Method for predicting camber of rolled material, method for manufacturing hot-rolled steel sheet, method for generating camber prediction model of rolled material, and camber prediction device for rolled material

Abstract

To provide a method or the like for predicting camber of a rolled material that predicts camber formed in the rolled material during rough rolling.SOLUTION: There is provided a method for predicting camber of a rolled material that predicts camber of the rolled material when rough rolling is performed on the rolled material that has undergone heating and width reduction. The method for predicting camber of the rolled material comprises: a temperature acquisition step of measuring surface temperatures of the rolled material in a width direction and a longitudinal direction before the rolled material is subjected to the heating and the width reduction; a position information acquisition step of acquiring position and shape of the rolled material on the basis of the surface temperature obtained in the temperature acquisition step; an image data generation step of generating image data associated with the position, the shape, and the surface temperature of the rolled material; and a prediction step of predicting camber of the rolled material that has undergone the rough rolling step, using a camber prediction model learned by machine learning with the image data as input data and the camber of the rolled material as output data.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method for predicting the camber of a rolled material when the material has been heated and width-reduced and then rough-rolled, a method for manufacturing hot-rolled steel plate, a method for generating a camber prediction model for a rolled material, and a camber prediction device for a rolled material. [Background technology]

[0002] An example of a metal sheet manufacturing line is a hot rolling line, which includes a heating furnace for heating the rolled material, a descaling device for descaling the rolled material, a width reduction press for width reduction of the descaled rolled material, a roughing mill for rough rolling the width-reduced rolled material, and a finishing mill for finish rolling of the rolled material.

[0003] In hot rolling lines, a curve called camber may occur in the material being rolled. If camber occurs in the material being rolled, the material may come into contact with equipment such as side guides of the roughing mill or finishing mill, causing problems.

[0004] In particular, as the camber of the material to be rolled increases, the possibility of poor engagement with the rolling rolls of the roughing mill or the finishing mill and damage to associated equipment increases. Thus, the camber of the material to be rolled is problematic in that it reduces the product yield and the availability of the hot rolling line.

[0005] Therefore, efforts are being made to reduce the camber of the material to be rolled. For example, Patent Document 1 discloses that when there is a temperature deviation in the width direction of the material to be rolled, a pair of dies are moved relative to each other in the direction of the conveying line, and the material is reduced in width while applying a bending force to the material to be rolled.

[0006] Furthermore, Patent Document 2 discloses that the angle of incidence of the rolled material relative to a width reduction press device is changed based on information about the rolled material obtained at least either before or after width reduction.

[0007] Patent Document 3 discloses that the leveling amount in a roughing mill is calculated based on the sum of the predicted load on the work side and the predicted load on the drive side of the roughing mill, the difference between the predicted load on the work side and the predicted load on the drive side, and the temperature distribution in the width direction of the material to be rolled. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] Japanese Patent Application Publication No. 3-254301 [Patent Document 2] Patent No. 6103158 [Patent Document 3] Japanese Patent Application Publication No. 2019-123004 Summary of the Invention [Problem to be solved by the invention]

[0009] However, with the technology disclosed in Patent Document 1, there is a risk that the thickness shape near both ends in the width direction of the cross section of the rolled material, the so-called dog-bone shape, may become uneven between the work side and the drive side, which may result in camber when the rolled material after width reduction is rolled by a roughing mill.

[0010] In Patent Document 2, the information on the rolled material includes the temperature distribution in the width direction of the rolled material before width reduction, and the angle of incidence of the rolled material is changed depending on the temperature distribution. However, the temperature distribution in the width direction of the rolled material itself is not eliminated even when width reduction is performed using a width reduction press device, so there is a problem that camber occurs when the rolled material after width reduction is rolled using a roughing mill.

[0011] In Patent Document 3, not only does temperature deviation occur in the width direction of the rolled material after width reduction, but there is also a risk of the dogbone shape being formed unevenly between the work side and the drive side, which can result in the occurrence of camber when the rolled material after width reduction is rolled by a roughing mill.

[0012] However, when a width-reduced rolled material is subjected to a temperature distribution not only in the width direction but also in the longitudinal direction. When a temperature distribution occurs in the longitudinal direction, the dog-bone shape of the rolled material may change in the longitudinal direction. Therefore, there is room for improvement in that the camber that occurs at the leading and trailing ends of the rolled material may not be sufficiently reduced.

[0013] As described above, the relationship between the temperature deviation in the width direction of the rolled material and the camber of the rolled material is affected by various factors, which makes it difficult to predict the camber of the rolled material in advance, and therefore makes it difficult to reduce the camber of the rolled material.

[0014] The present invention has been made in consideration of the above-mentioned problems, and has an object to provide a method for predicting the camber of a rolled material that can be formed in the rolled material by rough rolling. [Means for solving the problem]

[0015] In order to solve the above problems, the present invention has the following features.

[0016] [1] A method for predicting the camber of a rolled material when rough rolling is performed on a rolled material that has been heated and width-reduced, comprising: a temperature acquisition step of acquiring surface temperatures in the width direction and the longitudinal direction of the rolled material after the heating and before the width reduction; a position information acquisition step of acquiring the position and shape of the rolled material based on the surface temperature acquired in the temperature acquisition step; an image data generating step of generating image data in which the position and shape of the rolled material and the surface temperature are associated with each other; a prediction step of predicting the camber of the rolled material when the rough rolling is performed using a camber prediction model trained by machine learning, with the image data as input data and the camber as output data. [2] The width reduction is performed using a plurality of operational parameters that can affect the deformation of the rolled material, The method for predicting the camber of a rolled material described in [1], wherein the camber prediction model is learned by machine learning using at least one of the plurality of operational parameters of the width reduction and the image data as the input data. [3] The method for predicting camber of a rolled material according to [1] or [2], wherein the temperature acquisition step is performed using a scanning radiation thermometer. [4] A method for manufacturing a hot-rolled steel sheet using the camber prediction method for a rolled material according to any one of [1] to [3], a reference data acquisition step of acquiring reference data of the camber, which is a criterion for determining whether or not the rough rolling is to be performed; A rough rolling process of performing the rough rolling in accordance with the camber predicted in the prediction process and the reference data. [5] A rough rolling suitability information generating step is provided for generating rough rolling suitability information that determines whether or not the rolling target material is suitable for rough rolling based on the camber predicted in the prediction step and the reference data, [4] The method for manufacturing a hot-rolled steel sheet according to [4], wherein in the rough rolling step, the rough rolling is performed by excluding specific rolled materials from the target of the rough rolling based on the rough rolling feasibility information. [6] A method for generating a camber prediction model for a rolled material, the method comprising: generating a camber prediction model for predicting the camber of a rolled material when rough rolling is performed on the rolled material that has been heated and width-reduced; a temperature acquisition step of acquiring surface temperatures in the width direction and the longitudinal direction of the rolled material after the heating and before the width reduction; a position information acquisition step of acquiring the position and shape of the rolled material based on the surface temperature acquired in the temperature acquisition step; an image data generating step of generating image data in which the position and shape of the rolled material and the surface temperature are associated with each other; A method for generating a camber prediction model for a rolled material, comprising a camber prediction model generation step of generating a camber prediction model that predicts the camber of the rolled material that has been rough rolled by machine learning using a plurality of learning data, with the image data as input data and the camber as output data. [7] A camber prediction device for a rolled material that predicts the camber of the rolled material when rough rolling is performed on the rolled material that has been heated and width-reduced, a temperature acquisition unit that acquires the surface temperatures of the rolled material in the width direction and the longitudinal direction after the heating and before the width reduction; a position information acquisition unit that acquires the position and shape of the rolled material based on the surface temperature acquired by the temperature acquisition unit; an image data generating unit that generates image data in which the position and shape of the rolled material and the surface temperature are associated with each other; A camber prediction device for a rolled material, comprising: a prediction unit that predicts the camber of the rolled material after the rough rolling using a camber prediction model learned by machine learning, with the image data as input data and the camber as output data. [Effects of the Invention]

