ISIJ International, Vol. 50 (2010), No. 1, pp.

9–16

Mathematical Modeling of the Melting Rate of Metallic Particles

in the Electric Arc Furnace

O. J. P. GONZÁLEZ,1) Marco A. RAMÍREZ-ARGÁEZ2) and A. N. CONEJO1)

1) Graduate Program in Metallurgy, Morelia Technological Institute (ITM), Av. Tecnológico 1500, Col. Lomas de Santiaguito,
58120, Morelia, Michoacán, México. 2) Department of Metallurgy, Facultad de Química, Universidad Nacional
Autónoma de México (UNAM), Edificio “D” Circuito de los Institutos s/n, Cd. Universitaria, Del. Coyoacán, 04510 México D. F.,
México. E-mail: [email protected]
(Received on May 25, 2009; accepted on September 25, 2009 )

A computational fluid dynamics model coupled to a lagrangian model of melting/solidifying particles has
been developed to describe the melting kinetics of metallic particles in an industrial Electric Arc Furnace
(EAF), assuming that liquid steel occupies the entire computational domain. The metallic particles represent
Direct Reduced Iron (DRI). The use of two previous models, an arc model and a fluid flow model has made
possible to evaluate the melting rate of injected DRI in a three phase-EAF, evaluating the influence of the
initial particle size, the initial DRI temperature, feeding position, feeding rate, arc length and some of the
metallurgical properties of DRI. The frozen shell formed in the early stage of the melting process has also
been evaluated in this model.
KEY WORDS: DRI melting; CFD; melting rate.

coefficient was associated with higher melting rates.

1. Introduction
The first mathematical model on the melting rate of DRI
Direct Reduced Iron (DRI) is an important feedstock in was reported by Elliot et al.3) in 1978. They also evaluated
Electric Arc furnaces (EAF). Its melting and dissolution the melting rate of DRI under laboratory conditions in fer-
processes involve heat, mass and momentum transfer, how- rous silicate slags. They found that by increasing both stir-
ever in addition to the complex physical and chemical com- ring conditions and slag temperature, the thickness of the
position of DRI, the melting and dissolution phenomena frozen shell decreases. According with their model, a parti-
also involve a multiphase system, which comprises liquid cle of 10 mm requires approximately 35 s to melt, this time
slag, liquid steel, evolving gases and solid particles. A full is almost the same time it takes to remelt the frozen shell.
understanding of the DRI melting process remains today, as They also reported the influence of density and particle
a challenge. size. In small particles (⬍5 mm), its density doesn’t affect
The melting process of direct reduced iron (DRI) is simi- the melting time, however for large particles (⬎10 mm)
lar to the dissolution kinetics of ferroalloys in steelmaking. those particles with a density higher than that of the slag
Lee et al.1) outlined the influence of several process vari- melt faster.
ables on the dissolution kinetics of ferroalloys, in the ab- Ehrich et al.4) reported results on the melting rate of
sence of turbulence, which can be summarized as follows: sponge iron spheres as a function of initial temperature and
(i) a shell of frozen liquid is formed around the additions, particle size, the heat transfer coefficient and thermal diffu-
the thickness of this shell increases as the thermal conduc- sivity. A mathematical model was developed by solving the
tivity of the addition also increases, (ii) the dissolution heat flow equation for spherical particles. An important fea-
mechanism depends on the melting point of the addition. ture of this work was the model validation by immersing
When the melting point is lower than that of the steel, the dense iron spheres in its own melt. This mathematical
core melts before the shell, in this case the controlling model was subsequently reviewed by Zhang.5) One of the
mechanism is by heat transfer, (iii) an increase in the parti- most important equations in the model is the boundary con-
cle size of the addition increases the dissolution time, (iv) dition which describes the heat balance at the solid–liquid
the time required for the melting of the frozen shell is much interface during the heating of one single particle. In this
higher than the time required for the core. Seaton et al.2) equation the difference between the convective heat flux
developed a heat transfer model to compute the temperature entering the interface and the conductive heat flux leaving
profile during DRI melting in liquid steel, neglecting the the interface is equal to the rate of melting or solidification,
formation of the frozen shell. They found a low heating rate which depends on the sign of the rate of change of the par-
of the inner radius during the early stage and a high heating ticle radius (either increasing or decreasing), furthermore,
rate at the end of the molten process. A higher heat transfer in order to simplify this particular equation it is assumed

