Available online at www.sciencedirect.

com

Minerals Engineering 21 (2008) 178189

This article is also available online at:
www.elsevier.com/locate/mineng

Continuous scrap melting in a short rotary furnace

Yanjun Zhang *, Peter V. Barr 1, T. Ray Meadowcroft
Department of Materials Engineering, University of British Columbia, Frank Forward Building, 309-6350 Stores Road, Vancouver, BC, Canada V6T 1Z4
Received 14 June 2007; accepted 25 September 2007
Available online 5 November 2007

Abstract
A study was conducted on steel scrap melting in a continuous process rotary melting furnace, which was envisioned as a replacement
for the electric arc furnace in steelmaking operations. A bench-scale rotary furnace was designed and constructed, followed by a campaign of copper melting trials to explore aspects of furnace operation such as oxygen enrichment and slag thickness. The operational data
was employed to validate the heat-transfer model developed as a part of this work. The model was then used to evaluate commercial
viability of industrial-scale furnaces with melting rates in order of 100 ton h1 and 200 ton h1.
 2007 Elsevier Ltd. All rights reserved.
Keywords: Pyrometallurgy; Process synthesis; Process optimization

1. Introduction
The electric arc furnace (EAF) process is a typical steel
scrap processing method. Although the EAF was originally
designed for both scrap melting and steelmaking, it has
become focused on scrap melting since the metallurgical
adjustments have been largely shifted downstream to the
ladle furnace station. Another important revolution in
the EAF process is to reduce tap-to-tap time (from 3 h to
less than 1 h Evenson and Guerard, 1997) by hydrocarbon
addition via oxy-fuel burners and post-combustion (Warty
and Mathur, 2003; Gitman, 1995; Mathur and Daughtridge, 1993; Kournetas et al., 1997). As a result, the
electricity consumption has been reduced dramatically to
5060% of the total energy input while hydrocarbon fuels
contributing 35% or more (Haissig et al., 1999a; Grant,
2000). This trend will be continued in future because
hydroelectricity is becoming progressively expensive and
*

Corresponding author. Tel.: +1 250 231 2568.

E-mail addresses: [email protected] (Y. Zhang), pvbarr@
in terchange.ubc.ca (P.V. Barr), [email protected] (T.R. Meadowcroft).
1
Oce: Room 213, Frank Forward Building, 309-6350 Stores Road,
Vancouver, BC, Canada V6T 1Z4. Tel.: +1 604 822 3666.
0892-6875/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mineng.2007.09.009

nuclear power generation impacts the environment. However, heat transfer to the bath from the EAF freeboard
combustion is constrained by the combination of operating
practice (foamy slag) and furnace shape (small freeboard
volume and surface area for heat transfer). The heat-transfer eciency has been determined to be 6070% during
scrap melting and 2040% during at bath conditions (Jepson, 2000; Sarma et al., 1996; Grieshaber et al., 1995; Jones
and Oliver, 1995). This encourages people to explore
alternatives.
A potential alternative is the continuous process rotary
melting furnace (CPRMF) proposed by Barr and Meadowcroft (2001). As shown in Fig. 1, the process includes an
oxy-fuel red rotary melting furnace (RMF) and an
optional scrap preheater. Scrap is charged continuously,
while the hot metal is tapped in batches to match the batch
nature of the ladle furnace. By employing a bath weight
much greater than that of the metal tapped, for example
an average bath weight of 200 tons for 50-ton tapping
per 30 min, uctuations in bath volume and temperature
(hence conditions for heat transfer) are minimized and,
overall, the RMF approaches a continuous process in
terms of furnace control. This is the key dierence from
the batch RMFs. The continuous process not only makes

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

179

Table 1
Comparison of energy consumptions by the EAF and the equivalent
CPRMF

Fig. 1. Schematic of the CPRMF process with scrap preheater.

the furnace control easier but also provides a sucient

melting capability to supplant the EAF in minimill steelmaking (where at least 100 ton h1 melting capability
would be needed).
In addition to simplicity, low capital cost and low noise
levels, the advantages of the CPRMF relative to the EAF
might include:
Improved heat transfer due to the regenerative action of
the refractory which also provides a cooling mechanism for the refractory hot-face.
Because only a thin, non-foamed, slag layer is required
to protect the bath surface, the thermal resistance of slag
is minimized and the dust loss associated with foamyslag practice is eliminated.
Distribution of the refractory-slag interaction over the
entire refractory surface.
Improved process control due to near steady-state
operation.
Improved product quality; e.g. low nitrogen and sulphur
pickup.

Without scrap
preheating

With scrap
preheating

Typical EAF a
Electrical energy (kW h t1)
Bath reaction/post combustion (kW h t1)
Burner energy (kW h t1)
Scrap preheating (kW h t1)
Sub total (kW h t1)
Oxygen (N m3 t1)
Electricity for O2 (kW h t1)
Total direct energy (kW h t1)
At-source energy (kW h t1)

380
190
80
NA
650
33
12
662
1250

285
190
50
55
580
33
12
592
1038

Oxy-fuel CPRMF b
Hydrocarbon fuel (kW h t1)
Oxygen (N m3 t1)
Electricity for O2 (kW h t1)
Total direct energy (kW h t1)
At-source energy (kW h t1)

737
149
54
791
872

456
92
33
489
539

Saving of at-source energy

30%

48%

a
b

Typical operating data.

Model predictions.

tion and transmission (Smyth, 1997; Bhatt et al., 1999),

the CPRMF can save at-source energy by 30%. With scrap
preheating, the CPRMF consumes less direct energy and
can save at-source energy by about 50%.
Although promising, the model predictions were not
validated against experimental data and the model itself
was relatively simplistic in certain key assumptions, for
example, the bath and freeboard gas were each assumed
to be well-mixed in both the radial and axial directions
which renders the model essentially 0-dimensional. In this
study, a campaign of copper melting trials was conducted
on a bench-scale rotary furnace. Also, an improved heattransfer model was developed to include the eects of axial
temperature variations within the gas phase. The model
was partially validated against the operational data then
was employed to predict performance of industrial-scale
furnaces under dierent design criteria.

