Method for predicting wedge of rolled material, wedge control method, manufacturing method, method for generating wedge prediction model, and wedge control device

Abstract

To provide a method for predicting a wedge of a rolled material and a method for generating a wedge prediction model which can accurately predict a wedge of a rolled material, without requiring much labor, in a rolling step of performing reverse rolling of a rolled material through a plurality of rolling passes using a rolling machine.SOLUTION: A method for predicting a wadge of a rolled material according to the present invention, which predicts a wedge pf a rolled material at an outlet side of a rolling machine in a pass to be predicted which is a rolling pass selected from a plurality of rolling passes, in a rolling step of reverse-rolling the rolled material through the plurality of rolling passes using the rolling machine, includes a step of predicting the wedge of the rolled material at the outlet side of the rolling machine in the pass to be predicted, using a wedge prediction model learnt by machine learning, in which the wedge of the rolled material at an inlet side of the rolling machine in the pass to be predicted, at least one of rolling operation parameters and data concerning a rolling direction of the rolled material are included in input and the wedge of the rolled material at the outlet side of the rolling machine in the pass to be predicted is included in output.SELECTED DRAWING: Figure 9

Description

[Technical Field]

[0001] The present invention relates to a wedge prediction method for rolled material, a wedge control method, a manufacturing method, a method for generating a wedge prediction model, and a wedge control device. [Background technology]

[0002] Plate rolling mills and roughing mills in hot strip rolling lines use multiple passes to reverse-roll the material. Reverse rolling is sometimes called a reversing mill. Because the time between passes is longer than in continuous rolling, reverse rolling is typically performed when the material is thick and the temperature drop is small. When the material is thick, even if there is a roll gap difference between the work side (WS) and drive side (DS) of the rolling mill, camber or meandering of the material is unlikely to become apparent. However, when there is a roll gap difference between the work side and drive side of the rolling mill, a wedge, which is the difference in thickness between one end and the other end of the material in the width direction, is likely to form. Even if a wedge forms in the material, it rarely directly causes operational problems when the material is thick. However, if the wedge formed on the rolled material becomes large, camber or meandering may occur as the rolling pass progresses and the thickness of the rolled material becomes thinner. As a result, the rolled material is more likely to encounter threading problems, such as contact with equipment such as side guides of the rolling mill. Furthermore, even if the product thickness of the rolled material is thick, if the wedge of the rolled material becomes large, the thickness accuracy of the product deteriorates, and the product yield decreases due to poor dimensional accuracy.

[0003] Given this background, a technology has been proposed for adjusting the leveling of a rolling mill (the difference in the roll gap between the work side and drive side of the rolling mill) to control the wedge of the rolled material in reverse rolling. Specifically, Patent Document 1 describes a method for setting and controlling wedges in rolling metal sheets and other plate materials. Specifically, the method described in Patent Document 1 first measures and stores the wedge and width-center thickness of the plate according to the distance from the plate tip during odd-numbered passes of the roughing mill. Next, during even-numbered passes of the roughing mill, the stored wedge and width-center thickness, the roll gap leveling influence coefficient on the wedge, and the delivery thickness calculated by mill setting are used to determine a roll gap leveling control amount according to the distance from the plate tip. The determined roll gap leveling control amount is then applied to the roll gap leveling of the roughing mill by feedforward control. Patent Document 1 claims that this method enables plate materials of the same thickness to be rolled on the work side and drive side.

[0004] Furthermore, Patent Document 2 describes a wedge control method for hot-rolled material that can control the wedge and camber during hot rolling within an allowable range. Specifically, the method described in Patent Document 2 first determines a roll gap control amount sufficient to eliminate at least the amount of wedge exceeding the threshold when the wedge of the hot-rolled material measured by a wedge meter installed on the delivery side of a finishing rolling mill exceeds a threshold. The roll gap control amount is determined from the relationship between the roll gap control amount and the wedge change amount, which are determined in advance according to the thickness, width, and material quality of the hot-rolled material. Next, the camber amount on the delivery side of the roughing mill is predicted, and the leveling amount of the next hot-rolled material in the roughing mill is controlled so that the predicted camber amount does not exceed a predetermined threshold. Patent Document 2 claims that this method makes it possible to control the wedge and camber during hot rolling within an allowable range. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 4685777 [Patent Document 2] Japanese Patent Application Publication No. 11-10215 Summary of the Invention [Problem to be solved by the invention]

[0006] The method described in Patent Document 1 calculates a roll gap leveling control amount according to the distance from the leading edge of a sheet material using a roll gap leveling influence coefficient on a wedge in even-numbered rolling passes of a roughing mill. However, in this case, the roll gap leveling influence coefficient on a wedge means the amount of change in the sheet material wedge relative to the amount of change in roll gap leveling, but a specific method for specifying this coefficient is not disclosed. In general, the roll gap leveling influence coefficient on a wedge varies depending on various conditions, such as the thickness, width, and reduction of the rolled material, as well as the rigidity difference between the work side and the drive side of the rolling mill. Therefore, it requires a great deal of effort to specify in advance the roll gap leveling influence coefficient on a wedge that corresponds to these conditions.

[0007] On the other hand, Patent Document 2 discloses that the relationship between the roll gap control amount and the wedge change amount is specified in advance depending on the thickness, width, and material quality of the hot-rolled material. The method described in Patent Document 2 groups the thickness, width, and material quality of the hot-rolled material, and specifies the relationship between the roll gap control amount and the wedge change amount for each group based on actual measurement results of the relationship between the roll gap control amount and the wedge change amount. In this case, the thickness and width of the hot-rolled material are parameters related to the dimensions of the hot-rolled material, and the material quality is a parameter correlated with the rolling load because it affects the deformation resistance. However, the relationship between the roll gap control amount and the wedge change amount varies not only depending on parameters related to the rolling conditions of the hot-rolled material, but also depending on the characteristics of the rolling mill, such as the difference in rigidity between the work side and the drive side of the rolling mill. For this reason, it is difficult to achieve high-precision wedge control simply by specifying the relationship between the roll gap control amount and the wedge change amount depending on the thickness, width, and material quality of the hot-rolled material, which are parameters related to the rolling conditions of the hot-rolled material.

[0008] The present invention has been made to solve the above-mentioned problems, and its object is to provide a method for predicting the wedge of a rolled material and a method for generating a wedge prediction model, which are capable of accurately predicting the wedge of a rolled material without requiring much effort in a rolling process in which the rolled material is reverse rolled through multiple rolling passes using a rolling mill. Another object of the present invention is to provide a method for controlling the wedge of a rolled material and a wedge control device, which are capable of reducing the wedge of the rolled material. Another object of the present invention is to provide a method for manufacturing rolled material, which is capable of producing rolled material with small thickness deviation. [Means for solving the problem]

[0009] The method for predicting the wedge of a rolled material according to the present invention is a method for predicting the wedge of a rolled material at the exit side of a rolling mill in a prediction target pass, which is a rolling pass selected from a plurality of rolling passes in a rolling process in which the rolled material is reverse rolled through a plurality of rolling passes using the rolling mill, and includes a step of predicting the wedge of the rolled material at the exit side of the rolling mill in the prediction target pass using a wedge prediction model learned by machine learning, which includes as inputs the wedge of the rolled material at the entry side of the rolling mill in the prediction target pass, at least one of the rolling operation parameters, and data related to the rolling direction of the rolled material, and outputs the wedge of the rolled material at the exit side of the rolling mill in the prediction target pass.

