JPH08243614A - Reversing rolling method excellent in accuracy of shape and thickness - Google Patents

Abstract

PURPOSE: To obtain a new rolling method for obtaining a product sheet having good flatness and highly accurate thickness in rolling using reversing rolling mill. CONSTITUTION: At the time of executing reversing rolling according to a predetermined pass schedule, at a 1st pass of the reversing rolling, initialization of shape control function in accordance with the target rolling load which is determined from a load estimating equation and correction of screw down position for making actual rolling load during rolling to coincide with the target rolling load are executed. Rolling is executed while respectively executing learning of parameters of the load estimating equation. As for passes on and after a 2nd pass, rolling is executed while repeatedly executing the correction of the pass schedule to the final pass based on the load estimating equation learned at the preceding pass, initialization of the shape control function in accordance with the target rolling load determined from the load estimating equation, correction of the rolling reduction position for making the actual rolling load to coincide with the target rolling load and learning of the parameter of the load estimating equation during rolling or just after rolling of each pass.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a reverse rolling method which is effective as a technique for improving flatness of sheets and accuracy of final sheet thickness in reverse rolling suitable for rolling difficult-to-roll materials such as stainless steel. It is about.

[0002]

2. Description of the Related Art As a conventional rolling method using a reverse rolling mill, a technique disclosed in Japanese Patent Publication No. 5-41332 is known.

In this conventional method, the delivery side plate thickness when the rolling load in each pass is set to the maximum allowable rolling load is sequentially calculated from the first pass, and the number of passes is determined under the condition that the finished plate thickness is less than that. Furthermore, the rolling schedule is determined by correcting the rolling reduction of each pass so that the delivery thickness at the number of passes matches the finishing thickness.Here, the allowable maximum rolling load is the reverse rolling mill body. Is defined as the load within the load limit of the mechanical device and within the range where the plate shape is good, and is set within the rolling reduction range in which abnormal rolling phenomena such as heat streak and chattering do not occur.

This method realizes a good plate shape by rolling at a load equal to or lower than the maximum allowable rolling load, and P = P (H, h, μ, k _{f, t f, t b,} R) ( where, P: rolling load, H: thickness at entrance side, h: thickness at delivery side, mu: friction factor, k _f: deformation resistance, t _f: exit side tension,
The function (rolling load formula) of t _b : entrance side tension, R: roll radius) was used.

In the above method, as the shape control means for obtaining a good shape, the operation amount of the shape control function (shape actuator) is y = y (P, Cw, B, ...) (where , Y: manipulated variable of shape control function, P: rolling load,
(C _W : initial crown of work roll, B: plate width)
And that it is given as a function form of rolling conditions.

Further, Japanese Patent Publication No. 58808/1993 discloses a method for setting a rolling position in consideration of preventing accidents due to poor flatness of rolled material when rolling cold-rolled material having a thin plate thickness. .

According to the method disclosed herein, first, a pass schedule is calculated, a delivery side plate thickness, a load, and a tension at a steady portion of each pass are calculated, and then a load at the start of rolling (target load) is calculated by a rolling load model formula. The initial position is set by correcting the reduction position based on the deviation between the target load and the actual load.By this method, the reduction position is corrected so that rolling can be performed at the target load. Since it is rolled by rolling, it is said that the defective shape of the plate due to the increased load can be avoided.

[0008]

However, the above-mentioned conventional method has the following problems. That is, according to the method disclosed in Japanese Patent Publication No. 5-41332, the rolling conditions of the next rolled material are the inlet plate thickness, the outlet plate thickness friction coefficient, the deformation resistance, the outlet tension, the inlet tension, and the roll radius value. The predicted rolling load of the next rolled material is calculated from the rolling load formula, but the deformation resistance and friction coefficient of the load formula differ depending on the steel type of the rolled material, the difference in the previous process, the lubrication conditions, etc. As a result of the calculation, an error was inevitable in the actual operation between the predicted load value calculated by this load prediction formula and the actual load value when actually rolled.