9 © 2010 ISIJ
 ISIJ International, Vol. 50 (2010), No. 1

that the thickness of the frozen shell is negligible in com- convective boundary. This approach is able to properly
parison with the radius of the particle, therefore, in the ther- compute both the temperature distribution and the particle
mal balance the radius employed is not the radius at the liq- radius evolution, however, since the computational domain
uid–particle interface but the initial particle radius, as fol- is a single solid sphere of variable radius, the convective
lows: boundary condition is oversimplified and consequently the
local temperature and stirring conditions of the melt are not
∂Tp dr taken into account. It is the objective of this work to present
kp ⫽ ρp Δ H m ⫹ h(TC ⫺ TP ) ..............(1)
∂r dt some results describing the melting process of iron parti-
cles in its own melt in an industrial three phase EAF includ-
Sato et al.6,7) reported laboratory experimental results from ing the local temperature and stirring conditions. This new
the melting of pre-reduced pellets in liquid steel and in approach losses some details in the computation of the DRI
molten slags. They found an increase in the melting rate by melting kinetics, however its main advantage is the cou-
increasing metallization of pellets and by increasing the pling of the local temperature and stirring conditions in the
temperature of the molten bath, however, once the tempera- furnace under specific process parameters with the melting
ture reaches 1 570°C a further increase has no significant kinetics of DRI.
effect on the melting rate.
Aboutalebi et al.8) reported the influence of particle size,
2. Description of Mathematical Model
temperature and stirring conditions on the melting rate of
metallic particles in a ladle. In this work the formation of Frame of Reference: The simulation is restricted to two
the solid shell was neglected. They concluded that the melt- phases; the solid to be melted and the liquid phase. The two
ing rate increases by decreasing the particle size and by in- phase system can be described by establishing an Eulerian
creasing both superheat temperature and stirring conditions. frame of reference to describe the fluid flow and the heat
Similar results have been reported by Jiao and Themelis9) transfer in the continuous phase by solving the turbulent
and Ji et al.10) Navier–Stokes, the continuity and the energy conservations
Zhang and Oeters11) described a mathematical model to equations, additionally, a Lagrangian frame of reference is
represent the melting process of ferromanganese particles used to track the pellets trajectories by using the Newton’s
thrown into a steel ladle. They found a short time, less than second law of motion and allowing exchanges of momen-
1 s, to reach the terminal velocity inside the melt and such tum and heat between the liquid and the particle. The parti-
velocity was used to compute the heat transfer coefficient. cles are allowed to undergo phase transformations such as
A melting distribution time to represent the melting rate of melting or solidification. This Eulerian–Lagrangian ap-
all particles was used to define the melting time of a vari- proach is able to analyze the two phase system and also the
able particle distribution. It was found a higher melting rate time to solve the system of equations is shorter than the
as both the terminal velocity and melt temperature in- Eulerian–Eulerian approach.
creases. The two-phase model to simulate the melting behavior of
Several works have been reported on scrap melting in solid particles in an electric arc furnace bath is composed
Electric Arc Furnaces (EAF). Matson and Ramirez12) inves- of three sub-models, (a) an arc model, (b) an Eulerian fluid
tigated the melting process of scrap assuming spherical iron flow model and (c) a Lagrangian melting model. The first
particles. Gaye et al.13) suggested a maximum scrap size of two models have been reported elsewhere16,17) and only a
120 mm in the converter to avoid unmelted scrap at the end brief description will be outlined below. An important fea-
of the blow. Szekely et al.14) attributed a key role to carbon ture of these models is the computational domain shown in
dissolved in liquid steel to facilitate scrap melting. Li Fig. 1, which considers an industrial size, 3-phase electric
et al.15) used steel bars with various sizes to describe scrap arc furnace of 220 tons.
melting. Arc Model: This model16) is used to define the instanta-
An integral approach coupling heat, mass and fluid flow neous electric power delivered by each phase in a 3-phase
phenomena to understand the DRI melting and dissolution electric arc furnace, as a function of the main electrical pa-
processes in electric arc furnaces has not yet been carried rameters, such as arc length and arc voltage. The arc model
out. It is important to point out that most of the previous is based on the Channel Arc Model modified by Larsen.18)
models reported in the literature deal with numerically Fluid Flow Model: This model17) was developed to
solving a transient heat conduction problem with a mobile compute velocity and temperature fields in the entire com-

