You are on page 1of 36

REVIEW

Electric Arc Furnace www.steel-research.de

Review on Modeling and Simulation of the Electric Arc


Furnace (EAF)
Hans-Jürgen Odenthal,* Andreas Kemminger, Fabian Krause, Lukas Sankowski,
Norbert Uebber, and Norbert Vogl

availability of the charge material is variable


The exponential growth of computing power in the last two decades opens regionally and in terms of time. Conse-
up entirely new options for numerical simulations of the Electric Arc Furnace quently, the charge mix is determined by
(EAF). Simulations can be used to analyze physical phenomena resisting economic considerations.
In the EAF, the energy required to
direct observation or operational measurement even to this very day. This
meltdown the scrap is predominantly
paper gives an overview of the state-of-the-art of the Computational Fluid introduced by electricity and converted
Dynamics (CFD) simulation on the EAF as well as an outlook on future fields into heat by the electric arc (simply arc in
of application, while being well aware that by far not all phenomena and the following). A distinction is made
literature can be used. The paper makes no claim to exhaustiveness, between alternating-current (AC) and di-
especially since three subjects of simulation technology have to be excluded: rect-current (DC) furnaces.
With the advent of powerful rectifiers
Process models for furnace controlling, stress calculations, electromagnetic
from about the end of the 1980s, DC
simulations. Thus, the focus is on the fluid- and thermo-dynamic furnace furnaces have been increasingly used. The
processes and on fundamental methods that can be applied to examine these DC furnace has fewer limitation as to the
processes. The basic EAF functionalities and selected fluid-dynamic simu- increase in the electrical power input, for
lations are presented, for example, on multiphase flow, thermal loading of example, because there are less flickers in
refractory lining and wall panels, chemical reactions and post-combustion, the power mains. Other advantages are the
low specific electrode consumption, homo-
oxygen injection technology, and bottom tapping.
geneous meltdown behavior due to the
centrally-arranged arc, and quieter operat-
ing practice. Disadvantages are the high
investment costs, extensive changing of the
1. Introduction bottom electrode, and the larger hot heel volume in the furnace
In 2016, the crude steel production worldwide amounted to shell. With the continuous optimization of the AC EAF, for
about 1.629 billion t; the figure corresponds to a hardly example, with regard to electrode cooling, life of the refractory
imaginable steel cube with a side length of approx. 600 m. lining, increase in electrical power input, etc., the AC EAF has
The European Union (EU) is the second- largest steel producer established itself all over the globe. Currently, there are about 980
in the world behind China; Germany takes seventh place with AC EAFs and 100 DC EAFs in operation worldwide, the latter
approx. 42.1 million t.[1] predominantly in North America.[1] Therefore, this paper mainly
Approximately two-thirds of the crude steel worldwide is melted deals with the AC EAF.
in integrated steel plants via the blast furnace  converter route as The EAF process is energy-intensive and accounts for about
oxygen steel. The remaining one-third is produced via the electric 3% of the electrical energy consumption in industry world-
steelmaking route in the EAF. Last year, 39% of the crude steel in wide.[2] Behind the costs for the charge materials (scrap, alloys,
the EU was produced by the EAFroute. While steel scrap is used in etc.), electrical energy has the second largest share in the total
the converter only up to approx. 30%, up to 100% can be melted in costs of EAF steel production. However, with regard to energy
the EAF. The EAF route is the main recycling route, it can be aspects, the EAF route is more favorable than the blast furnace
adjusted well to local operating conditions and required product route because the energy density of the arc is extremely high
ranges. Until around 1970, the EAF was merely used as meltdown converting electrical energy efficiently into melting heat; a
unit for the production of structural steel grades. Nowadays, it is temperature of 10 000–15 000  C[3] is generated in the arc,
possible to charge Direct Reduced Iron (DRI), Hot Briquetted Iron thereby heating the melt to approx. 1650  C. Melting is
(HBI), hot metal, and pig iron. It has to be observed that the accelerated by additional injection of oxygen and/or fuel/oxygen
mixtures. As a result of the high energy density, the EAF route
H.-J. Odenthal, A. Kemminger, F. Krause, L. Sankowski, N. Uebber, N. Vogl requires approx. 30% less energy and saves about 1.5 t of iron ore
SMS group GmbH, Eduard-Schloemann-Straße 4, 40237 Düsseldorf, and 0.65 t of carbon versus the blast furnace route.[4] The EAF’s
Germany CO2 emission of 395 kg tCS1 is lower by 77% compared with
E-mail: hans-juergen.odenthal@sms-group.com
1744 kg tCS1 for the blast furnace route (incl. sinter and coke
DOI: 10.1002/srin.201700098 oven plant).[5]

steel research int. 2018, 89, 1700098 1700098 (1 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

The steel industry is currently going through a difficult phase


with low demand and low steel prices compared with the production Prof. Hans-Jürgen Odenthal received his diploma and
costs. Energy and resource efficiency are crucial factors for the doctoral degree in mechanical engineering from University
economic production of electric steel. Against this background, of Siegen. After an industrial activity at BMW Rolls-Royce
technical developments are required which improve the sustain- Aero Engines he went to RWTH Aachen University,
ability and productivity of the EAF process in the longer term. Institute for Industrial Furnaces and Heat Engineering,
Plant trials on the EAF are costly and can often not be where he qualified as associate professor in High
implemented for safety-related reasons. Trials in general may not Temperature Fluid Dynamics and as adjunct professor. In
interrupt ongoing operation, thus offering only short time 2005, he joined SMS group GmbH and is now head of
windows with few options for variation. An alternative is the Steelmaking Technology department. He is chairman of
numerical simulation. Due to the enormous progress of High the technical committee Fluid Mechanics and Fluid Flow
Performance Computing (HPC), complex physical phenomena Simulation of the German Steel Institute (VDEh) and
can now be examined. Where reference is made to such member of the sri advisory board. Since 2004 he gives
phenomena in this paper, this means fluid- and thermodynamic lectures at the Faculty of Georesources and Materials
processes based on the convective and diffusive transport of mass, Engineering of RWTH Aachen University.
momentum, energy, and species; their processing is referred to as
Computational Fluid Dynamics (CFD). CFD is complex and Dr. Andreas Kemminger received his masters and doctoral
should be performed on clusters with adapted architecture. CFD degree in ferrous and non-ferrous metallurgy from
helps to make the EAF process transparent, comprehensible and University of Leoben, Austria. He held a position at the
secure, and to identify improvement potential. department for Simulation and Modelling of Metallurgical
In contrast to applications in the aircraft or motor-vehicle Processes at University of Leoben before joining SMS
industries, metallurgical units are often not flow-optimized because group GmbH in 2015, where he is now working in the
the focus is on robust and cost-efficient design and not on minimal Steelmaking Technology department. He is member of
pressure loss or non-separated flow. Hence, flow effects are induced the technical committee Fluid Mechanics and Fluid Flow
which are a big numerical challenge and require long computing Simulation of the German Steel Institute (VDEh).
times. The parallelization by breaking down the computational
domain does not always correlate with shorter computing times in Dr. Fabian Krause received his diploma and doctoral
case of metallurgical flows. The reason is that parallelization of the degree in mechanical engineering from University of
submodels (e.g., multi-phase model þ turbulence model þ particle Bremen. From 2006 to 2010 he was research assistant at
model þ combustion model þ . . .) requires high computing times the Bremen IWT Foundation Institute of Material Science
even at small grids (e.g., 1 million cells) as compared with a single in the department of “Multiphase Flow, Heat and Mass
submodel at a large grid (e.g., 50 million cells). Transfer” and member of the Research Training Group
A CFD model for the EAF should be capable of describing the “Graduiertenkolleg PORENET” funded by Deutsche
following phenomena using relevant submodels, cf.[6]: Forschungsgemeinschaft (DFG). In 2010 he joined SMS
group GmbH and is now working in the Steelmaking
1. Consideration of the three-dimensional, time-dependent, Technology department. He is member of the technical
non-isothermal, multi-phase, turbulent flow of melt, slag, committee Fluid Mechanics and Fluid Flow Simulation of
and off-gas based on the convective-diffusive transport of the German Steel Institute (VDEh).
mass, momentum, energy, and species.
2. Consideration of heat transfer, for example, during scrap Norbert Uebber received his diploma degree in physics
preheating, including the heat transfer mechanisms of from University of Cologne. He has been working for SMS
conduction, convection and radiation. group GmbH for more than 27 years with different tasks
3. Consideration of chemical reactions (C, Si, P, S, combustion) in the field of R&D. He is head of the working group
and phase transformations (melting) of the scrap/DRI/HBI Models & Process Control in the Steelmaking Technology
charge. department.
4. Formation and movement of gas bubbles (O2, CO, Ar, N2,
etc.) and/or particles in all phases; formation of foamy slag.
5. Consideration of electromagnetic forces and fields and the
interaction between fluid flow field and magnetic field.
6. Consideration of varying furnace domains, for example, by subdomains and submodels, however, requires knowledge of
tilting the furnace shell (short-term) or refractory wear (long- the initial boundary conditions which are either derived from
term). experiments or are ad hoc assumptions. In any case, it is
necessary to identify the relevant aspects before the CFD
Even in the foreseeable future, it will not be possible to simulation is carried out.
consider all phenomena in a single CFD model. The submodels The list of CFD software is long, ranging from commercial
for individual processes are not available or are so complex that codes (e.g., ANSYS FLUENT/CFX, AVL Fire, COMSOL,
they would far exceed the computing capacity. Therefore, a priori PHOENICS, Star CCM þ) via open source codes (e.g., Open-
calculations, such as the drag coefficient of a motor-vehicle, are FOAM) to individual codes used in the university field. It is
not possible for the EAF. Breaking down the EAF into especially here that open source solutions have attracted large

steel research int. 2018, 89, 1700098 1700098 (2 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 1. Mechanical design of an AC EAF. Reproduced with permission.[15] 2017, SMS group GmbH.

numbers of users, because they are free and do not entail follow- The furnace shell consists of a steel shell and is located in a
up costs. Despite high acquisition and license costs, commercial tilt-able structure. The lower shell (furnace hearth) is located
codes are frequently used in the professional business sector below the tilting mechanism, is equipped with a projecting
because questions have to be answered quickly. Many submodels structure (EBT) and contains the refractory bricks. The furnace
that often can be linked to one another are integrated into hearth is made of a permanent lining of baked MgO–C bricks
commercial codes. However, the use of open source codes is and a multi-layer ramming mix. The tap hole for tapping crude
expedient, if several submodels have already been combined to steel into the ladle is located in the EBT. The upper shell rests on
form an overall model used for parameter variation studies. the bottom shell, it is made of water-cooled copper or steel tubes.
Right into the 1980s, the EAFs were completely refractory lined.
With the growing melting capacity, the lining was exposed to
2. Design of common Electric Arc Furnaces extensive wear and became an additional cost factor. Nowadays,
the water-cooled wall and roof panels have replaced the uncooled
(AC EAF)
refractory lining. The liquid slag solidifies at the wall panels and
From the authors’ point of view, it is necessary to briefly explain forms a heat-insulating protective layer.
the mechanical and electrical design, possible interactions Water-cooled panels are sensitive components. Heavy scrap
between both systems, and the operating practice of an AC EAF. pieces may damage the panels during charging. If the wall
thickness of the tubes is designed insufficiently or the material is
too soft, water leaks may occur. However, at specific apparent
2.1. Mechanical Design powers of 1000 MVA t1 the panels are predominantly subjected
to thermal loading. The highest heat radiation caused by the
Figure 1 shows the mechanical design of an AC EAF: arcs  and consequently the thermal loading in the pipework 
occurs in the hot spots. The prediction of this loading is a typical
1) Hydraulic tilting platform. question related to the EAF. Mechanical and thermal loading
2) Lower shell: Eccentric Bottom Tapping (EBT), tap-hole, result in contradictory demands on the panels, because usually
refractory lining. copper tubes with smaller wall thicknesses would have to be
3) Upper shell: Water-cooled panels, furnace roof, off-gas elbow, used at high thermal loading, but these thinner tubes would not
post-combustion chamber. withstand the mechanical loading. This is why compromises
4) Power supply, gantry system with roof lifting and swing have to be made during the design, depending on the position
mechanism, electrode supporting arms, water-cooled and loading exposure of the panels. The same applies to the
electrodes. orientation of the pipework. Vertically arranged pipework are
5) Burner and injector systems, additional equipment, for less susceptible to damage during charging, while horizontal
example, lance manipulator for oxygen and carbon arrangements are advantageous regarding to the adhesion of
injection. slag. A number of design variants have resulted with regard to

steel research int. 2018, 89, 1700098 1700098 (3 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 2. a) Horizontal orientated pipes of water-cooled panels inside upper EAF vessel. b, c) Panel box with equipped burner/injector. Reproduced with
permission.[15,78] 2015, Steel Institute VDEh.

the panel resistance. Widespread are tubular structures with a The installed electrical power of the transformers is typically
gap between the pipes in which slag is selectively solidified, thus 100–150 MVA. The current flows through flexible cables to the
offering thermal protection, Figure 2a. bus tubes and/or the electrode supporting arms. The latter can
Since the 1990s, gas burners (CxHy, LPG, oil, etc.) have been be moved hydraulically and the electrodes are clamped at their
used for scrap melting to reduce the electrical energy ends. The electrodes are arranged in a pitch circle and controlled,
consumption. While the melting power is highest in the furnace so that the arc burns freely between electrode and melt. The
center, cold spots may occur near the furnace walls, in which alternating current flows from one electrode via the charge
melt and slag can build up in layers up to meters thick. Here, gas material and/or the melt to the next electrode. The arc supplies
burners selectively introduce chemical energy. Modern burners the process heat with the energy being transferred to the charge
are combined burner/injector systems, Figure 2b and c. Usually, material by conduction, convection, radiation, and Joule’s
several burners are integrated into the furnace panels for resistance heating. Both the arc acting on the melt and the
melting the scrap, and injectors are used to add oxygen and coke. current flowing through the melt induce moderate agitation and
Some furnaces contain additional oxygen injectors that post- mixing; optionally inert gas or electromagnetic stirring
combust CO inside the furnace. Lance manipulators injecting equipment can be installed.[12]
oxygen and coke via, for example, self-consuming lances are The smallest EAFs used in the foundry industry have tapping
used at the slag door. weights of 5 t of crude steel, the largest ones worldwide about
The water-cooled roof has two further openings in addition to 320 t (300 MVA) with an annual crude steel production of approx.
the electrode ports: one opening for material charging (DRI, 3.0 million tpy. The electrical energy required for melting and
slag formers, alloys), and one for off-gas extraction. The off-gas superheating is approx. 320 kWh t1, tap-to-tap times are about
(time-averaged analysis approx. 30% CO, 10% CO2, 55% N2, 5% 40 min.
H2,[7] T ¼ 1200–1750  C[8]) is post-combusted at the elbow gap
(4th roof hole) with secondary air, partly in the off-gas duct,
predominantly in the post-combustion chamber. The dust- 2.2. Electrical Design
laden primary gas is quenched, mixed with ambient air from
the roof area of the steel plant, and the resulting gas flow The electrical design of the AC EAF is made up of transformer,
cleaned in the filter system.[9–11] electrode supporting arms, electricity supply lines and

steel research int. 2018, 89, 1700098 1700098 (4 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 3. Electrical design of an AC EAF. Reproduced with permission.[13] 2009, Universitätsbibliothek der Helmut-Schmidt-Universität.

electrodes; and for the DC EAF single graphite and bottom introduced into the charged material results from the arc voltage
electrodes, Figure 3. The electric circuit is closed by the arc and Uarc and the current rating Iarc (Parc ¼ Uarc  Iarc). Uarc is
the power conduction via the melt. The transformer reduce the controlled by the arc length Larc, that is, the electrode position.
high-voltage mains current (e.g., 110 kV, 500 A) into a low- Foamy slag protects the upper shell and the roof from arc
voltage (e.g., 500 V, 110 kA) mains current. Control of the AC radiation.[13]
EAF is via different transformer steps, however, it is not possible
to control current and voltage independently. The arc current for
the DC EAF is controlled via the thyristors of the rectifier unit.[13]
The electrode supporting arm holds the electrode, conducts
the current and controls the electrode position. If high electric
powers are conducted via the supporting arm, it will have to be
water-cooled to reduce the heat induced in the material. There
are many historical types of construction, for example, steel
arms with water-conducting copper tubes as current conduc-
tors. Today, the supporting arm is made of either aluminum or
copper-plated steel. With the latter design, the steel arm bears
the mechanical loading while the current flows via the copper
plates with their thickness of several centimeters. The current
for the AC EAF flows via the copper outer surfaces as a result of
the skin effect.[3] Discontinuities in the copper lining, for
example, at edges or attachment points, cause the current to
flow via the steel and thus cause local heating on it. Inadequate
connections between steel and copper also cause problems. The
aluminum supporting arm combines mechanical strength and
electric conductivity, with the strength being lower than with
steel and the conductivity not as good as with copper
(σAl ¼ 10  106 S m1, σCu ¼ 58  106 S m1); advantages are the
low weight and the homogeneous material.
The flexible cables connect the transformer to the supporting
arm, so that the electrode can be moved up and the roof to the
side. The cables consists of a bundle of twisted copper cables
routed around a water-filled core hose; up to four cables are used
for each phase. The electrical energy is transformed into thermal
energy by the ohmic resistance of the electrodes, the scrap and
the melt. The thermal energy is emitted to the environment at Figure 4. Twisted connecting cable (phase 2) during boring phase on a
arc temperatures of 10 000–15 000  C by convection and 250 t AC EAF, Imax ¼ 90 kA. Reproduced with permission.[15] 2017, SMS
radiation of the ionized gases.[3,8,11] The power Parc of the arc group GmbH.

