Accepted Manuscript

New methodology for CFD three-dimensional simulation of a walking beam type

reheating furnace in steady state
Jos Manuel Casal, Jacobo Porteiro, Jos Lus Mguez, Alfonso Vzquez
PII:

S1359-4311(15)00351-8

DOI:

10.1016/j.applthermaleng.2015.04.020

Reference:

ATE 6538

To appear in:

Applied Thermal Engineering

Received Date: 30 December 2014

Revised Date:

15 March 2015

Accepted Date: 4 April 2015

Please cite this article as: J.M. Casal, J. Porteiro, J.L. Mguez, A. Vzquez, New methodology for CFD
three-dimensional simulation of a walking beam type reheating furnace in steady state, Applied Thermal
Engineering (2015), doi: 10.1016/j.applthermaleng.2015.04.020.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

NEW METHODOLOGY FOR CFD THREE-DIMENSIONAL SIMULATION OF A

WALKING BEAM TYPE REHEATING FURNACE IN STEADY STATE
Jos Manuel Casal a*, Jacobo Porteiro a, Jos Lus Mguez a, Alfonso Vzquez b
a

Industrial Engineering School, University of Vigo, Lagoas-Marcosende, s/n 36310 Vigo,

Spain.
AIMEN, Relva, 27 A. Torneiros - 36410 Porrio (Pontevedra- Espaa).

RI
PT

* Corresponding author. Telf: +34986818624. E-mail address: [email protected]

Abstract

SC

This study presents a new modelling methodology for three-dimensional simulation of a

reheating furnace of steel billets with natural gas burners. The movement of the billets inside

M
AN
U

the furnace is a periodically transient phenomenon. A modification in the energy transport

equation is introduced through source terms in the billet region to convert the transient
movements of the billets into a steady-state simulation, which significantly reduces the
computational time without any loss of information.

The combustion simulation inside the burners is performed using a global combustion
mechanism that takes into account the combustion of the main gaseous species present in

TE
D

natural gas following a kinetic and turbulent control of the reaction through the Eddy
Dissipation Concept (EDC).

The results are contrasted and validated with the data obtained in a real facility though the

EP

SCADA of the plant. Some parameters, such as the surface temperatures of billets at the outlet
and the distribution of power supplied by the fuel inside the furnace are reasonably close to

AC
C

the real case data obtained in temperature fields of both billets and the gas phase as well as
the energy balance, leading to the conclusion that the proposed methodology is adequate for
the purpose of simulating this system.
Keywords: Reheating furnace, billets, computational fluid dynamics, steady state, natural gas
combustion.

Nomenclature




Arrhenius pre-exponential factor

specific heat (Jkg-1K-1)

RI
PT

SC

M
AN
U

absorption coefficient (m-1)

temperature exponent
stoichiometric coefficient for reactant  in reaction  (-)
stoichiometric coefficient for product  in reaction  (-)
turbulent dissipation (m2s-3)
wall emissivity (-)
rate exponent for reactant species  in reaction  (-)
rate exponent for product species  in reaction  (-)
molecular viscosity (kgm-1s-1)
turbulent viscosity (kgm-1s-1)
kinematic viscosity (m2s-1)
density (kgm-3)
Stefan-Boltzmann coefficient (Wm-2K-4)
scattering coefficient (m-1)t
viscous stress tensor (Nm-2)
scattering function (-)
residence time of billet in a position (s)
solid angle (sr)

AC
C

'
(
*
) ,
**
) ,
,
,
*
- ,
**
- ,
.
./
0
1
2
345/
2
6

TE
D

Greek

centroid coordinates in current cell (m)

centroid coordinates in previous cell (m)
molar concentration of species j in reaction (kmol m-3)
activation energy (Jkmol-1)
specific enthalpy (kJkg-1)
irradiation intensity (Wm-2sr)
diffusion flux of species  (kgm-2s-1)
turbulent kinetic energy (m2s-2)
forward reaction rate
backward reaction rate
effective conductivity (W.K-1m-1)
conductivity (W.K-1m-1)
equilibrium constant (-)
molecular weight of the species  (kgkmol-1)
refractive index (-)
pressure (Pa)
position vector (m)
position vector of the wall boundary (m)
gas constant (JK-1mol-1)
mass rate of generation of species  (kgm-3s-1)
molar rate of species  (kmolm-3s-1)
direction vector (m)
scattering direction vector (-)
Temperature (K)
velocity component (ms-1)

EP

,,


,,
 ,





,
,



,




!
!
!" ,
#
#
%
&

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
1. INTRODUCTION.
The Spanish steel industry is the third-largest steel producer in Europe, making this industrial
sector one of the largest consumers of energy resources. The main energy consumption in this
sector comes from non-renewable resources. One of the greatest challenges, therefore,
involves the optimisation of energy resources for this sector, which is associated with a

reduction in emissions of greenhouse gases.