[0010] The rolling operation parameters may include a leveling amount of the rolling mill.

[0011] The method for controlling the wedge of a rolled material according to the present invention includes an operation amount calculation step of calculating an operation amount of the rolling mill in the pass to be predicted based on the wedge of the rolled material at the exit side of the rolling mill predicted using the wedge prediction method for rolled material according to the present invention, so that the size of the wedge of the rolled material at the exit side of the rolling mill in the pass to be predicted is equal to or less than a predetermined threshold value.

[0012] A method for producing a rolled material according to the present invention includes a step of producing a rolled material using the method for wedge control of a rolled material according to the present invention.

[0013] The method for generating a wedge prediction model for a rolled material according to the present invention is a method for generating a wedge prediction model for a rolled material at the exit side of a rolling mill in a rolling process in which a rolled material is reverse rolled through a plurality of rolling passes using the rolling mill, the method predicting the wedge of the rolled material at the exit side of the rolling mill in a prediction target pass which is a rolling pass selected from the plurality of rolling passes, and includes the steps of acquiring a plurality of data sets each consisting of a wedge of the rolled material at the entry side of the rolling mill in the plurality of rolling passes, at least one of the rolling operation parameters, data related to the rolling direction of the rolled material, and a wedge of the rolled material at the exit side of the rolling mill, and generating a wedge prediction model using the acquired plurality of data sets as training data as input, the wedge of the rolled material at the entry side of the rolling mill, at least one of the rolling operation parameters, and data related to the rolling direction of the rolled material, and outputting the wedge of the rolled material at the exit side of the rolling mill.

[0014] The wedge control device for a rolled material according to the present invention is a wedge control device for a rolled material that controls a wedge of the rolled material at the exit side of a rolling mill in a prediction target pass that is a rolling pass selected from a plurality of rolling passes in a rolling process in which a rolled material is reverse rolled through a plurality of rolling passes using the rolling mill, and the device includes as inputs the wedge of the rolled material at the entry side of the rolling mill in the prediction target pass, at least one of rolling operation parameters, and data related to the rolling direction of the rolled material, The method includes a wedge prediction unit that predicts the wedge of the rolled material at the exit side of the rolling mill in the prediction target pass using a wedge prediction model learned by machine learning, which outputs the wedge of the rolled material at the exit side of the rolling mill, and an operation amount calculation unit that calculates an operation amount of the rolling mill in the prediction target pass based on the wedge of the rolled material at the exit side of the rolling mill predicted by the wedge prediction unit so that the size of the wedge of the rolled material at the exit side of the rolling mill in the prediction target pass is equal to or less than a predetermined threshold. [Effects of the Invention]

[0015] According to the method for predicting the wedge of a rolled material and the method for generating a wedge prediction model of the present invention, it is possible to accurately predict the wedge of a rolled material without requiring much effort in a rolling process in which a rolling mill is used to perform reverse rolling of the rolled material through multiple rolling passes. Also, according to the method for controlling the wedge of a rolled material and the wedge control device of the present invention, it is possible to reduce the wedge of the rolled material. Also, according to the method for producing a rolled material of the present invention, it is possible to produce a rolled material with a small thickness deviation. [Brief explanation of the drawings]

[0016] [Figure 1] FIG. 1 is a schematic diagram showing the configuration of a plate rolling line that produces steel plates using rolling mills according to one embodiment of the present invention. [Figure 2] FIG. 2 is a diagram showing a detailed configuration of the rolling mill shown in FIG. [Figure 3] FIG. 3 is a diagram showing the configuration of a control device that controls the rolling mill shown in FIG. [Figure 4] FIG. 4 is a diagram for explaining the wedge of the rolled material. [Figure 5] FIG. 5 is a block diagram showing the configuration of a rolled material wedge prediction model generating device according to one embodiment of the present invention. [Figure 6] FIG. 6 is a block diagram showing a detailed configuration of the model generating unit shown in FIG. [Figure 7] FIG. 7 is a diagram showing the contact state between the work roll and the backup roll. [Figure 8] FIG. 8 is a block diagram showing the configuration of a rolled material wedge prediction device according to one embodiment of the present invention. [Figure 9] FIG. 9 is a block diagram showing a detailed configuration of the data acquisition unit and the wedge prediction unit shown in FIG. [Figure 10] FIG. 10 is a block diagram showing the configuration of a wedge control device for rolling material according to one embodiment of the present invention. [Figure 11]FIG. 11 is a flowchart showing the flow of a method for controlling the wedge of a rolled material according to one embodiment of the present invention. [Figure 12] FIG. 12 is a block diagram showing a detailed configuration of the rolling material wedge control device shown in FIG. DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, a rolled material wedge prediction method, wedge control method, manufacturing method, wedge prediction model generation method, and wedge control device according to one embodiment of the present invention will be described with reference to the drawings.

[0018] [Rolling mill] First, the configuration of a rolling mill according to one embodiment of the present invention will be described with reference to FIGS.

[0019] FIG. 1 is a schematic diagram showing the configuration of a plate rolling line that produces steel plates using a rolling mill according to one embodiment of the present invention. As shown in FIG. 1, the plate rolling line 1 includes a heating furnace 2, a descaling device 3, a rolling mill 4, a cooling device 5, and a leveling machine 6. A cast slab (not shown) is charged into the heating furnace 2, heated to a predetermined temperature, and then extracted from the heating furnace 2 as a hot slab. The hot slab extracted from the heating furnace 2 has primary scale formed on its surface removed by the descaling device 3, and is then reverse-rolled to a predetermined thickness in the rolling mill 4. Once the hot slab is fed to the rolling mill 4 and rolling begins, the material to be rolled is simply referred to as the "rolled material." A cooling device 5 is provided downstream of the rolling mill 4, where the rolled material S is cooled to a predetermined temperature and then leveled by the leveling machine 6. The rolled material S leveled by the leveling machine 6 is transported to a cooling bed 7 for air cooling and cooled to a predetermined temperature. The rolling mill 4 that performs reverse rolling is sometimes called a reversing rolling mill.

[0020] Generally, in a plate rolling line, a rolling mill is used to reduce the thickness of the rolled material to the product thickness through multiple rolling passes of 5 to 50 passes. This rolls the thickness of a hot slab from 200 to 300 mm to 8 to 150 mm. On the other hand, in a reversing rolling mill used as a roughing mill installed in a hot rolling line that produces hot-rolled steel plates, multiple rolling passes of 3 to 5 passes are performed, and the rolled material after rough rolling is sent to a finishing rolling mill where it is rolled to the product thickness. In this case, the rolled material is rolled to 30 to 80 mm by the reversing rolling mill.