steel research int. 2018, 89, 1700098 1700098 (5 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

2.3. Electro-Mechanical Interaction The same electromagnetic effects cause the arcs to be forced
toward the outside in the flat-bath phase. The arc angle results
On the EAFs secondary side there are currents up to 100 kA from the equilibrium of the deflecting forces and the attempt to
and more, thus causing electromagnetic (Lorentz) forces in follow the path of the least electrical resistance. This results in
the cables, supporting arms, electrodes, and arcs. These arcs being deflected horizontally by 120 relative to one another,
interactions result in mechanical and thermal loading, and vertically by 30 –35 .
movements, and a deflection of the arcs.[3] The Lorentz As the alternating current flows through the conductors, it
forces become visible on powerful furnaces and at high generates self-induction, causing the current to flow predomi-
current ratings in form of movement and twisting of the nantly at the outer material layer. The skin effect improves the
cables. Figure 4 shows the twisted connecting cable at the conductivity of copper-plated supporting arms and reduces the
center supporting arm during the boring phase on a 250 t AC radial temperature gradient in the electrodes of AC EAFs to
EAF. The cables are usually arranged in an equilateral 200  C compared to 800  C for DC EAFs.
triangle, the forces acting on the cables are compensated. At a larger distance from the AC EAF conductors, hardly any
However, especially in the boring and melting phase the magnetic field is induced as the fields of the three phases
current flow is asymmetric because the arcs jump between compensate each other. However, in the vicinity of the supply
scrap pieces, and the scrap itself shields the magnetic fields of lines, there is induction in all conductive materials. If these
the adjacent electrodes. Consequently, the center cable tends materials are thermally and electrically insulated, local over-
to twist, while the external cables swing toward the outside. heating may occur.
The forces F acting between the cables are proportional to the The DC EAF forms an electric circuit from the transformer via
cable length L, the electrode pitch circle diameter D, and the the electrode supporting arm and electrode to the bottom
square of the current rating Iarc, that is, F ¼ f ðL; D; I2arc Þ. electrode and back to the transformer. This circuit (D  10 m)
Analog periodic forces also act on the supporting arms and the together with the high current rating induce a magnetic field of
electrodes. approx. 10 000 A m1. The field strength magnetizes ferritic steel

Figure 5. Typical process steps of an AC EAF. Reproduced with permission.[15] 2015, SMS group GmbH.

steel research int. 2018, 89, 1700098 1700098 (6 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

grades. However, as the electric and/or magnetic field is constant


most of the time, eddy currents are not induced in the DC EAF. ðFeOÞ þ C ! ½Fe þ fCOg ð1Þ
Melt stirring by the arcs is a desired interaction. The
alternating electric fields of the AC EAF generate a rotating forming metallic iron and CO. The latter produces gas bubbles,
magnetic field thus causing a slight stirring action in the melt. thus foaming up and increasing the slag volume. As a result, the
The current on the DC EAF induces a magnetic field agitating arcs are covered and the thermal energy is better transferred to the
and stirring the melt more intensely.[14] melt. In this phase, the slag flows out of the slag door. Typical slags
contain high amounts of FeO (e.g., 36% FeO, 27% CaO, 10% SiO2,
5% MnO, 4% Al2O3, all data in percent by weight wt%), reducing
2.4. EAF Operating Practice the slag viscosity and increasing the solubility of MgO.
Tapping: After the melt is superheated and the target temperature
A sequence of typical process steps is shown in Figure 5[15]: reached, the furnace tilts into tapping position. At the end of tapping,
1) Charging. 2) Melting: Ignition, boring, melting. 3) Super- a vortex is produced above the tap-hole (vortexing) causing the
heating. 4) Tapping. undesirable effect that floating slag is entrained and tapped. To
Usually the furnace is charged with the two scrap buckets; detect the slag carryover in time and tilt the furnace back into
further buckets may be necessary depending on the scrap density. position, slag detection systems are used at the EBT.
Single-bucket charging is ideal because energy is saved by the short
loading time. A scrap menu is defined for each steel grade, 3. CFD Simulation
specifying scrap quantity, scrap type, and charging sequence. The
scrap mix is selected according to the target analysis, and the The challenges faced by CFD simulations of the EAF are huge,
sequence results from the predicted meltdown behavior. Lime and because the phenomena described are three-dimensional,
dolomite are added during charging to produce a reactive slag. transient, friction-affected, turbulent, non-isothermal, and
Material charging and process sequences are tuned toward multi-phased. Despite that, it is already possible today to
achieving the steel grade. Approx. 5 kg tsc1 of charged carbon is simulate some phenomena by using specific models. Fluid- and
loaded in the bucket to minimize the loss of iron during melting. thermodynamic processes are described by the Navier-Stokes
Charging: Before charging some residual melt (hot heel, (NS) equations balancing the fluxes of mass, momentum,
10–15% of the starting weight) is in the furnace (mandatory with energy, and species; the Lattice-Boltzmann method is still
the DC EAF) serving as a heat reservoir and accelerating the insignificant in metallurgical fluid mechanics. Based on the NS
meltdown process. At the beginning of the sequence, the furnace equations, four approaches are available:
roof is lifted and swung out. The scrap bucket is positioned by
the crane and unloaded. The bulk density ρsc of the scrap is 1) Reynolds-Averaged Navier-Stokes (RANS): RANS describes
0.7–1.5 t m3. The furnace shell is almost completely filled with statistical turbulence modeling, where the scales are not
scrap after charging. Only after the scrap has been melted there directly calculated, but modeled. Due to the Reynolds
is again a larger free domain above the melt. averaging approach, the coupled, non-linear system of
Melting: After the roof has been closed and the electrode lowered, partial differential equations has more variables than
the arc is ignited. The boring phase starts with a short arc to protect equations; the Reynolds’ stresses, heat, and species fluxes
the roof. With the electrode entering into the scrap pile, the crater must be modeled by turbulence models (e.g., Standard k-e,
and the arc length grow, the main melting phase starts. The burners RNG k-e, Realizable k-e, Standard k-ω-, SST k-ω, Transition
are used simultaneously with the ignition, they have their highest SST-SAS, Reynolds stress, SAS model, etc.).[16–18] Unsteady
efficiency as long as the flames are burning inside the scrap pile. (U) RANS refers to transient RANS simulations; RANS/
If required, the oxygen lance is used at the slag door to URANS methods are most frequently used for metallurgical
accelerate melting. The CO burns with the oxygen to form CO2 flows.
supplying additional energy. The scrap melts down early in the 2) Large Eddy Simulation (LES): Direct simulation of the large
burner area and accumulates in the hot heel. The slag formers and modeling of the small turbulence scales (e.g., Smagor-
dissolve and produce the initial slag. Next to the walls, the scrap insky-Lilly model). The NS equations are low-pass filtered
melts down with a delay. according to location and time, that is, the large vortex
Superheating: After the scrap has been melted down the flat- structures are resolved directly, while the small structures are
bath phase starts. The scrap resting in front of the wall does not mapped by means of a subgrid-scale turbulence model.
longer protect the wall from radiation. Consequently, hot spots 3) Detached Eddy Simulation (DES): RANS method with spatial
are generated, thus causing heat loading and wear and, in the and temporal highly resolved flow regions in which the same
worst case, destruction of lining or panels. To counteract this calculations are applied as with LES.
effect, these parts are protected by decreasing the arc length 4) Direct Numerical Simulation (DNS): Solution of the NS
and foaming up the slag (foamy slag practice). Carbon equation without any parameterization; DNS is, however,
(10–15 kg tcharge1) is added by injectors and door lance into currently limited to geometrically small flow domains, but
the slag. If too much injector coke is used, carbon escapes first DNS analyses have already been carried out for an EAF
unused as CO and/or CO2 in the off-gas, if the amount of injector injector.[19]
coke is too small, not enough foamy slag is produced. Carbon
reacts with the high iron oxide contents of the slag according to In comparing expenditure with benefit regarding the EAF
the endothermic reaction CFD simulation, the RANS and/or URANS method is without

steel research int. 2018, 89, 1700098 1700098 (7 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

doubt the fastest variant to get reliable results. It is an advantage considered by the Discrete Ordinates (DO) radiation model. The
that most RANS solvers can be linked very well to the following DO model solves the Radiative Transfer Equation (RTE) for a
submodels: finite number of discrete spatial angles. It can not only be used to
Multiphase models: The Euler-Euler method (space-fixed) calculate the radiation exchange between any surfaces by
assumes that the phases (e.g., melt, slag, gas) are interpenetrating considering the lines of sight, but also the inner gas radiation
continuous fluids interacting with each other. The Volume of Fluid with absorption, diffusion, and refraction phenomena.
(VoF) model is a special Euler-Euler method for non-mixing fluids
with free phase boundaries and belongs to the interface capturing
methods.[20] The interface is characterized by control volumes 4. CFD Simulation of Individual EAF
containing mixed phase fractions with averaged fluid properties. Components
In addition to the equation of mass, momentum, and energy, a
scalar transport equation for the void fraction αs is solved for the The previous chapters illustrate that many coupled fluid- and
secondary (and each further) phase in the control volume: thermodynamic processes occur in the furnace, which cannot be
calculated to the very last detail and not simultaneously. In the
@αs @αs following, excerpts from literature are presented simulating
þ ui ¼0 ð2Þ selected topics and furnace components and providing an
@t @xi
overview of the state of the art.
The volume fraction of the primary phase is αp ¼ 1 –αs.
Depending on the quantity of the secondary phase in the
control volume, αs takes any state between 0 < αs < 1. The phase 4.1. Furnace Shell, Furnace Roof
boundary between the primary and secondary phase is
interpreted as discontinuity of the fluid properties. With increasing electric power, the spatial dimension of the arc
On the other hand, the Euler-Lagrange method (mass-fixed) is has become considerably larger. If the arc was freely burning on
used to describe the movement of bubbles and particles. top of the melt bath, this would lead to high energy losses and
Looking, for example, at the argon injection via bottom stirrers, damage to the furnace wall.[21] Premature wear of the bottom
argon is considered as a set of discrete bubbles, the movements shell lining and failures of the slag-covered panels in the upper
of which  under the influence of the ambient melt forces  are shell cause production interruption. It was therefore consequent
superimposed on the main melt flow. All forces (buoyancy, drag, to define a parameter for designing and controlling the furnace
inertia, masses, pressure forces, etc.) acting on the bubble are process characterizing the radiation emission and wear of the
considered. Integrating the force balance equation twice leads to refractory lining, respectively.
the trajectory of the bubble. An example of an Euler-Lagrange In 1962, Schwabe[22] defined the Refractory Erosion Index RE
model is the Discrete Particle Model (DPM).[16] as a measure for thermal wall loading and an indicator for the
Reaction kinetics: Combustion models combine the changes of wear rate. The central statement is that the arc radiation is
the species concentrations of the reactants with their volumetric proportional to the active power Parc and the arc length Larc. Since
generation/destruction rate and the changes of enthalpy, the arc voltage Uarc is a measure for the arc length Larc, the
temperature, and mixing density. A transport equation of following applies:
convection-diffusion type for the mass fraction of the respective
gas component is set up, in which, amongst others, the reaction RE ¼ Parc Uarc : ð3Þ
rate R is unknown. At turbulent combustion, R is determined by
the competing processes of the turbulent, mixture-controlled Today, the Refractory Wear Index (RWI) is used as the measure
reaction on the one hand, and the kinetics-controlled reaction on for thermal wall loading. By transformation of Parc and Uarc and
the other hand. Both processes are modeled: The Finite Rate integration of the distance r between arc and wall, RWI is
Model (FRM) is based on a reaction kinetics approach with expressed as:
forward and backward reactions. The Eddy Dissipation Model
(EDM) assumes, analogously to the turbulent dissipation during Iarc U2arc
RWI ¼ const: : ð4Þ
vortex break-up, a very fast mixture-controlled irreversible r2
reaction. The smaller of the calculated R values is applied in
Over the years, RE and/or RWI had been modified several times,
the transport equation as the source term.[16]
and according to Dittmer and Krüger[21] with contradictory
The Eddy Dissipation Concept (EDC) model is an enhance-
results. Advanced models show a relationship between the
ment of the EDM model. The EDC model bases on the _ of the arc and
thermal wall loading by the radiation power Q arc
assumption that chemical reactions occur within the smallest
the current rating Iarc and/or resistance Rarc.[3] Dittmer and
turbulent structures. The length of the fine scales ξ ¼ f(k, ν, e) is
Krüger[21] consider the voltage drop at the anode and cathode as
introduced into the calculation of R along with the time scale
well as the melt depression caused by the jet effect of the arc. The
τ ¼ f(ν, e). Because the EDC model considers multistep
authors derive the modified radiation model:
reactions in turbulent flows, it needs a lot of computing power.
Radiation: Off-gas components, for example, CO2, H2O,
Uarc  80 V
absorb and emit radiation energy in spectral bands. The dust in Q_ arc  p8
ffiffiffiffiffiffiffi : ð5Þ
Iarc
the off-gas with a mean loading rate of 30–50 g m3 also
contributes to heat radiation.[4] Such effects can, for example, be

steel research int. 2018, 89, 1700098 1700098 (8 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

At asymmetrical meltdown of the scrap pile or excessive wall post- combustion, Figure 6b. Compared to the case without
loading, the radiation intensity can be corrected according to recirculation, about 78% more CO2 is generated.
Equation 5 strand by strand, for example, by modification of the Figure 7 shows the temperature distribution at the inner
impedance set-point. The goal is to homogeneously melt down furnace wall. Hot spots are produced by the superimposed
the scrap and to superheat the melt. radiation with different lengths of sight in the furnace. The
Equations 3–5 deliver an integral parameter which does not maximum temperature in the upper shell is 2139 K. This value is
provide any information on the location of the hot spots. Local significantly above the permissible temperature of the refractory.
overheating can only be predicted with CFD, that is, if the entire The elevated temperature, however, is explained with the partly
furnace is discretized. freely burning arcs.
Gruber[23] and Gruber et al.[24] developed a CFD model that Although the model of Gruber et al.[24] is one of the models
simulates the influence of electrical energy (arcs) and considering the largest number of phenomena in the EAF, there
chemical energy (fuel gas, charging, and injection carbon) is further need for optimization. The authors state that the
on the transport of mass and energy in the freeboard of 100 t exchange of energy between melt and/or slag on the one hand,
AC EAF (90 MVA). Melt and slag are modeled as solid and gas on the other hand would have to be modeled. The
domains, that is, not calculated. The region between slag thermal energy produced by the reaction C þ 1/2 O2 ! CO is
surface and inlet into the post-combustion chamber is transferred away from the phase boundary by both the
meshed. The interface between slag and furnace gas is circulating melt and the circulating furnace gas; the first-
considered as plane surface via which the superheat enters the mentioned phenomena is not considered in the model.
furnace. The secondary air at the slag door, electrode port, Moreover, modeling of the thermic processes in the electrode
furnace gap, and 4th roof hole are estimated. Burner, lances would have to be improved. Answers to such questions can only
and injectors are not considered. It is assumed that the be given by transient multiphase simulations with Euler-Euler
injector oxygen reacts completely with the entire carbon to models.
form CO, which is homogeneously added via the slag-gas Guo and Irons[26] examine the arc radiation in an AC EAF of
interface (CO source). The carbon consumption of the unknown size. An approach according to Alexis et al.[27] is used,
electrode and evaporation of electrode cooling water are also in which 18% of the arc energy is directly emitted to the melt and
considered. A low-pressure of 350 Pa (against ambience) is 2% is absorbed by the electrodes; 80% is available as radiation
assumed at the transition to the post-combustion chamber. energy. The arc is modeled by a CAM model. The radiating arc
Simulations for the flat-bath phase are carried out with surface is resolved by 1440 surface elements. Each element acts
ANSYS FLUENT (3D-RANS, Realizable k-e model, standard-wall as a black body and is transparent with regard to absorption of
function, FRM/EDM, DO model, approx. 4 million hexahe- radiation. Movement of the arc, gas radiation, and particle
drons). The heat emitted at the furnace shell is considered by absorption are not taken into account. The furnace wall is
heat transfer coefficients. The model contains three submodels: defined as a gray body. The effective radiation is calculated as a
Arc model, electrode model, reaction/radiation model. function of emissivity, irradiation and radiation energy of the arc,
Arc model: The arc is modeled by a time-averaged Channel Arc melt bath surface, and furnace wall. The authors perform
Model (CAM) according to Pfeifer,[25] cf. section 4.5 Electric arc separate calculations for the upper and bottom shells. The
and graphite electrode, and regarded as steady-state cylinder overlapping region of both domains is linked by a block coupling
(Darc ¼ 0.045 m, Larc ¼ 0.4 m, Tarc ¼ 5500 K). The kinetic energy technique, that is, an additional layer of control volumes is used
conversion in the arc is considered in so far as the arc absorbs in which the boundary conditions are stored and delivered to the
m_ ¼ 0.44 kg s1 of furnace gas in the plasma column, heats it up relevant domain as required.
and releases it again downstream. The lower part of the arc Figure 8a shows the heat flux density Earc of the arcs on the
immerses into the slag layer (Hsl ¼ 0.17 m, i.e., 43% shielding). extended furnace wall at Hsl ¼ 0, Larc ¼ 0.557 m, Pel ¼ 94.8 MW.
The electrode surface is a thermally coupled wall with e ¼ 1. The Earc is high in the arc area and decreases with increasing
surface temperature results from the thermal conditions distance. The hot spots result from missing slag shielding.
(conduction, convection, radiation). Analogously, Figure 8b shows the heat flux density Em emitted by
Electrode model: Joule’s thermal energy produced at Iarc ¼ 63 the melt to the furnace wall. The area surrounding the slag door
kA is considered as a volumetric heat source within the is exposed to less radiation than the area surrounding the tap-
electrode. hole. The total heat radiation emitted by the melt of about
Reaction/radiation model: The FRM/EDM model is used for 400 kW m2 is significantly lower than the arc radiation. For
the post-combustion calculation of O2, CO2, CO, H2O, and N2. Larc ¼ 0.452 m, the influence of the slag height on the radiation
The reaction of CO to CO2, the dissociation of CO2 to CO, and heat absorbed by the furnace is shown in Figure 9. As the wall
radiation (DO model) are considered. and roof are water-cooled, the total heat of both is additionally
The authors show that the entrainment and release of gas into introduced as heat extracted by cooling water. The heat
the arc (Case 1), that is, internal recirculation, considerably transferred to the wall generally decreases with increasing
affects the temperature, and CO distribution in the furnace as height of the slag.
compared to the situation without gas entrainment (Case 2), Fallah et al.[28] investigate a 200 t AC EAF to optimize the wall
Figure 6. With internal recirculation, the freeboard is heated up cooling panels and in particular the EBT area by means of CFD.
more and a hot gas strand is produced between slag door and The furnace walls are treated as gray bodies, the arcs are
off-gas elbow, Figure 6a. CO and secondary air mix at the completely covered by slag. Internal gas radiation and particle
approximate level of the slag door and intensify the absorption are neglected. ANSYS FLUENT (3D-RANS, Standard