RI
PT

reduction production in costs and beneficial side-effects on the environment, such as, the

One of the largest consumers of energy in the steel industry is the reheating furnace. These
furnaces are used to raise the temperature of the steel introduced into them in billet form, to

SC

facilitate the subsequently processing of the steel in rolling mills. The process of heating the
billets requires two specific thermal conditions a minimum temperature and a homogeneous

M
AN
U

temperature in the billet. Of the various types of reheating furnaces, in the present work, a
walking beam furnace with several top fuel burners is studied. The main heating mechanism of
the billets is due to thermal radiation generated by the high temperature flames and the
surfaces of the wall that form the enclosure of the furnace [1]. To a lesser extent, another
heating mechanism of the billets, consisting of convection heating with the combustion gases,
is also at work. The knowledge of the behaviour of the heating mechanism might be

TE
D

interesting from the standpoint of the improved energy efficiency in this type of furnaces.
Computational Fluid Dynamics (CFD) can be a useful tool to study heat-transfer mechanisms
and to make improvements in the energy efficiency of equipment, without the need to test the
installation and stop or jeopardise the production process. To this end, it is necessary to

EP

validate a complete simulation model of the furnace with a real case. There is much work
described in scientific literature related to the CFD modelling of the billet heating process in

AC
C

reheating furnaces. Among these models, perhaps the first published studies are the simplified
model of Zhang et al. [2, 3], in which the furnace geometry was simplified and the billets were
treated as a single slab of steel placed on the floor of the furnace. This concept has been
applied recently by Huang et al. [4] who modelled a reheating furnace in which the billets were
modelled in a steady-state as a continuous sheet that crosses the furnace, and the mass and
energy transport in the sheet was modelled through a high-viscosity laminar flow.
Other more complex three-dimensional models, such as that reported by Kim et al. [5]
considered combustion in turbulent flows and radiative heat transfer to predict, in a steady
state, the temperature distribution inside the reheating furnaces and the heat flux through the
surfaces of the billets in which the surface temperature was fixed based on experimental data.
3

ACCEPTED MANUSCRIPT
Another study by Kim and Huh [6] has related a steady-state analysis of the heat transfer at
the boundary of the billets and has calculated the temperature profile within the billets, using
the finite difference solver method. Hsieh et al [7] calculated the combustion in turbulence
flows in the same manner as Kim and Huh applied in a more complex furnace geometry. There
are other simplified models, such as that of Kim [8], in which the furnace was divided in zones

RI
PT

with different wall and gas temperatures. In these models, the furnace was modelled as a
radiating medium and the temperature field of the billets was evaluated with a transient heat
conduction equation. Two of the most recent works in this area were published by Han et al.
[9, 10], in which a detailed 3D transient simulation was performed.

SC

In these works, the movement of billets was simulated to reach a periodically transient
solution, moving the field of temperature of the previous position to the next position at the
end of the residence time. A different approach in the reheating furnace field, was used by

M
AN
U

Martin et al. [11], where a low computational-cost model was performed. In their work, during
the pre-processing phase, a database was created with the stored information about the
behaviour of the physical properties of the system. The information in this database was used
to predict the behaviour of the system for a random selection of furnace operating
parameters.

TE
D

As describe above, there is a large body of work on the simulation of reheating furnaces; some
are simplified models focusing on the steady-state with a consequent lack of accuracy, while
others are more complex models solved for the transient state, which leads to better
resolution and more detailed results but at a higher computational cost. Generally, to achieve

EP

a reasonable computational time, transient simulations have to implement further

simplification of the geometry and the size of the mesh. In the present work, a different

AC
C

strategy is proposed to solve the problem mentioned by Hang et al [9], by performing a

steady-state calculation that converges to the periodic solution and avoids the lengthy
calculations of transient simulations. Among recent works, it is worth to mention the studies
which were focussed on the determination of the transient temperature fields and the heat
flux through the surfaces of the billets( Tan et al [12], Emadi et al [13], Li et al [14]) . In the first
two, mathematical models based on the "zone" method of radiation analysis were used to
simulate a transient furnace behaviour in production. In the last work the transient
temperature field was determined using the finite volume method and the alternating
direction implicit algorithm.

ACCEPTED MANUSCRIPT
The proposed solution in this paper consists in modification of the energy transport equation
of the CFD code ANSYS Fluent using and ad-hoc routine programmed as User-Defined Function
(UDF) to take into account the energy transported in the movement of the slabs from one
stage to the next. The simulation results allow us to obtain an energy balance for the whole
furnace, as well as the field of temperature in the furnace and the temperature field of the

RI
PT

billets. The simulation results are contrasted with real data collected by the control system of
the industrial facility. This work demonstrates that this methodology is adequate for the
purpose of simulating this kind of system.

SC

2. FURNACE DESCRIPTION

In this study, a steel reheating furnace is simulated using the commercial software FLUENT.

M
AN
U

The furnace simulated in this work is a walking-beam reheating furnace that use steel billets to
produce steel wire rods. The nominal power of the furnace is 23.89 MW and the approximate
dimensions are: length 20.9 [m], amplitude 13 [m] and height: 12 [m]. Inside the furnace, four
zones are distinguished: Preheating, Heating, Equalisation Right, Equalisation Left as shown in
the Fig. 1. The billets are heated by 56 vertical burners placed on the top of the furnace. These
56 burners are divided into: 16 burners in the Preheating zone, 24 in the Heating zone, 8 in the

TE
D

Equalisation Right zone and 8 Equalisation Left zone. The fuel used for the burners is natural
gas. The fuel and air mass flow of the burners in each zone is different, but the same for all the
burners in the zone. The production capacity of the furnace is approximately 90 t/h. The steel
is introduced in the furnace in the form of billets whose dimensions are 160 mm x 160 mm x

EP

12000 mm and the weight of a billet is approximately 2.4 Ton. The space between the billets

AC
C

inside the furnace is 90 mm and the number of billets that fit in the furnace is 75.
Figure. 1 (column fitting image 2)

As detailed in Fig. 1 the billets are placed in the furnace through a side gate using a roller
mechanism on which the billets slide, then the billets are placed over the floor of the furnace
using a pusher. The extraction process also using rollers and a mechanism that positions the
billets on the rollers. All of this apparatus is cooled by water to preserve the mechanical
properties of the materials. As shown in Fig. 2, the extraction mechanism has a window on the
rear of the furnace required for the movement of the mechanism. To prevent furnace gases
from entering the machinery room housing the walking beam system, there is a sealing
mechanism between the floor slabs of the furnace. The seal is formed by a water film that is
constantly recirculated.
5

ACCEPTED MANUSCRIPT
The exhaust gases are led passed through a heat exchanger to recover energy for combustion
air pre-heating. The inlet temperature of the preheated air in the burners is 300 0C. The gas
flow inside the furnace is measured with two thermocouples per zone. The flue gas
temperature is measured with two thermocouples before and after the heat exchanger. There
is also a pyrometer for measuring the temperature of the billet at the furnace outlet. Each

spent of each billet at each position in the furnace is 97 seconds.