[0017] The method for predicting the camber of a rolled material of the present invention includes an image data generation step of generating image data that associates the position, shape, and surface temperature of the rolled material. The method also includes a prediction step of predicting the camber of the rough-rolled material using a camber prediction model trained by machine learning, with the image data as input data and the camber of the rolled material as output data. Therefore, the camber of the rolled material can be predicted based on the overall temperature distribution of the rolled material. That is, the camber of the rolled material can be predicted taking into account the temperature distribution in the width direction and the longitudinal direction (i.e., the conveying direction) of the rolled material. Therefore, the camber can be predicted with higher accuracy than conventional methods. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is an explanatory diagram showing an overview of a hot rolling line and a camber prediction device for a rolled material. FIG. [Figure 2] FIG. 2 is a functional block diagram of a camber prediction device for a rolled material. [Figure 3] FIG. 10 is an explanatory diagram showing a manner in which a camber prediction model is generated. [Figure 4] 1 shows a process flow of a method for manufacturing a hot-rolled steel sheet and a method for predicting the camber of a rolled material. [Figure 5] FIG. 5 is an explanatory diagram showing an aspect of the position information acquisition step in step S202 of FIG. 4. [Figure 6] 5 is an example of image data generated in the image data generating step in step S203 of FIG. 4. [Figure 7] FIG. 2 is an explanatory diagram schematically showing the camber of a rolled material. [Figure 8] FIG. 2 is an explanatory diagram showing a cross section of the camber of a rolled material. [Figure 9] FIG. 10 is an explanatory diagram showing an aspect in which a camber prediction model according to the second embodiment is generated. [Figure 10] FIG. 2 is an explanatory diagram showing the configuration of a width reduction press device. DETAILED DESCRIPTION OF THE INVENTION

[0019] (First embodiment) DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will now be described with reference to the accompanying drawings. Fig. 1 shows a hot rolling line 10 and a rolled material camber prediction device 30 for predicting the camber of a rolled material.

[0020] The hot rolling line 10 includes a heating furnace 11, a descaling unit 12, a width reduction press unit 13, a roughing mill 14, a finishing mill 15, a cooling unit 16, and a coiler 17. The heating furnace 11, the descaling unit 12, the width reduction press unit 13, the roughing mill 14, the finishing mill 15, the cooling unit 16, and the coiler 17 are arranged in this order in the conveying direction of the material to be rolled. The hot rolling line 10 includes a hot rolling line control unit 18 that controls the operation of the heating furnace 11, the descaling unit 12, the width reduction press unit 13, the roughing mill 14, the finishing mill 15, the cooling unit 16, and the coiler 17.

[0021] The descaling device 12 is a device for descaling the material to be rolled by injecting high-pressure water onto the material to be rolled that has been heated in the heating furnace 11. The high-pressure water injected from the descaling device 12 has a volume of 5.0 to 10.0 m 3 / min and the pump pressure should be 10 to 50 MPa.

[0022] The width reduction press device 13 has a pair of width reduction dies (not shown) arranged on one side and the other side of the width direction of the material to be rolled. The pair of width reduction dies are arranged opposite each other. The pair of width reduction dies press down the material to be rolled by sandwiching it between them.

[0023] The width reduction press device 13 has a pair of pinch rolls (not shown) that transport the material to be rolled. The width reduction press device 13 transports the material to be rolled using the pair of pinch rolls, while pressing down the material from both sides in the width direction using width reduction dies. The width reduction press device 13 can change the feed pitch, which is the transport distance of the material to be rolled per press pass, by changing the drive amount of the pinch rolls.

[0024] The roughing mill 14 is composed of, for example, a reversing mill 14a capable of reverse rolling and a non-reversing mill 14b capable of rolling only in the conveying direction downstream. The reversing mill 14a typically performs about 5 to 11 rolling passes to reduce the plate thickness. The reduction passes by the reversing mill 14a are performed from the upstream side to the downstream side or from the downstream side to the upstream side. In the final rolling pass, rolling and conveyance to the next rolling mill are performed simultaneously. For this reason, the number of rolling passes by the reversing mill 14a is typically an odd number.

[0025] The finishing mill 15 is not particularly limited, but for example, a continuous rolling mill having 5 to 7 stands can be used.

[0026] The cooling device 16 is provided in a facility called a run-out table. The run-out table is provided downstream of the finishing rolling mill 15. The cooling device 16 cools the rolled material by spraying cooling water.

[0027] The coiler 17 is a winding machine that winds up the material to be rolled that has been cooled by the cooling device 16. The material to be rolled that has been wound up by the coiler 17 is transported to a predetermined location as a hot-rolled steel sheet.

[0028] The hot rolling line control unit 18 is a computer including a CPU. The hot rolling line control unit 18 has a storage unit 19 in which data for controlling the operation of the hot rolling line 10 and the like are stored.

[0029] The storage unit 19 is a writable non-volatile memory such as an EPROM. There are no particular limitations on the storage unit 19, and for example, a storage device such as an HDD or SSD can be used. The storage unit 19 stores the time when the rolled material arrives at each part of the hot rolling line 10, reference data for the camber that serves as a criterion for determining whether rough rolling by the roughing mill is possible, and the like.

[0030] The hot rolling line 10 is provided with a temperature measuring unit 21 for measuring the temperature of the rolled material between the descaling device 12 and the width reduction press device 13. The temperature measuring unit 21 measures the surface temperatures of the rolled material in the width direction and longitudinal direction after descaling by the descaling device 12 before width reduction by the width reduction press device 13.

[0031] When the material to be rolled is heated in the heating furnace 11, primary scale is generated. This primary scale can cause errors in measuring the surface temperature of the material to be rolled. By performing descaling, the primary scale can be removed from the material to be rolled. Therefore, errors in measuring the surface temperature of the material to be rolled can be reduced.

[0032] Furthermore, when the width of the rolled material is reduced, processing heat is generated due to plastic deformation. Therefore, by measuring the surface temperature of the rolled material before the width reduction, the influence of processing heat due to plastic deformation can be reduced, thereby improving the accuracy of temperature measurement.