[0021] In the plate rolling line 1 shown in FIG. 1, the side of the rolling mill 4 on which the heating furnace 2 is located is sometimes referred to as the "front side," and the side of the rolling mill 4 on which the cooling device 5 is located is sometimes referred to as the "rear side." When reverse rolling of the rolled material S is performed by the rolling mill 4, the upstream side of each rolling pass in the direction of travel of the rolled material S is called the "entrance side," and the downstream side of the direction of travel of the rolled material S is called the "exit side." In other words, when rolling is performed by the rolled material S traveling from the front side to the rear side of the rolling mill 4, the left side of the rolling mill 4 shown in FIG. 1 is the entry side, and the right side is the exit side. On the other hand, when rolling is performed by the rolled material S traveling from the rear side of the rolling mill 4 to the front side, the right side of the rolling mill 4 shown in FIG. 1 is the entry side, and the left side is the exit side. In addition, the rolling pass in which the rolled material S is transported from the front to the rear of the rolling mill 4 and rolled is called the "forward pass," and the rolling pass in which the rolled material S is transported from the rear to the front of the rolling mill 4 and rolled is called the "reverse pass."

[0022] 2(a) and 2(b) are diagrams showing the detailed configuration of the rolling mill 4. As shown in FIGS. 2(a) and 2(b), the rolling mill 4 is a single-stand, four-high rolling mill. The rolling mill 4 is equipped with a pair of work rolls 41a, 41b arranged vertically across the pass line PL. The work rolls 41a, 41b are supported by backup rolls 42a, 42b, respectively. One end of each of the work rolls 41a, 41b is connected to a drive motor via a coupling or a reducer. The drive motor rotates the work rolls 41a, 41b. In this case, the side of the rolling mill 4 on which the drive motor is located is called the drive side (DS), and the opposite side is called the work side (WS).

[0023] The backup roll 42a (42b) is supported by bearing housings (backup roll chocks) 43a1, 43b1 (43a2, 43b2) arranged at the axial ends. The rolling load applied to the rolled material S is transmitted to the housings 44a, 44b via the backup roll chocks 43a1, 43b1 (43a2, 43b2). Load cells 45a, 45b, which serve as load detectors, are arranged between the housings 44a, 44b and the backup roll chocks 43a2, 43b2, making it possible to measure the rolling load applied to the rolled material S. In this case, the sum of the measurement values measured by the load cell 45a and the load cell 45b is sometimes called the rolling load or sum load, and the difference between the measurement values measured by the load cell 45a and the load cell 45b is sometimes called the differential load.

[0024] Screw down devices 46a, 46b are arranged on the work side and drive side of the rolling mill 4. The screw down devices 46a, 46b adjust the gap (also called the roll gap or roll opening) between the work rolls 41a and 41b by vertically displacing the backup roll chocks 43a1, 43b1, respectively. The vertical positions of the backup roll chocks 43a1, 43b1 are usually configured so that they can be measured by a displacement meter (not shown). Because it is difficult to actually measure the gap between the work rolls 41a and 41b, the roll opening is conventionally set by the vertical positions of the backup roll chocks 43a1, 43b1.

[0025] The screw down devices 46a, 46b are equipped with electric or hydraulic screw down mechanisms. The hydraulic screw down mechanism (also called hydraulic screw down) displaces the backup roll chocks 43a1, 43b1 in the up and down direction by controlling the hydraulic pressure of a hydraulic cylinder. The electric screw down device (also called electric screw down) displaces the backup roll chocks 43a1, 43b1 in the up and down direction by rotating an electric motor for the screw down device and moving a screw down screw up and down via a gear. The screw down devices 46a, 46b are not limited to a type that displaces the backup roll 42a in the up and down direction from the upper part of the housing 44a, 44b, but may be a type that displaces the backup roll 42b in the up and down direction from the lower part of the housing 44a, 44b.

[0026] The screw down devices 46a, 46b typically set the vertical positions of the backup roll chocks 43a1, 43b1 so that the gap between the work rolls 41a and 41b is the same on the work side and the drive side. However, the vertical positions of the backup roll chocks 43a1, 43b1 may be set differently. Setting the vertical positions of the backup roll chocks 43a1, 43b1 to be different is called leveling, and the difference in the vertical positions of the backup roll chock 43a1 on the work side and the backup roll chock 43b1 on the drive side is sometimes called the leveling amount. Leveling is intended to impart a difference in roll gap between the work side and the drive side work rolls, but in practice, it is performed by imparting a difference in the vertical positions of the backup roll chocks 43a1, 43b1 as described above.

[0027] FIG. 3 is a diagram showing the configuration of a control device that controls the rolling mill 4. As shown in FIG. 3, front conveying rolls 8 and rear conveying rolls 9 are arranged at the front and rear of the rolling mill 4. The front conveying rolls 8 and rear conveying rolls 9 are configured to be capable of rotating forward and reverse depending on the direction of travel of the rolled material S in each rolling pass. The plate rolling line 1 is equipped with a programmable logic controller (PLC) 11 for controlling each device that constitutes the rolling mill 4, and a control computer (process computer) 12 that issues control commands to the control controller 11. The plate rolling line 1 is also equipped with a host computer 13 that issues production instructions to each piece of equipment, including the rolling mill 4. Rolling control in the rolling mill 4 is performed by the control computer 12 setting operating conditions for the rolled material S based on the host computer 13 or production instructions from the host computer 13. In addition, the control controller 11 has the function of collecting information obtained from various sensors (thickness gauge, width gauge, temperature gauge, wedge gauge, etc.) installed on the rolling mill 4 at a predetermined sampling period and outputting the information to the control computer 12.

[0028] The rolling mill 4 is equipped with wedge gauges on its front and rear surfaces for measuring the wedge of the rolled material S. Here, the wedge gauge arranged on the front surface of the rolling mill 4 is referred to as the front wedge gauge 21. Furthermore, the wedge gauge arranged on the rear surface of the rolling mill 4 is referred to as the rear wedge gauge 22. This makes it possible to measure the wedge of the rolled material S at the entry side and exit side of the rolled material S being rolled by the rolling mill 4. In this case, the wedge of the rolled material S measured at the entry side in the traveling direction of the rolled material S is referred to as the entry wedge, and the wedge of the rolled material S measured at the exit side in the traveling direction of the rolled material S is referred to as the exit wedge. In other words, when the rolled material S travels from the front surface to the rear surface of the rolling mill 4 and is rolled, the wedge of the rolled material S measured by the front wedge gauge 21 is the entry wedge, and the wedge of the rolled material S measured by the rear wedge gauge 22 is the exit wedge. On the other hand, when the rolled material S advances from the rear surface of the rolling mill 4 to the front surface and is rolled, the wedge of the rolled material S measured by the rear surface wedge meter 22 becomes the entrance wedge, and the wedge of the rolled material S measured by the front surface wedge meter 21 becomes the exit wedge.

[0029] FIG. 4 is a diagram illustrating the wedge of the rolled material S. The wedge of the rolled material S refers to the thickness deviation in the width direction of the rolled material S. Specifically, the wedge is defined as the difference between the thickness h1 of the rolled material S at a predetermined distance We from the end of the width direction on the work side and the thickness h2 of the rolled material S at the distance We from the end of the width direction on the drive side in the cross-sectional shape of the rolled material S shown in FIG. 4. In this case, a positive wedge may be defined as a case where the thickness h1 at the end of the width direction on the work side is greater than the thickness h2 at the end of the width direction on the drive side, and a negative wedge may be defined as a case where the thickness h1 at the end of the width direction on the work side is smaller than the thickness h2 at the end of the width direction on the drive side. The distance We from the end of the width direction, which serves as the reference for the wedge, is set, for example, in the range of 15 to 200 mm.