steel research int. 2018, 89, 1700098 1700098 (9 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 6. 100 t AC EAF (90 MVA): Influence of arc region on a) Temperature distribution, b) Mass transport of CO from slag surface. Case 1: With gas
entrainment into arc. Case 2: Without gas entrainment into arc. Reproduced with permission.[24] 2016, Wiley-VCH Verlag GmbH & Co. KGaA.

k-e model) is used for the calculation. The heat flux density E at the lower bath distance and smaller elbow, Figure 11a, a clear
the furnace wall is determined for different slag temperatures of temperature gradient of 100  C results at the roof-heart level.
Tsl ¼ 1550–1700  C. This is why this region is susceptible to thermal shocks,
Figure 10 shows the heat flux density E in the refractory lining formation of cracks, and spalling of refractory material. An
and at the EBT panel for Tsl ¼ 1630  C and the water volume flow explanation might be that the extraction capacity at the off-gas
rate of V_ w ¼ 19 m3 h1. The highest loading results at the EBT elbow is too low, thus hot off-gas is recirculating in the furnace
panel (E  200 kW m2). The average heat flux density at the and additionally heats up the roof-heart. When optimizing the
furnace roof is E  100 kW m2, the values at the sidewall are furnace by increasing the distance between roof and melt as
about half as large. This behavior is explained by the fact that the
EBT panel is installed horizontally and located closer to the slag.
Further simulations confirm that the thickness Dsl of the
solidifying cover-slag increases with an increasing volume flow
rate of water V_ w. If V_ w is increased by 1 m3 h1, the water
temperature drops by about 1.25 K.
Al-Harbi et al.[29] examine two AC EAFs of unknown size. The
furnaces can be distinguished in so far as the distance between
roof and melt differs by 0.4 m, and the dimension of the off-gas
duct is larger by approx. 40%. The melt is to be seen as porous
medium and considered as oxygen sink and CO source,
respectively. Slag is neglected, the slag door is closed. A CO-
air mixture is assumed as furnace gas. The entire oxygen reacts
with carbon, the CO produced is implemented at the phase
boundary as source term (3D-RANS, Standard k-e model, 0.60
million cells). The numerical approach of defining the oxygen
injector as a tube of length L ¼ 1 m to be able to simulate the
coherent supersonic jet in a better way and to prevent its early
spreading is quite interesting. That way the required compress-
ible calculation can be replaced with good accuracy by an
incompressible calculation, cf. section 4.7 Gas burner, injector.
Figure 7. 100 t AC EAF (90 MVA): Temperature distribution on the inner
Figure 11 shows the temperature distribution at the uncooled wall, case 1: With gas entrainment into arc. Reproduced with
roof-heart and the off-gas elbow of both furnaces. In the case of permission.[24] 2016, Wiley-VCH Verlag GmbH & Co. KGaA.

steel research int. 2018, 89, 1700098 1700098 (10 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

well as the off-gas elbow, Figure 11b, the thermal loading is the simulation. Side wall and roof are assumed to be
reduced and the temperature gradient drops down to 60  C. smooth, slag-covered walls. Bottom shell and roof-heart are
Although the extraction capacity is higher, the off-gas velocity modeled adiabatically because the removed heat can be
decreases, thus increasing thermal stability and reducing neglected as compared to the heat in the water- cooled area.
erosion. Melt and slag (Hsl ¼ 0.2 m) are considered as being fixed
Sanchez et al.[30] simulate the hot spot distribution for a areas in terms of space and are not calculated. The furnace
210 t AC EAF with ANSYS FLUENT (3D-RANS, Standard k-e entrains secondary air through slag door, cover gap, and
model, DO model, 72 604 cells). The arc is simulated by a electrode port. Three minutes of the flat-bath phase are
CAM model (Uarc ¼ 1210 V, Parc ¼ 120.6 MW, Larc ¼ 0.45 m), it considered and it is assumed that m _ C ¼ 0.317 kg s1 escapes
emits 25% of its energy into the furnace of 1945 K. In the case from the furnace.
of the freely burning arc (Hsl ¼ 0 m), Figure 12a, a tempera- Figure 13a shows the temperature at the lower shell, on the
ture of T  2100 K arise in the hot spots at the wall, and electrodes, and on the arcs; the maximum temperature is
T  1920 K in the hottest spots of the roof-heart. If the arc is Tmax ¼ 1870 K. The temperature level at the upper shell is lower
completely covered (Hsl ¼ 0.45 m), Figure 12b, the tempera- as a result of water cooling, Figure 13b, with Tmax ¼ 1350 K
ture is only about 1660 K and/or 1760 K. Hot spots cannot be here. Figure 13c illustrates the situation at the furnace roof.
noticed at the furnace wall any longer. The authors conclude While the temperature of the cooled roof panel is only
that at least 75% of the arc has to be shielded for reliable approx. 1300 K, it increases up to approx. 1610 K in the
operation. roof-heart.
Pfeifer et al.[31] simulate an AC EAF to be used as 120 t Logar et al.[32–37] developed mathematical models to describe
furnace for stainless steel or as 140 t furnace for structural most of the physical phenomena of the EAF process. The models
steel production. As in refs.,[26,30] a CAM model for the arc consider heat and mass transfer and thermo-chemical reactions
(Uarc ¼ 340 V, Iarc ¼ 64 kA, L arc ¼ 0.3 m, Darc ¼ 0.045 m) is inside the furnace. The modeling of the radiative heat transfer
used. ANSYS FLUENT (3D-RANS, Realizable k-e model, mainly based on the time-dependent evaluation of the
standard wall function, FRM/EDM, DO model) is taken for temperature, area and view factors between five zones (roof,
upper walls, solid steel, liquid steel, arcs). The arc-heat
distribution model is based on a modified CAM model. The
required variables are estimated by empirical equations and
results from MHD models. The resulting calculation algorithms
are solved with Matlab 2013b in approx. 77 μs and reproduce well
the measured EAF data. These simplified and adapted models
have the advantage of short computational times, but their
accuracy seems to be not comparable with the accuracy of latest
CFD models.
To summarize, the thermal loading of the furnace wall can
already be calculated well by CFD. The loading is predominantly
caused by the arc radiation. Appropriate countermeasures are
the optimized foamy slag practice, and the adaption of the
furnace geometry to the prevailing flow conditions.

Figure 8. a) Settlement of heat flux density Earc of arc radiation on furnace


walls, Hsl ¼ 0, Larc ¼ 0.557 m, Pel ¼ 94.8 MW, b) Settlement of heat flux
density Em of bath surface radiation on furnace wall, Hsl ¼ 0, Tm ¼ Figure 9. Effect of increasing slag height on absorbed radiation heat for
1500  C. Reproduced with permission.[26] 2003, Commonwealth Scientific Larc ¼ 0.452 m. Reproduced with permission.[26] 2003, Commonwealth
and Industrial Research Organisation (CSIRO). Scientific and Industrial Research Organisation (CSIRO).

steel research int. 2018, 89, 1700098 1700098 (11 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

tap-hole to wear quickly. Typical life times are a few hundred


charges. The shape of the tapping orifice determines the velocity
and turbulence of the melt, and consequently the erosion rate.
The inlet section should be nozzle shaped in order to achieve an
erosion-less flow.
Kirschen et al.[38] use ANSYS FLUENT (3D-URANS, Realizable
k-e model, 3-phase VoF, 0.700 million cells) to examine conical and
cylindrical tapping devices for five types of EAFs with tapping
weights from 85 to 250 t. Figure 14 illustrates that the rounded,
conical tap-hole, Figure 14a, induces a homogeneous erosion-less
flow with a thinner boundary layer as compared to the sharp-edge
cylindrical tap-hole, Figure 14b. The turbulence is reduced by
Figure 10. 200 t AC EAF: Heat flux distribution E at inner furnace walls 30–45% at mean Reynolds numbers of Re  15  105, thus
(Tsl ¼ 1630  C, V_ w ¼ 19 m3h1). Reproduced with permission. 2008, Polish
[28]
resulting in a more stable melt stream and reduced wear. At the
Association of Metallurgical Engineers and Technicians. same time the tapping rate is increased. The smaller the orifice
diameter, the later the slag will be entrained into the tap-hole via the
free melt surface; that is, thrombus effects and vortexing above the
4.2. Bottom Tapping
tapping orifice are reduced. The static pressure in the tapping
orifice is on a higher level with the conical version, hence
Eccentric Bottom Tapping (EBT) has been frequently used since
increasing the life time of the MgO–C bricks.
the 1980s. In combination with semi-automatic slag detection
systems, it prevents slag carryover and nitrogen pickup. The
furnace can be tapped within very short time (50–80 t min1). 4.3. Refractory Lining, Cooling Panel
The geometry of the tapping orifice varies, depending on the
furnace size. Typical values are Ltap ¼ 0.8–1.2 m, and Dtap ¼ 0.15– Both the refractory bricks and the water-cooled wall or roof
0.25 m. The high flow velocity um, with um ¼ f(Hm, Hsl, Ltap, panels are exposed to extreme mechanical and thermal loading.
Dtap,..) < 5 m s1, and high melt temperature Tm causes the The lining has to withstand the aggressive FeO containing slag.

Figure 11. Horizontal temperature distribution at roof heart and off-gas duct. a) Small distance between melt and roof heart. b) Large distance between
melt level and roof heart. Reproduced with permission.[29] 2011, Steel Institute VDEh.

steel research int. 2018, 89, 1700098 1700098 (12 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 12. 210 t AC EAF: Hot spot formation at furnace shell and furnace roof, delta, and 4th roof hole section for slag height a) Hsl ¼ 0 m, b)
Hsl ¼ 0.45 m. Reproduced with permission.[30] 2012, The Iron and Steel Institute of Japan.

From production and safety-related viewpoints, it is important to the EAF flat-bath phase. However, until today such technologies
know the temperature gradient in the refractory lining. to compensate the power loss failed to establish themselves.
Mombeni et al.[39] examine a water-cooled roof panel of an AC
EAF with ANSYS FLUENT (3D-RANS, Standard k-e model, 3.57
million cells). For this purpose, the water tubes of one roof panel 4.4. Melt and Slag Movement
of the entire roof structure consisting of eight units are
examined. The heat flux density E ¼ 108 kW m2 transferred to Calculation of the melt and slag movement involves considerable
the panel is calculated from the water flow rate, the specific heat effort due to the large furnace geometry and the resulting
capacity, and the temperature difference between water inlet and number of grid cells. In addition, extensive modeling of the
outlet. Figure 15a shows the water temperature of the phases (melt, slag, furnace gas, stirring gas, combustion
conventional roof panel. The distribution is inhomogeneous, components) and metallurgical processes is required. Therefore,
the highest loading results in the panel center (Tw ¼ 380 K), the many authors[12,14,43–51] forgo holistic CFD simulations for the
maximum temperature gradient at the panel is ΔTw ¼ 50 K. EAF and concentrate on individual phenomena.
Possible consequences are thermal fatigue, formation of cracks, Foamy slag: Foamy slag protects the refractory lining from the
and water leaks. However, when using a concentric pipework arc radiation and allows high melting capacities.[48] It protects
concept, Figure 15b, the heat loading is better distributed, the the melt from cooling due to thermal insulation and increases
maximum temperature gradient decreases to ΔTw ¼ 20 K, and furnace efficiency.[49] Foamy slags are created by injecting carbon
the service life of the roof panel is extended. particles into the slag via a carrier gas (N2, air). CO bubbles
Thermal overloads which may cause water leaks cannot be nucleate at the surface of the carbon particles and after
excluded, particularly in the hot spots. However, to be able to separation grow to an average bubble size of D  2 mm.[52]
continue operating with a damaged wall panel, Fabbro[40,41] Slag foaming can, in certain ways, be compared with milk
reports on a concept verified by 2D-CFD simulations. With two boiling over on a kitchen stove. However, even for such a small
panel rows arranged behind each other, the rear row can also be computational domain, a CFD simulation will not be possible in
operated alone in case of damage, Figure 16. Moreover, the the near future. Among others, this is due to the fact that
power loss is reduced thanks to the thicker protective slag layer. individual air bubbles in the milk froth would have to be
Borlée et al.[42] propose measures to reduce the power loss. For discretized in terms of space and time. Additionally, the
this purpose, a layer of refractory material with high thermal chemical composition has a high influence on slag properties.
resistance is applied on the water-cooled wall panels. Special Although a great deal of research has been conducted toward the
anchors, which have to resist the high temperature are necessary fundamental understanding of slag properties on the stability of
to ensure that the layer remains on the tubes. CFD is used to slag foaming,[53–55] at this moment a holistic CFD model that
calculate the temperatures at the pipework and the anchoring for combine chemical and fluid dynamic phenomena does not exist.

steel research int. 2018, 89, 1700098 1700098 (13 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Sattar et al.[56] develop a foamy slag model to


predict the slag height. For a small cylinder filled
with slag (Hsl ¼ 0.04 m), argon is blown on top via
a nozzle (D ¼ 0.0016 m). The argon velocity varies
between 1  102 and 3  102 m s1. The model
determines the slag height as a function of gas
velocity and FeO content. To do this, an Euler-Euler
multiphase model is combined with a Population
Balance model. The growth of the foam bubbles is
considered as well as the coalescence according to
Prince and Blanch[57] and the decomposition of
large foam bubbles according to Luo and
Svendsen[58] PHOENIX is used as CFD frame-
work, the slag model is programmed in FORTRAN
and integrated by subroutines. With the argon
velocity increasing the height of the foam layer
rises for a certain slag analysis.
Mixing and homogenization: The AC EAF is a
good example of a reactor with minor mixing
ability. The turbulence intensity in the melt is
proportionately to the scrap meltdown rate and the
temperature distribution.[14] The melt and slag
flow results from several superimposed mecha-
nisms, some of which had been examined in
literature:

Buoyant forces resulting from thermal gradients


(free convection)

Gonzáles et al.[14] calculate velocity and tempera-


ture of the melt as well as the mixing time of a 210 t
AC EAF. The study is done with PHOENICS (3D-
RANS, Standard k-e model, Boussinesq ap-
proach[16]), the mixing time τ99 is determined by
a FORTRAN subroutine. The model assumes that
the domain is completely filled with melt of
constant density. Radiation, electromagnetic
forces, and mixing caused by the injector oxygen
are neglected. The arcs are modeled with the CAM
model according to Sanchez et al.[50] (Uarc ¼ 1210
V, Parc ¼ 95–120 MW, Larc ¼ 0.25–0.45 m).
Figure 17 shows the influence of the arc length
Larc on the melt flow pattern. It is characterized by
two spacious vortices and a mean velocity of
um ¼ 0.05 m s1. The melt flows up in the furnace
center, is radially deflected, and flows down again at
the wall. While the arc length has hardly any
influence on the flow field, the temperature field is
clearly affected. If the arc length increases from
Larc ¼ 0.25 to 0.45 m, the mean melt temperature
will rise from Tm ¼ 1675  C to 1716  C, that is, by
41  C. This is important for the operator because
Figure 13. 120 t AC EAF: Temperature distribution. a) Lower vessel, electrode surfaces, superheating allows charging of DRI. If the arc
arcs. b) Upper vessel, electrode surfaces, arcs. c) Slag surface, electrodes, arcs, roof. length Larc is short, undesirable thermal layering can
Reproduced with permission.[31] 2012, Publications Office of the European Union. be observed in the melt which should be counter-
acted by using bottom stirring equipment.
In the further course, Gonzáles et al.[14]
determine the convective heat transfer coefficient

steel research int. 2018, 89, 1700098 1700098 (14 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 14. 85 t AC EAF, tap diameter 0.170 m: Melt velocity in a vertical cross section through the tap area. a) Conical shape. b) Conventional cylindrical
shape. Reproduced with permission.[38] 2008, Polish Association of Metallurgical Engineers and Technicians.

