Method for measuring the amount of warping of rolled materials, method for reducing the amount of warping, device for measuring the amount of warping and device for reducing the amount of warping, and, method for learning a warping measurement model and a warping measurement model

Abstract

We provide a technology for measuring the amount of bending of rolled materials that can measure the amount of bending of rolled materials with high precision. A method for measuring the amount of bending of a rolled material before and after a rolling mill in a hot rolling line, wherein an image of the rolled material after rolling is captured from an oblique upper angle using a camera capable of measuring luminance in the wavelength range from visible light to near-infrared light, the captured image of the rolled material is input into a bending measurement model prepared in advance by machine learning, the area where the plate width of the leading end of the rolled material decreases is estimated, a measurement image area provided for measuring the amount of bending is specified, the plate width edge of the rolled material is detected based on the luminance value of the image of the rolled material existing in the measurement image area, the image of the rolled material is divided along the rolling direction of the rolled material, the shape of the plate width edge in each divided image is approximated by a quadratic equation, and the amount of bending of the rolled material is quantified as curvature or bending height based on the quadratic approximated shape of the plate width edge.

Description

Technology Field

[0001] The present invention relates to a method for measuring the amount of warping of a rolled material before and after a rolling mill in a hot rolling line, a method for reducing the amount of warping, a device for measuring the amount of warping, and a device for reducing the amount of warping, and also to a method for learning a warping measurement model and a warping measurement model used therein. Background Technology

[0002] In the hot rolling line, a cast slab with a thickness of about 200 to 300 mm is rolled by multiple or single rough rolling mills, and if necessary, reverse rolled to become a steel plate, so-called sheet bar, with a thickness of about 20 to 80 mm. Then, the sheet bar is finally finished rolled into a steel plate with a thickness of about 0.6 to 30 mm by multiple finishing rolling mills arranged in series. In addition, the steel plate may be cooled with cooling water during hot rolling to suppress surface defects and control the microstructure.

[0003] However, during rough rolling or finish rolling, if there is asymmetry in the rolling state of the upper and lower surfaces of the steel plate, bending may occur in the height direction. Specifically, it is said that bending occurs in the steel plate when there is asymmetry in factors such as the coefficient of friction, roll speed, roll conditions, steel plate temperature, and the angle of incidence to the rolling mill. In particular, since the leading edge or trailing edge of the steel plate corresponds to an abnormal rolling zone, it is difficult to control the aforementioned factors, and thus the amount of bending tends to increase at the leading edge or trailing edge of the steel plate. If large bending occurs at the leading edge of the steel plate, there is a possibility of trouble occurring where the steel plate cannot be fed into the next stage rolling mill and cannot be rolled, or where the steel plate damages surrounding equipment. In particular, for thick steel plates, i.e., sheet bars, the possibility of damaging surrounding equipment is higher because the rigidity of the steel plate is high. In order to suppress the occurrence of such troubles, it is essential to measure the bending of the steel plate occurring in the stand or the preceding stand and to take measures to suppress it.

[0004] Patent Document 1 discloses a method for measuring the amount of bending of a rolled material before and after a rolling mill in a hot rolling line. The measurement method is configured as shown in FIG. 7, that is, using a camera, an image of the rolled material after rolling is captured from an obliquely upward angle of the rolled material, and the plate width edge portion of the rolled material is detected based on the brightness value of the captured image of the rolled material. Then, the shape of the plate width edge portion is approximated by a quadratic equation, and the amount of bending of the rolled material is quantified as curvature based on the quadratic approximated shape of the plate width edge portion. Prior art literature

[0005] Japanese Patent Publication No. 2019-181562 The problem to be solved

[0006] However, each of the above conventional technologies has the following problems that must be solved.