Figure. 2 (column fitting image 1.5)

SC

3. NUMERICAL SIMULATION

RI
PT

billet takes 7275 seconds (approximately 2 h) to pass through the whole furnace, and the time

The movement of the billets in the furnace is a periodic transient phenomenon. The main

M
AN
U

feature of the model employed in this work is the conversion of a periodic transient
phenomenon to a steady-state solution.

The furnace will be solved under the hypothesis that it has reached a steady state in which the
temperature of the billet at the outlet is steady, and therefore the furnace has reached it is
stable, nominal operation therefore no transient terms, apart from those corresponding to the

3.1 Flow and turbulence.

TE
D

movement of the billets are required.

EP

The flow in the furnace is a mixture of gases from the combustion of natural gas, which takes
place in the burners. This mixture is modelled in a steady state as single-phase incompressible

AC
C

Newtonian fluid and is governed by Reynolds average Navier-Stokes (RANS) equations (1-4).
Continuity:

:
<& = = 0
:;

(1)

Momentum:
1

:
:
:
+
D2.F 1& * & * G
@& & A =
:;
:; :;

(2)

ACCEPTED MANUSCRIPT
where Sij is the mean strain-rate tensor:
1 :& :&
F = I
+
J
2 :;
:;

(3)

The 1& * & * term of the equation (2) (Reynold Stress Tensor) is modelled using the
Boussinesq (4) hypothesis in the development of turbulence modelling.
./
F
1

RI
PT

1& * & * = 2

(4)

SC

The turbulence is modelled using the method proposed by Shih et al.[15] using realizable k- as
shown in equations (5) and (6).
Turbulence:

(5)

:
:
./ :,
,
T
KL. + N
O + 1 FQ 1R
+  Q U Q P
@1,& A =
:;
:;
2Q :;
 + 0,


(6)

TE
D

M
AN
U

:
:
./ :
:&
KL. + N
O 1& * & *
+ P 1,
@1& A =
:;
:;
2M :;
:;

where P is production of turbulent kinetic energy due to buoyancy and

]

 = max Y0.43, ]^_` ,

M

c =

EP

./ = 1c a ,

- = F , F = b2F F

d ^e

fg
i

AC
C

The values of the model constants used in this work are  Q = 1.44 , R = 1.9,
2M = 1.0 , 2Q = 1.2

To resolve the viscous sub-layer of the near-wall region enhanced wall treatment
(EWT) is used in combination with realizable k-.

3.2 Combustion of the natural gas.

In this work, combustion is modelled by The Eddy Dissipation Concept Model (EDC), developed
by Magnussen [16]. This model is an extension of the Eddy Dissipation Model (EDM) that uses
7

ACCEPTED MANUSCRIPT
the Arrhenius rate and can be used with detailed chemical mechanisms in turbulent flows. This
model assumes that reactions take place at the small turbulent scale at which a small reactor,
which follows the chemical mechanisms proposed, is solved.

Natural gas combustion is modelled through the kinetic mechanism developed by Jones and

RI
PT

Lindstedt [17]. This multi-step mechanism is developed for the natural gas combustion with
air, and it is comprises four reactions. As shown in the (Table 1) this mechanism is formed by
two irreversible reactions of hydrocarbons dissociation with O2 and H2O (JL1 and JL2) and two
reversible reactions, namely, of Hydrogen oxidation (JL3) and finally the water gas shift
reaction (JL4). The fuel used in the furnace is a commercial natural gas and the composition is

M
AN
U

depending on the number of carbons in the hydrocarbon.

SC

detailed in Table 2. As shown in Table 1 each compound of natural gas has its own kinetic

These reactions take place over a time scale and are governed by the Arrhenius rate as shown
in equations (7) and (8). In the irreversible reactions (the hydrocarbons dissociation with O2
and H2O), the molar rate of creation/destruction is given by equation (7), while in the
reversible reactions (hydrogen oxidation and the water gas shift reaction) the molar rate is
given by (8). The mass rate is evaluated by equation (9).
, )

, A k , lm , n

q
qq
D]o,p
^ ]o,p
G

TE
D

!" , = @)

**

s
r

AC
C

EP

!" , = @) , ) , A k , lm , n
rw

! = , v !" ,

q
]o,p

(7)

 , lm , n
s

qq
uo,p

(8)

(9)

s

The forward and backward reaction rate constant is calculated by equations (10) and
(11).
, =
 %xp exp L


N
!%

(10)

, =

ACCEPTED MANUSCRIPT

,


(11)
Table 1

3.3 Conservation of energy.

RI
PT

The heat transfer in the steady state in the fluid region of the furnace is governed by the
general equation of energy transport (12, 13 and 14), where the source term (14) is composed
of three terms: the term due to radiation, the term due to convective heat transfer with the
solid region and the term due to the heat reaction of the chemical species. The terms on the

SC

right-hand side of the equation (12) represent energy transfer due to conduction, chemical
species diffusion, and viscous dissipation, respectively. For a solid region, the terms of species

M
AN
U

diffusion and viscous dissipation are null in addition to the term on the left-hand side
(advective term).