Fig. 1. Computational domain.

© 2010 ISIJ 10
 ISIJ International, Vol. 50 (2010), No. 1

putational domain totally filled with liquid steel. It was as- U C ⫽ U C ⫹ U C′ ..............................(4)
sumed that the main driving forces are due to buoyancy
forces. Localized heating is quantitatively described by the
arc model. For a given set of electrical parameters, the in- 2.2. Equation of Thermal Energy Conservation
stantaneous electric power is used as boundary condition in The change of temperature in the solid with respect to
this model. time is obtained by an energy balance, included in the sub-
Melting Model: This model defines the melting rate of routine GENTRA of the commercial code Phoenics, which
DRI pellets in the electric arc furnace as a function of sev- assumes that the particle temperature changes with time
eral process variables. In order to simplify the computa- and the fraction of solid remaining in the particle, T(fs,t),
tions, the following assumptions are made. but at any instant, the temperature is uniform in the whole
– Free surface (no slag). particle.
– Constant thermo-physical properties. ∂Tp df
– Walls at constant temperature of 1 500°C. mpC p ⫽ mp ΔH m s ⫹ h(TC ⫺ TP ) ...........(5)
∂t dt
– Radiation from the walls and roof is not taken into ac-
count. where: mp is the mass of the particle, Cp is the heat capacity
– Steady state conditions in the flow. of the particle, Tp and Tc represent the temperatures for the
– Porous iron spheres of low density and uniform size to particle and the surrounded liquid, respectively, DHm is the
simulate DRI. latent heat of solidification, fs is the fraction of solid phase
– Liquid steel is the continuous phase and DRI the dis- remaining in the particle, h is the convective heat transfer
persed phase. coefficient.
– Uniform temperature in DRI particles. In the previous equation, the left hand side term repre-
– Temperature is uniform across the section of a single pel- sents the accumulation term. The two terms on the right
let but this temperature changes with time. hand side represent the amount of heat released or absorbed
Governing Equations: The equations take into consid- due to solidification or melting and the heat transfer ex-
eration heat and momentum exchange among the two change from the surrounding liquid steel to the particle. At
phases as well as the formation of a frozen shell around the the beginning of computation the Lagrangian model com-
solid phase at the beginning of the melting process. putes the size of the frozen shell. Once the shell is known,
the particle size, Rp, and mass mp remain constant. The
2.1. Equations of Motion melting rate is now controlled by the solid fraction which
The momentum equation associated with each particle is initially is unity and when melting has been completed is
given by the following expression19): zero, finishing with a fully liquid particle of mass mp, a par-
ticle size Rp and a temperature equal to the liquidus temper-
dmpU p 4 ature of the particle. In order to describe the melting rate of
⫽ πRp3 ( ρ p ⫺ ρl )g ⫹ 2πRp2CD ρ U C ⫺U p
dt 3 the particle, the radius of the remaining solid phase is di-
4 dU p rectly related to the fraction of solid (fs). This calculation
⫺ C A πRp3 ρ ...(2)
3 dt differs from previous models where only the solid particle
was considered in the computational domain and as melting
This is the second law of motion, i.e., mass times accelera- occurs the particle size decreases. Although the former
tion equals all forces acting on the particle. The term on the computational approach accurately describes the instanta-
left side of the previous equation represents the rate of neous particle radius and temperature profile within the
change of momentum associated with the particle. The particle, convection is oversimplified, only treated as a
three terms on the right side correspond to the various boundary condition for a single solid sphere. On the other
forces acting on the particle; buoyancy force, drag force hand with the approach followed in our work, there is an
and force associated with the added mass. oversimplification in the radius and temperature evolution
The instantaneous velocity of the particle is given by the of the particles, however the convection governed by the
following equation. heat transfer coefficients are included based on local bath
temperatures, turbulence and stirring conditions prevailing
dx p in a more realistic multiparticle system. The convective
Up ⫽ ..................................(3) heat transfer coefficient is computed based on the
dt
Ranz–Marshall equation20):
Where: xp represents the particle position vector, Up the
particle instantaneous velocity, mp is the mass of each parti- k c ....................(6)
h ⫽ (2 ⫹ 0.6 Re0.5 1/ 3
p Prp )
cle, Rp is the particle radius, CD is the drag coefficient, CA is Dp
a coefficient of added mass, g is the gravity constant, r p and
r l represent the densities of particle and liquid, respectively. Where: Rep and Prp represent the Reynolds number and
UC represents the instantaneous velocity associated with the Prandtl number for the particle, respectively, kc is the ther-
liquid continuous phase, UC is the time averaged velocity of mal conductivity for liquid steel and Dp is the diameter of
liquid and U⬘C is the fluctuating velocity of the liquid due to the particle.
the turbulence. These three velocities are related by the fol- The drag coefficient, CD, is taken from the correlation
lowing expression. provided by Clift, Weber and Grace,21) for spherical parti-
cles and Reynolds numbers lower than 3⫻105.

