Bending angle compensation method in material machining process and material bending angle machining method and system

Abstract

The invention belongs to the technical field of material processing, and discloses a bending angle compensation method in a material processing process, and according to the thickness, the width, the target bending angle and the elastic modulus of a processing material, a springback angle, namely a compensation angle, is obtained through a multiple linear regression model. The invention further provides a material bending angle machining method and system, the target bending angle of the machined material is obtained, the compensation angle of the machined material is determined through the bending angle compensation method, the sum of the target bending angle and the compensation angle serves as the preset machining angle of the machined material, and the material is machined through a bending machine. According to the device, the springback angle generated in the material machining process can be accurately compensated, and the material machining bending angle can be accurately controlled.

Description

Technical Field

[0001] The present invention belongs to the technical field of material processing, and relates to bending angle control during material processing, and in particular to a bending angle compensation method, a material bending angle processing method and a system during metal material processing used in electrical engineering. Background Art

[0002] With the rapid development of the power industry, the demand for installation and maintenance of power equipment is increasing. In the installation and maintenance of power equipment, various busbars are widely used as electrical connectors. In order to ensure that the assembly meets the process requirements and the insulation distance requirements between equipment, the copper or aluminum busbar needs to be bent into a specific angle that meets the on-site use conditions and punched for bolt connection. In addition, during the installation of the grounding grid, the grounding flat iron also needs to be bent. Therefore, the precise control of the bending angle of the material is one of the key factors that determine whether the material can be installed in place.

[0003] At present, the bending of materials is mainly achieved by motor-driven bending machines. Motor-driven bending machines mainly use motors to drive the moving shaft to rotate around the fixed axis to achieve bending of materials (such as busbars). When bending busbars of different thicknesses, due to the different gaps between the busbars and the moving shaft, even if the machine is set to a fixed angle, the actual angle of the material after bending will still be different.

[0004] Moreover, most current bending machines use chains or gear transmissions to obtain greater torque, but the angle adjustment methods are limited. One is to use a delay switch to control the motor working time, and the other is to use a limiter to limit the range of motion of the gear or chain. Due to the different bending resistance of different materials and sizes, and the rebound characteristics of metal materials, the busbar will rebound to a certain angle after bending; resulting in these two methods being unable to accurately control the angle error and making it difficult to achieve fine adjustment of the bending angle. For example, when a traditional bending machine bends a copper busbar, when the setting is 90°, the actual measured angles of a 60×6mm copper busbar and an 80×6mm copper busbar are 96.5° and 91° respectively; when bending a 60×6mm copper busbar, aluminum busbar, or flat iron, the actual measured angles are 96.5°, 107.5°, and 94° respectively.

[0005] In order to solve the problem that electric bending machines cannot adapt to busbars of different sizes, most existing solutions use replaceable moving shafts, fixed shafts and fixtures to adapt to busbars of different sizes. However, these solutions are all fully manual or semi-automatic, that is, when changing the size of the busbar to be bent, it is necessary to manually change or manually select the size of the busbar to be bent. However, in actual construction sites, conditions are insufficient, and the thickness and width of commonly used substation busbars are mostly 2mm and 10mm apart, so the size of the busbar cannot be accurately judged by the naked eye alone. Summary of the invention

[0006] The purpose of the present invention is to address the deficiencies in the above-mentioned prior art and to provide a method for compensating bending angles in a material processing process. Firstly, by real-time detection and classification of materials, the springback angles of materials of different sizes and materials can be effectively compensated.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions to achieve it.

[0008] The present invention provides a method for compensating bending angles in a material processing process, which comprises the following steps:

[0009] Step 1: Obtain a processing material image, and obtain the material size according to the processing material image, including thickness and width along the bending axis;

[0010] Step 2: Input the processed material image into the recognition model, classify the material, and determine the material properties and elastic modulus;

[0011] Step 3: Based on the thickness, width, target bending angle and elastic modulus of the processed material, the springback angle, i.e. the compensation angle, is obtained through a multivariate linear regression model.