 &R
 = +
1 2

Ik  T v  + D6 &

|G J

TE
D

@&
| <1 + =A


F

(12)

(13)

(14)

EP


F
= F5 + F4 + F4

The charge and discharge process of the billets in the furnace occurs periodically at fixed
intervals of time, therefore the movement of the billets can be described as transient periodic

AC
C

phenomenon. If we consider the movement of the billets as a flow, it could not be considered
steady. In this work, a CFD code has been developed to allow the constant flow of discrete
solid parts that move at a constant speed from one position to another. To represent
numerically their continuous movement, the periodic transient movement of the billet has
been transformed into a steady-state operation.
To solve the problem of generating movement in solid regions in a steady state framework, the
movement of the billets is performed virtually. For this purpose, we kept the position of the
billets in the furnace fixed, and the thermal energy contained in a billet is moved from one
position to the next. This is analogous to subtracting the energy balance between the energy

ACCEPTED MANUSCRIPT
due to transport from the previous position to the current position and the energy contained
in a billet. As shown in Fig. 3, the energy subtracted in every current cell is the energy due to
transport in the analogous cell in the previous position minus the energy due to transport in
the cell in the current position.
Figure. 3 (column fitting image 1.5)

RI
PT

The developed algorithm modifies the general equation of energy transport in the solid region
of the billets while in the rest of the solid regions of the furnace and the fluid regions the
energy equation remains unchanged.

In ANSYS Fluent [18-20] solid flow is not allowed in solid regions such that the advective term

SC

is null. For this reason the energy transport equation only has diffusive and source terms (15).
In the equation (15), the source term is given by three contributions: radiation, convective

of the billets (16).

<K T=


F3

Diffusive term

M
AN
U

heat transfer with the fluid region and the new term introduced to account for the movement

(15)

Source term

(16)

TE
D


F3
= F5 + F4 + F5

Hence, to introduce the thermal transport generated by the movement of billets, an advective
term in a solid region of billets is created through an equal source but with opposite sign (17).

EP

F5 = <1&
|=

(17)

AC
C

If we consider an infinitesimal element with sides  ,  and  and apply the <1&
|=, as

shown in the Fig. 4:

Figure. 4 (column fitting image single)[21]

The net flow of energy that crosses the boundary of the unit element is given by:
D1& +
D1

G  
R 

<= 


<= 


D1&

 G   + D1 +

G  
R 

<= 


< = 


+ D1 +

G  
R 

<= 

 G   D1

< = 


 G  

(18)

10

ACCEPTED MANUSCRIPT
The terms resulting from the energy balance are arranged and divided by the unit element

F5 = I

Where,

<=


=&

:<1&= :<1= :<1=

+
+
J
:;
:
:

<=
+


<=
<=
;



<=
+


(19)

<=
< =
;



Arranging equation (19) this yields the following:

Where <1&
|= = 0.

<=
< =
+




SC

:<1=
:<1=
:<1=
+
+
+ <1&
|=
:;
:
:

M
AN
U

F5 = &

RI
PT

   (19).

Transforming this source to a Lagrangian framework, &, , are

  
/ / /

(20)

. Arranging the

terms of equation (20) and applying a temporal discretization, we finally obtain that the source
given by equation (21):

:<1=
:<1=
:<1=
+
+
N
<1= / <1= /
<1=
:;
:
:
=
=

TE
D

F5 =

L ;

(21)

When particularised for our case, this yields:

EP

(22)

AC
C

F5

1 ,,
,,
=
 %
 %
d
d

3.4 Radiation model.

The radiative heat transfer was modelled using the discrete ordinates model (DO). This model
solves the radiative heat transport equation (RTE) by discretising all directions of radiation in a

finite number of solid angles [22, 23]. Every solid angle is associated with a vector direction #
fixed in the global Cartesian system. For the fluid region, the RTE is calculated by equation (23)
while for the boundary conditions for emitting and reflecting walls, the RTE is solved by
equation (24). Equation (23) represents the energy balance of radiative energy, in which the
intensity obtained is the sum of the emission, absorption and scattering away from direction #,

11

ACCEPTED MANUSCRIPT

and scattering in the direction of #. Equation (24) represents the energy emitted and reflected

by radiation from a surface.

< , #= =

, R 2%

'R 2% 2 345/
+
<, #=<#, #=

1,

<, #= # |

|
3q .

(23)

(24)

RI
PT

<, #= + <' + 2 345/ =<, #= =

The absorption coefficient is calculated using the Weight Sum of Grey Gases Model (WSGGM)

M
AN
U

fraction of water vapour and carbon dioxide in the gases.

SC

[24, 25], where the local value of the absorption coefficient is a function of the local mass

4. CFD MODEL
4.1 Mesh.

The total volume of the furnace was discretised in a mesh of approximately 11.4 millions of
cells, of which 5.4 million belong to the fluid region of the furnace, 1.2 million belong to the

TE
D

solid region of the billets and 4.8 million belong to the solid region of the furnace walls. The
solid region of the furnace walls was meshed extruding prism to form all insulation layers
forming the walls as shown in Fig. 5(c). The movement of billets inside the furnace is modelled
using the algorithm described in section 2.3. As shown in Fig. 5(a) for the proper functioning of

all billets.

