ISIJ International, Vol. 45 (2005), No. 8, pp. 1173–1178
Mathematical Model for the Thin Strip Cold Rolling and Temper
Rolling Process with the Influence Function Method
Yuli LIU and Won-Ho LEE1)
Dept. of Process Eng. & Dev., QUAD Eng. Inc., 75 Scarsdale Road, Toronto, Ontario, M3B 2R2, Canada.
1) POSCO Tech. Res. Lab., Pohang P.O. Box 36. 790-785, Korea.
(Received on February 8, 2005; accepted on April 26, 2005 )
A mathematical model for the thin strip cold and temper rolling process has been developed using the influence function method. By solving the equations describing the roll gap phenomena in a unique procedure
and considering more influence factors, the model offers significant improvements in accuracy, robustness
and generality of the solution for the thin strip cold and temper rolling conditions. The relationship between
the shape of the roll profile and the roll force was also discussed. Calculation results show that any change
increasing the roll force may result in or enlarge the central flat region in the deformation zone. Applied to
the temper rolling process, the model can well predict not only the rolling load but also the large forward
slip. Therefore, the measured forward slip, together with the measured roll force, was used to calibrate the
model. The model was installed in the setup computer of a temper rolling mill to make parallel setup calculations. The calculation results showed good agreement with the measured data and the validity and precision of the model were proven.
KEY WORDS: steel cold rolling; mathematical model; roll force.
rolling process in which there are relatively large entry and
exit elastic zones.
(2) The pressure profile iterative loop was taken as the
basic iterative loop. The boundaries of different zones are
determined with an extra iterative loop. Under some circumstances, it is very difficult to determine the boundaries
of different zones5) because the iterative processes to determine the boundaries would diverge or not converge to
unique values.
(3) The solution procedure was divided into three
regimes, namely, Regime I (with no central flat zone),
Regime II (with a small central flat zone and one pressure
peak) and Regime III (with a large central flat zone and two
pressure peaks). For the different regimes there are different
algorithms. Before using the model to calculate a certain
rolling case, we have to decide which regime the certain
rolling case belongs to.
Although the boundaries between the different regimes
were given through multi-variable regression analysis in the
foil rolling case,3) it can not be applied to other cases.
Several applications of Fleck’s work have since followed
by Dixon et al.,6) Domanti et al.,7) Gratacos et al.,8)
Montmitonnet et al.,9) Haeseling et al.,10) and Liu et al.11)
Some improvements were made, but the above mentioned
problems remained unsolved.
In this paper, the aforementioned shortcomings were
eliminated. The strip elastic deformation at the entry and
exit was taken into consideration. The roll profile iterative
loop was taken as the basic loop, so that the boundaries of
different zones can be directly determined according to the
1. Introduction
Severe cases of strip rolling, such as thin strip cold
rolling and temper rolling etc., cause well known problems
with simplified classic circular arc models. They may have
convergence difficulties or bad sensitivity to process parameters. In order to solve these problems, some researchers
have attempted to develop more realistic models by the use
of influence functions describing the roll deformation (see,
for example, Jortner et al.,1) Grimble et al.2) These models
gave some prediction improvement, but they failed to converge for thinner or harder materials as ever.
A major development in modeling of the thin strip and
foil rolling process with the influence function method was
achieved by Fleck et al.3) By assuming that there is a region
of roll flattening where the roll surface is flat and parallel,
they solved the non-convergence problems. In the flat region, no further reduction takes place and the shear or frictional stresses at the roll/strip interface remain at values
below that predicted by the Coulomb friction law.
Therefore, the roll pressure in this region can be obtained
by inverting the roll profile to roll pressure relationship defined by the roll deformation influence function.
This model has provided useful reference results, however it has several shortcomings:
(1) The strip elastic deformation zones at the entry and
exit were neglected, which results in an under estimation of
the load. Because of the ‘elastic plug’ effect,4) the predicted
load by this model may be much lower than the actual one.
Besides, the model could not be used in light reduction
1173
© 2005 ISIJ
ISIJ International, Vol. 45 (2005), No. 8
roll profile. The extra boundary determination iterative loop
was eliminated, and it was relatively easy to determine the
boundaries of the different zones. Besides, which regime
the rolling case belongs to could also be directly determined by observing the shape of the roll profile. Therefore,
it would not be necessary to determine the regimes and
choose the calculation procedure before hand.
After calibration using the production data, the newly developed model was applied to the actual temper rolling
process and the validity and precision of the model were
also proven.
2. Main Equations
A geometric sketch of the deformation zones is shown in
Fig. 1. The geometric parameters are defined with a polar
coordinates system.