[0007] For example, an example of a bending measurement system described in Patent Document 1 is schematically illustrated in FIG. 7. FIG. 8 shows an example of an image in which the amount of bending could be measured normally with this device configuration. FIG. 9 also shows an example of an image in which the measured value of the amount of bending was abnormal in a conventional method. As shown in FIG. 8, when the shape of the leading edge of the steel plate is an upward bend close to a rectangle, the amount of bending can be quantified with relatively good precision from the image of the steel plate. The width edge of the steel plate on the operator side (OP) can be correctly extracted so that the amount of bending can be measured. In addition, the width edge on the motor side (DR) cannot be used for measuring the amount of bending because it is obstructed by surrounding equipment. On the other hand, when the shape of the leading edge of the steel plate is a complex shape such as a downward bend or a fishtail, there is room for improvement in the precision of the amount of bending measurement. In particular, when there are facilities such as side guides or descaling equipment around the exit side of the rolling mill that interfere with the image of the steel plate, the steel plate is photographed from an oblique upward angle. In such cases, depending on the shooting angle, it becomes difficult to extract the shape of the end of the steel plate in the width direction from the image. For example, as shown in Fig. 9, the amount of bending shows an abnormal value, and it becomes difficult to measure with high precision.

[0008] The present invention was made to solve the above problem, and its purpose is to provide a technology for measuring the amount of bending of a rolled material that can measure the amount of bending of a rolled material with high precision. means of solving the problem

[0009] A method for measuring the amount of warping of a rolled material according to the present invention, which advantageously solves the above problem, is a method for measuring the amount of warping of a rolled material before and after a rolling mill in a hot rolling line, wherein an image of the rolled material after rolling is captured from an oblique upper angle of the rolled material using a camera capable of measuring luminance in the wavelength range from visible light to near-infrared light, the captured image of the rolled material is input into a warping measurement model prepared in advance by machine learning, the region where the plate width of the leading end of the rolled material decreases is estimated, a measurement image region provided for measuring the amount of warping is specified, the plate width edge of the rolled material is detected based on the luminance value of the image of the rolled material existing in the measurement image region, the image of the rolled material is divided along the rolling direction of the rolled material, the shape of the plate width edge in each divided image is approximated by a quadratic equation, and based on the quadratic approximated shape of the plate width edge, the of the rolled material It is characterized by quantifying the amount of bending as curvature or bending height.

[0010] In addition, the method for measuring the amount of bending of a rolled material according to the present invention is,

[0011] (a) The above bending measurement model is machine learned using the past images of the rolled material captured by the camera and the manufacturing conditions of the past rolled material as explanatory variables, and the region where the plate width of the leading end of the rolled material decreases in the past images of the rolled material as the objective variable,

[0012] (b) The above bending measurement model is machine learned using logistic regression, crystal tree, neural network, and deep learning.

[0013] The back becomes a more desirable embodiment.

[0014] The method for reducing the amount of warping of a rolled material according to the present invention, which advantageously solves the above problem, is characterized by performing control to suppress the amount of warping of the rolled material based on the quantified amount of warping using the method for measuring the amount of warping of the rolled material of any of the above.

[0015] A device for measuring the amount of bending of a rolled material according to the present invention, which advantageously solves the above problem, is a device for measuring the amount of bending of a rolled material before and after a rolling mill in a hot rolling line, comprising: a camera capable of measuring luminance in the wavelength range from visible light to near-infrared light, which captures an image of the rolled material after rolling from an oblique upper angle of the rolled material; and an information processing device that, using a bending measurement model prepared in advance by machine learning, estimates an area where the plate width of the leading end of the rolled material decreases from the captured image of the rolled material, specifies a measurement image area provided for measuring the amount of bending, detects the plate width edge of the rolled material based on the luminance value of the image of the rolled material existing in the measurement image area, divides the image of the rolled material along the rolling direction of the rolled material, approximates the shape of the plate width edge in each divided image using a quadratic equation, and quantifies the amount of bending of the rolled material as curvature or bending height based on the quadratic approximated shape of the plate width edge. It is characterized by having.

[0016] The device for reducing the amount of warping of a rolled material according to the present invention, which advantageously solves the above problem, is characterized by being composed of a means for suppressing warping, which includes either a rolling condition that is changed or a crop cutting amount that is changed, based on the amount of warping quantified using the warping amount measuring device.