Potential diculties could include:

2. Experimental
Spalling at the refractory hot-face due to thermal
cycling.
Although much reduced relative to the EAF, the thermal resistance of the slag might still constrain the
achievable melting rate.
Using predictions from the preliminary modelling study
(Barr and Meadowcroft, 2001), energy consumptions by
the CPRMF without and with scrap preheating were compared with those by typical EAFs and a shaft furnace
(Haissig et al., 1999b) (EAF with scrap preheating). As
seen in Table 1, without scrap preheating, the CPRMF
requires more direct energy than the EAF. However, considering an overall eciency of 40% for electricity genera-

2.1. Apparatus
As shown in Fig. 2, the bench-scale CPRMF consists of
a refractory lined 0.51 m diameter steel shell and a natural
gas burner at one end with provision for scrap charging
and tapping at the other end. The furnace is rotated by
an electric motor with variable speed gearbox and a chain
drive to the furnace. For discharging the liquid metal, a
hydraulic jack located at the burner end tilts the entire unit
about the pivot located near the discharge-end. Type-S (Pt
PtRh) thermocouples were placed at various depths and
axial locations within the refractory structure in order to
measure radial temperature gradients and thus calculate

180

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

Fig. 2. Conguration of the bench-scale CPRMF 7.

heat transfer through the wall. A portable gas analyzer

(PG-250) was employed to continually measure the concentrations of O2, CO, CO2, SO2 and NOx in the o-gas and
the data was recorded at 1-min intervals with a computer
data logger. After exiting the furnace, air is introduced into
the exhaust gas to complete the combustion process before
releasing it into the building fume collection system.
Although steel scrap would be the material of choice, all
trials were performed using copper because the small shell
diameter (hence refractory thickness) limited the maximum
bath temperature allowable for the trials. In all experiments the bath temperature was maintained at about
1250 C in order to at least approach the 1600 C that
would be required for melting steel scrap. Because some trials were performed with no protective slag layer, thus
exposing the bath surface to oxidizing conditions, copper,
being less active than iron, was deemed the safer material.
The furnace capacity was designed based on a maximum
melting rate of 25 kg h1, which was designed according to
physical constraints. As shown in Fig. 2, the working volume consists of a cylindrical central section (25.4 cm diameter by 40 cm long) with 15 cm long conical dams forming
the ends. The small length to diameter ratio (<3:1) was
selected in order to encourage a well-mixed bath condition.
At the maximum ll level of 20%, the furnace capacity is
about 50 kg of liquid copper (8930 kg m3). For the trials,
the tapping/charging process was carried out at 30-min

intervals during which 10 kg (depending on the designed

melting rate) of liquid copper was tapped into a clay
graphite crucible (Fig. 2) and a corresponding amount of
solid scrap charged into the furnace. The bath weight
was about 4 times greater than that of the tapped metal
so that uctuations in the bath temperature and volume
were minimized and a quasi-continuous process achieved.
The furnace design was based on a two-layer refractory
structure. Due to its high maximum service temperature
(1870 C) and good resistance to corrosion by basic slags,
NARMAG-FG castable (containing 62.4% MgO, 18.8%
Cr2O3, 11.3% Fe2O3 and 5.6% Al2O3) was selected for
the working lining. The insulating layer was PLICAST
LWI-24R (47.6% SiO2 + 37.5% Al2O3 + 9.3% CaO), also
a castable with a low thermal conductivity of 0.5 W m1
C1, a low density of 1360 kg m3, and a maximum service temperature of 1371 C. With the xed total refractory
thickness of 12 cm, a working layer of 2.5 cm combined
with an insulating layer of 9.5 cm was the best compromise
to minimize the shell temperature (174 C, determined
using the preliminary model) with a necessary thickness
of the working lining to prevent penetration of the liquid
metal.
Two burner designs were employed for the trials. For
the initial trials carried out using air with low levels of oxygen enrichment, the burner did not include an auxiliary
cooling circuit as shown in Fig. 3a. Natural gas is supplied
centrally to the nozzle (with combustion air owing in the
annulus) and jetted at an angle of about 15 to the furnace
axis into the surrounding combustion air. The gas and air
ows were measured by rotameters calibrated using a positive displacement, diaphragm-type gas meter and all rotameter readings were corrected to standard temperature
and pressure. The error in reading values was estimated
at about 25% of the increment between the scribe marks
on the rotameter body leading to an overall error in the

Fig. 3. Schematic diagram of the two burner designs: (a) without cooling
and (b) including air-cooling.

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

measured ows of <10% of the corresponding readings as

shown in Table 2. For the trials with high levels of oxygen
enrichment, the burner design was modied to incorporate
an air-cooling circuit as shown in Fig. 3b in order to control the temperature and oxidation rate of the nozzle head.
A suction pyrometer (Fig. 4) was employed to measure
the gas temperature within the furnace freeboard. It consists of a Type-S thermocouple within an alumina tube connected to a vacuum pump by exible tubing. Suction from
the pump produced a reasonable gas velocity (around 75 m
s1) at the thermocouple junction. The velocity was determined roughly using the gas ow rate measured by a gas
meter and was checked by the pressure drop through the
tubing measured by a pressure gauge. The measurement
procedure will be shown later in this paper.

Table 2
The errors in the gas ow measurements

Natural
gas
Oxygen
Air

Rotameter

Readings
in the
trials
(N m3 h1)

Increment
(N m3 h1)

Error
(N m3 h1)

Error/
reading
(%)

Single

1.952.60

0.65

0.16

6.28.2

Single
#1
#2

2.934.04
4.8511.48
1.34

0.73
1.62
0.30

0.18
0.40
0.07

4.56.1
3.58.2
5.2

Fig. 4. Schematic diagram of the suction pyrometer.

