JPH11104721A - Plate crown/shape controlling method in hot rolling - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain good plate crown and plate shape from the front end portion of the next rolled steel by learning the correction term within the model on the basis of the deviation between the measured value and the predicted value of the plate crown and the deviation between the measured value and predicted value of differential expansion, and calculating the predicted value of differential expansion from the predicted value of plate crown after learning. SOLUTION: When plate crown learning and shape learning are used jointly, the shape learning for achieving agreement between the measured plate shape and the predicted value of the model, and the plate crown learning which changes the relations among stands as a result of distributing the learning amount of each stand to achieve agreement between the measured plate crown and the predicted value of the model, interfere with each other. To prevent the problem, the calculated value after the plate crown learning is used when the differential expansion is calculated by the formula. In the formula, Δε is the differential expansion on exit side of stand (i), Cri is the calculated value of the plate crown on an exit side of stand i after learning, Hi is plate thickness on an exit side of stand (i), and ξ is a correction factor.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for controlling a sheet crown and a shape in a hot finishing mill, and more particularly to a method for measuring a sheet crown and a flatness meter disposed between at least one of stands and at the exit side of a final stand. By learning the prediction model based on the actual measurement values, it is possible to improve the setup accuracy of the sheet crown and shape, and to apply the sheet crown in hot rolling, which is preferable to be applied when realizing stable threading and improvement in yield. The present invention relates to a shape control method.

[0002]

2. Description of the Related Art In order to produce high-quality products having a uniform thickness distribution in the width direction of a sheet by using a hot finishing rolling mill (equipment) comprising a tandem rolling mill, a sheet crown is required to improve the yield. Needs to be controlled with high accuracy. or,
In order to realize a stable operation by a stable passing through a good plate shape, it is necessary to accurately control the plate shape (flatness).

[0003] In recent years, with the development of measurement technology, a sheet crown meter for measuring the thickness distribution in the width direction of a steel sheet and a flatness meter for measuring a sheet shape have been introduced on the exit side of the rolling mill. High-quality steel sheets are being manufactured using dynamic feedback control and the like utilized in the field.

[0004] However, since dynamic feedback control by a sheet crown meter or a flatness meter cannot be applied to the foremost part of a rolled material (coil), the quality and threadability of the foremost part are still set at the start of rolling. It depends on the setup accuracy of the plate crown and shape. Therefore, in order to obtain a higher quality steel plate and a better plate shape than the leading end of the coil, it is important to improve the accuracy of the plate crown / shape prediction model.

[0005] Regarding the plate crown / shape prediction model,
See, for example, J. Shoet et al. Ironand Steel Inst
. , 206 (1968) 11, 1088, 33rd Joint Lecture on Plastic Working, p. 143, Plasticity and Working: Vol. 3
6, No. 417 (1995-10) and the like, various studies have conventionally been made.

Generally, a sheet crown prediction model includes a rolling mill deformation model, a sheet deformation model for correcting three-dimensional deformation of a sheet,
It consists of a roll thermal expansion and wear model that predicts the roll profile. Among them, it is particularly difficult to predict a work roll profile that changes every moment due to thermal expansion, wear, and the like, and it is difficult to accurately predict the work roll profile.

As a method of compensating for such a prediction error, a setup method incorporating learning control is effective, and many studies have been made so far. For example,
JP-A-61-283407, JP-A-07-32331
5. The above-mentioned setup method is disclosed in JP-B-63-025845 and the like.

In other words, Japanese Patent Application Laid-Open No. 61-283407 discloses a method of dividing the thickness profile in the width direction at a set pitch and learning the thickness profile of each stand in the width direction using the genetic characteristics of the thickness profile. is there. Also, JP-A-07-
323315 proposes a method of separating and controlling plastic deformation of a plate and an error in roll profile change during rolling in a plate crown / shape prediction model. This method
Both the roll profile error and the plastic deformation error of the plate are mixed at the edge of the plate to be rolled, and only the roll profile error is generated at the non-edge portion to separate them. Also, Japanese Patent Publication No. 63-02584
5 measures both or any one of the sheet crown and the sheet shape, and learns the difference from the calculated sheet crown and / or the sheet shape as being caused by the roll profile estimation error, This is a method to be reflected in the setting calculation of the next rolled material.

