PREDICTION OF WORK ROLL INITIAL CROWN ACCORDING

TO DESIRED STRIP PROFILE IN HOT ROLLING

H. S. Salavati Pour
Department of Mechanical Engineering, Amirkabir University of Technology, Hafez Ave., Tehran 15916-34311, Iran
[email protected]
H. Khademhosseini-Beheshti, M. Poursina*
Department of Mechanical Engineering, Faculty of Engineering, University of Isfahan, Isfahan, 81746-73441, Iran.
[email protected], [email protected]

* Corresponding Author

(Received: November 02, 2010 – Accepted in Revised Form: December 15, 2011)

doi: 10.5829/idosi.ije.2011.25.01c.09

Abstract The final hot strip profile is a superposition of the roll initial crown, the roll bending and
flattening crown, the roll wearing crown and the roll thermal crown. In this research, linear regression
models are proposed to predict the roll force, the roll wearing crown and the roll thermal crown using
experimental data provided by Mobarake Steel Complex (MSC). The Euler-Bernoulli beam theory
based on elasticity approach is also used to predict the roll bending and flattening crown. Finally, the
work roll initial crown in order to obtain desired strip profile in hot rolling is predicted utilizing a
computer programming. The obtained initial crowns are then used for seven stands of an actual roll
schedule for hot strip mill of MSC. The measured strip profile shows a good agreement with the
desired one for the mentioned mill.

Keywords Hot rolling; Initial crown; Wearing crown; Thermal crown; Bending crown; Empirical
data; Regression

‫ ﺗﺎج ﺳﺎﯾﺸﯽ ﻏﻠﺘﮏ و‬،‫ ﺗﺎج ﺧﻤﺸﯽ و ﺗﺨﺖ ﺷﺪﮔﯽ ﻏﻠﺘﮏ‬،‫ﭼﮑﯿﺪه ﭘﺮوﻓﯿﻞ ﻧﻬﺎﯾﯽ ورق ﺣﺎﺻﻞ ﺗﺎج اوﻟﯿﻪ ﻏﻠﺘﮏ‬
‫ در ﺗﺤﻘﯿﻖ ﺣﺎﺿﺮ ﺑﺎ اﺳﺘﻔﺎده از دادهﻫﺎي ﺗﺠﺮﺑﯽ اﺧﺬ ﺷﺪه از ﻣﺠﺘﻤﻊ ﻓﻮﻻد ﻣﺒﺎرﮐﻪ‬.‫ﺗﺎج ﺣﺮارﺗﯽ ﻏﻠﺘﮏ ﻣﯽﺑﺎﺷﺪ‬
‫ ﺗﺎج ﺳﺎﯾﺸﯽ و ﺗﺎج ﺣﺮارﺗﯽ ﻏﻠﺘﮏ و ﺑﺎ اﺳﺘﻔﺎده از ﺗﺌﻮري‬،‫ﻣﺪلﻫﺎي رﮔﺮﺳﯿﻮن ﺧﻄﯽ ﺑﺮاي ﭘﯿﺶﺑﯿﻨﯽ ﻧﯿﺮوي ﻧﻮرد‬
.‫ﺗﯿﺮ ﮐﻮﺗﺎه ﺑﺮ ﻃﺒﻖ اﺻﻮل اﻻﺳﺘﯿﺴﯿﺘﻪ ﺗﺎج ﺧﻤﺸﯽ و ﺗﺨﺖ ﺷﺪﮔﯽ ﺳﻄﺢ ﺗﻤﺎس ﻏﻠﺘﮏ ﺑﺪﺳﺖ آورده ﺷﺪه اﺳﺖ‬
‫ﻧﻬﺎﯾﺘﺎً ﺗﺎج اوﻟﯿﻪ ﻏﻠﺘﮏ ﺟﻬﺖ اﯾﺠﺎد ﭘﺮوﻓﯿﻞ دﻟﺨﻮاه ورق در ﻧﻮرد ﮔﺮم ﺑﺎ اﺳﺘﻔﺎده از ﯾﮏ ﺑﺮﻧﺎﻣﻪ ﮐﺎﻣﭙﯿﻮﺗﺮي ﭘﯿﺶ‬
‫ ﻣﻘﺎدﯾﺮ ﭘﯿﺶ ﺑﯿﻨﯽ ﺷﺪه ﻓﻮق ﺑﻪ ﻋﻨﻮان ﺗﺎجﻫﺎي اوﻟﯿﻪ ﻫﻔﺖ ﻗﻔﺴﻪ ﺧﻂ ﻧﻮرد ﮔﺮم ﻣﺠﺘﻤﻊ ﻓﻮﻻد‬.‫ﺑﯿﻨﯽ ﺷﺪه اﺳﺖ‬
‫ﻣﺒﺎرﮐﻪ در ﯾﮏ ﺑﺮﻧﺎﻣﻪ واﻗﻌﯽ ﻧﻮرد اﻧﺘﺨﺎب ﺷﺪﻧﺪ و ﭘﺮوﻓﯿﻞ ﺑﺪﺳﺖ آﻣﺪه ورق در ﺗﻄﺎﺑﻖ ﺑﺴﯿﺎر ﺧﻮﺑﯽ ﺑﺎ ﻣﻘﺪار اﻧﺪازه‬
.‫ﮔﯿﺮي ﺷﺪه ﺗﻮﺳﻂ ﻣﺠﺘﻤﻊ ﻓﻮﻻد ﻣﺒﺎرﮐﻪ ﻣﯽﺑﺎﺷﺪ‬

