived in this paper is intended. for the testing of control strategies using off-gas variables in
the control of an Electric Arc Furnace (EAF). The derivation of a multivariable non-linear state
space model
of an EAFprocess is treated. Assumptions are madethat facilitate the modeling effort. First principles of
thermodynamics, together with empirical relationships, are used to derive the model equations. Astate-space
model is developed with physical variables such as carbon content in the steel as model states. Rates of
changesuch as decarburization are also considered. Typical operating conditions for a furnace are discussed
and initial conditions for a simulation are derived from this. Atimeiine of events corresponding to operating
conditions is determined for the simuiation. A simulation is conducted starting with the initial conditions
and proceeding with the line of events. The results of the simulation are shown and discussed in terms of
the purpose of the model,

KEYWORDS:
electric arc furnace; EAFsimulation; off-gas control; state-space model.

control system can either be implemented for the en-

l. Introduction
tire EAFor as a separate control system for the off-gas
The emphasis on the operation of Electric Arc system only.
Furnaces (EAFS) shifted over the last decade to more This work investigates the use of off-gas variables in
efficient and environmentally friendly operation. Accord- the control of an EAF. In contrast, other attempts at
Ing to Jones,i) productivity increases are directly related EAFcontrol have focused mainly on electrode control
to the increased use of oxygen in the EAF, where the (Reuter et al.4)). King and Nyman5)showedhowa neural
exothermic reaction energy replaces a large amount of network is used to control the electrode voltages and
the electric energy input. Advantages of using in-vessel currents of a three-phase arc system.
off-gas post-cornbustion are that it leads to To apply advanced control to a process, a real-time
a potentially
higher scrap charge ratio (Zhulkovsky2)) and lower dynamic process model is needed to approximate the
overall energy waste. Both practices allow shorter tap- conditions in the EAFover a wide range of operating
to-tap times, but it also lead to greater demandson the conditions. A first principle model is preferred because
off-gas system. only limlted measurementsof the dynamic conditions
The off-gas system of an EAF plays an important role inside the EAFare available. Austin et al.6) developed
in EAF operation. It is used to extract and combust a two-dimensional model for blast furnaces that repre-
hazardous gases such as CO, and It also influences the sents the major chemical reactions and physical struc-
steelmaklng process. The rate of gas extraction has a tures. Dynamicmodels and simulatorsare becomingin-
direct influence on the EAF pressure, thus affecting air creasingly popu]ar for
types of furnaces,
all including
entrainment, and hence the process energy requirement EAFs, as process engineers are working to optimize these
and nitrogen pick-up in the steel (Lankford3)). processes.
There are limitations on the allowable temperature Static EAF
mass- and energy-balance models are cited
and pressure of the waste gas where it passes through in the literature (see K6hle7)). In contrast there are
the off-gas system bag-house. Wlth greater demand
on only limited references to dynamic EAF
models in the
present off-gas systems, as a result of economic and en- literature, since
many such models tend to be proprietary.
vironmental consideratlons, it is considered worthwhile Morales et al.8)
are working on a complete mathematical
to investigate the possibility of introducing advanced simulator for EAF operation. Petersohn et al.9) derived
control of the off-gas system. The advantage of an ef- a dynamic EAF model based on mass and energy
fective control system could be balances. They show what the principal measurement
a lower cost to waste
gas treatment, and also improvement in overa]1 opera- points required by a comprehensive model are.
tion of the EAF. This can be achieved without design Sridhar and Lahiri I o) developed modelsto predict EAF
changes, by applying real-time dynamic control. The temperature distributions. Papers that focus on gas EAF
23 O 1999 ISIJ
 ISIJ International. Vol. 39 (1999), No. 1
fiow include Gardin et al.,1 l) which study gas flow above
the liquid steel bath, and Zhulkovsky et al.,2) which Slip-