[0017] The learning method of a bending measurement model according to the present invention, which advantageously solves the above problem, is a learning method of a bending measurement model used to measure the amount of bending of a rolled material before and after a rolling mill in a hot rolling line, characterized by using a camera capable of measuring luminance in the wavelength range from visible light to near-infrared light to an image of the rolled material after rolling taken from an oblique upper angle of the rolled material and the manufacturing conditions of the rolled material as explanatory variables, and machine learning the area where the plate width of the leading end of the rolled material decreases in the image of the rolled material as an objective variable.

[0018] The bending measurement model according to the present invention, which advantageously solves the above problem, is a model learned by the learning method of the bending measurement model, and is characterized by having at least a photograph of a rolled material as an input variable and an area where the plate width of the leading end of the rolled material decreases as an output variable. Effects of the invention

[0019] According to the present invention, even if there is a region at the end of the rolled material where the plate width is reduced, the amount of bending can be accurately measured, so it contributes to reducing the amount of cutting crop and also contributes to reducing equipment damage accidents, making it industrially useful. Brief explanation of the drawing

[0020] FIG. 1 is a schematic block diagram showing the system configuration of a bending measuring device for rolled material according to one embodiment of the present invention. Figure 2 is a schematic diagram showing an example of constructing a bending measurement model. FIG. 3 is a schematic diagram illustrating a method for specifying an image provided for measuring the amount of bending from a captured image using a bending measurement model. FIG. 4 is a flowchart illustrating a method for measuring the amount of bending using a bending measuring device for rolled material according to the above embodiment. FIG. 5 is a schematic diagram showing an overview of a hot rolling line applying the above embodiment. Figure 6 is a schematic diagram illustrating an example of equipment trouble caused by upper bending of a steel plate. Figure 7 is a schematic conceptual diagram showing an example of a conventional bending measurement system. Figure 8 is an image of the leading edge of a rolled material in which the amount of bending was normally observed using a conventional method. FIG. 9 is an image of the leading edge of a rolled material, where the amount of bending was greater than in the conventional method. FIG. 10 is a graph comparing the amount of bending measured by the method according to the present invention with the amount of bending measured by the conventional method. Specific details for implementing the invention

[0021] (Form for carrying out the invention)

[0022] Embodiments of the present invention will be described in detail below. Furthermore, each drawing is schematic and may differ from the actual product. Also, the following embodiments are intended to illustrate devices or methods for embodying the technical concept of the present invention and do not limit the configuration to the following. That is, the technical concept of the present invention may be modified in various ways within the technical scope described in the claims.

[0023] [Hot Rolling Line]

[0024] First, with reference to FIG. 5, the configuration of a hot rolling line (100) to which a method for measuring the amount of bending of a rolled material, which is an embodiment of the present invention, is applied will be described.

[0025] In the hot rolling line (100) to which the method for measuring the bending amount of a rolled material according to the present embodiment is applied, first, a slab with a thickness of about 200 to 300 mm, a width of about 600 to 2200 mm, and a length of about 5 to 15 m, manufactured in the casting process which is the previous process, is heated to a temperature of about 1000 to 1250°C and then extracted from the heating furnace (110).

[0026] Next, in the rough rolling process, the steel plate (S) is rough rolled to a thickness of approximately 20 to 80 mm using multiple rough rolling mills (R1 to R5). If necessary during the rough rolling process, a cooling device for the steel plate may be formed. Examples of cooling devices include a device that cools the steel plate (S) using rod-shaped water, referred to as laminar cooling, or a device that strongly cools the steel plate (S) by vigorously spraying a large volume of water. Furthermore, in order to control the amount of bending of the steel plate (S), it is desirable to configure the cooling device so that the temperature of the upper and lower surfaces of the steel plate (S), that is, the deformation resistance of the upper and lower surfaces of the steel plate (S), can be adjusted. Specifically, it is desirable to configure the cooling device with a structure that allows the flow rate of cooling water on the upper and lower surfaces of the steel plate (S) to be adjusted, or with specifications that allow cooling of either the upper or lower surface of the steel plate (S) to be stopped.

