CFD modeling of complete walking hearth reheating furnaces

V. Battaglia1, E. Malfa1, M. Fantuzzi2, E. Filippini3

1
Centro Sviluppo Materiali - Italy,
2
Tenova LOI Italimpianti - Italy,
3
Feralpi Siderurgica SpA - Italy
[[email protected]]

Abstract: The paper describes the Computational

in accordance with the NOx emissions limits
Fluid Dynamics (CFD) approach employed to model permitted by the environmental standards.
100 tph walking heart reheating furnace of Feralpi All those targets have to be achieved with
Siderurgica (LAM2), including case set up and “reasonable” costs (capital and operating).
validation by comparison between calculated and Therefore a great deal of attention is needed both
measured heating curves. The two main constrains in optimizing the furnace design and in reducing
in approaching the simulation of a complete furnace, unnecessary energy consumption.
the large number of computational cells necessary to Since it is well known that radiative heat
reproduce the geometry details and the modelling of transfer is the major mechanism to heat-up the
moving steel stock have been addressed. The results
stock inside the reheating furnace, simplified
allow to know in details the flow characteristics, the
species concentration, the temperature gradients,
models are commonly used to analyze the heating
within both the WH reheating furnace and each billet, characteristics and efficiency of reheating
and more in general to evaluate furnace heat balance. furnaces. The state of art is zonal methods, one or
two-dimensional for preliminary calculation of
Keywords – reactive flow simulation, CFD, reheating
stock temperature or three dimensional for taking
furnace, scale formation
into account the effect of skid pipe and the
temperature of stock edges [1-4].
I. INTRODUCTION On the other side the existing measurement
methods do not provide continuous measurement
The main target of a reheating furnace is of temperature: trailing thermocouples can be used
always the production that is directly linked to the only inside test stock to measure the reheating at
efficiency of heating process of stock (slabs, few measurement points, while optical pyrometers
blooms, billet, thin slag, tubes, strip). or thermal camera only give information about the
surface temperature and they are affected by
presence of scale.
Therefore in the last years two main
development are on-going thanks to increasing
computational power:

3D online control tool based on existing
zonal model extended to simulate the transient
radiative heating [5];

3D Computational Fluid Dynamics
(CFD) simulation of complete furnace coupling
the reacting flow combustion models developed
for the single burner and conjugated heat transfer
Figure 1. Targets priority for furnaces in the [6,7].
steel industry The work presented in this paper is focused on
the set-up of CFD model to simulate complete
Of course other important aspects became more furnace with the goal to reproduce in detail the
and more relevant such as: fluid dynamic, the thermal and the species fields

the product quality; inside the furnace together with the stock heating

the energy saving (the minimum energy curve. The 3D detailed information should allow
requirement compatible with the production target to evaluate the effect of different burners
and the best product quality); equipment and furnace design on the heating

the environmental impact (the minimum characteristics and efficiency. Moreover, knowing
pollution compatible with the previous topics and the stock surface temperature and local

1
concentration of combustion products, the scale Since in the past to simplify the zonal methods
growth can be evaluated. the steel stock movement was represented by a
continuous solid moving at fixed velocity
(equivalent strip concept) this simplified approach
II. THE CFD APPROACH AT COMPLETE has been used as starting point of the CSM
REHEATING FURNACE SIMULATION development work [7].
The physical and numerical models based for
single burner simulation [7,8] have been the III. THE FERALPI FURNACE TEST CASE
staring point for extending to complete furnace
(multiple burners). The main problems to To test and validate the CFD model for
overcome have been: reheating furnace the Feralpi walking heart type

the large number of computational cells furnace LAM 2 has been chosen as representative
necessary to discretise the different geometry scale test case. In fact in frame of the RFCS CO2RED
(burners .vs. furnace .vs. stocks); project, the furnace has been extensively