[0033] The temperature measurement unit 21 is preferably capable of measuring a temperature range of 1000 to 1250° C. The temperature measurement unit 21 is not particularly limited, but may be, for example, a thermal imaging camera, a spot radiation thermometer, a scanning radiation thermometer, or the like. In this embodiment, an example will be described in which a scanning radiation thermometer is used as the temperature measurement unit 21.

[0034] A scanning radiation thermometer is a thermometer that detects infrared rays emitted from a measurement object while scanning a detection position and measures the temperature distribution of the measurement object corresponding to the detection position. As a scanning radiation thermometer, for example, a mirror-type scanning radiation thermometer that scans the measurement object area by adjusting the angle of a rotating mirror can be used.

[0035] The scanning radiation thermometer is installed so as to scan the width direction of the rolled material. The rolled material is transported at a constant speed in the transport direction in the hot rolling line 10. In other words, by matching the scanning speed in the width direction with the transport speed of the rolled material, the scanning radiation thermometer can measure the temperature of the entire top or bottom surface, which is the measurement surface of the rolled material.

[0036] The hot rolling line 10 is provided with a camber measurement unit 22 that measures the camber of the rolled material between the reversible rolling mill 14a and the non-reversible rolling mill 14b. When the camber measurement unit 22 measures the camber of the rolled material, it stores the measured camber in the memory unit 32 as actual data of the camber of the rolled material.

[0037] Here, camber refers to the distance in the width direction between the center position in the width direction at the front end of the rolled material and the center position in the width direction located a predetermined distance in the longitudinal direction from the front end. Alternatively, camber may be the distance in the width direction between the center position in the width direction at the tail end of the rolled material and the center position in the width direction located a predetermined distance in the longitudinal direction from the tail end. Hereinafter, camber will be mainly explained with reference to the front end, but it can also be measured at the tail end in the same way as the front end. Incidentally, camber is also commonly referred to as the amount of lateral bow.

[0038] The camber measurement unit 22 measures the camber of the rolled material that has been subjected to at least one pass of horizontal rolling by the roughing mill 14. The camber measurement unit 22 measures the camber of the rolled material by processing, for example, an image captured by a camera that includes the leading or trailing end of the rolled material. The image processed by the camber measurement unit 22 is an image that includes the leading or trailing end of the rolled material downstream of the roughing mill 14. The image is an image captured from the top or bottom surface of the rolled material.

[0039] In the hot rolling line 10 shown in FIG. 1, camber measuring units 22 are arranged downstream of the reversing rolling mill 14a and downstream of the non-reversing rolling mill 14b.

[0040] The camber measuring unit 22 can be installed at any position as long as it can measure the camber of the rolled material. For example, the camber measuring unit 22 may be installed at a position where it can measure the camber of the rolled material when rough rolling of the pass to be predicted is completed.

[0041] A rolled material camber prediction device 30 is connected to the hot rolling line 10, which predicts the camber when the rolled material is rolled in the roughing mill 14. The rolled material camber prediction device 30 is connected to a temperature measurement unit 21 and a camber measurement unit 22 so as to be able to communicate data with each other.

[0042] 2 is a functional block diagram of the rolled material camber prediction device 30. As shown in Fig. 2, the rolled material camber prediction device 30 has an input / output unit 31 which is an interface with external devices, a memory unit 32 which stores various data, and a control unit 33 which controls the camber prediction device 30.

[0043] The input / output unit 31 is connected to the temperature measurement unit 21, the camber measurement unit 22, the hot rolling line control unit 18, and the terminal 23. Therefore, the temperature measured by the temperature measurement unit 21 and the camber of the rolled material measured by the camber measurement unit 22 are connected so as to be input to the camber prediction device 30 for the rolled material via the input / output unit 31.

[0044] The hot rolling line control unit 18 and the terminal 23 are also connected to be able to output data via an input / output unit 31. That is, the camber measured by the camber measuring unit 22 is stored in the memory unit 32 together with the time of measurement. The temperature measured by the temperature measuring unit 21 is also stored in the memory unit 32 together with the time of measurement via the input / output unit 31.

[0045] The storage unit 32 is a writable non-volatile memory such as an EPROM. The storage unit 32 is not particularly limited, but may be, for example, a storage device such as an HDD or SSD. The storage unit 32 stores position information of the rolled material and the shape of the rolled material. The shape of the rolled material is stored together with, for example, the dimensions of the rolled material.

[0046] The control unit 33 is a computer including a central processing unit. The control unit 33 includes a temperature acquisition unit 35 that acquires the surface temperatures in the width and length directions of the rolled material after heating and before width reduction. The control unit 33 also includes a position information acquisition unit 36 ​​that acquires the position and shape of the rolled material based on the surface temperature acquired by the temperature acquisition unit 35.

[0047] The control unit 33 includes an image data generation unit 37 that generates image data that associates the position, shape, and surface temperature of the rolled material. The image data generation unit 37 stores the generated image data in the storage unit 32.

[0048] The control unit 33 includes a camber prediction model 38 that is trained by machine learning and that uses image data as input data and the camber of the rolled material as output data.

[0049] The control unit 33 includes a prediction unit 39 that uses a camber prediction model 38 to predict the camber of the material to be rolled after rough rolling.

[0050] 3 shows how the camber prediction model 38 is generated. As shown in FIG. 3, the camber prediction model 38 is generated by a camber prediction model generating unit 40.

[0051] The camber prediction model generation unit 40 only needs to be connected to the camber prediction device 30 so that data communication is possible, and may be provided in the camber prediction device 30 or in a device different from the camber prediction device 30.

[0052] The camber prediction model generation unit 40 includes a data acquisition unit 41 that acquires data from the memory unit 32 of the camber prediction device 30, a database unit 42 that stores the data acquired by the data acquisition unit 41, and a machine learning unit 43 that performs machine learning based on the data stored in the database unit 42.

[0053] The data acquisition unit 41 reads and acquires image data D1 from the storage unit 32. The data acquisition unit 41 also reads and acquires, from the storage unit 32, actual data D2 of the camber of the rolled material that corresponds to the image data D1. The image data D1 and the actual data D2 of the camber of the rolled material are associated with each other, for example, by the measurement times of the two. That is, when a specific rolled material is transported in the hot rolling line 10, the times at which the rolled material arrives at the temperature measurement unit 21 and the camber measurement unit 22 are stored in the storage unit 32, and the data acquisition unit 41 associates and acquires measurement data that is close to those times.

[0054] The data acquisition unit 41 generates a data set that associates the actual data of the image data D1 with actual data D2 of the camber of the rolled material that has been subjected to rough rolling (hereinafter also referred to as the rough rolled material). The image data D1 and the actual data D2 of the camber are used as learning data.

[0055] The camber performance data D2 acquired by the data acquisition unit 41 is acquired corresponding to the camber of the rough-rolled material output by the camber prediction model 38. For example, when the camber prediction model 38 outputs the camber after the first pass of rolling by the reversing rolling mill 14a is completed, the data acquisition unit 41 correspondingly acquires performance data of the camber after the first pass of rolling by the reversing rolling mill 14a is completed.

