Abstract
A heat conduction model developed to predict the temperature distribution within the oxide layer of carbon steel being rolled is
presented. This model takes into account the different physical properties of the three oxide species, and the parabolic growth of the layer.
The thickness of the layer is divided into 40 nodes or elements of which 36 are considered to be of wustite, three of magnetite and only one
of hematite to comply with their proportions. It is found that this particular model is too complicated when the aim is to evaluate the effect
of the oxide crust in thin oxide layers, such as those encountered during strip rolling, because similar results can be obtained using a single
node model based on the properties of wustite; whereas with the behaviour of thick layers, such as those encountered after reheating or
during roughing passes, is modelled, the present model becomes valuable. The thermal gradients predicted by the model can be employed
to predict the integrity of the oxide layer. # 2000 Elsevier Science B.V. All rights reserved.
0924-0136/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 4 - 0 1 3 6 ( 0 0 ) 0 0 5 6 9 - 0
M. Torres, R. ColaÂs / Journal of Materials Processing Technology 105 (2000) 258±263 259
where t is the time and a the thermal diffusivity of the The proportion of the different species within the whole
material, layer is taken from experimental work [6,7,11] which report
k that the wustite represents around 90% of the layer, whereas
a (3) the magnetite is around 8% and the hematite occupies only
Cp r
2% of the full layer.
where Cp is the heat capacity and r the density. Heat conduction within the oxide layer and strip or plate is
Heat losses during air cooling are caused by convection approximated by ®nite differences in one dimension, Fig. 2.
and radiation, the former can be described by To start with, the oxide thickness is divided into 40 nodes or
q elements, of which 36 are considered to be wustite, three
h
Ts ÿ T1 (4) magnetite and one hematite, in order to comply with the
A
reported proportions [6,7,11], while half the thickness of the
where h is the convective coef®cient, and Ts and T1 are the strip is divided into 20 nodes, this being done after assuming
temperatures on the surface and that of the surrounding symmetrical cooling on the top and bottom surfaces [5]. A
media. Heat losses due to radiation are calculated by thermal pro®le or a constant temperature value can be
q speci®ed on both the strip and the oxide layer. An initial
es
Ts4 ÿ T1
4
(5) oxide thickness of 1 mm (the thickness of each node is set to
A
2.510ÿ8 m), and constant temperature within the layer are
where e is the emissivity and s5.669910ÿ8 W/m2 K4 is assumed when the case of a plate or strip coming from
the Stefan±Boltzmann constant. descaling is considered. Since the size of the elements on
Hollander [1] assumed that heat loss by convection layer and strip are different (normally the thickness of the
in a strip cooling from 1200 to 9008C are only about elements in the strip will be around 610ÿ4 m), an implicit
4±6% of the total loss and, therefore, is negligible. Other ®nite difference method [16,17] was chosen.
authors [2±5] use an empirical equation to calculate the Once the size of the elements in the layer and strip is
effect of convection and radiation on a cooling piece of established, a stability time, which assures that the thermal
steel, gradient does not penetrate more than one node per iteration,
H a bTs c
Ts 2734 (6) is calculated in a way that will be shown later. With the
stability time it is possible to calculate a dimensionless
where H is the heat lost per unit area, a6746 W/m2, parameter (Z) for the strip,
b21.2 W/m2 8C, c4.76310ÿ8 W/m2 8C4, and Ts is
expressed in 8C. ks Dtm2
Zs (8)
Other phenomena that the model has to comply with are l2
the growth of the layer and the distribution of the different
oxide species. In the present work it is assumed that the
growth follows a parabolic regime [7,15],
Dx kp t0:5 (7)
where Dx is the oxide thickness at a given time (t), and kp the
growth coef®cient, which depends on the type of oxide and
temperature. Eq. (7) implies that the growth is very rapid on
oxide-free surfaces, but decreases as the oxide builds
up. During actual rolling the layer is removed with high
pressure water jets (descaling), exposing the free surface of
the metal to the surrounding media. Fig. 2. The nodes used to compute heat ¯ow.
260 M. Torres, R. ColaÂs / Journal of Materials Processing Technology 105 (2000) 258±263
Fig. 4. Temperature pro®le within the oxide layer of the 254 mm thick Fig. 6. Temperature pro®le within the oxide layer of the 25.4 mm thick
slab. plate.
from a reheating furnace, after being heated uniformly to within the oxide layer formed on top of the plate after 40 s is
12508C. It was also assumed that the oxide layer thickness shown in Fig. 6. The computed ®nal thickness of 67 mm was
was equal to 6.25 mm at the moment at which the slab leaves obtained in a manner similar to that for computing the
the furnace. growth on the slab.