The roll-gap model is based on two relationships which
enable a gauge profile to be calculated from a known pressure profile and conversely a pressure profile to be calculated from a gauge profile. The elastic deformation of the
work roll is related to the roll pressure via the linear integral equation
∫
Fig. 1.
differential equations. In the entry and exit elastic regions it
is2):
dp
2a µpν ⫹ p(1⫺2ν ) tan β
2(h⫺ h1 ) tan βE
⫽⫺
⫹
dθ
h
1⫺ν
h1 (1⫺ν 2 )
...........................................(4)
φ
a(θ ) ⫽
U (θ ⫺ t ) p(t )dt ⫹ R ...................(1)
0
where t :
a(q ) :
R:
p(t) :
U(q ⫺t) :
and, applying the yield criteria of the difference of vertical
and horizontal stresses equal to yield stress in the plastic regions2):
integration variable, which varies from 0 to f ;
work roll radius at point q ;
undeformed radius of work roll;
Normal pressure profile;
Jortner’s influence function9,10);
dp
dy
2a
................(5)
⫽⫺
( µp⫺ y tan β ) ⫹
dθ
h
dθ
where n : Poisson’s ratio;
E : Young’s modulus of elasticity;
h1 : unstrained gauge in a elastic region; (h⫺h1)/h1
is the elastic strain in vertical direction;
R∆φ
U (θ ⫺t ) ⫽
πE
1⫺ cos(θ ⫺t )
⫻ (1⫺ν 2 )cos(θ ⫺t ) ln
⫹
2
1⫹ cos(θ ⫺t )
β⬇
⫺ (1⫺ν ⫺2ν 2 ) sin(θ ⫺t )
dh
;
2adθ
y : yield stress;
m : friction coefficient between rolls and the strip.
In the central flat region, the discrete pressure profile can
be obtained by solving the following linear equation system
(all parameters are expressed in discrete form):
1⫺ cos(θ ⫺t )
1⫹ cos(θ ⫺t )
⫻ tan⫺1
⫹ tan⫺1
sin(θ ⫺t )
sin(θ ⫺t )
...........................................(2)
j2
where D f : segmental angle.
The gauge profile is determined by the deformed roll radius according to the equation
∑ (U
h(q )⫽Rs⫺2a(q ) cos(q ⫺f 0) ...................(3)
⫺
i , j ⫺U e , j ) p j ⫽
j1
Rs ⫺ hi
Rs ⫺ he
⫺
2 cos(θi ⫺φ0 ) 2 cos(θ e ⫺φ0 )
j1⫺1
∑
d
where h(q ) : gauge profile;
Rs : distance between centers of the work rolls;
f 0 : angle measured to the line of work roll centers.
In above equations, the f 0 is a pre-determined value,
which must be larger than possible contact angle after roll
centers considering roll flattening and elastic recovery.
Otherwise, the calculation will fail. However, larger f 0 will
increase the calculation time. Normally take f 0⫽
(0.5⬃1.5)√苶
(h1⫺h2)/R will be sufficient.
The pressure profile is related to the gauge profile via
© 2005 ISIJ
Sketch of the deformation zones in the roll gap.
e
(U i , j ⫺U e , j ) p j ⫺
∑ (U
i , j ⫺U e , j ) p j
i ⫽ j1,L, j2
j 2⫹1
...........................................(6)
where j1 : the first node in the flattened region;
j2 : the final node in the flattened region;
e : the node at entry plane;
d : the node at exit plane.
The intersections of roll profile, determined by a(q ), and
strip profile h(q ), will determine the node at entry plane
and exit plane during iteration.
Once the pressure and gauge profiles are determined, the
roll force, P, and torque, T, can be calculated through inte1174
ISIJ International, Vol. 45 (2005), No. 8
Fig. 2.
The calculation procedure of new model.
gration of the pressure and friction force distribution:
∫
φ
P⫽
a(θ )
0
∫
Before applying the present model to actual temper
rolling process, the reliability and convergence of calculation result were checked with various rolling conditions.
Here, some typical calculated results are introduced. Figure
3 is an example of foil, which is ultra thin material, cold
rolling process calculation. Three reduction rates were used
in the calculation.
Though very thin entry thickness was used, there was no
convergence problem, in this model. It can be seen that as
the reduction increases from 50 to 70%, the pressure profile
changes from one peak (regime II) to two peaks (regime
III) , and the length of the flat region and the roll force are
also increase more than two times. It reveals that the same
calculation procedure could be used regardless of deformation regimes in the present model.