[0056] Furthermore, the data acquisition unit 41 may acquire a camber corresponding to the position of the camber of the rough-rolled material output by the camber prediction model 38. For example, if the camber prediction model 38 outputs the camber of the leading end of the material to be rolled, the data acquisition unit 41 may acquire actual data of the camber of the leading end of the rough-rolled material.

[0057] The database unit 42 stores data sets including the actual data of the image data D1 acquired by the data acquisition unit 41 and the actual data D2 of the camber of the rough-rolled material. The database unit 42 stores 1,000 or more data sets. Preferably, 5,000 or more data sets, and more preferably, 20,000 or more data sets. The data sets stored in the database unit 42 may be screened as necessary.

[0058] Furthermore, the number of data sets stored in the database unit 42 may be limited to a certain number, and the data sets stored in the database unit 42 may be updated as appropriate within that limit, or the data sets stored in the database unit 42 may be updated using the latest operational data obtained over a certain period of time (e.g., six months).

[0059] The machine learning unit 43 generates the camber prediction model 38 using the data set stored in the database unit 42, and executes the camber prediction model generation step.

[0060] The machine learning unit 43 may use any machine learning model that can provide sufficient prediction accuracy for the camber of the rough-rolled material for practical use. For example, the machine learning model may be a commonly used neural network (including deep learning, convolutional neural network, etc.), decision tree learning, random forest, support vector regression, etc.

[0061] Furthermore, the machine learning model may be an ensemble model that combines multiple models. In particular, when deep learning is used, the problem of multicollinearity does not need to be taken into consideration, and other operational parameters that are correlated with the temperature information of the rough-rolled material and the camber of the rough-rolled material can also be freely selected as inputs, thereby improving the prediction accuracy of the camber of the rough-rolled material.

[0062] A convolutional neural network is preferably used for the machine learning unit 43. Specifically, a convolutional neural network having a convolutional layer, a pooling layer, a fully connected layer, and an output layer, with the performance data of the image data D1 as input, is preferably used for the machine learning unit 43. By using such a machine learning unit 43, a first feature map is generated in which the information of the image data D1 is compressed into one-dimensional information while maintaining the feature amounts of the performance data of the image data D1.

[0063] In the convolution layer, a nonlinear function is preferably used as the activation function, and more preferably, the Relu function is used as the nonlinear function to suppress the gradient vanishing problem during training.

[0064] The purpose of the pooling layer is to reduce the size of the output data while preserving the features of the image data D1. The pooling layer takes the first feature map output by the convolution layer as input and compresses the information in the first feature map. Max pooling or average pooling can be applied to the pooling layer. Such a pooling layer compresses the amount of information while maintaining the features related to the performance data of the input image data, thereby generating a second feature map.

[0065] The fully connected layer transforms the second feature map generated in the pooling layer, arranging the values ​​of the second feature map in a row and consolidating the output from the pooling layer. An example of a preferred form of the fully connected layer is a connection layer with 16 to 2048 nodes. The output layer combines the information of the neurons transmitted by the fully connected layer and outputs the camber of the rough-rolled material.

[0066] The machine learning unit 43 may improve the accuracy of estimating the camber of the rough-rolled material by dividing the data set stored in the database unit 42 into training data and test data and performing learning on the data. For example, the machine learning unit 43 may use the training data to learn the weighting coefficients of the neural network, and generate the camber prediction model 38 while appropriately changing the structure of the neural network (the number of intermediate layers and the number of nodes) so as to improve the accuracy of predicting the camber of the rough-rolled material using the test data. Note that the weighting coefficients can be updated using an error propagation method.

[0067] Fig. 4 is a flow chart showing a method for manufacturing a hot-rolled steel sheet. As shown in Fig. 4, the material to be rolled is charged into a heating furnace 11, and then heated to a predetermined set temperature, thereby carrying out a heating step (step S101).

[0068] Next, the rolled material extracted from the heating furnace 11 is subjected to a descaling process by injecting high-pressure water into the descaling device 12 (step S102). By carrying out the descaling process of step S102, primary scale formed on the surface of the rolled material is removed.

[0069] When the descaling process in step S102 is performed, the surface temperature of the rolled material is measured by a scanning radiation thermometer, which is the temperature measurement unit 21. The temperature measured by the temperature measurement unit 21 is stored in the memory unit 32 together with the measurement time via the input / output unit 31.

[0070] The scanning frequency of the scanning radiation thermometer should be set according to the conveying speed of the rolled material. For example, if the conveying speed of the rolled material is 100 to 150 m / min, the scanning frequency should be set to 100 to 200 Hz. By setting the scanning frequency in this way, the surface temperature can be measured at intervals of 0.8 to 25.0 mm in the longitudinal direction of the rolled material.

[0071] Furthermore, it is preferable to use a scanning radiation thermometer with a measurement wavelength of 0.60 to 0.90 μm. By using a measurement wavelength of 0.60 μm or more, the intensity of the light reflected by the rolled material can be ensured even when the surface temperature of the rolled material is low, and measurement accuracy can be maintained.

[0072] Furthermore, the water sprayed onto the rolled material from the descaling device 12 may remain on the rolled material as a water film. In this case, the water film may cause fluctuations in light transmittance. By setting the measurement wavelength to within 0.90 μm, errors in temperature measurement can be reduced even when such fluctuations in light transmittance occur.

[0073] The surface temperature of the rolled material is preferably measured by the scanning radiation thermometer at any timing between 2 seconds and 20 seconds after the rolled material passes through the descaling device 12 .

[0074] During descaling using the descaling device 12, high-pressure water is sprayed to lower the surface temperature of the rolled material. The surface temperature of the rolled material then rises due to recuperation. The temperature of the rolled material can be measured appropriately by measuring the temperature at least two seconds after the elapsed time from descaling.

[0075] Furthermore, by measuring within 20 seconds of descaling, the temperature can be measured before the formation of secondary scale, which is the oxide scale that forms after descaling, and this prevents a decrease in the accuracy of the temperature measurement.

[0076] When the temperature is measured by the temperature measurement unit 21, the temperature acquisition unit 35 of the camber prediction device 30 reads and acquires the measured temperature from the memory unit 32 and executes a temperature acquisition step (step S201). That is, the temperature acquisition unit 35 acquires the surface temperatures in the width direction and the length direction of the rolled material after heating and before width reduction.

[0077] The position information acquisition unit 36 ​​acquires the position and shape of the rolled material based on the surface temperature acquired in the temperature acquisition step of step S101, and executes the position information acquisition step (step S202). Specifically, the position information acquisition unit 36 ​​acquires the position information of the rolled material and the shape of the rolled material based on the surface temperatures in the width direction and length direction of the rolled material.

[0078] Figure 5 shows how the positional information of the rolled material and the shape of the rolled material SS are acquired in the positional information acquisition process of step S202 in Figure 4. Figure 5 shows an example in which width direction temperature distributions T1 to T4 of the rolled material SS measured by a scanning radiation thermometer are arranged along the longitudinal direction of the rolled material SS.