The temperature pro®le developed within the oxide layer Temperature evolution on both surfaces of the oxide layer,
after 40 s is shown in Fig. 4, in which the individual as well as its thickness, for the simulation of 2.54 mm thick
thickness of the different oxide species is marked. It is strip being produced in a six-stand continuous mill is shown
interesting to note the sharp gradients that develop in such in Fig. 7. In this case, the temperature at the metal±oxide
a small distance, as a result of the low thermal conductivity interface is considered to be that calculated by a two-
of the compounds [8±10]. The ®nal thickness of the oxide dimensional ®nite difference program described elsewhere
layer (6.30 mm) was calculated employing Eq. (7), with the [5], whilst the temperature on the oxide surface is calculated
coef®cients [5,7,15] corresponding to the average tempera- by the present model. The seven temperature drops shown in
ture in the individual layers. Fig. 7(a) correspond to the cooling due to descaling and
Simulation of the thermal gradients which develop in a contact with the work-rolls of each stand. The change of
25.4 mm thick plate after descaling is shown in Fig. 5: as thickness of the oxide layer, Fig. 7(b), is computed after
with Fig. 3, the curves plotted correspond to the temperature assuming that the oxide is fully plastic and will deform to the
evolution at the centre and surface of the plate (the metal± same extent as the steel during rolling, and that it will be able
oxide interface) and the oxide surface. The calculations were to grow, when the strip is in air, following the parabolic
conducted assuming a homogeneous temperature distribu- regime given by Eq. (7). The computed thermal gradient
tion within the plate at time zero, and a descaled oxide-free within the oxide layer at the exit of the ®rst stand, when the
surface. This last condition implies that the oxide layer will thicknesses of the outgoing stock and of the oxide layer are
grow at fast rate following Eq. (7). The temperature gradient 13 mm and 6 mm, respectively, is shown in Fig. 8.
Fig. 5. Cooling predicted by the model for a 25.4 mm thick plate; the
temperatures shown correspond to the centre of the plate, the metal±oxide Fig. 7. Computer simulation of: (a) the temperature evolution; and (b) the
interface and the oxide surface. oxide growth during the production of 2.54 mm thick strip.
262 M. Torres, R. ColaÂs / Journal of Materials Processing Technology 105 (2000) 258±263
References
Table 2
Results of the simulations
[1] F. Hollander, Mathematical models in metallurgical process devel-
Slab Plate Strip opment, Iron Steel Institute Publication 123, London, 1970, p. 46.
[2] R. Harding, Ph.D. Thesis, University of Shef®eld, UK, 1976.
Steel thickness 254 mm 25.4 mm 13 mm
[3] L.A. Leduc, Ph.D. Thesis, University of Shef®eld, UK, 1980.
Oxide thickness 6.3 mm 67 mm 6.1 mm
[4] C.M. Sellars, Mater. Sci. Technol. 1 (1985) 325.
DT a 187.58C 2.28C 58.08C
[5] R. ColaÂs, Mater. Sci. Technol. 14 (1988) 388.
Thermal gradient 29.88C/mm 32.88C/mm 9.58C/mm
[6] L.S. Darken, R.W. Gurry, J. Am. Chem. Soc. 68 (1946) 798.
Heat flow 250 kW/m2 130 kW/m2 30 MW/m2
[7] F. Lorang, Rev. Univ. Mines 17 (1961) 514.
a
Across the oxide after 40 s of simulation in the slab and plate, and at [8] G.V. Samsonov (Ed.), The Oxide Handbook, Plenum Press, New
the exit of the ®rst stand for the strip. York, 1973.
M. Torres, R. ColaÂs / Journal of Materials Processing Technology 105 (2000) 258±263 263
[9] R. Taylor, C.M. Fowler, R. Rolls, Int. J. Thermophys. 1 (1980) 225. [17] S.V. Pantakar, Numerical Heat Transfer and Fluid Flow, Hemisphere,
[10] J. Slowik, G. Borchard, C. KoÈhler, R. Jeschar, R. Scholz, Steel Res. 7 New York, 1980.
(1990) 302. [18] C.M. Sellars, J.A. Whiteman, Met. Technol. 8 (1981) 10.
[11] J. Paidassi, Rev. MeÂtall. 54 (8) (1957) 2. [19] M.P. Guerrero, C.R. Flores, A. PeÂrez, R. ColaÂs, J. Mater. Process.
[12] L. Himmel, R.F. Mehl, C.E. Birchenall, Trans. AIME 197 (1953) Technol. 94 (1999) 52.
889. [20] M. Torres, R. ColaÂs, Modelling of Metal Rolling Processes, Institute
[13] H. HoÈfgen, G. Zouhar, F. Birnstock, J. Bathelt, in: Proceedings of the of Materials, London, 1993, p. 629.
Fourth International Steel Rolling Conference on Science and [21] R. ColaÂs, in: J.J. Jonas, T.R. Bieler, K.J. Bowman (Eds.), Advances in
Technology of Flat Rolling, IRSID, 1987, p. B.2.1. Hot Deformation Textures and Microstructures, TMS-AIME, War-
[14] C. Devadas, I. Samarasekera, Ironmaking and Steelmaking 6 (1986) rendale, 1994, p. 63.
311. [22] W. Jaenicke, S. Leistikow, A. StaÈdler, J. Electrochem. Soc. 111
[15] C. Wagner, Atom Movements, ASM, Metals Park, 1951. (1964) 1031.
[16] R.I. Burden, J.D. Farues, Numerical Analysis, 3rd Edition, Prindle, [23] A.G. Crouch, J. Am. Ceram. Soc. 55 (1972) 558.
Weber & Schmidt, Boston, MA, 1985. [24] T.E. Mitchell, D.A. Voss, E.P. Buther, J. Mater. Sci. 17 (1982) 1825.