181

2.2. Experimental procedure and materials

Based on the preliminary modelling work the key variables to the furnace operation would be the combustion
temperature achieved and the thermal resistance of the
slag. Thus the variables for the trials were:
(1) Oxygen content in the combustion air; i.e. 37.1%,
53.0%, and 80.3%. The trial with 53.0% oxygen was
repeated to check the repeatability. During these trials there was no slag present.
(2) Thickness of slag present layer; i.e. 1.3 and 2.5 cm.
For all trials the furnace rotation was 1 rpm and the ll
set at 20%. It should be noted that this 20% ll included the
slag when present.
Each trial consisted of a warm-up phase followed by a
steady-state operation. As shown in Fig. 5, to begin each
experiment the empty furnace was slowly heated to about
1200 C (hot-face temperature). Then the ow rates of natural gas (96.4% CH4, 1.6% C2H6 and 0.6% C3H8), oxygen
and air were set to the predetermined levels for each trial
(Table 3) and held at these levels for the duration of the
trial. After setting the ring rate, copper scrap (>99.9%
Cu) was charged into the furnace and was melted to form
a full bath (about 47 kg of liquid copper). Once the bath
temperature reached 1250 C, the tapping/charging process
(previously described) was commenced at 30-min intervals.
All conditions were then maintained constant (with tapping/charging every 30 min) until all furnace temperatures
stabilized. This indicated steady-state which was then continued for 1 h during which time all available operating
data were recorded. All experimental results reported in
this paper are based on the measured data during the
steady-state phase.
Once the tapping/charging process was commenced,
temperatures of the bulk metal bath (Tbath), exposed bath
surface (Teb), exposed refractory hot-face (Tew) and freeboard gas (Tg) were measured at 60-min intervals so that
the process was not interrupted too frequently as acquiring
these temperatures required stopping furnace rotation.

Fig. 5. Experimental schedule.

182

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

Table 3
Summary of operating conditions for the bench-scale furnace trials (Trial 2-b replicates conditions for Trial 2-a)
Trial

Slag (cm)

Natural gas (N m3 h1)

Oxygen (N m3 h1)

Air (N m3 h1)

Total (N m3 h1)

O2 level (%)

Excess O2 (%)

1
2-a
2-b
3
4
5

0
0
0
0
1.3
2.5

2.60
1.95
1.95
1.95
1.95
1.95

2.93
3.30
3.30
4.04
3.30
3.30

11.48
4.85
4.85
1.34
4.85
4.85

17.01
10.10
10.10
7.32
10.10
10.10

37.1
53.0
53.0
80.3
53.0
53.0

3.0
10.9
10.9
10.8
10.9
10.9

Tbath was measured by immersing an alumina-sheathed

thermocouple into the liquid copper to a depth of about
3 cm. Teb and Tew were measured both by the thermocouple (inserted through the back-end and brought into contact with each surface) and a dual colour optical
pyrometer. Dierences between the thermocouple and
pyrometer readings were 610 C. Data for determining
Tg were obtained with the suction pyrometer positioned
on the furnace centreline about 18 cm into the furnace
from the back-end. In performing this measurement, a stable thermocouple reading was rst obtained with the suction pump o, which was assumed to be the temperature
of the alumina tube (Ttube) enclosing the thermocouple
junction. The vacuum system was then started and maintained until the indicated temperature stabilized (Ttc). Tg
was calculated based on Ttube and Ttc as shown later. During each trial the shell temperature (Tshell) and refractory
temperatures were recorded on a continuous basis, both
for assessing when steady-sate had been achieved and for
use in the heat balance calculation.
For the trials performed with a slag layer, tail slag from
Teck Comincos zinc-fuming furnace was used as a starting
point but modied by CaO, MgO and SiO2 addition to the
composition of 29.7% SiO2, 23.0% CaO, 29.1% FeO, 7.3%
MgO and 3.5% Al2O3in order to obtain a low melting point
(about 1220 C) (Diaz, 1974), a reasonable slag basicity


wt% CaOwt% MgO

1:0
, and a low viscosity (Elliott et al.,
wt% SiO2
1963; Muan and Osborn, 1965). During the experiments
the slag thickness was measured after tapping by immersing a copper rod (4 mm in diameter) through the slag layer
and into the copper bath for 2 or 3 s. Due to the diering
thermal diusivities of the copper and slag, any portion
of the rod immersed in the liquid copper is melted but a
layer of frozen slag protects the rod within the slag. The
slag thickness was maintained approximately constant by
adding make-up slag as required to compensate for any
loss during tapping. The make-up slag was included in
the energy balance as described later.
2.3. Heat ow calculations
For steady-state operating conditions the furnace heat
balance is:
Q_ comb Q_ bath Q_ shell Q_ open Q_ off-gas

The energy input by combustion was calculated from the

low heating value (Hv = 35,570 kJ N m3) of the gas fuel
and the measured natural gas ow (Fr):
Hv  Fr
Q_ comb
3600

Using a commercial HSC software (HSC Chemistry), the

low heating value was determined based on the gas fuel
composition and a product temperature of 150 C.
For the trials performed without slag, the energy input
to the bath was calculated based on the average melting
rate; i.e. the energy required to bring the scrap charge
(mCu) up to the bath temperature divided by the charging
interval (1800 s):
mCu H bath
Q_ bath
1800

where Hbath is the enthalpy increment in going from 25 C

to liquid bath temperature (obtained using the HSC software). For the trials performed with slag allowance was
made for the additional energy, about 68% of the energy
consumed by the copper scrap, necessary to bring any
make-up slag to the bath temperature. Thermodynamic
data for the slag was obtained using the HSC software
assuming a composition of 2FeO SiO2.
The heat loss through the refractory was calculated
using the measured refractory temperatures on the basis
of steady-state heat transfer through a cylindrical wall with
an additional term to cover the heat loss by conduction
through the end walls 2Q_ end;c ; i.e. for thermocouples 1
and 5 in Fig. 2 (Kreith and Black, 1980a):
T1  T5
Q_ shell lnr5 =r 2Q_ end;c
1