11 © 2010 ISIJ
 ISIJ International, Vol. 50 (2010), No. 1

24 0.42
CD ⫽ (1⫹ 0.15 Re0.687
p )⫹
Re p 1⫹ 4.25 ⫻ 104 Re⫺p 1.16
...........................................(7)

The solid fraction of the particle is determined from the

lever’s rule:
m
⎛ Tl ⫺ Tp ⎞
fs ⫽ ⎜ ⎟ .............................(8)
⎝ Tl ⫺ Ts ⎠

Where fs is the solid fraction, Tl is the particle’s liquidus

temperature, Ts is the particle’s solidus temperature, Tp is
the particle temperature, and m represents a solidification
index (a value of one was employed in the simulations). Fig. 2. Frozen shell formation and melting time as a function of
arc length.
2.3. Physical Properties of Liquid Steel and DRI
The following physical properties were employed to sim- the thickness of the frozen shell is independent of arc
ulate liquid steel22): density, 72 000 kg/m3; thermal conduc- length. In this case, the particle size increases by about
tivity, 15 W/m K; heat capacity, 670 J/kg K; viscosity, 0.0065 41%. The total melting time is in the range of approxi-
kg/m s. mately 12–17 s for a particle size of 12 mm as a function of
The following physical properties were employed to sim- arc length. As arc length increases the electric power in-
ulate sponge iron3,23–25): density, 2 500–3 500 kg/m3; ther- creases and consequently the temperature of liquid steel
mal conductivity, 0.58139 W/m K; heat capacity, 745 J/kg also increases which yields faster melting rates, therefore,
K; heat of solidification, 274 214.35 J/kg; liquidus tempera- in order to increase the melting rate an operation with long
ture, 1 495°C; solidus temperature, 1 435°C; particle weight, arcs is recommended. It can also be observed the impor-
3–4.4 g; porosity, 30–80%. tance of the frozen shell formation. Its formation process is
The commercial CFD code PHOENICS version 3.4 was instantaneous; however, its remelting process takes approxi-
used to solve the complete set of partial differential equa- mately 50% of the total melting time. This result reinforces
tions, describing conservation of energy, mass and momen- the importance of including the frozen shell formation
tum by the control volume method for the liquid phase. The stage during the melting process. Engh26) reported a theo-
trajectory of DRI particles were simulated under a La- retical analysis which concluded that at high Nusselt num-
grangian frame of reference and solved by the code GEN- bers (strong agitation) and high melt superheat, the frozen
TRA included in the same version of PHOENICS. Each shell formation could be ignored, but according with the re-
calculation took approximately 8 h to converge the fluid sults reported in this work, ignoring this shell represents
flow problem and an additional 10 min to run the particle 50% of error in the melting time. This difference is proba-
track subroutines on a PC with a Pentium-4 processor. bly due to the fact that in multiparticle systems energy bal-