1. INTRODUCTION wearing crown and the roll forces on the work roll.
This causes quality of product and decreasing of
Hot rolling is one of the most common methods in cultch. Investigation of rolling was started from
steel production. In fact, more than 80 percent of early decades of 20th century. Some of these
steel production is produced by this method. The investigations are mentioned in this article.
efficiency and delicacy of rolling production Liu [1] used finite element method for
makes this industry to be a precise process against modeling the rolling process. Chakraborti et al.
other methods of metal forming. Among different [2] studied the surface profiles of hot rolled slabs,
methods of rolling, hot rolling mill encounters quantified in terms of crown, by genetic
many parameters to predict the work roll initial algorithms. They used two different models and
crown in order to obtain desired strip profile. carried out the simulations in a multi-objective
Precise prediction of strip profile is possible based mode to generate the relevant Pareto fronts, which,
on the initial crown, the thermal crown, the in turn, was tested against the operational data of

IJE Transactions C: Aspects Vol. 25, No. 1, March 2012 - 79

an integrated steel plant. Tudball et al. [3] used 2. CROWN
a transient 3D finite element model to capture
thermal variations during hot rolling of steel. They The aim of an ideal rolling is to produce a strip
demonstrated the versatility of their code by a with homogeneous width and thickness. However,
series of validation exercises. John et al. [4] the product usually is not flat and has some wave
minimized the flatness defects in HR strips using a on it. One of the parameters that deals with flatness
novel hybrid model with the combination of a of strip is crown. Crown in each point is defined as
predictive artificial neural network (ANN) and a the variation of central thickness to the thickness
genetic algorithm (GA). Nandan [5] et al. used a of that point. The parameters for defining the
genetic algorithms-based multi optimization for the crown are shown in Figure 1. So, central crown is
hot rolling practice in an integrated steel plant. defined as shown in Eq. 1 [10]:
Their aim was to identify the parameter settings
and rolling schedules that would result in the
H ( 0)
optimum values of crown and flatness. H (−W / 2 + a ) H (+W / 2 − a )
Rossomando et al. [6] presented a real application
of optimal control of a hot rolling mill. They used x
the state space model formulation in the
minimization of the strip thickness variations and
presented the simulation results of the control a a