[0009]

However, in the methods disclosed in the above-mentioned JP-A-61-283407 and JP-A-07-323315, in which the sheet crown is learned by measuring only the sheet thickness profile, the sheet crown meter is used. The calculated value of the plate crown on the exit side of the installed stand matches the measured value, but the plate shape of the same stand or its upstream stand changes as the plate crown is changed. However, in this method, since the plate shape is not particularly controlled, there is a problem that the plate shape may be adversely deteriorated.

In the method disclosed in Japanese Patent Publication No. 63-025845 for measuring and learning both the plate profile and the plate shape, both the plate crown prediction error and the plate shape prediction error are learned as roll profile errors. To change the plate crown by modifying the plate shape,
Further, since the plate shape is changed by correcting the plate crown, mutual interference between the plate crown and the plate shape is unavoidable, and it is difficult to improve both the plate crown and the plate shape at the same time. Was.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned conventional problems, and is intended to provide a hot-rolled sheet which can be rolled to the next rolled material from the foremost portion with a good sheet crown and sheet shape. It is an object to provide a crown / shape control method.

[0012]

According to the present invention, when rolling is performed by a tandem rolling mill in which a plurality of rolling stands are continuously arranged, the rolling stands are provided on at least one of an exit side of a final stand and between the stands. A learning control method of learning a correction term in each prediction model based on a deviation between a value measured by a sheet crown meter and a flatness meter and a prediction value by a corresponding prediction model, and reflecting the correction term in a setup of a next rolled material. And a sheet crown learning control for learning a correction term in the sheet crown measurement value measured by the sheet crown meter and a correction term in the model based on a deviation of the sheet crown prediction value calculated by the sheet crown prediction model. Based on the difference between the measured difference in elongation measured by the flatness meter and the predicted difference in elongation calculated by the predicted difference in elongation model, With a combination of shapes learning control for learning the Seiko, the elongation difference ratio predicted value, by using the plate crown predicted value after learning, the following equation _{Δε i = ξ × (Cr i} / H i -Cr i- ₁ / H _i-1 ) (1) (Δε _i : percentage difference in elongation on the exit side of i-stand, Cr _i , C
r _{i -1} : i, i -1 stand exit side sheet crown predicted value after learning, H _i , H _{i -1} : i, i -1 stand exit side plate thickness,
ξ: a correction coefficient determined by the rolled material dimensions, etc.) to solve the above problem.

That is, in the present invention, the model is learned based on the difference between the measured value of the sheet crown measured by the sheet crown meter and the value calculated by the sheet crown prediction model, and at the same time, the measurement is performed by the flatness meter. Based on the difference between the measured difference in elongation and the calculated difference in elongation calculated by the difference in elongation prediction model, the model is learned. The calculation by the elongation difference prediction model of (1) is performed using the plate crown calculation value calculated by the learned plate crown prediction model. This can be reflected in the value. As a result, it is possible to avoid mutual interference between the sheet crown and the sheet shape, thereby improving both the sheet crown and the sheet shape from the foremost part of the coil, and improving the yield and the operation rate by stabilizing the sheet passing. It can be realized.

[0014]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a schematic diagram showing a hot finish rolling equipment applied to one embodiment of the present invention.

The above-mentioned rolling equipment is a tandem rolling mill consisting of seven stands F1 to F7, and a strip crown meter (profile meter) 10 is provided at an outlet of a fourth stand F4, and an outlet of a final seventh stand F7. Are provided with a sheet crown meter 10 and a flatness meter 12, respectively.

Each stand constituting the rolling equipment is a pair cross type 4HI rolling mill, and includes a roll cloth device (not shown) and a work roll bending device 14 as shape control actuators.

In this embodiment, after learning a sheet crown prediction model and a sheet shape (difference in elongation) prediction model in accordance with the flowchart shown in FIG. 2, the sheet crown / shape for the next rolled material is learned. Perform setup calculations.

That is, the sheet crown is measured by the sheet crown meter 10 installed on the exit side of the stand, and the calculated value (predicted value) of the sheet crown on the exit side of the stand is calculated.
The measured value and the calculated value of the above-mentioned sheet crown are compared, the error is distributed to the crown learning term of each stand, and the learning value obtained for each stand is smoothed (steps 1 to 4). The processing of each step described above is performed for all the stands on which the sheet crown meter is installed (step 5).

Thereafter, the difference in elongation is measured by the flatness meter 12 installed on the exit side of the stand, and the calculated value (predicted value) of the difference in elongation on the exit side of the stand is calculated as the sheet crown value after learning. Calculate using the above, compare the measured value and the calculated value of the differential expansion rate, correct the learning term in the differential expansion rate prediction model of the stand using the error, and smooth the obtained learning value. A conversion process is performed (steps 6 to 9). When a plurality of flatness meters 12 are installed, the processing in each of the above steps is performed for all the stands where the flatness meters are installed on the output side (step 10). Then, a setup calculation of a sheet crown and a shape for the next rolled material is performed based on the above learning results.

