A welding control method for copper and aluminum substrates

Abstract

The present invention discloses a welding control method for copper and aluminum base materials, which relates to the technical field of welding. A welding control method for copper and aluminum base materials includes weld seam image acquisition and processing, dynamic welding control, and weld seam quality inspection and feedback. By accurately extracting the volume characteristics of the three-dimensional model of the weld seam, the present invention can accurately output the corresponding amount of brazing material, which helps to reduce material waste and ensure the economy and efficiency of the welding process; determining the initial welding parameters according to the weld seam volume characteristic value and the amount of brazing material enables the welding process to be more precisely adjusted according to the specific welding part and requirements, which not only improves the welding quality but also makes the welding process more controllable; during the welding process, the welding parameters are adjusted in real time, and through the max pooling of the first parameter adjustment strategy and the second parameter adjustment strategy, the most comprehensive and optimal welding parameter adjustment strategy can be obtained, realizing the comprehensive monitoring and optimization of the welding environment.

Description

Technical Field

[0001] The present invention relates to the technical field of welding, and particularly to a welding control method for copper and aluminum base materials. Background Art

[0002] Copper-aluminum welding technology is a technology widely used in the fields of electrical engineering, refrigeration, air conditioning, and other fields that require metal bonding. In the field of electrical engineering, the most commonly used copper-aluminum welding technology is brazing technology, especially soft brazing using silver-based filler metals. This method is widely used because of its excellent electrical conductivity, good mechanical strength, and low processing temperature.

[0003] In the existing welding technology, there are many parameters, such as current, voltage, welding speed, flow rate of shielding gas, etc. However, even for professional technicians, relying on manual setting of the same parameters may result in slight deviations, reducing the repeatability of the welding process and affecting the quality of welded products. Moreover, during the welding process, it is necessary to adjust the welding parameters according to the changes in the brazing material and the base material. If only relying on manual adjustment of parameters, it is difficult to achieve precise control of welding.

[0004] Therefore, there is an urgent need to provide a welding control method for copper and aluminum base materials that can precisely control and adaptively adjust welding parameters. Summary of the Invention

[0005] The present invention aims to provide a welding control method for copper and aluminum base materials that can precisely control and adaptively adjust welding parameters, so as to improve the precise control of welding parameters and thus improve the quality of welded products.

[0006] A welding control method for copper and aluminum base materials includes:

[0007] S1. Acquisition and processing of weld images:

[0008] Acquire the weld images of the copper base material and the aluminum base material to obtain the copper weld image to be processed and the aluminum weld image to be processed; perform edge detection on the copper weld image to be processed and the aluminum weld image to be processed to obtain the copper weld edge image and the aluminum weld edge image; based on the copper weld edge image, the aluminum weld edge image, and the brazing material selection model, determine the amount of brazing material used and the initial welding parameters; perform the welding work of the copper and aluminum base materials according to the initial welding parameters; the welding parameters include laser power, focal position, laser beam diameter, flow rate of shielding gas, welding speed, and welding path.

[0009] S2. Dynamic welding control:

[0010] Obtain the welding environment temperature and the weld surface image at each time interval f; compare the welding environment temperature with the temperature warning range. If the welding environment temperature reaches the temperature warning range, obtain the first parameter adjustment strategy; otherwise, continue welding; perform feature extraction based on the weld surface image to obtain the brazing contact angle eigenvalue, match the corresponding wetting degree based on the brazing contact angle eigenvalue. If the wetting degree is lower than the set wetting degree, output the corresponding second parameter adjustment strategy in the wetting degree improvement library based on the wetting degree; perform max pooling on the first parameter adjustment strategy and the second parameter adjustment strategy to obtain the welding parameter adjustment strategy, and control the welding process based on the welding parameter adjustment strategy; the first parameter adjustment strategy and the second parameter adjustment strategy include adjustments to various welding parameters;

[0011] S3. Weld quality inspection and feedback:

[0012] Obtain the external image and the internal image of the weld joint; perform quality inspection on the weld joint based on the external image of the weld joint, the internal image of the weld joint, and the weld joint quality inspection model to obtain the weld joint quality inspection result; if the weld joint quality inspection result is defect-free, record the actual consumption of the brazing material and the welding parameter adjustment table, and optimize the brazing material selection model based on the actual consumption of the brazing material and the welding parameter adjustment table; otherwise, judge the welding problem based on the external image of the weld joint, the internal image of the weld joint, and the weld joint quality inspection result, and adjust the welding control method based on the welding problem.

[0013] As a preferred technical solution of the present invention, the brazing material selection model includes a feature recognition layer, a material consumption layer, a parameter output layer, and a result output layer;

[0014] The feature recognition layer is used to utilize stereo vision technology for the copper weld edge image and the aluminum weld edge image to construct a weld stereo model; perform volume feature extraction on the weld stereo model to obtain the weld volume eigenvalue;

[0015] The material consumption layer is used to output the corresponding brazing material consumption according to the weld volume eigenvalue;

[0016] The parameter output layer is used to match from the brazing material standard library according to the brazing material consumption to obtain the welding standard parameters corresponding to the brazing material consumption with the highest similarity, which are the initial welding parameters; the brazing material standard library contains the brazing material consumption and the corresponding welding standard parameters;

[0017] The result output layer is used to output the brazing material consumption and the initial welding parameters.