h (cal  cm2  s1   C1) as a measure for the intensity of the nodes, P ¼ 120.6 MW, Uarc ¼ 1210 V, Larc ¼ 0.45 m). Without
convective flow. The parameter h is a function of particle size slag, the mean melt velocity is um ¼ 0.05 m s1, which coincides
Dp, thermal conductivity λ, Re-number, and Pr-number. Based on with the data of Gonzales et al.[14] Assuming Hsl ¼ 0.45 m,
this, three furnace flow domains can be classified: low um ¼ 0.02 m s1 results. The reduced velocity can be attributed to
mixing (h ¼ 0.5–0.7 cal  cm2  s1   C1), intermediate mixing higher friction at the melt-slag interface. As a rule, the increase
(h ¼ 0.7–1.1 cal  cm2  s1   C1), strong mixing (h > 1.5 of slag viscosity, for example, caused by changes of the chemical
cal  cm2  s1   C1). composition, causes convection in the melt to decrease and the
Figure 18a and b show the h-value at Larc ¼ 0.45 m. Although thermal layer formation to increase; figuratively speaking the
the highest temperature occurs in the vessel center, the h- cover-slag dampens the movement of the free melt surface.
values are low here due to the low velocity. Figure 18c and d Thermal conductivity of the slag, however, only has a minor
show the Residence Time Distribution (RTD) curves of a influence on convection and thermal layer formation of the melt.
scalar tracer added to the computational domain. The local
tracer concentration is detected at ten monitor points and
normalized with the equilibrium concentration. As soon as all Oxygen and gas jets impinging the melt surface
points have reached unity, the mixture is considered being
homogenized. The investigation shows that the mixing time An empirical model for the penetration depth of a supersonic
τ99 is longer with a short arc (τ99 ¼ 9.5 min) than with a long oxygen jet in melt is presented by Koria and Lange.[59] Although
arc (τ99 ¼ 9.0 min). the approach was originally developed for converters, it can be
Guzmán et al.[46] get similar results examining the influence transferred to the EAF as the inclination angle of the gas jet is a
of the slag height Hsl on convection effects in a 220 t AC EAF model parameter. The approach shows good coincidences with
using ANSYS FLUENT (Mixture multiphase model, 93 534 the research of Odenthal et al.[60,61]

Figure 15. Temperature of cooling water in the hot side of the a) conventional panel, b) proposed panel. Reproduced with permission.[39] 2016, Elsevier
B. V., Amsterdam.

steel research int. 2018, 89, 1700098 1700098 (15 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 16. Long-life, energy-saving panels, thermal profile study, and in operation filled with slag. Reproduced with permission.[41] 2017, Danieli.

Chen et al.[44] develop an ANSYS FLUENT model (3D- produced that increase turbulence and improve mixing. The
URANS, Standard k-e model, 3-phase-VoF, 1.025 million cells) highest velocities are found in the area of the oxygen impact
for the melt agitation in a 150 t AC EAF with oxygen injectors. points, the lowest close to the furnace bottom. The authors
Special attention is given to the time- dependent behavior of slag, recommend installing bottom stirring equipment.
melt, and oxygen in the vicinity of the supersonic oxygen jet. All
fluid properties are constant, oxygen is considered as being
compressible. Free convection and electromagnetic forces are Stirring by inert gas (N2, Ar) and/or CO gas in the
not considered. The pressure–velocity correction is done via the decarburization zone
PISO algorithm. The computing time is 120 h for 2 s of process
time. Liu et al.[47] develop a multiphase model for a bottom-stirred 75 t
Figure 19 shows the melt velocity at V_ ox ¼ 1450 m3 h1 and AC EAF and compare six different stirring concepts. The mixing
2000 m3 h1. Vortices recirculating the slag are induced below time is measured at the downsized water model, the melt
the impact point of the oxygen jet. The location of the vortex movements are simulated using ANSYS FLUENT (3D-URANS,
center is a function of the blowing rate V_ ox and is located at a Standard k-e model, 3-phase-VoF, 43 722 cells). The furnace walls
distance of 0.75–1.93 m from the oxygen impact point just below are adiabatic, electromagnetic forces, and chemical reactions are
the melt surface. At V_ ox ¼ 2000 m3 h1 secondary vortices are not considered. All stirring concepts have in common that three

Figure 17. 210 t AC EAF: Velocity and temperature field as a function of arc length a) Larc ¼ 0.25 m, b) Larc ¼ 0.45 m. Reproduced with permission.[14]
2010, The Iron and Steel Institute of Japan.

steel research int. 2018, 89, 1700098 1700098 (16 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 18. 210 t AC EAF, Larc ¼ 0.45 m: a, b) Contours of convective heat transfer coefficient h [cal cm2 s1  C1], cross sectional and top view. c) Position of
tracer injection and monitoring points. d) Concentration curves. Reproduced with permission.[14] 2010, The Iron and Steel Institute of Japan.

stirrers are concentrically arranged and one stirrer position is 4RB1. The greater the distance between stirrer and furnace wall,
varied. The distance of the stirrers from the center point varies the higher the velocity and the smaller the mixing time,
for the different stirring concepts and is between 40 and 60% of respectively. Boakuan and Jicheng[43] come to similar results.
the furnace radius. They observe that the velocity of the melt and/or mixing is not so
Figure 20 shows the velocity fields with different stirring much depending on the stirrer diameter, but rather on the
concepts. Designation 4RB1 means that the stirrer B is set at number and position of the stirrers. In general, stirring concepts
position 1 and the distance to the center point is 40% of the using eccentrically arranged stirrers show improved mixing
furnace radius. The highest velocity is observed for concept behavior, cf.[62]

Figure 19. 150 t AC EAF: Velocity distribution in melt domain at 0.2 m down the bath surface and 2 s after start of blowing; injector oxygen flow rate
a) V_ ox ¼ 1450 m3h1, b) V_ ox ¼ 2000 m3 h1 . Reproduced with permission.[44] 2012, Trans Tech Publications Inc.

steel research int. 2018, 89, 1700098 1700098 (17 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 20. 75 t AC EAF: Comparison of different bottom stirring concepts based on velocity flow field in longitudinal section. 4RB1 means three stirrers
uniformly distributed at 40% of furnace bath radius R at locations A, B1, C. Flow regions A, B indicate high melt velocity, flow regions C, D moderate melt
velocity. Reproduced with permission.[47] 2015, The Iron and Steel Institute of Japan.

The overall flow field as a rule influences mixing and short arc and high current. Compared with this, the AC EAF
homogenization of the melt. High velocities increase the scrap superficially stirs between the electrodes only. Here, the stirring
meltdown efficiency and reduce the formation of thermal layers, effect is the higher the longer the arcs are. In general, the
but at the same time increase wear of the refractory lining in the relationship between arc length and melt temperature is
freeboard. described in a study of Gonzales et al.[14] and can be seen in
Figure 21. Looking at the isothermal T ¼ 1665  C, it can be seen
4.5. Electric Arc and Graphite Electrode for the short arc, Figure 21a, that this area is inhomogeneous and
shows low average velocities and temperatures (Larc ¼ 0.25 m:
Due to the complex physics, it is difficult to observe and um ¼ 0.48 m s1, Tm ¼ 1675  C). For the long arc, Figure 21b, the
investigate the arc as the name-giving part of the EAF. As long as area is homogeneous and shows higher values (Larc ¼ 0.45 m:
these phenomena are not known to every detail, CFD results um ¼ 0.51 m s1, Tm ¼ 1716  C).
involving the arc should be examined critically. However, two Wang et al.[63] develop a rotationally symmetric arc model for
methods have emerged for modeling the arc so far: the DC EAF. The electrode diameter is D ¼ 0.16 m, the arc length
Larc ¼ 0.07–0.25 m, the melt surface is smooth, and the material
1) CAM models approximate the arc in a simplified way as a properties are independent of temperature. Below the electrode,
cylinder fixed in space and time with constant boundary T ¼ 4000 K is assumed, the furnace ambience is T1 ¼ 1800 K,
conditions. They assume that the arc consist of two zones; a the average current density is I ¼ 4.4  107 A m2. PHOENICS
hot electrically conductive and cylindrical zone in the inner (2D-RANS, rotational symmetric, Standard k-e model, standard
arc referred to as the electrical conductivity zone, and an wall function, 3696 cells) is used for the simulation. The
outer energy emitting zone. In the electrical conductivity calculated temperature in the arc (Iarc ¼ 2160 A, Larc ¼ 0.07 m) is
zone the arc temperature T is assumed being independent of T ¼ 10 000–16 000 K, the velocity is 1600 m s1. These findings
the location coordinates r, ϕ, and z, and calculated from the are consistent with experimental results of Bowmann et al.[3]
energy balance of the supplied and discharged energy; inside Direct below the arc the heat flux density is E ¼ 120 MW m2. A
the cylinder the temperature is T ¼const., thermal conduc- vortex is induced in the melt as a result of the electromagnetic
tivity λ ¼ 1, and electrical conductivity σ ¼ σ (T, p). In the field. Compared with the electromagnetic stirring, the melt
energy discharge zone outside the cylinder applies λ ¼ λ(T, p) agitation by the thermal buoyant force and the jet momentum
and σ ¼ 0. The temperature drops down from the inside out can be neglected from the authors’ viewpoint.
until reaching furnace temperature T1. The energy is Ramirez-Argaez[64] investigates a cylindrical subdomain of a
transported from the electrical conductivity zone to the DC EAF (D ¼ 3 m, Hm ¼ 0.5 m, Iarc ¼ 40 kA, Larc ¼ 0.25 m). The
energy discharge zone by radiation and convection.[3,25] hot furnace wall of 1773 K is considered as being adiabatic, the
2) Alternatively, the arc can be simulated in terms of space and bottom shell is used as zero potential (anode). The coupled NS
time with all the physical phenomena. This is possible so far and Maxwell equations are solved with PHOENICS (2D-RANS,
for small arc domains only, with the rest of the furnace rotational symmetric, Standard k-e model, standard wall
remaining unconsidered. In most cases, the DC EAF is function, 1444 cells for arc domain, 2400 cells for melt domain).
investigated because it can be regarded rotationally symmet- Temperatures up to T ¼ 27 000 K are calculated below the
ric and therefore, the NS and Maxwell equations can be electrode. The arc emits 3 MW in total, approximately half via
simplified. Contrary, the solution for the AC EAF has to be convection and a quarter via radiation and electric conduction
three-dimensional and highly resolved in terms of time due into the melt. Vortices are produced in the melt that move
to the mains frequency. downward at u  1.2 m s1 below the electrode and upward at the
wall. Despite the intensive convection induced by the electric
The DC EAF intensively stirs the melt as the electric current field, a high temperature gradient of T ¼ 210 K is prevailing in
flows directly through the melt. The stirring effect is biggest with the melt. Ramirez-Argaez et al.[10] examine the influence of the

steel research int. 2018, 89, 1700098 1700098 (18 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 21. 210 t AC EAF: Temperature fields as a function of arc length. Reproduced with permission.[14] 2010, The Iron and Steel Institute of Japan.

gas composition on the arc power Parc with a similar setup. These Rafiei et al.[66] examine the influence of temperature on the
results show that for Larc ¼ 0.3 m and Iarc ¼ 44 kA the arc power electrode consumption (D ¼ 0.6 m, L ¼ 7.2 m). The basis is a
varies from Parc ¼ 24.64 MW (for argon) via Parc ¼ 33.92 MW steady-state ANSYS-Mechanical APDL simulation (3D, 1800
(for N2) to Parc ¼ 60.8 MW (for air), Figure 22. elements). At the electrode tip the temperature decreases over a
Frittella et al.[65] calculate the spatially and temporally resolved length of about 0.4 m from 3600 to 1500 K. The results show that
arc behavior for a 100 t AC EAF (Pel ¼ 75 MW, Uarc ¼ 629 V, the electrodes influence each other with respect to their surface
Iarc ¼ 48 kA). The electrodes (D ¼ 0.6 m) are arranged on a pitch temperature.
circle D ¼ 1.150 m, their distance to the melt surface is One conclusion is that the physical effects in the arc can only
H ¼ 0.6 m. ANSYS FLUENT (3D-URANS, species transport be simulated by submodels. Unless it is intended to simulate a
model, net emission coefficient radiation model, 0.90 million small area around the arc only, CAM models offer a good
polyhedrons, Δt ¼ 2 ms) is linked with the in-house code UNIBO compromise between physical accuracy and computing effort.
for separate plasma modeling approaches. A mixture of CO,
CO2, and Fe-vapor is taken as furnace gas, with the latter
component increasing the plasma conductivity and reducing the 4.6. Electrode Supporting Arms
voltage between the electrodes. The Lorentz force steers the arc
radially toward the outside, Figure 23, thus causing the arc to A component seldom simulated is the electrode supporting arm.
become longer. Four MW are emitted as radiation at a Amongst others, this is due to the fact that the interaction
temperature of Tarc ¼ 15 000 K and an arc power of Parc ¼ 140 between the electromagnetic field and the mechanical structure
MW, which is very low compared with the data of other authors. is complex and requires high computing effort. Farschtschi[67]
The measured values of current and voltage from the field test uses the Finite Network Method (FNM) developed to calculate
could be reproduced by the simulation, but only when the iron three-dimensional fields in electrically conductive materials
(Fe) vapor is considered. without the surrounding space. The volume of the current-
conducting conductor is discretized and the system transferred
in an electro-dynamically linked network with a large number of

Figure 22. Integrated radial heat flow below the arc as a function of gas Figure 23. Voltage potential [V] on the 6000 K temperature iso-surface.
species. Reproduced with permission.[10] 2009, The Iron and Steel Reproduced with permission.[65] 2012, Austrian Society for Metallurgy and
Institute of Japan. Materials (ASMET).

steel research int. 2018, 89, 1700098 1700098 (19 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

injectors. As a basic rule the injector should not be arranged too


close to the melt because it may be destroyed by thermal loading.
Consequently, higher installation positions are desirable.
However, the injector gets further away from the melt surface,
for example, more than two meters, particularly in the flat-bath
phase, or at the end of the furnace campaign. If the injector
nozzle is in such case not designed correctly, gas-dynamic effects
(e.g., straight and oblique shock waves, expansion waves) inside
and outside the nozzle will cause wear of the nozzle edge and
significant reduction of the jet velocity, cf.[61,70] The oxygen jet
spreads, the jet momentum is lost, and the amount of oxygen
reaching the melt is insufficient. Possibly, the jet momentum
may become so low that the foamy slag is  depending on its
viscosity  no longer displaced and the oxygen does not reach
the melt any longer. The main requirement on the injector is that
Figure 24. Three dimensional current density distribution I. Reproduced
it delivers large amounts of oxygen, is located in retracted
with permission.[67] 2011, Quartz Business Media Ltd, Redhill. position in the wall panel, and so far away from the melt that no
melt/slag spatters get stuck.
To maximize the jet momentum, supersonic nozzles (Laval
ohmic resistors and inductivities. An equivalent circuit diagram nozzle, convergent-divergent nozzle) are used. The Laval nozzle
is prepared for this network. It is possible that way to calculate consists of a convergent inlet section and a divergent outlet
the current density distribution in detail for an overall furnace section (0: Inlet;  : critical, 1: outlet), Figure 25a. In the subsonic
system with electrode supporting arms, electrodes, and supply part oxygen is accelerated, initially reaches supersonic speed
lines, Figure 24. (M ¼ 1) in the throat, and further accelerates in the divergent part
It is possible to simulate the electrode support arm system as a result of expansion. The Mach number is the ratio of the
with the Finite Element Method (FEM) using high-performance local gas velocitypffiffiffiffiffiffiffiffi u ffi and the speed of sound a
computers. The Maxwell equations for the system and the (M ¼ u=a with a ¼ κRT ). The supersonic jet enters with
affected environment are completely computed in the process. M  2 into the hot furnace. The jet is characterized by the
Brusca[68] describes such a model used to calculate current flow, supersonic region (supersonic jet length L ) and the subsonic
temperature, and resulting forces at electrode supporting arms, region downstream. In the potential core u, p, T are constant
masts, and electrodes. along the jet axis. L /D1  10 applies as the approximation
formula for a cold oxygen jet entering into a cold environment.
L /D1  20–30 applies for a cold oxygen jet entering into a hot
4.7. Gas Burner, Injector environment of approx. 1600  C.
Supersonic nozzles are designed using the isentropic theory
Gas burners, oxygen, and coke injectors support the meltdown assuming one- dimensional, steady-state, frictionless flow of gas
and superheating process in the EAF. Irons[69] gives a detailed without heat exchange with the environment. As a result,
overview of the state of the art. pressure p, temperature T, density ρ, speed of sound a are
Currently combined gas burner–oxygen injectors are used in calculated as a function of the Mach number M, and the
the EAF, which are always referred to in the following as stagnation values (0). As the isentropic theory considers