Next, the learning of the plate crown prediction model in steps 1 to 5 will be described in detail.

Generally, a plate crown Cr on the exit side of the i-stand
_i is the roll crown CrR when the stand is loaded
_{Using i} and the sheet crown Cr _{i- 1} on the entrance side of the stand (exit side of the preceding stand), the prediction is made by the following equation (2).

Cr _i = α · CrR _i + β · Cr _i-1 (2) where α is called a transcription rate, and β is called a genetic coefficient, and is an influence coefficient of the roll crown on the outgoing sheet crown, and an incoming sheet crown, respectively. Is an influence coefficient on the exit side plate crown.

The roll crown CrR under load is defined as a deformation model of a rolling mill having factors such as bending and flatness of a work roll due to a rolling load, and a roll profile model having factors such as initial profile, thermal expansion, and wear of a work roll. It is calculated by

The sheet crown learning method includes a method of learning the sheet crown itself by adding the learning term L as shown in equation (3) and a method of giving the roll profile (roll crown) by giving it as shown in equation (4). There is a way to learn.

Cr _i = α · CrR _i + β · Cri ₋₁ + L _i (3) Cr _i = α · (CrR _i + L _i ) + β · Cri ₋₁ (4)

Although the effect of the present invention does not change when either learning method is used, the learning of the roll profile, that is, the learning method using equation (4) will be described here.

In the strip crown learning, the difference between the _strip crown measurement value Cr _mes measured by the strip crown meter installed on the exit side of the stand and the calculated value Cr _cal calculated from the crown prediction model for the same stand is given by: to compensate for this error, it from obtaining the learning term L _i of each stand upstream.

That is, in the rolling equipment shown in FIG. 1, since the strip crown on the exit side of the fourth stand F4 can be measured, the strip crown error on the exit side of the finishing mill (the exit side of the seventh stand) is the fifth. The roll profile error of the stands F5 to the seventh stand F7 may be compensated for, and the crown error on the exit side of the fourth stand F4 may be compensated for by the first stand F1 to the fourth stand F4. (Therefore, if there is a sheet crown meter only on the exit side of the last stand, the first stand F1
To the roll stand learning terms of all the stands of the seventh stand F7. )

In the roll barrel direction, the sheet crown meter measures the sheet thickness at a predetermined dimension position from the end of the sheet. Therefore, the learning point of the rolled material is adjusted according to the change in the sheet width configuration of the rolled material in the rolling cycle. Changes with respect to the barrel position, and the profile error of each point may be learned in accordance with the change.

First, a description will be given of sheet crown learning in the subsequent stands of F5 to F7 using the measurement by the sheet crown meter on the exit side of the final stand. Plate crown error ΔCr ₇ = Cr _mes on the exit side of the final stand F7
−Cr _cal is represented by the following equation (5). Where A _i
Is an influence coefficient of the roll crown of the i-stand on the sheet crown on the exit side of the final stand.

ΔCr ₇ = A ₅ · L ₅ + A ₆ · L ₆ + A ₇ · L ₇ (5)

The influence coefficient A _{i of the} equation (5) can be easily calculated by simultaneously applying the equation (4) for i = 5 to 7. The influence coefficient A is represented by a transcription rate and a genetic coefficient, for example, A ₅ = α ₅ β ₆ β ₇ .

[0034] Also, the learning term L _i of equation (5), the fifth stand F5, the inability to measure the plate crown out is the sixth stand F6, the error of the strip crown at respective stand delivery side is not known Therefore, to calculate the learning term for each of these stands, L ₅ , L ₆ ,
Relationship between L _7, it is necessary to put the assumptions as follows.

L ₅ = k ₅ · L ₇ , L ₆ = k ₆ · L ₇ (6) where k ₅ and k ₆ are constants, and L ₅ and L ₇ , and L ₆ and L ₇ , respectively. The ratio is shown. How to determine these constants are arbitrary, profile error of each stand may be k ₅ = k ₆ = 1 think the same level, and, as a profile error of each stand is proportional to the plate thickness, k ₅ = H
May be _{5 / H 7, k 6 =} H 6 / H 7.