2pkL

where T1 and T5 are the measured refractory temperatures

at radius r1 and r5, k is the thermal conductivity, and L is
the furnace length. From Fig. 2 it is evident that several
thermocouple pairings can be used for this calculation
and the values reported are the average of the values obtained for all pairings. Using the simplied geometry
shown in Fig. 6 the rate of conduction loss through the
back-end refractory was determined by
T bath  T 1
Q_ end;c DL end
h1 A1 end
k end Aend

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

Q_ off-gas

N_ i

183

T off-gas

cp;i dT

298 K

where the summation is made over the species H2O, CO2,

CO, H2, O2 and N2. Again this was done using the HSC
software. For model simulations on industrial-scale furnaces with To-gas > 1600, energy loss as residual combustibles induced by uncompleted combustion was also included
in Q_ off-gas .
In assessing furnace performance, the thermal eciency
was calculated on the basis of:
g 100
Fig. 6. Schematic diagram showing the treatment of the furnace back-end.

where T1 is the environmental temperature at the refractory cold-face, DLend, Aend and kend are the thickness, area
and thermal conductivity of the end-wall, and h1 is the
convection heat-transfer coecient between the cold-face
and environment. Values of the parameters used to determine Q_ end;c for both experiential and model calculations
are summarized in Table 4. For the bench-scale furnace
the cold-face of the back-end wall was enclosed by the exhaust collection hood with a temperature of Texh so
T1 = Texh. For all calculations the rate of conduction loss
through the front-end refractory was assumed to be same
as that through the back-end.
Besides Q_ end;c , at the furnace back-end energy is also lost
by radiation through the opening as shown in Fig. 6.
Assuming a small black surface within a large enclosure
at Texh, the rate of heat loss by radiation was estimated
from Modest (1993a):
"
#
4
T

T
eb
ew
off-gas
4
 T exh
Q_ open Aopen r
6
3

Q_ bath
Q_ comb

In determining Q_ open and Q_ off-gas the value of To-gas was taken according to Tg. In order to calculate Tg from the suction pyrometer data Ttube and Ttc, heat balance for the
thermocouple junction was used as the starting point:
hc T g  T tc hr T tc  T tube

The convection heat-transfer coecient hc was evaluated

using the correlation for forced convection over a sphere
(Kreith and Black, 1980b):
Nu 0:37  Re0:6

17 < Re < 70; 000

10

The radiation heat-transfer coecient was given as:

hr

ePt rT 4tc  T 4tube

T tc  T tube

11

where ePt is the normal emissivity of the platinum thermocouple junction. Calculation result indicated that, based on
the gas velocity of 75 m s1, the thermocouple readings
with the vacuum pump in operation were about 100 C below the true gas temperature.
3. Mathematical model

where Aopen is the area of the opening and To-gas is the ogas temperature (determined to be 250 C lower than the
freeboard gas temperature Tg measured about 18 cm into
the furnace from the back-end). At the burner-end, no radiation loss is accounted because there is no physical
opening.
The rate of energy loss as sensible heat in the o-gas was
calculated using the integrated specic heat expressions and
molar ow, N_ i , for each gaseous component:
Table 4
Parameters used to determine Q_ end;c and Q_ open

Bench scalea
Industrial scaleb
a

T1
(C)

Texh
(C)

End-wall refractory

Opening

Thickness
(m)

Conductivity
(W m1 C1)

Diameter
(m)

500
25

500
1200

0.1
0.5

2.0
0.5

0.1
1.5

Measured data or equivalent values estimated according to

measurements.
b
Assumed.

As shown in Fig. 7, heat-transfer processes within the

rotary melting furnace are as follows:
(1) In the furnace freeboard heat is transferred from the
hot gas to the exposed wall and bath surfaces by both
radiation and convection (Q_ g-ew and Q_ g-eb ). Radiative
exchange also occurs between the wall and bath surfaces Q_ ew-eb . For the furnaces operating at >1600 C
thermal radiation is the dominant mechanism for
heat transfer while convection from the gas-phase
plays only a minor role and can be neglected (Watkinson and Brimacombe, 1978; Barr et al., 1989;
Tscheng and Watkinson, 1979).
(2) Due to furnace rotation, a portion of the energy
absorbed by the refractory during exposure to the
hot freeboard passes into the bath during the period
in contact with the liquid metal (regenerative heat
transfer Q_ cw-cb ) with the remainder being lost through
the refractory by conduction Q_ shell .

184

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

Fig. 9. Eective heat of methane combustion as a function of ame

temperature.

mined by simultaneous solution of the energy balance

equations:
m
X
~
Q_ comb;i
G~
i Gj E b;i  Gj Gi Eb;j
j1

Fig. 7. Heat transfer in the transverse plane of the rotary furnace.

n
X

~
G~
i S j E b;i  S j Gi Eb;j DH s i

12

j1

(3) The energy received by the exposed bath (slag top

surface) is transferred to the bulk metal bath by a
temperature dierence between the slag/freeboard
and slag/metal interfaces.

m
X
~
G~
j S i Eb;j  S i Gj E b;i
j1

n
X

~
_
S~
j S i E b;j  S i S j E b;i Qloss;i

13

j1

The model treats the furnace as three domains; i.e. the

freeboard space, the liquid metal/slag bath and the
refractory.
3.1. The freeboard space
The freeboard space is treated as an enclosure formed by
the exposed bath and refractory surfaces (radiatively opaque, diuse and gray (Susa and Mills, 1995; Kreith and
Black, 1980c) containing an emitting/absorbing gas mixture resulting from CH4 combustion. Radiative exchange
within the enclosure was solved using the zone method
(Hottel and Cohen, 1958) in combination with a clearplus-3-gray-gas model (Modest, 1993b) for the gas phase
(assuming a uniform partial pressure over the entire freeboard space). The extinction coecients and emissivity/
absorptivity weighting coecients were taken from Smith
et al. (1982) In terms of gas ow, the freeboard was divided
into two regions, i.e. a recirculation region induced by the
conned jet burner system followed by a plug-ow region.
Fuel/oxygen mixing was assumed to occur within the recirculation region (where mixed-is-burnt) where recirculation was assumed sucient to approach well-mixed
conditions both radially and axially. In the plug-ow
region, the gas was assumed well-mixed only in the radial
direction but to have a temperature gradient in the axial
direction. Further combustion is under thermodynamic
control.
Using temperature-dependent total exchange areas,
temperature distribution within the enclosure was deter-