Figure 25. Notation: a) Conventional nozzle. b) Coaxial nozzle with primary (p) supersonic nozzle (Mp > 1) and secondary (s) subsonic nozzle (Ms < 1);
cold oxygen is supplied to the primary nozzle and hot CO to the secondary nozzle; Mcsp: Convective Mach number between secondary (s) and primary
(p) jet; Mc1s: Convective Mach number between ambience (1) and secondary (s) jet. Reproduced with permission.[78] 2015, Steel Institute VDEh.

steel research int. 2018, 89, 1700098 1700098 (20 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 26. Combustion flame structure for supersonic oxygen jet. Reproduced with permission.[75] 2010, Association for Iron & Steel Technology.

cross-sections ( ) and (1) only, it is more advantageous to use the Allemand et al.[73] examine, pressure, and temperature of the
method of characteristics for nozzle design. This is a method for oxygen and the ambient temperature on the supersonic jet
solving the partial gas-dynamic differential equation for axially length L of a single nozzle. For this, adjusted and unadjusted
symmetrical, steady-state, frictionless, and irrotational flows. It nozzles are simulated and compared. As the isentropic theory
uses the Mach lines, that is, the lines of weak pressure does not give information on the opening angle of the divergent
disturbances which propagate at sonic speed and are arranged at nozzle section, the nozzle contour is optimized by the method of
Mach’s angle relative to the local velocity vector, as the basis for characteristics. The contour along the nozzle axis is approxi-
the right-running and left-running characteristics. It is possible mated by radii, thus resulting in a non-separated, low-turbulence
to solve the gas-dynamic differential equation along these nozzle flow. The authors establish that a long nozzle reduces
characteristics.[70,71] turbulence at the nozzle outlet and increases L . For a nozzle
It is the physical principle of a supersonic nozzle that its with V_ ox ¼ 2500 Nm3 h1, pox ¼ 10 bar, M ¼ 2.1, ANSYS FLU-
contour can only be designed for a single design point. If, during ENT is used to simulate a domain of the length L/D1 ¼ 100.
the blowing process, just one of the operating variables p0, T0, When comparing a conventional with an optimized nozzle, the
V_ 0 , p1 deviates from the design parameters, the nozzle will work latter case shows a 50% greater supersonic jet length. If T1 is
inefficiently and the outlet edges will wear extremely fast. increased from 300 to 2000 K, L triples. The increase of pox
While formerly single supersonic nozzles were used, now the extends L only slightly. If T0 is reduced, L becomes shorter. In
trend is toward coaxial nozzles, Figure 25b. The coaxial nozzle is the field trial on a 145 t AC EAF (80 MVA) an optimized injector
characterized by a central primary (p) supersonic oxygen nozzle (V_ ox ¼ 2500 Nm3 h1, M ¼ 2.1) is applied at 50 to the horizontal
surrounded by one or several secondary (s) subsonic coaxial jets. and the distance of 1.5 m to the melt; the theoretical supersonic
The coaxial and/or secondary nozzle consists of circularly jet length is L ¼ 1.4 m. The injector was used for nine months
arranged bore holes or a consistent coaxial gap, refer to the detail and produced a more homogeneous melt temperature, less tap-
sketch in Figure 26. The coaxial jet principle has been known for hole clogging, reduced the power-on time by one minute, and the
a long time. It is based on the shrouding of the primary jet, and power consumption at the EAF by 4 kWh t1.
consequently the density and the viscosity of the primary jet, Malfa et al.[74] examine a coaxial jet injector with V_ ox ¼
secondary jet, and environment can be adjusted to each other via 900–2500 Nm3 h1 and pox ¼ 9–15 bar. The supersonic jet length
the respective velocity and temperature of the jets.[72] The relative is increased up to L ¼ 2.5 m by a freely burning subsonic coaxial
velocity of individual vortices in the shear layer between the jet. The shrouding gas is generated by two concentric ring gaps;
primary and secondary jet is reduced and less fluid is entrained the flow through the inner ring gap is V_ CH4 ¼ 90–135 Nm3 h1,
into the primary jet from the environment, thus reducing the jet and V_ ox ¼ 90 Nm3 h1 through the outer ring gap. The calculated
growth rate and increasing the supersonic jet length L . For the domain (L ¼ 2.5 m, D ¼ 0.5 m) is calculated with ANSYS
identical mounting location and flow rate of the injector, the FLUENT (3D-URANS, Standard k-ω model, EDM (CH4, O2,
longer L goes along with a higher jet momentum. Hence, it is CO, CO2, H2O, N2)).
conceivable arranging the coaxial nozzle injector further away If V_ CH4 ¼ 90 Nm3 h1 and/or 135 Nm3 h1 are added through
from the melt bath because it still delivers the same amount of the inner ring gap, L increases by 30% and/or 36%. If the outer
oxygen per time unit into the melt as a single nozzle injector. ring gap is additionally charged with V_ ox ¼ 90 Nm3 h1, L
The phenomena of a supersonic oxygen jet in a hot furnace increases by additional 5%, respectively. Three optimized injectors
environment can neither be measured in situ nor modeled in the (V_ ox ¼ 1600 Nm3 h1, pox ¼ 11 bar, V_ CH4 ¼ 125 Nm3 h–1) are used
laboratory. Therefore, CFD is an integral part of injector for 920 melts on the 100 t AC EAF. Using the injectors, the input of
development. oxygen could be improved by approx. 50%.

steel research int. 2018, 89, 1700098 1700098 (21 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Alam et al.[75,76] simulate a coaxial jet injector for natural gas potential core to get longer by four- times. If nitrogen is
and oxygen. 16 holes (Ds ¼ 0.00287 m) for natural gas and 16 injected as the primary gas instead of oxygen, L gets longer by
holes (Ds ¼ 0.00408 m) for oxygen are arranged on two three-times only. The reason for the jet lengthening is that the
concentric circles around the primary nozzle (Dp ¼ 0.0147 m), flame suppresses the mixing with the ambience and reduces
Figure 26. The holes in the rotationally symmetrical domain the growth of the turbulent shear layer, Figure 26. The flow
(x/Dp ¼ 105, r/Dp ¼ 20) are assumed to be consistent ring gaps remains parallel in the potential core and spreads downstream
with an equivalent surface area. The inner section of the only.
supersonic nozzle is not calculated. For the defined inlet Further studies on the coaxial supersonic jet in a hot furnace
variables (Primary nozzle: Mox,p ¼ 2.1, pox,p ¼ 9.145 bar, atmosphere (1) are carried out by Klioutchnikov et al.[19] It is the
Tox,p ¼ 298 K, secondary nozzle: m _ ox;s ¼ 3.488  105 kg s1, objective to maximize the supersonic jet length L of the primary
5 1
m_ CH4;s ¼ 1.833  10 kg s ) the outlet variables are computed (p) cold supersonic oxygen jet with the aid of the secondary (s) hot
by the isentropic theory (uox,p ¼ 504 m s1, Tox,p ¼ 159 K). AVL subsonic coaxial jet. The slower secondary jet (CO) protects the
FIRE (2D-URANS, rotational-symmetric, Standard k-e model, shear layer of the faster primary jet (oxygen) toward the hot furnace
one-step combustion model (CH4, O2, CO, CO2), Δt ¼ 105 s, atmosphere, thus causing L to increase by approx. 35%. The
WSGGM model) is used for the simulation. The grid sensitivity studies of Murakami and Papamoschou[72] are used as the basis for
is checked by three grids, however, a total of 28 000 cells is DNS simulations for different combinations of a cold oxygen jet
finally considered. The coherent flame causes the supersonic and a hot coaxial jet. The analysis shows that L is a function of

Figure 27. 3D-LES simulation, instantaneous velocity distribution in hot furnace atmosphere. a) Standard injector with flat walls, designed by isentropic
theory. b) Advanced injector (coaxial jet with primary and secondary nozzle) designed by CARD and COAX software; both cases: V_ ox ¼ 2500 Nm3 h1,
p0,ox ¼ 10 bar, T0,ox ¼ 20  C, p1 ¼ p1 ¼ 1.013 bar, T 1 ¼ 1650  C. Reproduced with permission.[78] 2015, Steel Institute VDEh.

steel research int. 2018, 89, 1700098 1700098 (22 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

1. the Mach number Ms and/or velocity us, oxygen jet against excessive friction, so that it remains
2. the temperature Ts of the hot gas at the outlet of the coaxial concentrated over a longer distance. As a consequence, the
nozzle, and supersonic jet length increases from L /D1 ¼ 26.92 to L /
3. the ratio between the gap height Hs of the coaxial nozzle and Dp ¼ 35.23, that is, by 31%. The prototype of the advanced
the diameter Dp of the oxygen nozzle, Figure 25b. injector was successfully tested at a 125 t AC EAF and operated
well for a furnace campaign.[78]
It is interesting that L is not at the maximum length if the One conclusion of the study is that the supersonic jet length
velocities up and us of the primary and secondary jets are almost L cannot be determined in a satisfactory way with RANS and/
the same, but if the convective Mach numbers Mc are equal, that or URANS simulations. It was not possible to obtain
is, Mc1s ¼ Mcsp (1: furnace atmosphere, p: primary, s: satisfactory coincidence with literature data and experimental
secondary) cf.[78]. The convective Mach number describes the examinations at the Shock Wave Laboratory of RWTH Aachen
relative convection speed of large-scale flow structures and the University for 2D/3D approaches with domains of up to 8
corresponding speed of sound. Mc characterizes compressible million cells and different turbulence models (Standard k-e, k-
effects in mixing layers. New design principles for an injector ω, SST k-ω model). From the authors’ point of view, coincidence
is possible only with LES and DNS.
with a maximum supersonic jet length are defined on the basis
of this result.
Based on ref.[19] Odenthal et al.[77,78] carried out LES
simulations with ANSYS FLUENT (3D-LES, Smagorinky-Lilly 4.8. Scrap
model, species transport model, L ¼ 4 m, D ¼ 2.5 m, 8.202
million hexahedrons, Δt ¼ 2.5  106 s). The species transport There are a lot of different scrap types as regards the size, shape,
model is used to consider the gaseous mixing between oxygen chemical properties, tramp elements, and weight. Typical scrap
and CO.[16] Pressure boundary conditions are applied to the types are, for example, heavy demolition scrap, shear scrap, sheet
computational domain at the inlets and ambient outlets. metal wastes, steel, and cast scrap or solid crude iron. Table 1
Second-order upwind schemes are used to discretize the lists some scrap types with typical bulk densities and specific
governing equations, that is, pressure, density, momentum, surfaces. The specific surface varies by factor 40, while the
species, and energy. The time-step size amounts to Δt ¼ 2.5 spectrum differs in the same scrap type by factor 4. This list
 106 s, for an advanced injector. Figure 27 shows the result already shows how difficult it is to define reliable data for a
for the standard injector with single nozzle and flat wall, mathematical preheating or meltdown model.
Figure 27a, and the advanced injector, Figure 27b. Both Different scenarios with different numbers of phase and heat
injectors were designed for V_ 0;ox ¼ 2500 Nm3 h1, p0,ox ¼ 10 transfer mechanisms involved are conceivable for scrap melting.
bar, T0,ox ¼ 20  C, p1 ¼ p1 ¼ 1.013 bar, T1 ¼ 1650  C. The One useful categorization is the way how the thermal energy gets
standard nozzle was designed by the isentropic theory, the to the location of the melting process; this may be directly or
advanced coaxial nozzle by the software CARD and COAX, indirectly.
cf.[71,78]; the contour of primary and secondary nozzle is For indirect scrap melting, the thermal energy is conveyed by
designed, so that the optimization criterion (Mc1s ¼ Mcsp) is the superheated melt. Heat and scrap contact each other and the
fulfilled. The species transport model is used to consider energy required for phase transformation is conveyed to the
gaseous mixing between oxygen and CO.[16] Pressure phase boundary by thermal conduction of the melt. In addition
boundary conditions were applied to the computational to the temperature of melt and scrap, the carbon content of the
domain, second-order upwind schemes used to discretize charged materials also plays a role as it influences the melting
the equations, that is, pressure, density, momentum, species, temperature of the latter. As Szekely et al.[79] describe, the scrap
and energy. Figure 27 shows the instantaneous velocity will melt significantly faster if the carbon content in the melt is
distribution at a random point in time. The supersonic jet has higher than that in the scrap. Li and Brooks state in refs.[80,81]
a high frequency, stochastic structure, and consists of vortices that the melt freezes on the cold scrap in areas with low flow
at different size. Hot CO is entrained into the cold oxygen jet velocity, thus considerably prolonging melting.
via the free shear layer increasing the mass flow rate In direct scrap melting the arc and/or flame acts on the
transported downstream; hence, the velocity decreases. The location of melting and/or the scrap. Heat is transferred by
subsonic hot shrouding gas protects the supersonic cold convection and radiation provided that the molten material can

Table 1. Specific surface of individual types of scrap. Reproduced with permission.[11] 1997, Verlag Stahleisen.

Bulk density [t m3] Specific surface [m2 tswell1] Specific surface [m2 m3swell]
Heavy demolition scrap 1–1.2 10–30 10–36
Shear scrap 0.7–1 20–80 14–80
Sheet scrap 0.6–0.8 100–200 60–160
Steel and cast filings 0.7–1.2 200–400 140–480
Solid iron 3–4 6–8 18–32

steel research int. 2018, 89, 1700098 1700098 (23 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 28. Cross sections of multibar samples after immersion for approx. 10 s. Initial temperature of samples T0 ¼ 25  C, initial liquid bath temperature
Tm ¼ 1650  C. Reproduced with permission.[80] 2008, Springer US.

discharge. When melting scrap with gas burners, oxidation of does not calculate the melt flow; a heat transfer coefficient is
the scrap through CO2 has to be considered.[82] Finally, it is also assumed to consider the influence of the flow at the phase
possible to heat scrap by Joule heating, that is, by thermal energy boundary. Buoyant force effects are also neglected. First, a
resulting from the voltage drop at the electric resistance of the solidified layer is produced on the rods during immersion. If
scrap pile and/or melt. the rods are closely located to each other, these layers will
A CFD model for melting down scrap with indirect transfer of coalesce, Figure 28. This effect influences the total melting
the thermal melting energy should describe the following time. The authors also examine the influence of rod
phenomena: temperature and oxide layer on the melting process. The oxide
layer does not influence melting. The scrap porosity is the
1. Phase transformation solid–liquid and liquid–solid. critical factor for the melting time.
2. Carbon balance in the liquid and solid state. Arzpeyma et al.[85] use ANSYS FLUENT to predict the
3. Electrical-resistance heating. transient meltdown behavior of an individual cylinder in the
melt of a 150 t AC EAF. The phase change is modeled by a
In addition, a CFD model for melting down scrap with direct porosity approach, which considers the solid and liquid phases
transfer of the thermal melting energy should describe the together as a continuum, and the fraction of the solid phase is
additional phenomena: introduced as a scalar quantity. The scrap is modeled in the last
minutes of melting down as a cylinder with T ¼ 300 K next to the
1. Chemical reactions (e.g., decarburization, dephosphoriza- EBT tap-hole. The grid around the cylinder is refined to
tion) in the liquid, solid, and gaseous phases, and at the scrap minimize numerical diffusion. The melt is stirred electromag-
surface. netically to reduce the melting time, Figure 29. The induced
2. Heat balance (conduction, convection, radiation) in the scrap forces are calculated by an external software program and
pile. considered as source terms in the momentum equation.
3. Movement and/or sliding-down of the scrap pile. Electromagnetic stirring significantly increases the melt velocity
and intensified mixing resulting from this in turn leads to faster
Indirect transfer of melting heat: As early as in 1972, Szekely melting of the cylinder.
et al.[79] investigated the meltdown behavior of individual metal Direct transfer of melting heat: The study by Guo and Irons[86]
rods in a hot steel melt. When comparing the one-dimensional can be considered as the early stage of a simulation. It was
thermal conduction calculation including moving wall and the attempted at determining the movement of the scrap pile in a
measurement results, it can be shown that the influence of the container of 0.3 0.3 0.6 m3 as a model of an EAF. To do this, a
temperature and carbon content of the melt on the scrap melting 2-phase VoF model was used. The momentum equation of the
rate can be determined correctly. NS equations is replaced by the momentum equation for a
Li and Provatas,[80] Li and Brooks,[81] and Li et al.[83] examine flowing granular material. This approach is well suitable for
the meltdown behavior of rods using the phase field model.[84] modeling the monitored behavior of furnace scrap piles.
Li and Provatas[80] simulate melting of randomly distributed Guo et al.[82] examine melting down of scrap using a gas flame
scrap pieces to determine the effect of the bulk porosity on the and an arc. The scrap is considered as porous medium with
melting process. Good coincidence with experimental immer- constant porosity. The thermal energy introduced by the arc into
sion trials could be obtained. The phase field model, however, the scrap is predefined geometrically. The convective heat

steel research int. 2018, 89, 1700098 1700098 (24 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 29. 150 t AC EAF with bottom stirring: Comparison between velocity distribution of a) natural convection, b) forward stirring, c) backward stirring.
Reproduced with permission.[85] 2013, The Iron and Steel Institute of Japan.