By simultaneously solving the equations (5) and (6), the learning terms L ₅ , L ₆ and L ₇ of each stand can be calculated. In addition, when using the sheet crown meter 10 installed on the exit side of the fourth stand F4, the first stand F1
For each stand of the fourth stand F4, the fifth stand
By using the same method as the case of the stands F5 to the seventh stand F7, the learning amount of each stand can be calculated.

Although the learning value calculated as described above changes for each rolled material, it may change greatly due to measurement noise or the like. It is desirable that the learning value is subjected to a smoothing process so as not to cause a trouble such as a passing plate. As a method of smoothing, for example, the following (7)
As shown in the equation, the calculated instantaneous value L of the learning value is multiplied by the learning gain γ (= 0 to 1), and the learning value L up to the previous time is further calculated.
_Old is multiplied by 1-γ and added, and the learning value L of the next rolled material is
A method of calculating _new may be used. Of course, other smoothing methods may be used.

L _new = γ · L + (1−γ) · L _old (7)

Next, the learning of the plate shape (difference in elongation) prediction model in steps 6 to 10 in the flowchart of FIG. 2 will be described.

In general, if the ratio Δl / l of the difference Δl of elongation to l in the center of the sheet width and the vicinity of the sheet edge (shape evaluation point) in the fixed section l to l is defined as the elongation difference rate Δε, It is known that the elongation difference rate on the delivery side is represented by the following equation (8).

Δε _i = ξ · (Cr _i / H _i −C _i−1 / H _i−1 ) (8)

As an index indicating the degree of the plate shape defect, steepness is usually used. This is the wave height δ of the plate
Λ = δ / p using the pitch e and the pitch p, and there is a well-known relationship expressed by the following equation (9) between the elongation difference rate Δε and the steepness λ.

Λ = (2 / π) ｛| Δε |｝ ^1/2 × 100 [%] (9)

The shape learning in the present embodiment is to learn the elongation difference rate shown in the above equation (8). In learning, the sheet crowns Cr _i and Cr in the equation (8) are learned.
The calculated value after learning is used as _i-1 . That is, in learning the elongation difference rate of the rolled material in the stand where the flatness meter is installed on the delivery side, the elongation difference rate Δε _mes actually measured by the flatness meter and the elongation difference rate calculated by the prediction model are used. when the difference [Delta] [epsilon] _{ca l} and learning value M, the differential expansion rates after learning the next rolling material, so that calculated using the following equation (10).

Δε _i = ξ × (Cr _i / H _i −C _i−1 / H _i−1 ) + M _i (10) M _i = Δε _{i, mes} −Δε _{i, cal} (Δε _i : I stand out Side elongation difference, Cr _i , C
r _i-1 : the calculated value of the crown of the i, i-1 stand exit side after learning, _Hi , H _i-1 : i, the thickness of the i-1 stand exit side plate thickness,
ξ: Correction coefficient determined by rolling material dimensions etc.)

In the case of the rolling equipment shown in FIG. 1 in the present embodiment, the flatness gauge 12 is installed on the exit side of the seventh stand, and therefore, the above-mentioned shape learning is performed on the final stand 7th.
It will be applied to the stand (F7).

The learning value M in the shape learning is desirably smoothed and used in the same manner as in the above-described plate crown learning. The method of smoothing may be a method of multiplying the instantaneous value M of the learning by a learning gain of 0 to 1 and adding it to the previous value of the learning as in the case of the plate crown learning shown in the equation (7). May be used.

By using the elongation difference rate prediction model learned in this way, the measured value of the elongation difference rate by the flatness meter 12 and the predicted value match. As a result, the learning result is reflected in the setting calculation of the next rolled material, and a rolled material having a target shape can be obtained.

Using the learning terms of the stands calculated as described above, the setting calculation of the shape control actuator for obtaining the desired flatness and strip crown for the next rolled material is performed according to a conventional method.

According to the embodiment described above, the following effects can be obtained.

As described above, when the sheet crown learning is performed, the learning amount is distributed to each stand in order to compensate for an error between the learned value and the measured value of the sheet crown at the stand for setting the sheet crown meter. Naturally, the relation between stands changes before and after learning, and as a result, the plate shape changes. That is, when the sheet crown learning and the shape learning are used in combination, the shape learning for matching the actually measured sheet shape and the model predicted value and the actual sheet crown and the model predicted value Plate crown learning, which distributes the learning amount to the stands and consequently changes the relationship between the stands, interferes with each other.