study in-vessel off-gas post-combustion. Fruehanl2) did Electrodes

a study on the benefits of gas stirring in an EAFto
Furnace roof
investigate interaction between the FeOin slag and the
carbon in the steel melt. Sarma et al,13) provided ex- slag
perimental results of the reduction of FeOin smelting Door side door
slag by graphite. Equrvalent offgas fan
Tap hole
The emphasis of this work is to obtain a model that
is appropriate for closed-loop control system simulation Fig. l. Process description.
and initial controller verification. The model therefore
needs to approximate the physical process in terms of its (3)
&
FromPlockinger Etterichi4) (p. 89 Fig. 7) it is
dynamic response due to control and disturbance input seen that 48 '/o of the arc heat flows to the steel melt by
changes. The aim is to obtain the sametype of response, of
means convection and 170/0 by meansof radiation.
with the same order of magnitude as is found in an There can be heat transfer from the arcs to the scrap,
industrial EAF. but since the greatest fraction of heat flows to the steel
Section 2 treats a process description.
Section gives 3 melt it is more convenient to makethe assumption that
all assumptions and simplifications.Section treats the 4 all heat transfer from the
arcs is to the liquid metal.
non-linear model derivation. Subsections 4, to 4, giveI 15 (4) A
power division coefficient describes heat
the derivations of the state equations and auxiliary transfer from the fluid group to the solid group. The
equations. Section 5
discusses typical initial conditions, heat transferred to the scrap can be utilized to increase
a timelipe of events and operating conditions for the the temperature of the solid group, or to melt part of
EAFselected for simulation in this paper. Section 6 the scrap. The coefficient (xl 3/xl 2) is used for the melting
discusses simulation results. rates of the scrap and the solid slag, while the coefficient
(1 -xl3/xl2) is used for the heating rate.
2. Process Description (5) Fluxes added to the EAF
do not melt imme-
diately, but remain in an unmixed, solid condition
EAFStypically
use three-phase electrical supply to
until melted by the heat of the fluid
produce steel by melting scrap. Wherehot metal (high group to form part
of the liquid slag.
in carbon and silicon)is charged (practiced at a South
African producer) to the EAFtogether with scrap,
steel
(6) Becausethe concentration of Fe in the steel melt
is
very muchgreater than that of carbon or silicon, all
oxygen blowing can lower the concentrations of carbon
and silicon, and provide an additional energy source. oxygen blown into the steel reacts with Fe to form FeO
in the s]ag.
Oxygenis typically blown into the metal both through
one or more side lances.
(7) All FeO and Si02 that form in the EAFare
immediately dissolved in the molten slag.
With the EAFin operation, the roof is closed, and
off-gas is extracted through the duct in the roof. The
(8) A carrier gas blows graphite into the slag to
off-gas system employs an extraction fan to pass the
reduce the FeOcontent. The carrier gas is typically air
and contains oxygen. All graphite is assumedto either
off-gas through a baghouse filtration system. Figure l
gives a schematic description. react with FeOinstantly or with the carrier gas during
For simplicity an equi-
the ascent to the surface. This is based on the result that
valent fan without the baghouse and extraction fan
is shown in Fig. l. Since off-gas should not carry no carbon solid remain in the slag. The typical residence
time of a carbon particle in the slag is 8-lO sec (Morales
combustible gases into the baghouse, air is entrained a!.8)).
c't
through the slip-gap to combust CO. The slip-gap size
varies from 0,1-0.4m. At the end of the batch steel-
(9) Experimental evidence by Sarma et al.13) gives
making process, slag is removed through the slag door
the FeOreaction rate proportional to FeOin the slag.
and steel is tapped through the tap hole.
(lO) The DRI has 93 "/o metalization, and whenDRI
is introduced into the EAF, 93 o/. of the mass is added
3. Assumptions to Facilitate Modeling 7
as iron to the steel melt, and o/o is added as FeOto the
liquid slag.
The following assumptions facilitate the modeling (11) The iron in the DRImelts instantaneously. This
effort, and allow simplification. is based
on the observation that the temperature EAF
(1) The liquid metal, molten slag and gas-phase are can be controlled by matching the DRI feed rate with
all at the
same temperature. These elements form the
fluid group. The mass of gas-phase elements is small in
the electrical power input: 33kg/min per yields a
constant temperature (Maschlankal5)). By implication
MW
relation to the other fluids. the DRI effectively me]ts immediately
upon entry to the
(2) The solid is at the
sametemperature as the solid EAF, with the associated latent heat requirement ex-
scrap. These elements form the solid group. The solid erting its effect immediately.
group melt rates and the rate of increase of the solid (12) The scrap does not contain impurities that will
temperature (xl 3) is proportional to the difference be oxidized. In steel-making practice graded scrap is
group
between the fluid group temperature (xl2) and the solid needed (Geldenhuisl6)) for low tramp e]ement requlre-
group temperature. ments in steel, to prevent high impurity content.

@1999 ISIJ 24
 ISIJ International. Vol. 39 (1999). No. 1
(13) The average value (for temperature from 600 to heat multiplied by the temperature difference between
l K) for the specific heat of iron is used for the scrap.
800 the liquid group and the solid group.
Cp(Fe(s))=0.039kJ/(mol•K). Theequation for the scrap melt rate in units of (kg s~ I
K) for liquid iron.
Cp(Fe(L))=0.046kJ/(mol-
is given by Eq. (1):
'
)
(14) Typical slag additives are 2/3 burnt line and l/3
MF*kth"Ik*...1()cl)(xl2 -xl3)(xl3/xl2)
dolomite (Geldenhuisl6)). Burnt lime contains 97 "/o CaO ......(1)
xi = (AF.
and 3"/. MgO,while dolomite contains 570/0 CaOand + Cp(Fe(s))(xl 2 - xl 3))

43 o/o MgO (Turkdoganl7) pp. 209244). 4.2. Rate of Changeof Liquid Metal Mass
(15) The CaOand MgO parts ofthe slag are lumped
The liquid metal mass (;~2) is determined by six
and treated as a single state variable, with an equivalent
mechanisms.Firstly the melt rate of the steel scrap, same
heat of fusion, heat of solution, molar massand specific
heat. as Eq. (1), but with opposite sign. Secondly the rate of
oxidation of Fe by 02 to FeO. The third, fourth and
(16) All elements coming into the EAFare heated fifth mechanismsare the reduction of FeO to Fe by
to the respective temperatures. Reagents are heated up graphite injection into the slag and by dissolved carbon
to 1680K before they participate in reactions. A11 re-
and silicon in the steel melt. The sixth mechanismis DRI
actions take place at a temperature of 1680 (average
K addition rate, which contains 93 o/o Fe (see assumptions
temperature) to avoid using temperature as variable in
(lO) and (11)). The rate of Fe oxidation by
chemical reaction power. oxygen
injection (see assumption (6)) is given by Eq. (2). The
(17) All energy is exchanged with the fiuid group.
rate of reduction of FeO by the graphite is given by
The solid group receives energy from the fluid group.
Eq. (3). It is proportional to the rate of graphite injection
This meansthat the electric arc power is assumedto heat
and the content of FeOin the slag (see assumption (8)).
up the liquid metal first, and then the liquid metal heats
up and melts the other componentsin the EAF. This is 2MF*dl
(2)
a convenient simplification. Mo, .........
(18) The only gas-phase elements are CO, C02 and
N2' All 02 reacts on entry, and H2 is ignored. x7kg.MF*d5
(3)
(19) The pressure in the EAFcan be obtained from (x6 + x7 + x8)Mc
the ideal gas law.
The rates of FeOreduction by dissolved carbon and
4. Non-linear Model Derivation dissolved silicon in the steel melt are discussed in sub-
sections 4.3 and 4.4. Please refer to these subsections
The EAFis described in a Multi-Input Multi-Output for the development of these two rate equations. Equa-
(MIMO) state space format with 14 states. There are tions (8) and (9) give Xc and Xc'q respectively and
two other vectors besides the state vector x, namely input Eqs. (15) and (16) give Xsi and Xs'iq respectively. Thus
vector u and disturbance vector d. The model is a vector the rate of change of liquid metal is given by Eq. (4):
function (D of 14 ordinary differential equations (ODE)
MF.kth** Ik**.* (xl)(x 3)(xl 3/x 2)
that describes the time rate of change of each state i 12 - x 1 1
variable. Thedifferential equations are mostly non-linear,
and are part of the set that contains all functions of x,
x2 =
- (A.