transfer and the thermal energy introduced by oxidation of the


scrap surface by CO2 are also modeled. The exchange of
radiation with the furnace shell, the movements of melt and
scrap, and freezing of the melt on the scrap is not considered.
The EAF off-gas can be used to preheat the steel scrap, for
example, in shaft furnaces. This will reduce the energy required
for melting down the scrap.
Mandal[87] examines transient heating of a bulk pile of
irregularly shaped scrap pieces by means of a self-developed
approach. He models the scrap using an adapted porosity model
and the gas flow using a modified k-e model. The thermal
conductivity of the scrap pile, the heat transfer coefficient
between gas and scrap, and the radiation in the scrap pile are also
required. Experiments at a scrap pile indicate that the porosity
from the wall to the center of the scrap pile decreases
exponentially. Figure 30 shows an exemplary calculated
temperature distribution in the scrap pile.
Giavani et al.[88] investigate scrap preheating for the Consteel
Figure 30. Temperature contour of the furnace after 90 min. of the process. ANSYS FLUENT is used for the transient calculation of
experiment with small shredded scrap with top cover on, burner the scrap in the first step as a porous motionless pile using the
P ¼ 12.8 kW. Reproduced with permission.[87] 2010, McMaster University, Darcy-Brinkman-Forschheimer model. By analogy with Man-
Hamilton. dal[87] this approach considers the effect of a burner flame in the

steel research int. 2018, 89, 1700098 1700098 (25 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Investigations on the melting behavior of DRI pellets have


been done, cf.[45,51,90–94] So called mathematical simulators have
been developed which are able to predict the composition of
slags and the reaction kinetics of melting particles. The
simulators consider the change of slag properties with time
and several process operating aspects by using different input
types of the raw materials.
González et al.[93] use PHOENICS to simulate the melting
behavior of DRI pellets in the liquid sump of a 220 t AC EAF. The
model links three sub-models; a CAM model that takes into
account different arc lengths (Pel ¼ 90 MW, Larc ¼ 0.25 m;
Pel ¼ 120 MW, Larc ¼ 0.45 m), a bath model to simulate the
steady-state melt flow driven by natural convection, and a pellet
melting model. The latter considers the DRI pellets as
Lagrangian particles with uniform temperature tracked during
their flow through the melt. The particles are calculated in a post-
processing manner. Drag forces and energy required for the
melting process are not coupled back to the fluid regime. During
the initial contact between pellet and melt a frozen shell builds
up on the pellet which increases the diameter about 41%. The
thickness of the frozen shell is directly related to the pellet size
and does not dependent on the heat input. The melting rate of
the pellets is analyzed depending on the electric power, the
Figure 31. a) 600 kW burner on a moving bed of shredded scrap, b) Result
particle size, the particle temperature, and the particle feeding
of CFD simulation, c) Simulated heating curve of solid and shredded
(porous) scrap. Reproduced with permission.[88] 2012, South East Asia
position. A result of the study is that the melting time is
Iron and Steel Institute (SEAISI). decreased by using small pellet diameters, long arcs, hot
charging, and a feeding position as close as possible to the
hottest furnace region. González et al.[45] added the pellet
scrap interstitial volumes. The combustion of natural gas and the porosity to the parameters studied so far. Recently, Ramírez-
heat transfer by conduction, convection, and radiation are Argáez[51] extended the models to form a two-phase model with
considered. In order to account for the scrap movement, the liquid steel and slag layer.
scrap pile is in the second step replaced by a groove-shape, Martínez et al.[94] investigate the influence of the properties of
uniformly moved domain, Figure 31. The groove dimension is non-reactive slags, pellets, and melt bath conditions on the
selected in such way that both the penetration depth of the flame melting kinetics and the energy consumption. Using a one-
and the porosity of the domain correspond to the preceding dimensional finite-difference temperature model for the
transient calculation. Hence, the heating of an equally moved spherical pellet, the freezing, and melting process of the pellet
scrap pile is modeled in a useful way. is described in detail. Because the real melt and slag movement
To sum up: The processes for indirect scrap melting can be induced by oxygen injection or CO formation is neglected in the
calculated well by CFD on the precondition that the scrap is not model, the forced convection between melt/slag bath and pellet
moving in the melt. Far more complex are the processes during is approximated by a heat transfer coefficient. The model is
direct melting, and consequently only the heat transfer to the validated by experimental data, recommendations for energy
scrap pile had been simulated until now.[87] If it is intended to savings are listed.
consider the arc and the scrap movement, empirical approaches Although individual models are available for the melting of
will have to be applied.[82] Similar statements can also be made DRI pellets, these very special thermo-kinetic models have not
for the scrap preheating process. yet found their way into commercial CFD solver. Thus, there is
still a need to rely on the integration of such models into self-
developed and open source solver, respectively.

4.9. DRI
4.10. Off-Gas Elbow, Post-Combustion
There is the trend to reduce the share of scrap in the furnace and
to replace it by sponge iron. In standard furnaces up to 30% of The dust-laden exhaust gas is extracted through the off-gas elbow
DRI are charged, in special furnaces up to 100%.[89] Typical DRI into the post-combustion chamber and routed to the secondary
pellets have sizes of D ¼ 4–20 mm and a bulk density of dedusting equipment. The walls of the off-gas elbow are water-
ρ ¼ 1600–1900 kg m3 and/or an appeared density of ρ ¼ cooled to avoid overheating. The elbow gap (4th roof hole) is located
3400–3600 kg m3; the density is small compared to solid iron between off-gas elbow and post- combustion chamber. If the
due to the porous structure of the pellets. The melting rate of the extraction capacity is sufficiently high, a slight low-pressure is
pellets controls the tap-to-tap time, therefore, the knowledge of produced here, secondary air is entrained and mixed with the off-
this process value is of great interest. gas. The reactions referred to as post-combustion take place

steel research int. 2018, 89, 1700098 1700098 (26 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

(relative to environment). ANSYS FLUENT (3D-RANS, Standard


k-e model, 1-step reaction mechanism for CO, H2, FRM/EDM,
DO model, 0.340 million hexahedrons, ρ, η, λ ¼ f (p, T)) is used
for the simulation. Figure 32a shows the velocity and
temperature distribution for s ¼ 0.15 m and p ¼ 680 Pa. The
mean flow velocity in the off-gas elbow is approx. 30 m s1.
Ambient air is entrained, mixes with the primary gas, and
produces a hot off-gas strand which may cause an undesired
thermal wall loading. A recirculation area can be found on the
post-combustion chamber side, that is, characterized by cool and
slow secondary air. The flow regions inactive for post-
combustion could be avoided by adequate deflection equipment.
The maximum temperature in the hot gas strand determined to
be about 2500  C is relatively high. At the outlet of the post-
combustion chamber average amounts of 0.64% CO and 5.68%
CO2 are still present. Further findings are that it could be useful
to install guide vanes at the elbow for critical operating
conditions and that primary gas escapes from the 4th roof hole
into the steel plant if the low-pressure in the post-combustion
chamber is not high enough.
The studies of Gruber[23,24] have been already discussed in the
section 4.1 Furnace shell, furnace roof. This CFD model considers
the gas flow in both the furnace and the off-gas elbow, and the
heat fluxes of electrode, arc, melt, and slag. The amount of
secondary air is predefined, gas radiation has been considered.
The gas reactions are modeled by single-step models and the
FRM/EDM model, cf. Figure 6.
Figure 32. 180 t AC ConArc EAF: 3D-CFD-Simulation of post-combustion in
The study of Pfeifer et al.[31] has also been discussed in the
off-gas elbow and combustion chamber, pressure combustion chamber p ¼
–680 Pa, gap width s ¼ 0.150 m. a) Velocity and temperature distribution. b)
section 4.1 Furnace shell, furnace roof. The authors examine the
CO distribution. Reproduced with permission.[15] 2017, SMS group GmbH. chemical reactions in a 140 t AC EAF (D ¼ 6.9 m). For
simplification, the computational domain of the off-gas elbow
is separated from the EAF and solved separately. Gas and
downstream in the pipe routing and determine temperature and secondary air volume flow rates are pre-defined, the pressure-
composition of the off-gas. It is important that the low-pressure at dependent inflow at the 4th roof hole is permitted. CH4 is
the 4th roof hole is high enough to prevent off-gas from escaping added in the domain to consider the gas burners. ANSYS
into the melting shop. The furnace off-gas temperature may even FLUENT (3D-RANS, Realizable k-e model, FRM/EDM, 4-step-
reach 1750  C. When integrating the off-gas volume over a charge reaction mechanism) is used for the simulation, the thermal
and referring this value to a ton of produced steel, the specific off- NOx formation is calculated as well. The boundary conditions
gas volume is in the range of approx. 0–11 m3 t1 CO, 16–27 m3 t1 are taken from a measuring campaign. The objective of the
CO2, 8–13 m3 t1 O2, 95–98 m3 t1 N2, 0.05 m3 t1 CH4, 0.5 m3 t1 study is the determination of CO and NOx at the elbow outlet as
H2, as well as traces of H2O. The off-gas contains about 17 kg t1 of a function of the furnace parameters. The authors conduct
dust (Fe2O3, SiO2, CaO, MnO, ZnO etc.) per ton of produced steel; seven calculations with different conditions for the secondary
this corresponds to a 7% mass fraction of dust in the off-gas stream. air openings in the furnace. That way, the influence of the
For such mass fractions, however, the dust particles should be secondary air openings on the flow structure and the gas
introduced in the NS equations. In addition to internal gas composition in the EAF can be shown. However, when
radiation, radiation exchange between walls, also the particle comparing simulation and measurement, differences are to
radiation is involved, and therefore the particle radiation should be noticed in the gas composition. The authors explain this with
also be considered in the RTE equation. differences between the real and calculated flow fields, as well
In ref.,[15] the off-gas elbow of a 180 t AC ConArc EAF as uncertain CFD boundary conditions.
(145 MVA) was examined. The objective was to calculate the Labiscsak et al.[95] simulate a water-cooled post-combustion
amount of secondary air as a function of the elbow offset, the gap chamber (V ¼ 390 m3, S ¼ 430 m2) using an unknown CFD
width and the static pressure in the post-combustion chamber tool (3D-RANS, Standard k-e model, 1-step reaction mecha-
including the reaction kinetics. On the side of the furnace, the nism, EDC model, DO model, 0.5 million cells). The boundary
elbow diameter is D ¼ 2.20 m, on the side of the combustion conditions are derived from the measured data. Post-
chamber D ¼ 3.32 m, the gap width s varies from s ¼ combustion is simulated for three gap widths (s ¼ 0.2, 0.3,
0.05–0.15 m. The primary gas flow rate is assumed to be 0.4 m). The authors demonstrate that the off-gas and
V_ ¼ 56 000 Nm3 h1, T ¼ 1900  C, 18% CO, 20% CO2, 3.45% O2, secondary air is mixed efficiently, and that the CO can be
58.5% N2, 0.05% H2. The static pressure p in the post- reduced from 30% (34 300 ppm) to 0.0065% (64 ppm) by
combustion chamber varies between p ¼ 80 and 1000 Pa increasing the gap width.

steel research int. 2018, 89, 1700098 1700098 (27 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 33. 140 t AC EAF: Temperature distribution of the off-gases in the section from voltage joint to segment III. Reproduced with permission.[96]
2007, Instituto Argentino de Siderurgia.

Karbowniczek and Kawalkowski[96] simulate the post-com- desirable to disconnect the computational domains of post-
bustion for a 140 t AC EAF to predicting the gas composition and combustion chamber and furnace. The realization is difficult
NOx portion at the outlet of the off-gas duct. ANSYS FLUENT because mass flow rate, pressure, velocity, temperature,
(3D-RANS, Standard k-e model, FRM/EDM, 1.4 million cells) is composition, and dust content at the inlet of the off-gas elbow
used for the simulation. The pressure boundary conditions are have to be known. It is hardly possible to get reliable boundary
adjusted for two operating states in such a way that the measured conditions, especially due to the entrainment of secondary air at
volumetric flow rate is reproduced. At the inlet to the the slag door, furnace gap, electrode port, and elbow gap. The
computational domain, T ¼ 1893 K applies for the off-gas, low-pressure in the furnace, and/or the secondary air amount is,
operating point 1: 19% CO, 19.3% CO2, 0% O2, operating point amongst others, depending on the exhaust fan power, gas
2: 3.2% CO, 1.1% CO2, 17.7% O2. The numerical results in temperature, and gas amount induced by melt and slag
Figure 33 matches the measured temperatures in the duct. reactions. Along with the metallurgical processes widely varying
Tang et al.[97] calculates post-combustion in the horizontal in the course of the operating cycle, it must be concluded that
exhaust duct of a 145 t AC EAF (135 MVA) for stainless steel CFD boundary conditions cannot be specified a priori for the
grades, and in the chamber of a 30 t AC EAF (15 MVA) both EAF; calculations are rather depending on empirical data and
without and with supply of energy. ANSYS FLUENT (3D-RANS, measured values. It is for this reason why gas amount and
Standard k-e model, 1-step reaction mechanism for CH4, H2, composition at the inlet of the off-gas elbow are in most cases
CO, FRM/EDM, DO model, 0.428 million cells) is used as solver. specified empirically. None of the publications presented here
Among others, the author conducts sensitivity analyses on the considered the dust in the mass, energy, and radiation balance.
mesh size. In general, the findings from plant trials are Frequently, an irreversible single-step reaction model is used for
reproduced well by all simulations. The maximum off-gas reaction kinetics in the form of CO þ 1/2 O2 ! CO2, and the
temperature in the horizontal exhaust duct, for example, is about calculations are in most cases carried out steady-state. All
2100 K. The temperature in the post-combustion chamber is authors applied the DO model to consider radiation effects in the
approx. 2080 K, the post-combustion degree is 99.5%. energy equation because it considers shadowing effects in
Arzpeyma et al.[98] simulate the off-gas elbow of an AC EAF complex geometries and internal gas radiation.
using ANSYS FLUENT (3D-RANS, Standard k-e model, 1-step
reaction mechanism, FRM/EDM). At the elbow inlet, they
assume T ¼ 1800 K, 25% CO, 15% CO2, 17.7% O2. The off-gas
5. Holistic EAF CFD Simulation
composition is determined downstream for various operating
conditions by modifying the low-pressure in the elbow gap in the This chapter discusses aspects of a holistic EAF CFD simulation,
range from 200 to 50 Pa. Objective was to calculate the recently published by Odenthal et al.[99,100] The subjects of the
furnace gas composition before addition of secondary air based studies are a 120 t AC EAF and a 140 t DC EAF, and some of the
on a CO measurement downstream of the elbow gap, that is, it operating data are shown in Table 2.
was intended to determine a look-up table for the off-gas. The 120 t AC EAF (150 MVA) features PCA aluminum electrode
The determination of off-gas temperature and composition is supporting arms, electrodes (L  5 m, D ¼ 0.71 m), four burner/
often a purpose of a CFD simulation for the off-gas elbow. It is injector systems, a door burner/injector, a lance manipulator with

steel research int. 2018, 89, 1700098 1700098 (28 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

self-wearing oxygen lance and carbon lance, three carbon injectors preheating shaft by 34 water-cooled fingers. The fingers can be
in the furnace wall, and a slide gate tapping system. The overall moved back and forth either together or individually, so that the
design is close to the equipment at a German steel-plant.[78] The hot scrap is distributed homogeneously in the lower shell. The
furnace is equipped with special injectors which assist with scrap hot off-gas is removed from the furnace at both sides directly
melting and melt superheating. Each injector is a combined burner- below the charge openings. The injectors (3 devices: V_ ox ¼ 3320
injector system of the coaxial jet type, which can be operated either in Nm3 h1, p0,ox ¼ 11 bar (IM); P ¼ 6 MW (BM); 1 EBT device:
burner mode (BM) or injector mode (IM), cf.[71,78] In burner mode, V_ ox ¼ 2210 Nm3 h1, p0,ox ¼ 10 bar (IM); P ¼ 4 MW (BM)) are
oxygen and fuel gas mix outside the injector and ignite inside the hot installed in the copper boxes. In addition, two burners are fitted
environment. The oxygen flows through the central supersonic in the furnace, and these are operated in burner mode only. The
nozzle, the fuel gas flows through the coaxial subsonic nozzle. In furnace is tilted for deslagging and tapping purposes. The power-
injector mode, the oxygen also flows through the central nozzle. on time is around 30 min, and the duration of the flat-bath phase
Simultaneously, the oxygen jet is shrouded by a hot off-gas jet which is approx. 15 min.
exits the coaxial nozzle, cf. section Gas burner, injector. In addition to For the flat-bath sequence of both furnaces, the behavior of
the injectors (3 devices: V_ ox ¼ 2500 Nm3 h1, p0,ox ¼ 10 bar (IM); melt, slag, and gas is investigated with high spatial and temporal
P ¼ 6 MW (BM)), which are housed in copper boxes, a door injector resolution. The authors list some questions, which can be
(1 device: V_ ox ¼ 3000 Nm3 h1, p0,ox ¼ 10 bar (IM)) above the slag answered by the simulation, for example, the optimum
door is used. Carbon injectors are not included in the study. arrangement of the injectors with regard to installation height
The 140 t DC EAF (140 MVA) is a shaft-type furnace with two and installation angle. The oxygen jets must not aim at the
preheating shafts, in which the charging material is heated to electrode and hot melt must not splash against electrode or wall.
700–800  C, Figure 34a. As the scrap is preheated, the operation At the same time the oxygen jets should not penetrate too deeply
of the shaft-type furnace is both cost-effective and eco-friendly. In into the melt bath as otherwise the bottom shell would be
addition to scrap a large proportion of DRI or HBI can be used in damaged. Furthermore, the air entrainment through slag door,
order to meet the relevant purity requirements of the steel being electrode port, furnace gap, 4th roof hole, the gas composition,
produced. The furnace consists of an EBT lower shell, a water- the degree of post-combustion, the wall temperature (hot spots),
cooled upper shell, and two vertical preheating shafts of identical and the formation of hot strands can be simulated.
design. The electrode (L  5 m, D ¼ 0.70 m) moves between the Figure 34 shows the pre-processing procedure on the example
shafts. From the electrode, the electric current flows over the of the DC EAF. Starting from a CAD model, Figure 34a, the fluid
charging material into the pin-type bottom electrode. The scrap domains are simplified, Figure 34b, and approximated by a
is charged from above and held in the bottom section of the tetrahedron mesh with 17.391 million cells, Figure 34c. The