[0012] In the above step 1, image acquisition can use a high-resolution industrial camera to shoot the processed material from different angles to obtain image data of the processed material. In a specific implementation, the processed material can be photographed by a camera set directly above and on the side of the material, and images of the processed material can be obtained from two directions. The two images are then preprocessed and feature extracted to obtain the size of the processed material; the image collected from the side of the material can obtain the thickness of the material; the image collected from directly above the material can obtain the width of the material along the bending axis (that is, when the processed material is fixed on the bending machine, the direction in which the processed material is parallel to the fixed axis of the bending machine).

[0013] Preprocessing and feature extraction of the processed material image to obtain the size of the processed material includes the following steps:

[0014] Step 11: convert the processed material image into a grayscale image and perform filtering and denoising;

[0015] Step 12 performs edge detection on the denoised image;

[0016] Step 13 extracts the edge straight line to obtain the size of the processed material, including the thickness and the width along the bending axis.

[0017] In the above step 11, a Gaussian filter is used to process the grayscale image to reduce noise.

[0018] In the above step 12, the Canny edge detection method is used to perform edge detection on the denoised image; this step includes the following sub-steps:

[0019] Step 121 uses the Sobel operator to calculate the gradient magnitude and direction of the smoothed image;

[0020] Step 122 uses a non-maximum suppression method to remove pixels that are not local maxima in the gradient direction to accurately locate the edge;

[0021] Step 123 determines edge points based on dual threshold detection, and connects edges according to the hysteresis method to complete edge detection.

[0022] In the above step 13, the edge straight lines are extracted using the Hough transform algorithm, and the distances between the line segments are calculated to obtain the thickness of the processed material and the width along the bending axis.

[0023] In the above step 2, the recognition model is a convolutional neural network (CNN). The convolutional neural network (CNN) can be obtained through pre-training.

[0024] In the above step 3, the multivariate linear regression model is:

[0025] Y=β0+β1X1+β2X2+β3X3+β4X4+ε

[0026] Where Y represents the springback angle, X1 represents the thickness of the processed material, X2 represents the width of the processed material, X3 represents the target bending angle of the processed material, X4 represents the elastic modulus of the processed material, β0, β1, β2, β3, and β4 represent regression coefficients; ε represents the random error.

[0027] In the present invention, the regression coefficients β0, β1, β2, β3, and β4 are determined by using the least square method based on the material processing history data.

[0028] Through the above-mentioned bending angle compensation method, the springback angle generated during the material processing can be effectively compensated, thereby ensuring that the target bending angle is achieved.

[0029] The present invention also provides a material bending angle processing method, which comprises the following steps:

[0030] S1 obtains the target bending angle of the processed material;

[0031] S2 determines the compensation angle of the processed material by the bending angle compensation method;

[0032] S3 takes the sum of the target bending angle and the compensation angle as the predetermined processing angle of the processing material;

[0033] S4 processes the material through a bending machine;

[0034] S5 collects the bending angle of the material in real time and determines whether the bending angle reaches a predetermined angle. If so, the bending machine is stopped and the process proceeds to step S6; otherwise, step S4 is repeated;

[0035] S6 determines whether the deviation between the actual angle of the material after springback and the target bending angle is within the allowable error range. If so, the material processing is completed; otherwise, proceed to step S7;

[0036] S7 optimizes the multivariate linear regression model and then returns to step S2.

[0037] The above-mentioned material bending angle processing method can obtain the compensation angle based on the target bending angle of the processed material, detect the bending angle of the processed material in real time, and adjust the compensation angle in time, thereby improving the bending angle processing accuracy of the processed material and also improving work efficiency, providing reliable guarantee for the installation and maintenance of power equipment.

[0038] In the above step S4, a bending machine disclosed in the art may be used for processing.

[0039] In the above step S5, the processing material in the process is collected in real time by a camera located directly above the bending machine, and then the bending angle of the processing material is obtained by edge detection. The Canny edge detection method can be used for edge detection, and the angle formed by the detected edge is used as the bending angle.