The cause is, for example, that the deformation resistance changes due to subtle differences in the composition of the same steel type, or subtle differences in the processing in the previous process, and changes in lubrication conditions due to wear of the outside air temperature roll. It is because the wear coefficient changes. Then, in this method, in order to determine the operation amount of the shape control function based on the predicted rolling load value obtained by the load prediction formula, the initial crown of the work roll, and the strip width (setup of the shape actuator), the load prediction formula and There was a problem that the error from the actual load would lead to a defective shape setup (a defective plate shape).

Further, Japanese Patent Publication No. 58803/1993 discloses that the rolling position is corrected so that the actual load matches the predicted load (target load) obtained by the load prediction formula to prevent the plate shape defect due to the increased load. However, Japanese Patent Fairness 5-41
Similar to Japanese Patent No. 332, since an error occurs in load prediction,
In order to match the actual load with the predicted load (target load) obtained from the load prediction formula, it is essential to correct the rolling position, so it is possible to maintain the condition set in advance to obtain the desired plate thickness. There was a big problem that the sheet thickness disappeared, leading to defective thickness.

An object of the present invention is to propose a novel rolling method capable of obtaining a product plate having good plate flatness and high plate thickness accuracy in rolling using a reverse rolling mill. .

[0012]

The present invention is optimal based on the base plate thickness, finish thickness, plate width, material, etc. of the material to be rolled and the load limit, reduction ratio limit, motor power limit, etc. for each pass. The number of passes, the optimal rolling reduction in each pass, the optimal delivery side plate thickness, and other pass schedules are determined. When performing reverse rolling according to this pass schedule, the first pass of reverse rolling is the target obtained from the load prediction formula. The initial setting of the shape control function according to the rolling load, the correction of the rolling position that matches the actual rolling load during rolling to the target rolling load, and the learning of the parameters of the load prediction formula, and rolling Regarding, the correction of the pass schedule up to the final pass based on the load prediction formula learned in the previous pass and the initial setting of the shape control function according to the target rolling load obtained from the load prediction formula. In addition, the shape and plate thickness control is characterized in that the rolling position is corrected while the rolling position is corrected so that the actual rolling load matches the target rolling load, and the parameters of the load prediction formula are repeatedly learned during or immediately after rolling in each pass. This is an excellent reverse rolling method (first invention), and in the above reverse rolling, learning parameters of the load prediction equation are provided with a learning term in the deformation resistance equation, and the learning term of this learning term is set so that the target rolling load matches the actual rolling load. Changing the value (second invention), and changing the value of the friction coefficient so that the target rolling load matches the actual rolling load in learning the parameters of the load prediction formula (third invention).
Is particularly advantageous.

Further, according to the present invention, the base plate thickness, the finished thickness, the plate width, the material, etc. of the material to be rolled, the load limitation for each pass, the rolling reduction limitation,
The first pass of reverse rolling is performed when determining the optimum number of passes, the optimal reduction ratio in each pass, the optimal delivery side plate thickness, and other pass schedules based on motor power limits, etc., and performing reverse rolling according to this pass schedule. The initial setting of the shape restriction according to the target rolling load obtained from the load prediction formula, the correction of the rolling position to match the actual rolling load with the target rolling load, and the rolling while performing the learning of the parameters of the load prediction formula, For the first pass after the second pass, the pass schedule is corrected to the final pass using the learned load prediction formula, and the initial setting of the shape control function according to the target rolling load calculated by the load prediction formula, Correct the rolling position to match the actual rolling load to the target rolling load and learn the parameters of the load prediction formula during or immediately after rolling. Rolling is performed while rolling back, and for the second half pass, the initial setting of the shape control function according to the target rolling load calculated by the learned load prediction formula, and the correction of the rolling position to match the set plate thickness and the actual plate thickness, This is a reverse rolling method (fourth invention) excellent in shape / thickness control, which is characterized in that rolling is performed while repeatedly learning the parameters of the load prediction formula during or immediately after rolling the pass. In the learning of the parameters of the load prediction equation, a learning term is provided in the deformation resistance equation, and the value of this learning term is changed so that the target rolling load matches the actual rolling load (fifth invention), or regarding the friction coefficient, It is particularly advantageous to change the value of the friction coefficient so that the target rolling load matches the actual rolling load (sixth invention).