[0079] The scanning radiation thermometer measures the width direction temperature distribution of the rolled material SS at a predetermined time interval according to a preset measurement frequency. Therefore, by using data related to the conveying speed of the rolled material SS, the width direction temperature distribution for each longitudinal position of the rolled material SS can be obtained from the width direction temperature distribution of the rolled material SS measured at the predetermined time interval.

[0080] That is, the width direction temperature distributions T1 to T4 correspond to positions in the longitudinal direction of the rolled material SS. In this case, a threshold value Ts for the surface temperature of the rolled material SS is set in advance, and when the surface temperature of the rolled material SS is equal to or higher than the threshold value Ts, it can be determined that the surface temperature of the rolled material SS has been detected.

[0081] On the other hand, if the surface temperature of the rolled material SS is less than the threshold value Ts, the surface temperature of the rolled material SS has not been detected, that is, the width direction temperature distribution can be considered as measurement data that is not of the rolled material SS.

[0082] In the example shown in Fig. 5, the width direction temperature distributions T2 and T3 both exceed the threshold value TS, so the surface temperature of the rolled material SS is detected. In addition, the width direction temperature distributions T1 and T4 both do not exceed the threshold value TS, so they are measurement data of temperatures other than that of the rolled material SS.

[0083] In this way, the longitudinal position information and shape of the rolled material SS can be identified. Furthermore, in the width direction temperature distributions T2 and T3, the range where the surface temperature of the rolled material SS is equal to or greater than the threshold value Ts is the detected surface temperature of the rolled material, and the range where it is less than the threshold value Ts is not the surface temperature of the rolled material SS. Therefore, by defining the boundary between the parts that exceed the threshold value Ts and the parts that do not exceed the threshold value Ts, the width direction position information and shape of the rolled material SS can be identified.

[0084] This allows the in-plane positional information of the top or bottom surface of the rolled material SS to be identified, and the shape of the rolled material, i.e., the contour shape of the top or bottom surface of the rolled material, to be obtained as the planar shape of the rolled material SS.

[0085] In this way, the position information acquisition unit 36 ​​acquires the position information of the rolled material SS and the shape of the rolled material SS based on the surface temperatures in the width direction and length direction of the rolled material SS. The threshold value Ts can be set arbitrarily depending on the temperature range of the rolled material SS, and is preferably set in the range of 300 to 750°C, for example.

[0086] The image data generating unit 37 generates image data that associates the position and shape of the rolled material with the surface temperature, and executes the image data generating step (step S203).

[0087] The image data generation unit 37 acquires the surface temperatures of the rolled material in the width direction and length direction measured by the temperature measurement unit 21, and generates image data that correlates the surface temperatures with positions within the plane of the rolled material. Specifically, the image data generation unit 37 generates image data by extracting the surface temperatures of the rolled material measured within the range of the contour shape of the rolled material acquired in the position information acquisition step of step S202. Since the position information of the rolled material is acquired in the position information acquisition step, image data that correlates the surface temperatures with positions within the plane of the rolled material is generated.

[0088] The prediction unit 39 uses the camber prediction model 38 to predict the camber of the rolled material when rough rolling is performed, and executes the prediction step (step S204). The prediction unit 39 stores the camber predicted in the prediction step of step S204 in the memory unit 32. The control unit 33 reads out the camber predicted in the prediction step of step S204 from the memory unit 32 and transmits it to the hot rolling line control unit 18.

[0089] When the hot rolling line control unit 18 acquires the camber transmitted from the camber prediction device 30, it executes a reference data acquisition step of reading and acquiring reference data from the storage unit 19 (step S103). The reference data is data indicating the camber that serves as a criterion for determining whether or not rough rolling can be performed. For example, the reference data can be an upper limit value of the camber that allows rough rolling to be performed.

[0090] The hot rolling line control unit 18 executes a rough rolling feasibility information generation process to generate rough rolling feasibility information that determines whether the material to be rolled is suitable for rough rolling based on the camber predicted by the camber prediction device 30 and the reference data (step S104).

[0091] In the possibility information generating process of step S104, the hot rolling line control unit 18 generates rough rolling possibility information, which determines that the material to be rolled is suitable for rough rolling (hereinafter also referred to as "possible to be charged") when the magnitude of the camber predicted by the camber prediction device 30 is within the reference data. Note that since the camber can be positive or negative, the magnitude is an absolute value.

[0092] In the feasibility information generation process of step S104, the hot rolling line control unit 18 generates rough rolling feasibility information, which determines that the material to be rolled is not suitable for rough rolling (hereinafter also referred to as unloadable) if the magnitude of the camber predicted by the camber prediction device 30 exceeds the reference data.

[0093] If the rough rolling feasibility information generated in the feasibility information generation step of step S104 indicates that the material cannot be loaded into the roughing mill, the material to be rolled is transported to a predetermined position without being loaded into the roughing mill 14. For example, a return process is performed in which the material to be rolled is returned to a position where it will be heated again in the heating furnace 11. In this way, if the rough rolling feasibility information indicates that the material cannot be loaded into the roughing mill, the material to be rolled is excluded from the rough rolling targets and rough rolling is performed.

[0094] If the rough rolling possibility information generated in the possibility information generation step of step S104 indicates that the material can be charged into the roughing mill, the material to be rolled proceeds to the next step for producing a hot-rolled steel plate.

[0095] That is, a width reduction step is performed in which the material to be rolled is reduced in width to a predetermined set width by the width reduction press device 13 (step S105). The material to be rolled whose width has been reduced in the width reduction step of step S105 is transported to the roughing mill 14.

[0096] The material to be rolled whose width has been reduced in the width reduction step of step S105 is rolled in the roughing mill 14, and the rough rolling step is executed (step S106). In the rough rolling step of step S106, it is preferable to perform rough rolling by excluding specific materials to be rolled from the targets of rough rolling based on the rough rolling feasibility information.

[0097] The material to be rolled that has been rough rolled in the rough rolling step of step S106 is transported to the finishing mill 15.

[0098] In the finishing mill 15, a finish rolling step is carried out in which the material to be rolled is rolled (step S107). In the finish rolling step of step S107, the material to be rolled is rolled to a predetermined product thickness.

[0099] In the cooling device 16, a cooling step is carried out in which the material to be rolled is cooled to a predetermined temperature (step S108). In the coiler 17, the material to be rolled cooled in step S108 is wound up, and a winding step is carried out (step S109).

[0100] Fig. 6 is an example of image data generated in the image data generation process of step S203 in Fig. 4. As shown in Fig. 6, the image data is generated as a two-dimensional image in a manner in which the surface temperature is measured over the entire width and length directions of the top surface of the rolled material.

[0101] The image data is created by associating the measured temperatures with the shape of the rolled material stored in the memory unit 32. That is, the image data is data in which the surface temperatures are associated with positions within the surface of the rolled material by measuring the surface temperatures in the width and length directions of the rolled material.

[0102] In other words, the image data is a two-dimensional image in which surface temperatures are assigned to two-dimensional coordinates set within the surface (top and bottom surfaces) of the rolled material. The image data generator 37 generates the image data as a color image, for example, by assigning a color to the surface temperature at each position within the surface of the rolled material.