[0040] In the above step S6, the camera is used again to measure the actual angle of the processed material after rebound. The measured actual angle is compared with the target angle to see if the angle deviation is within the allowable error range. If so, the material processing is completed. Otherwise, the actual rebound angle in the current iteration is determined based on the actual angle, target angle and compensation angle determined in step S2:

[0041] Actual rebound angle = target angle + compensation angle - actual angle.

[0042] Then, the actual springback angle and the thickness, width, target bending angle and elastic modulus of the processed material determined in step S2 are added to the processed material history data to update the processed material history data.

[0043] In the above step S7, the regression coefficient is determined by the least square method according to the updated historical data of the processed material, so as to optimize the multivariate linear regression model. Then, the process returns to step S2, and the compensation angle is re-obtained by using the optimized multivariate linear regression model, and the material is continuously bent.

[0044] Repeat the above steps S2-S7 until the bending angle of the material meets the set requirements, that is, whether the deviation between the actual angle of the material after springback and the target bending angle is within the allowable error range.

[0045] The present invention also provides a material bending angle processing system, which comprises:

[0046] Bending machine, used to process the bending angle of materials;

[0047] The bending angle control subsystem is used to compensate the bending angle, generate the predetermined processing angle and monitor; it includes:

[0048] The bending angle compensation module is used to obtain the material compensation angle, including an image preprocessing and feature extraction unit, a recognition model and a compensation angle acquisition unit; the image preprocessing and feature extraction unit is used to obtain the size of the processed material according to the processed material image, including the thickness and the width along the bending axis; the recognition model is used to classify the material according to the processed material image, and determine the material properties and elastic modulus; the compensation angle acquisition unit is used to obtain the springback angle, which is the compensation angle, through a multivariate linear regression model according to the thickness, width, target bending angle and elastic modulus of the processed material;

[0049] A processing preset angle acquisition module is used to take the sum of the target bending angle and the compensation angle as the processing preset angle of the material;

[0050] The bending angle monitoring module is used to monitor the bending angle in real time, and includes a first judgment unit and a second judgment unit; the first judgment unit is used to judge whether the predetermined angle is reached according to the bending angle of the material collected in real time, and if so, the bending machine is stopped, otherwise the bending machine continues to process the material; the second judgment unit is used to judge whether the deviation between the actual angle of the material after the material is processed and rebounded and the target bending angle is within the allowable error range, and if so, the material processing is completed, otherwise, the actual rebound angle is determined, and the actual rebound angle and the thickness, width, target bending angle and elastic modulus of the processed material are added to the historical data of the processed material;

[0051] The optimization module is used to optimize the multivariate linear regression model and then push the optimized multivariate linear regression model to the bending angle compensation module.

[0052] In the above bending angle compensation module, the image preprocessing and feature extraction unit, the recognition model, and the compensation angle acquisition unit perform operations according to the above steps 1 to 3.

[0053] The above-mentioned processing predetermined angle module performs the operation according to the above-mentioned step S3;

[0054] In the bending angle monitoring module, the first judgment unit and the second judgment unit perform operations according to the above steps S5-S6.

[0055] The above optimization module performs operations according to the above step S7.

[0056] The above-mentioned bending machine is a conventional device already disclosed in the art.

[0057] The material bending angle processing system further includes a camera, which is located above the bending machine and is used to collect the processed material in real time during the processing and obtain the bending angle of the processed material. The camera adopts conventional equipment.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1) The bending angle compensation method in the material processing process provided by the present invention determines the compensation angle through a multivariate linear regression model according to the material size, material properties and target bending angle, and can accurately compensate for the springback angle generated in the material processing process to ensure that the target bending angle is achieved;

[0060] 2) The material bending angle processing method and system provided by the present invention only need to determine the target bending angle to realize automatic processing of the material bending angle with high precision, which not only improves the work efficiency but also reduces the manual error, and is particularly suitable for busbar bending operations in large-scale production and complex environments;

[0061] 3) The material bending angle processing method and system provided by the present invention, combined with the bending angle compensation method and real-time angle monitoring, can achieve high-precision measurement and angle control of the material processing bending angle. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the process of the bending angle compensation method in the material processing process;