Table 2. Original EAF equipment and operating data; some volume flow rates have been modified in CFD study as compared to original EAF
equipment.[99]

120 t AC EAF 140 t DC EAF

Production portfolio Beam steel and support steel Rail steel


Furnace Type EBT/125 t EBT/140 t
Transformers 150 MVA 140 MVA
Air-cooled reactor 15 MVar –
Furnace diameter 7.7/7.9 m 6.9/7.1
Tapping system Slide gate Slide gate
Roof swinging system Roller bearing with cylinder drive –
Electrode supporting arms PCA aluminum PCA aluminum
Charging system Scrap bucket with single or twin bucket charging Scrap bucket with 4 bucket charging
Regulation system Ferrotron DECTEQ DC regulation
1
Burner/injector 3 devices: Burner P ¼ 6 MW Injector V_ ox ¼ 3000 Nm h
a) 3
3 devices: Burner P ¼ 6 MW Injector V_ ox ¼ 3320 Nm3 h1
p0,ox ¼ 10 bar p0,ox ¼ 11 bar
EBT burner/injector _ ox ¼ 3000 Nm3 h1
1 EBT deviceb): Burner P ¼ 6 MW Injector V _ ox ¼ 2210 Nm3 h1
1 EBT device Burner P ¼ 6 MW Injectord) V
p0,ox ¼ 10 bar p0,ox ¼ 10 bar
Door burner/injector/lance Door burnerc), lance manipulator (oxygen, carbon) Door þ EAF burner: P ¼ 4 MW
Carbon injector 3 tubes through wall panels 2 tubes through wall panels

For CFD study: a)


V_ ox ¼ 2500 Nm3 h1, p0,ox ¼ 10 bar; b)
Not used; c)
Used as subsonic door injector with V_ ox ¼ 3000 Nm3 h1, p0,ox ¼ 10 bar; d)
V_ ox ¼ 3320 Nm3 h1,
p0,ox ¼ 11 bar.

steel research int. 2018, 89, 1700098 1700098 (29 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

mesh is locally refined in areas of high velocity and temperature to the high speed (M > 0.3). If, for example, the oxygen injector is
gradients, for example, along the jet axes and in the arc region. seen as having V_ ox ¼ 2500 Nm3 h1, p0,ox ¼ 10 bar, p1 ¼ 1.1013
Since transient simulations with large grids are time-consum- bar, then, according to the isentropic theory, the following will
ing, the tetrahedron mesh is converted to a polyhedron mesh apply at the exit of the Laval nozzle: M1 ¼ 2.15, p1 ¼ 1.1013 bar,
with 3.682 million cells. u1 ¼ 505.8 m s1, T1 ¼ 120.2  C (152.9 K). Since the incom-
The simulation is carried out with ANSYS FLUENT based on pressible melt and slag phases cannot be adequately coupled
the URANS solver with the SST-SAS model and enhanced near- with the compressible oxygen phase in the VoF model, the
wall treatment.[18] With both furnaces, the arc is regarded as a authors describe a numerical variant.
cylinder with a surface temperature of Tarc ¼ 5500 K. In the case The phenomenology of the compressible, cold oxygen jet,
of the AC EAF, the arcs (Larc ¼ 0.22 m, Darc ¼ 0.040 m) are shrouded by a hot off-gas jet which enters the hot CO furnace
inclined outwards by 30 , and are fully immersed in the slag (T1 ¼ 1650  C) is determined by means of a preceding, separate
(100% slag-enveloped). With the DC EAF, the arc (Larc ¼ 0.60 m, LES simulation, cf. section 4.7 Gas burner, injector. The
Darc ¼ 0.055 m) is aligned vertically on the melt surface, and the momentum force Iox ¼ ρox  uox 2 A along the oxygen jet axis is
top part of the arc radiates heat (67% slag-enveloped). The time-averaged. The key piece of data is the momentum value at
electrodes are subjected to a linear temperature drop (2773 K the transition region between jet and melt. The value from the
< Twall < 573 K). Slag door, cover gap, and electrode ports are LES simulation of the coaxial jet is considered in the URANS
open. The time-dependent behavior of melt, slag, and oxygen is simulation of the entire furnace in a way that the oxygen density
calculated with the implicit VoF approach (Δt ¼ 5  104 s).[20] The is set constant and selected so that the momentum near the melt
heat radiated from the arcs onto the furnace wall is realized by surface is approximately in the same order for the compressible
the DO model (asl ¼ 0 m1, σsl ¼ 0.5 m1, nsl ¼ 1). Pressure (LES) and the incompressible (URANS) simulation. A source of
boundary conditions are applied to the computational domain at error which has to be accepted is that the incompressible jet
inlets and outlets. The low-pressure in the off-gas duct is spreads due to entrainment effects more than the compressible
specified in such a way that the design-point of the off-gas flow jet.
rate is roughly attained. Second-order upwind schemes are used Figure 35a shows the instantaneous velocity distribution in
to discretize the governing equations, that is, flow, turbulent the entire furnace. It can be seen that the pressure at the off-gas
kinetic energy, and dissipation rate. The cases are executed on a elbow is too high, thus causing furnace gas to flow outside
Linux cluster with 64 parallel processes. through the slag door. The oxygen jets are marked as blue iso-
However, while melt and slag are to be seen as incompress- surfaces, displacing the light slag (red), and penetrate into the
ible, the oxygen would have to be regarded as compressible due melt (yellow) at different depths. The typical penetration depth

Figure 34. 140 t DC EAF: a) CAD model, b) Geometry simplified with ANSYS ICEM-CFD 17.0. c) Mesh consisting of 17.391 million tetrahedrons later
transferred into 3.682 million polyhedrons. Reproduced with permission.[99] 2017, SMS group GmbH.

steel research int. 2018, 89, 1700098 1700098 (30 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 35. 120 t AC EAF, URANS simulation: Three injectors, Vox _ ¼ 2500 Nm3 h1, p0,ox ¼ 10 bar, 1 door burner/injector, V_ ox ¼ 3000 Nm3 h1,
p0,ox ¼ 10 bar. Instantaneous velocity and temperature distribution at t ¼ 14.080 s. a) Velocity in melt domain (umelt), gas domain (ugas), electrode wall
temperature (Twall). Velocity iso-surface, melt, slag. b) Velocity vectors in melt level (hidden slag layer). c) Hot spot distribution at inner furnace
wall. Reproduced with permission.[99,100] 2017, SMS group GmbH.

of an oxygen jet is 0.1–0.2 m which is in good agreement with heat toward the furnace wall and induce hot spots near to the
the findings by Koria and Lange.[59] The jets induce a slight slag door with Tmax  1580  C, thus probably causing
agitation in the melt bath and at the melt surface, Figure 35b. overheating of the refractory lining and cooling panels,
Splashing of the melt against the electrodes does not occur. The Figure 35c. The CFD model is able to detect time-variable
injectors are arranged in such way that the oxygen jets aim hot spots and where the local wall temperature Twall in
between the electrodes and high velocities below the injector conjunction with the wall shear stresses τwall,n, τwall,t may cause
box are avoided. The slag layer is penetrated by the jets and premature wear. The authors suggest to define a new
partly displaced from the arcs. Hence, the arcs temporarily emit number for classification, the Numerical Refractory Wear

steel research int. 2018, 89, 1700098 1700098 (31 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Figure 36. 140 t DC EAF, URANS simulation: 3 injectors, 1 EBT injector, ¼ , V_ ox ¼ 3320 3320 Nm3/h1, p0,ox ¼ 11 bar. Instantaneous velocity and
temperature distribution at t ¼ 7.278 s. a) Velocity in melt domain (umelt), electrode wall temperature (Twall). Velocity iso-surface, melt, slag.
Reproduced with permission.[99] 2017, SMS group GmbH.

Index RWInum: For the DC EAF, the model has been enhanced by the
pressure loss Δploss in the scrap pile. The scrap pile is considered
RWI num ¼ a1 τ wall;n þ a2 τ wall;t þ a3 T wall þ ::: ð6Þ as a porous medium, and the flow resistance of the hot furnace
gas is determined empirically. The porous media model includes
Here, ai represents weighting factors still to be defined, an additional sink term in the momentum equation and
τwall,n is the normal stress, and τwall,t the tangential stress calculates the superficial velocity in the scrap pile as a function of
which is caused by the respective fluid on the wall surface the gas flow volume and the porosity of the scrap pile. The heat
element under review. Twall is the wall temperature, that is, the transfer in the scrap pile is taken into account, hence a heat sink is
temperature in the wall cell. While RE (Eq. 3) and RWI (Eq. 4) defined in this domain. Assuming the furnace shafts are 80% filled
provide only integral information on the refractory wear, with a mixture of scrap, pig iron, and HBI, (ρmix ¼ 0.7 t m3),
RWInum would better illustrate the spatial and temporal then there are approx. 81 t of material in the shafts. The porosity Ψ,
varying wall loading. that is, the void to total volume ratio, is Ψ ¼ 0.913. This value and a

Figure 37. 140 t DC EAF, RANS simulation: Three injectors, one EBT injector, V_ ox ¼ 3320 Nm3 h1, p0,ox ¼ 11 bar. a) CO concentration. b) CO2
concentration. Velocity iso-surface. Reproduced with permission.[99,100] 2017, SMS group GmbH.

steel research int. 2018, 89, 1700098 1700098 (32 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

series of other variables are used to determine Δploss in the scrap While the proprietary CFD tools frequently used already come
pile as a function of the gas flow rate. with a variety of submodels, some of these submodels are
By analogy with Figure 35, Figure 36 shows the results for the currently being incorporated into open source tools. Even today,
DC EAF. The oxygen jets displace the slag and penetrate, however, CFD is often carried out on the basis of statistical
depending on the installation height, up to 0.1–0.2 m into the turbulence modeling (RANS/URANS þ two-equation turbu-
melt, thus inducing velocities of about 0.1 m s1. However, the lence models etc.). With highly unsteady or anisotropic flows,
movement of the melt is far more intensive in reality as a result or if the scope of application of the statistical turbulence
of the electrical current flow which is neglected. modeling is exceeded, this procedure may result in errors that
If the focus is on the post-combustion, the authors point out has initiated, at least within a university context, a paradigm shift
that a transient, multi-phase URANS approach is no longer from statistical RANS/URANS to LES and DES methods; this
expedient and should be substituted by a steady- state, single- paradigm shift has not yet been achieved for the EAF simulation.
phase RANS approach. Additionally, the EDC model is used for Among other things, this paper describes the furnace
the post-combustion simulation taking into account multi-step simulation with holistic CFD approaches. The designation
reactions; 13 species (H2, H, O2, O, OH, H2O, HO2, H2O2, CO, holistic means that the simulation takes into account as many of
CO2, HCO, N2, Ar) and 35 reactions (Yetter algorithm) are the fluid dynamic and thermo-dynamic phenomena occurring
considered.[16] Figure 37a and b show the result of such a RANS inside the furnace as possible (e.g., discrete, dispersed, and
simulation. Assuming that during the flat-bath phase 0.93 kg s1 continuously interacting fluid phases; compressibility; turbu-
of carbon is blown into the furnace via carbon injectors for lence; phase change; species transport; chemical reaction;
15 min, and the carbon content in the pig iron is still 4%, and radiation; electromagnetic interaction; electric arc/plasma
0.5% in the scrap, the amount of CO formed is estimated to be formation, etc.). Such a holistic model can give answers
7.7 kg s1. This CO mass flow rate is added to the slag surface as regarding the movement of the phases, the penetration depth
a CO source. Openings in the furnace (slag door, roof gap, of the oxygen jets, and the thermal wall loading. However, there
electrode port) entrain ambient air which is combusted with CO are still limitations to the holistic simulation, and some of the
to form CO2. reasons for this are:
In the present case, 16.2 kg s1 of false air is taken in at the
slag door entering the furnace at a velocity of approx. 14.9 m s1. 1. Substantially simplified mathematical-numerical models or a
Consequently, the mass flow rates through both shafts are not lack thereof (e.g., models for foamy slag, scrap melting,
uniform, that is, the slag door should be closed during operation. radiation of particle-laden off-gas), which can only be coupled
The gas temperature in the left-hand shaft is higher than in the and parallelized to a certain degree.
right-hand shaft, with a mean temperature of approx. 1040  C. 2. Imprecise starting and boundary conditions (u, p, T, Tu,. . .)
Due to this, the CO and CO2 concentration in the shafts is not and fluid properties (ρ, η, cp, λ, e,...).
uniform, Figure 37a and b. 3. Variable furnace sequences (preheating, charging, melting,
The authors show that already the simulation of around 20 s of superheating, tapping) with spatially and temporarily
real furnace process time can offer important information on the changing process conditions, scale-transition flow phenom-
design and operation of modern furnaces. Mixing times, ena over ten to the power of six or eight (e.g., supersonic
however, cannot be determined. According to the authors, the oxygen jet (M ¼ 2)) with shock wave: Scale size Δx ¼ 106 m,
next step in EAF CFD modeling should be the coupling of the Δt ¼ 105 s; melt flow in the furnace: Scale size Δx ¼ 100 m,
multiphase and the compressible solver under consideration of Δt ¼ 103 s).
combustion models. 4. A lack of access to high-performance computers, thus
insufficient computational capacity with complex, coupled
submodels that are difficult to parallelize.
6. Summary 5. Missing data or operating experience data, which is used to
During the last decades a large number of CFD models have validate the CFD results, that cannot be accessed as a
been developed to increase the transparency of the EAF process. consequence of extreme process conditions; model tests are
Due to the complex phenomena with a wide range of the time only transferable to a certain degree.
and length scales it is, even today, a big challenge to simulate the
furnace process in detail. Against this background, scientific It has been shown, however, that the CFD simulation is
literature deals mostly with partial aspects of the furnace very efficient, if the respective model is specifically adapted
spectrum, that is, CFD is either carried out for individual to the question under review, and this also allows results
components (furnace shell, furnace cover, off-gas elbow, obtained within an acceptable period of time. In this case,
refractory lining, fluid phases, arc, electrode, electrode support- CFD is the first choice to simulate, predict and visualize the
ing arm, burner/injector, scrap bed, bottom tap, etc.) or for EAF process.
individual phenomena (post-combustion, melting, movement of
phases, etc.). With validated submodels and reliable boundary
conditions, CFD simulations for the furnace process are
Conflict of Interest
promising and beneficial, thus making the process more
transparent, comprehensible, and safe. The authors declare no conflict of interest.

steel research int. 2018, 89, 1700098 1700098 (33 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