[0027] Generally, the shape of the leading and trailing ends of a steel plate (S) after rough rolling is such that only the end portion in the width direction, referred to as a fishtail, extends in the length direction of the steel plate (S). For instance, if the fishtail shape differs in the width direction of the steel plate (S), there is a possibility that the steel plate (S) will become meandered during finishing rolling and develop into a problem. For this reason, from the perspective of operational stability, it is common practice to cut part or all of the leading and trailing ends of the steel plate (S) after rough rolling using a cropper (FSC) to form a rectangular shape.

[0028] Since the surface temperature of the steel sheet after rough rolling is high, ranging from 700 to 1000°C, iron oxide (scale) is formed. To prevent defects remaining in the final product, the iron oxide is removed by hydraulic pressure using a finisher scale breaker (FSB) before finish rolling.

[0029] In the finishing rolling process, a plurality of finishing rolling mills (F1 to F7) are arranged in series with a pitch of 2000 to 4000 mm, and each finishing rolling mill is equipped with a large frame called a housing and a plurality of rolling rolls arranged in the height direction. The finishing rolling mills have multiple types, such as a type called a 4-stage type in which two pairs of rolling rolls are arranged in the vertical direction of the steel plate, or a type called a 6-stage type in which three pairs of rolling rolls are arranged in the vertical direction of the steel plate. FIG. 5 shows examples of the configuration of 4-stage and 6-stage finishing rolling mills (F1 to F7). In the finishing rolling mills (F1 to F7) shown in FIG. 5, the steel plate (S) is conveyed by a conveying device not shown and is rolled from the right direction to the left direction in FIG. 5. At this time, the steel plate (S) is rolled by an upper and lower pair of work rolls arranged within the housing, and backup rolls are arranged above and below each of the work rolls.

[0030] Generally, the rolling load of the steel plate (S) is reduced by applying tension to the steel plate (S). Therefore, to apply tension between the finishing rolling mills, the tension of the steel plate (S) is controlled by raising and lowering the angle of the looper formed between the finishing rolling mills. In the finishing rolling, the steel plate (S) is finished rolled to a thickness of about 0.6 to 30 mm.

[0031] In addition, although varying according to the design specifications of each factory, thermometers (FET, FDT) for measuring the temperature of the steel plate (S) are installed at least on the entry and exit sides of the finishing rolling. The thermometer (FET) on the entry side of the finishing rolling may be placed before the finisher scale breaker (FSB) to eliminate its influence, or it may be installed before the cutting by the cropper (FCS). In particular, since the thermometer (FET) on the entry side of the finishing rolling is an important thermometer for controlling the amount of warping of the steel plate (S), it is desirable to use a thermometer capable of measuring both the upper and lower surfaces of the steel plate (S).

[0032] In hot rolling performed using equipment such as the tactical method, when measuring and quantifying bending, it is necessary to optimize optical conditions, such as radiation from the heated steel sheet. At the same time, it is required to quantitatively indicate bending in an area of ​​about 1.0 m or more at the leading edge of the steel sheet while dealing with cooling water sprayed between the scale breaker and stand at the finishing entry side, and reflections from equipment other than the looper.

[0033] In order to determine the bending behavior of the steel plate (S) due to rolling, it is necessary to determine the amount of bending of the steel plate (S) before and after rolling, the conditions for entry into the rolling mill, that is, the position of the rolling rolls when the steel plate (S) enters, and the positions of the steel plate (S) in the height and width directions. Before finishing rolling, there may be cases where a leveler is formed or the leading edge of the steel plate is cut by a cropper. Therefore, before finishing rolling, there is no bending of the steel plate (S), and it can be assumed that the steel plate (S) is straight. On the other hand, after rough rolling, there exists an area of ​​the leading edge of the steel plate (S) where the plate width is reduced, referred to as the crop. It is desirable to measure the amount of bending of the steel plate (S) in practice. In addition, since the shape of the leading or trailing end of the steel plate (S) may become a factor in the increase of bending during subsequent rolling, it is desirable to also determine the bending shape in a length range of about 200 to 400 mm in the rolling direction during measurement. As the steel plate (S) becomes thicker and more rigid after rough rolling or before the finishing rolling mill, and the risk of major troubles such as equipment damage increases, it is particularly important to measure the amount of bending of the steel plate (S) after rough rolling or before the finishing rolling mill.