[0063] Figure 2 Schematic diagram of the process of preprocessing and feature extraction for processed material images;

[0064] Figure 3 The busbar image edge detection result; (a) corresponds to the busbar thickness, (b) corresponds to the busbar width;

[0065] Figure 4 It is a schematic diagram of the process flow of the material bending angle processing method. DETAILED DESCRIPTION

[0066] The following will clearly and completely describe the technical solutions of various embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0067] Example 1

[0068] This embodiment provides a method for compensating bending angle during material processing. Figure 1 As shown, it includes the following steps:

[0069] Step 1: Obtain a processed material image, and obtain the material size, including thickness and width along the bending axis, based on the processed material image.

[0070] The processing material in this embodiment is a material with uniform thickness. When the processing material is a busbar, its width is the width along the bending axis; when the processing material is a special-shaped material, the width along the bending axis refers to the width of the contact part between the processing material and the fixed axis of the bending machine along the fixed axis.

[0071] Two high-resolution industrial cameras can be used to photograph the processed material from different angles to obtain images of the processed material. In a specific implementation, the processed material can be photographed by cameras set directly above and on the side of the material, and images of the processed material can be obtained from two directions. The two images are then preprocessed and feature extracted to obtain the size of the processed material. The image collected from the side of the material can obtain the thickness of the material; the image collected from directly above the material can obtain the width of the material along the bending axis (that is, when the processed material is fixed on the bending machine, the direction in which the processed material is parallel to the fixed axis of the bending machine).

[0072] The processed material image is preprocessed and feature extracted to obtain the size of the processed material, such as Figure 2 As shown, it specifically includes the following steps:

[0073] Step 11 converts the processed material image into a grayscale image and performs filtering and denoising.

[0074] The processed material image is converted into a grayscale image to simplify subsequent processing. The grayscale image is then processed using a Gaussian filter to reduce noise.

[0075] Gaussian filtering is expressed as follows:

[0076]

[0077] Where g(x,y) is the value of the filter at point (x,y), and σ is the standard deviation, which determines the width of the filter.

[0078] Step 12 performs edge detection on the denoised image.

[0079] This step uses the Canny edge detection method to perform edge detection on the denoised image; this step includes the following sub-steps:

[0080] Step 121 uses the Sobel operator to calculate the gradient magnitude and direction of the smoothed image.

[0081] In this step, based on the Sobel operator, the convolution template is first applied to the x and y directions respectively, and then the gradient amplitude and direction are calculated. The calculation formula is as follows:

[0082] d (x) =[[-1,0,1],[-2,0,2],[-1,0,1]]*A;

[0083] d (y) =[[-1,-2,-1],[0,0,0],[1,2,1]]*A;

[0084]

[0085] Among them, S is the gradient amplitude, θ is the direction, and A represents the smoothed image.

[0086] Step 122 uses a non-maximum suppression method to remove pixels that are not local maxima in the gradient direction and accurately locate the edge.

[0087] In this step, the gradient magnitude of the pixel and its two adjacent pixels are checked in the gradient direction. For example, if the gradient direction is horizontal, the pixels to the left and right of the pixel are checked. If the gradient magnitude of the current pixel is greater than the gradient magnitudes of the two adjacent pixels in the gradient direction, it is considered to be a local maximum. Only pixels that are considered to be local maxima will be retained, and other pixels will be suppressed (or removed).

[0088] In this way, non-maximum suppression can effectively thin the edges, retaining only the true edges while removing unnecessary noise and minor lines.

[0089] Step 123 determines edge points based on dual threshold detection, and connects edges according to the hysteresis method to complete edge detection.

[0090] In this step, two thresholds, high and low, are set. Points with gradient amplitudes lower than the low threshold are marked as non-edge points, and points with gradient amplitudes higher than the high threshold are marked as edge points, so as to distinguish strong edges from weak edges.

[0091] Connecting edges with hysteresis: The points between two thresholds are judged as edges based on their connectivity. Ensure edge continuity.

[0092] Step 13 extracts the edge straight line to obtain the size of the processed material, including the thickness and the width along the bending axis.