[0014]

According to the present invention, the rolling is performed so that the target rolling load obtained from the load predicting formula and the actual rolling load match, so that the target rolling load can be set to the allowable maximum rolling load or less and the plate shape is good. It can be anything.

Further, since the shape control function is initially set based on the target rolling load, not only the optimum setting can be performed, but also the parameters of the load prediction formula are learned for each pass, so that the material to be rolled is It is possible to improve the load prediction accuracy with respect to, and the plate thickness accuracy improves as the number of passes increases (first invention).

Further, in learning the parameters of the load prediction equation, a learning term is provided in the deformation resistance equation, and the value of this learning term is changed so that the target rolling load matches the actual rolling load. Even if the deformation resistance of the material changes due to subtle differences in components and subtle differences in the processing in the previous process, the shape of the plate is not impaired, and
The plate thickness accuracy is improved (second invention).

When there is a concern that the friction coefficient may change due to changes in the lubrication condition due to the outside air temperature, roll wear, etc., regarding the friction coefficient, the value of the friction coefficient may be adjusted so that the target rolling load matches the actual rolling load. Can be changed, and the intended purpose can be advantageously achieved by this as well (third invention).

If the target rolling load calculated by the load prediction formula and the actual rolling load are matched in the first and second passes of the reverse rolling after the second pass, the target rolling load is set to the allowable maximum rolling load or less. It is possible to improve the shape of the plate.

Further, by making the initial setting of the shape control function based on the target rolling load, optimum setting is possible, and further, in the latter half pass, the reduction position is set so that the set plate thickness and the actual plate thickness match. By correcting and learning the parameters of the load prediction formula for each pass, the load prediction accuracy for the material to be rolled is improved, and the plate thickness accuracy is improved by the optimal initial setting of the shape control function based on the target rolling load. (4th invention).

Subtle differences in the composition of the same steel type,
Alternatively, if the deformation resistance changes due to subtle differences in the processing in the previous process, a learning term is provided in the deformation resistance equation when learning the parameters of the load prediction equation,
The value of this learning term may be changed so that the target rolling load matches the actual rolling load (fifth invention).

When the wear coefficient changes due to changes in lubrication conditions due to wear of the outside air temperature roll, the friction coefficient value is changed so that the target rolling load matches the actual rolling load. The flatness is improved and the plate thickness accuracy is improved (sixth invention).

[0022]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described more specifically below with reference to the drawings. Regarding the First Invention FIG. 1 is an example of a reverse rolling mill to which the present invention is applied, showing a configuration of a 12-high cluster rolling mill. In the first pass, the rolled material S is the payoff reel 1
Is paid out and wound on the right tension reel 2. After the second pass, the material S to be rolled is rolled through the mill 4 while being alternately wound by the tension reels 2 and 3. At this time, the rolling load is measured by the load meter 5, and the plate thickness is measured by the plate thickness meter 6. In this example, cold rolling of SUS 430 steel sheet that has annealing and pickling processes in the previous process is targeted, and an example of rolling a material with a sheet width of 1208 mm and a mother sheet thickness of 2.5 mm to a finish thickness of 0.610 mm in all 7 passes is explained. To do.

FIG. 2 shows a flowchart according to the present invention. In this example, first, the pass schedule up to the finish plate thickness (the output plate thickness of the final pass) is calculated. In the calculation of this pass schedule, the base plate thickness of the material to be rolled, the finished thickness, the plate width, the steel grade, and the like, the load limit for each pass, the rolling reduction limit, the optimal number of passes within the range of the motor power control, and Determine the rolling reduction and the outlet plate thickness for each pass.

