Method for generating shape control actuator setting model of rolling equipment, method for setting shape control actuator of rolling equipment, method for controlling shape of steel sheet, shape control device for steel sheet, and method for manufacturing steel sheet

Abstract

To provide a generation method for shape control actuator setting model of rolling equipment that can generate a shape control actuator setting model which can control a shape of a steel plate in a longer direction with good accuracy and can improve stability of the shape of the steel plate.SOLUTION: A generation method for shape control actuator setting model of rolling equipment includes a setting model generation step at which a shape control actuator setting model is generated by a neural network method with use of a plurality of pieces of learning data in which one or more actual operation data selected from rolling operational data of rolling equipment and actual shape data acquired from a shape detector are so made as to be actual input data, and setting change information as to whether setting change of a shape control actuator is performed or not and a set value of the shape control actuator are so made as to be actual output data.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present invention relates to a method for generating a shape control actuator setting model of a rolling facility, a method for setting a shape control actuator of a rolling facility, a method for controlling the shape of a steel sheet, a shape control device for a steel sheet, and a method for manufacturing a steel sheet.

Background Art

[0002] In a rolling facility for reducing the thickness of a thin steel sheet, a steel strip obtained by winding the steel sheet in a coil shape is loaded into the rolling facility, and the steel sheet is continuously unwound and rolled while being conveyed. In the rolling process, it is required to control the steel sheet to a predetermined dimension and appropriately control the shape of the steel sheet. The shape of the steel sheet means flatness, and ear waves and center elongation are typical. The shape of the steel sheet is such that buckling occurs and a wave shape (shape defect) becomes apparent because the elongation strain (longitudinal strain) distributed in the width direction of the steel sheet during rolling cannot be maintained as an in-plane stress distribution. When a shape defect of the steel sheet occurs, not only is the quality of the product inferior, but the rolling speed (line speed) of the rolling facility has to be reduced, resulting in a decrease in production efficiency. Furthermore, the shape defect of the steel sheet can also cause operating troubles such as breakage and narrowing of the steel sheet. For this reason, in the rolling process, it is required to control the shape of the steel sheet to a desired shape over the longitudinal direction.

[0003] On the other hand, a rolling mill constituting a rolling facility is provided with a shape control actuator that controls the elastic deformation of the work roll to adjust the elongation strain distributed in the width direction of the steel sheet to a desired distribution. The shape control actuator varies depending on the mill type of the rolling mill, but a work roll bender that applies a bending force to the work roll to adjust the deflection of the work roll is typical. In addition, there are also those of a type that adjusts the deflection of the work roll using a roll shift that moves the work roll and intermediate roll in the axial direction or a split backup roll. Further, the rolling facility is provided with a shape detector for the steel sheet, and the shape control actuator is set so that the shape of the steel sheet detected by the shape detector matches the target shape (target shape).

[0004] A shape control device that sets a shape control actuator often has a function of performing an initial setting before starting a rolling pass and a feedback control function of modifying (resetting) the setting of the shape control actuator with reference to the output from a shape detector after the start of the rolling pass. By performing an appropriate initial setting before the start of the rolling pass, the shape can be made good from the tip of the steel sheet, and even if a disturbance occurs, the shape can be made good over the entire length of the steel sheet by the feedback control function. However, it may be difficult to quantitatively predict the influence of the setting value of the shape control actuator on the elastic deformation of the work roll of the rolling mill. For example, in order to quantitatively predict the deflection of the work roll, it is necessary to predict the behavior of the thermal expansion of various rolls constituting the rolling mill, and various errors often occur in these predictions. On the other hand, a technique for appropriately adjusting the setting value of the shape control actuator of the rolling equipment has been proposed.

[0005] Specifically, Patent Document 1 describes a method for appropriately performing the initial setting of a shape control actuator. This is a method of constructing a neural network using the influencing factors related to the flatness of the rolled material as input data. Patent Document 1 describes that for the actual shape value obtained when rolling with a certain set value, the optimal set value of each operation end is calculated by an optimal set value calculation device, and the paired data of the influencing factors and the set value of each operation end at this time are accumulated and used as learning data for the neural network. In this case, the neural network has a hierarchical structure, the influencing factors are input, and learning is performed by error backpropagation learning with the set value of each operation end as the output.

[0006] In addition, Patent Document 2 also describes a method for appropriately performing the initial setting of the shape control actuator. This is a method of accumulating in advance the rolling conditions of the steel sheet, the setting values of the shape control actuator, and the shape data as shape control performance data, and extracting typical data (prototypes) through learning using a neural network. Then, before executing the rolling pass, the similarity between the rolling conditions of the steel sheet and the target shape and the typical data is calculated, and based on the calculated similarity, the setting values of the shape control actuator according to the typical data are combined to perform the initial setting of the shape control actuator.

[0007] On the other hand, Patent Document 3 exemplifies the shape control of a steel sheet by rolling equipment, and describes a method of constructing a neural network that inputs shape deviation data, which is the difference between the actual shape detected by a shape detector and the target shape, and outputs the operation amount of the shape control actuator. Patent Document 3 also describes shape feedback control for resetting the shape control actuator according to the detected shape deviation. Furthermore, Patent Document 3 describes a method of optimizing the control output of the feedback control by providing a control result good / bad determination unit for determining the quality of the control output of the shape control actuator.

Prior Art Documents

Patent Documents

[0008]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0009] However, the methods described in Patent Document 1 and Patent Document 2 relate to the initial setting method of the shape control actuator and do not relate to the shape feedback control for controlling the shape in the longitudinal direction of the steel sheet. Therefore, there are cases where the shape of the steel sheet cannot necessarily be controlled to the target shape over the longitudinal direction. On the other hand, the method described in Patent Document 3 can be applied to dynamic shape control (shape feedback control) for controlling the shape of the steel sheet in the rolling equipment, and the shape in the longitudinal direction of the steel sheet can be controlled to the target shape. However, in the method of constructing a neural network that inputs the shape deviation data, which is the difference between the actual shape and the target shape by the shape detector, and outputs the operation amount of the shape control actuator, the shape deviation data changes for each control cycle of the shape control. For this reason, the operation amount of the shape control actuator changes at any time. That is, as long as the actual shape and the target shape do not fall within a predetermined deviation, the set value of the shape control actuator always fluctuates during the rolling of the steel sheet, and such fluctuations become disturbances when rolling the steel sheet, resulting in instability of the shape of the steel sheet in the longitudinal direction. Furthermore, due to the fluctuation of the set value of the shape control actuator, fluctuations in the lubrication and friction states occur at the contact portion between the steel sheet and the rolling roll, resulting in appearance defects such as color unevenness on the surface of the rolled steel sheet, and a decrease in the yield of the steel sheet. Also, Patent Document 3 describes a control result pass / fail determination unit for determining the quality of the control output of the shape control actuator. However, this determines whether the actual data (in this case, the shape deviation data) obtained as a result of changing the set value of the shape control actuator has become better than before the change. Therefore, the problem that the set value of the shape control actuator fluctuates during the rolling of the steel sheet cannot be solved.