[0030] The wedge gauges (front wedge gauge 21, rear wedge gauge 22) can be thickness gauges that measure the thickness in the width direction of the rolled material S. For example, multiple wedge gauges using X-rays or gamma rays are installed in the width direction of the rolled material S to measure the thickness h1 of the rolled material S at the width direction end of the work side of the rolled material S and the thickness h2 of the rolled material S at the width direction end of the drive side. In addition, the wedge gauge may scan the thickness gauge along the width direction of the rolled material S to measure the thickness distribution in the width direction of the rolled material S, and calculate the wedge from the measured thickness distribution. Furthermore, laser scanning type range finders may be placed on the front and back sides of the rolled material S, and the wedge may be calculated from the difference between the distance information obtained on the front side and the distance information obtained on the back side using distance information measured by the range finders at each position in the width direction of the rolled material S.

[0031] As described above, the rolling mill 4 can acquire actual measurement data of the entry wedge and delivery wedge of the rolled material S for each rolling pass of reverse rolling in which multiple rolling passes are executed. The acquired actual measurement data of the entry wedge and delivery wedge may be configured to be collected by the control computer 12.

[0032] [Rolled material wedge prediction model generation device] Next, the configuration of a rolled material wedge prediction model generating device according to one embodiment of the present invention will be described with reference to FIGS.

[0033] Figure 5 is a block diagram showing the configuration of a rolled material wedge prediction model generation device according to one embodiment of the present invention. As shown in Figure 5, a rolled material wedge prediction model generation device 30 according to one embodiment of the present invention includes a data acquisition unit 31 and a model generation unit 32. The data acquisition unit 31 acquires actual data on rolling operation parameters for each rolling pass and actual data on the rolling direction of the rolled material S for each rolling pass from the control computer 12. The data acquisition unit 31 acquires actual data on the wedges of the rolled material S detected at the front surface of the rolling mill 4 from the front wedge meter 21. The data acquisition unit 31 acquires actual data on the wedges of the rolled material S detected at the rear surface of the rolling mill 4 from the rear wedge meter 22.

[0034] The data acquisition unit 31 identifies the entry side and exit side for each rolling pass based on actual data regarding the rolling direction of the rolled material S acquired from the control computer 12. That is, when the actual data regarding the rolling direction of the rolled material S is data representing a forward pass, the data acquisition unit 31 identifies the wedge of the rolled material S detected by the front wedge meter 21 as an entry wedge, and identifies the wedge of the rolled material S detected by the rear wedge meter 22 as an exit wedge. When the actual data regarding the rolling direction of the rolled material S is data representing a reverse pass, the data acquisition unit 31 identifies the wedge of the rolled material S detected by the rear wedge meter 22 as an entry wedge, and identifies the wedge of the rolled material S detected by the front wedge meter 21 as an exit wedge. Then, the data acquisition unit 31 creates a data set that associates, for each rolling pass, actual data of the inlet wedge of the rolled material S, actual data of the outlet wedge of the rolled material S, actual data regarding the rolling direction of the rolled material S, and actual data of the rolling operation parameters acquired from the control computer 12.

[0035] The model generation unit 32 accumulates the data set constructed by the data acquisition unit 31 and generates a wedge prediction model that predicts the exit wedge of a rolling pass through learning using machine learning. The configuration of the model generation unit 32 will be described in detail with reference to Fig. 6. Fig. 6 is a block diagram showing the detailed configuration of the model generation unit 32. As shown in Fig. 6, the model generation unit 32 includes a database unit 32a and a machine learning unit 32b. The database unit 32a accumulates data sets including actual data on the entry wedge of the rolled material S acquired by the data acquisition unit 31, actual data on the exit wedge of the rolled material S, actual data on the rolling direction of the rolled material S, and actual data on rolling operation parameters.

[0036] The database unit 32a stores 1,000 or more data sets as data sets of rolling passes in a rolling process using the rolling mill 4. Preferably, 5,000 or more data sets, and more preferably, 20,000 or more data sets are stored. The data sets stored in the database unit 32a may be screened as needed. Furthermore, the data sets stored in the database unit 32a may be updated as appropriate within a certain upper limit of the number of data sets. Furthermore, the data sets stored in the database unit 32a may be updated using the latest operational data obtained within a certain period (for example, six months).

[0037] The machine learning unit 32b uses the data set stored in the database unit 32a to generate a wedge prediction model M that receives input data including the entry wedge of the rolled material S, at least one of the rolling operation parameters, and data related to the rolling direction of the rolled material S, and outputs the exit wedge of the rolled material S. The machine learning model used to generate the wedge prediction model M may be any machine learning model that provides sufficient prediction accuracy for the exit wedge for practical use. For example, commonly used neural networks (including deep learning and convolutional neural networks), decision tree learning, random forests, support vector regression, etc. may be used. An ensemble model combining multiple models may also be used. In particular, deep learning allows other operational parameters that are correlated with the exit wedge of the rolled material S to be freely selected as inputs without considering the problem of multicollinearity, thereby improving the prediction accuracy for the exit wedge. For example, a neural network with two to three intermediate layers, three to five nodes, and an activation function using a sigmoid function or a ramp function may be used.

[0038] The machine learning unit 32b may improve the estimation accuracy of the exit wedge by dividing the data set stored in the database unit 32a into training data and test data and performing learning. For example, the machine learning unit 32b may use the training data to learn the weighting coefficients of the neural network, and generate the wedge prediction model M while appropriately changing the structure of the neural network (the number of intermediate layers and the number of nodes) so as to improve the prediction accuracy of the exit wedge using the test data. The weighting coefficients can be updated using an error propagation method.

[0039] The model generation unit 32 may perform re-learning every time the work rolls and backup rolls in the rolling mill 4 are rearranged, and may update to a new wedge prediction model M. Rearranging the work rolls and backup rolls may change the contact state between the housing of the rolling mill 4 and the roll chocks, which may change the relationship between the set value of the leveling amount and the actual gap difference between the work rolls on the work side and the drive side. For this reason, the data sets stored in the database unit 32a may be stored in association with the roll numbers and chock numbers (control numbers) of the backup rolls, the roll numbers and chock numbers (control numbers) of the work rolls, etc., and a wedge prediction model M may be generated for each backup roll and chock and each work roll and chock incorporated in the rolling mill 4.

[0040] [Rolling operation parameters] Next, the rolling operation parameters used to input the wedge prediction model M will be explained.

[0041] The rolling operation parameters are operation parameters that specify the rolling conditions for each rolling pass of the rolled material S in the rolling mill 4. The leveling amount can be used as the operation parameters that specify the rolling conditions. Parameters used in the setting calculations performed by the control computer 12 for each rolling pass, such as the roll gap, rolling load, entry thickness, exit thickness, reduction, width, temperature and deformation resistance of the rolled material S, may also be used. In addition to the above, the operation parameters that specify the rolling conditions may also include the bender load as a setting condition for the actuator that controls the crown and shape of the rolled material S. Furthermore, in a rolling mill that can shift the work rolls in the axial direction, the work roll shift amount may be used, and in a rolling mill that can cross the upper and lower work rolls, the cross angle may be used. Furthermore, the rolling distance, which is the cumulative rolling length after rearranging the work rolls of the rolling mill 4, and the rolling amount, which is the cumulative rolled weight, may also be used.