[0103] The image data generating unit 37 may acquire data on the conveying speed of the rolled material from the hot rolling line control unit 18. The conveying speed may be the conveying speed at which the rolled material passes through the position where the temperature measuring unit 21 is installed.

[0104] The scanning radiation thermometer measures the width direction temperature distribution of the rolled material at a predetermined time interval according to a preset measurement frequency. Therefore, by using data related to the conveying speed of the rolled material, the width direction temperature distribution of the rolled material measured at the predetermined time interval can be converted into the width direction temperature distribution for each position in the longitudinal direction of the rolled material.

[0105] In this way, the image data generating unit 37 associates the temperature distribution in the width direction of the rolled material with the position on the surface of the rolled material using data related to the conveying speed.The image data generating unit 37 then generates a color image as a two-dimensional image, in which 50 to 100 levels of color are assigned to represent the surface temperature of the rolled material in a range of, for example, 750 to 1200°C.

[0106] Fig. 7 shows a schematic representation of the camber of the rolled material. In the example shown in Fig. 7, the distance in the width direction between the center position in the width direction at the tip of the rolled material and the center position in the width direction that is a predetermined distance from the tip in the longitudinal direction is taken as the tip camber C1.

[0107] The distance in the width direction between the center position in the width direction at the tail end of the rolled material and the center position in the width direction located a predetermined distance in the longitudinal direction from the tail end is defined as tail end camber C2.

[0108] In the example shown in Figure 7, both the leading and trailing ends of the rolled material are bent so as to camber toward the work side (OP side). Note that positive and negative camber may be defined depending on the direction of the curvature of the rolled material. For example, a rolled material bent toward the work side (OP side) may be considered a positive camber, and a rolled material bent toward the drive side (DR side) may be considered a negative camber.

[0109] If there is camber at the leading end of the material to be rolled, it may collide with a guide or the like when it is charged into the finishing rolling mill 15, causing operational problems. If there is camber at the tail end of the material to be rolled, it may collide with a guide or the like when it leaves the finishing rolling mill 15, causing operational problems.

[0110] The camber output from the camber prediction model 38 may be at least one of the leading edge camber C1 and the trailing edge camber C2, or the leading edge camber C1 or the trailing edge camber C2, whichever is larger in absolute value, may be used.

[0111] The camber of the rough-rolled material is formed with a non-uniform curvature in the longitudinal direction of the rough-rolled material. The curvature of the camber is large near the front and rear ends of the rough-rolled material, and small in the steady portion. Furthermore, the camber does not have a uniform curvature in the longitudinal direction of the rough-rolled material.

[0112] When the steady portion of the material to be rolled is horizontally rolled, the material is restrained by the edgers and side guides, so the rough-rolled material is less likely to bend. On the other hand, at the leading and trailing ends of the rough-rolled material, the edgers and side guides do not provide sufficient restraint, so the rough-rolled material is more likely to bend.

[0113] The dog bones formed in the rolled material by the width reduction press device 13 have different shapes at the steady portion, the leading end, and the trailing end of the rolled material even if the amount of width reduction is constant in the longitudinal direction.

[0114] If there is a temperature deviation in the width direction of the rolled material, the height of the dog bones will be uneven between the work side (OP side) and the drive side (DR side), and this unevenness will also have a distribution in the longitudinal direction.

[0115] In the width reduction press device 13, the material to be rolled is constrained on both the upstream and downstream sides by pinch rolls. When width reduction is performed on the leading end of the material to be rolled, the material has not yet reached the downstream side, so it is constrained only by the upstream pinch rolls. Similarly, when width reduction is performed on the tail end of the material to be rolled, it is constrained only by the downstream pinch rolls.

[0116] In the portions of the material not restrained by the pinch rolls, the amount of width reduction differs between the work side and the drive side of the material, causing the height of dog bones formed in the material to be uneven across the width and resulting in bending of the material. In other words, even if the temperature deviation across the width of the material is constant, the unevenness of the height of dog bones formed in the material and the amount of bending of the material will vary across the length.

[0117] Figure 8 shows a schematic diagram of the rolled material after width reduction by the width reduction press. More specifically, Figure 8(a) shows a cross section of line AA located at the front end of the rolled material. Figure 8(c) shows a cross section of line CC located at the tail end of the rolled material. Figure 8(b) shows a cross section of line BB located in the steady portion between the front end and the tail end of the rolled material.

[0118] In the example shown in Figure 8, the surface temperature of the work side (OP side) in the width direction of the rolled material is measured to be higher than the surface temperature of the drive side (DR side). As shown in Figures 8(a) to (c), the work side (OP side) and the drive side (DR side) are higher, i.e., thicker, than the center in the width direction of the rolled material. This type of rolled material shape is known as a dog-bone shape.

[0119] As can be seen from Figure 8, at the front and rear ends of the rolled material, the dog bone height formed on the work side (OP side) is greater than the dog bone height formed on the drive side (DR side).

[0120] Furthermore, compared to the dog-bone shape formed on the drive side (DR side), the dog-bone shape formed on the work side (OP side) has an increased thickness over a wider area from the width edge of the rolled material toward the center of the plate width.In contrast, in the steady part of the rolled material, even if there is a temperature deviation in the width direction of the rolled material, there is no significant difference between the dog-bone shapes formed on the work side (OP side) and the drive side (DR side).

[0121] Therefore, when rough rolling is performed on the rolled material after width reduction, the steady part of the roughly rolled material will mainly form a bend that corresponds to the temperature deviation in the width direction, whereas the front and tail ends of the roughly rolled material will form a bend that corresponds to the temperature deviation in the width direction and a bend that corresponds to the unevenness of the dogbone shape.

[0122] As described above, when there is a temperature deviation in the material to be rolled being loaded into the width reduction press device 13, even if the temperature deviation in the width direction is uniform in the longitudinal direction, the bending behavior will change depending on the longitudinal position of the roughly rolled material.

[0123] Therefore, when the temperature deviation in the width direction of the rolled material charged into the width reduction press device changes in the longitudinal direction, the bending behavior of the rough rolled material becomes even more likely to change depending on the longitudinal position. In particular, near the front and rear ends of the rolled material, the deformation caused by width reduction by the width press device and rough rolling by the rough rolling mill differs from that in the steady state.

[0124] Furthermore, the temperature of the material to be rolled is likely to decrease near the front end and near the tail end before it is extracted from the heating furnace 11 and becomes a rough-rolled material, which makes the temperature of the material to be rolled uneven. Therefore, sufficient accuracy cannot be obtained by simply predicting the camber of the rough-rolled material based on the temperature distribution in the width direction at a specific position in the longitudinal direction of the material to be rolled.

[0125] In contrast, in the present embodiment, image data is used that associates the position of the measurement surface of the rolled material with the surface temperature. This image data represents the temperature non-uniformity across the entire measurement surface. This image data is used as input data for the camber prediction device 30, allowing for a detailed understanding of the camber of the rolled material. As a result, it becomes possible to accurately determine whether or not the rough rolling of the rolled material is appropriate.