[0010] The present invention has been made to solve the above problems, and an object thereof is to provide a method for generating a shape control actuator setting model for a rolling facility that can accurately control the shape of a steel plate in the longitudinal direction and improve the stability of the shape of the steel plate. Another object of the present invention is to provide a method for setting a shape control actuator of a rolling facility, a method for controlling the shape of a steel plate, and a shape control device for a steel plate, which can accurately control the shape of the steel plate in the longitudinal direction and improve the stability of the shape of the steel plate. Still another object of the present invention is to provide a method for manufacturing a steel plate that can efficiently manufacture a steel plate having a desired shape in the longitudinal direction.

Means for Solving the Problems

[0011] A method for generating a shape control actuator setting model for a rolling facility according to the present invention is a method for generating a shape control actuator setting model for a rolling facility including a shape control actuator for controlling the shape of a steel plate and a shape detector for detecting the shape of the steel plate, the method comprising a setting model generation step of generating a shape control actuator setting model by a neural network method using a plurality of learning data, wherein the input performance data includes one or more operation performance data selected from the rolling operation parameters of the rolling facility and shape performance data obtained from the shape detector, and the output performance data includes setting change information indicating whether to change the setting of the shape control actuator and the setting value of the shape control actuator.

[0012] The rolling operation parameters may include target shape data of the steel plate.

[0013] The neural network may include, as an output layer, a first output layer that outputs the setting change information and a second output layer that outputs the setting value of the shape control actuator, and the first output layer and the second output layer may have a network structure branched from a common intermediate layer.

[0014] The method for setting a shape control actuator of a rolling facility according to the present invention is a method for setting a shape control actuator of a rolling facility including a shape control actuator for controlling the shape of a steel sheet and a shape detector for detecting the shape of the steel sheet, wherein one or more operating data selected from the rolling operation parameters of the rolling facility and shape data obtained from the shape detector are input data, setting change information indicating whether to change the setting of the shape control actuator, and a set value of the shape control actuator are output data, and when the setting change information is for changing the setting of the shape control actuator, the method includes a resetting step of resetting the shape control actuator to the set value using a shape control actuator setting model generated by a neural network method.

[0015] The neural network includes, as an output layer, a first output layer that outputs the setting change information and a second output layer that outputs a set value of the shape control actuator, and the first output layer and the second output layer have a network structure branched from a common intermediate layer. The resetting step may reset the shape control actuator to the set value when the output of the first output layer is for changing the setting of the shape control actuator.

[0016] The method for controlling the shape of a steel sheet according to the present invention includes a step of resetting the shape control actuator during rolling of the steel sheet using the method for setting a shape control actuator of a rolling facility according to the present invention.

[0017] The steel sheet shape control device according to the present invention includes means for resetting the shape control actuator during rolling of the steel sheet using the method for setting a shape control actuator of a rolling facility according to the present invention.

[0018] The method for manufacturing a steel sheet according to the present invention includes a step of manufacturing a steel sheet using the method for controlling the shape of a steel sheet according to the present invention.

Effects of the Invention

[0019] According to the method for generating a shape control actuator setting model of a rolling facility according to the present invention, a shape control actuator setting model can be generated that accurately controls the shape of a steel plate over the longitudinal direction and improves the stability of the shape. Further, according to the method for setting a shape control actuator of a rolling facility, the method for controlling the shape of a steel plate, and the shape control device for a steel plate according to the present invention, the shape of a steel plate can be accurately controlled over the longitudinal direction and the stability of the shape can be improved. Further, according to the method for manufacturing a steel plate according to the present invention, a steel plate having a desired shape over the longitudinal direction can be manufactured with good yield.

Brief Description of the Drawings

[0020]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Embodiments for Carrying Out the Invention

[0021] Hereinafter, with reference to the drawings, an embodiment of the present invention will be described.

[0022] 〔Rolling Equipment〕 First, with reference to FIGS. 1 and 2, the configuration of a rolling equipment according to an embodiment of the present invention will be described. In this embodiment, equipment including at least a rolling mill and a shape detector will be referred to as rolling equipment.

[0023] FIG. 1 is a side view showing the configuration of a rolling equipment according to an embodiment of the present invention. As shown in FIG. 1, a rolling equipment 1 according to an embodiment of the present invention is composed of a 12 - stand cluster - type rolling mill. The 12 stands of rolls include a pair of upper and lower work rolls 2, two pairs of upper and lower intermediate rolls 3, a pair of upper and lower small backup rolls 4, two pairs of upper backup rolls 5 on the upper side, and two pairs of lower backup rolls 6 on the lower side. The steel sheet S as the rolled material is supplied from any one of the pay - off reel 7, the right tension reel 8, and the left tension reel 9, rolled, and wound around the tension reel on the opposite side. The shape detector 10 detects the elongation difference in the width direction of the steel sheet S.

[0024] FIG. 2 is a front view showing the schematic configuration of a shape control actuator (operation end) and a rolling mechanism. The steel sheet S is rolled and stretched from the upper and lower directions by the work rolls 2. The upper backup roll 5 is divided into seven parts in the axial direction, and the center 51a, quarter - in 51b, quarter - out 51c, and edge 51d can adjust the roll crown by performing extrusion in order from the axial center position. The lower backup roll 6 is divided into six parts in the axial direction, and the center 61a, quarter 61b, and edge 61c can adjust the roll crown by performing extrusion in order from the axial center position.

[0025] The rolling equipment is provided with, as shape control actuators, a roll crown adjustment function for the upper backup roll 5 and the lower backup roll 6, an intermediate roll bender 11, and a work roll bender 12 that applies bending force to the upper and lower work rolls 2. As a shape control actuator, a leveling mechanism that tilts the axial center of the upper backup roll 5 or the lower backup roll 6 in the width direction may be provided. Further, in addition to these shape control actuators, an intermediate roll shift in which the upper and lower intermediate rolls 3 shift in opposite directions with respect to the axial direction of the roll, or a work roll shift in which the upper and lower work rolls 2 shift in opposite directions may be provided.