EP

the algorithm, the billets were discretised using a structured mesh that is exactly the same in

AC
C

The fluid region of the furnace was discretised with an unstructured mesh formed by
tetrahedral elements. The fluid region is formed by two volumes, one volume corresponding to
the interior of the furnace and the other corresponding to a volume created in the exterior in
the rear of the furnace. This second volume is created to evaluate energy losses through the
windows of the discharge mechanism.
Figure. 5 (column fitting image 2)
4.2 Boundary conditions.
4.2.1 Burners.

12

ACCEPTED MANUSCRIPT
The burner geometry is modelled as shown in Fig. 5(b), and the air flux is introduced by the
ring placed on top of the burner in the axial direction of the burner, while the gas is introduced
by the ring placed bellow the air inlet perpendicular to the axial direction of the burner. The
two flows are modelled as a mass flow inlet condition. The composition and the technical
specification of the natural gas used are shown in Table 2. There are three types of burners,

RI
PT

whose size depends on the location (Preheating, Heating and Equalisation zones). The mass

10% excess air with respect to the stoichiometric ratio.

Table 2
Table 3

M
AN
U

4.2.2 Thermal boundary condition and material properties.

SC

flow rates introduced in the burners are detailed in Table 3. The natural gas is burned with a

As detailed in Fig. 5(c), the furnace walls are made up of different layers. The characteristics of
the materials of these layers were provided by the furnace manufacturer. The heat flow
through the furnace walls, is evaluated on the exterior surface, and is modelled with mixed
thermal conditions that consider combined radiation and natural convection between the wall

TE
D

and the exterior, in which the operation conditions are indicated in Table 4. The heat flux
dissipated in the rollers and the discharged mechanism are modelled assuming a convection
heat flux with a water flux into the rollers and a discharge mechanism with a fixed
temperature, and the water sealing of the walking beam system is modelled as a wall with the

EP

properties of the water with a fixed temperature because the experimental furnace is

AC
C

regulated to ensure these conditions.

Table 4

Fig. 6 shows the specific heat of the billets that was obtained experimentally. For the correct
operation of the algorithm to calculate the movement of the slabs, the specific heat values
have been adjusted to a function. This function must be differentiable and continuous in the
whole domain of the working temperature range to avoid numerical instabilities in the
calculation. This function is composed of three polynomials. The thermal conductivity of the
billets in Table 5, is obtained for steel with similar properties to those of the billets [26]. The
emissivity of the walls and the floor into the furnace is set to 0.8, and the emissivity of the
billets is set to 0.6.

13

ACCEPTED MANUSCRIPT
Figure. 6 (column fitting image 1.5)
Table 5
5. RESULT AND DISCUSSION
5.1 Temperature field of the billets.

RI
PT

The methodology presented here has been applied to simulate the whole furnace under
nominal operation. The fine mesh employed in the billets allows a very detailed resolution of
their temperature fields. Fig. 7 shows the temperatures field of the billets in which there is two
important phenomena are observed in the heating of the billets. On the one hand, as shown in

SC

Fig. 7, the heating of the billets is not homogeneous in the longitudinal direction of the billets.
The surface temperature is higher in areas just below the burners. This feature is also shown in
Fig. 8, where the composed image represents the power absorbed by the billets, which is

M
AN
U

higher in the vicinity of the burners. This fact implies that the arrangement of burners might
not be optimal for the heating process and would suggest that a staggered arrangement of
burners would homogenise the temperature of the billets in the longitudinal direction. On the
other hand, the temperature of the billets increases in the region near the walls. This
phenomenon is mainly due to two factors: Firstly, the heat exchange surface is greater than

TE
D

the rest of the billet and secondly, the radiation absorbed is higher due to the proximity of the
walls. This also influences the first and last lines of burners, which are closer to the furnace
walls, thereby favouring overheating on this area of the billets. It is common practice during
the furnace operation to correct the overheating of the heads of the billets by manually

simulation.

EP

reducing the power of certain burners, but this has not been taken into account in the

AC
C

Figure. 7 (column fitting image 2)

5.2 Temperature field of the furnace gases.

As mentioned above, the furnace is divided in to four zones (Preheating, Heating, Equalisation
Right, and Equalisation Left). As detailed in Fig. 8, the temperature in the first half of the
preheating zone is significantly lower than in the other areas. However, the power absorbed in
this part by the billets is due to convection, and is greater than the power absorbed in other
areas due to the higher thermal gradient. At the back of the furnace, the power absorbed by
convection is very low due to the small thermal gradient between the billet and the furnace
gases. As shown in Fig. 8, thermal radiation is the main mechanism transferring heat to the
billets. The figure also shows the influence of the hot walls of the furnace on the power
14

ACCEPTED MANUSCRIPT
absorbed by the billets due to radiation, which causes an increase in the power absorbed due
to radiation in the first and last billets, which are located near the furnace walls.
As shown in Fig.11, the main power consumption is due the heat absorbed by the billets,
which is 16,621 kW. The power supplied by the fuel is 24,345 kW, therefore the efficiency of
the furnace is 68.3%. The second biggest power consumption is due to the heat dissipated as

RI
PT

exhaust gases (5,438 kW). As a part of this energy is used to preheat the air employed by the
burners to raise its temperature to 300 0C. Part of the power recovered by the heat exchanger
and introduced back into the system (2,578 kW). The power dissipated by the flue gases at the
outlet of the heat exchanger is 2,860 kW (11.7% of the power supplied by the fuel).

SC

Figure. 8 (column fitting image 2)

5.3 Validation of the model.

M
AN
U

The real case data have been obtained from a furnace control system, which records hundreds
of variables for the whole facility. Other parameters have been measured ad-hoc or have been
estimated, such as the power dissipated by the walls to be used for the energy balance of the
real furnace. This energy balance was calculated based on data obtained in a thermography
study.