model. Kazeminezhad et al. [7] developed a

slab analysis to predict the rolling pressure Strip Width, W

distribution and rolling force in the wire flat rolling

process. Zhang Guo-min et al. [8] developed a Figure 1. Parameters for defining the crown
three-dimensional model for strip hot rolling, in
which plastic deformation of strip, thermal crown H1 + H 2
of rolls, roll deflection and flattening were Central Crown= H m − ( ) (1)
2
calculated by rigid-plastic finite element method,
finite difference method, influential function In this equation, Hm is the thickness in the central
method and elastic finite element method, width of strip (mm) and H1 and H2 are the
respectively. The roll wear was taken into thickness at the edge of strip (mm) with the
consideration in their work. Younes et al. [9] distance a. The distance a is about 19-40 mm and
presented the application of parameters design to depends on the strip width (W ) , strip thickness (t ) ,
improve both the product quality and the and the adopted standard [21]. There are some
equipment performance in a hot sheet and they parameters that affect the profile of strip. These
developed three empirical models to predict and parameters are the work roll initial crown, the
explain the relationship between variations in some work roll thermal crown, the work roll wearing
process parameters. The three process parameters
crown and the work roll elastic deformation due to
considered were the reduction ratio (Rd), the slab the bending and flattening. The effect of all these
temperature (Ts), and the exit strip speed (Sp). parameters in order to obtain desired strip profile
In this research, first new empirical models must be considered. So, the total profile of strip is
based on measured data by Mobarake Steel calculated using Eq.2:
Complex (MSC) and linear regression method are
obtained to predict the roll force, the roll wearing y x = ycvc− prof + ytherm−exp + ywe + ydef (2)
crown and the roll thermal crown. In addition,
Euler-Bernoulli theory based on elasticity In this equation ycvc − prof is the work roll initial
approach is presented to predict the roll bending
crown, ytherm − exp is the work roll thermal crown,
and flattening crown in hot rolling milling. Then, a
computer program was written in Delphi to predict ywe is the work roll wearing crown, ydef is the work
the roll initial crown according to desired strip roll elastic deformation due to bending and
profile. flattening.

80 - Vol. 25, No. 1, March 2012 IJE Transactions C: Aspects

 The strip profile is defined as: The roll force is measured by load cells of each
stand and is recorded by ibaAna5.7.1 software.
=C y( w / 2 − 40) − y0 (3) Figure 3 shows the measured roll force in a sample
The effects of all above parameters are rolled schedule for stand No. 3. As this figure
investigated in this article. shows, the roll force is not constant during a rolled
schedule and may be a function of strip width (W),
In this equation, C is the strip profile and
subscripts ( w / 2 - 20 ) is the value of y at 40mm from strip initial thickness (H), reduction (r), roll
the edge and the subscript (0) is the value of y at velocity (Vr), strip initial temperature (T) and strip
strip center. exit length (Le). Table 1 contains some portion of
the measured roll force as a function of mentioned
parameters during a rolled schedule for stand No.
3. EXPERIMENTAL WORK 3.
The experimental work was carried out in the hot
strip mill of MSC illustrated in Figure 2. In the
selected hot strip mill, slabs from the slab caster
are rolled into strips, with a corresponding
reduction in thickness from an input value of 203
mm to about 1.6–16 mm, while the width ranges
from 800 to 1600 mm. During this process, the
slabs are heated and soaked in a walking beam
type reheating furnace up to a temperature of Figure 3. Measured roll force for one of the rolling
1250°C. The heated slab is then reduced to a schedules of stand No 3.
desired thickness at the finishing mill with seven
stands. From the finishing stand, the strip goes to In order to measure the wearduring a roll
the run out table (ROT), where it is water cooled program, the roll profile before and after a rolling
and finally coiled in down coiler. The finishing schedule are measured by Herkules WS450
mill is equipped with continuous variable crown machine shown in Figure 4.
(CVC) technology and work roll bending for shape
control. The roll force, the roll wearing crown and
the roll thermal crown were measured in this plant.