In the present embodiment, in order to prevent this mutual interference, the calculated value after the sheet crown learning is used as the sheet crown value used when calculating the elongation difference rate in the shape learning by the equation (1). Because of this, the effect of the shape change caused by the plate crown learning,
This can be considered in shape learning, and the above-mentioned interference can be avoided.

In the conventional shape learning method, a prediction error measured by a shape meter (flatness meter) installed on the exit side of the stand is determined based on a roll profile prediction error of the stand, and the roll profile is learned. Was. In this case, the shape of the stand may be changed by changing the crown on the stand side by learning the shape.

However, in this embodiment, since the relationship between the elongation differential rate, that is, the change in the crown ratio at the entrance and exit of the stand is directly learned, it is said that the shape learning is performed so that the exit-side sheet crown is not changed. There are advantages. Also, the shape of each stand upstream of the stand is not changed.

Next, an example which is a specific example of the present embodiment will be described.

[0056]

EXAMPLE Hot rolling was carried out using a continuous finishing rolling equipment as shown in FIG. 1 in order to verify the effects of the present invention. In this finishing mill, all seven stands are formed by a pair cross mill (maximum angle 1.5 degrees), and each stand is provided with a work roll bender (maximum 1200 kN / c). The strip crown meter 10 is disposed on the exit side of the final stand and the fourth stand as described above, and the flatness meter is disposed on the exit side of the final stand.

FIG. 3 shows the configuration of the plate thickness, plate width, and steel type of the rolling cycle actually rolled in this embodiment. The first half is a wide material made of ultra low carbon steel sheet,
The latter half is a rolling cycle of a narrow material made of low carbon steel, which is a coffin cycle with a total of 84 rolling.

In the rolling cycle shown in FIG. 3, the sheet crown learning and the shape learning according to the present embodiment were simultaneously applied. In the sheet crown learning, the roll profiles of the first to fourth stands F1 to F4 are determined based on the deviation between the measured value and the calculated value (predicted value) at the exit side of the fourth stand F4, and the deviation between the measured value and the calculated value at the finishing output side. The above-described method of learning the roll profiles of the fifth to seventh stands F5 to F7 based on the above is used. Learning starts from the board edge (edge in the figure) 10
The measurement was performed at two points: a point of 0 mm and a point of 25 mm. At this time, the learning gain was set to 0.5, and the distribution of the learning amount of each stand was set to the equal distribution (the learning amount of each stand was the same).

FIGS. 4 and 5 also show the change of the calculated values of the seventh stand F7 by the prediction model when crown learning is performed and when crown learning is not performed, together with the actual crown value (solid line). From FIGS. 4 and 5, it can be seen that the calculated value by the learned model matches the actually measured value. Although not shown, similar results were obtained with the fourth stand F4.

FIGS. 6 to 8 show changes in the calculated shape (steepness) in the subsequent stands F5 to F7 at this time. The plate shape was evaluated at the point 100 mm from the plate edge by the above-mentioned formulas (8) and (9). These figures 6
From 〜8, it can be seen that the calculated value fluctuates more when the learning is not performed than when the learning is not performed. Therefore, it is understood that the relationship between the stands is changed by performing the crown learning, and the calculated shape is changed.

FIGS. 9 and 10 show the change in the actual value of the elongation difference rate and the calculated value based on the prediction model with the shape learning (learning gain 0.5).

FIG. 9 shows the calculated values predicted by the method of the present invention (plotted with ○) in comparison with the measured values of the elongation difference (solid line), and FIG. The calculated values (similarly plotted with a circle) are shown in comparison with the actual measured values.

Here, the method of the present invention corresponds to the case where the crown calculated value after crown learning is used for the plate crown values Cr _i and Cr _i-1 applied to the above equation (1). This corresponds to a case where the crown calculation value after learning is not used.

9 and 10 show the results when the learning value is not considered in the equation (10) used for the method of the present invention and the comparative example A, that is, when M _i = 0. , Respectively, as Comparative Examples B and C.

9 and FIG. 10, a comparative example A
In the area where the calculated shape changes due to crown learning shown in a circle in the figure, the predicted value varies due to interference with the shape change due to crown learning, and the rolled material whose actual value and calculated value do not match was generated. . On the other hand, in the case of the present invention, it can be seen that the actual growth difference rate and the calculated value are in good agreement. In each case, the shape learning value M
_{If i} is not considered, it can be seen that the predicted value (値) does not match the actual value. Thus, the reason for the consideration of the shape learning M _i in the present invention is understood.