[0042] The leveling amount set for each rolling pass has a direct effect on the exit wedge of the rolled material S. The entry thickness, exit thickness, reduction, and width affect the plastic flow (transverse flow) near the widthwise ends of the rolled material S when it is rolled, and therefore affect the thickness difference formed at the widthwise ends of the work side and drive side of the rolled material S. The roll gap is set according to the exit thickness, and therefore indirectly affects the exit wedge. The rolling load changes the roll gap difference between the work rolls on the work side and drive side due to the difference in rigidity between the work side and drive side of the rolling mill 4, and affects the exit wedge of the rolled material S. In addition, the entry thickness, exit thickness, reduction, width, temperature and deformation resistance of the rolled material S all affect the rolling load, and therefore a correlation occurs between the entry thickness, exit thickness, reduction, width, temperature and deformation resistance of the rolled material S.

[0043] On the other hand, the settings of the actuators that control the crown and shape of the rolled material S, such as the bender load, work roll shift amount, and cross angle, affect the load distribution in the width direction of the rolled material S when it is rolled. This changes the difference in roll gap between the work rolls on the work side and the drive side, affecting the exit wedge of the rolled material S. The rolling distance and rolling amount affect the wear and thermal expansion of the work rolls, which in turn affect the roll gap distribution in the width direction of the work rolls, resulting in a correlation with the exit wedge of the rolled material S.

[0044] The input of the wedge prediction model M includes at least one of the above rolling operation parameters. In this case, it is preferable that the rolling operation parameters used as inputs to the wedge prediction model M include the leveling amount. This is because the leveling amount has a direct effect on the delivery wedge of the rolled material S. Furthermore, it is preferable that the rolling operation parameters used as inputs to the wedge prediction model M include, in addition to the leveling amount, the rolling load, entry thickness, width, and rolling temperature. This improves the prediction accuracy of the delivery wedge by the wedge prediction model M.

[0045] The input to the wedge prediction model M may include parameters relating to the dimensions of the rolling mill 4, such as the work roll diameter, work roll barrel length, backup roll diameter, and backup roll barrel length, as necessary. This is because these affect the elastic deformation of the rolling mill 4 that occurs when rolling the rolled material S, and thereby affect the outlet wedge of the rolled material S. These are not included in the operation parameters that specify the rolling conditions of the rolled material S, and are not operation parameters that change for each rolling pass, so they may be used as input to the wedge prediction model M along with the above rolling operation parameters.

[0046] [Data regarding rolling direction] Next, data relating to the rolling direction of the rolled material S used as input for the wedge prediction model M will be described.

[0047] The data regarding the rolling direction of the rolled material S is data that specifies the rolling direction for each rolling pass of the rolled material S in the rolling mill 4. In other words, it is data that identifies the forward pass and the reverse pass by the rolling mill 4. For example, a numerical value, a symbol, or the like that can identify a forward pass and a reverse pass may be used, such as "1" for a forward pass and "2" for a reverse pass. In a rolling process in which the rolled material S is hot rolled through multiple passes using the rolling mill 4, the control computer 12 sets a pass schedule for the rolling passes, thereby specifying the rolling directions for the first pass to the Nth pass (where N is an integer of 2 or more), which are the multiple rolling passes performed by the rolling mill 4. This specifies the data regarding the rolling direction of the rolled material S. Furthermore, the control controller 11 that controls each device that constitutes the rolling mill 4 controls the operating conditions of each rolling pass according to the data regarding the rolling direction of the rolled material S.

[0048] The reason for using data related to the rolling direction of the rolled material S as input for the wedge prediction model M will be explained. Figure 7 is a diagram showing the contact state between the upper work roll 41a and the upper backup roll 42a of the rolling mill 4. In rolling mills, the work rolls and backup rolls are generally arranged so that the axis of the work roll is offset horizontally from the axis of the backup roll. The position of the bearing box (backup roll chock) that supports the backup roll can be stabilized by pressing it to one side of the housing. Figure 7(a) shows an example in which the work roll 41a is arranged offset toward the front of the rolling mill 4 from the backup roll 42a.

[0049] At this time, when the reaction force P received by the work roll 41a during rolling is transmitted to the backup roll 42a, its horizontal component acts in a direction that moves the backup roll 42a toward its rear surface, regardless of whether it is a forward or reverse pass. Meanwhile, as the work roll 41a rotates, the direction of the frictional force that rotates the backup roll 42a reverses between a forward pass and a reverse pass. In the example shown in Figure 7(a), during a forward pass, the backup roll 42a is subjected to a frictional force Fn acting from the work roll 41a toward its front surface. Meanwhile, during a reverse pass, the backup roll 42a is subjected to a frictional force Fr acting from the work roll 41a toward its rear surface.

[0050] Figure 7(b) shows an enlarged view of the contact state between the work roll 41a and the backup roll 42a during a forward pass and a reverse pass. As shown in Figure 7(b), during a forward pass, the backup roll 42a is subjected to a frictional force Fn acting on its front side from the work roll 41a, so that the contact point Cn between the backup roll 42a and the work roll 41a is located above the work roll 41a. On the other hand, during a reverse pass, the backup roll 42a is subjected to a frictional force Fr acting on its rear side from the work roll 41a, so that the contact point Cr between the backup roll 42a and the work roll 41a is located below the contact point Cn of the work roll 41a. As a result, the inclination of the tangent line Tn between the backup roll 42a and the work roll 41a during a forward pass is gentler than the inclination of the tangent line Tr between the backup roll 42a and the work roll 41a during a reverse pass.

[0051] Therefore, when leveling is performed to provide a difference in the vertical positions of the backup roll chocks 43a1 and 43b1, even if the same leveling amount is set for the drive side and the work side, the inclination of the tangent to the backup roll 42a and the work roll 41a differs between the forward and reverse passes, resulting in different reduction amounts for the drive side and the work side of the work roll 41a between the forward and reverse passes. In other words, even if a constant leveling amount is set, the difference in the opening between the drive side and the work side of the work roll differs between the forward and reverse passes, resulting in differences in the delivery wedge of the rolled material S. Furthermore, if the backup roll 42a and the work roll 41a are perfectly aligned, the contact points Cn and Cr between the backup roll 42a and the work roll 41a should be maintained at a constant position axially. However, in reality, misalignment between the backup roll 42a and the work roll 41a occurs, so the positions of the contact points Cn and Cr usually change axially. As a result, the directions of the tangents Tn and Tr between the backup roll 42a and the work roll 41a also vary in the axial direction. As a result, even if the leveling amount is set by the position of the backup roll chock, the difference in the opening between the drive side and the work side of the work roll differs between the forward pass and the reverse pass, resulting in a difference in the delivery wedge of the rolled material S.