the modelling of moving steel stock (in monitored and billets have been instrumented with
present applications the billets). thermocouples to characterize the billets heating
To represent the furnace elements with curve. Experimental data are, therefore, available
sufficient details and to have, at the same time, to validate the CFD model and test its sensitivity
sufficient resolution of flame interaction, to operative conditions and furnace production.
computational mesh with tens of million cells is The Feralpi furnace, sketched in Figure 2,
required. Taking into account that the main part of consists of four control zones plus a pre-heating
the cells (more than 50%) are concentrated in the tunnel separated by a baffle.
burners region, where very high temperature,
velocity and species concentration gradients
Rolling mill no. 2
request high grid resolution, a procedure has been
developed to reduce the grid size [7]. Reheating furnace: TS Impianti
 Type Walking hearth
The representation of stock heating requires  Length 23.000 mm
 Width 12.900 mm
time dependent calculation due to the movement  Fuel Natural gas
inside the furnace and consequently very high  Burners: Roof
 Control zones: 4
computation time. To overcome this limitation  Nominal cap. 80 t/h
Charging T. Ambient
CSM has simplified the problems coupling the 
 Discharging T. 1.150-1200°C
CFD calculation, that iteratively performs the  Temperature control: thermocouple
stationary simulation of the reacting flow inside  Heat recovery: centralized air preheating
 Input Billets (carbon steel):
the furnace (furnace model) with the non 140x140x12000 mm – 1 row 140x140x6000 mm – 2 rows
stationary simulation of the stock during its 130x130x12000 mm – 1 row 130x130x6000 mm – 2 rows

advancement in the furnace (stock model). The

selected approach is possible due to the different
Figure 2. Outline of the Feralpi reheating furnace
characteristic time scales of the two processes: the LAM2
fluid dynamic time is of the order of tenth of
seconds while the solid thermal time is of the order The preheating and heating zone are equipped
of seconds. Hence the billets surface temperature with 24 NG roof burners (750kW each one). The
can be assumed constant in time during the fluid- soaking zone consists of two control zones with 12
dynamic calculation while the resulting heat flux NG roof burners (400kW each one). The installed
on the billets can be assumed as boundary total thermal power is 43 MW. The combusting
condition during the simulation of stock movement air is preheated by a central recuperator.
inside the furnace. The solid simulation returns The nominal furnace production is 80 tph,
updated billets surface temperature to the fluid however the furnace is presently often used up to
simulation in an iterative process that terminates heat up 125 tph. Table 1 reports the conditions
when the difference between the billets surface during January and April 2009 monitoring
temperature in two successive iterations is less campaign corresponding to 121 tph and 126 tph
than a tolerance value. The stock model has been furnace production of 140x140x12000mm billet,
developed in the Ansys FLUENT UDF [9] and it is respectively.
applicable for all types of steel stock geometry and
handling system.

2
 Table 2. Feralpi WHF heat balance
Table 1. Operative condition of Feralpi furnace (legends: PRO=production, TTPI=Total thermal power
(NG composition : CH4=98.895%mol; (gas+hot air), ST=stock exit temperature, STG=power to
CO2=0.173%mol; N2=0.932%mol) stock, FGH= flue gas heat losses, FHL= furnace heat losses).
3
FR air 35441 Nm /h
zone1 zone2 zone3 zone4
17287 11000 3577 3577
TEST PRO TTPI ST STG FGH FHL η
[tph] [kW] [°C] [kW] [kW] [kW] [%]
3
FR gas 3552 Nm /h 01/09 121 40101 1170 26366 12006 1729 65
zone1 zone2 zone3 zone4
1747 1058 373 373 04/09 126 42293 1150 27733 12767 1793 65

air/gas ratio
9.90 10.40 9.59 9.59
IV. TWO DIMENSIONAL FURNACE
Power (kW)
zone1 zone2 zone3 zone4 THERMAL MODEL
17370.6 10519.8 3708.8 3708.8
a) January 2009: combustion air @ 370°C To set-up and test the stock model preliminary
3
work has been conducted by means of two
FR air 39024 Nm /h
zone1 zone2 zone3 zone4
dimensional ANSYS FLUENT representation of
17450 13141 4150 4283 Feralpi WHF (figure 3). The 2D model considers
3
only energy and radiation (Discrete Ordinates
FR gas 3923 Nm /h
zone1 zone2 zone3 zone4 model) equations. The zone temperatures have
1758 1279 436 450 been imposed as wall boundary conditions. The
air/gas ratio approach is equivalent to the zonal method.
9.93 10.27 9.52 9.52
Tunnel wall temperature from CFD

Power (kW)
zone1 zone2 zone3 zone4 Furnace Set Temperature
17480.0 12717.2 4335.2 4474.4
b) April 2009: combustion air @ 370°C 1130°C 1210°C 1220°C