x7kgrMFed5
F. + Cp(Fe(s))(xl 2- xl 3))
MFe
u, d, i.e. i 9714 * l
= f(x, u, d) where the dimensions are: x e + (x6 +
Mc
kdc(Xc
- Xeq)
c
ue 9?2 * 1, de91:4* 1 + x7 + x8)Mc
See Bekker et al,18) for further detail. The Appendix 2MF. 2MF.dl
contains a list of symbols. + (v
kdsikAsi
-
Xs',q)
- +0.93d2 ""'(4)
Msi M02
4.1. Rate of Changeof Solid Scrap
4.3. Rate of Changein Dissolved Carbon
Scrap melting involves both mass transfer (diffusion
of carbon) and heat transfer. In view of the aim of this
The rate of change of dissolved carbon (;~3) in the steel
melt is due to reaction with FeOIn the slag. The reaction
model a formulation that yields a realistic melt rate based
with FeOin the slag is driven by the difference from
only on heat transfer is acceptable. The solid scrap rate
equilibrium that is approximated as a hyperbolic rela-
of change (;~l) is proportional to the heat transfer rate
tionship between the carbon in the steel melt and the
from the liquid group to the solid group, which is
proportional FeOin the slag (mass percentages). For low-carbon melts
to the temperature difference between the (in which yc=y~), and for basic slags (in which yc~~ l)
two groups. The heat transfer coefficient is k*h*,1
= 0.4 the equilibrium product varies with temperature from
kW•K-1•m~2 that can be adjusted depending on the l.8 at 1500'C to O.89 at 1700'C. Since
observed melting time. The area of contact is propor- 1600'C is close
to the final tap temperature, a constant equilibrium
tional to the mass of scrap in the EAFand a specific
area of k....1 =0.005m2,kg~ I is used that corresponds
product wasdetermined for 1600'C and used throughout
the simulation. The reaction between Cand FeOis far
to sheets of thickness O.05m. A
partition coefficient is
from equilibrium throughout the steelmaking process,
used to allocate power transferred to melt the scrap, as
and therefore the exact value of the equilibrium constant
opposed to power transferred to heat the scrap, as in is not critical. At 1600'C Turkdoganl7) (Eq. 6.6, p. 183)
assumption (3). The scrap that is being melted must
gives equilibrium by Eq. (5).
receive latent heat of fusion (~F. (kJ/mol)) and specific

25 (c) 1999 ISIJ

 ISIJ International, Vol. 39 (1999). No. 1
(o/oFeO)(o/oC)*q~ilibr*um l 25=kc% "-' "-""
(5) fraction. These are given by Eqs. (15) and (16) re-
spectively.
The relationship between kc'/, and the molar fractions
is approximated by Eq. (6). Equation (7) gives the re- x4/Msi
lationship for the dimensionless equilibrium concentra-
Xsi
= x2/MF* x3/Mc x4/Msi
+ +
. .. . .
.... .
(15)

tion constant kxc'

X6MF.O+X8MF.O )2
kc~
MpeoMc 2) Xs'iq=kxsi
+I ...
(16)
v MslagM**.*~ 100
,. = XFeoAc . ..
(6) x 7M~1'g x7Msio.
The reaction controlled by the difference of
rate is
MslagM,*. the actual sillcon content from the equilibrium, similar
kxc
= XF.o Xc=kc~~
..

.*
(7)
MFeoMc 1002 "' to Eq. (1 1). The rate constant was determined as kdsi
=
144 kg/s. The rate of change in silicon due to reaction
kxc 491 10~6 was obtained. The average molar
with FeOin the slag is given by Eq. (17). Equations
massof steelmaking slag is slag =0.0606 kg/mol (Turk-M
doganl7)). Equation (8) gives the molar fraction of car-
(15) and (16) give Xsi and Xs'iq'
bon in the steel Xc' Equation (9) gives the equilibrium ~4 = rv
kdsi~Asi Xs'iq) """-"(17)
molar fraction of carbon Xceq. Equation (lO) gives the - ~ """-'
molar fraction of FeOin the slag XFeo' 4.5. Rate of Changeof Solid Slag Mass
The rate of change of solid slag (~s) is dependent on
x3/M__~: the addition of slag (d3) and the slag melting rate. The
Xc= (8)
X2/MFC+ x3/M! + x4/Msi melting rate is proportional to the temperature difference
. . . , , . . , , ,

between the liquid group and the solid group. The heat
X6MFeo ' X8MF*O '
transfer coefficient is klh..5 =0.2kW'K'm~2 that
Xceq
= kxc x + x7Msi02 + l (9) can
7M
"- be adjusted depending on the observed slag melting rate.
stag

x7/MF,o
The area of contact is assumedto be proportional to
XFeo
= x6/Mslag + x7/MF,o + x8/Msi02 . . . . . . . .
(IO) the massof solid slag in the EAF, and a specific area of
k....s =0.12m2,kg~1 is used that corresponds to cubic