[0026] Returning to FIG. 1, the shape detector 10 is a measuring instrument that measures the shape of the steel sheet S rolled by the rolling mill. As the shape detector 10, means usually used for measuring the shape of the steel sheet S may be selected. The shape of the steel sheet S is caused by the distribution of the line segment lengths in the longitudinal direction of the steel sheet S in the width direction, and the difference in the length in the width direction is usually evaluated in units of elongation difference rate (%) or elongation difference (I-unit (1×10 -5 )). Specifically, when the length of the steel sheet S is distributed in the width direction, the contact pressure (vertical load on the contacting roll) is distributed in the width direction with respect to the roll that contacts the steel sheet S at a constant winding angle. The shape detector 10 calculates the elongation difference by measuring the distribution of the contact pressure. At this time, the width direction distribution of the contact pressure is detected by load detectors arranged by being axially divided inside the contact roll (split load cell method).

[0027] However, as the shape detector 10, a method of detecting the out-of-plane displacement of the steel plate S by a non-contact distance meter (such as a laser distance meter) and converting it into an elongation difference, or a method of calculating the elongation difference from the moiré fringes irradiated on the steel plate S may also be used. Using any of the shape measuring instruments, the elongation difference rate or elongation difference (I-unit) shall be measured from the line segment length in the width direction of the steel plate S. The number of channels (number of divisions in the width direction) for detecting the elongation difference in the width direction is preferably larger than the number of shape control actuators used for shape control. Specifically, the shape detector 10 shall be divided into 24 to 64 parts in the width direction so that the elongation difference rate or elongation difference at each position can be measured.

[0028] 〔Shape control system〕 Next, with reference to FIG. 1, the shape control system of the rolling equipment 1 will be described.

[0029] As shown in FIG. 1, the shape control system of the rolling equipment 1 includes a control computer 100 and a rolling control controller 110. The control computer 100 acquires information on the base material dimensions (such as plate thickness, plate width, length, etc.) of the steel plate S and information on the deformation resistance and product plate thickness of the steel plate S from the host computer, and calculates the predicted value of the rolling load at each rolling pass when performing rolling with the number of rolling passes set according to the manufacturing specifications of the steel plate S. Then, the control computer 100 sends the initial setting value of the shape control actuator for each rolling pass to the rolling control controller 110 based on the predicted value of the rolling load. The rolling control controller 110 operates each device of the shape control actuator based on the acquired initial setting value. The rolling control controller 110 also acquires information on the target shape of the steel plate S in each rolling pass from the control computer 100.

[0030] After the initial setting values of the shape control actuator are set and the rolling pass starts, the rolling control controller 110 resets the setting values of the shape control actuator as needed (shape feedback control). The shape feedback control is performed using the actual shape of the steel sheet S obtained from the shape detector 10 and the target shape of the steel sheet S obtained from the control computer 100. By executing the shape feedback control, even if there are disturbances such as fluctuations in the plate thickness of the base material in the longitudinal direction of the steel sheet S, the shape of the steel sheet S can be brought closer to the target shape, and a steel sheet S having a uniform shape in the longitudinal direction can be manufactured.

[0031] 〔Shape control actuator setting model〕 Next, with reference to FIGS. 3 to 6, a shape control actuator setting model according to an embodiment of the present invention will be described.

[0032] A shape control actuator setting model according to an embodiment of the present invention is provided in the rolling control controller 110 and is used to send appropriate setting values to the shape control actuator with reference to the actual shape of the steel sheet S obtained from the shape detector 10. As shown in FIG. 3, the shape control actuator setting model M is generated by a shape control actuator setting model generation unit 140. In this embodiment, the data acquisition unit 120 selects, in the rolling pass of the steel sheet S, one or more operation performance data selected from the shape performance data of the steel sheet S acquired during rolling, the rolling operation parameters of the rolling facility 1 in the rolling pass, the performance data of the setting values of the shape control actuator (hereinafter sometimes referred to as actuator setting values), and the performance data of the setting change information of the shape control actuator.

[0033] The setting change information of the shape control actuator (hereinafter sometimes referred to as "setting change information") refers to information indicating whether to change the setting value of the shape control actuator in the next control cycle (next step) from the current setting value of the shape control actuator during the operation of the rolling equipment. The setting change information can be represented by binary information, for example, "1" when the actuator setting value is changed and "0" when the actuator setting value is not changed. Also, the actual performance data of the setting value of the shape control actuator refers to the setting values of at least one or more shape control actuators selected from the above shape control actuators. The various data acquired by the data acquisition unit 120 constitute a set of data sets and are stored in the database unit 141 of the shape control actuator setting model generation unit 140, and the shape control actuator setting model M is generated by the machine learning unit 143.

[0034] The shape actual performance data acquired by the data acquisition unit 120 is the shape data measured by the shape detector 10 (see FIG. 1), and is, for example, as illustrated in FIG. 4. The shape data is information specified by the width direction position of the steel plate S and the elongation difference rate or elongation difference (I-unit) at that position. The elongation difference rate is a value representing the line segment length of the steel plate S at other width direction positions in a strain form with reference to the position where the line segment length in the width direction of the steel plate S is the shortest. Also, since the shape data becomes a very small value even when the elongation difference (I-unit) is expressed as a percentage of the elongation difference rate, the elongation difference rate is customarily converted to a value 100,000 times and expressed in a unit called I-unit. In this case, when the elongation difference rate or elongation difference at the width direction end of the steel plate S is larger than other parts, it is often called ear wave, and when the elongation difference rate or elongation difference at the width direction center part is larger than other parts, it is often called middle elongation. Also, when there is a large elongation difference rate or elongation difference between the width direction end and the width direction center part, it is called quarter elongation, and a form in which the above shapes are combined is called composite elongation.

[0035] The operation performance data acquired by the data acquisition unit 120 together with the shape performance data of the steel plate S is one or more pieces of operation performance data selected from the rolling operation parameters. The rolling operation parameters are parameters representing the rolling state when executing a rolling pass, and are parameters representing factors that can affect the shape of the steel plate S, excluding the shape performance data of the steel plate S and the setting performance data of the shape control actuator. As the rolling operation parameters, for example, as the dimensional information of the steel plate S, information on the plate thickness and plate width of the base material held by the host computer that gives instructions to the control computer 100 may be used. Also, the inlet side plate thickness, outlet side plate thickness, and plate width in the rolling pass for acquiring the shape performance data may be used. Another example of the rolling operation parameters is the attribute information of the base material. As the attribute information of the base material, for example, indices representing the characteristics of the base material such as the deformation resistance of the steel plate S, steel type, component system, surface roughness of the base material, and distribution state of surface oxides may be used. Also, as the attribute information of the base material, the manufacturing conditions in the process (previous process) before the rolling process is executed by the rolling equipment 1 may be used. For example, when the previous process is an annealing process, information such as annealing temperature, soaking time, and cooling rate, and when the previous process is a hot rolling process, the coiling temperature on the hot rolling line may also be included in the attribute information of the base material. This is because these are parameters that indirectly affect the rolling load when performing shape control with the rolling equipment 1.