FIG. 3 is an example of the outgoing side plate thickness of each pass obtained by calculation of the pass schedule. In this schedule, there are a total of 7 passes from the base plate thickness to the finish thickness, and rolling is performed according to this pass schedule. In rolling of each pass, first, a rolling load predictive calculation for the pass is performed according to a preset pass schedule. There are various load prediction formulas for load calculation, but the form is: P = P (H, h, μ, k _f , t _f , t _b , R) (where P: rolling load, H: thickness at entrance side, h: thickness at delivery side, mu: friction factor, k _f: deformation resistance, t _f: exit side tension,
t _b : entrance side tension, R: roll radius).

Next, the set values of the mill such as the rolling position in the pass are calculated. Then, based on the obtained predicted load, the shape control setting calculation for calculating the set value of the shape control actuator is performed.

As the shape control setting calculation, a table is created in advance from actual data, and an optimum setting value is searched from the table, and a relationship between the setting value of the shape control actuator and rolling conditions such as load is calculated. There is a method of defining it as the following function. A = A (P, B, H, h, R) (where A: shape control actuator set value, P:
Rolling load, B: Strip width, H: Inlet strip thickness, h: Outlet strip thickness,
R: Roll radius)

After the set value of the mill is calculated as described above, the mill is set and the rolling is started. At this time,
The rolling load is adjusted by adjusting the rolling position such that the rolling load P ′ actually measured by the load meter 5 matches the load P (target load) calculated by the prediction formula.

The accuracy of the load prediction formula is insufficient, and there is a difference between the predicted load for the plate thickness of the pass schedule (target plate thickness) and the actually measured load when the rolling position is set so that the plate thickness of the pass schedule is set. In such a case, the plate thickness h ′ actually measured by the plate thickness meter 6 has an error from the target plate thickness h.

In such a case, after the rolling of this pass is completed (if this is the final pass, the rolling of the coil is completed), then the rolling of the next pass is carried out. At this time, the parameters of the load prediction formula are learned based on the results of the previous pass.

For learning parameters, a learning term k is provided, and at P = P (k, H, h, μ, k _f , t _f , t _b , R), the actual load and the predicted load before the rolling position correction are Determine the value of k to match. As a method of determining the learning k, for example, in P = k × P (H, h, μ, k _f , t _f , t _b , R), k = α × P / P ′ (where α: learning The learning term k is determined by a coefficient, P: predicted load, P ′: actual load before rolling position correction, or at P = P (H, h, μ, k, t _f , t _b , R) + k, The learning term k can be defined by k = α × (PP ′).

The load prediction formula may be learned during rolling, and the pass schedule is corrected after the learning is completed.

FIG. 4 shows the plate thickness of the outgoing side plate of each pass after the second pass, which is obtained by correcting the pass schedule. In FIG. 4, ◯ is the pass schedule calculated before the first pass rolling (the same as in FIG. 3), and ● is the first pass actual thickness measured by the plate thickness gauge,
There is an error compared to the initially planned thickness of the first pass. This error is caused by an error in the accuracy of the load prediction formula, because the rolling position is adjusted so that the actual load matches the target load.

In the correction of the pass schedule, the pass schedule for the second and subsequent passes is calculated with the actual thickness of the first pass on the exit side as the second pass entrance thickness. In this pass schedule calculation method, as described above, the rolling reduction and the exit side plate thickness of each pass are determined based on the base plate thickness, finished thickness, plate width, steel type, etc. of the material to be rolled. The exit side plate thickness after the second pass obtained by the pass schedule correction is indicated by ⊚ in FIG.

Thereafter, rolling of each pass is sequentially performed in the same procedure. By learning the load prediction formula, the accuracy of the predicted load improves as the path progresses, and the consistency between the predicted load (target load) and the actual load increases, so it is not necessary to adjust the reduction position, and the plate thickness accuracy is significantly improved. It can be improved, and the shape of the plate is extremely good because it is rolled at a target load.