Figure 2. Hot rolling mill of Mobarake Steel Complex Figure 4. Herkules WS450 machine for measuring the
(MSC) rolls wearing crown

IJE Transactions C: Aspects Vol. 25, No. 1, March 2012 - 81

 First, cooled rolls are put in this machine and measured wearing crown and the obtained thermal
measuring sensors measure the roll diameter. crown in a sample rolled schedule for stand No. 3.
Then, the control system programs the machining Downfall of these diagrams is due to changing the
steps. After machining, the grinded roll profile is work roll during the program. The wearing crown
measured and is recorded in the system. The roll and the roll thermal crown also vary during a
profile before and after a rolling program measured rolled schedule and may be a function of strip
by this machine is shown in Figures 5 and 6, width, strip initial thickness, reduction, roll
respectively. The roll wearing crown can be velocity, strip initial temperature, strip exit length
measured by comparing the diameter profile of roll and some other parameters. These empirical data
which is shown in these figures. were used to obtain a linear regression model for
calculating the roll force, the roll wearing crown
and the roll thermal crown in the mentioned mill.
Roll Profile (mm)

Roll Length (mm)

Figure 7. Measured wearing crown for one of the
Figure 5. Measured roll profile before one of the rolling rolling schedules of stand No. 3.
schedules for stand No. 6.
Roll Profile (mm)

Figure 8. Measured thermal crown for one of the

rolling schedules of stand No. 3.

Roll Length (mm)

Figure 6. Measured roll profile after one of the rolling

schedules for stand No. 6. 5. LINEAR REGRESSION

Linear regression refers to any approach to model

The thermal crown model provided by hot the relationship between one or more variables
rolling stands manufacturer which is used in Level denoted y and one or more variables denoted x,
2 control of the roll line is used for obtaining the such that the model depends linearly on the
thermal crown data. Figures 7 and 8 show the unknown parameters to be estimated from the data.

82 - Vol. 25, No. 1, March 2012 IJE Transactions C: Aspects

Such a model is called a “linear model”. Linear model quality, R-Square coefficient is used. This
regression focuses on the conditional probability coefficient is a simple method for measuring the
distribution of y given x, rather than on the joint regression fitting line. When the model fits to data
probability distribution of y and x, which is the precisely, then the value of R 2 is equal to 1. Using
domain of multivariate analysis. Linear regression this coefficient as a criterion is not an appropriate
was the first type of regression analysis to be method to investigate the model quality. That is
studied rigorously, and to be used extensively in because of the fact that it rises by adding xi to the
practical applications. This is because models model. Therefore, the adjusted R-Square
which depend linearly on their unknown coefficient is used instead [15].
parameters are easier to fit than models non-
linearly related to their parameters. The statistical
properties of the resulting estimators are 6. MODELING THE WORK ROLL CROWN
determined easier. Regression models are usually EFFECTIVE PARAMETERS
expressed as an algebraic function. If the equation
is expressed as a linear structure due to the 6.1 Roll Force Linear Regression Model Roll
described variables, xi , a good model can be force is one of the main rolling parameters which
written as below: affect the strip profile. Due to the nature of rolling,
the roll force may be a function of different
y = β 0 + β1 x1 + β 2 x2 + ... + β k xk + ε (4) parameters. This was investigated for different
MSC rolled schedules and it showed that the strip
In this equation, βi are unknown constants that are width, initial thickness, and initial temperature, the
called regression coefficients and ε expresses the reduction, the roll diameter, the roll velocity and
models errors. For investigation of model quality, strip exit length are the main parameters which
nomination coefficient is used. This coefficient is affect the roll force in the mill. Different
shown by R 2 and is a simple method for measuring parameters for some portion of a sample rolled
the regression fitting line. Considering Equation schedule (stand No. 3) are shown in Table 1. These
(4), if total variation in data (SSTotal) is divided to empirical data were investigated for this stand
explained variation by dependent variables using SPSS software and linear regression method.
(SSReg) and the variations that are not explained Finally, the results showed that the roll force is just
by them (SSError), then: a function of strip width, the reduction and the
strip temperature for stand No. 3. The following
=
SS SS Re g + SS Error (5) formula is then proposed for stand No. 3:
Total

So, F = A+ B×W +C× r + D×T (7)