The reactionrate is controlled by the difference of the blocks (density 2500kg/m3) with 0.02m sides. The slag
actual carbon content from the equilibrium carbon that is being melted must receive latent heat of fusion
content as in Fruehanl2) (Eq, lO, p. 37). The decar-
()~sl'g) and specific heat multiplied by the temperature
burization rate constant was determined as kdc= difference between the liquid group and the solid group.
As].g and Cp(Slag(x))
12kg s1 to give a reasonable decarburization rate. It
'
has units of kJ/g-atom (kJl(g-
has the sameunits as the differential equation for car- atom'K)) instead of kJ/g-mol (kJ/(g-mol'K)). The
bon (kg s~ 1). The rate of decarburization due to reac-
'
equation for the solid slag melt rate in units of (kg s~ 1) '

is given by Eq. (18):

tion with FeOin the slag is given by Eq. (1 l). Xc and
Xecq are given by Eqs. (8) and (9) respectively:
Msl*gkth**5k****5(x5)(xl 3)(xl 3/xl 2)
2 - xl
~3 kdC(Xc Xc'q) (1 l)
x5 = + d3
=- ~ "-"""
()~stag
+ Cp(Slag(s))(xl
2- xl
3))
.(18)
4.4. Rate of Changein Dissolved Silicon
The rate of decrease of
the silicon (~4) in the steel melt 4.6. Rate of Changeof Liquid Slag Mass
is due to reaction with FeOin the slag. The reaction The rate of increase of molten slag (;e6) is equal but
with FeOin the slag is driven by the difference from of opposlte sign to the melting rate of the solid slag.
equilibrium that is approximated as a hyperbolic rela-
tionship between the Si in the steel melt and the (FeO)2 M~1*gkth**5k*..*5(x5)(xl 3)(xl 3/xl 2)
2 - xl
in the slag (mass percentages).
x6~ ().sl'g
+ Cp(Slag(s))(xl
2 - xl
3)) " ' .
(19)
This is similar to the treatment of the carbon in the
steel melt. Equilibrium is taken as: 4.7. Rate of Changeof lron-oxide Massin Slag
(o/oFeO)2 (o/oSi)=0.00057=ksi'/.
.
""""-(12) The FeOin the slag (Jt7) is determined by five rate
mechanisms. Firstly the rate of Fe oxidation by oxygen
The relationship
between ks~'/. and the molar fractions injection and secondly the reduction of FeOby graphite.
Is approximated by Eq. (13). Equation (14) gives the These two mechanismsare the sameas Eqs. (2) and (3)
relationship for the dimensionless equilibrlum concen- respectively, but with opposite sign, and MF.o replaces
tration constant kxsi' The third
Mlr-*. rate is the DRI addition rate with 7010
2 ll
MFeolvlsi 3) FeO. The fourth and fifth rates are the FeOreduction
2 vsi
= XFeoA 100
ksi"/. (13)
Ms2lagM***** ..,,.
by carbon and silicon from the steel melt. The rate of
change of iron-oxide in the slag is then given by Eq.
Ms2lagM*^_.~ (20). Equations (8) and (9) give Xc and Xc*q and Eqs.
kxsi=X2FeoXsi~-k s*~(;~~ (14) (1 5) and (]6) give Xsi and Xs'~q respectively.
M2
F*o M
1003 si
,,,

kxsi = 808 '

• l0~8 wasobtained Xsi '
is the molar fraction

of silicon in the steel and Xs'iq is the equilibrium molar

C 1999 ISIJ 26
 ISIJ International, Vol. 39 (1999), No. 1
ratio of the respective characteristic dimension to the
~
2MFeodl x 7kgr MFeod5 + 0.07d2 of the characteristic dimensions. Expressed as
M02 _ (x6 + x7 + x8)Mc sum
fraction of ul the rate of gas extraction from the
, EAFis:
MFeo
kdc(Xc Xeq)
-( 2MFeo
kdsi(Xsi Xseiq) hd/(kUu2 +hd) (25)
- Mc ~c Msi ~
To improve the approximation by adjustment, a di-
........ .........

.(20)
mensionless constant kU is used. The extraction rate of
4.8. Rate of Changeof Silicon-dioxide Massin Slag
COfrom the EAFas a fraction of the total gas extraction
The rate of change of Si02 in the slag (i8) is equal rate equals the COfraction of the gas composition in
but of opposite sign to Eq. (17). Since the molar masses the EAF. Combiningthe three componentsas discussed,
are different this equation must be multiplied by the ratio the equation for the rate of changeof in the gas-phase CO
of Msio, to Msi' Equations (1 5) and (16) give Xsi and Xs'iq within the EAF
is obtained.
When the relative pressure
respectively. The rate of change of Si02 in the slag is is negative, Eq. (26) is used, however
whenthe relative
given by Eq. (21): pressure is positive, Eq. (27) is used. This accounts for
the difference in composition between EAFgas and air.
Msio
x 8= ' rv
kds~\^si Xs'iq) ......(21) Xc and Xc'q are given by Eqs. (8) and (9).
Msi - ' .
"
Mcod5 hdulx9
4.9. Carbon-monoxidein EAFGas-phase ,e

The CO(i9) in the

Mc ~ (kUu2 +hd)(x9 + xlo + xl 1)
gas phase is determined by four
rate mechanisms. Firstly decarburization of the steel Mco kdC(Xc
bath. Secondly reaction of FeOwith graphite injected + Xeq)+2McokAtRlkpRxl4_"'(26)
~c
into the slag. Thirdly combustion of COin leak-air and
Mc
fourthly extraction of gas through the off-gas duct. Mcod5 hdulx9
~
The rate of COproduced by decarburization is the Mc (kuu2 + hd)(x9 + xlo + xl 1)

sameas Eq. (1 l), but with opposite sign and multiplied Mco kdc(Xc kpR^14"9
by the ratio of Mcoto Mcsince the molar masses are v ~
+ -
Xc*q)
-
.......(27)
different. This is given by Eq. (22): Mc (x9 + xl o + xl 1)

kdc(Mco/Mc)(Xc Xc'q) (22)

4.lO. Carbon-dioxide in EAFGas-phase
~ -"-" '