The eective combustion energy Q_ comb;i was determined

from Fig. 9 (obtained by HSC calculation on pure methane) by the gradient of local temperature. The uniform
temperature of the recirculation zone was determined by
performing an overall energy balance on this zone:
"
I
m
X
X
_Qcomb;rec
~
G~
i Gj E b;i  Gj Gi E b;j
i1

n
X

j1

~
G~
i S j E b;i  S j Gi Eb;j DH s rec

14

j1

where I is the number of discrete gas volume zones forming

the recirculation zone.
The system of simultaneous equations was solved by
Jacobi relaxation (Vesely, 2001) with assigned temperatures of the bath surface zones. Although the temperatures
of the refractory surface zones are assumed to be known,
an iterative procedure (to be described) is necessary to
ensure that the refractory temperatures produce closure
of the refractory energy balance.
3.2. The refractory structure
Conduction within the refractory was assumed to be
transient, one-dimensional in the radial direction. Thus,
at any axial position along the furnace, the refractory problem is reduced to solving (with implicit nite-dierence
method) the 1-D transient conduction equation


1 o
oT
cp q oT
r
15

r or
or
k ot

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

Boundary condition at the hot-face is the net radiative ux

to that surface zone:
Pn
Pm
~
~
~
~
j1 Gj S i E b;j  S i Gj E b;i
j1 S j S i Eb;j  S i S j Eb;i
q_ i
Ai
16

185

where Tcw,x is the local refractory surface temperature as

determined by the refractory sub-model. The local heattransfer coecient was evaluated using the correlation for
turbulent liquid metal ow over a at plate (Churchill, 1976):
(
)4=5
Rex Pr2=3
Nux 0:032
19
1 0:0468=Pr2=3 1=2

3.3. The metal bath and slag

Due to small furnace length/ID ratios of 65:1, the liquid
metal bath is assumed to be well-mixed in both the transverse and axial directions, the justication being a combination
of
good
thermal
conductivity
(about
43 W m1 C1 for 1%C steel) (Kreith and Black, 1980d)
and rotation-induced stirring along with the tumbling
motion of the solid scrap component within the bath.
The slag is assumed to have a temperature dierence
between the slag/freeboard and slag/metal interfaces
(DTslag, which is specied) while heat transfer in the axial
direction was neglected. Both the metal bath and slag are
assumed to be at steady-state, i.e., scrap melting occurs
at a constant metal bath temperature Tbath (specied) with
a constant DTslag over the entire melting period.
At steady-state the radiative ux to the slag top surface
q_ g-eb q_ ew-eb is transferred to the bulk metal bath by the
assigned DTslag; i.e.
k eff DT slag
q_ g-eb q_ ew-eb
DLslag

17

3.4. Solution procedure

The three models are linked by shared boundary conditions and the requirement that the furnace itself operates at
steady-state. As shown in Fig. 8, in the freeboard model the
net radiative ux q_ i to each refractory surface zone is calculated using its temperature (Ts,i) as determined by the
refractory model. Using these heat ux values fresh values
for Ts,i are then calculated by the refractory model which
are returned to the freeboard model; etc. The steady-state
condition can be viewed in the way that at a certain circumferential position temperatures of the refractory elements
converge to constant values.
3.5. Energy balance and melting rate
Model-predicted heat release rate from combustion is:
Q_ comb Q_ comb;rec

I
X

Q_ comb;i

20

i1

Heat ux on the liquid side of the bath-refractory interface

is determined by

where I is the number of the gas zones forming the plugow region. The total heat-transfer rate to the metal bath is

q_ cw-cb;x hcw-cb;x T cw;x  T bath

Q_ bath Q_ eb Q_ cw-cb

18

Fig. 8. Schematic diagram of the three calculation domains.

21

186

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

where Q_ eb is the total radiant energy to the exposed bathsurface zones and Q_ cw-cb is the total regenerative heat-transfer rate to the covered bath. Although Q_ shell ; Q_ open and
Q_ off-gas are determined by the similar procedures as those
in the experimental calculations, the measured temperatures are replaced by the model-predicted surface-zone
and gas-zone temperatures.
Scrap melting in a liquid metal bath is controlled by heat
transfer and the eect of mass transfer can be neglected
because the scrap has a similar composition to the liquid
bath (Li et al., 2005). At steady-state the total heat-transfer
rate to the metal bath Q_ bath is equal to the rate of energy
supply for scrap melting. With no scrap preheating, the
model-predicted melting rate is
_
_ Qbath  3600
M
H bath

22

The furnace thermal eciency is dened by the same

expression as that in the experimental calculation. If the
scrap is preheated by the furnace o-gas, the melting rate
becomes
_
_ Qbath  3600
M
H bath  H scrap

refractory hot-face, the gas temperature was measured at

only one axial position (0.53 m). The hot-face temperature
was measured at the same axial position. Also, it was backcalculated at the two axial thermocouple positions within
the refractory (Fig. 2) using the measured internal temperatures to calculate the rate of heat transfer which could
then be used to calculate the hot-face temperature. As
shown in Fig. 11, the model predictions are reasonably
good for the hot-face temperatures but high on the gas
temperature by about 80 C. Based on the model predictions, the gas temperature drops by about 250 C between
the point where the measurement was made and the exit
from the furnace.
Validation of the model was also made against individual components of the furnace heat balance as summarized
in Table 5. It can be seen that all of the model predictions,
except for those in Trial 5, are in good agreement with the
experimental values with the largest error being about 16%.