[0036] Another example of the rolling operation parameters is the rolling condition information in the rolling pass for acquiring the shape performance data. Examples of the rolling condition information include rolling load, rolling torque, rolling speed, inlet side tension, outlet side tension, and work roll diameter. Also, as the rolling condition information, parameters related to the lubrication conditions during rolling such as coolant flow rate, coolant temperature, and coolant concentration may be used. For each parameter of the dimensional information of the steel plate S, the attribute information of the base material, and the rolling condition information, either the set value preset by the host computer or the control computer 100 or the actual performance value measured by the rolling equipment 1 may be used. When the rolling equipment 1 is equipped with measuring instruments such as a thickness gauge for measuring the plate thickness of the steel plate S, a load cell for measuring the rolling load, and a tension gauge for measuring the tension applied to the steel plate S, it is preferable to use their actual measured values.

[0037] The rolling operation parameters preferably include data on the target shape of the steel sheet S set in the rolling control controller 110. The target shape of the steel sheet S may be the target value of the elongation difference rate at each position in the width direction of the steel sheet S, as exemplified as shape performance data in FIG. 4. However, as the target shape of the steel sheet used for the rolling operation parameters of the present embodiment, a polynomial approximation may be performed on the target of the elongation difference as a function of the width direction position, and the values of the coefficients of the polynomial may be used. For example, as the target shape of the steel sheet S, the width direction distribution of the target elongation difference (I-unit) is approximated by a fourth-order function and expressed as the following mathematical formula (1). However, z in the mathematical formula (1) is the width direction position obtained by normalizing the plate width in the range of -1 to +1. The coefficients λ1 to λ4 in this case may be included in the rolling operation parameters as data on the target shape of the steel sheet.

[0038]

Number

[0039] For the operation performance data of the present embodiment, at least one parameter is selected from the above rolling operation parameters, and the operation performance data regarding this parameter is used. On the other hand, as the rolling operation parameters, identification data of an operator (operator) who manually operates the shape control actuator of the rolling equipment 1 may be included. When the operator manually operates to control the shape of the steel sheet S, numbers, symbols, names, etc. may be included in the rolling operation parameters as identification data that can identify the operator who performed the operation. The shape control of the steel sheet S exhibits unique control performance according to the characteristics of the shape control actuator of the rolling equipment 1. Therefore, there is a difference in the superiority and inferiority of the shape control of the steel sheet S between an operator with high proficiency and an operator with low proficiency. Therefore, a shape control actuator setting model M can be generated by separately evaluating the shape control performance of the steel sheet S based on the difference between the target shape and the actual shape, and selectively outputting the operation data of the operator with good evaluation results.

[0040] The set performance data of the shape control actuator of the rolling equipment 1 refers to the setting information of the shape control actuator specified according to the rolling mill to be used. When the rolling mill is equipped with a plurality of shape control actuators, any one of them can be selected and used as the set performance data of the shape control actuator of the present embodiment. As the setting information of the shape control actuator, a leveling setting value for operating the tilting state of the axial center of the upper backup roll 5 or the lower backup roll 6 can be used. Also, setting values for the bender force of the intermediate roll bender 11 or the work roll bender 12 can be used. Further, a roll shift setting value representing the axial shift amount of the intermediate roll 3 or the work roll 2 can be used. Also, the 12-stage cluster type rolling mill shown in FIG. 1 is provided with a shape control actuator for adjusting the position of the split backup roll. Therefore, the setting position information of each backup roll, such as the center position, quarter-in position, quarter-out position, and edge position of the upper backup roll 5, and the center position, quarter position, and edge position of the lower backup roll 6, may be included in the setting information of the shape control actuator. As the set performance data of the shape control actuator, it may be the set value (command value) of each operation end exemplified above, or the actual measured value if it is actually measured. Also, when the bender force or the like can be calculated from the measured value such as the hydraulic pressure, it is preferable to use the actual measured value instead of the set command value.

[0041] It is preferable to select a shape control actuator for use in dynamic control during rolling of a steel sheet from among all shape control actuators. Thereby, a steel sheet having a desired shape in the longitudinal direction can be produced with good yield. The performance data of the setting change information of the shape control actuator is information representing the performance of whether or not the actuator setting value has been changed in the subsequent control cycle (next step) from the actuator setting value at the current time when the rolling equipment is in operation. It can be represented by binary information where "1" indicates that the setting value of the shape control actuator has been changed and "0" indicates that the setting value has not been changed. The performance data of the setting change information of the shape control actuator may be obtained by the operator of the rolling equipment acquiring the performance data of whether or not the shape control actuator has been changed by an operation switch or the like connected to the rolling control controller 110.

[0042] On the other hand, after obtaining the actual data of the actuator set value as time-series data at regular intervals (for example, for each control cycle), it is also possible to specify the actual data of the setting change information. FIG. 5 is a diagram for explaining a method of specifying the setting change information from the time-series data of the actuator set value. FIG. 5 shows the actual data of the actuator set value at times t1, t1+Δt, t1+2Δt, and t1+3Δt. However, the time t1 is an arbitrary time during the rolling of the steel plate, and Δt is the control cycle (for example, 0.2 to 10 seconds) for performing shape control. In this example, the actual data of the actuator set value at time t1 and time t1+Δt are compared. As a result of the comparison, if any of the actuator set values has changed, it is determined that there is a setting value change, and if none of the actuator set values has changed, it is determined that there is no setting value change. Then, the determination result is assigned as the actual data of the setting change information at time t1. For example, in the example shown in FIG. 5, at time t1 and time t1+Δt, since none of the actuator set values has changed, the actual data of the setting change information at time t1 is set to "0". On the other hand, at time t1+Δt and time t1+2Δt, since any of the actuator set values has changed, the actual data of the setting change information at time t1+Δt is set to "1". In this way, based on the continuously acquired actual data of the actuator set value, the actual data of the setting change information can be specified. Therefore, in the data acquisition unit 120 shown in FIG. 3, the actual data of the setting change information may be acquired as continuous data for each fixed period, and a data processing unit for specifying the actual data of the setting change information as described above may be provided.

[0043] The shape performance data, operation performance data of rolling operation parameters, performance data of actuator setting values, and performance data of setting change information acquired by the data acquisition unit 120 are associated based on the manufacturing management number of the steel sheet S and the rolling pass information, and are accumulated in the database unit 141 as a set of data. The data sets accumulated in the database unit 141 can be generated in plural along the longitudinal direction of the steel sheet S even in the same rolling pass of the same steel sheet S. For example, when acquiring performance data at a predetermined cycle (e.g., 100 ms) during rolling, the number of data sets acquired per rolling pass becomes enormous. Therefore, when accumulating data sets in the database unit 141, the number of data sets accumulated in the database unit 141 may be reduced by performing a screening process for extracting only the data sets necessary for generating the shape control actuator setting model M.