The C02 (;elo) gas-phase is determined by 2

in the
The rate of COproduced by injection of graphite into mechanisms. The rate of COcombustion to C02 and
the slag is proportional to the rate of graphite injection the extraction of gas through the off-gas duct. The rate
(d5)' Following from assumption (8) all the carbon of COcombustion to C02 is the negative of Eq. (24),
introduced by the graphite injection reacts instantly with where Mc02 replace Mco' The rate of C02 extraction
FeOto form COand iron. This rate of production is from the EAFis the same fraction of the total gas
represented by the following equation: extraction as the C02 fraction of gas composition in
the EAF. The total gas extraction
d5MCO/MC (23) was discussed in the
""""""-""-"""" previous section. Combining the two components as
The rate of COcombustion is proportional to the rate discussed, the equation for the rate of change of C02in
of leak-air entry. This is proportional to the relative EAF the gas-phase is obtained. Equation (28) is used with
pressure (xl4) relative to outside air pressure. Relative negative re]ative pressure and Eq. (29) with positive
pressure will normally be negative, becauseof the draught relative pressure.
force. Leak-air has the normal air compositlon of 79010
nitrogen and 21 o/o oxygen. This amountsto 7.3 mole 02 hdulxl
o
and 27.4 mole N2 per kilogram of air. Based on the
xl o =
~ (kUu2 + hd)(x9 + xlo + xl l) _ 2MC02kAIR
IkpRxl
4
assumption that all the oxygen in the leak-air combusts .(28)
with CO in the EAFfreeboard, the rate of
com- CO hdulxlo kpR"-i4xlo
bustion is: ;tlo
= ~ (kUu2
-2McokA~RlkpRxl4 ..... .........
(24) + hd)(x9 + xl o + xl 1) (x9 + xl o + xl l)
.(29)
The of gas extraction from the EAF is pro-
rate
portional to the draught force of the fan. For this 4.ll. Nitrogen in EAFGas-phase
modelthe draught force is approximated by an equivalent The N2 (ill) in the gas-phase is determined by three
massflow. Referred to as ul, this approximation is one rates. The N2 Ieaking in with the leak air, the N2 that
of the manipulated variables of the off-gas process, and is blown in with the carrier
gas and the extraction of gas
has the units of (kg' s~ 1). At the slip-gap air is entrain- through the off-gas duct.
ed from the atmosphere to combust the COremaining in The rate of N2 Ieaking in is directly proportional to
the extracted EAFgas, and to cool the off-gas. The duct the relative pressure:
area at the slip gap has a characteristic dimension hd'
MN=k*i*2kpR^14
v ..........(30)
The slip
gap width (u2) is the other manip~lated variable .....

and is the characteristic dimension of the slip-gap area. The rate of N2 blown into the EAFby the graphite
The flow rate through each area is proportional to the injection system is a fraction of the graphite injection

27 @1999 ISIJ
 ISIJ International, Vol. 39 (1999). No. 1
rate. For every 150kg of
graphite that is injected kg I 4.14. Relative Pressure
of N2 with
is injected
the carrier gas. The atmospheric pressure remains constant. Then
The N2 extraction rate is proportional to the N2 the rate of change of the relatlve pressure is only de-
fraction of the off-gas composition. The total gas
extraction was discussed in subsection 4.9. The three
pendent on the rate of change of the EAFpressure and
independent of the atmospheric pressure. To obtain an
componentscombined give the rate equation for N2 in equation for the rate of changeof pressure, one approach
the gas-phase. Equation (31) is for negative relative is to
use the ideal gas law, to solve it for pressure, and
pressure, and Eq. (32) for positive relative pressure then take the first derivative with respect to time. In
;tl doing this, for simplicity it Is assumedthat the free-
I = board volume remains constant. Since the freeboard
hdulxl d5
I volume (175m3) is large with respect to the volume of
(kUu2 + hd)(x9 + xl + xl
o
1) -MN=k *i*2 kPRx 14 +
150 the solids and liquids in the EAF, this is acceptable.
.(31)
Nowonly the molar quantity and the temperature
;~l rates have to be considered. Whenthe temperature rate
I = is considered, the heat capacity is approximated
by a
hdulxll ~PR"
1~
* 14x 11 d5 constant equivalent heat capacity for the EAF(CEAF).
+ The molar quantity is changed by three mechanisms.
(kUu2 +hd)(x9 +xlo + xi l) (x9 + xlo +xl 1)
150
.(32) Firstly the oxygen injection rate and the associated CO
release rate. Secondly the rate of gas exchange with the
4.12. Temperature of Liquid Metal atmosphere through leaks due to relative pressure.
The total power generated in the EAFis the sumof Thirdly the gas extraction rate given by Eq. (37):
all
sources and losses. The power loss through the
water-cooled wall proportional to the temperature
is hdul
(37)
difference between the EAFtemperature and the outside MGAs(kUu2
.

+ hd)
temperature. The total power is equal to (pt+d4-
kvTxl2); pt is a function of chemical reactions, heat
The average molar massof the gases that are extracted
capacities, and scrap and slag melting, d4 is the electrlc
is approximately MGAS =0.03 kg/mol. The COrelease
rate caused by the oxygen is 2dl (two moles of COform
arc power. The function p* will be discussed in Subsec. for every mole 02) and the molar mass of 02 must be
1
4. 5. The liquid metal temperature rate is the total
power used to convert the units of dl from (kg/s) to (mol/s).
divided by the heat capacity of all the fluids. It is assumed
that the heat transfer between the liquid metal and the
The gas that exchange because of the pressure dif-
ference would normally be air, since the pressure dif-
slag is sifficiently fast that the slag and the liquid metal
ference should be negative. The molar mass of air is
are at the same temperature. Equation (33) gives the approxirnately that of N2' Whenall the componentsof
temperature rate. the equation are combined, the rate of change for the
![ x2Cp(Fe(L)) x3Cp(C) relative pressure is obtained:
;~l
2 ::= pt( + d4 - kvT(xl
2-
Tair))
MFe
+
Mc
x4Cp(Si) (2x6 +2x7 + 3x8) R x9 + xio +xll (pt+d4-kvTxl2)
+ + Cp(Slag(L)) Mco Mc02 XN2
xl 4 =
Msi Mstag -- Vol CEAF '