[0126] When measuring the camber of the rolled material, the camber measuring unit 22 measures the camber at the position of the leading or trailing end of the rolled material in correspondence with the position of the camber predicted by the camber prediction device 30.

[0127] For example, the contour shape of the rolled material is extracted from an image captured by a camera by image processing, and the camber of the rolled material is identified based on the extracted contour shape of the rolled material. In this case, it is advisable to identify in advance the correspondence between the distance between pixels and the actual distance in order to convert the dimensions of the image acquired by the camera into an actual scale.

[0128] Both the leading and trailing ends of the rolled material are bent toward the work side (OP side). In this case, the positive and negative camber can be defined according to the direction of the bend of the rolled material, with positive camber being when the rolled material is bent toward the work side (OP side) and negative camber being when the rolled material is bent toward the drive side (DR side).

[0129] The camber measurement unit 22 can be a measuring device that can measure the width direction position of the rolled material at the leading or trailing end and the center position in the longitudinal direction of the material. The camber measurement unit 22 can, for example, be provided with a camera that captures an image of the leading or trailing end of the rolled material from the top or bottom surface downstream of the roughing mill 14, and can identify the camber of the rolled material by applying image processing to the image captured by the camera that includes the leading or trailing end of the rolled material.

[0130] For example, the contour shape of the rolled material is extracted from an image captured by a camera by image processing, and the camber C of the rolled material is identified based on the extracted contour shape of the rolled material.

[0131] As described above, the camber prediction method for a rolled material of the present invention includes an image data generation step of generating image data that associates the position, shape, and surface temperature of the rolled material. It also includes a prediction step of predicting the camber of the rolled material after rough rolling using a camber prediction model trained by machine learning, with the image data as input data and the camber of the rolled material as output data. Therefore, the camber of the rolled material can be predicted based on the overall temperature distribution of the measurement surface (top or bottom surface) of the rolled material. That is, the camber of the rolled material can be predicted taking into account the temperature distribution in the width direction and the longitudinal direction (i.e., the conveying direction) of the rolled material. Therefore, the camber can be predicted with higher accuracy than conventional methods.

[0132] In the above embodiment, an example has been described in which a possibility information generating step of generating rough rolling possibility information is executed. However, the present invention is not limited to this example, and for example, rough rolling may be executed according to the camber predicted in the prediction step and the reference data. That is, rough rolling possibility information may be generated only when the magnitude of the camber predicted by the camber prediction device 30 exceeds the reference data.

[0133] In the above embodiment, an example has been described in which a scanning radiation thermometer is used as the temperature measurement unit 21. The temperature measurement unit 21 is not limited to this mode, and a spot radiation thermometer may also be used.

[0134] For example, a plurality of spot-type radiation thermometers may be arranged in a direction perpendicular to the conveying direction of the rolled material, and the surface temperature of the rolled material may be measured at a preset measurement interval while the rolled material is being conveyed. This allows the surface temperature of the rolled material to be measured in both the width direction and the length direction.

[0135] Furthermore, a thermal imaging camera may be used as the temperature measurement unit 21. The surface temperature of the entire surface of the rolled material may be acquired as two-dimensional data. The thermal imaging camera can acquire the surface temperature of the entire top or bottom surface of the rolled material in a single image. When a thermal imaging camera is used as the temperature measurement unit 21, a typical thermal imaging camera has the functions of the temperature measurement unit 21 and the position information acquisition unit 36, and therefore the thermal imaging camera constitutes the image data generation unit 37.

[0136] Furthermore, scanning radiation thermometers can measure the temperature distribution in the width direction of the rolled material at shorter intervals than spot-type radiation thermometers. Furthermore, scanning radiation thermometers are less expensive than thermal imaging cameras, have a high degree of freedom in selecting measurement wavelengths, and are suitable for temperature measurement on production lines.

[0137] Furthermore, since a scanning radiation thermometer can measure temperature faster than these thermometers, it can accurately measure the temperature of the entire measurement surface of the rolled material even when the conveying speed of the rolled material is high.

[0138] In the above embodiment, an example has been described in which the camber prediction model 38 is provided in the camber prediction device 30. The camber prediction model 38 may be provided in a device different from the camber prediction device 30 as long as it is connected to the camber prediction device 30 so as to be able to communicate data with the camber prediction device 30.

[0139] (Second embodiment) In the above-described embodiment, an example has been described in which camber is predicted using a camber prediction model that inputs image data and outputs camber. The data input to the camber prediction model may be at least one of a plurality of operation parameters of the width reduction press device and image data. Note that the same components as those in the first embodiment are assigned the same reference numerals and descriptions thereof will be omitted.

[0140] Fig. 9 shows an example of how a camber prediction model 50 according to the second embodiment is generated. The storage unit 51 shown in Fig. 9 differs from the storage unit 32 described in the first embodiment in that it stores operation parameters D3 of the width reduction press device 13. Here, the operation parameters of the width reduction press device refer to the operation conditions when width reduction is performed on the rolled material.

[0141] The camber prediction model 50 is trained by machine learning using at least one of a plurality of width reduction operation parameters D3, image data D1, and camber performance data D2 as input data.

[0142] A plurality of operational parameters of the width reduction press device 13 can affect the deformation of the rolled material. Examples of the operational parameters include a parameter related to the amount of width reduction of the rolled material and an operational parameter related to the shape of the die.

[0143] Examples of operational parameters related to width reduction include the width reduction of the material to be rolled and the feed rate of the material to be rolled, which is the feed pitch of the material between width reduction passes. These operational parameters related to width reduction have a greater effect on camber than other parameters, and are therefore preferably used as input data.

[0144] Fig. 10 shows the configuration of the width reduction press device 13. As shown in Fig. 10, the width reduction press device 13 transports the rolled material SA in a transport direction D1. The width reduction press device 13 is provided with a pair of width reduction dies 61 that face each other in the width direction of the rolled material SA.

[0145] The width reduction die 61 has a parallel portion 61a extending in the conveyance direction of the material SA to be rolled, and an inclined portion 61b formed so as to widen in the width direction of the material SA from one end side of the parallel portion 61a.

[0146] Examples of operational parameters related to the die shape include the length of the parallel portion 61a of the width reduction die 61 in the conveying direction D1 and the angle of the inclined portion 61b with respect to the conveying direction D1. These parameters affect the deformation state of the rolled material during width reduction. Therefore, changes in these parameters affect the state of camber formation.

[0147] As described above, the non-uniformity of the dogbone shape formed after width reduction varies depending on the operating conditions of the width reduction press. Furthermore, the operating parameters of the width reduction press affect the camber of the rolled material. By generating a camber prediction model 50 using the operating conditions of the width reduction press and inputting the operating conditions and image data, it becomes possible to output a camber prediction with higher accuracy. [Example]

[0148] As an example, in the hot rolling line shown in Fig. 1, the camber of the material to be rolled was predicted using the camber prediction device 30. Furthermore, based on the camber prediction, a decision was made as to whether or not the material to be rolled could be charged into the roughing mill.