[0093] The Hough transform algorithm is used to extract edge lines and calculate the distance between line segments to obtain the thickness and width of the processed material.

[0094] The Hough transform algorithm is used to extract edge lines, calculate the distance between line segments, and calculate the width and thickness of the busbar.

[0095] r=xcosθ+ysinθ

[0096] Where (x, y) is a point on the image, and (r, θ) is a point in the Hough space. By accumulating votes in the Hough space, the local maximum is found, which corresponds to the straight line in the original image.

[0097] Figure 3 The image of the rectangular busbar after preprocessing and feature extraction is given. From the straight lines extracted from the image, it can be determined that the thickness of the material is 10 mm and the width along the bending axis (that is, the width of the busbar) is 125 mm.

[0098] Step 2: Input the processed material image into the recognition model, classify the material, and determine the material properties and elastic modulus.

[0099] In this embodiment, the recognition model is a convolutional neural network (CNN), which takes the processed material image as input and the material properties (such as copper) and elastic modulus as output. The convolutional neural network (CNN) can be obtained by pre-training, see Xie Fengyun, Dong Jiankun, Fu Yu, et al. Transfer learning and CNN motor fault diagnosis method [J]. Mechanical Science and Technology, 2024, 43(03): 513-519. DOI: 10.13433 / j.cnki.1003-8728.20220196.

[0100] Then, the images of the actual busbars are used as training data to perform transfer learning on the pre-trained convolutional neural network (CNN) and fine-tune the model parameters to adapt it to the busbars of specified materials and sizes.

[0101] For example, inputting the image of the copper busbar processing material into the trained recognition model can directly output the material properties and elastic modulus of copper and 115×10 9 Pa.

[0102] Step 3: Based on the thickness, width, target bending angle and elastic modulus of the processed material, the springback angle, i.e. the compensation angle, is obtained through a multivariate linear regression model.

[0103] In this embodiment, a multiple linear regression method is adopted to fit the springback angle of busbars of different materials with respect to material thickness, width, target bending angle and material elastic modulus according to actual test data.

[0104] The multiple linear regression model is:

[0105] Y=β0+β1X1+β2X2+β3X3+β4X4+ε;

[0106] Where Y represents the springback angle, X1 represents the thickness of the processed material, X2 represents the width of the processed material, X3 represents the target bending angle of the processed material, X4 represents the elastic modulus of the processed material, β0, β1, β2, β3, and β4 represent regression coefficients; ε represents the random error.

[0107] Taking busbar processing as an example, the construction process of the above multivariate linear regression model is explained, including the following steps:

[0108] (1) Obtain busbar data of different materials, including thickness, width, target bending angle, and corresponding springback angle.

[0109] (2) Use the least squares method to determine the regression coefficient.

[0110] Represent the multiple linear regression model in matrix form:

[0111] Y=Xβ+ε

[0112] Wherein, Y represents an n×1 dependent variable vector, which is composed of the actual springback angle of the processed material; X represents an n×(k+1) independent variable matrix (including a column of 1 for the intercept term), which is composed of processing material parameters (including thickness, width, target bending angle and elastic modulus); β represents a (k+1)×1 regression coefficient vector, ε represents an n×1 error vector, n represents the number of samples, k represents the number of independent variables, and in this embodiment, k=4.

[0113] The estimated values ​​of the regression coefficients are obtained based on the least squares estimation:

[0114]

[0115] in, represents the estimated value vector of regression coefficients, X' represents the transpose of X, (X'X) -1 represents the inverse matrix of (X'X).

[0116] Then, the estimated equation is obtained based on the estimated value vector of the regression coefficient:

[0117]

[0118] in, express Vector elements.

[0119] By inputting the material thickness, width, target bending angle and material elastic modulus into the multivariate linear regression model constructed above, the springback angle can be calculated, which is the compensation angle that needs to be compensated in material processing.

[0120] When processing the material, the sum of the target bending angle and the compensation angle is used as the predetermined processing angle, and the material is processed by the bending machine. Through the above bending angle compensation method, the springback angle generated during the material processing can be effectively compensated, thereby ensuring that the target bending angle is achieved.