,(33)
R~)c~l2 2Mdo ulhd 2)
kpRxl4
Cp(Slag(L)) is obtained from Turkdogan19) for liquid slag: + MGAs(kUu2 hd) - MN ..
(38)
2 +
HT=-356.31 +0.0405T kJ/g-atom .,,.....
(34)
4.15. Thermo-chemical Energy Exchange
4.13, Scrap and Solid Slag Temperature Table I gives the average enthalpy of formation for
The liquid metal heats the scrap and the solid slag. the various species used here. Enthalpies of formation,
The heat transfer is proportional to the temperature molar heats of fusion, and molar heat capacities were
difference between the liquid metal and the scrap. The obtained from Kubaschewski.20) These are temperature
heat transfer coefficients are kth.,1 =0.4kWK~ I m~2 ' ' dependent, and average values for the relevant tem-
for scrap and kth.,s =0,4kW'K-1 'm~2 for solid slag, perature range are used.
The area constants k.,**i =0.005 m2 kg~ I for scrap and . A11 species are at a temperature of 1680K (see as-
k****5 =0.12m2.kg~i for solid slag
are used. The heat sumption (16)). AH** is the enthalpy of formation.
that is transferred divided by the heat capacity gives the The enthalpy of solution must be accounted for with
rate of changefor the scrap and solid slag temperature: species in solution. An underscore (s) is used for
enthalpy of solution. For Oto 4o/o Cin liquid Fe solution,
~l
3=
[(k
th*,i k*,**lxl(xl2-xl3) AHc-s=27kJ/mol (Rist &
Chipman21)). The activity
+ kth.,5k.,,.5xs(xl 2 - xl
3))(
I - x 13/xl 2)]/ coefficient of FeOis a weak function of temperature
[xlCp(Fe(s))/MF. + 2x5Cp(Slag(s))/M~1'g] "'(35) (Turkdoganl7)). Then it follows that the enthalpy of
solution for FeOin slag is small and can be ignored.
Cp(Slag(s)) is obtained from Turkdogan19) for solid slag:
The enthalpy of solution for Si02 in slag is AHsios=
HT=-348.33+0.03T kJ/g-atom ..,,.....(36) -45 kJ/mol. AHsi -s = 132 kJlmol for low Si concentra-
~
@1999 ISIJ 28
 ISIJ International, Vol, 39 (1999), NO. 1

Table l. Enthalpies of formation. 02 Fan Arc Slip-gap Sla DRj: Cinto slag

Smbol Value
AHCO -1 17 kJ/mol
-3 96 kJ/mol
AHc02 o 1 2 3 4 5 6 7 8 9 Io Time(*in)
AHF*a -243 kJ/mol
Fig. 2. Timehneof simulation.
AHsi02 -946 kJ/mol

• 60t liquid metal with 4"/. Cand I "/. Si

tion in steel (Sigworth & Elliott22)). e 20t DRI, with 93
"/o metalization
The following equationsare the auxiliary equations, lOt burnt line and 5t dolomite
e
allwith units of (kW). Equations (8) and (9) give Xc and
e Oxygen, 30Nm3/t
Xc'q respectively and Eqs. (1 5) and (16) give Xsi and Xs'~q
Typical tap-to-tap times are 100 to 120 min. The EAF
res pectively .
diameter is lOmand the freeboard volume 175 m3. The
"hot heel" and foaming slag practices are commonly
pl = reaction([C] + [FeO] ~,Fe + CO)
pl = (A Hc- s+ AHF*o~ AHco)(kdc/Mc)(Xc ~ Xc'q)
employed in EAFoperations. With the foaming slag
practice graphite is blown into the slag, where it reacts

p2 = reaction(Fe + ~02~ FeO) with FeOto form CO that bubbles and foams the slag,
improving the arc efficiency. DRI is used to control the
p2 = - (AHF.o)2dl/Mo, temperature. For detailed discussion of EAFpractices
p3 = reaction(CO + ~02->C02) see e,g. Mclntyre and Landry23) and Jones.1)

p3 = - (A Hco. ~ AHco)(2k*i.1 )(kpRxl 4) 5.2. Initial Conditions

p4 = reaction([Si] + [FeO] H, Fe+ [Si02]) For this simulation the instant after the EAFdoor
was closed is taken as the starting point. At this time it
p (AHs* s AHsio,~AHsio.-s)(kdsi/Ms*)(Xs* Xs'*q) is assumedthat the following
componentshave already
p5 = heating loss to oxygen stream been added:
e 20t hot heel, 0.1
o/o carbon O~/o Si.

p5 = - (dl/Mo,)(xl2 - To,)Cp(02) 70t scrap, I o/* carbon, Oo/o Si.

e
p6 = heating loss to oxygen in leak-air stream e 60t hot metal, 40/0 carbon, I
~/o Si.

e 6t burnt lime and 3t dolomite.