SS Re g SS Error (6) where F is roll force in (MN), W is strip width in
R2 = = 1−
SSTotal SSTotal (mm), r is reduction in (%) and T is temperature in
(°C). A, B, C and D are unknown coefficients and
must be predicted by linear regression. Table 2
It is obvious that 0 ≤ R 2 ≤ 1 . When the model shows the model summary. As this table shows,
fittsthe data precisely, R 2 = 1 . The nomination the model quality is 99.3%. It means that the
coefficient is not an appropriate method for quality independent variables can explain 99.3% of the
investigation of a model. It rises due to adding xi dependent variable variations. Durbine-Watson test
to the model. Therefore, edited nomination value is obtained 1.737. That is, in the range of 1.5
coefficient is used instead. and 2.5 and shows non-auto-correlation between
Linear regression has some assumptions. These the errors. Fisher test is also performed for this
assumptions must be satisfied in order to obtain an model and the results are shown in Table 3.
acceptable model. There are some tests which Considering the value of this test and the
investigate these assumptions. These tests are probability value shown in the fifth and in the sixth
Durbine-Watson d test, t-student test, Fisher test columns of this table, the model significance is
and Multicolinearity test. For investigation of the satisfied up to 100%. Table 4 shows the model

IJE Transactions C: Aspects Vol. 25, No. 1, March 2012 - 83

coefficients and t-student test provided for each of Normal P-P plot of regression is shown in Figure
variables. As this table shows, the exact value of A 9. As this figure depicts, the actual data and
is equal to 9.650446004387064, B is equal to predicted data are close to each other.
0.014217686021280157, C is equal to -
0.030358343243534713 and D is equal to
0.4827355398395615. The obtained model is so
sensitive to its coefficients. Therefore, exact values
of A, B, C and D must be obtained and used in the
model. It is obvious that values of B and D are
positive and the value of C is negative. This shows
that the roll force has a direct relation with strip
width and reduction and a reverse relation with
temperature. These relations are adopted with the
nature of hot rolling process based on hot rolling
experts’ experience as well. Considering the fifth
and the sixth columns of Table 4, the t-student test
and the probability values are shown. Therefore, it
can be concluded that the model coefficients
significance is satisfied up to 100%. Table 5 shows
the eigenvalue and condition index of the model.
As this table shows, condition index is below 30 Figure 9. Normal P-P plot of regression standardized
and it shows non-autocorrelation in this model. residual for roll force model

TABLE 1. Different parameters for a part of a sample rolled schedule (stand No. 3)

Program W (mm) H (mm) T (°C) r (%) Vr (m/s) Le (m) F (MN)

16001 1041.8 14.59 944.471 33.072 2.499 139.1364 12.277
16002 1046.6 14.589 987.246 33.07 3.226 278.2823 14.616
16003 1045.6 14.594 1007.479 33.077 3.779 417.3806 12.401
16004 1043.5 14.592 935.269 33.074 2.156 556.4979 11.993
16005 1042.9 14.59 934.589 33.071 2.386 695.6343 12.299
16006 1025 14.59 924.213 33.072 2.375 834.7707 12.732
16007 1049.9 14.592 1021.432 33.074 2.116 973.888 12.466
16008 1044.4 14.591 937.561 33.073 2.305 1113.015 12.152
16009 1025 14.587 981.419 33.066 4.553 1252.18 10.776
16010 1045.5 14.59 938.047 33.071 2.482 1391.316 13.104
16011 1191.8 13.526 928.494 36.616 2.896 1541.398 16.737
16012 1190.9 13.527 937.681 36.616 2.761 1691.468 14.836
16013 1201.8 13.856 943.912 35.883 3.26 1837.975 16.08
16014 1200.2 13.858 944.098 35.886 3.546 1984.461 15.101
16015 1231.1 15.596 936.458 31.045 2.802 2114.622 16.036
16016 1232.6 15.596 956.44 31.045 2.659 2244.784 13.19
16017 1232.6 15.593 956.475 31.04 2.826 2374.97 13.425
16018 1232.8 15.596 960.795 31.045 2.8 2505.132 13.297
16019 1231 15.597 949.166 31.046 2.826 2635.285 13.948
16020 1232.8 15.595 948.33 31.044 2.741 2765.455 13.269
16021 1232.2 15.596 958.459 31.045 2.505 2895.617 12.757
16022 1232.6 15.595 966.24 31.043 2.698 3025.787 13.236
16023 1289.7 15.998 927.753 30.254 2.701 3152.677 17.283
16024 1541.5 15.999 951.609 30.236 3.27 3279.56 18.67
16025 1542.2 16.001 980.504 30.239 2.884 3406.427 14.627