[0052] As described above, the contact state between the work roll and the backup roll changes between the forward pass and the reverse pass, affecting the delivery wedge of the rolled material S. For this reason, in this embodiment, data related to the rolling direction of the rolled material is included as an input to the wedge prediction model M.

[0053] Unlike the above-mentioned rolling operation parameters which specify the rolling conditions for each rolling pass of the rolled material S, the data relating to the rolling direction of the rolled material S can be said to be parameters relating to the structural characteristics of the rolling mill 4, including the positional relationship between the work rolls and backup rolls which configure the rolling mill 4, and the mechanical backlash between the backup roll chocks and the housing. In other words, this embodiment is characterized in that, unlike the prior art, the delivery wedge is predicted using only parameters which specify the rolling conditions for each rolling pass of the rolled material S, but the delivery wedge is predicted using information relating to the structural characteristics of the rolling mill 4 in addition to the operational parameters relating to the rolling conditions for each rolling pass of the rolled material S.

[0054] [Rolled material wedge prediction device] Next, the configuration of a rolled material wedge prediction device according to one embodiment of the present invention will be described with reference to FIGS.

[0055] Fig. 8 is a block diagram showing the configuration of a wedge prediction device for rolled material according to one embodiment of the present invention. Fig. 9 is a block diagram showing the detailed configuration of a data acquisition unit 51 and a wedge prediction unit 52 shown in Fig. 8. As shown in Figs. 8 and 9, a wedge prediction device 50 for rolled material according to one embodiment of the present invention includes a data acquisition unit 51 and a wedge prediction unit 52.

[0056] Before a pass to be predicted is executed, the data acquisition unit 51 acquires setting data for the rolling operation parameters for the pass to be predicted and setting data related to the rolling direction of the rolled material S for the pass to be predicted from the control computer 12. The pass to be predicted is a rolling pass arbitrarily selected from a plurality of rolling passes performed using the rolling mill 4. The pass to be predicted may be any rolling pass selected from the plurality of rolling passes performed by the rolling mill 4, from the first pass to the Nth pass (where N is an integer of 2 or more), which is the final pass. The pass to be predicted may be set by the control computer 12. Furthermore, the pass to be predicted may be specified in advance by the wedge prediction device 50 for the rolled material, or may be set by an operator who manages the operation of the rolling mill 4.

[0057] The data acquisition unit 51 acquires data related to the inlet wedge in the pass to be predicted before the pass to be predicted is executed. The control computer 12 sets rolling operation data for the pass to be predicted and specifies setting data related to the rolling direction of the rolled material S in the pass to be predicted at the latest before the pass to be predicted is executed. Therefore, the data acquisition unit 51 can acquire setting data for the rolling operation parameters for the pass to be predicted and setting data related to the rolling direction of the rolled material S in the pass to be predicted from the control computer 12 before the pass to be predicted is executed.

[0058] The data acquiring unit 51 can measure the entry wedge in the pass to be predicted of the rolled material S and acquire the actual measured value as the data regarding the entry wedge in the pass to be predicted. If the pass to be predicted is a rolling pass other than the first pass, the actual measured value of the exit wedge in the rolling pass Pi-1 immediately preceding the pass to be predicted Pi may be used as the data regarding the entry wedge in the pass to be predicted. Furthermore, as the data regarding the entry wedge in the pass to be predicted, it is also possible to use the predicted value of the exit wedge in the rolling pass Pi-1 predicted using the wedge prediction model M with the rolling pass Pi-1 immediately preceding the pass to be predicted Pi as the prediction target pass. In either case, the data regarding the entry wedge in the pass to be predicted can use the actual measured value or predicted value regarding the entry wedge of the rolled material S immediately before rolling is performed in the pass to be predicted. The data acquired by the data acquisition unit 51 is the same as the data acquired by the data acquisition unit 31, so the data acquisition unit 31 may be configured as the same device as the data acquisition unit 51 for predicting the exit wedge.

[0059] The wedge prediction unit 52 predicts the exit wedge of the rolled material S in the pass to be predicted using the wedge prediction model M generated by the wedge prediction model generation device 30. The wedge prediction unit 52 can acquire the exit wedge in the pass to be predicted as an output by inputting input data acquired by the data acquisition unit 51 into the wedge prediction model M. The input data acquired by the data acquisition unit 51 includes the entry wedge in the pass to be predicted, at least one of the rolling operation parameters for the pass to be predicted, and data related to the rolling direction of the rolled material S in the pass to be predicted.

[0060] The prediction of the delivery wedge of the rolled material S by the wedge prediction unit 52 can be performed for each pass to be predicted. Before the pass to be predicted is executed, the control computer 12 sets the rolling operation data for the pass to be predicted and data related to the rolling direction of the rolled material S. Therefore, by setting each rolling pass as a pass to be predicted according to the progress of the rolling pass, the delivery wedge of the rolled material S for all rolling passes can be predicted before each rolling pass is executed. Furthermore, two or more consecutive rolling passes may be set as prediction target passes from the first pass to the Nth pass, which are the multiple rolling passes performed by the rolling mill 4, to predict the delivery wedge of the consecutive rolling passes.

[0061] The prediction result of the delivery wedge in the prediction target pass by the wedge prediction unit 52 may be displayed on a monitor screen provided in the operator's cab of the rolling mill 4. The operator of the rolling mill 4 can appropriately change the operating conditions of the rolling mill 4 to appropriate ones in accordance with the prediction result of the delivery wedge displayed on the monitor screen.

[0062] [Rolling material control device, wedge control method, and manufacturing method] Next, a rolled material wedge control device, wedge control method, and manufacturing method according to one embodiment of the present invention will be described with reference to FIGS.

[0063] FIG. 10 is a block diagram showing the configuration of a rolled material wedge control device according to one embodiment of the present invention. As shown in FIG. 10, a rolled material wedge control device 60 according to one embodiment of the present invention includes a data acquisition unit 51 and a wedge prediction unit 52 for predicting the exit wedge in a target pass using information acquired from a control computer 12. The rolled material wedge control device 60 also includes a manipulated variable calculation unit 61. When the exit wedge in a target pass predicted by the wedge prediction unit 52 exceeds a predetermined threshold, the manipulated variable calculation unit 61 calculates the manipulated variable of the rolling mill 4 in the target pass so that the exit wedge in the target pass is equal to or smaller than the threshold. The manipulated variable of the rolling mill 4 in the target pass calculated by the manipulated variable calculation unit 61 is output to the control computer 12, which sets the manipulated variable of the rolling mill 4 in the target pass. In this case, the manipulated variable of the rolling mill 4 in the target pass calculated by the manipulated variable calculation unit 61 may be output to the control controller 11, which then sets the manipulated variable.

[0064] 11 is a flowchart showing the flow of a method for controlling the wedge of a rolled material according to one embodiment of the present invention. The method for controlling the wedge of a rolled material according to one embodiment of the present invention is started after the control computer 12 identifies a pass to be predicted and the rolling pass immediately preceding the pass to be predicted has been completed, but before the pass to be predicted is executed. At this time, an upper limit value for the absolute value of the wedge of the rolled material S in the pass to be predicted (size of the wedge) is set in advance as a threshold value in the control computer 12. The threshold value for the wedge of the rolled material S is set, for example, to 0 to 400 μm.