[0034] FIG. 6 schematically illustrates an example of equipment trouble caused by bending at the leading edge of a steel plate (S). For example, when the upper surface of the steel plate (S) is the low-temperature side (LT) and the lower surface is the high-temperature side (HT) before entering the upstream rolling mill (Fig. 6(a)), the deformation resistance of the upper surface of the steel plate (S) is greater, and the deformation resistance of the lower surface of the steel plate (S) is smaller. Consequently, at the exit side of the upstream rolling mill shown in FIG. 6(b), the leading edge bends upward. If this bending becomes excessive, the steel plate may be unable to enter the downstream rolling mill, or it may get caught in the work roll (12) or backup roll (13) at the downstream rolling mill as shown in FIG. 6(c), resulting in equipment trouble. FD in FIG. 6 indicates the conveying direction of the steel plate (S).

[0035] [Method for Measuring the Bending Amount of Rolled Material]

[0036] When photographing the bending state of the steel plate (S) before and after the rolling mill from the side of the steel plate (S), it is difficult to determine the bending state of the steel plate (S) immediately after rolling or the bending state of the steel plate (S) immediately before rolling because the roll bite located between the work rolls (12) is inside the housing. Therefore, in this embodiment, the area immediately after the rolling mill or cropper, or between the housings of the rolling mill, is selected as the imaging field of view (21) of the imaging device (3). FIG. 7 is a schematic diagram for explaining the imaging angle (θ) of the imaging device (3). As shown in FIG. 7, the imaging angle (θ) of the imaging device (3) is preferably set within the range of 10 to 60° with respect to the horizontal direction, specifically as an angle for photographing an image of the steel plate (S) from an oblique upward direction. More preferably, θ is within the range of 30 to 60° with respect to the horizontal direction. When the imaging angle (θ) is less than the lower limit, when the height direction position of the steel plate (S) differs from the width direction of the steel plate (S), the steel plate (S) is captured at an angle, making it difficult to detect the edge portion of the steel plate (S). On the other hand, when the imaging angle (θ) is greater than the upper limit, it becomes difficult to obtain information in the height direction of the steel plate (S).

[0037] The distance (L) between the steel plate (S) and the imaging device (3), that is, the distance from the leading edge of the imaging device (3) to the center of the steel plate in the width direction of the steel plate (S), is not set to a particularly good range. However, if the distance (L) is extremely short, the imaging device (3) may be heated by radiant heat from the steel plate (S), which may cause a malfunction. A heat shield may be installed between the imaging device (3) and the steel plate (S). The measurement precision of the amount of bending of the steel plate (S) is also affected by the clarity of the image. On the other hand, the measurement precision also changes depending on the distance (L) and the number of pixels. If the distance (L) is extremely long, the measurement precision of the amount of bending of the steel plate (S) may decrease. It is possible to enlarge the captured image using a telephoto lens. On the other hand, considering the measurement precision of the amount of bending of the steel plate (S), it is desirable to set the shooting length per pixel to at least 5 mm or less. In FIG. 7, the height (H) of the imaging device (3) relative to the pass line of the steel plate (S) is preferably such that it can secure a field of view (W) in the width direction of the steel plate (S). Depending on the size of the steel plate (S), the field of view (W) is preferably about 2000 to 3200 mm. Depending on the steel plate temperature and equipment configuration, the distance (L) from the camera (3) to the center of the steel plate (S) in the width direction is preferably 3 to 10 m, and the height of the camera (3) from the pass line is preferably 800 to 4000 mm.

[0038] The imaging device (3) may be a color camera or a black and white camera. In the case of a color camera, it is good to detect the edge portion of the steel plate (S) using any of R (red), G (green), B (blue), or grayscale values.