Nomenclature L Lp , Supersonic jet length


α Angle, volume/void fraction LES Large Eddy Simulation
Δ Difference LPG Liquid Pressurized Gas
e Dissipation rate, Emissivity M Mach number
ζ Fine scales length _
m Mass flow rate
η Dynamic viscosity Mc1s Convective Mach no. between ambience and
θ Fluid flow angle, that is, angle between the local secondary nozzle
velocity vector and the system of coordinates Mcsp Convective Mach no. between secondary and
κ Specific heat ratio primary nozzle
λ Thermal conductivity coefficient MHD Magneto-hydrodynamics
ν Kinematic viscosity n Refractive index
ρ Fluid density, porosity NS Navier-Stokes
σ Electric conductivity, scattering coefficient OBM Oxygen Bottom Maxhütte
τ Shear stress p Pressure
τ Time scale P Power
τ99 Mixing time PCA Power Conductive Arms
ϕ Mass fraction, cylindrical coordinates Pr Prandtl number
ψ Porosity Q_ Radiation power
ω Specific dissipation rate R Electric resistance, reaction rate
a Absorption coefficient, speed of sound R0 Specific gas constant
A Area r Radial coordinate, distance between arc and wall
AC/DC Alternating Current/Direct Current RANS Reynolds-Averaged Navier-Stokes
AOD Argon Oxygen Decarburization RE Refractory Erosion Index
BM Burner mode Re Reynolds number
BOF Basic Oxygen Furnace RNG Re-Normalization Group
c Mass concentration RWI Refractory Wear Index
c, cþ Right/left- running characteristic line RWTH Rheinisch-Westfälische Technische Hochschule
cp Specific heat capacity (constant pressure) S Surface area
CAD Computer Aided Design s Gap width
CAM Channel Arc Model SHARC Shaft arc furnace
CD Convergent- divergent SAS Scale Adaptive Simulation
CFD Computational Fluid Dynamics SST Shear Stress Transport
D Diameter, thickness t Time
DES Detached Eddy Simulation T Temperature
DNS Direct Numerical Simulation Tu Turbulence intensity
DO Discrete Ordinates U Electric voltage
DPM Discrete Phase Model u, v Flow velocity in axial and radial direction
DRI Direct Reduced Iron UDF User Defined Function
E Heat flux density URANS Unsteady Reynolds-Averaged Navier-Stokes
EAF Electric Arc Furnace V Volume
EBT Eccentric Bottom Tapping V_ Volume flow rate
EDC Eddy Dissipation Concept VoF Volume of Fluid
EDM Eddy Dissipation Model WSGGM Weighted-sum-of-gray-gases model
EU European Union wt% Weight per cent
F Force x, y, z Coordinates
FEM Finite Elements Method
FNM Finite Network Model
FRM Finite Rate Model
H Height Indices
h Convective heat transfer coefficient
HBI Hot Briquetted Iron — Average and/or mean value
HPC High Performance Computing σ Electric conductivity
Hs/Dp Ratio of the secondary nozzle height Hs to primary 1 Furnace atmosphere, ambience

nozzle diameter Dp Critical value (nozzle throat)
I Electric current, momentum 0 Referred to standard fluid flow conditions (nozzle inlet)
i Electric current density 1 Outlet, (nozzle exit)
IM Injector mode air Air
k Turbulent kinetic energy Al Aluminum
L Length, length scale arc Electric arc

steel research int. 2018, 89, 1700098 1700098 (34 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

C Carbon [10] M. A. Ramírez-Argáez, C. González-Rivera, G. Trápaga, ISIJ Int.


c Convective 2009, 49, 796.
cc Combustion Chamber [11] K.- H. Heinen, Elektrostahlerzeugung, Verlag Stahleisen, Düsseldorf,
CS Crude steel Germany 1997.
[12] O. Widlund, U. Sand, O. Hjortstam, X. Zhang, 15th METEC &
CH4 Methane
InSteelCon, Düsseldorf, Germany 2011, 1.
CO CO
[13] V. Haverkamp, Prädiktive dynamische Lastkontrolle für ein Elektros-
Cu Copper tahlwerk, Dissertation Helmut-Schmidt- Universität/Universität der
d Design point Bundeswehr Hamburg 2009.
el Electric [14] O. J. P. Gonzáles, M. A. Ramírez-Argáez, A. N. Conejo, ISIJ Int. 2010,
g Gas 50, 1.
loss Loss [15] SMS group GmbH, Internal Information, 2017.
m Melt [16] ANSYS FLUENT 17.0 Users’ Guide, 2016.
max Maximum [17] J. H. Ferziger, M. Peric, Computational Methods for Fluid Dynamics,
n Normal Springer-Verlag, Berlin, Heidelberg 1999.
num Numerical [18] F. R. Menter, Y. Egorov, Int. Aerospace CFD Conf., Paris, France
2007, 1.
ox Oxygen
[19] I. Klioutchnikov, H. Olivier, H.-J. Odenthal, Comput Fluids 2013,
P Particle 86, 490.
P primary; primary (central) nozzle [20] C. W. Hirt, B. D. Nichols, J. Comput. Phys. 1981, 39, 201.
s secondary; secondary (shroud gas) nozzle [21] B. Dittmer, K. Krüger, Elektrowärme Int. 2009, 4, 195.
sc Scrap [22] W. E. Schwabe, Electric Furnace Conf. AIME 1962, 20, 195.
sl Slag [23] J. C. Gruber, Development of a Numerical Model for the Heat and
t Tangential Mass Transport in an Arc Furnace Freeboard, Dissertation, RWTH
tap Tapping Aachen University 2015.
tur Turbulent [24] J. C. Gruber, T. Echterhof, H. Pfeifer, Steel Res. Int. 2016, 87, 15.
w Water [25] H. Pfeifer, Energietechnische Untersuchung der Plasmatechnik bei der
Stahlerzeugung, Habilitation, Universität GH Siegen 1992.
wall Wall
[26] D. Guo, G. A. Irons, 3rd Int. Conf. on CFD in the Minerals and Process
x, y, z Components in direction of coordinates Industries, CSIRO, Melbourne, Australia 2003, 223.
[27] J. Alexis, M. A. Ramírez-Argáez, G. Trápaga, P. Jönsson, ISIJ Int.
2000, 40, 1089.
Keywords [28] A. Fallah, M. Meratian, H. Edris, B. Zamani, 9th European Electric
Steelmaking Conf., Krakow, Poland 2008, 1.
AC/DC electric arc furnace, bottom tapping, burner, chemical reaction, [29] M. N. Al-Harbi, S. K. Mishra, M. Maity, S. Al-Turky, 15th METEC &
coaxial jet, computational fluid dynamics (CFD), graphite electrode, InSteelCon, Düsseldorf, Germany 2011, 1.
multiphase flow, post-combustion, refractory lining, scrap melting, shaft- [30] J. L. G. Sanchez, A. N. Conejo, M. A. Ramírez-Argáez, ISIJ Int. 2012,
type furnace, simulation, supersonic flow, thermal loading, wall panel 52, 804.
[31] H. Pfeifer, T. Echterhof, L. Voj, J. Gruber, H.-P. Jung, S. Lenz,
Received: March 15, 2017 C. Beiler, F. Cirilli, U. De Miranda, N. Veneri, E. Bressan, EUR 25078
Revised: July 3, 2017 - Control of Nitrogen Oxide Emission at the Arc Furnace - CONOX,
Published online: November 28, 2017 European Commission, Luxembourg 2012, 92.

[32] V. Logar, D. Dovzan, I. Skrjanc, ISIJ Int. 2011, 51, 382.

[33] V. Logar, D. Dovzan, I. Skrjanc, ISIJ Int. 2012, 52, 402.

[34] V. Logar, D. Dovzan, I. Skrjanc, ISIJ Int. 2012, 52, 413.
[1] www.stahl-online. de, 25.01.2017. 
[35] V. Logar, I. Skrjanc, ISIJ Int. 2012, 52, 1225.
[2] International Energy Agency (IEA), World Energy Outlook 2014, IEA 
[36] A. Fathi, Y. Saboohi, I. Skrjanc, V. Logar, ISIJ Int. 2015, 55, 1353.
Publications, Paris, France 2014. 
[37] V. Logar, A. Fathi, I. Skrjanc, Steel Res. Int. 2016, 87, 330.
[3] B. Bowman, K. Krüger, Electric Arc Furnace Physics, Verlag
[38] M. Kirschen, C. Rahm, J. Jeitler, G. Hackl, 9th European Electric
Stahleisen, Düsseldorf, Germany 2009.
Steelmaking Conf., Krakow, Poland 2008, 365.
[4] T. Meier, Modellierung und Simulation des Elektrolichtbogenofens,
[39] A. G. Mombeni, E. Hajidavalloo, M. Behbahani-Nejad, Appl. Therm.
Dissertation, RWTH Aachen University, Verlag Mainz, Aachen,
Eng. 2016, 98, 80.
Germany 2016.
[40] M. Fabbro, Kosco-Kroman-Tokyo-MMK Projects, 27. Aachener
[5] Jahrbuch Stahl 2017, Verlag Stahleisen, Düsseldorf, Germany Stahlkolloquium, Aachen, Germany 2012, 141.
2017. [41] Danieli Service Center, water-cooled elements for steelmaking
[6] A. Buchholz, World Metal.  Erzmetall 2008, 61, 146. furnaces and equipment, http://web.danieli.it/dsmailbulletin/
[7] F. Stahl, M. Kühnemund, K. Krüger, M. Hergt, W. Hartmann, 15th DSMB_0101/download_files/Dan_WaterCooled_panels_ENp.
METEC & 2nd ESTAD Conf., Düsseldorf, Germany 2015, 1. pdf, 2017.
[8] M. Kirschen, Energieeffizienz und Emissionen der Lichtbogenöfen in [42] J. Borlee, M. Wauters, L. Zampetti, EUR 22404EN  New Cooling
der Stahlindustrie, Habilitation, RWTH Aachen University, Verlag Panels for Reduction of Heat Losses in EAF Steelmaking, European
Stahleisen, Düsseldorf, Germany 2006. Commission, Luxembourg 2006.
[9] H.- J. Eckstein, Stahlerzeugung, VEB Deutscher Verlag für [43] L. Baokuan, H. Jicheng, Proc. of the 14th PTD Conf., Nashville, USA
Grundstoffindustrie, Leipzig, Germany 1980. 1995, 181.

steel research int. 2018, 89, 1700098 1700098 (35 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com www.steel-research.de

[44] S. Chen, R. Zhu, J. Li, C. He, Adv. Mater. Res. 2012, 479–481, 1750. [75] M. Alam, J. Naser, G. Brooks, AISTech, Iron and Steel Technology
[45] O. J. P. González, M. A. Ramírez-Argáez, A. N. Conejo, ISIJ Int. 2010, Conf., Pittsburgh, USA 2010, 885.
50, 9. [76] M. Alam, J. Naser, G. Brooks, A. Fontana, Metall. Mater. Trans. B
[46] Y. I. C. Guzmán, M. A. Ramírez-Argáez, A. N. Conejo, 2010, 41B, 1354.
G. M. Trápaga, Advances in Theory of Ironmaking and Steelmaking, [77] H.-J. Odenthal, J. Bader, R. Nörthemann, M. Reifferscheid,
IISc, Bangalore, India 2009, 341. I. Klioutchnikov, H. Olivier, AISTech, Iron and Steel Technology
[47] F. Liu, R. Zhu, K. Dong, X. Bao, S. Fan, ISIJ Int. 2015, 55, 2365. Conf., Indianapolis, USA 2014, 1.
[48] H. Matsuura, R. J. Fruehan, ISIJ Int. 2009, 49, 1530. [78] H.-J. Odenthal, S. Buess, P. Starke, R. Nörthemann, M. Lohmeier,
[49] E. B. Pretorius, R. C. Carlisle, 56th Electric Furnace Conf. ISS-15-18, 17th METEC & 2nd ESTAD  European Steel Technology and
New Orleans, USA 1998, 275. Application Days, Düsseldorf, Germany 2015, 1.
[50] J. L. G. Sanchez, M. A. Ramírez-Argáez, A. N. Conejo, Steel Res. Int. [79] J. Szekely, Y. K. Chuang, J. W. Hlimka, Metall. Trans. 1972, 3, 2825.
2009, 80, 113. [80] J. Li, G. Brooks, N. Provatas, Metall. Mater. Trans. B 2008,
[51] M. A. Ramírez-Argáez, A. N. Conejo, M. S. L opez-Cornejo, ISIJ Int. 39B, 268.
2015, 55, 117. [81] J. Li, G. A. Brooks, AISTech, Iron and Steel Technology Conf.,
[52] H. H. Rodríguez, A. N. Conejo, R. D. Morales, Steel Res. Int. 2001, Nashville, USA 2004, 834.
72, 298. [82] D. Guo, G. A. Irons, AISTech, Iron and Steel Technology Conf.,
[53] M. S. Bafghi, H. Kurimoto, M. Sano, ISIJ Int. 1992, 32, 1084. Cleveland, USA 2006, 425.
[54] S. Hara, K. Ogino, ISIJ Int. 1992, 32, 81. [83] J. Li, G. Brooks, N. Provatas, Metall. Mater. Trans. B 2005,
[55] R. Urquhart, W. Davenport, Can. Metall. Q. 1972, 12, 507. 36B, 293.
[56] M. Sattar, J. Naser, G. Brooks, Procedia Eng. 2013, 56, 421. [84] N. Moelans, B. Blanpein, P. Wollants, Calphad 2008, 32, 268.
[57] M. J. Prince, H. W. Blanch, AIChE J. 1990, 36, 1485. [85] N. Arzpeyma, O. Widlund, M. Ersson, P. Jönsson, ISIJ Int. 2013,
[58] H. Luo, H. F. Svendsen, AIChE J. 1996, 42, 1225. 53, 48.
[59] S. C. Koria, W. Lange, Steel Res. Int. 1987, 58, 421. [86] D. Guo, G. A. Irons, 133rd TMS Annual Meeting & Exhibition,
[60] H.-J. Odenthal, U. Falkenreck, J. Schlüter, European Conf. on Charlotte, USA 2004, 149.
Computational Fluid Dynamics (ECCOMAS CFD), Egmond aan Zee, [87] K. Mandal, Modeling of Scrap Heating by Burners, Dissertation,
Netherlands 2006, 220. McMaster University, Canada, June 2010.
[61] H.- J. Odenthal, P. Grygorov, M. Reifferscheid, J. Schlüter, Iron Steel [88] C. Giavani, E. Malfa, V. Battaglia, 10th European Electric Steelmaking
Technol. 2014, 11, 142. Conf., Graz, Austria 2012.
[62] H.- J. Odenthal, W. H. Emling, J. Kempken, J. Schlüter, Iron Steel [89] DRI products & applications - Providing flexibility for steelmaking,
Technol. 2007, 4, 71. www.midrex.com, January 2015.
[63] F. Wang, Z. Jin, Z. Zhu, Ironmaking Steelmaking, 2006, 33, 39. [90] R. Lule, F. Lopez, R. Torres, A. N. Conejo, R. D. Morales, 58th Electric
[64] M. A. Ramirez-Argaez, ISSTech, Ironmaking-Process Technology-Wolf Furnace Conf., Orlando, USA 2000, 47.
Symp. Indianapolis, USA 2003, 1059. [91] R. D. Morales, H. Rodríguez-Hernández, A. N. Conejo, ISIJ Int.
[65] P. Frittella, V. Colombo, E. Ghedini, G. Solari, 10th European Electric 2001, 41, 426.
Steelmaking Conf., Graz, Austria 2012. [92] R. D. Morales, A. N. Conejo, H. H. Rodríguez, Metall. Mater. Trans.
[66] R. Rafiei, A. Kermanpur, F. Ashrafizadeh, Ironmaking Steelmaking B 2002, 33B, 187.
2008, 35, 465. [93] O. J. P. González, Y. I. C. Guzmán, M. A. Ramírez-Argáez,
[67] A. Farschtschi, Steel Times Int. 2011, 9, 32. A. N. Conejo, Arch. Metall. Mater. 2008, 53, 359.
[68] E. Brusa, Meccanica 2014, 49, 2979. [94] E. P. Martínez, C. A. Hernández-Bocanegra, A. N. Conejo,
[69] G. A. Irons, 35th McMaster University Symposium on Iron and M. A. Ramírez-Argáez, ISIJ Int. 2015, 55, 1906.
Steelmaking, Hamilton, USA 2007. [95] L. Labiscsak, G. Straffelini, C. Corbetta, M. Bodino, WIT Trans.
[70] J. D. Anderson, Modern Compressible Flow, McGraw-Hill, New York, Modell. Simul. 2011, 51, 205.
USA 2004. [96] M. Karbowniczek, M. Kawałkowski, 16th IAS Steelmaking Conf.,
[71] H.-J. Odenthal, P. Bui, M. Reifferscheid, E. Hovestädt, J. Nies, Rosario, Argentina 2007, 126.
I. Klioutchnikov, H. Olivier, 4th Int. Conf. on Modelling and [97] X. Tang, M. Kirschen, M. Abel, H. Pfeifer, Steel Res. Int. 2003,
Simulation of Metallurgical Processes in Steelmaking, Düsseldorf, 74, 201.
Germany 2011, 1. [98] N. Arzpeyma, M. Ersson, P. Jönsson, 10th Int. Conf. on CFD in Oil &
[72] E. Murakami, D. Papamoschou, AIAA J. 2002, 40, 1131. Gas, Metallurgical and Process Industries, Trondheim, Norway
[73] B. Allemand, P. Bruchet, C. Champinot, S. Melen, F. Porzucek, 17th 2014, 1.
Process Technology Conf., Orlando, USA 2000, 927. [99] H.-J. Odenthal, A. Kemminger, F. Krause, N. Vogl, AISTech, Iron and
[74] E. Malfa, F. Maddalena, C. Giavani, F. Memoli, 2nd Int. Conf. on Steel Technology Conf., Nashville, USA 2017, 1101.
Process Development in Iron and Steelmaking, Lulea, Sweden [100] H.-J. Odenthal, 18th Int. UIE-Congress on Electrotechnologies for
2004, 429. Material Processing, Hannover, Germany 2017, 9.

steel research int. 2018, 89, 1700098 1700098 (36 of 36) © 2017 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

You might also like