[0044] For example, it takes about 0.2 to 10 seconds from the resetting of the shape control actuator of the rolling equipment 1 until the steel sheet S reaches the shape detector 10. Therefore, the sampling period for acquiring the data set may be set within the range of 0.2 to 10 seconds, and only the data set extracted in this way may be accumulated in the database unit 141. In this case, the sampling period may be changed according to the rolling speed. Also, in a region where the fluctuations in the plate thickness of the base material and the deformation resistance are large, such as near the tip or the tail end of the steel sheet S, during the period from the resetting of the shape control actuator of the rolling equipment 1 until the steel sheet S reaches the shape detector 10, the rolling state may change significantly over time. And in this case, the correspondence between the set value of the shape control actuator and the shape actual result may become unclear. In such a case, a process of excluding the data set acquired at the tip or the tail end of the steel sheet S from the database unit 141 may be performed. Furthermore, when the output regarding the shape actual result of the shape detector 10 fluctuates greatly in a short time, there may be a detection error in the shape detector 10. Therefore, when the shape actual result fluctuates in a short cycle, screening processing may be performed to exclude it from the database unit 141. On the other hand, screening may be performed to leave the data set in which the actual result data of the setting change information is "1" (there is a setting change of the shape control actuator) and preferentially delete a part of the data set in which the actual result data of the setting change information is "0" (there is no setting change of the shape control actuator).

[0045] The shape control actuator setting model generation unit 140 generates a shape control actuator setting model M in the machine learning unit 143 using the learning data accumulated in the database unit 141. The shape control actuator setting model generation unit 140 may be inside the control computer 100 that oversees the rolling process of the steel sheet S by the rolling equipment 1, or may be constituted by separate hardware from the control computer 100. Also, it may be arranged in the shape control actuator setting unit 150 described later.

[0046] The number of data sets stored in the database unit 141 is 10,000 or more, preferably 100,000 or more, and more preferably 1,000,000 or more. Also, the data sets may be classified based on the attribute information of the steel plate S and the rolling pass number, and the shape control actuator setting model M may be generated according to those classifications. The data sets stored in the database unit 141 may be appropriately updated within the database unit 141 with a certain upper limit on the number of data sets as the upper limit.

[0047] After the learning data is stored in the database unit 141, the preprocessing unit 142 performs preprocessing on the data sets stored in the database unit 141 as necessary before executing machine learning to generate the shape control actuator setting model M. Examples of the preprocessing executed by the preprocessing unit 142 include normalization of the shape data and the actual performance data of the setting values of the shape control actuator, data shuffling, and the like.

[0048] The preprocessing unit 142 may perform normalization processing on the shape actual performance data, which is the input actual performance data stored in the database unit 141, the operation actual performance data selected from the rolling operation parameters, and the actual performance data of the setting values of the shape control actuator, which is the output actual performance data. The normalization processing is a process for efficiently updating the weight coefficients during network learning and is effective in avoiding the vanishing gradient problem during the learning of the neural network. When normalizing the input variables and output variables, it is preferable to specify the minimum value and the maximum value of each variable and normalize them in the range of 0 to 1. Also, the average value of the variables may be calculated and normalized in the range of -1 to +1.

[0049] In the database unit 141, data sets composed of shape performance data which is input performance data, operation performance data selected from rolling operation parameters, setting performance data of the shape control actuator which is output performance data, and performance data of setting change information are accumulated in the order in which the data is acquired. Alternatively, data shuffling may be performed to randomly rearrange the order of the data sets accumulated in the database unit 141. By changing the order of the data sets acquired in time series, it is effective in improving the generalization performance of the neural network and suppressing overfitting.

[0050] The data sets accumulated in the database unit 141 may be divided into learning data and test data. The learning data is set to 70 - 80% of all the data sets for use in learning the neural network. On the other hand, the remaining data sets are used to confirm whether the neural network obtained after the learning by the machine learning unit 143 can also exhibit good performance with respect to the test data.

[0051] The machine learning unit 143 generates a shape control actuator setting model M by means of a neural network method using a plurality of learning data, using the dataset stored in the database unit 141 (setting model generation step). The learning data uses, as input performance data, one or more operation performance data and shape performance data selected from the rolling operation parameters, and uses, as output performance data, the setting change information of the shape control actuator and the actuator setting value. As the machine learning model for generating the shape control actuator setting model M, any machine learning model can be used as long as a practically sufficient estimation accuracy of the setting value of the shape control actuator is obtained. For example, commonly used neural networks, decision tree learning, random forest, support vector regression, etc. Also, an ensemble model combining a plurality of models may be used. Further, in the present embodiment, a neural network is used as a machine learning method, and includes a first output layer that outputs setting change information as an output layer, and a second output layer that outputs an actuator setting value, and it is preferable that the first output layer and the second output layer have a network structure branched from a common intermediate layer.

[0052] Referring to FIG. 6, the neural network of the present embodiment will be described. In the neural network of the present embodiment, the shape data acquired from the shape detector 10 and the operation data selected from the rolling operation parameters of the rolling facility 1 are input from the input layer. The neural network of the present embodiment includes one or more intermediate layers and two output layers branched from a common intermediate layer for these input data. In the first output layer, setting change information of the shape control actuator is output, and in the second output layer, the set value of the shape control actuator is output. The first output layer is a classifier that outputs whether to change the setting of the shape control actuator (\"1\") or not (\"0\") as the setting change information, and a sigmoid function may be used as the activation function. On the other hand, the second output layer is a regressor that outputs the set value of the shape control actuator used as the output actual result data when generating the shape control actuator setting model M, and an identity function may be used as the activation function. However, when the actuator set value was normalized when generating the shape control actuator setting model M, it is preferable to convert the output from the second output layer into a physical quantity that is the set value of the shape control actuator. The intermediate layer of the neural network is preferably in the range of 1 to 10 layers, and the number of nodes in the intermediate layer is preferably 32 to 2048. This is because if the number of intermediate layers and nodes is too small, the setting accuracy of the shape control actuator setting model M will decrease, and if there are too many, overfitting will become prominent and the prediction accuracy may instead decrease. From the viewpoint of prediction accuracy, it is preferable to use the ReLU function as the activation function in the intermediate layer.