[0039] In a hot rolling mill, fine iron oxide is scattered during rolling or descaling. For this reason, it is desirable to protect the imaging device (3) by placing it in a dustproof case. When photographing a steel plate (S) in an environment where a large amount of steam is generated, it is effective to capture only the luminance of wavelengths longer than visible light, such as near-infrared wavelengths, where the diffuse reflection of steam is reduced. It is also acceptable to use information only in the near-infrared wavelength range by attaching a filter that cuts off visible light to an imaging device (3) capable of measuring luminance of wavelengths ranging from visible light, for example, light described as having a wavelength of 360 to 830 nm in JIS Z 8120: 2001, to near-infrared light, for example, light described as having a wavelength of 700 to 2500 nm in JIS 0134: 2002. The same effect can be obtained by using a mid-infrared or far-infrared camera. On the other hand, there is a concern that the imaging device (3) may be expensive. However, since the brightness of the infrared wavelength decreases as the temperature of the steel plate (S) decreases, it is desirable to select the wavelength range to be measured according to the environment.

[0040] In calculating the amount of bending of a steel plate (S) from an image captured by an imaging device (3), the method described in Patent Document 1 may be used for the relationship between pixels and actual dimensions or for the selection of the captured image. In this embodiment, as shown in FIG. 1, the leading edge of the steel plate (S) is captured by an imaging device (3) installed at the exit of the roughing mill (R1 to R5), the cropper (FSC), and the upstream finishing mill (F1, F2). Then, as shown in FIG. 2, a bending measurement model is constructed by machine learning to detect the crop by labeling the shape of the range (crop) where the plate width of the leading edge of the steel plate narrows for the group of captured images. The bending measurement model is constructed, for example, on a server (6). Here, crops labeled from various image groups with different reflection methods, such as the width, thickness, presence or absence of warping, and shape of the crop, are treated as training data for the warping measurement model (Fig. 2(a)). Here, the labeling is performed by defining the range as a crop from the point where the boundary (edge) between the steel plate (S) and the background in the image changes from the width direction to the length direction, until it changes back to the width direction (Fig. 2(b)). Then, the constructed warping measurement model is applied to the training data to infer and estimate the crop (Fig. 2(c)). At this time, for image data that could not be inferred, it is desirable to learn the crop shape again to improve the accuracy of the warping measurement model.

[0041] FIG. 3 is a schematic diagram illustrating a method for specifying an image provided for measuring the amount of bending from a captured image using a bending measurement model. FIG. 4 is a flowchart illustrating a method for measuring the amount of bending using a bending measurement device for rolled material. In Step 1 (S01), the leading edge of a steel plate (S) is captured by a camera (3) installed on the exit side of a rolling mill, etc. Then, the newly captured image of the steel plate (S) is sent to a server (6) that operates the constructed bending measurement model through a machine-side relay panel (4) and a bending gauge (5). In Step 2 (S02), a crop is estimated from the image constructed in the server (6). After that, the image after applying the bending measurement model and the range data of the detected crop are sent to the bending gauge (5). For the measurement of the amount of bending, for example, the method of Patent Document 1 can be applied. In Step 3 (S03), the bending gauge (5) masks the cropped portion from the image based on the data (Fig. 3(c)). In Step 4 (S04), the shape of the plate width edge portion of the steel plate (S) is extracted from the remaining image, and the amount of bending is measured. In Step 5 (S05), the measurement result of the amount of bending is transmitted to an upper computer (7), for example, a rolling process computer, to calculate the amount of bending. The upper computer (7) transmits the amount of bending to a bending suppression device (8).

[0042] It is desirable to approximate the shape of the measured edge portion by a quadratic approximation and to quantify the amount of bending of the steel plate (S) as curvature or bending height. Specifically, first, a general quadratic formula is defined as Equation (1) shown below. The height in the plate thickness direction at the position where the edge portion is furthest from the pass line in the conveying direction (FD) of the steel plate (S) is defined as the bending height, with the upper bending being positive and the lower bending being negative. Additionally, the curvature of the edge portion is calculated using Equation (2) shown below, utilizing coefficients a, b, and c obtained by the least squares method. Furthermore, although Equation (2) can calculate the curvature of the edge portion, since the value changes in the rolling direction, a representative value can be obtained by taking the average value in the calculated rolling direction or by taking the curvature at the center of the rolling direction. Additionally, the value calculated by Equation (2) is positive. If the bending of the steel plate (S) is upward, a positive value is taken, and if it is downward, a negative value is taken. In addition, the shape of the edge may be approximated by other polynomials, but as the order increases, the influence of noise becomes greater, making it difficult to correctly calculate the curvature. Furthermore, if the approximation is performed using the least squares method with the general formula of a circle, it is not desirable because in cases where the bending is close to a straight line, a solution with a large curvature may be obtained.