23

where Hscrap is the sensible heat in unit weight of the preheated scrap. The overall thermal eciency is
g 100

_
Q_ bath H scrap  M=3600
Q_ comb

24

4. Partial validation of the model

Using the operating conditions in the trials, the model
was used to simulate the bench-scale furnace. For the simulations the emissivity of the refractory, the liquid copper
and the high iron slag were assumed to be 0.7, 0.14 and
0.9, respectively (Baukal, 2000a,b; Markham et al., 1992).
The model predictions were compared with the experimental results in terms of both temperature values and the
energy balance. As shown in Fig. 10, the model predictions
for the refractory temperatures at the axial position of
0.53 m are in good agreement with the data measured at
the same axial position. Although the model develops the
axial temperature prole for the gas phase and the exposed

Fig. 11. Validation against gas and refractory hot-face temperatures

(Trial 2-a).
Table 5
Validation against terms in the furnace heat balance
Trial
1

2-a

2-b

5
Fig. 10. Validation against refractory temperatures (Trial 2-a with 53%
O2, no slag).

From
data
Model
Error
From
data
Model
Error
From
data
Model
Error
From
data
Model
Error
From
data
Model
Error
From
data
Model
Error

Q_ comb
(kW)

Q_ bath
(kW)

Q_ shell
(kW)

Q_ open
(kW)

Q_ off-gas
(kW)

g
(%)

25.7

4.20

7.8

2.8

10.6

16.3

25.7
614%
19.3

4.79

7.9

2.8

10.6

18.6

4.25

7.8

2.8

6.6

22.0

19.3
616%
19.3

3.76

7.8

2.1

5.8

19.5

4.20

8.0

2.3

5.7

21.8

19.3
65%
19.3

4.07

7.8

2.1

5.7

21.1

5.22

7.9

2.4

4.5

27.0

19.3
613%
19.3

5.59

8.0

1.8

4.5

29.0

3.40

7.9

2.4

5.6

17.6

19.3
65%
19.3

3.29

7.8

2.2

5.6

17.0

2.26

7.8

2.6

5.9

11.7

19.3
628%

2.89

7.8

2.4

5.7

15.0

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

In Trial 5 a thick slag layer was applied resulting in a large

error in the measured Q_ bath . This large error combined with
the small Q_ bath value results in a large drift (28% of the
measured data) between the measured and model-predicted
results. In Trial 3 (80% O2, no slag) a melting rate of
25 kg h1 was achieved with a thermal eciency of 27%.
This performance is not bad for the bench-scale furnace.
5. Model predictions for industrial-scale furnaces
Table 6 summarizes the baseline design and operating
conditions assumed for industrial furnaces ring with
100% oxygen to melt steel scrap. For all simulations the
length of recirculation zone Lrec (fuel/oxygen mixinglength) was assumed to be 30% of the furnace length (L).
The partial pressure (P) of CO2 + H2O in the freeboard
gas was determined by HSC calculation on CH4/O2 combustion at 2000 C.
Without scrap preheating, a melting rate in order of
100 ton h1 can be achieved by a 4 m ID 16 m furnace
operating at a ring rate of 6000 N m3 h1. Furnace performance and energy dispositions are summarized in Table 7.
Using the calculation procedure described previously, this
furnace consumes less direct energy (625 kW h t1, including the electricity for oxygen generation) than the EAF
without scrap preheating (about 660 kW h t1) which
translates into an at-source energy saving of 45%.
Although competitive in terms of saving energy, the 4 m
ID 16 m furnace has a critical working condition for the
refractory. Fig. 12 shows the circumferential hot-face temperature prole at 1.6 m into the furnace from the burner
end which, under these operating conditions, is the location
for the maximum refractory hot-face temperatures

187

(Thf,Max = 2020 C) and also the maximum thermal cycling

of the hot-face (DThf,Max = 393 C). The maximum temperature is pushing the limit for conventional refractories in
terms of Thf,Max and the high level of thermal cycling might
well lead to thermally induced spalling.
With the xed ring rate of 6000 N m3 h1, an increase
of furnace geometry does not improve the melting rate as
shown in Fig. 13 because the gas temperature (1773 C)
at the back-end of the 4 m ID 16 m furnace is close to
those of the exposed bath surface (1650 C) and refractory
hot-face (1649 C, averaged in the circumferential direction) so energy extraction from the gas is close to the maximum possible. However, for the xed energy release rate
from combustion, the increased freeboard gas volume
results in a low gas temperature. The low gas temperature
combined with an enlarged refractory hot-face area leads
to a low peak temperature (Thf,Max) and a small thermal
cycling (DThf, Max) in the hot-face as shown in Fig. 14, indicating a better working condition for the refractory.
If the scrap is preheated by the o-gas energy (sensible
heat plus residual combustibles CO and H2), both the melting rate and overall thermal eciency are improved significantly. As shown in Fig. 15, for the 4 m ID 16 m furnace
operating with the ring rate of 6000 N m3 h1, the melting
rate increases to 151 ton h1 if 100% of the o-gas energy is
used to preheat the scrap. At the same time, the overall
thermal eciency increases to 96% with only minor energy
(4% of Q_ comb ) being lost through the furnace refractory and
the opening. With a specic fuel consumption (SFC) of
397 kW h t1 and a specic oxygen consumption (SOC)
of 80 N m3 t1, this conguration saves direct energy by
28% and at-source energy by 55% relative to the EAF with
scrap preheating.