TE
D

The real surface temperature is measured using a pyrometer, which is placed on top of the
furnace discharge door. This pyrometer records the temperatures of the billet during the
discharge process and is used as a control parameter for the whole process. In this work, the
average value of the temperature profile of 10 billets during nominal operation of the furnace

EP

has been used. As shown in Fig. 9, this profile is compared with CFD results. As seen from the
graph, the temperature pattern obtained in the CFD simulation is reasonably similar to that

AC
C

registered in the real operation. Both patterns (CFD and real case) have temperature peaks
that can be associated with the areas aligned with the rows of burners. The difference
between the two patterns with regard to the temperature at the ends of the billet might be
due to the manual regulation of the furnace, as that has been previously mentioned. In
addition, the temperature pattern is displaced with respect to the CFD pattern in this process,
which is probably because the extraction velocity of the billet from the furnace is not
determined by the action of a pusher. Instead extraction is performed by means of rollers that
might slip and therefore do not attain a constant speed, which impedes the correct
superposition of the patterns.

15

ACCEPTED MANUSCRIPT
Figure. 9 (column fitting image 2)
Fig. 10 shows the difference between the surface temperatures pattern of the CFD simulation
versus the minimum, maximum and standard deviation of the patterns of all real case samples.
Figure. 10 (column fitting image 1.5)

RI
PT

A set of thermocouples are distributed inside the furnace at the locations shown in Fig. 8.
Table 6 shows the temperatures recorded by these thermocouples. These thermocouples are
fixed to the furnace walls, such that the real case measurement could be slightly
overestimated due to the radiation emitted from the hot furnace walls, which increases the
surface temperature of the thermocouple, leading to a higher value than the gas passing over

M
AN
U

Table 6

SC

the tip of the sensor. No correction has been introduced in the CFD for this comparison.

The difference in the temperature of the exhaust gases between CFD and real data is 36.4 K,
which, bearing in mind all the uncertainties involved and the complexity of the simulation is
considered reasonable. The temperatures of the gases measured by thermocouples located in
the preheating zone are in the order of 1,250 K, while the temperatures estimated by CFD are
between 1,234 and 1,268 K. In the heating zone, the temperature recorded by the first

TE
D

thermocouple and the temperatures estimated by the model are very close. The temperatures
recorded by the thermocouples in the equalisation left zone are around 1,473 K (field
measurement) and 1,462 K (CFD) and in the equalisation right zone the temperatures of the
field measurements range between 1,423 K and 1,439 K while the temperatures obtained in

EP

CFD range between 1,438 K and 1,441 K. The deviation between the recorded temperatures by
the second thermocouple of the heating zone and the CFD is 2.35%. This is the largest

AC
C

deviation between the field measurements and the model, therefore the methodology
developed can be considered reasonably accurate in regards of the gas temperature field.
The main source of this discrepancy is probably that the walls of the furnace are perfect in the
simulation, while the real walls have gaps, in addition to thermal bridges, which were not
included in the simulation.
Fig. 11 shows the overall energy balance of the furnace. The power absorbed by the billets in
the CFD simulation is 16,621 kW (68.3% of the power supplied by the natural gas) while in the
real case, the measured power absorbed is 16,454 (67.8% of the power supplied by the natural
gas). Therefore, the difference between the power absorbed by the billets in the CFD
simulation and in the real case is only 0.5 % with respect to the power supplied by the natural
16

ACCEPTED MANUSCRIPT
gas. The power absorbed by the billets in the real case was calculated based on the average
volume temperature of the billet output. The average volume temperature for the real case
was estimated from the temperature measured at the surface and the gradients estimated by
the CFD. The power difference between the power stated in the energy balance of the real
case, and that given by the CFD simulation does not exceed 1.5% of the total power supplied

Figure. 11 (column fitting image 2)

6. CONCLUSIONS

RI
PT

by the natural gas except for power losses due to the apertures, which amount to 2%.

This study presents a new methodology for converting the steady periodic operation of a

SC

metal-gas system into a steady-state problem. This new methodology has been developed to
simulate reheating furnace.

M
AN
U

This model is implemented in the commercial software ANSYS Fluent. The main advantage of
this new methodology is that it allows the calculation to be solved in the steady state, and thus
achieves a significant reduction in calculation time compared with a transient modelling
method. Hence, computational effort can be employed for detailed discretisation of the
furnace geometry and correct simulation of the combustion process, which generates results

TE
D

that are more accurate.

The simulation results show detailed maps of the variables of primary interest such as the
surface temperature of the billets at the outlet, the temperatures of the exhaust gases and in
different areas of the furnace, and the distribution of the power consumption which reflects

EP

the energy balance. Comparison the billets and gases temperatures and energy balance
between the real case data those derived from the simulation demonstrates that the

AC
C

predictions of the simulation are reasonably accurate. The main differences between the real
case and the simulation are in the temperature difference of the exhaust gases, where the
temperature in the CFD simulation is 4.3% greater than the real case and the losses shown in
the energy balance, where the losses in the CFD simulation are 2% lower than the real case in
terms power supplied by the natural gas. In both cases it is believed that the discrepancy is
due to imperfections in the walls of the furnace that were not included in the virtual model.
The methodology described here, once validated, can be applied to the prediction and
quantification of the potential benefits of geometrical and operational modifications of a
furnace, leading to better performance in terms of productivity, energy efficiency and other
features of reheating furnaces.
17

ACCEPTED MANUSCRIPT
7. ACKNOWLEDGEMENTS
This research was partially supported by the Spanish Government, Ministerio de Innovacin
Economa y Competitividad under the program FEDER INTERCONNECTA EXP 00064768 / ITC20133025. The participation of RUSSULA and CELSA ATLANCTIC SL. is also gratefully

RI
PT

acknowledge.