[0053] On the other hand, in the neural network of the present embodiment, a batch normalization layer for performing batch normalization of data may be provided immediately before the operation of the activation function in the intermediate layer. Batch normalization refers to an operation of normalizing data using the mean and variance calculated for each batch. This improves the efficiency of learning the weight coefficients. Also, a dropout layer may be provided on the downstream side of the activation function in the intermediate layer. The dropout layer is an operation that invalidates the output with a certain probability (for example, 0.1 to 0.4). This is because it is possible to suppress overfitting.

[0054] As described above, the output of the shape control actuator setting model M is divided into two parts: setting change information and actuator setting value, because a shape control system for a rolling facility can be configured as follows. That is, when the setting change information, which is the first output of the shape control actuator setting model, is "0" (no change in the setting of the shape control actuator), the setting value of the shape control actuator is not changed, and control is performed to maintain the current setting value. On the other hand, when the setting change information, which is the first output of the shape control actuator setting model, is "1" (there is a change in the setting of the shape control actuator), the setting of the shape control actuator of the rolling facility is changed according to the setting value of the shape control actuator (actuator setting value), which is the second output. This solves the problem that, as in the prior art, the setting value of the shape control actuator is frequently changed during rolling, the variation in the rolling state becomes a disturbance, and the shape of the steel plate in the longitudinal direction is not stable. In addition, it is possible to suppress fluctuations in the lubrication and friction conditions at the contact portion between the steel plate and the rolling roll, and to suppress a decrease in yield due to appearance defects and the like. In particular, when an operator with rich operating experience of a rolling facility manually operates the shape control actuator, for small fluctuations in the rolling operation parameters and shape performance data during operation, the operation of the shape control actuator is not actively performed. On the other hand, when the rolling operation parameters and shape performance data during operation change significantly, the setting value of the shape control actuator is actively corrected to perform the operation so that the shape of the steel plate is not greatly disturbed. The above embodiment realizes the shape control of a rolling facility with a stable longitudinal shape by specifying the performance data in which the shape control actuator operation is performed based on the rolling operation performance data together with other operating conditions, similar to the operation of an operator with rich operating experience of a rolling facility. Therefore, by including the identification data of the operator who manually operates the shape control actuator in the input data of the shape control actuator setting model M, it becomes possible to set the shape control actuator to reproduce the operation of an excellent operator.

[0055] On one hand, the reason why the network structure in which the first output layer and the second output layer branch from a common intermediate layer is adopted is that the determination of whether to change the setting of the shape control actuator by the operator is associated with the amount of setting change to be made. That is, when attempting to change the setting of the shape control actuator, it is a case where a certain degree of shape variation of the steel plate is expected, and in that case, the shape control actuator is often changed to a certain extent. Therefore, since it is difficult to separately evaluate the setting change information and the actuator setting value, a common intermediate layer is used. Since it is preferable that the first output layer is configured as a classifier and the second output layer is configured as a regressor, it is advisable to adopt the structure of a neural network having two different output layers as described above.

[0056] By using the shape performance data acquired from the shape detector 10 as the input performance data of the shape control actuator setting model M, when the shape of the steel plate S is greatly disturbed, the setting of the shape control actuator will be greatly changed. Also, by using the target shape data as the input performance data of the shape control actuator setting model M, depending on the deviation between the shape performance data and the target shape data, it is determined whether to change the setting of the shape control actuator and the amount of change in the actuator setting value is set. Thereby, high-precision shape control can be realized.

[0057] The reason for including, as the input to the shape control actuator setting model M, not only the shape data of the steel plate S but also the operation performance data selected from the rolling operation parameters is as follows. That is, the rolling operation parameters affect the rolling load during the rolling of the steel plate S, and since the deflection deformation of the work roll changes complexly due to the rolling load, there is a certain correlation between the rolling operation parameters and the shape performance data. Also, when performing shape control of the rolling equipment 1, even if the setting value of the shape control actuator is the same, it is known that the shape of the rolled steel plate S will be different if the rolling conditions such as the plate thickness, plate width, reduction ratio, and deformation resistance of the steel plate S are different.

[0058] By dividing the data set stored in the database unit 141 into training data and test data and performing learning, it is possible to improve the setting change information of the shape control actuator and the setting accuracy of the actuator setting value. For example, using the training data, the weight coefficients of the neural network are learned, and while appropriately changing the structure of the neural network (the number of hidden layers and the number of nodes), a shape control actuator setting model M may be obtained so that the correct answer rate of the setting change information and the actuator setting value in the test data becomes high. For updating the weight coefficients, the error backpropagation method can be used.

[0059] The shape control actuator setting model M may be updated to a new model by relearning, for example, every one month or every one year. This is because as the data stored in the database unit 141 increases, it becomes possible to perform highly accurate shape control actuator setting. Further, by updating the shape control actuator setting model M based on the latest data, a shape control actuator setting model M that reflects changes in operating conditions over time can be generated.

[0060] 〔Steel Plate Shape Control Method〕 Next, with reference to FIG. 7, a steel plate shape control method according to an embodiment of the present invention will be described.

[0061] In the steel plate shape control method according to an embodiment of the present invention, the shape control actuator of the rolling facility 1 is set using the shape control actuator setting model M generated by the machine learning unit 143. FIG. 7 shows a configuration example of a steel plate shape control system including a shape control actuator setting unit 150 in the rolling facility 1. The shape control actuator setting unit 150 shown in FIG. 7 may be provided in a general-purpose computer such as a workstation or a personal computer. Further, the shape control actuator setting unit 150 has an input unit for acquiring and a storage unit for storing the shape control actuator setting model M generated by the shape control actuator setting model generation unit 140 via a network. However, the shape control actuator setting unit 150 may be configured inside the control computer 100 or the rolling control controller 110 for controlling the rolling facility 1.

[0062] In the shape control of the steel sheet S using the shape control actuator setting unit 150 shown in FIG. 7, an initial set value for rolling the steel sheet S is calculated by the control computer 100. After the shape control actuator is initially set by the rolling control controller 110, the rolling pass is started. Then, during the execution of the rolling pass, the data acquisition unit 120 acquires operation result data selected from the rolling operation parameters of the rolling facility 1 and shape result data acquired from the shape detector 10. The data acquired by the data acquisition unit 120 is sent to the shape control actuator setting unit 150. The shape control actuator setting unit 150 inputs the input data acquired by the data acquisition unit 120 to the shape control actuator setting model M acquired from the shape control actuator setting model generation unit 140 and stored in the storage unit, thereby calculating setting change information and an actuator set value. Then, based on the calculated setting change information and actuator set value, a new shape control actuator setting command value is generated and sent to the rolling control controller 110. When the setting change information indicates that the setting of the shape control actuator is to be changed, the calculated actuator set value is used as the new shape control actuator setting command value. On the other hand, when the setting change information indicates that the setting of the shape control actuator is not to be changed, no change is made to the new shape control actuator setting command value so as to maintain the current shape control actuator set value. Thereby, the rolling control controller 110 can operate the shape control actuator of the rolling facility 1 by updating or maintaining the current shape control actuator setting according to the shape control actuator setting command value acquired from the shape control actuator setting unit 150.