850°C 1050°C
700°C 1200°C
The temperature of a billet was monitored 950°C gas absorption coefficient=0

during monitoring period by 30 meters long five K Billets (140x140mm)

type thermocouples (figure 3). In both tests the

billet rotated at baffle zone (180° and 270° in (Tin=20°C) Reference billet for post-procesing Adiabatic condition out

respectively), therefore the thermocouple fixed in

the upper side moved to lower and lateral side of Figure 3. 2D Fluent furnace model set temperature
the billet. The billet rotation phenomena is well
Simulations has been performed in case of 80 tph
know in Feralpi and results from the strong
nominal furnace production both representing
thermal gradients in correspondence of the baffle
directly the billets movement and using the
zone in the furnace (see results of CFD simulation
equivalent strip concept based on similitude
in figure 12, 16). Recorded temperatures are
criteria already defined in [7]:
reported in figures 8, 10.
1. equal production:
m& strip = ρ steel ⋅ S strip ⋅ vstrip = furnace production
where
Sstrip is the strip transversal section
vstrip is the equivalent strip velocity

2. equal radiant flux area:

∑ (S billet ⋅ Fbillet ) = k ⋅ S up − strip ⋅ Fstrip
billets
Figure 3. Thermocouple scheme for instrumented where
billet in Feralpi furnace Sstrip is the upper strip surface
F is the surfaces view factors
The furnace heat balance (table 2) reports the
k is the equivalent strip radiant factor equal to
calculated furnace efficiency.
∑ ( S billet ⋅ Fbillet ) / S up−strip ⋅ Fstrip
billets

3
 obtain the profiles for representing a specific
3. equal thermal resistivity: furnace operative conditions by interpolating
k strip k billet between those available in a DB for the simulation
= of a single burner. Moreover the profile can be
δ strip δ billet
rotate and translate in the position required as
where boundary conditions to the complete furnace
kstrip is the thermal conductivity of the equivalent domain.
strip
δstrip is the thickness of the equivalent strip
kbilett is the thermal conductivity of the billet
δbillet is the thickness of the billet

In figure 4 the heating curves obtained with the

two approaches are compared demonstrating that
the equivalence strip reproduces the billet heating
curve with very good accuracy.
1400.00

1300.00
T mid - 2D billet

T up - 2D billet
ZONE 1 ZONE 2 ZONE 3-4
Figure 5. Burners schematization for domain
1200.00

1100.00
T mid - 2D strip eq.

T up - 2D strip eq.
optimization
1000.00

In the specific case the two size of roof

NOSE

900.00

800.00
TUNNEL
burners, 750kW in preheating and heating zone
y [m]

700.00

600.00

500.00 and 400kW in soaking zone, have been simulated

400.00

300.00
as installed into the a reference furnace (CSM roof
200.00 burner test facility) for the operative condition of
100.00

0.00
Table3.
0.00 5.00 10.00 15.00 20.00
temperature [°C]

Table 3. Operative condition for burner simulation

Figure 4. Two dimensional model: billet model vs.
equivalent strip Tair 300 oC 450 oC 550 oC
1% 1% 1%
O2
2% 2% 2%
(DFG)
3% 3% 3%
V. OPTIMIZATION OF THREE 20% 20% 20%
DIMENSIONAL COMPUTATIONAL DOMAIN Power 50% 50% 50%
100% 100% 100%
The representation of a complete reheating
furnace, based on grid sensibility analysis
Figure 6 shows an example of the simulation
performed for single zone, requests a number of
results for 750kW burner.
computational cells in the order of tens of million °C

cells [7]. The run time for such grid dimension,

using present CSM computational resource (cluster
of four Dell 2 CPU dual core cluster with 32 GB
RAM) results in several weeks. Power 100% Power 50% Power 20%
Since, as stated previously the main part of the
cells are concentrated in the burners region, a
procedure has been set–up to avoid detailed
burners representation. The profiles extracted at
the burner outlet from a dedicated simulation of Figure 6. Temperature contour (°C) for 750kW
single burner are applied as boundary condition in burner. Tair= 450°C – 2%O2 (DFG)
the complete furnace model. The procedure has
been described and validated in [7] and it is used Using this approach and exploiting the symmetry
also in other works [4]. to represent the Feralpi furnace only 10M cells
To facilitate the application of this approach in have been necessary. A non structured zone is
case on multiple burners a dedicated software has used in front of every burner while in the remain
been developed [7]. Therefore, it is possible to