[0149] The camber was predicted when the rough rolling of the rolled material was performed by the rough rolling mill. The predicted position of the camber in the hot rolling line 10 was set to after the first pass of the rough rolling pass. In other words, the camber was predicted when one rolling pass was performed by the reversing rolling mill of the hot rolling line 10 shown in FIG.

[0150] The rolled material was rectangular in top view, with a thickness of 150 to 300 mm and a width of 900 to 2000 mm. The rolled material removed from the heating furnace was descaled using a descaling device 12. Before width reduction was performed on this rolled material using a width reduction press device 13, the surface temperature was measured using a scanning radiation thermometer as a temperature measurement unit. The surface temperature was measured across the width and length of the measurement surface of the rolled material, i.e., over the entire measurement surface.

[0151] (Camber prediction model generation) Actual data of image data correlating the in-plane position of the rolled material with the surface temperature was generated. In addition, the rolled material was width-reduced by the width press device 4, and the camber of the rolled material, which had undergone one pass of rough rolling by the roughing mill, was measured by the camber measurement unit. The camber prediction model generation unit generated a camber prediction model by hot rolling 2,000 sheets of the rolled material.

[0152] The machine learning algorithm used to generate the camber prediction model is a convolutional neural network, in which image data is input from the first input layer, and one-dimensional array data in which the image data information is compressed is generated using two convolutional layers and two pooling layers.Operating parameters of the width reduction press device are input from the second input layer, and a neural network with three intermediate layers is constructed for the one-dimensional array data in which the image data information is compressed and the operating parameters of the width reduction press device.

[0153] In Example 1, image data generated by the image data generating unit was used as input to the camber prediction model, and the output of the camber prediction model was the camber at the leading end and the camber at the tail end of the rolled material.

[0154] In Example 2, the image data generated by the image data generating unit and the set values ​​of the width reduction amount and feed pitch as operation parameters of the width reduction press device were used as inputs to the camber prediction model 38. The output was set in the same manner as in Example 1.

[0155] As a result, the deviation between the predicted camber value and the actual value for the rolled material in Example 1 was 10 mm on average and 5 mm on standard deviation for the leading edge camber, and 12 mm on average and 5 mm on standard deviation for the trailing edge camber.

[0156] The deviation between the camber prediction results and the actual values ​​for the rolled material in Example 2 was 8 mm on average for the leading edge camber and 4 mm on standard deviation. The deviation for the tail edge camber was 8 mm on average and 4 mm on standard deviation. From the above, it was confirmed that the camber prediction model can accurately predict the camber at the leading edge and tail edge of the rolled material.

[0157] A hot rolling line was operated by applying the camber prediction model generated as Invention Example 1 to determine whether or not material could be charged to a roughing mill. For the camber prediction model generated as Invention Example 1, hot-rolled steel sheets were newly produced using 1,000 sheets of rolled material.

[0158] To determine whether or not the material could be loaded into the roughing mill, an upper limit was set for the tip camber of the material formed after the first pass of rolling by the roughing mill. The upper limit for the camber was set to 100 mm.

[0159] As a comparative example, a hot rolling line was operated using a conventional method without determining whether or not the material could be charged to the roughing mill. As a result, in the comparative example, threading troubles in the finishing mill occurred with a probability of 0.1% due to excessive camber of the material being rolled.

[0160] In contrast, in the inventive example, no threading trouble occurred in the finishing mill, which was caused by an excessively large camber in the material being rolled. [Explanation of symbols]

[0161] 10 Hot Rolling Line 30 Camber prediction device 35 Temperature acquisition section 36 Location information acquisition unit 37 Image data generation unit 38 Camber prediction model 39 Prediction Department 40 Camber prediction model generation unit 50 Camber Prediction Model

Claims

1. A method for predicting the camber of a rolled material when rough rolling is performed on a rolled material that has been heated and width-reduced, comprising: a temperature acquisition step of acquiring surface temperatures in the width direction and the longitudinal direction of the rolled material after the heating and before the width reduction; a position information acquisition step of acquiring the position and shape of the rolled material based on the surface temperature acquired in the temperature acquisition step; an image data generating step of generating image data in which the position and shape of the rolled material and the surface temperature are associated with each other; a prediction step of predicting the camber of the rolled material when the rough rolling is performed using a camber prediction model trained by machine learning, with the image data as input data and the camber as output data.

2. The width reduction is performed using a plurality of operational parameters that can affect the deformation of the rolled material, 2. The method for predicting camber of a rolled material according to claim 1, wherein the camber prediction model is learned by machine learning using at least one of the plurality of operational parameters of the width reduction and the image data as the input data.

3. The method for predicting camber of a rolled material according to claim 1 , wherein the temperature acquisition step is performed using a scanning radiation thermometer.

4. The method for predicting camber of a rolled material according to claim 2 , wherein the temperature acquisition step is performed using a scanning radiation thermometer.

5. A method for manufacturing a hot-rolled steel sheet using the camber prediction method for a rolled material according to any one of claims 1 to 4, a reference data acquisition step of acquiring reference data of the camber, which is a criterion for determining whether or not the rough rolling is to be performed; A rough rolling process of performing the rough rolling in accordance with the camber predicted in the prediction process and the reference data.

6. A rough rolling suitability information generating step is provided, which generates rough rolling suitability information that determines whether the rolled material is suitable for the rough rolling based on the camber predicted in the prediction step and the reference data, The method for manufacturing a hot-rolled steel sheet according to claim 5, wherein in the rough rolling step, the rough rolling is performed by excluding specific rolled materials from targets of the rough rolling based on the rough rolling feasibility information.

7. A method for generating a camber prediction model for a rolled material, the method comprising: generating a camber prediction model for predicting the camber of a rolled material when rough rolling is performed on the rolled material that has been heated and width-reduced; a temperature acquisition step of acquiring surface temperatures in the width direction and the longitudinal direction of the rolled material after the heating and before the width reduction; a position information acquisition step of acquiring the position and shape of the rolled material based on the surface temperature acquired in the temperature acquisition step; an image data generating step of generating image data in which the position and shape of the rolled material and the surface temperature are associated with each other; A method for generating a camber prediction model for a rolled material, comprising a camber prediction model generation step of generating a camber prediction model that predicts the camber of the rolled material that has been rough rolled by machine learning using a plurality of learning data, with the image data as input data and the camber as output data.

8. A camber prediction device for a rolled material that predicts the camber of the rolled material when rough rolling is performed on the rolled material that has been heated and width-reduced, a temperature acquisition unit that acquires the surface temperatures of the rolled material in the width direction and the longitudinal direction after the heating and before the width reduction; a position information acquisition unit that acquires the position and shape of the rolled material based on the surface temperature acquired by the temperature acquisition unit; an image data generating unit that generates image data in which the position and shape of the rolled material and the surface temperature are associated with each other; A camber prediction device for a rolled material, comprising: a prediction unit that predicts the camber of the rolled material after the rough rolling using a camber prediction model learned by machine learning, with the image data as input data and the camber as output data.

Images