[0065] As shown in Fig. 11, in a method for controlling a wedge of a rolled material according to one embodiment of the present invention, first, before the start of a pass to be predicted, a wedge prediction unit 52 predicts the delivery wedge of the rolled material S in the pass to be predicted using a wedge prediction model M (prediction step, S1). Next, a wedge control device 60 for the rolled material determines whether the magnitude of the predicted delivery wedge (prediction value) is equal to or less than an upper limit value (threshold value) (step S2). If the result of the determination is that the prediction value is equal to or less than the upper limit value (step S2: Yes), the wedge control device 60 for the rolled material determines to maintain the set values of the operating conditions for the current pass to be predicted (step S3).

[0066] On the other hand, if the predicted value exceeds the upper limit value (step S2: No), the operation amount calculation unit 61 calculates the operation amount of the rolling mill 4 as follows (step S4). Then, the set value of the operation condition determined in the processing of step S3 or the operation amount of the rolling mill 4 calculated by the operation amount calculation unit 61 is sent to the control computer 12, and the control computer 12 sets the operation condition for the pass to be predicted (step S5). Note that the control controller 11 may set the operation condition instead of the control computer 12.

[0067] The operation amount of the rolling mill 4 calculated in the processing of step S4 is preferably the set value of the leveling amount for the pass to be predicted. This is because the leveling amount has a large effect on the delivery wedge of the rolled material S in the pass to be predicted. The operation amount of the rolling mill 4 calculated in the processing of step S4 should be such that only the set value of the leveling amount is changed from the set value of the operating conditions for the current pass to be predicted, while the set values of the other operating conditions are maintained.

[0068] In the processing of step S4, first, the operation variable calculation unit 61 identifies whether the predicted value of the delivery wedge of the rolled material S is positive or negative, and determines whether the thickness of the rolled material S is thicker at the end of the work side or at the end of the drive side. If the delivery wedge of the rolled material S is predicted to be thicker on the work side than on the drive side, the operation variable calculation unit 61 calculates a new leveling operation variable so that the roll gap on the work side is reduced or the roll gap on the drive side is increased relative to the leveling set value for the current prediction target pass. On the other hand, if the delivery wedge of the rolled material S is predicted to be thicker on the drive side than on the work side, the operation variable calculation unit 61 calculates a new leveling operation variable so that the roll gap on the drive side is reduced or the roll gap on the work side is increased relative to the leveling set value for the current prediction target pass. When the input of the wedge prediction model M includes a leveling amount, a new leveling setting value calculated by the operation amount calculation unit 61 may be input to the wedge prediction model M, and the above steps may be repeated until the predicted value of the exit wedge becomes equal to or less than the threshold value.

[0069] On the other hand, in the above-described method for controlling the wedge of the rolled material, the wedge of the rolled material S in the pass to be predicted can be made to approach zero by making the threshold value of the wedge of the rolled material S in the pass to be predicted approach zero. For example, when the input of the wedge prediction model M includes a leveling amount, the operation amount calculation unit 61 sets a plurality of candidate values as setting values of the leveling amount, inputs each candidate value to the wedge prediction model M, selects a candidate value that satisfies the upper limit value from the predicted values obtained as the output, and calculates this as the operation amount of the rolling mill 4.

[0070] A method for producing rolled material according to one embodiment of the present invention uses the above-described method for controlling the wedge of a rolled material to produce the rolled material by multiple-pass reverse rolling. This makes it possible to produce a rolled material S with a small deviation in thickness in the width direction. In this case, two or more prediction target passes are selected from the multiple rolling passes, and the threshold value of the exit wedge in each prediction target pass may be set so as to gradually decrease as the rolling passes progress. Alternatively, the threshold value of the exit wedge in each prediction target pass may be set so as to decrease the exit wedge ratio (the value obtained by dividing the exit wedge by the plate thickness) as the rolling passes progress. By decreasing the threshold value of the exit wedge in the latter rolling passes of the rolling mill, it is possible to suppress wedging of the rolled material S at the stage when the rolling process is completed.

[0071] Fig. 12 is a block diagram showing a detailed configuration of the wedge control device 60 for rolled material shown in Fig. 10. As shown in Fig. 12, the wedge control device 60 for rolled material, which is one embodiment of the present invention, includes a memory unit 62, a data acquisition unit 51, a wedge prediction unit 52, an operation amount calculation unit 61, and an output unit 63.

[0072] The memory unit 62 stores the wedge prediction model M generated by the rolled material wedge prediction model generation device 30. The memory unit 62 also stores programs and data related to wedge control. The memory unit 62 may store an upper limit value related to the size of the delivery wedge corresponding to the pass to be predicted as a threshold. The memory unit 62 may store various information obtained by wedge control. The memory unit 62 may include any memory device such as a semiconductor memory device, an optical memory device, or a magnetic memory device. The semiconductor memory device may include, for example, a semiconductor memory. The memory unit 62 may include multiple types of memory devices.

[0073] The data acquiring unit 51 acquires data relating to the entry wedge in the pass to be predicted, the set values of the rolling operation parameters in the pass to be predicted, and the rolling direction of the rolled material S in the pass to be predicted before the pass to be predicted is started. The data acquiring unit 51 may acquire a threshold value of the exit wedge corresponding to the pass to be predicted.

[0074] The wedge prediction unit 52 uses the input data and the wedge prediction model M to predict the delivery wedge of the rolled material S in the prediction target pass.

[0075] When the exit wedge in the prediction target pass predicted by the wedge prediction unit 52 exceeds a threshold value, the operation amount calculation unit 61 calculates the operation amount of the rolling mill 4 in the prediction target pass so that the exit wedge is equal to or smaller than the threshold value.

[0076] The output unit 63 outputs the operation amount of the rolling mill 4 calculated by the operation amount calculation unit 61 to the control computer 12 or the control controller 11. The output unit 63 may present the operation amount of the rolling mill 4 calculated by the operation amount calculation unit 61 on the display unit (terminal device) 70 as a guidance operation amount for the pass to be predicted. When the operation amount of the rolling mill 4 calculated by the operation amount calculation unit 61 is a leveling amount, the leveling amount for the pass to be predicted is displayed as the guidance operation amount.

[0077] When the operation amount of the rolling mill 4 is output from the output unit 63 to the control computer 12, the operation amount of the rolling mill 4 may be automatically updated by the control computer 12. In addition, the operator may change the rolling operation conditions in the pass to be predicted based on the guidance operation amount shown on the display unit 70.

[0078] The rolling material wedge control device 60 can be realized by, for example, a computer as described above. The computer includes, for example, a memory, a hard disk drive (storage device), a CPU (processing device), and the like. The program can be stored in the hard disk drive, and when executed by the CPU, it is read from the hard disk drive to the memory. Data during processing is stored in the memory, and if necessary, stored in the HDD. The storage unit 62 can be realized by, for example, a storage device. The data acquisition unit 51, the wedge prediction unit 52, the operation amount calculation unit 61, and the output unit 43 can be realized by, for example, a CPU that reads and executes the program. [Example]

[0079] As an example, an example of controlling the wedge of a rolled material using a plate rolling mill that performs hot rolling of the rolled material using a plurality of rolling passes will be described. In the example, first, the rolling mill performed 3 to 50 rolling passes of reverse rolling, and hot rolling of 20,000 steel plates having a thickness of 8 to 100 mm and a width of 1,500 to 4,500 mm after the final pass was performed, and a wedge prediction model generation device 30 for rolled material generated a wedge prediction model M. The machine learning algorithm used to generate the wedge prediction model M was a neural network, with three intermediate layers and 100 nodes each. A ramp function was used as the activation function.