4
part of the domain structured mesh guarantees the accuracy and CPU time to simulate the NG
better resolution close to stock. combustion process [10,11] when NOx or CO are
not the target.
VI. CFD MODELLING SET UP FOR THE Table 4. CFD modelling set up
FERALPI FURNACE SIMULATION CFD model set-up
Fluid Ideal gas
The CFD simulation has been performed by the Turbulence k-ω with shear stress correction (Wilcox)

commercial code ANSYS FLUENT release 6.3 Chemistry 2 steps reduced mechanism:
CH4+1.5O2 = CO+2H2O
and 12.0 [9]. The code allows the parallel CO+O2 = CO2
Rate of reaction: Eddy Dissipation/ Arrhenius *
calculation by means of MPI domain A=4 & B=0.5
decomposition. The ANSYS FLUENT release Material properties For each species
12.0 increases the parallel efficiency respect to 6.3, - Specific heat for each species taking into account dissociation (K.
Pelders)
- Thermal conductivity and viscosity by kinetic theory
speeding up the calculation of about 10%. The Mixture
- Density ideal gas
total CPU time using the present CSM - Specific heat, viscosity and thermal conductivity by ideal gas mixing
law
computational resource is 120 hours. - Mass Diffusivity by kinetic theory

The boundary conditions of the complete Radiation DOM (Discrete Ordinate Model – Domain based)
furnace CFD model are the following (Figure 7): Absorption coefficient: WSGGM

Numerical approach
1. furnace walls: mixed (radiative and
Mesh type Hybrid
convective heat transfer) wall boundary condition
Solver Segregated
with material conductivity and thickness for lateral Discretization Implicit/finite volume
wall, roof, bottom and tunnel; Non linear term Second order scheme
2. burners: velocity inlets with profile of
velocity components, turbulence, temperature and
species concentrations extracted by single burner VII. COMPLETE FURNACE SIMULATION
simulations; WITH EQUIVALENT STRIP MODEL (ESRF)
Wall holes:
wall transparent to
wall holes: wall trasparent to radiation
radiation
Flue gas The simulation of the complete Feralpi furnace
exit
with stock represented by equivalent strip
TUNNE
L approach (ESRF) has been carried out in the frame
charging:
wall trasparent to radiation
of RFCS CO2RED project. Herein the main
burner
s
Discharging
results are summarised while detail has been
discharging system: fixed T (100°C)
system:
Steel billet cha
rge fixed T(100°C) already reported in [7].
The setting for each furnace zones correspond
to January data in table 3a. Each pair of adjacent
roof burners are considered with contra-rotating
Figure 7. Boundary conditions for furnace CFD swirl. The heating curve obtained by the
model simulation has been compared with billet core
temperature measured by thermocouples (figure
3. flue gas exit: pressure outlet; 8).
4. charging port: transparent "window" 1400
T lat right - exp ZONE 3-4
ZONE 1 ZONE 2
ZONE 1 ZONE 2 ZONE 3-4
exchanging radiation heat flux with outside 1200
T lat left - exp
T lat - mid - CFD

ambient at 25°C; T lat - up - CFD

CFD up
(°C)

1000
NOSE

5. furnace wall holes: trasparent "window" TUNNEL

TUNNEL
temperature [°C]
Temperature

exchanging radiation heat flux with outside 800

ambient at 25°C 600

CFD mid
6. the discharge handling roll system: fixed 400

temperature surface (T=100°C) at discharge 200

TDX TSX
position.
0
0 2 4 6 8 10 12 14 16 18 20 22
x [m]
The modelling set-up for the simulation of Position along the furnace (m)
Figure 8. Feralpi WHF heating curve, exp vs. CFD
single natural gas (NG) burner (Table 4) [7,8] has
data (January conditions)
been adopted. In particular the two steps reduced
mechanism with Westbrook&Dryer kinetic has The difference between calculated and
been selected being the best compromise between experimental data in the nose zone can be

5
explained by the rotation of the billet observed 10), demonstrate that the CFD model is able to
during the measurement (figure 9) due to strong reproduce the different discharging temperatures
thermal gradient (figure 15). observed experimentally

VIII. COMPLETE FURNACE SIMULATION

WITH COUPLED MODEL (CMRF)