ances are totally different than in single particle systems
where the liquid temperature is assumed to be constant
3. Model Results and Analysis
while in a real process, temperature gradients above 100°C
The results reported in this work correspond to those em- may exist, furthermore, the strong agitation assumed by
ployed in the operation of an industrial electric arc furnace Engh is not valid in the electric arc furnace. This is a reac-
of 220 ton of nominal capacity with the following general tor with low stirring conditions and large thermal stratifica-
dimensions: top radius of 3.474 m and bath depth of 1.5 m. tion.
Steel is produced from 100% DRI, which is continuously The non realistic flat profile of the shell formed around
injected into the furnace at a predefined feeding rate. the solid particle implies instantaneous formation and an
The variables investigated and their range are as follows: extremely low melting rate in the initial stage of the melting
particle size (1–35 mm), porosity (30–90%), density (2 000– process of the frozen shell. This abnormal behavior may be
9 000 kg/m3), solid’s feeding rate (600–5 400 kg/min), solid’s due to the oversimplification of assuming a uniform tem-
temperature (25–1 000°C), inlet position, arc length (25– perature distribution in the whole solid sphere.
45 cm). A standard case was defined for the following set of The effect of the initial particle size on the melting rate
conditions: particle size, 12 mm; porosity, 0%; feeding rate, and thickness of the frozen shell is shown in Fig. 3. This
3 500 kg/min; solid’s temperature, 45°C; voltage phase, figure indicates that when the initial particle size increases,
1 210 V; arc length, 45 cm. Due to software limitations, a the mass and diameter of the frozen shell also increases and
fixed amount of particles was defined in 500. consequently the melting time also increases. When sponge
iron is produced, exhibits a particle size distribution which
3.1. Frozen Shell Formation around DRI Particles ranges from powder to large particles. Typically, almost
The results displayed in Fig. 2 correspond to the standard 40–50% of the total corresponds to particles of 12.5 mm. In
case and various arc lengths. This figure indicates the general, more than 95% of the total covers a range from
growth of a frozen shell around DRI particles as soon as 5–15 mm. Particles below 6.3 mm are considered fines and
they came in contact with liquid steel. It is observed that particles above 16 mm are considered too large particles.

© 2010 ISIJ 12
 ISIJ International, Vol. 50 (2010), No. 1

Fig. 3. Frozen shell formation and melting time as a function of

the initial particle size.

Fig. 5. (a) Melting time as a function of arc length; (b) arc input
power as a function of the arc length for different arc
voltages.16)