[0043] y = ax 2 +bx+c (1)

[0044] 1 / r=2a / {1+(2ax+b) 2} 1.5 (2)

[0045] In Step 6 (S06), when the amount of bending is a predetermined value, the bending suppression device (8) first sends an instruction to change the rolling conditions when rolling with the downstream, particularly, immediately after, rolling. By rolling according to that instruction, bending is suppressed during subsequent rolling. The rolling conditions to be changed include the cooling conditions of the cooling device caused by the temperature difference between the upper and lower surfaces of the steel plate, the amount of cutting in the cropper, the line speed, etc., and the system is configured to automatically control the amount of change in those rolling conditions based on the magnitude of the amount of bending. Also, in Step 6 (S06), when the amount of bending is a predetermined value, the bending suppression device (8) sends an instruction to change the rolling conditions when rolling with the target rolling mill during the next rolling after the target rolling mill. By rolling according to that instruction, bending is suppressed during subsequent rolling by the target rolling mill. The rolling conditions to be changed include the cooling conditions of the cooling device or line speed caused by the temperature difference between the upper and lower surfaces of the steel plate, and the system automatically controls the amount of change in those rolling conditions based on the magnitude of the bending amount.

[0046] In this embodiment, the rough bending gauge, the finishing bending gauge, the server, the upper computer, the bending suppression device, etc., may be constructed as separate computers. On the other hand, multiple functions may be organized and constructed into a single computer.

[0047] Examples

[0048] The effects of the present invention were verified using the hot rolling equipment shown in FIG. 5. A continuously cast slab was heated, and after performing sizing press, rough rolling, cropping, and finisher scale breaker, hot finishing rolling was performed. A bending measurement system for the leading edge of the steel plate shown in FIG. 7 was installed on the exit side of the roughing mill (R5), and the amount of bending was calculated offline for the image of the leading edge of the steel plate (S) captured. A telephoto lens was attached to the imaging device and adjusted so that the size per pixel of the captured image was 2 mm. The frame rate of the imaging device was set to 20 fps, and an image in which the steel plate appeared as long as possible was selected for analysis. The length of the steel plate in the rolling direction obtained as the imaging field of view was 1.7 m. The amount of bending was evaluated in mm by approximating the edge shape using a quadratic equation and taking the height in the plate thickness direction at the position where the edge is furthest from the pass line in the conveying direction (FD) of the steel plate (S) as the bending height. In addition, the imaging angle (θ) was set to 50° with respect to the horizontal direction.

[0049] FIG. 10(a) shows the relationship between the measured bending amount calculated by the conventional method described in Patent Document 1 and the actual bending amount. FIG. 10(b) shows the relationship between the bending amount and the actual bending amount when the crop of the leading edge is estimated and masked by a bending measurement model from a captured image, and the edge portion in the width direction of the steel plate (S) is extracted from the remaining image. The bending measurement model was created by using 499 leading edge images with different leading edge shapes, widths, and thicknesses as teaching data to estimate the crop shape. Furthermore, leading edge images for which estimation could not be performed were trained as teaching data. In addition, the number of the above teaching data is preferably 100 or more to increase the precision of the bending measurement model characterized by the present invention, and furthermore, it is more preferable to have 499 or more. Although no upper limit is specifically set, it is preferable to have 100,000 or fewer.

[0050] Comparing FIG. 10(a) and FIG. 10(b), it can be seen that the measurement precision improves as the amount of bending increases with the application of the present invention. In addition, it can be seen that even in the vicinity of 0 mm, where the amount of bending is small, the deviation from the actual measured value is small, and a more high-precision measurement is possible.