In the above-mentioned embodiment, when the subtle difference in the process in the previous process is likely to occur, or when the deformation resistance of the rolled material s is likely to fluctuate, the parameters of the load prediction formula are learned. The deformation resistance is learned as. The following equation can be used as the deformation resistance equation. k _f = (2 / √3) × k ₀ × (e + e ₀ ) ⁿ (where k ₀ : work hardening coefficient, e ₀ : deformation resistance constant,
n: strain hardening index, e: equivalent strain) However, the work hardening coefficient, the deformation resistance constant, and the non-hardening index are
It is recursively calculated based on past rolling results.

The modification resistance equation may be learned by providing the modification resistance equation k _f with a learning term k as follows. k _f = (2 / √3) × k × k ₀ × (e + e ₀ ) ⁿ (where k ₀ : work hardening coefficient, e ₀ : deformation resistance constant,
(n: strain hardening index, e: equivalent strain, k: learning term)

That is, after the rolling of the previous pass is completed,
If this is the final pass, the rolling of the coil is completed, but otherwise the rolling of the next pass is performed. At this time, the value of k is determined based on the result of the pass so that the actual load before the reduction position correction and the predicted load match. Needless to say, this learning may be performed during rolling.

Third Invention The present invention is particularly convenient in the first invention when the roll is easily worn and the friction coefficient is easily changed. That is, after the rolling of the previous pass is completed,
If this is the final pass, the rolling of the coil is completed, but otherwise the rolling of the next pass is performed. At this time, the value of the friction coefficient is changed based on the result of the previous pass so that the actual load before the reduction position correction and the predicted load match. Further, the learning of the load prediction formula may be performed during rolling.

Regarding the Fourth Invention FIG. 5 shows a flowchart for rolling one coil. First, calculate the pass schedule up to the finish plate thickness (outlet plate thickness of the final pass). This pass schedule is based on the method described above. The outgoing plate thickness of each pass obtained by the calculation of the pass schedule is the same as that shown in FIG. 2 above. Further, the load prediction calculation, mill set value calculation, shape control set value calculation, and mill setting are performed in the same manner as in the first invention described above. Next, it is determined whether the pass is the first half pass or the second half pass.

As a criterion for this judgment, when the number of passes is n, it is before the (n / 2) th pass, that is, 3
The first half pass up to the pass, and the second half pass after the (n / 2) th pass, that is, the fourth and subsequent passes. However,
This criterion is appropriately set according to the steel type of the material to be rolled, the required plate thickness accuracy, and the like.

If the result of the determination here is that in the first half pass, rolling is performed by adjusting the reduction position so that the target load and the actual load match. After rolling, the parameters of the load prediction formula are learned and the pass schedule is corrected. This is performed by the same method as the above-mentioned first invention. In the latter half pass, rolling is performed so that the target plate thickness obtained by the pass schedule and the actual plate thickness match.

FIG. 6 shows the plate thickness of each pass at the stage when the rolling of the first half pass (third pass) is completed. The actual value of the outlet plate thickness up to the third pass is indicated by ●. The exit side plate thickness due to the modification of the pass schedule after the fourth pass is indicated by ⊚.

In the rolling after the fourth pass, the rolling is performed so that the actual strip thickness matches the exit strip thickness set value according to the pass schedule. After rolling in each pass, the load prediction formula is learned based on the actually measured load values. At this time, since the actual output plate thickness value matches the plate thickness according to the pass schedule, it is not necessary to correct the pass schedule.

Thereafter, the load prediction calculation, the mill set value calculation, the shape control set value calculation, and the mill setting are performed in the same manner as in the above-described first invention, and rolling is performed at the target plate thickness until the final pass. By learning the parameters of the load prediction formula, the accuracy of the predicted load improves as the path progresses, and the match between the predicted load (target load) and the actual load increases, so highly accurate shape control settings can be set in the second half pass. Therefore, the plate shape can be improved. Further, in the latter half pass, rolling is performed so that the target plate thickness and the actual plate thickness match, so that the plate thickness accuracy is significantly improved.