84 - Vol. 25, No. 1, March 2012 IJE Transactions C: Aspects

 TABLE 2. Roll force model summary
Adjusted R Std. Error of the
Model R R Square Durbin-Watson
Square Estimate
1 99.3% 98.6% 98.5% 0.37386 1.737

TABLE 3. Roll force model anova and fisher test

Model Sum of Squares df Mean Square F Sig.
Regression 589.08 3 196.36 1405 .000
Residual 8.526 61 0.14
Total 597.607 64

TABLE 4. Roll force model coefficients and t-student test

Standardized
Unstandardized Coefficients
Model Coefficients t Sig
B Std. Error Beta
(Constant) 9.56 3.001 3.216 .002
W .014 .000 1.256 63.231 .000
T -0.03 .003 -0.156 -9.642 .000
r 0.483 0.01 0.964 48.806 .000

TABLE 5. Roll force collinearity diagnostics test

Condition Variance Proportions
Model Dimension Eigenvalue
Index (Constant) W T r
1 3.924 1.000 .00 .00 .00 .00
1 2 .068 7.605 .00 .23 .00 .14
3 .008 22.374 .01 .76 .01 .82

6.2 Work Roll Wearing Crown Linear six millimeter. Obtaining final strip thickness more
Regression Model Work roll wearing crown than six millimeters, the wearing crown is almost
causes the changes in its crown so that it affects constant. Therefore, the formula for predicting the
the strip profile. Investigation of different MSC wearing crown in stand No. 3 is obtained as below:
rolled schedules showed that the wearing crown
may be a function of strip width, strip initial  A3 A6 A8
 A1 + A2T + W + A4 Le + A5 LnLe + L + A7 Le + P hf  4
3

thickness, strip initial temperature, reduction, roll  e

WC =  A9 + A10T + A11 Le + A12 LnLe + A13 r 3 + A14 P 3 4 ≤ hf ≤ 6
diameter, roll velocity and strip exit length. This  0.23 hf  6
was done for stand No. 3 using SPSS software and 

empirical data by MSC. The results show that the
(8)
wearing crown is extremely dependent on final
strip thickness which is the exit strip thickness where WC is wearing crown in (mm), P is roll
from stand No. 7. Therefore, three different models force in (MN), T is temperature in (oC) (), Le is strip
were obtained to predict the work roll wearing exit length in (mm), W is strip thickness in (mm), r
crown based on the final strip thickness. The is reduction in (%) and hf is strip exit thickness
results show that the wearing crown is a function from stand No. 7 in (mm). The model coefficients
of temperature, strip width, exit strip length and were obtained utilizing regression method, and are
roll force for final strip thickness, which is below shown in Table 6. All the regression tests
four millimeter. This is a function of temperature, explained in part 6.1 were also performed for the
strip exit length, reduction and roll force for final wearing crown model and the model could satisfy
strip thickness in the range of four millimeter and all of them.