The "equivalent strip" approach has allowed to

develop and test the CFD “furnace model” for
complete 3D geometry, however the intrinsic
Figure 9. Billet rotation at baffle zone (the rotating limitation to the application to rotating handling
billet edge is circle out in red) system has required its extension.
The “stock model” has been developed using
The calculate furnace heat balance, reported in both ANSYS FLUENT User-Defined-Function
table 5, shows a good agreement with that obtained (UDF) and macros. The model implements a
during the furnace monitoring (table 2 – January process that iteratively performs the stationary
09) simulation of the fluid dynamic inside the furnace
Table 5. Feralpi WHF heat balance and the time dependent simulation of the solid
body (billet in Feralpi case) advancement inside
PRO TTPI ST STG FGH FHL η the furnace (figure 11).
[tph] [kW] [°C] [kW] [kW] [kW] [%] First process iteration
Furnace CFD simulation with billets surface temperature

Heat flux on the billets

121 40102 1175 26378 11801 1923 65 fixed on the basis of expected discharge temperature

Iterative process
Moreover from CFD results the main contribution Heat flux on the billets

to heat losses can be discriminated (table 6). Furnace CFD simulation with billets (1) (2) Billets solid body simulation with
surface temperature defined by solid
heat flux defined by CFD heat flux
body simulation results
for every position
Table 6: Furnace heat losses Te
m
pe
ra
Test 31/01/09 Heat losses tu
re
on Convergence check
th
e
bil Tbill_it-Tbill_it-1<0.5°C
Furnace wall 351.4 kW le
ts
Tbill_it-Tbill_it-1>0.5°C

Discharging roll way 1177.0 kW

Output data
Charging port 219.5 kW •Furnace thermo fluid dynamic field
•Billets solid temperature
Wall holes 107.4 kW
Figure 11. Coupled model scheme
The CFD model has been also applied to the
simulation of the operative condition The solid body time dependent calculation (2)
corresponding to the field test conducted in April is performed by ANSYS FLUENT, too. The
2009 (table 3b). computational domain represents a single billet
and the heat flux from (1) is set as boundary
1400
T - exp 01/09 ZONE 1 ZONE 2 ZONE 3-4
condition at billet surface. The billet advances in
1200 T - CFD 01/09 time and finds at every position the corresponding
T mid - CFD
1000 T - exp 04/09 heat flux calculated by “furnace model”. The
temperature [°C]

BAFFLE

800
ANSYS FLUENT solves only the energy equation
TUNNEL

600
(the momentum equations is deactivated) for each
400
time step. The billet surface temperature is written
in the Fluent profile file at every position and it is
200
passed back to the Fluent “furnace model” (1).
0
0 2 4 6 8 10 12 14 16 18 20 22 The furnace model, updated with new billets
x [m]
temperature, runs for a fixed number of iteration.
Then heat flux on billets is extracted again and
Figure 10. Heating curve (billet mid point) 01/2009
passed to the solid bodies calculation (2). The
vs. 04/2009 test case
convergence was obtained performing the solid
The resulting heating curve, plotted versus the calculation every 25 iterations of the furnace CFD
one corresponding to January 2009 case (Figure model. The iterative process is numerically robust

6
and the convergence has been obtained after 350
furnace model iteration stating from the first step billets
(tentative solution) that can be obtained running a
first CFD furnace model step setting as boundary
°C
condition reasonable billets surface temperature.
The billet heating curve can be obtained from
experimental data, such as in the case of Feralpi
furnace, or running zonal model or through the
simplified two dimensional model.
The coupled model has been applied for the
Feralpi furnace with operative condition of April equivalent strip
2009 (table 3b). The boundary conditions setting is
the same of the equivalent strip case reported in
the previous paragraph.

billets

Figure 13. Temperature on a furnace section.

°C
billets

m/s

equivalent strip

equivalent strip

Figure 12. Billets vs. eq. strip surface temperature

The difference is the steel stock representation:

each billet has been represented in its position. The
simulation results are qualitatively similar to the
equivalent strip model ones, as shown by Figure 14. Velocity on a furnace section
temperature comparison in Figure 12. The flow
field in the near billets region is also very similar. The heat flux on the stock, figure 15 is not
In a zoom close to the billet (figures 13,14) affect by the stock model assumption confirming
temperature and velocity fields are plotted on a that the similitude criteria for the radiation
furnace section in the near baffle zone. parameter (view factor and emissivity, see
paragraph IV) is correct.