About the Fifth Invention In the fourth invention, the parameters of the load predicting equation are learned when a slight difference in the process in the previous process is likely to occur and therefore the deformation resistance of the rolled material is likely to change. The deformation resistance is learned as.

The following equation can be used as the deformation resistance equation. k _f = (2 / √3) × k ₀ × (e + e ₀ ) ⁿ (where k ₀ : work hardening coefficient, e ₀ : deformation resistance constant,
n: strain hardening index, e: equivalent strain) However, the work hardening coefficient, the deformation resistance constant, and the strain hardening index are
It is recursively calculated based on past rolling results.

The deformation resistance type learning learns the deformation resistance equation k _f provided learning term k as follows. k _f = (2 / √3) × k × k ₀ × (e + e ₀ ) ⁿ (where k ₀ : work hardening coefficient, e ₀ : deformation resistance constant,
(n: strain hardening index, e: equivalent strain, k: learning term)

That is, after the rolling of the previous pass is completed,
If this is the last pass, the rolling of the coil is completed, but otherwise the rolling of the next pass is performed. On this occasion,
Based on the results of the previous pass, the value of k is determined so that the actual load before the reduction position correction and the predicted load match. The learning of the load prediction formula can be performed during rolling.

Sixth Invention In the fourth invention, the friction coefficient is learned when the roll is easily worn and therefore the friction coefficient is easily varied. That is, after the rolling of the previous pass is finished, if this is the final pass, the rolling of the coil is finished,
Otherwise, the next pass is rolled. At this time, based on the results of the previous pass, the value of the friction coefficient is changed so that the actual load before the rolling position correction and the predicted load match, but the learning of this load prediction formula may also be performed during rolling. Good.

[0050]

According to the first aspect of the invention, the accuracy of the predicted load is improved as the pass progresses by learning the parameters of the load prediction formula based on the actual rolling value, and the consistency between the predicted load (target load) and the actual load is improved. Therefore, it is not necessary to adjust the pressing position to match the both, and it is possible to stably manufacture a plate having good plate thickness accuracy. Further, the shape is extremely good because it is rolled under the target load.

According to the second aspect of the invention, even if the deformation resistance changes due to, for example, a subtle difference in the composition of the same steel type or a subtle difference in the treatment in the previous step, it is not affected.

According to the third aspect of the invention, even if the friction coefficient changes due to a change in the lubrication condition due to, for example, the outside air temperature or the wear of the roll, it is not affected.

According to the fourth aspect of the invention, the accuracy of the predicted load is improved as the path progresses by learning the parameters of the load prediction formula, and the consistency between the predicted load (target load) and the actual load is increased. It is possible to set the shape control with high accuracy in the pass and obtain a good shape. Further, in the latter half pass, rolling is performed so that the target plate thickness and the actual plate thickness match, so that the plate thickness accuracy is remarkably improved.

According to the fifth aspect of the invention, even if the deformation resistance changes due to, for example, a subtle difference in the composition of the same steel type or a subtle difference in the treatment in the previous step, there is no effect. .

According to the sixth aspect of the invention, even if the friction coefficient changes due to changes in the lubrication conditions due to, for example, the outside temperature or the wear of the rolls, it is not affected and the plate thickness accuracy and shape are good. The board can be obtained.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a reverse rolling mill suitable for use in implementing the present invention.

FIG. 2 is a diagram showing a flowchart of the first invention.

FIG. 3 is a diagram showing an example of the exit side plate thickness of each pass obtained by calculating a pass schedule.

FIG. 4 is a diagram showing the exit-side plate thickness of each pass after the second pass, which is obtained by modifying the pass schedule.

FIG. 5 is a diagram showing a flowchart of a fourth invention.