8. REFERENCE

AC
C

EP

TE
D

M
AN
U

SC

[1] S.H. Han, S.W. Baek, S.H. Kang, C.Y. Kim, Numerical analysis of heating characteristics of a
slab in a bench scale reheating furnace, International Journal of Heat and Mass Transfer, 50
(2007) 2019-2023.
[2] C. Zhang, T. Ishii, S. Sugiyama, Numerical modeling of the thermal performance of
regenerative slab reheat furnaces, Numerical Heat Transfer, Part A: Applications, 32 (1997)
613-631.
[3] C. Zhang, T. Ishii, Y. Hino, S. Sugiyama, The numerical and experimental study of nonpremixed combustion flames in regenerative furnaces, Journal of Heat Transfer, 122 (1999)
287-293.
[4] M.J. Huang, C.T. Hsieh, S.T. Lee, C.H. Wang, A coupled numerical study of slab temperature
and gas temperature in the walking-beam-type slab reheating furnace, Numerical Heat
Transfer, Part A: Applications, 54 (2008) 625-646.
[5] J.G. Kim, K.Y. Huh, I.T. Kim, Three-dimensional analysis of the walking-beam-type slab
reheating furnace in hot strip mills, Numerical Heat Transfer, Part A: Applications, 38 (2000)
589-609.
[6] J.G. Kim, K.Y. Huh, Prediction of transient slab temperature distribution in the re-heating
furnace of a walking-beam type for rolling of steel slabs, ISIJ International, 40 (2000) 11151123.
[7] C.T. Hsieh, M.J. Huang, S.T. Lee, C.H. Wang, Numerical modeling of a walking-beam-type
slab reheating furnace, Numerical Heat Transfer; Part A: Applications, 53 (2008) 966-981.
[8] M.Y. Kim, A heat transfer model for the analysis of transient heating of the slab in a directfired walking beam type reheating furnace, International Journal of Heat and Mass Transfer, 50
(2007) 3740-3748.
[9] S.H. Han, D. Chang, C.Y. Kim, A numerical analysis of slab heating characteristics in a
walking beam type reheating furnace, International Journal of Heat and Mass Transfer, 53
(2010) 3855-3861.
[10] S.H. Han, D. Chang, Optimum residence time analysis for a walking beam type reheating
furnace, International Journal of Heat and Mass Transfer, 55 (2012) 4079-4087.
[11] E. Martn, M. Meis, C. Mourenza, D. Rivas, F. Varas, Fast solution of direct and inverse
design problems concerning furnace operation conditions in steel industry, Applied Thermal
Engineering, 47 (2012) 41-53.
[12] C.-K. Tan, J. Jenkins, J. Ward, J. Broughton, A. Heeley, Zone modelling of the thermal
performances of a large-scale bloom reheating furnace, Applied Thermal Engineering, 50
(2013) 1111-1118.
[13] A. Emadi, A. Saboonchi, M. Taheri, S. Hassanpour, Heating characteristics of billet in a
walking hearth type reheating furnace, Applied Thermal Engineering, 63 (2014) 396-405.
[14] Y. Li, G. Wang, H. Chen, Simultaneously estimation for surface heat fluxes of steel slab in a
reheating furnace based on DMC predictive control, Applied Thermal Engineering, 80 (2015)
396-403.
18

ACCEPTED MANUSCRIPT

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

[15] T. H. Shih, W. W. Liou, A. Shabbir, Z. Yans, J. Zhu, A new k-[epsilon] eddy viscosity model
for high Reynolds number turbulent flows: model development and validation, National
Aeronautics and Space Administration ; National Technical Information Service, distributor,
Washington, D.C. : Springfield, Va., 1994.
[16] B.F. Magnussen, On the structure of turbulence and a generalized eddy dissipation
concept for chemical reaction in turbulent flow, in: 19th AIAA Aerospace Science Meeting, St
Louis, Missouri, USA, 1981.
[17] W.P. Jones, R.P. Lindstedt, Global reaction schemes for hydrocarbon combustion,
Combustion and Flame, 73 (1988) 233-249.
[18] I. ANSYS, Fluent 14.0 Theory Guide, Lebanon, New Hampshire (USA), 2013.
[19] I. ANSYS, FLUENT 14.0 User's Guide, Lebanon, New Hampshire (USA), 2013.
[20] I. ANSYS, UDF Manual, Lebanon, New Hampshire (USA), 2013.
[21] H. K. Verstegg, W. Malalasekera, An introduction to computational fluid dynamics. the
finite volume method, Essex CM20 2JE, England, 1995.
[22] J.C. Chai, H.S. Lee, S.V. Patankar, Finite volume method for radiation heat transfer, Journal
of Thermophysics and Heat Transfer, 8 (1994) 419-425.
[23] J.Y. Murthy, S.R. Mathur, Finite volume method for radiative heat transfer using
unstructured meshes, Journal of Thermophysics and Heat Transfer, 12 (1998) 313-321.
[24] A. Coppalle, P. Vervisch, The total emissivities of high-temperature flames, Combustion
and Flame, 49 (1983) 101-108.
[25] T.F. Smith, Z.F. Shen, J.N. Friedman, Evaluation of coefficients for the Weighted Sum of
Gray Gases Model, Journal of Heat Transfer, 104 (1982) 602-608.
[26] VDI, Heat Atlas, Dsseldorf, 2010.