7
 240 1400
ZONE 1 ZONE 2 ZONE 3-4
Billets T mid - exp
210 1200 T up - exp
Eq. strip

B AFFLE
T mid - CFD
[kW/m2]

180 1000 T up - CFD

te m p e ra tu re [°C ]
heat flux [kW]

150 800
TUNNEL
120
600
90
400
60
200
30

0 0
0 2 4 6 8 10 12 14 16 18 20 22
0 2 4 6 8 10 12 14 16 18 20 22
x [m]
x[m]
Figure17. Heating curve:
Figure15. Heat flux on the stock upper surface simulated (coupled model) vs. experimental data
(billets vs. equivalent strip) 1400
T mid - bill ZONE 1 ZONE 2 ZONE 3-4
1200
In Figure 16 the heat flux for the single billet in T up - bill
T mid - eq.strip
the baffle region is plotted putting in evidence that:

B A FFLE
1000

tem perature [°C ]

T up - eq.strip

the roof burners produce strong non 800
uniformity; 600
TUNNEL

the heat flux are non uniform between

400
upper and lateral billet surface

200

kW/m 2 0
0 2 4 6 8 10 12 14 16 18 20 22
x [m]
Figure18. Heating curve:
coupled model vs. equivalent strip model

The main difference between calculated and

measured data still remain in the zone of baffle
where experimental results are affected by the
billet rotation due to the strong heat flux non-
uniformity. Since the coupled model allows to
introduce the rotation of the billet in any point of
the furnace domain it has been possible to
reproduce the observation during the test, as sum
of two rotations imposed to the billet (180°
rotation at 8.9 m, 90° at 9.7m). The resulting
flowfield inside the furnace does not qualitatively
∆ heat flux (upper-lateral)= 150kW/m2 differ from the previous case while the effect of
the billet rotation is evident on the surface
temperature (figure 19). Taking into account
Figure16. Heat flux in the near baffle zone rotation effect calculated heating curve (figure 20)
is very close to the measured one, reproducing the
localised temperature discontinuity in the baffle
The simulated billet heating curve is plotted in zone.
figure 17 versus experimental data and in figure 18
°C
versus equivalent strip data. The coupled model 1° rotation (180°)
reproduces the experimental curve slope better
2° rotation (90°)
than the equivalence strip model in the tunnel-
baffle zone and in the soaking zone. The
temperature delta between center and upper billet
point at discharge is 35°C respect to the 44°C of
the equivalent strip simulation.

Figure 19. Billets surface temperature

8
 1400
T mid - exp ZONE 1 ZONE 2 ZONE 3-4
This is due to the furnace high production (121 –
1200 T up - exp
126 tph) during the test respect the nominal
production of this furnace (80 tph).

B A F F LE
T mid - CFD
1000
tem peratu re [°C ]

T up - CFD

800 T (°C)
TUNNEL
up
Tflue gas: 745 °C (exp. 740 - 753°C)
600 mid

400
Velocity
(m/s)
up mid
200

0
0 2 4 6 8 10 12 14 16 18 20 22
x [m]
O2 (%)
a) complete heating curve O2 flue gas: 0.7%
Zone 1 Zone 2 Zone 3/4

up 9.59
Air/gas ratio 9.89 10.39
mid
Figure 21. Temperature, velocity and oxygen
concentration iso-line on furnace symmetry plane
up mid

T mid - exp
T up - exp IX. SCALE GROWTH POST PROCESSING
T mid - CFD
T up - CFD
The temperature, velocity and species
4.00 6.00 8.00 10.00 12.00 14.00 concentrations (T, v, Xi) 3D field on the billets
b) detail close to buffle zone surface obtained with the CFD simulation allow to
perform calculation of scale growth on the billets.
Figure 20. Simulated vs. experimental heating curve The CSM model [12], that considers the scale
with billet rotations imposed growth in term of mass transport of oxidizing
species from the bulk of the gas atmosphere to the
The reduction of temperature in zone 2 is steel surface and diffusion of reacting species
peculiar of this data set and it is not reproduced by inside the formed scale, figure 22, has been
model. One possible explanation can be the lateral implemented as post-processing in ANSYS
door opening for long time during the FLUENT by User-Defined-Function (UDF).
instrumented billet to allow taking movie of the
billet rotation. O2 2Fe+O2=2FeO
Scale:
In general the simulation results allows to FeO - wustite
CO2
2FeO+1/2O2=Fe2O3
H2O
underlain/justify the main phenomenon taking Fe2O3 - hematite H2 Fe+CO2=FeO+CO
Fe3O4 - magnetite Fe+H2O=FeO+H2
place inside the furnace:

the baffle produce two separate macro Fe++
zones in the furnace both in term of velocity (low
velocity with big “horizontal” recirculation in the Figure 22. Scale growth controlling phenomena.
tunnel, high velocity with “vertical” recirculation
due to swirled roof burner), temperature and gas In Figure 23, the scale growth curve on the central
composition (figure 21); point of the upper billet surface is plotted.

zone 3 and 4 is under-stoichiometric in
agreement with the air/gas ratio adopted (table 1) 450
while the oxygen in the flue gas is 0.8% (O2 in 400
scale thickness (micron)

350
figure 21);
300

the non uniform transversal distribution 250
of the heat flux and temperature on the strip is 200

generated by “in line” burners arrangement on the 150

100
roof of the furnace ; 50

the temperature difference between the 0
0 2.5 5 7.5 10 12.5 15 17.5 20 22.5
core and the billet surface indicates that the stock
billet position (m)
temperature is not well equalized. Figure 23. Scale growth curve on the billet upper
center point

9
 The coupled model has been validated in case
The scale thickness field on the billet surface is of rotating handling system simulating the
shown in Figure 24 for two different billet localized billet rotation in the baffle region
position. The zoom shows the scale thickness of observed during the furnace monitoring. Taking
the billet at discharging position. into account this specific behaviour of the Feralpi
furnace the comparison of the experimental and
m/s
°C
simulated heating curve shows very good
agreement.
Comparing the results obtained with the ESRF
and CMRF models the similitude criteria of
equivalent strip approach has been verified,
showing its applicability to a non rotating handling
µm system.
Moreover being the stock surface temperature
and local concentration of combustion products
know by CFD, the CSM scale growth model has
been implemented as post-processing option, by
FLUENT Ansys UDF in the furnace model.
The good results of validation work and the
a) Scale thickness at billet #73 and #110 required CPU time, with present CSM HW
µm capability, indicates that CFD furnace coupled
model is ready to be applicable to the simulation
of reheating furnace where it is necessary the
analysis of fluid dynamic, flames interaction,
detailed representation of furnace wall and steel
stock temperature and heat fluxes.

XI. REFERENCES
[1] R,J. Tucker et others: British-Franch Flame Day, Lille
9/10 March 2009
[2] L.Ballarino et others: 7th International Symposium on
High Temperature Air Combustion and Gassification, Phuket,
Thailand, January, 2008
b) Scale thickness at discharging position
[3] Sang Heon Han et others: International Journal of Heat
Figure 24. Scale thickness field on billets and Mass Tranfer 52 (2009) 1005-1011
[4] P. Pronk: AiSTech 2006 Proceedings – Volume I

X. CONCLUSION [5] A. Jaclic et others: Materiali in Tehnologije 39 (2005),

ISSN 1580-2949
The CFD modelling developed for single [6] Bin Wu et others, AISTech 2009 Proceedings -
burners simulation has been extended to complete Volume II
reheating furnace including stock movement. A [7] Battaglia et others: 16th IFRF Members’ Conference,
specific procedure has been applied to reduce the Boston, USA, June 08-10, 2009.
extension of the computational grid without
[8] E. Malfa et others 7th High Temperature Air Combustion
loosing results accuracy. and Gasification International Symposium, Phuket, Thailand,
The first version of the furnace model (ESRF) January 2008
has been implemented in the frame of RFCS
[9] ANSYS FLUENT USER’S MANUAL. Fluent Inc. 2009
CORED project representing the steel stock by
equivalent strip concept. This model has been the [10] C. Westbrook and F. Dryer..Progress Energy
Combustion Science. 10:1-57, 1984
base for implementing an up-dated version
(CMRF) that extends the tool applicability to all [11] M. Mancini: Ph.D. Thesis. Clausthal University of
type of stock geometry and handling system. Technology
Both models have been tested on the 100tph [12] C. Borgianni and others. ECSC agreement nr.7210-EA-
Feralpi walking heart reheating furnace. 407 (1980)

10

View publication stats