[0121] For example, based on the historical data of busbar bending processing, the constructed multivariate linear regression model is:

[0122] Y=3.0X1-0.6X2-2.5X3+2.361×10 -9 X4;

[0123] In the formula, Y represents the springback angle, X1 represents the thickness of the processed material (unit, mm), X2 represents the width of the processed material (unit, mm), X3 represents the target bending angle of the processed material (unit, degree), X4 represents the elastic modulus of the processed material (unit, Pa), ε represents the random error, and in this embodiment, β0=ε=0.

[0124] Then, the material dimensions determined in step 1 (thickness X1 = 10 mm and width X2 = 125 mm), the material elastic modulus determined in step 2 (X4 = 115 × 10 9 Pa) and the target bending angle (X3 = 90°) are substituted into the above multivariate linear regression model, and it can be determined that the springback angle is 1.5°; therefore, the actual processing predetermined angle is X3 + 1.5°.

[0125] Example 2

[0126] This embodiment is a further improvement on the embodiment 1. This embodiment provides a material bending angle processing method, such as Figure 4 As shown, it includes the following steps:

[0127] S1 obtains the target bending angle of the processed material.

[0128] This step determines the target bending angle of the processed material (eg busbar) according to user requirements.

[0129] S2 determines the compensation angle of the processed material by the bending angle compensation method.

[0130] In this step, the compensation angle during the material bending process is determined by the bending angle compensation method provided in Example 1.

[0131] S3 uses the sum of the target bending angle and the compensation angle as the predetermined processing angle of the processing material.

[0132] S4 processes the material through a bending machine.

[0133] The bending machine disclosed in the art can be used for processing. According to the size information of the processing material (such as the busbar) obtained in the above step S2, the clamp and the fixed axis of the bending machine are automatically adjusted to automatically clamp the processing material. Then the bending machine is started and the material is bent by driving the moving axis of the bending machine.

[0134] S5 collects the bending angle of the material in real time and determines whether the bending angle reaches a predetermined angle. If so, the bending machine is stopped and the process proceeds to step S6; otherwise, step S4 is repeated.

[0135] The camera located directly above the bending machine collects the processed material in real time during the processing, and then obtains the bending angle of the processed material through edge detection. The Canny edge detection method can be used for edge detection, and the angle formed by the detected edge is used as the bending angle.

[0136] S6 determines whether the deviation between the actual angle of the material after springback and the target bending angle is within the allowable error range. If so, the material processing is completed; otherwise, enter step S7.

[0137] The camera can be used again to measure the actual angle of the processed material after rebound. Compare the measured actual angle with the target angle to see if the angle deviation is within the allowable error range. If so, the material processing is completed. Otherwise, the actual rebound angle in the current iteration is determined based on the actual angle, target angle, and compensation angle determined in step S2:

[0138] Actual rebound angle = target angle + compensation angle - actual angle.

[0139] Then, the actual springback angle and the thickness, width, target bending angle and elastic modulus of the processed material determined in step S2 are added to the processed material history data to update the processed material history data.

[0140] S7 optimizes the multivariate linear regression model and then returns to step S2.

[0141] In the above step S7, the regression coefficient is determined by the least square method according to the updated historical data of the processed material, so as to optimize the multivariate linear regression model. Then, the process returns to step S2, and the compensation angle is newly obtained by using the optimized multivariate linear regression model to continue bending the material.

[0142] Repeat the above steps S2-S7 until the bending angle of the material meets the set requirements, that is, whether the deviation between the actual angle of the material after springback and the target bending angle is within the allowable error range.