Table 6
Summary of baseline design and operating conditions assumed for the industrial-scale furnaces
L/ID

Fill (%)

Fuel

Lrec/L

Rotation (rpm)

Tbath (C)

DTslag (C)

P (atm)

Emissivities
Wall

Slag

4:1

20

CH4

0.3

1.0

1600

50

0.85

0.7

0.9

Refractory specications
Thickness (m)
Conductivity (W m1 C1)
Density (kg m3)
Specic heat (kJ kg1 C1)

Working layer

Intermediate

Backup

0.20
5.0
3400
1.1

0.15
1.25
1500
1.0

0.15
0.75
1000
0.95

Properties of steel scrap (HSC Chemistry; Kreith and Black, 1980d; Yaws, 1994)
Liquidus (C)
Hbath (kJ kg1)
Conductivity (W m1 C1)

Density (kg m3)

Viscosity (m2 s1)

1500

7800

7.43 107

1350

43

Table 7
Model-predicted furnace performance
Q_ off-gas (MW)

g (%)

_ (ton h1)
M

SFC (kW h t1)

SOC (N m3 t1)

4 m ID 16 m furnace with a ring rate of 6000 N m h

59.8
39.1
1.6
0.9

18.3

65.4

103

583

117

6 m ID 24 m furnace with a ring rate of 11,500 N m3 h1

114.6
77.1
3.1
0.9

33.1

67.3

202

567

114

Q_ comb (MW)

Q_ bath (MW)

Q_ shell (MW)

Q_ open (MW)
3

1

188

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

Fig. 12. Circumferential hot-face temperature prole at the axial position

of 1.6 m (4 m ID 16 m furnace).

Fig. 13. Melting rate versus furnace size.

For the xed ring rate of 11,500 N m3 h1, a smaller

furnace of 5 m ID 20 m also results in a similar melting
rate, 194 ton h1. However, both the maximum hot-face
temperature and the thermal cycling (2068 C and 442 C,
respectively) are greater than those in the 6 m ID 24 m
furnace (2005 C and 385 C, respectively), indicating a
worse working condition for the refractory. A farther
decrease in the furnace geometry to 4 m ID 16 m leads
to a signicant decrease in both the melting rate
(169 ton h1) and thermal eciency (56.1%). Temperature
of the gas mixture leaving the furnace is as high as
2097 C, meaning that a large portion (41%) of the combustion energy is lost with the o-gas. Therefore, at the ring rate of 11,500 N m3 h1 energy extraction from the
freeboard gas in the 4 m ID 16 m furnace is far away
from the maximum possible.
Again, the performance of the 6 m ID 24 m furnace is
improved signicantly by scrap preheating. The melting
rate and thermal eciency increase to 289 ton h1 and
96%, respectively when 100% of the o-gas energy is used
for scrap preheating. The SFC and SOC (396 kW h t1
and 80 N m3 t1, respectively) are same as those for the
4 m ID 16 m furnace operating with 100% of the o-gas
energy for scrap preheating, indicating a same energy saving relative to the EAF with scrap preheating.

6. Conclusions

Fig. 14. Thf,Max and DThf,Max vs furnace size.

Fig. 15. Improvement of furnace performance by scrap preheating.

With no scrap preheating, a melting rate in order of

200 ton h1 would require a 6 m ID 24 m furnace operating with a ring rate of 11,500 N m3 h1. As shown in
Table 7, the SFC and SOC are close to those for the 4 m
ID 16 m furnace with no scrap preheating, meaning an
energy saving in the same order compared with the EAF.

In this study a campaign of copper melting trials was

conducted on a bench-scale CPRMF. Also, an improved
heat-transfer model was developed based on the previous
modelling study. After validation by the operational data,
the model was employed to predict performance of industrial-scale CPRMF for steel scrap melting. The results
can be summarized as follows:
(1) When melting copper with 80% oxygen in the combustion air and no slag present, the bench-scale
CPRMF achieved a melting rate of 25 kg h1 with
a thermal eciency of 27%. This supports the viability of the CPRMF concept which oers signicant
advantages over existing batch-type rotary furnaces.
(2) Model predictions indicated that with no scrap preheating, a melting rate in order of 100 ton h1 can
be achieved by a 4 m ID 16 m furnace operating
at a CH4 ring rate of 6000 N m3 h1. Doubling the
melting rate would require a double ring rate combined with an enlarged furnace size of 6 m
ID 24 m. Compared with the EAF, saving of atsource energy by the CPRMF is about 45%.
(3) The furnace performance can be improved signicantly by scrap preheating. If 100% of the o-gas
energy is employed to preheat the scrap, the melting
rate will be increased by 4050%. Saving of at-source
energy relative to the EAF with scrap preheating will
be about 55%.

Y. Zhang et al. / Minerals Engineering 21 (2008) 178189

(4) For the 4 m ID 16 m furnace, energy extraction

from the freeboard gas is close to the maximum possible. An increase in furnace size will not improve the
melting rate, but will lower the hot-face temperature
and reduce the thermal cycling of the refractory.
Acknowledgements
Financial support for this work by the Natural Sciences
and Engineering Research Council of Canada (NSERC) is
gratefully acknowledged. The authors would also like to
thank Mr. Milaim Dervishaj, Research Technician, for
assistance with the experimental work to provide the validation data.
References
Barr, P.V., Meadowcroft, T.R., 2001. Scrap melting in an oxy-fuel red
rotary furnace. In: Electric Furnace Conference Proceedings, pp. 451
463.
Barr, P.V., Brimacombe, J.K., Watkinson, A.P., 1989. A heat-transfer
model for the rotary kiln: Part II development of the cross-section
model. Metallurgical Transactions B 20B, 403419.
Baukal Jr., Charles E., 2000a. Heat Transfer in Industrial Combustion.
Chapter 3 Heat Transfer Modes, Table 3.2 Normal Total
Emissivities of Refractories as a Function of Temperature. CRC Press
LLC, New York, E-book from the Library at University of British
Columbia.
Baukal Jr., Charles E., 2000b. Heat Transfer in Industrial Combustion.
Chapter 3 Heat Transfer Modes, Table 3.1 Normal Total Emissivities
of Various Surfaces as a Function of Temperature. CRC Press LLC,
New York, E-book from the Library at University of British Columbia.
Bhatt, M.S., Mandi, R.P., Jothibasu, S., Rajkumar, N., 1999. Performance enhancement in coal red thermal power plants. Part IV:
Overall system. International Journal of Energy Research 23, 1239
1266.
Churchill, S.W., 1976. A comprehensive correlating equation for forced
convection from at plates. AIChE Journal 22 (2), 264268.
Diaz, C., 1974. The Thermodynamic Properties of Copper-Slag System.
In: Incra Monograph III: The Metallurgy of Copper. International
Copper Research Association, Inc., pp. 105.
Evenson, E.J., Guerard, J., 1997. Energy optimization and continuous
fume analysis at CO-STEEL LASCO. In: Electric Furnace Conference
Proceedings, pp. 435453.
Elliott, J.F., Gleiser, M., Ramakrishna, V., 1963. Thermochemistry for
Steelmaking: Thermodynamic and Transport Properties. AddisonWesley Publishing Company, Inc, pp. 673.
Gitman, G., 1995. EAF thermal enhancement with the use of gases and
solid fuels. In: Electric Furnace Conference Proceedings, pp. 185198.
Grant, M., 2000. Principles and strategy of EAF post-combustion. In:
58th Electric Furnace Conference Proceedings, November 1215, pp.
1528.
Grieshaber, K., Feldermann, C., Jones, J., 1995. Considerations for a
post-combustion strategy. In: Electric Furnace Conference Proceedings, pp. 219234.
Haissig, M., Fuchs, G., Auer, W., 1999a. Electrical arc furnace technology
beyond the year 2000. Metallurgical Plant and Technology International 22 (1), 5663.
Haissig, M., Fuchs, G., Auer, W., 1999b. Electrical arc furnace technology
beyond the year 2000. Metallurgical Plant and Technology International 22 (1), 5663.
HSC Chemistry for Windows, Outokumpu Research, Finland, Fax: +358-2626-5310, Telephone: +358-2-626-6111, E-mail: [email protected].