[0051] Industrial applicability

[0052] According to the present invention, even if there is a region at the end of the rolled material where the plate width is reduced, the amount of bending can be accurately measured, so it contributes to reducing the amount of cutting crop and also contributes to reducing equipment damage accidents, making it industrially useful. Explanation of the symbols

[0053] 100 : Hot rolling line 110 : Heating furnace 1: Bending Amount Measurement System 3 : Imaging device (camera) 4: Aircraft relay team 5: Bending gauge 6 : Server 7 : Higher computer 8: Bending restraint device 12: Workroll 13: Backup Roll W : Imaging field of view R1~R5: Rough mill F1~F7: Finishing rolling mill S : Steel plate FSC: Cropshire FSB: Finisher Scale Breaker FET, FDT: Thermometer HT: High temperature side LT: Low temperature side FD: Return direction OP: Operator side DR: Motor side

Claims

Claim 1 A method for measuring the amount of warping of a rolled material before and after a rolling mill in a hot rolling line, comprising: capturing an image of the rolled material after rolling from an oblique upper angle using a camera capable of measuring luminance in the wavelength range from visible light to near-infrared light; inputting the captured image of the rolled material into a warping measurement model pre-created by machine learning; estimating the region where the plate width of the leading end of the rolled material decreases; specifying a measurement image area provided for measuring the amount of warping; detecting the plate width edge of the rolled material based on the luminance value of the image of the rolled material existing in the measurement image area; dividing the image of the rolled material along the rolling direction of the rolled material; approximating the shape of the plate width edge in each divided image using a quadratic equation; and quantifying the amount of warping of the rolled material as curvature or warping height based on the quadratic approximated shape of the plate width edge. Method for measuring bending amount. Claim 2 A method for measuring the amount of bending of a rolled material according to claim 1, wherein the bending measurement model uses the past image of the rolled material captured by the camera and the manufacturing conditions of the past rolled material as explanatory variables, and machine learns the region where the plate width of the leading end of the rolled material decreases in the past image of the rolled material as the objective variable. Claim 3 A method for measuring the amount of bending of a rolled material according to claim 1, wherein the bending measurement model is machine learned using any one of logistic regression analysis, crystal wood, neural network, and deep learning. Claim 4 A method for reducing the amount of warping of a rolled material, wherein control is performed to suppress the amount of warping of the rolled material based on the quantified amount of warping using a method for measuring the amount of warping of the rolled material described in any one of claims 1 to 3. Claim 5 A device for measuring the amount of bending of a rolled material before and after a rolling mill in a hot rolling line, comprising: a camera capable of measuring luminance in the wavelength range from visible light to near-infrared light, which captures an image of the rolled material after rolling from an oblique upper angle of the rolled material; and an information processing device that, using a bending measurement model pre-created by machine learning, estimates an area where the plate width of the leading end of the rolled material decreases from the captured image of the rolled material, identifies a measurement image area provided for measuring the amount of bending, detects a plate width edge portion of the rolled material based on the luminance value of the image of the rolled material existing in the measurement image area, divides the image of the rolled material along the rolling direction of the rolled material, approximates the shape of the plate width edge portion in each divided image using a quadratic equation, and quantifies the amount of bending of the rolled material as curvature or bending height based on the quadratic approximated shape of the plate width edge portion. Claim 6 A device for reducing the amount of warping of a rolled material, comprising means for suppressing warping, which includes either a rolling condition that changes or a crop cutting amount that changes, based on the amount of warping quantified using a warping amount measuring device for the rolled material described in paragraph 5. Claim 7 A learning method for a bending measurement model used to measure the amount of bending of a rolled material before and after a rolling mill in a hot rolling line, wherein the image of the rolled material after rolling, taken from an oblique upper angle of the rolled material using a camera capable of measuring luminance in the wavelength range from visible light to near-infrared light, and the manufacturing conditions of the rolled material are used as explanatory variables, and machine learning is performed using the area where the plate width of the leading end of the rolled material decreases in the image of the rolled material as the objective variable. Claim 8 A bending measurement model trained by the training method of the bending measurement model described in claim 7, wherein at least a photograph of a rolled material is used as an input variable, and the region where the plate width of the leading end of the rolled material decreases is used as an output variable.

Images