[0143] Example 3

[0144] This embodiment is a further improvement on the embodiment 2. This embodiment provides a material bending angle processing system, which includes:

[0145] Bending machine, used to process the bending angle of materials;

[0146] The bending angle control subsystem is used to compensate the bending angle, generate the predetermined processing angle and monitor; it includes:

[0147] The bending angle compensation module is used to obtain the material compensation angle, including an image preprocessing and feature extraction unit, a recognition model and a compensation angle acquisition unit; the image preprocessing and feature extraction unit is used to obtain the size of the processed material according to the processed material image, including the thickness and the width along the bending axis; the recognition model is used to classify the material according to the processed material image, and determine the material properties and elastic modulus; the compensation angle acquisition unit is used to obtain the springback angle, which is the compensation angle, through a multivariate linear regression model according to the thickness, width, target bending angle and elastic modulus of the processed material;

[0148] A processing preset angle acquisition module is used to take the sum of the target bending angle and the compensation angle as the processing preset angle of the material;

[0149] The bending angle monitoring module is used to monitor the bending angle in real time, and includes a first judgment unit and a second judgment unit; the first judgment unit is used to judge whether the predetermined angle is reached according to the bending angle of the material collected in real time, and if so, the bending machine is stopped, otherwise the bending machine continues to process the material; the second judgment unit is used to judge whether the deviation between the actual angle of the material after the material is processed and rebounded and the target bending angle is within the allowable error range, and if so, the material processing is completed, otherwise, the actual rebound angle is determined, and the actual rebound angle and the thickness, width, target bending angle and elastic modulus of the processed material are added to the historical data of the processed material;

[0150] The optimization module is used to optimize the multivariate linear regression model and then push the optimized multivariate linear regression model to the bending angle compensation module.

[0151] In the above bending angle compensation module, the image preprocessing and feature extraction unit, the recognition model, and the compensation angle acquisition unit perform operations according to the above steps 1 to 3.

[0152] The above-mentioned processing predetermined angle module performs the operation according to the above-mentioned step S3;

[0153] In the bending angle monitoring module, the first judgment unit and the second judgment unit perform operations according to the above steps S5-S6.

[0154] The above optimization module performs operations according to the above step S7.

[0155] The above-mentioned bending machine is a conventional device already disclosed in the art.

[0156] The material bending angle processing system further includes a camera, which is located above the bending machine and is used to collect the processed material in real time during the processing and obtain the bending angle of the processed material. The camera adopts conventional equipment.

[0157] In summary, the present invention realizes the compensation design of material processing angle and automatic control of bending angle processing by adopting advanced image processing and deep learning technology. The present invention can realize real-time detection and classification of busbars through high-resolution industrial cameras and classification models, and realize accurate bending of busbars of different sizes and materials. At the same time, the springback angle of metal materials is predicted and compensated by machine learning algorithms to further improve the bending accuracy. The present invention not only reduces manual operations, improves work efficiency and accuracy, but also solves the problem of inconvenient operation when conditions on the construction site are insufficient, and provides reliable guarantee for the installation and maintenance of power equipment.

[0158] Those skilled in the art will appreciate that the embodiments herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for compensating bending angle in a material processing process, characterized in that: The following steps are involved: Step 1: Obtain a processing material image, and obtain the material size according to the processing material image, including thickness and width along the bending axis; Step 2: Input the processed material image into the recognition model, classify the material, and determine the material properties and elastic modulus; Step 3: Based on the thickness, width, target bending angle and elastic modulus of the processed material, the springback angle, i.e. the compensation angle, is obtained through a multivariate linear regression model.

2. The bending angle compensation method in material processing according to claim 1, characterized in that: Preprocessing and feature extraction of the processed material image to obtain the size of the processed material includes the following steps: Step 11: convert the processed material image into a grayscale image and perform filtering and denoising; Step 12 performs edge detection on the denoised image; Step 13 extracts the edge straight line to obtain the size of the processed material, including the thickness and the width along the bending axis.

3. The bending angle compensation method in the material processing process according to claim 2, characterized in that: In step 12, edge detection is performed on the denoised image using the Canny edge detection method; this step includes the following sub-steps: Step 121 uses the Sobel operator to calculate the gradient magnitude and direction of the smoothed image; Step 122 uses a non-maximum suppression method to remove pixels whose gradient directional entropy is not a local maximum, and accurately locates the edge; Step 123 determines edge points based on dual threshold detection, and connects edges according to the hysteresis method to complete edge detection.