Fig. 4. Melting time as a function of the initial particle size and

variable, confirming the advantages of increasing arc
arc length.
length. An increase in the transformer secondary voltage
increases arc length and also the electric power for melting,
3.2. Melting Rate as a Function of the Initial Particle consequently, more thermal energy is supplied to the bath
Size and thus enhancing the melting rate of DRI. Figure 5(b)
Figure 4 shows the melting rate as a function of the ini- shows the relationship between the arc power delivered to
tial particle size and arc length. It is observed that an in- the bath as a function of the arc length for various arc volt-
crease in the initial particle size increases the melting time, ages.16) In other words, in this model an increment in arc
for instance, particles of 12 mm melt in less than 17 s, but length corresponds to increasing the melt superheat. A long
when the particle size increases up to 20 mm, the melting arc operation can be detrimental to both, the furnace refrac-
time increases above 45 s. Elliot et al.3) computed longer tory due to the formation of severe hot spots, and the ther-
melting times for the same particle size; this difference is mal efficiency, if the radiation is out of control, however,
due to differences in the liquid phase employed in both the problem can be solved working with foamy slags.
cases. They immersed solid particles in liquid slag and in
this work the solid particles were immersed in liquid steel. 3.4. Melting Rate as a Function of the Initial DRI
Smaller melting times are expected using a liquid phase of Temperature
higher thermal conductivity. It is important to mention that, Figure 6 shows the effect of the initial temperature of
in spite of model predictions of faster melting times when DRI on melting time. It is observed that the initial tempera-
the particle size is smaller, that the addition of fines is an ture has a marked effect on melting time. The melting rate
extremely inefficient practice, since most of the fines in- increases as the initial temperature of DRI particles also in-
jected do not reach the required momentum to penetrate creases. This will in turn decrease tap-to-tap time, promot-
into the liquid and therefore end up trapped by the off-gas ing a higher productivity. Electric energy consumption
exhaust system or leave the furnace floating over the slag. would also be reduced with the associated savings. Figure
In practice, particles smaller than about 6 mm are consid- 7 reports the influence of the initial temperature of DRI on
ered fines which should be avoided. melting rate for the standard case. It is observed a higher
melting rate as the initial temperature of DRI increases. Ac-
3.3. Melting Rate as a Function of Arc Length cording with the model, changes in the initial DRI tempera-
The previous figure also shows that when arc length in- ture does not affect the thickness of the frozen shell. This is
creases, produces a decrease in melting time, with a more unrealistic, since a higher initial DRI temperature produces
pronounced effect when the particle size is above 10 mm. a lower thermal shock and thus a smaller thickness of the
Figure 5(a) describes in more detail the influence of this frozen shell should be expected. Again, this unrealistic be-

13 © 2010 ISIJ
 ISIJ International, Vol. 50 (2010), No. 1

Fig. 6. Effect of initial particle temperature on melting time. Fig. 9. Effect solids flow rate on the average temperature of the
molten bath.

Fig. 7. Effect of initial particle DRI temperature on melting

time.

Fig. 10. Effect of feeding position on the melting time; (a) tracer
injection, (b) melting time as a function of feeding posi-
tion.

molten bath increases. This behavior can be used to in-

crease the solids flow rate to keep the temperature of the
Fig. 8. Effect of initial particle DRI temperature on the average
molten bath at a constant temperature. The natural trend
temperature of the molten bath.
when the solids flow rate is increased, at a constant temper-
ature, is a decrease in the average temperature of the molten
havior is due to the uniform temperature assumption associ- bath, as shown in Fig. 9.
ated with the Lagrangian model.
As the initial temperature of DRI increases, electric en- 3.5. Melting Time as a Function of the Position of DRI
ergy consumption in the electric arc furnace decreases, evi- Feeding
dently, a higher initial DRI temperature yields multiple ben- In order to decrease the melting time, the optimum posi-
efits. Current technologies for DRI production are capable tion should provide for conditions of higher heating rates of
to provide with hot DRI. In general, temperatures in the the solid. The influence of position of the inlet stream of
order of 600°C have been reported for hot DRI. solid particles into the electric arc furnace is shown in Fig.
Figure 8 shows the effect of the initial DRI temperature 10. According with the model results the injection of DRI
on the average temperature of the molten bath. Increasing in the center of the pitch circle is the optimum position
the initial DRI temperature provides a higher amount of which yields the shortest melting time, this zone corre-
sensible heat, therefore, the average temperature of the sponds to the hottest zone in the furnace. In practice, the

© 2010 ISIJ 14
 ISIJ International, Vol. 50 (2010), No. 1

Fig. 13. Effect of DRI porosity on the melting time of DRI.

Fig. 11. Effect of DRI heat capacity on the melting time of DRI.

Fig. 14. Evolution of the melting process of one DRI particle.