Figure 4 shows the effect of the friction coefficient upon
the gauge and pressure profile under temper rolling conditions. It can be seen that not only the pressure profile but
also the gauge profile is greatly influenced by the friction
coefficient. At low friction conditions, the contact arc between the roll and the strip is nearly circular, but at high
friction conditions, a flat region appears in the central part
of the contact arc.
Similar situation can also be seen when we change the
yield stress of the strip. As shown in Fig. 5, the contact arc
changes its shape from nearly circular to non-circular and a
central flat region finally appears, as the yield stress increases. Besides, the elastic recover zone is also greatly in-
cos β
[ p(θ ) ⫹ τ (θ ) tan β ]dθ ..........(7)
cos ξ
φ
T⫽
a 2 (θ )[τ (θ ) ⫹ p(θ ) tan ξ ]dθ ...............(8)
0
where x : angle between un-deformed roll and horizontal
axis.
At the same time, the forward slip, f, can also be approximately, by neglecting the minor influence of elastic recovery to forward slip, calculated with the following formula:
f⫽hn /h2⫺1 ................................(9)
where hn : strip gauge at a flat region or a neutral point;
h2 : strip gauge at the exit side.
3. Calculation Procedure and Analysis of New Model
The main flow chart of the calculation procedure is
shown in Fig. 2. The iterative calculation procedure was introduced to solve the equations describing the roll-deformation and the strip-stress distribution. No matter which
regimes the rolling conditions belong to, the same calculation procedure can be used. That is one of the most important advantages in the present model.
1175
© 2005 ISIJ
ISIJ International, Vol. 45 (2005), No. 8
Fig. 3.
Thickness and pressure profile of foil cold rolling R⫽89 mm, friction coefficient⫽0.03, s y⫽230 MPa, entry
thickness⫽0.03 mm, E⫽230 GPa, (a) reduction⫽30 %, (b) reduction⫽50 %, (c) reduction⫽70 %.
Fig. 4.
Thickness and pressure profile under different friction coefficients R⫽205 mm, s y⫽196 MPa, entry thickness⫽
0.505 mm, reduction⫽0.46 %, rolling speed⫽37 m/min, entry tension⫽2 496 kg, exit tension⫽3 616 kg, width⫽
1 073 mm, (a) m ⫽0.12, (b) m ⫽0.15, (c) m ⫽0.18, (d) m ⫽0.21.
Fig. 5.
Thickness and pressure profile under different yield stress, R⫽205 mm, entry thickness⫽0.505 mm, reduction⫽0.46 %, width⫽1 073 mm, rolling speed⫽37 m/min, entry tension⫽2 496 kg, exit tension⫽3 616 kg,
m ⫽0.12, s y⫽210 MPa, (b) s y⫽240 MPa, (c) s y⫽270 MPa, (d) s y⫽300 MPa.
fluenced by the yield stress.
Another factor, which can influence the thickness and
pressure profile, is the entry and exit tension. Figure 6
shows the thickness and pressure profile under different
tensions. As the tensions decrease, the pressure increases
and the length of the central flat region also increases.
The above mentioned calculation results show that not
only the decreasing of entry thickness, but also other factors, such as increasing the reduction, friction and yield
stress as well as decreasing tension, can result in a central
flat region. That means any factor that can cause an increase of the roll force may result in a central flat region in
the deformation zone.
Through the analysis of rolling phenomena, it was confirmed that the new model has a calculating stability and
can give physically reasonable result for wide range of temper rolling conditions.
© 2005 ISIJ
4. Model Calibration and Application Result
In order to use the theoretical model as a setup model in
the elongation control system of a temper rolling mill in
POSCO, the model was calibrated against the production
data collected from the same mill. The friction coefficient,
m , and the constrained yield stress, K, were taken as the calibration factors. Because the roll force, P, and the forward
slip, f, can be measured in the temper rolling mill, and they
can also be calculated by the theoretical model, we used
these two measured data to determine the calibration factors K and m . Equating the measured roll force Pm and forward slip fm with their calculation formula, which were derived from the theoretical model, we got two coupled equations. Solving the coupled equations with the iterative
method, we obtained the constrained yield stress, K, and the
friction coefficient m for each coil.
1176
ISIJ International, Vol. 45 (2005), No. 8
Fig. 6.
Thickness and pressure profile under different tensions, R⫽205 mm, entry thickness⫽0.505 mm, reduction⫽
0.46 %, width⫽1 073 mm, rolling speed⫽37 m/min, s y⫽196 MPa, m ⫽0.12. (a) Entry and exit tension⫽8 100 kg,
(b) entry and exit tension⫽5 400 kg, (c) entry and exit tension⫽2 700 kg, (d) entry and exit tension⫽0 kg.