4. The bending angle compensation method in material processing according to claim 2, characterized in that: In step 13, the edge straight lines are extracted using the Hough transform algorithm, and the distances between the line segments are calculated to obtain the thickness and width of the processed material.

5. The bending angle compensation method in material processing according to claim 1, characterized in that: In step 2, the recognition model is a convolutional neural network (CNN).

6. The bending angle compensation method in material processing according to claim 1, characterized in that: In step 3, the multiple linear regression model is: Y=β0+β1X1+β2X2+β3X3+β4X4+ε; Where Y represents the springback angle, X1 represents the thickness of the processed material, X2 represents the width of the processed material, X3 represents the target bending angle of the processed material, X4 represents the elastic modulus of the processed material, β0, β1, β2, β3, and β4 represent regression coefficients; ε represents the random error.

7. The bending angle compensation method in material processing according to claim 6, characterized in that: In step 3, the regression coefficients are determined using the least squares method.

8. A material bending angle processing method, characterized in that: The following steps are involved: S1 obtains the target bending angle of the processed material; S2 determining the compensation angle of the processed material by the bending angle compensation method according to any one of claims 1 to 7; S3 takes the sum of the target bending angle and the compensation angle as the predetermined processing angle of the processing material; S4 processes the material through a bending machine; S5 collects the bending angle of the material in real time and determines whether the bending angle reaches a predetermined angle. If so, the bending machine is stopped and the process proceeds to step S6; otherwise, step S4 is repeated; S6 determines whether the deviation between the actual angle of the material after springback and the target bending angle is within the allowable error range. If so, the material processing is completed; otherwise, proceed to step S7; S7 optimizes the multivariate linear regression model and then returns to step S2.

9. The material bending angle processing method according to claim 8, characterized in that: In step S6, the actual springback angle in the current iteration is determined according to the actual angle, the target angle and the compensation angle determined in step S2; then the actual springback angle and the thickness, width, target bending angle and elastic modulus of the processed material determined in step S2 are added to the historical data of the processed material to update the historical data of the processed material; In step S7, the regression coefficient is determined by the least square method according to the updated processing material historical data, so as to optimize the multivariate linear regression model; Then, the process returns to step S2, and the compensation angle is re-acquired using the optimized multivariate linear regression model, and the material is continued to be bent.

10. A material bending angle processing system, characterized in that: include: Bending machine, used to process the bending angle of materials; Bending angle control subsystem, used for compensating bending angle, generating processing preset angle and monitoring; It includes: The bending angle compensation module is used to obtain the material compensation angle, including an image preprocessing and feature extraction unit, a recognition model and a compensation angle acquisition unit; the image preprocessing and feature extraction unit is used to obtain the size of the processed material according to the processed material image, including the thickness and the width along the bending axis; the recognition model is used to classify the material according to the processed material image, and determine the material properties and elastic modulus; the compensation angle acquisition unit is used to obtain the springback angle, which is the compensation angle, through a multivariate linear regression model according to the thickness, width, target bending angle and elastic modulus of the processed material; A processing preset angle acquisition module is used to use the sum of the target bending angle and the compensation angle as the processing preset angle of the processing material; The bending angle monitoring module is used to monitor the bending angle in real time, and includes a first judgment unit and a second judgment unit; the first judgment unit is used to judge whether the predetermined angle is reached according to the bending angle of the material collected in real time, and if so, the bending machine is stopped, otherwise the bending machine continues to process the material; the second judgment unit is used to judge whether the deviation between the actual angle of the material after the material is processed and rebounded and the target bending angle is within the allowable error range, and if so, the material processing is completed, otherwise, the actual rebound angle and the thickness, width, target bending angle and elastic modulus of the processed material are added to the historical data of the processed material; The optimization module is used to optimize the multivariate linear regression model and then push the optimized multivariate linear regression model to the bending angle compensation module.

11. The material bending angle processing system according to claim 10, characterized in that: It also includes a camera; the camera is located directly above the bending machine and is used to collect real-time data of the processed material during the processing process and obtain the bending angle of the processed material.

Images