IJE Transactions C: Aspects Vol. 25, No. 1, March 2012 - 85

 TABLE 6. Coefficients of wearing crown model
Coefficient Value Coefficient Value
A1 4.446398643639322 A8 1.7813654875651208
A2 -7.406187922551309×10-4 A9 -0.12254452561551915
A3 -217.9487385310034 A10 -6.095028372882652×10-4
A4 7.378810709361287×10-8 A11 -3.7482552082865447×10-8
A5 -0.2432463157114965 A12 0.09686673746775742
A6 -104035.62134222736 A13 -8.535071529717156×10-6
A7 -1.9516456622854154×10-22 A14 -1.3473217778631586×10-5

TABLE 7. Coefficients of thermal crown model

Coefficient Value Coefficient Value
A1 2.087995981026224 A8 -2.027408962491758×10-8
A2 -4.3215934755072826×10-4 A9 0.05454461791246326
A3 -0.010963153164038472 A10 2.237107286345748×10-22
A4 -5.899123888367267 A11 6.3766386654949×10-6
A5 -4.5219554744054533×10-4 A12 -0.5895695243278343
A6 -550.337563085544 A13 0.0043830898513108835
A7 -4.601218736687413×10-11

6.3. Work Roll Thermal Crown Linear length in (mm), W is strip width in (mm), h is strip
Regression Model Work roll thermal crown exit thickness in (mm) and hf is strip exit thickness
causes the changes in its crown so that it affects from stand No. 7 in (mm).
the strip profile. Investigation of different MSC The model coefficients were obtained utilizing
rolled schedules showed that this parameter also regression method and are shown in Table 7. All
may be a function of the strip width, initial regression tests explained in part 6.1 were also
thickness, and initial temperature, the reduction, performed for the thermal crown model and the
the roll diameter, the roll velocity and the strip exit model could satisfy all of them.
length. This was investigated for stand No. 3 using
SPSS software and provided data by MSC. The 6.4. Work Roll Bending and Flattening Crown
results show that the thermal crown is extremely Analytical Model The deformation of work roll
dependent on final strip thickness, which is the exit due to bending and flattening is calculated utilizing
strip thickness from stand No. 7. Therefore, two these equations:
different models were obtained to predict the work
roll thermal crown based on the final strip d 2 yb
EI =M (10)
thickness. The results show that the thermal crown dx 2
is a function of temperature, roll force, strip exit
thickness, width, length and the final thickness GA dys
=S
(11)
which is below six millimeters. Obtaining final K dx
strip thickness more than six millimeters, the
thermal crown is almost constant. Therefore, In this equation E is Young module (215GPa), G is
formula for predicting the thermal crown in stand shear module (82.7GPa), K is 4/3, and I is moment
No. 3 is obtained as below: of inertia ( I = π d 4 64 ). Meanwhile, d is the work
 A4 A6 roll diameter. Moreover, M and S are the bending
 A1 + A2T + A3 P + h + A5 h + W + A7W + A8 Le + h ≤ 6
3 3

 moment and the shear force in the beam cross

TC =  f

 A9 LnLe + A10 Le + A11 F + A12 Lnh f + A13 h f

3 3 3 section, respectively.

 0.46 hf  6 By solving these two equations, total deformation
(9) is calculated as:
where TC is wearing crown in (mm), P is roll force
in (MN), T is temperature in (oC), Le is strip exit ydef= yb + ys (12)

86 - Vol. 25, No. 1, March 2012 IJE Transactions C: Aspects

 7. PREDICTION OF WORK ROLL INITIAL 8. CONCLUSION
CROWN
The roll force, work roll wearing crown, work roll
Prediction of work roll initial crown according to thermal crown and work roll bending and
desired strip profile is calculated utilizing a flattening crown are the main factors of rolling
computer programing written in this research. This process. All of them cause changes in the roll
program is able to predict the work roll initial crown, and finally, they affect the strip profile.
crown according to desired strip profile for some Many analytical methods have been presented for
portion of rolled schedule with a specific error predicting them which did not match very well
obtaining by the management information system with the actual data in the MSC. Therefore, new
(MIS) of MSC. This was performed for one case methods based on the regression analysis were
with known desired crown, and the error and the employed for calculating the roll force and the roll
work rolls initial crown was calculated. Table 8 wearing in the mentioned mill. The results showed
shows the initial crown predicted for the that empirical models could predict physical
mentioned case. These crowns were implemented quantities better than mathematical models,
for a particular schedule and the measured strip particularly for the mentioned mill.
profile shown in Figure 10. The comparison Finally, the work roll initial crown according to
between the measured strip crown and the obtained desired strip profile was obtained. The results were
one is shown in Table 9, which shows a fairly good applied for a particular rolling schedule and the
agreement between them. obtained strip profile showed a good agreement
with the measured strip profile for that mill.