Fig. 12. Effect DRI thermal conductivity on the melting time of
DRI.
creases the melting time also increases.
typical feeding position of DRI is in the position number 1,
which is close to the hottest zone.
4. Model Validation
3.6. Melting Time as a Function of the Physical Prop- In order to validate the mathematical model of this work,
erties of DRI the experimental results reported by Ehrich et al.4) as well
DRI is a material highly heterogeneous, both physically as model predictions reported by Elliot et al.3) will be em-
and chemically, in addition to this, its metallurgical proper- ployed. Figure 14 describes the melting time of one parti-
ties are not well known, in particular its thermal properties. cle of 10 mm in diameter and the surface temperature of
In the present simulations, four properties were chosen to this particle. The current model defines a melting time of
evaluate the influence of those properties on the melting approximately 12 s. The surface temperature reaches ap-
time. Figure 11 shows the influence of heat capacity. As proximately 1 800°C. This high temperature results for the
the heat capacity increases also increases the melting time. standard case operating at the highest voltage. For the same
This is consistent with the definition of heat capacity. As particle size, Elliot defined a melting time of approximately
the heat capacity increases, it also increases the amount of 35 s. The results are consistent. Elliot immersed the parti-
heat required to increase its temperature. Figure 12 shows cles in molten slag and the current model involves liquid
the influence of DRI thermal conductivity on melting time. steel. The higher thermal conductivity of solid steel which
Experimental data on thermal conductivity of DRI was re- forms the frozen shell is responsible of higher melting
ported by Gudenau et al.,25) indicating values in the range rates.
from 2–5 W/m K, for the temperature range from In order to compare with the experimental values from
200–800°C. The thermal conductivity of DRI increases as Ehrich et al.,4) the arc length in the current model was ad-
temperature increases. This increment in thermal conduc- justed to yield an average temperature of 1 600°C, which al-
tivity has positive effects on higher melting rates. The sim- lows similar conditions for the melting rate of 15 mm parti-
ulations from the model suggest a strong influence of the cles. The results are shown in Fig. 15. It is observed the ex-
thermal conductivity on the melting time. Thermal conduc- perimental data are in good agreement with both models. A
tivity can be further increased by increasing DRI metalliza- mayor drawback in previous works is the poor knowledge
tion and decreasing porosity. The influence of porosity on on the convective heat transfer coefficient which excludes
melting time is shown in Fig. 13. DRI is a very porous ma- real furnace fluid flow conditions. This value has been esti-
terial, ranging from 50–70%. The model indicates a nega- mated based on empirical correlations, however, in the cur-
tive influence of porosity on melting time. As porosity in- rent model, a much better and realistic estimation of those

15 © 2010 ISIJ
 ISIJ International, Vol. 50 (2010), No. 1

T : Temperature
t : Time
X : Distance
U : Velocity
DH: Latent heat of solidification
r : Density
Subscripts
p : Particle
c : Continuous phase (liquid steel)
s : Solidus
l : Liquidus