[0063] The set command value of the shape control actuator is acquired for each control cycle of the shape control set in advance, and sent to the rolling control controller 110, so that the set value of the shape control actuator is updated in the longitudinal direction of the steel sheet S, and stable shape control of the steel sheet S can be performed. Since it takes a certain amount of time for the steel sheet S to reach the shape detector 10 after updating the set value of the shape control actuator of the rolling facility 1, the control cycle of the shape control is preferably set to a cycle of about 0.2 to 10 seconds. Further, in the conventional automatic shape feedback control, based on the shape performance data detected by the shape detector 10, the set value of the shape control actuator is often updated at any time by a shape control actuator setting model M constituted by a physical model or the like. Therefore, the shape control of the steel sheet using the shape control actuator setting unit 150 of the present embodiment can be realized by replacing a part of the system for executing the conventional automatic shape feedback control. Further, in parallel with the conventional automatic shape feedback control system, the shape control actuator setting unit 150 of the present embodiment may be provided, and the shape control of the rolling facility may be executed by switching as necessary. For example, in the first pass when rolling a steel sheet through a plurality of rolling passes, the material and thickness variation of the base material may be large, and it may be necessary to frequently operate the shape control actuator. Therefore, in the first pass or the initial rolling pass, shape control using the conventional automatic shape feedback control system is performed, and in the subsequent rolling passes, shape control of the steel sheet using the shape control actuator setting unit 150 of the present embodiment may be performed. Further, at the tip and tail ends of the steel sheet, the shape on the inlet side of the rolling mill may be disturbed, and it may be necessary to frequently operate the shape control actuator. Therefore, at the tip and tail ends of the steel sheet, shape control using the conventional automatic shape feedback control system is performed, and in the steady part, shape control of the steel sheet using the shape control actuator setting unit 150 of the present embodiment may be performed.

Example

[0064] Examples of the present invention are shown below. In this example, first, a shape control actuator setting model was generated for a 12-stage cluster rolling mill. The rolling mill used in this example is a rolling mill with a work roll diameter of 70 to 120 mm and a maximum rolling speed of 600 m / min. The steel plate to be rolled is high-carbon steel, with a base material plate thickness of 1.2 to 2.3 mm and a plate width of 400 to 1070 mm. The steel plate with a product plate thickness of 0.4 to 1.8 mm was rolled with 1 to 15 rolling passes. Also, the rolling mill used is a reversing rolling mill, and a shape detector is arranged between two tension reels on the left and right centered on the rolling mill, and the shape performance data of the steel plate can be obtained in any rolling pass, whether rightward or leftward. The shape detector is of the split load cell type, with 44 channels of load cells built into the roll axis direction, and the detection width per channel is 25.4 mm. From the shape detector, the value of I-unit at each position in the plate width direction is output as the elongation difference in the width direction of the steel plate. Also, in this example, 420 coils of the above steel plate were rolled. During that time, an operator with rich operating experience of this rolling mill manually operated the shape control actuator, and the learning data was accumulated in the database section.

[0065] In generating the shape control actuator setting model M, the inlet plate thickness, outlet plate thickness, and plate width were selected as the dimensional information of the steel plate from the rolling operation parameters. Also, the steel grade code of the steel plate was selected as the attribute information of the base material. Furthermore, from the rolling condition information, the rolling pass number, work roll diameter, rolling speed, coolant flow rate, and coolant temperature were selected. In this example, in addition to those rolling operation parameters, as the target shape of the steel plate, the coefficients λ1 to λ4 when the shape of the steel plate shown in formula (1) was approximated by a fourth-order function were selected. On the other hand, the shape performance data when the steel plate was rolled with the above rolling operation parameters was obtained using the I-unit output from the shape detector as described above. However, for the width direction position of the steel plate, the actual performance data of the I-unit at each position was collected using the value normalized by the maximum width that can be rolled by this rolling mill.

[0066] On the one hand, the shape control actuators used for the actuator installation values that are the output data of the shape control actuator setting model M are selected from six types: the work roll bender, the intermediate roll bender, the leveling, the edge position, the quarter position, and the center position of the lower backup roll, and the performance data of these setting values were obtained. In this embodiment, the performance data of the setting change information was specified from the performance data of the actuator setting values obtained as time series data by the method shown in FIG. 5. Then, a data set composed of the shape performance data, the operation performance data of the rolling operation parameters, the performance data of the setting change information of the shape control actuator, and the performance data of the setting values of the shape control actuator was accumulated in the database unit 141. When the number of data sets accumulated in the database unit 141 reached 100,000, the shape control actuator setting model generation unit 140 performed data division by preprocessing. As the learning data, 70% of all the data sets were used to perform the learning of the neural network, and the remaining data sets were used as the test data to evaluate the prediction accuracy.

[0067] The neural network having two output layers used in this embodiment used a network structure composed of an input layer, seven intermediate layers having 512 nodes, a first output layer, and a second output layer as shown in FIG. 6. The ReLU function was used as the activation function. In this embodiment, the output of the activation function was input to a dropout layer with a probability of 0.3. The number of nodes in the input layer was 64, the first output layer had 1 node, the second output layer had 6 nodes, and the normalized setting values of the shape control actuator were used.

[0068] As the loss function in the learning calculation, the mean squared error calculated from the predicted value and the actual measured value of the network was used. For the optimization calculation method, ADAM optimization was selected. Also, in machine learning, mini-batch learning (mini-batch gradient descent method) was used, with a batch size of 512 and 500 epochs. However, when the loss function for the test data did not improve over 30 epochs or more, the learning rate was decreased to 1 / 10 and the learning was continued. Using the shape control actuator setting model M generated as described above, the output value of the shape control actuator setting value and the actual result value were compared for the test data. In this case, when the output of the first output layer is to change the setting of the shape control actuator, the shape control actuator was updated to the value output by the second output layer. On the other hand, when the output of the first output layer does not change the setting of the shape control actuator, the output value of the shape control actuator setting value was maintained as the value of the previous step.

[0069] In contrast, as a comparative example, using the same dataset as the dataset stored in the database section above, learning was performed using a normal neural network with one output layer that outputs only the setting value of the shape control actuator, and the output value of the generated model and the actual result value were compared. In this case, the output value of the shape control actuator setting value used was the one updated in response to the update of the input data. Note that the input data to the neural network in the comparative example was the same as that in this example.