9. ACKNOWLEDGMENT

The authors appreciate the Hot Rolling Mill Section

Strip Profile (micron)

of Mobarake Steel Complex for their nice

cooperation.

10. REFERENCES
Strip Width (mm)
1. Liu, C., "Simulation of strip and slab Rolling using an
Figure 10. Variation of measured strip profile during elastic-plastic Finite-element Method", PhD thesis,
the width University of Birmingham, (1985).
2. Chakraborti, N., Siva Kumar, B., Satish Babu, V.,
Moitra, S. and Mukhopadhyay, A., "Optimizing surface
profiles during hot rolling: A genetic algorithms based
multi-objective optimization", Computational
Materials Science, No. 37, (2006), 159-165.
TABLE 8. Work roll initial crown in different stands 3. Tudball, A. and Brown, S.G.R., "Practical finite element
heat transfer modeling for hot rolling of steels",
Stand F1 F2 F3 F4 F5 F6 F7 Institute of Materials, Minerals and Mining, No. 33,
initial crown (2005), 61-66.
70 90 -110 -83 -65 -107 -6
(mm) 4. Johna, S., Sikdara, S., Swamya, P. K., Dasa, S. and
Maityb, B., "Hybrid neural–GA model to predict and
minimize flatness value of hot rolled strips", Journal of
Materials Processing Technology, No. 195, (2008),
314-320.
TABLE 9. Comparison of the measure crown and the
5. Nandan, R., Rai, R., Jayakanth, R., Moitra, S.,
obtained crown Chakraborti, N. and Mukhopadhyay, A., "Regulating
crown and flatness during hot rolling: A multiobjective
measured strip obtained strip Percent of optimization study using genetic algorithms", Journal
crown crown error of Materials Processing Technology, No. 20, (2005),
50 46 8% 459-478.

IJE Transactions C: Aspects Vol. 25, No. 1, March 2012 - 87

6. Rossomando, F.G. and Dentifilho, J., "Modeling and of Materials Processing Technology, Vol. 206, (2008),
Control of a Hot Rolling Mill", Latin American 264.
Applied Research, No. 36, (2006), 199-204. 11. Oike, Y., Okubo, I., Hirano, H. and Umeda, K., “Tetsu-
7. Kazeminezhad, M. and Karimi Taheri, A., "Calculation to-Hagane”, 63, S222, (1977).
of the rolling pressure distribution and force in wire flat 12. Sikadar, S., John, S., Pandit, A. and Dasu, R., "Analysis
rolling process", Journal of Materials Processing of roll stack deflection in a hot strip mill", Journal of
Technology, No. 171, (2006), 253-258. the Brazilian Society of Mechanical Science and
8. Guo-min, Z., Hong, X. and Chun-hua, W., "Three- Engineering, No. 3, (2008), 284-289.
Dimensional model for strip hot rolling", Journal of 13. Siciliano, F., Minami, K., Maccagno, T. M. and Jonas,
Iron and Steel Research, No. 13, (2006), 23-26. J. J., "Mathematical modeling of the mean flow stress,
9. Younes, M. A., Shahtout, M. and Damir, M. N., "A fractional softening and grain size during the hot strip
parameters design approach to improve product quality rolling of C-Mn steels", (1996), 1500-1506.
and equipment performance in hot rolling", Journal of 14. Gujarti, D. N., "Basiceconometric", 4th ed., McGraw-
Materials Processing Technology, (2006), 83–92. Hill, USA, (2003).
10. Malik, A.S. and Grandhi, R.V., "A computational 15. Rencher, A.C., "Linear models in statistics", Wiley,
method to predict strip profile in rolling mills", Journal USA, (2000).

88 - Vol. 25, No. 1, March 2012 IJE Transactions C: Aspects