189

Hottel, H.C., Cohen, E.S., 1958. Radiant heat exchange in a gas-lled

enclosure: allowance for nonuniformity of gas temperature. AIChE
Journal 4 (1), 314.
Jepson, S., 2000. Chemical energy in the EAF: benets and limitations. In:
58th Electric Furnace Conference Proceedings, November 1215, pp.
314.
Jones, J.A.T., Oliver, J.F., 1995. Post-combustion: a practical and
technical evaluation. In: Electric Furnace Conference Proceedings,
pp. 199210.
Kreith, F., Black, W.Z., 1980a. Basic Heat Transfer. Harper & Row,
Publishers, Inc., New York, pp. 55.
Kreith, F., Black, W.Z., 1980b. Basic Heat Transfer. Harper &
Row, Publishers, Inc., New York, pp. 275.
Kreith, F., Black, W.Z., 1980c. Basic Heat Transfer. Harper & Row,
Publishers, Inc., New York, pp. 299302.
Kreith, F., Black, W.Z., 1980d. Basic Heat Transfer. Harper & Row,
Publishers, Inc., New York, pp. 511.
Kournetas, N.G., McLean, A., Sommerville, I., Thomson, M., Guerard,
J., 1997. Eects of post combustion validated by steel mill
demonstration. In: Electric Furnace Conference Proceedings,
pp. 373380.
Li, J., Brooks, G., Provatas, N., 2005. Kinetics of scrap melting in liquid
steel. Metallurgical and Materials Transaction B 36B, 293302.
Mathur, P., Daughtridge, G., 1993. Oxygen injection for eective postcombustion in the EAF. In: Electric Furnace Conference Proceedings,
pp. 299303.
Markham, J.R., Best, P.E., Solomon, P.R., Yu, Z.Z., 1992. Measurement
of radiative properties of ash and slag by FT-IR emission and
reection spectroscopy. Journal of Heat Transfer 114, 458464.
Muan, A., Osborn, E.F., 1965. Phase Equilibria Among Oxides in
Steelmaking. Addison-Wesley Publishing Company, Inc.
Modest, M.F., 1993a. Radiative Heat Transfer. In: McGraw-Hill Series in
Mechanical Engineering. McGraw-Hill, Inc., New York, pp. 206.
Modest, M.F., 1993b. Radiative Heat Transfer. In: McGraw-Hill Series
in Mechanical Engineering. McGraw-Hill, Inc., New York, pp. 586
638.
PG-250 portable gas analyzer is manufactured by HORIBA INSTRUMENTS INCORPORATED, 17671 Armstrong Irvine Industrial
Complex, Irvine, California 92614, USA, Telephone: +1 (949) 2504811, Fax number: +1 (949) 250-0924.
Sarma, B., Mathur, P.C., Selines, R.J., 1996. Heat transfer rates and
mechanisms in EAF post-combustion. In: Electric Furnace Conference
Proceedings, pp. 519537.
Smyth, R., 1997. The use of high temperature heat exchangers to increase
power plant thermal eciency. Proceedings of the Intersociety Energy
Conversion Engineering Conference 3 (4), 16901695.
Susa, M., Mills, K.C., 1995. Slag Atlas, second ed.. In: Optical Properties
of Slags: Data for Refractive Indices and Absorption Coecients,
Chapter 16 Woodhead Publishing Limited, Abington, Cambridge,
England, pp. 602616.
Smith, T.F., Shen, Z.F., Friedman, J.N., 1982. Evaluation of coecients
for the weighted sum of gray gases model. Transactions of the ASME
104, 602608.
Tscheng, S.H., Watkinson, A.P., 1979. Convective Heat Transfer in a
Rotary Kiln. The Canadian Journal of Chemical Engineering 57, 433
443.
Vesely, F.J., 2001. Computational Physics: An Introduction, second ed.
Kluwer Academic/Plenum Publishers, New York.
Warty, S.K., Mathur, P., 2003. Praxairs CoJet gas injection system
delivering benets and gaining useful knowledge for future enhancements. Iron and Steelmaker 30 (3), 2330.
Watkinson, A.P., Brimacombe, J.K., 1978. Heat transfer in a direct-red
rotary kiln: Part II heat ow results and their interpretation.
Metallurgical Transactions B 9B, 209219.
Yaws, Carl L., 1994. In: Handbook of Viscosity: Inorganic Compounds
and Elements, vol. 4. Gulf Publishing Company, Houston, Texas, pp.
105.