REFERENCES

Fig. 15. Comparison of model predictions with experimental 1) Y. Lee, H. Berg and B. Jensen: Ironmaking Steelmaking, 22 (1995),
data.4) 486.
2) C. Seaton, A. Rodríguez, M. Gonzalez and M. Manrique: Trans. Iron
Steel Inst. Jpn., 23 (1983), 14.
3) J. F. Elliot, J. Nauman and K. Sadrnezhaad: Proc. Third Int. Iron and
coefficients are made due to the computation of velocity,
Steel Congress, ASM-AIME, Metals Park, Ohio USA, (1978), 397.
turbulence and temperature fields in the entire computa- 4) O. Ehrich, Y. K. Chuang and K. Schwerdtfeger: Arch. Eisenhütten-
tional domain. wes., 50 (1979), 329.
5) L. Zhang: Steel Res., 67 (1996), 466.
6) A. Sato, K. Kasashara, R. Nakagawa, S. Yoshimatsu, A. Fukuzawa,
5. Conclusion T. Ozaki, Y. Iwai, Y. Fukuzawa and T. Mitsui: Trans. Iron Steel Inst.
Jpn., 19 (1979), 490.
A two phase Eulerian–Lagrangian mathematical model 7) A. Sato, R. Nakagawa, S. Yoshimatsu, A. Fukuzawa and T. Ozaki:
describing the melting rate of metallic particles in liquid Trans. Iron Steel Inst. Jpn., 20 (1980), 33.
steel in an electric arc furnace has been reported. Operating 8) M. R. Aboutalebi and J. V. Khaki: Can. Metall. Q., 37 (1998), 305.
with smaller particles, injected at high temperatures and 9) Q. Jiao and N. J.Themelis: Can. Metall. Q., 32 (1993), 75.
10) P. Ji, T. Mo-Tang, P. Bing, Y. D. J. Kozinski and T. Chao-bo: J. Cent.
with a large superheat has been confirmed by the current
South Univ. Technol., 14 (2007), 32.
model. Furthermore, it has also been found that in order to 11) L. Zhang and F. Oeters: Steel Res., 70 (1999), 128.
increase the melt superheat that working with long arcs will 12) S. Matson and F. Ramirez: 57th Electric Furnace Conference Pro-
increase the heat supplied for melting and that higher feed- ceedings, ISS, Warrendale, PA, (1999), 719.
ing rates can be used, also, the feeding position affects the 13) H. Gaye, P. Destannes, J. Roth and M. Guyon: Proc. Sixth Int. Iron
and Steel Cong., ISIJ, Tokyo, (1990), 11.
melting rate as well as the metallurgical properties of the
14) J. Szekely, Y. Chuang and J. Hlinka: Metall. Trans. B, 3 (1972),
DRI. DRI with higher metallization has a higher thermal 2825.
conductivity, increasing its melting rate. DRI porosity nega- 15) J. Li, G. Brooks and N. Provatas: Metall. Mater. Trans. B, 36B
tively affects the melting rate. (2005), 293.
The frozen shell period plays a large role in the melting 16) J. L. G. Sanchez, M. A. Ramírez-Argaez and A. N. Conejo: Steel
Res., 80 (2009), 113.
rate of DRI, representing almost 50% of the total melting
17) O. J. P. Gonzalez, M. A. Ramírez-Argaez and A. N. Conejo: ISIJ Int.,
time. The thickness of this shell increases by increasing the 50 (2010), 1.
initial particle size. 18) H. L. Larsen, G. A. Sœvarsdottir and J. A. Bakken: 54th Electric
The melting time of metallic particles in liquid steel is Furnace Conf. Proc., ISS, Warrendale PA, Dallas USA, (1996), 157.
shorter in comparison with the case of melting in slags, this 19) CHAM, PHOENICS 2006. The GENTRA User Guide TR211, La-
grangian Equations, 46.
due to the higher thermal conductivity of the frozen shell.
20) C. Crowe, M. Sommerfeld and Y. Tsuji: Multiphase Flows with
Droplets and Particles, CRC Press LLC, Boca Ratón Florida, (1998),
Nomenclature 62.
CA : Coefficient of added mass 21) R. Clift, J. R. Grace and M. E. Weber: Bubbles, Drops, and Particles,
CD : Drag coefficient Academic Press, New York, (1978), 45.
22) J. Szekely, J. McKelliget and M. Choudhary: Ironmaking Steelmak-
Cp : Heat capacity ing, 10 (1983), 169.
D : Diameter 23) R. C. Gupta and A. A. Shaik: ISIJ Int., 45 (2005), 408.
fs : Fraction of solid phase 24) E. Hajidavalloo and A. Alagheband: J. Mater. Process. Technol., 208
h : Convective heat transfer coefficient (2008), 336.
k : Thermal conductivity 25) H. Gudenau, H. Friedrichs and P. Rademacher: Arch. Eissenhütten-
wes., 62 (1981), 261.
m : Mass 26) A. Engh: Principles of Metal Refining, Oxford Univ. Press, New
Pr : Prandtl number York, (1992), 344.
r,R : Radius
Re : Reynolds number

© 2010 ISIJ 16