19

ACCEPTED MANUSCRIPT
n

JL1

CnH2n+2 + n/2 O2 n CO + (n+1) H2

7,82. 10

7,47. 10

7,11. 10

6,76. 10

1-4

3,00. 10

4,45. 10

6,22. 10

5,33. 10

4,98. 10

CnH2n+2 + n H2O n CO + (2n+1) H2

JL3

H2 + O2 H2O

JL4

EA

Reaction orders

13

125,58

[CnH2n+2]

13

125,58

[CnH2n+2]

13

125,58

[CnH2n+2]

13

125,58

[CnH2n+2]

11

125,58

18

-1

167,44

[H2]

18

-1

167,44

[H2]

18

18

CO + H2O CO2 + H2

12

1-4

2,75. 10

[O2]

5/4

1/2

[O2]

5/4

1/2

[O2]

5/4

1/2

[O2]

5/4

1/2

[O2]

1/2

[O2]

1/2

[O2]

1/2

[O2]

-1

167,44

[H2]

-1

167,44

[H2]

1/2

[CnH2n+2] [H2O]

M
AN
U

JL2

Ar

RI
PT

Reaction

SC

Reaction
n.

83,72

9/4

[H2O]

-1

9/4

[H2O]

-1

9/4

[H2O]

-1

9/4

[H2O]

-1

[CO] [H2O]

Table 1 The Jones-Lindstedt global reaction mechanism for CnH2n+2 for n4.

Species

Mole fraction

AC
C

EP

TE
D

CH4
0.85190
C2H6
0.12464
0.00178
C3H8
0.00037
C4H10
N2
0.02131
3
HHV (MJ/Nm )
42.37
3
LHV (MJ/Nm )
38.18
3
Density (kg/Nm )
0.767
3
Wobbe Index (MJ/Nm )
54.68
Table 2 Composition of natural gas (data provided by
the gas supplier)

Zones

Mass flow
N burners

Preheating
Heating
Equalisation left
Equalisation right
Table 3 Mass flow of the burners

16
24
8
8

NG

Air

Mass flow per burner

NG
Air

(kg/s)

(kg/s)

(kg/s)

(kg/s)

0.17092
0.23620
0.03742
0.04363

3.15136
4.37016
0.69288
0.80680

0.01064
0.00984
0.00468
0.00545

0.19696
0.18209
0.08661
0.10085

ACCEPTED MANUSCRIPT

RI
PT

Parameter
Value
Units
hconv, air
17
(W/m2K)
Tair
293
(K)
ext,air
0.85
(-)
hconv, water
2500
(W/m2K)
296.5
(K)
Twater, rollers
Tsealing
301.4
(K)
Table 4 Operating parameters of the walls and the cooling system

M
AN
U

SC

Temperature [K]
W/mK
293
55
373
55
473
51
573
48
673
44
773
41
T>773
41
Table 5 Thermal conductivity of the billets

Thermo
couple 2

Thermo
couple 1

Thermo
couple 2

Thermo
couple 1

Thermo
couple 2

Thermo
couple 1

Thermo
couple 2

Thermo
couple

Zone 1

Zone 1

Zone 2

Zone 2

Zone 3

Zone 3

Zone 4

Zone 4

Exhaust
Gases

Experimental (K)

1253.1

1252.6

1456.8

1472.8

1474.8

1472.3

1423.9

1439.0

838.1

CFD (K)

1268.0

1234.1

1458.6

1438.3

1463.0

1461.0

1438.8

1441.4

874.5

T (%)

1.19

1.48

0.13

2.35

0.80

0.77

1.05

0.16

4.17

TE
D

Thermo
couple 1

AC
C

EP

Table 6 Temperatures of the furnace gases (CFD & real case)

AC
C

EP

TE
D

M
AN
U

Fig. 1 Installation scheme

SC

RI
PT

ACCEPTED MANUSCRIPT

Fig. 2 Detail of the domain at the back of the furnace

SC

RI
PT

ACCEPTED MANUSCRIPT

TE
D

M
AN
U

Fig. 3 Movement pattern of billets

AC
C

EP

Fig. 4 Components of the heat flux (adapted from [21])

M
AN
U

SC

RI
PT

ACCEPTED MANUSCRIPT

TE
D

Fig. 5 Computational mesh

1600

EP

1200
1000

800

AC
C

specific heat (J/kgC)

1400

600
400
200

200

400

600
800
Temperature (C)

1000

Fig. 6 Specific heat values of the billets

1200

SC

RI
PT

ACCEPTED MANUSCRIPT

AC
C

EP

TE
D

M
AN
U

Fig. 7 Temperature field of the billets

Fig. 8 Temperature field of the furnace gases and quantification of the mechanisms involved in heat transfer

TE
D

M
AN
U

SC

RI
PT

ACCEPTED MANUSCRIPT

1460
1440

CFD

Temperature (K)

AC
C

1420

EP

Fig. 9 Temperature pattern of the billet at the exit (real case vs simulated)

1400
1380
1360
1340
1320
1300

10

Distance along the longitudinal axis of the billet (m)

Fig. 10 Comparison between temperature patterns (CFD vs real case)

11

12

RI
PT

ACCEPTED MANUSCRIPT

AC
C

EP

TE
D

M
AN
U

SC

Fig. 11 Comparison of the energy balance between CFD simulation and the real case

ACCEPTED MANUSCRIPT
Research Highlights:

RI
PT

SC
M
AN
U
TE
D

EP

A new methodology is described in the CFD simulation of a reheating furnace in steadystate condition.
The movement of billets is performed by modifying the energy transport equation.
The method achieves a reduction in calculation time compared with transient modelling
method.
CFD simulation has been compared with a real case with reasonably close results.
The method can be applicable to predict performance following geometrical and
operational modifications.

AC
C