FIG. 11 shows the prediction accuracy of the sheet crown on the exit side of the final stand obtained in the present embodiment. FIG.
9 shows the effect of improving the model accuracy by the shape learning of the present embodiment. The comparative example of FIG. 12 is a case where the learning of the sheet crown is not performed similarly to the comparative example of FIG.

As described above, by applying the present invention, the prediction accuracy of the model is improved, and it becomes possible to obtain the desired sheet crown and sheet shape for each rolled material in the rolling cycle.

Although the present invention has been specifically described above, the present invention is not limited to the above-described embodiment, and can be variously modified without departing from the gist thereof.

For example, in the above-described embodiment, the method for learning the roll profile shown in the above equation (4) has been described as an example of the sheet crown learning method. However, the present invention is not limited to this, and in the present invention, the equation (3) is used. The same effect can be obtained by using a method for learning a sheet crown shown in FIG.

In the above embodiment, the case where the sheet crown meter and the flatness meter are installed on the exit side of the final stand (the exit side of F7) and the sheet crown meter is installed on the exit side of the fourth stand F4 has been described. The location of the flatness meter is not limited, and the same effect can be obtained regardless of the location.

[0071]

As described above, according to the present invention,
High-quality hot-rolled steel sheets with a uniform thickness distribution in the width direction of the coil can be manufactured from the foremost part of the coil, and the shape of the foremost part can be improved to improve the passability of the foremost part. It is possible to greatly improve, and it is possible to achieve an improvement in the yield and an improvement in the operation rate by stable passing.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram showing a hot finishing rolling equipment applied to an embodiment according to the present invention.

FIG. 2 is a flowchart showing a learning procedure.

FIG. 3 is a diagram showing a plate thickness and a plate width configuration of a rolling cycle adopted in the embodiment.

FIG. 4 is a diagram showing changes in actual values and predicted calculated values of a sheet crown at a point 25 mm from a sheet edge.

FIG. 5 is a diagram showing changes in actual values and predicted calculated values of a sheet crown at a point 100 mm from a sheet edge.

FIG. 6 is a diagram showing a shape change in a fifth stand caused by strip crown learning.

FIG. 7 is a diagram showing a shape change in a sixth stand caused by strip crown learning.

FIG. 8 is a diagram showing a shape change in a seventh stand caused by plate crown learning.

FIG. 9 is a diagram showing a change between a predicted value and a measured value of a shape according to the present invention.

FIG. 10 is a diagram showing a change between a predicted value and a measured value of a shape according to a comparative example.

FIG. 11 is a diagram showing a sheet crown prediction accuracy at the exit side of the final stand according to the present invention.

FIG. 12 is a diagram illustrating an effect of improving the prediction accuracy of a model by shape learning according to the present invention;

[Explanation of symbols]

 10: Plate crown meter 12: Flatness meter 14: Work roll bending device

Continued on the front page (72) Inventor Masanori Kitahama 1 Kawasaki-cho, Chuo-ku, Chiba-shi, Chiba Pref. In the Technical Research Institute of Kawasaki Steel Corporation (72) Inventor Katsuhiro Takebayashi 1-Kawasaki-cho, Chuo-ku, Chiba-shi, Chiba Company Chiba Works

Claims (1)

[Claims] When rolling is carried out by a tandem rolling mill in which a plurality of rolling stands are continuously arranged, a plate crown meter and a flatness meter are provided at least at one end of the final stand and between the stands. A learning control method for learning a correction term in each prediction model based on a deviation between a measurement value and a prediction value by a corresponding prediction model, and reflecting the correction term in a setup of a next rolled material, wherein the sheet crown meter The crown crown measured value measured by the sheet crown, and a sheet crown learning control for learning a correction term in the model based on the deviation of the sheet crown prediction value calculated by the sheet crown prediction model, and is measured by the flatness meter. Shape learning that learns the correction terms in the model based on the difference between the measured differential expansion rate and the predicted differential expansion calculated by the differential expansion prediction model. With a combination of control, the differential expansion ratio prediction value by using the plate crown predicted value after learning, the following equation _{Δε i = ξ × (Cr i} / H i -Cr i-1 / H i-1) (Δε _i : differential elongation difference at i-stand, Cr _i , C
r _{i -1} : i, i -1 stand exit side sheet crown predicted value after learning, H _i , H _{i -1} : i, i -1 stand exit side plate thickness,
ξ: a method for controlling the crown and shape of a sheet in hot rolling, which is calculated by a correction coefficient determined by the dimensions of a rolled material.