Table 1.
Fig. 7.
The values of the calibration factors for each group.
Calculation of constrained yield stress and friction coefficients.
Because present model can well predict the large forward
slip in the temper rolling process, we got reasonable results
using the above method. A flow chart of the calibration calculation is shown in Fig. 7.
After the constrained yield stress and friction coefficients
were calculated for all coils, we divided the coils into
groups according to the steel grades and the constrained
yield stress. The average constrained yield stress and average friction coefficient of each group were used as the final
values of the calibration factors for this group. Table 1
shows the values of the calibration factors for each group.
After calibration, the model was installed in the plant together with the data acquisition system to make parallel
calculations and comparisons. An evaluation sample that
consists of the valid data of 3 724 coils is adopted to make
the evaluation and comparison.
Figure 8 shows a comparison of the measured roll forces
and the roll forces calculated by a present model. It can be
seen that the predicted roll forces match the measured values well and the model has relative high prediction accuracy. A statistical analysis shows that the correlation coeffi-
Fig. 8.
Roll force measured and predicted by the present model.
cient between the measured roll forces and the roll forces
predicted by the present model equals 0.8899.
In order to show the prediction accuracy improvement
made by the present model, the regression model, which
had been used as the setup model in the same temper
rolling mill before the present model was developed, was
also evaluated with the same data. Figure 9 shows a comparison of the measured roll force and the roll force calculated by the regression setup model.
As we can see, the samples in Fig. 9 are more scattered
than those in Fig. 8. The correlation coefficient between the
measured roll force and the roll force calculated by the regression setup model is 0.7175. Besides, the average calculated roll force is lower than that of the measured roll force.
This indicates that the regression setup model underestimated the roll force in general. By comparison, we can see
that the present theoretical model is much more accurate
than the regression setup model. From a comparison with
1177
© 2005 ISIJ
ISIJ International, Vol. 45 (2005), No. 8
Fig. 9.
fore, it would not be necessary to determine the regimes
and choose the calculation procedure before hand.
By solving the equations describing the roll gap phenomena in a unique procedure, the new model offered significant improvement in the accuracy, robustness and generality of the solution for thin strip cold and temper rolling conditions.
The model was installed in the process computer of an
actual temper rolling mill to make setup calculations. The
calculation results showed good agreement with the measured data and the validity and precision of the model were
proven.
Roll force measured and predicted by the regression
model.
REFERENCES
the measured data and comparison with the old setup
model, the validity and precision of the present model are
proven.
1)
2)
3)
5. Conclusions
4)
5)
A new mathematical model for the thin strip cold and
temper rolling process has been developed by modification
of the influence function method. The main differences of
present model against to basic influence function method
are to be summarized as follows:
(1) The strip elastic deformation at the entry and exit
was taken into consideration.
(2) The roll profile iterative loop was taken as the basic
loop so that the boundaries of different zones can be directly determined according to the roll profile.
(3) The new calculation procedure was introduced to
apply the new model to various rolling conditions. There-
© 2005 ISIJ
6)
7)
8)
9)
10)
11)
1178
D. Jortner, J. F. Osterle and C. F. Zorowski: Int. J. Mech. Sci., 2
(1960), 179.
M. J. Grimble: GEC J. Sci. Technol., 43 (1976), No. 1, 3.
N. A. Fleck, K. L. Johnson, M. E. Mear and L. C. Zhang: Proc. Inst.
Mech. Eng. B, J. Eng. Manuf., 206 (1992), 119.
Q. Zhu: Proc. Adv. Technol. Plastic., 2 (1984), 1173.
W. Y. D. Yuen, D. N. Nguyen and D. L. Matthews: 37th MWSP
Conf. Proc., Vol. XXXIII, ISS, Warrendale, PA, (1995), 165.
A. E. Dixon and W. Y. D. Yuen: Proc. Computational Techniques and
Applications Conf. 95, Melbourne, Australia, (1995), 239.
S. A. Domanti, W. J. Edwards, P. J. Thomas and I. L. Chefneux:
Proc. 6th Int. Rolling Conf., VDEh, Duesseldorf, Germany, (1994),
422.
P. Gratacos and F. Onno: Proc. 6th Int. Rolling Conf., VDEh,
Duesseldorf, Germany, (1994), 441.
P. Montmitonnet, E. Massoni, M. Vacance, G. Sola and P. Gratacos:
Ironmaking Steelmaking, 4 (1993), 254.
G. Haeseling, S. Kastner, A. Kramer and H. G. Hartung: MPT Int.,
No. 6, (1998), 86.
Y. Liu and W. Lee: KSME Int. J., 2 (2001), 225.