[0070] The comparison results between the output values and the actual values of the shape control actuator setting values in this embodiment and the comparative examples are shown in FIG. 8. FIG. 8 shows, for this embodiment and the comparative examples, the differences between the output values and the actual values for six shape control actuators, and represents their mean squared error (MSE) as a loss function. From FIG. 8, it can be seen that the loss function of the prediction result using the neural network having two output layers in this embodiment is significantly improved compared to the comparative example. This indicates that the setting values of the shape control actuators reset by the shape control actuator setting method according to this embodiment are in relatively good agreement with the results manually set by the operator. That is, it is shown that the variation of the setting values of the shape control actuators during the rolling of the steel sheet can be suppressed.

[0071] Subsequently, in Comparative Example 2, rolling for shape control was performed on the same material to be rolled as in the embodiment by conventional automatic shape feedback control. In the conventional automatic shape feedback control of Comparative Example 2, the shape control actuator setting values were updated as needed according to the shape deviation data, which is the difference between the actual shape detected by the shape detector 10 and the target shape. Further, in Comparative Example 3, in the conventional automatic shape feedback control of Comparative Example 2, the frequency of updating the shape control actuator setting values was reduced by updating the shape control actuator setting values only when the shape deviation data exceeded a preset threshold value. In this case, the threshold value for updating the setting values of the shape control actuators was set to 10 (I-unit).

[0072] FIG. 9 shows the results of evaluating the shape deviation in the longitudinal direction of the rolled steel sheet, that is, the deviation between the target shape and the actual shape, by the mean absolute error (MAE) for the present embodiment, Comparative Example 2, and Comparative Example 3. The larger the value of the mean absolute error (MAE), the greater the deviation between the target shape and the actual shape. In Comparative Example 2, the set value of the shape control actuator was frequently changed during rolling, and the fluctuations in the rolling state became disturbances, resulting in unstable shape of the steel sheet in the longitudinal direction. Therefore, as shown in FIG. 9, the mean absolute error (MAE) in Comparative Example 2 was 8.71 (I-unit). On the other hand, in Comparative Example 3, since a threshold was set for the deviation between the target shape and the actual shape for determining whether to operate the shape control actuator, the degree of frequent change of the shape control actuator set value during rolling was alleviated, and the shape of the steel sheet in the longitudinal direction was somewhat stable. However, since the responsiveness to the shape deviation data decreased due to the setting of the threshold, the mean absolute error (MAE) remained at 9.2 (I-unit). In addition, when the shape deviation data was close to the set threshold, the shape control actuator set value was frequently updated, and there were parts where the shape of the steel sheet was unstable in a part of the longitudinal direction. In contrast, in the present embodiment, when a large fluctuation occurred with respect to the target shape, a significant disturbance of the steel sheet shape could be suppressed by actively operating the shape control actuator. In addition, control was realized in which the shape control actuator was not actively operated for small fluctuations in the rolling operation parameters and shape actual data during operation. Therefore, the mean absolute error (MAE) was improved to 7.51 (I-unit).

[0073] As described above, the embodiments to which the invention made by the present inventors is applied have been described. However, the present invention is not limited by the description and drawings that form a part of the disclosure of the present invention according to the present embodiment. That is, all other embodiments, examples, and operation techniques made by those skilled in the art based on the present embodiment are included in the scope of the present invention.

Description of Reference Numerals

[0074] 1 Rolling equipment 2 Work roll 3 Intermediate roll 4 Small backup roll 5 Upper backup roll 6 Lower backup roll 7 Payoff reel 8 Right tension reel 9 Left tension reel 10 Shape detector 11 Intermediate roll bender 12 Work roll bender 100 Control computer 110 Rolling control controller 120 Data acquisition unit 140 Shape control actuator setting model generation unit 141 Database unit 142 Preprocessing unit 143 Machine learning unit M Shape control actuator setting model S Steel plate

Claims

1. A method for generating a shape control actuator setting model for a rolling facility for setting a shape control actuator in a rolling facility including a shape control actuator for controlling the shape of a steel sheet and a shape detector for detecting the shape of the steel sheet, comprising: A neural network method using a plurality of learning data with one or more operation performance data selected from the rolling operation parameters of the rolling facility, shape performance data obtained from the shape detector, performance data of the setting value of the shape control actuator, and performance data of the setting change information of the shape control actuator indicating whether the operator changed the setting value of the shape control actuator during the operation of the rolling facility as input performance data, setting change information indicating whether to change the setting of the shape control actuator, and the setting value of the shape control actuator as output performance data, including a setting model generation step of generating a shape control actuator setting model. A method for generating a shape control actuator setting model for a rolling facility.

2. The method for generating a shape control actuator setting model for a rolling facility according to claim 1, wherein the rolling operation parameters include target shape data of the steel sheet.

3. The neural network includes, as an output layer, a first output layer that outputs the setting change information and a second output layer that outputs the setting value of the shape control actuator, and the first output layer and the second output layer have a network structure branched from a common intermediate layer. The method for generating a shape control actuator setting model for a rolling facility according to claim 1 or 2.

4. A method for setting a shape control actuator of a rolling facility including a shape control actuator for controlling the shape of a steel sheet and a shape detector for detecting the shape of the steel sheet, comprising: One or more operating performance data selected from the rolling operation parameters of the rolling equipment, shape performance data acquired from the shape detector, performance data of the set values of the shape control actuator, and performance data of the setting change information of the shape control actuator indicating whether or not the operator changed the set value of the shape control actuator during the operation of the rolling equipment are used as input performance data, setting change information indicating whether or not to change the setting of the shape control actuator, and the set value of the shape control actuator are used as output performance data. Using a shape control actuator setting model generated by a neural network method using a plurality of learning data, when the setting change information indicates that the setting of the shape control actuator is to be changed, the method includes a resetting step of resetting the shape control actuator to the set value. A method for setting a shape control actuator of a rolling equipment.

5. The neural network includes, as an output layer, a first output layer that outputs the setting change information and a second output layer that outputs the set value of the shape control actuator, and the first output layer and the second output layer have a network structure branched from a common intermediate layer. The resetting step resets the shape control actuator to the set value when the output of the first output layer indicates that the setting of the shape control actuator is to be changed. The method for setting a shape control actuator of a rolling equipment according to claim 4.

6. A method for controlling the shape of a steel sheet, including a step of resetting the shape control actuator during rolling of the steel sheet using the method for setting a shape control actuator of a rolling equipment according to claim 4 or 5.

7. A shape control device for a steel sheet, comprising means for resetting the shape control actuator during rolling of the steel sheet using the method for setting a shape control actuator of a rolling equipment according to claim 4 or 5.

8. A method for manufacturing a steel sheet, including a step of manufacturing the steel sheet using the method for controlling the shape of a steel sheet according to claim 6.

Images