p6 = - k.i,1kpRxl4(xl 2 ~ TAlR)Cp(02) The inltial conditions are chosen to represent typical
p7 = heating loss to oxygen in the leak-air stream conditions in the EAF. Whenthe simulation starts
approximately 40 t scrap is left since the electric arc
p7 = - k.i,2kpRxl4(xl 2 ~ TAIR)Cp(N2)
was used to melt a hole in the scrap for the hot metal.
p8 = heating loss to slag stream The following initial conditions for a simulation are
calculated:
p8 = - 2d3Cp(Slag(s))(xl 3 - T~1'g)/M~tag 40 OOOkg solid and I06 OOO kg liquid metal
e
p9 = heating loss to DRI stream e 4OOOkg Cand 750 kg Si in liquid metal
e 8OOO kg solid and I OOOkg liquid slag
~ (Fe)+ Cp(F~~M(s:i (xl2 TDRl)
. Okg FeOand Okg Si02 in slag
p9= -0.93d
. 11.7kg CO, 9.1 kg C02 and 17.4kgN2
plo = heating and melting loss to solid scrap and slag e Fluids at 1680Kand solids at 800K
e Relative EAFpressure starting at OPa
plo = ~ kth**1k****1xl(xl 2 - xl 3)
kth'*5k****5x5(xl 3) 5.3. Timeline
- 2- x i
pl I =heat of reaction (C + [FeO]~•Fe + CO) from d5
Figure 2
gives a graphical representation of actions.
Arrows indicate when disturbances or control actions
x7kg,d5(AHF*o AHco) occur or commence.The manipulated variables are ad-
pl I = ~ justed as follows:
+ x7 + x8)Mc
(x6
At time t = I min the off-gas fan flow is increased from
e
pt =pl +p2 +p3 +p4 +ps +p6 +p7 +p8 +p9 +plo +pl l O to 16.875 kg/s.
e 3
At time t = min the slip gap size is decreased from
5. Simulation 0.25 to 0.2m.
The discussion begins with typical operating condi- The following disturbance inputs are assumed:
tions, then leads on to initial conditions for a sirnulation, . At t=6s the oxygen fiow is turned to maximum
di
a timeline of events, and the results of the simulation. = 2.5 kgls
. 2
At t = min (120 s) the electric arc is switch on, at full
5.1. Typical EAFOperating Conditions
power d4 = 70 MW
A South African steel producer uses the following e At t=6min fluxes are added continuously at d3 =
typical EAFfeed materials. 20 kgls until the end of the simulation.
e 20 t hot heel
is left in the
EAF e 7
At t= min DRI is added continuously at d2 =40kgls
e 70 of
t scrap high purity type until the end of the simulation.

29 @1999 ISIJ
 ISIJ International, Vol. 39 (1999), No. 1

kg 104 kg
14
x 1200

12 1OOO

10 800

8 600

6 400

4 200

2 O
O I OO 200 300 400 500 600 O i OO 200 300 400 500 600
Time (s) Time (s)
Fig. 3. Liquid metal (dashed) and scrap mass. Fig. 5. SI02 (dashed) and FeOin liquid slag.

kg "/*FeO
4000 4
3500 3.5

3000

2500 2.5

2000 2
1500 1.5
1ooo 1
500 o.5

o o
o I oo 200 300 400 500 600 2.6 2.7 2.8 29 3
Time (s) "I.C
Fig. 4. Carbon (dashed) and sihcon in hquid metal. Fig. 6. o/oFeO in slag against o/oC in liquid metal,

e At t = 8min graphite injection into the slag commences kg

at a rate d5 I kg/s continuously until the end of the 12000

=
simulation. The carrier gas flow rate is approximately
O.007 kg/s. 1oooo

6. Simulation Results 8000

The equations are solved using a 4th order Runge- 6000

Kutta fixed step size numerical algorithm. The fourteen
state variables are then plotted against time, and the
4000
results are given here with discussion:
Figure 3
shows the liquid metal mass and the scrap
2000
mass on the samegraph. The addition of DRI directly
increases the liquid metal massfor the last three minutes.
Before DRI addition it is evident that the rate of in- o
o 1oo 200 300 400 500 600
crease of liquid metal mass is equal to the rate of de- Time (s)
crease of the scrap mass. Fig. 7. Liquid slag (dashed) and solid slag.
Figure 4
shows the mass of carbon and silicon in
solution in the liquid metal. Figure shows the Si02 5 The process is forced further
awayfrom equilibriurn by
and FeO in solution in the liquid slag. The carbon the oxygen injection and the DRI addition, but the
injection into the slag slows down the FeO increase. carbon injection brings it closer to equilibrium again.
Figure 6shows the relationship between the FeOmass Figure 7
shows the massesof the solid and the liquid
percentage in the slag and the carbon mass percentage slag. Take note that in this work the FeO and Si02
in the liquid metal. For this level of carbon content the
components are excluded from the liquid slag mass,
equilibrium FeOcontent is very small (close to zero). although they are included whenthermo-chemical calcu-

C 1999 ISIJ 30
 ISIJ International, Vol. 39 (1999), No. 1

kg
K
18 1800
l/
16 tl
, 1600
t
I\
14 I*
,
~
\ .** 1400
12 l \._
hJ
_
1200
10

8 i ooo

6 800
O i OO 200 300 400 500 600 O I OO 200 300 400 500 600
Time (s) Time (s)
Fig. 8. CO(dashed), N2 (dot-dashed) and C02' Fig, lO. Fluid tenrperature (dashed) and solid temperature.

selected such that the model responses

40 concur with those
kg of the actual EAF. In a preliminary investigation this

30
was done for EAF
steel temperature and carbon
com-
position data.
Simulation of the adjusted model gave
good agreement with the plant data. Such an adjustment
fails outside the
20 scope of this work, and is treated in Ref.
24) .

10
7. Conclusions

The model derivedin this paper is intended for the

P~l
o testing of control strategies using off-gas variables in
the control of an EAF. Derivation of
a multivariable
o non-linear state-space model for the process
*1
O was treated.
I OO 200 300 400 500 600
The assumptions that were madeand the state spacc
Time (s)
representation provided a comprehensive framework for
Fig. 9. Suln of the gas-masses and relative pressure.
the development of an EAFprocess model. First principles
of thermodynamicswere then used to derive the model,
lations are done. The meansthat the liquid slag mass ensuring behavior consistent with physical observations.
denotes CaOand MgOalone. Typical operating conditions for an EAFwere discussed,
Figure 8shows the three gas componentstogether, and
and a simulation was conducted with these operating
it
makesgood sense that the amountof CO
will increase
conditions. In order to apply the model to represent
dramatically with the introduction of the oxygen stream. a
While the fan is not yet on, the increase of COwill lead
specific industrial EAFthe adjustable parameters in the
model need to be selected such that the model response
to an increase in EAFpressure, which
in turn will force