FIG. 6 is a diagram showing a comparison between the actual plate thickness of the first half pass and the exit plate thickness of each pass of the latter half pass obtained by correcting the pass schedule after the end of the first half pass.

[Explanation of symbols]

 1 Payoff reel 2 Right tension reel 3 Left tension reel 4 Mill 5 Load meter 6 Plate thickness gauge S Rolled material

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Masanori Kitahama, 1 Kawasaki-cho, Chuo-ku, Chiba-shi, Chiba Prefecture Technical Research Institute, Kawasaki Steel Co., Ltd. (72) Masafumi Hoshino, 1 Kawasaki-cho, Chuo-ku, Chiba-shi, Chiba Kawasaki Steel Co., Ltd., Chiba Steel Works (72) Inventor Masaharu Saizaku, Kawasaki-cho, Chuo-ku, Chiba City, Chiba Prefecture Kawasaki Steel Co., Ltd., Chiba Steel Works

Claims (6)

[Claims] 1. An optimum number of passes based on the base plate thickness, finish thickness, plate width, material, etc. of the material to be rolled, load limit for each pass, roll reduction limit, motor power limit, etc., optimum for each pass When determining the pass schedule such as the reduction ratio and the optimum delivery thickness, and performing the reverse rolling according to this pass schedule, the first pass of the reverse rolling is the shape control function according to the target rolling load obtained from the load prediction formula. Of initial rolling, rolling of the actual rolling load during rolling to match the target rolling load, and the learning of the parameters of the load prediction formula, and the second and subsequent passes were learned in the previous pass. Correction of the pass schedule up to the final pass based on the load prediction formula, initial setting of the shape control function according to the target rolling load obtained from the load prediction formula, and target of the actual rolling load And correction of the pressing position to match the extended load, reverse rolling method is excellent in shape and accuracy of plate thickness, characterized by rolling while repeatedly learning parameters of the load prediction formula on during or immediately after the rolling roll of each path. 2. The reverse rolling according to claim 1, wherein a learning term is provided in the deformation resistance equation in learning the parameters of the load prediction equation, and the value of this learning term is changed so that the target rolling load matches the actual rolling load. Method. 3. The reverse rolling method according to claim 1, wherein in learning the parameters of the load prediction formula, the value of the friction coefficient is changed so that the target rolling load matches the actual rolling load with respect to the friction coefficient. 4. An optimum number of passes, an optimum number of passes in each pass based on the base plate thickness, finish thickness, plate width, material, etc. of the material to be rolled, load limitation for each pass, reduction ratio limitation, motor power limitation, etc. When determining the pass schedule such as reduction ratio and optimum delivery side plate thickness and performing reverse rolling according to this pass schedule, the first pass of reverse rolling is the initial setting of the shape restriction according to the target rolling load obtained from the load prediction formula. Then, the rolling position is corrected by correcting the rolling position to match the actual rolling load with the target rolling load and learning the parameters of the load prediction formula.
For the first half and subsequent passes, the pass schedule is corrected using the learned load prediction formula until the final pass, and the initial setting of the shape control function according to the target rolling load calculated by the load prediction formula and the actual results. Rolling is performed by repeatedly correcting the rolling position to match the rolling load with the target rolling load and learning the parameters of the load prediction formula during or immediately after rolling, and the target rolling load calculated by the learned load prediction formula for the latter half pass. Rolling while repeatedly performing the initial setting of the shape control function according to the above, the correction of the rolling position to match the set thickness and the actual thickness, and the learning of the parameters of the load prediction formula during or immediately after each pass rolling. Reverse type rolling method with excellent shape and thickness accuracy. 5. The reverse rolling according to claim 4, wherein in the learning of the parameters of the load prediction formula, a learning term is provided in the deformation resistance formula, and the value of this learning term is changed so that the target rolling load matches the actual rolling load. Method. 6. The reverse rolling method according to claim 4, wherein, in learning the parameters of the load prediction formula, the value of the friction coefficient is changed so that the target rolling load matches the actual rolling load with respect to the friction coefficient.