[0080] Next, the generated wedge prediction model M was stored in the wedge prediction unit 52, and the method for controlling the wedge of a rolled material according to the present invention was applied when hot rolling 100 new steel plates. In this case, for the wedge control of the rolled material, the final rolling pass of each steel plate was set as the prediction target pass. In addition, the upper limit value (threshold value) of the delivery wedge of the rolled material in the prediction target pass was set to 80 μm. In contrast, as a comparative example of the present invention, the method for controlling the wedge of a rolled material according to the present invention was not applied, and hot rolling was performed on 100 steel plates without changing the set values of the rolling operation conditions set by the control computer 12.

[0081] In Example 1 of the invention, the wedge prediction model M was generated by selecting, as inputs, the entry wedge in the pass to be predicted and data on the rolling direction of the rolled material in the pass to be predicted, as well as the leveling amount, rolling load, entry plate thickness, plate width, and temperature of the rolled material as rolling operation parameters in the pass to be predicted.

[0082] In Example 2 of the invention, the rolled material wedge prediction model generation device 30 divided the data set stored in the database 101 for each combination of the control number of the backup roll and the control number of the work roll installed in the rolling mill when performing reverse rolling, and generated a wedge prediction model M for each classification of the control number of the backup roll and the control number of the work roll. Then, the wedge prediction unit 52 referenced the control number of the backup roll and the control number of the work roll installed in the rolling mill in the pass to be predicted, and performed wedge control of the rolled material using the corresponding wedge prediction model M.

[0083] The results of wedge control are shown in Table 1. In the comparative example, the average wedge value for 100 rolled steel sheets was 22 μm, and the standard deviation of the wedge was 81 μm. In contrast, in invention example 1, the average wedge value for 100 rolled steel sheets was 14 μm, and the standard deviation of the wedge was 60 μm, indicating a reduction in wedge and a reduction in its variation. In addition, in invention example 2, the average wedge value for 100 rolled steel sheets was 10 μm, and the standard deviation of the wedge was 49 μm, indicating a further reduction in wedge and a reduction in its variation. This confirms that the wedge control method according to the present invention can reduce wedge in rolled material.

[0084] [Table 1]

[0085] Although the present invention has been described above as an embodiment, the present invention is not limited to the descriptions and drawings that form part of the disclosure of the present invention. In other words, other embodiments, examples, and operational techniques that can be made by those skilled in the art based on the present invention are all included in the scope of the present invention. [Explanation of symbols]

[0086] 1 Plate rolling line 2 Furnace 3 Descaling device 4. Rolling mill 5 Cooling device 6 straightening machine 7 Cooling bed 8 Front transport roll 9 Rear transport roll 11 Control Controller 12 Control computer 13 Upper computer 21 Front Wedge Gauge 22 Rear wedge gauge 30. Rolled material wedge prediction model generator 31 Data Acquisition Section 32 Model Generation Unit 32a Database Section 32b Machine Learning Department 41a, 41b Work rolls 42a, 42b Backup roll 43a1, 43a2, 43b1, 43b2 Backup roll chocks 44a, 44b Housing 45a, 45b Load cell 46a, 46b Screw-down device 50 Wedge prediction device for rolled material 51 Data Acquisition Section 52 Wedge Predictor 60 Wedge control device for rolling material 61 Operation amount calculation section 62 Storage section 63 Output section 70 Display unit (terminal device) PL Pass Line

Claims

1. A method for predicting a wedge of a rolled material at an outlet side of a rolling mill in a prediction target pass, which is a rolling pass selected from the plurality of rolling passes, in a rolling process in which a rolled material is reverse rolled through a plurality of rolling passes using the rolling mill, A method for predicting the wedge of a rolled material at the exit side of the rolling mill in the pass to be predicted, comprising a step of predicting the wedge of the rolled material at the exit side of the rolling mill in the pass to be predicted using a wedge prediction model learned by machine learning, the wedge prediction model including as input the wedge of the rolled material at the entry side of the rolling mill in the pass to be predicted, at least one of rolling operation parameters, and data related to the rolling direction of the rolled material, and outputting the wedge of the rolled material at the exit side of the rolling mill in the pass to be predicted.

2. The method for predicting a wedge of a rolled material according to claim 1 , wherein the rolling operation parameters include a leveling amount of the rolling mill.

3. 3. A method for controlling the wedge of a rolled material, comprising: a step of calculating an operation amount for the rolling mill in the pass to be predicted, based on the wedge of the rolled material at the delivery side of the rolling mill predicted using the wedge prediction method for rolled material according to claim 1 or 2, so that the size of the wedge of the rolled material at the delivery side of the rolling mill in the pass to be predicted is equal to or smaller than a predetermined threshold value.

4. A method for producing a rolled material, comprising the step of producing the rolled material using the method for controlling the wedge of the rolled material according to claim 3.

5. A method for generating a wedge prediction model of a rolled material, which predicts a wedge of the rolled material at an outlet side of a rolling mill in a prediction target pass that is a rolling pass selected from the plurality of rolling passes in a rolling process in which a rolled material is reverse rolled through a plurality of rolling passes using the rolling mill, a step of acquiring a plurality of data sets, each set including a wedge of the rolled material at the entry side of the rolling mill, at least one of the rolling operation parameters, data related to the rolling direction of the rolled material, and a wedge of the rolled material at the exit side of the rolling mill, and generating a wedge prediction model that uses input data including the wedge of the rolled material at the entry side of the rolling mill, at least one of the rolling operation parameters, and data related to the rolling direction of the rolled material as input and outputs the wedge of the rolled material at the exit side of the rolling mill through machine learning using the acquired plurality of data sets as training data.

6. 1. A rolling material wedge control device for controlling a wedge of a rolled material on an outlet side of a rolling mill in a prediction target pass, which is a rolling pass selected from the plurality of rolling passes, in a rolling process in which a rolled material is reverse rolled through a plurality of rolling passes using a rolling mill, a wedge prediction unit that predicts the wedge of the rolled material at the exit side of the rolling mill in the pass to be predicted using a wedge prediction model learned by machine learning, the wedge prediction model including as input the wedge of the rolled material at the entry side of the rolling mill in the pass to be predicted, at least one of rolling operation parameters, and data related to the rolling direction of the rolled material, and outputs the wedge of the rolled material at the exit side of the rolling mill in the pass to be predicted; an operation amount calculation unit that calculates an operation amount of the rolling mill in the prediction target pass based on the wedge of the rolled material at the delivery side of the rolling mill predicted by the wedge prediction unit so that the size of the wedge of the rolled material at the delivery side of the rolling mill in the prediction target pass is equal to or smaller than a preset threshold value; A wedge control device for rolling material comprising:

Images