a lot of EAFgas out, Ieading to the drop in N2 and concur with those of the actual EAF.
C02' Figure 9showsthe sumof the gas-masses, together
with the relative pressure of the EAF. Although the ideal REFERENCES
gas law determines that the total molar quantity remains l) J. A. T. Jones: Iron Stee! E,7g.. 72 (1995), 37.
constant in the EAF, the masswill vary because the gas 2) O. A. Zhulkovsky, A. G. Chernyatevich and A. V. Gress: Stee!
in Translation, 23 (1993), 73.
composition changes. C02 is a heavy gas and the total
3) The making, shaping and treating of steel. IOth ed., ed. by W.
gas massincrease with higher C02 content. The molar T. Lankford et a!., Association of lron and Steel Engineers,
quantity also varies with temperature. Pittsburgh, (1 985).
Figure 10 shows the variation of the fluid group 4) M. A.
Reuter, M. Oosthuizen, I. J. Barker, M. S. Rennie and A.
temperature and solid group temperature. Note that the DeWaal: 8th IFAC
97.
MMM
Conf., SunCity, South Africa, (1995),
solid group temperature approaches the fiuid group
5) P. E. King and M. D, Nyman:J. App!. Phys., 80 (1996), 1872.
temperature slowly as the solids are melted away. 6) P. R. Austin. H, Nogamiand J. Yagi: ISIJ In!., 37 (1997), 458.
The purpose of the derivation was to obtain a mod- 7) S. K6hle: Stah! Eisen, I12 (1992), 59,
el that is appropriate for closed-loop control system 8) R. D. Morales, H. Rodriguez-Hernandez, P. Garnica-Conzalez
simulation and initial controller verification. The model and J. A. Romero-Serrano: ISIJ Int., 37 (1997), 1072.
derived here approximates the dynamic response of the 9) C. Petersohn, E. Burstrom and P. Hahlin: 4th European Electric
Steel Congr., Madrid, Spain, (1992).
physical process sufficiently accurate for its purpose. In
order to apply the model to represent a specific industrial
l O) E. Sridhar and A. K. Lahiri: Stee/ Res., 65 (1994), 433.
l l) P. Gardin. C. Soide, A. Dez and I. Guillaume: Ironn7aking
EAFthe adjustable parameters in the model need to be S!ee!nlaking, 19 (1992), 306.

31
C 1999 ISIJ
 ISIJ International, Vol. 39 (1 999), No. 1
l2) R. J.Fruehan: ISS Trans., 12 (1991), 33. xl2 = T(fluid group), temperature (K)
l 3) B. Sarma, A. W. Cramband R. J. Fruedan: Metall. Trans. B,
xl3 = T(solid group), temperature (K)
27B (1996), 71 7.
14) E. Plbckinger and O. Etterich: Elektrostahlerzeugung, 3rd ed., xl4 = P(furnace) relative pressure (Pa)
Verlag Stahleisen m, b. H., Dusseldorf, (1979). ui =th(turb), off-gas turbine fiow (kg/s)
l 5) W. Masclanka: Int. Symp. on Modern Developments in u2 = w(slipgap), slip gap width (m)
Steelmaking, Jamshedpur. India, (1981) dl =th(Oxy), oxygen fuel flow (kg/s)
l 6) J. M. A. Geldenhuis:
SAIMMColloquium on Pyrometallurgy, cl2 =th(DRI),
DRI addition flow (kg/s)
Randburg, South Africa, (1997). d3 =th(slag), slag addition (kg/s)
l 7) E. T. Turkdogan: Fundamentals of Steelmaking, Institute of
d4 =p(Arc), electric arc power (kW)
Materials. London, (1996).
d5 =th(graphite), graphite injected (kg/s)
18) J G. Bekker, K. Craig and P. C. Pistorius: 9th IFAC
I.

Conf., K6ln, Germany, (1998).

MMM
l 9) E. T. Turkdogan: Physicochemical properties of molten slags and Table A-1 gives the values of dimensionless constants.
glasses, The Metals London. (1983).
Society, Table A-2 gives the air and area constants. Table A-3 gives
20) O. Kubaschewski, C. B. Alcock and P. J. Spencer: Materials the rate and thermal constants. Four constant tempera-
Thermochemistry, 6th ed., PergamonPress, Oxford, (1993). tures were assumedto be at 300 K: TDRl' T~~*g' To., TAlR'
21) A. Rist and J. Chipman: Rev. M~ta!!., 53 (1958), 796.
22) G. K. Sigworth and J. F. E1liott: Me!. Sci., 8 (1974), 298.
Thermodynamicdata such as latent heat, specific hcat or
23) E. H. Mclntyre and E. R. Landry: I,'on Stee!make,', 20 (1993), enthalpy of formation were obtained from Kubaschewski
61. et a/.20) except were a reference would indicate a dif-
24) J. G. Bekker: Master's Dissertation. University of Pretoria, South ferent source.
Africa, (1998). (in preparation)

Table A-1. Dimensionless constants.

Appendix
K kU kxc kxsi
xl =m(Fe(s))' mass of solid iron (kg) 15 O 844 491*10~ 8.08*10~
x2 =m(Fe(L)), massof liquid steel (kg)
x3 =m(C), carbon in solution in steel (kg) Table A-2. Alr and area constants.
x4=m(Si), silicon in solution in steel (kg) Air [mol,k ~
Area [m ,k ~
J
x5 =m(Slag(s))' solid slag mass(kg) kALRl kAJR2i kareal
k area2

x6 = m(Slag(L)), Iiquid slag mass(kg) 7.3 27.4 O005 O. 12

x7 =m(FeO), FeOin slag (kg)

x8 =m(Si02)' Si02 in slag (kg) Table A-3. Rate and thermal constants.
x9 =m(CO),carbon-monoxide (kg)
xlo =m(C02)' carbon-dioxide (kg)
Rate [k
kdc
s~

kdSi
J Thermal [kW.K~ ,m~ J
ktherl kther5

xli =m(N2)' nitrogen in gas-phase (kg) 12 144 0.4 02

@1999 ISIJ 32

View publication stats