[0018] As a preferred technical solution of the present invention, the specific steps for constructing the brazing material selection model include:

[0019] Obtain several weld edge images with known amounts of brazing materials. Each weld edge image contains a copper weld edge image and an aluminum weld edge image; Combine several weld edge images to obtain a brazing material selection training set;

[0020] Select a CNN model as the initial model of the brazing material selection model. Input the brazing material selection training set into the initial model of the brazing material selection model and train the model with the amount of brazing material, and output the loss function value; Calculate the accuracy of the loss function value and the true value to obtain an accuracy evaluation result; If the accuracy evaluation result is within the error standard range, output the trained initial model of the brazing material selection model, which is the brazing material selection model; Otherwise, continue to train the model using the brazing material selection training set.

[0021] As a preferred technical solution of the present invention, the specific steps of step S2 include:

[0022] S21. Obtain the welding environment temperature and the weld surface image at every time interval f, compare the welding environment temperature with the temperature warning range. If the welding environment temperature reaches the temperature warning range, enter step S22; Otherwise, continue welding;

[0023] S22. Calculate the difference between the welding environment temperature and the welding standard temperature to obtain the welding temperature difference ΔT; Input the welding temperature difference ΔT into the welding thermodynamics model for digital twin simulation to obtain a first adjustment strategy; The welding standard temperature is the optimal temperature during welding.

[0024] S23. Perform enhanced feature extraction on the weld surface image to obtain the brazing contact angle feature value; Based on the brazing contact angle feature value, perform wetting degree matching to obtain the wetting degree; If the wetting degree is lower than the set wetting degree, output the corresponding second parameter adjustment strategy in the wetting degree improvement library based on the wetting degree;

[0025] S24. Perform max pooling on the first parameter adjustment strategy and the second parameter adjustment strategy, and select the one with the largest adjustment value for each parameter to form a welding parameter adjustment strategy;

[0026] S25. Based on the welding parameter adjustment strategy, perform real-time control on the welding instrument.

[0027] As a preferred technical solution of the present invention, the specific steps of performing enhanced feature extraction on the weld surface image in step S23 include:

[0028] Denoise the weld surface image to obtain a denoised weld surface image;

[0029] Perform grayscale processing on the denoised weld surface image and enhance the edges to obtain an edge-enhanced weld surface image;

[0030] Extract the contour of the surface image of the reinforced edge weld to obtain the brazing contact angle contour image;

[0031] Analyze the characteristics of the brazing contact angle contour image to obtain the brazing contact angle characteristic value.

[0032] As a preferred technical solution of the present invention, the specific steps for quality inspection of the weld joint in step S3 include:

[0033] T1. Obtain the external image and internal image of the weld joint;

[0034] T2. Input the external image and internal image of the weld joint into the weld joint quality inspection model for quality inspection to obtain the weld joint quality inspection result;

[0035] T3. If the weld joint quality inspection result is qualified, record the actual amount of brazing material used and the welding parameter adjustment table, and enter step T4; otherwise, enter step T5;

[0036] T4. Take the average value of each parameter value in the welding parameter adjustment table to obtain the welding standard parameters, and input the actual amount of brazing material used and the corresponding welding standard parameters into the brazing material standard library;

[0037] T5. Based on the weld joint quality inspection result output by the weld joint quality inspection model, judge the welding problem and adjust the welding control method according to the welding problem.

[0038] As a preferred technical solution of the present invention, the weld joint quality inspection model includes an image classification layer, an external image detection layer, an internal image detection layer, and a result output layer;

[0039] The image classification layer is used to classify the input image, input the external image of the weld joint into the external image detection layer, and input the internal image of the weld joint into the internal image detection layer;

[0040] The external image detection layer is used to detect defects in the external image of the weld joint and output the external defect detection result;

[0041] The internal image detection layer is used to detect defects in the internal image of the weld joint and output the internal defect detection result;

[0042] The result output layer is used to output the weld joint quality inspection result. If both the external defect detection result and the internal defect detection result are defect-free, the weld joint quality inspection result is qualified; otherwise, the weld joint quality inspection result is unqualified, and the reason for the weld joint quality inspection is output.

[0043] As a preferred technical solution of the present invention, the specific steps for defect detection of the external image and internal image of the weld joint include:

[0044] In the external image detection layer, a weld joint external quality inspection model is deployed; the external image of the weld joint is input into the weld joint external quality inspection model for image analysis to obtain the external defect detection result;

[0045] Using the YOLOv5 model as the initial model of the weld joint external quality inspection model, m external weld joint defect images and external weld joint defect normal images are obtained, and the weld joint defect images are classified and labeled for defects, and the external weld joint defect normal images are labeled as non-defective; q external weld joint normal images are processed by data augmentation methods to obtain q new external weld joint defect images; r external weld joint defect images are processed by data augmentation methods to obtain p new external weld joint defect images; the p + q new external weld joint defect images are combined with m external weld joint defect images and external weld joint defect normal images to obtain an external weld joint defect training set, and the YOLOv5 model is trained using the external weld joint defect training set to obtain the weld joint external quality inspection model;

[0046] In the internal image detection layer, a weld joint internal quality inspection model is deployed, the internal image of the weld joint is obtained, the internal image of the weld joint is processed for image annotation, and the image in the annotation box is extracted to obtain the weld joint internal image feature image; the image annotation target is the darker or brighter area in the internal image of the weld joint; the weld joint internal image feature image is input into the weld joint internal quality inspection model for image analysis to obtain the internal defect detection result;

[0047] Using the YOLOv5 model as the initial model of the weld joint internal quality inspection model, g weld joint internal image feature images are obtained, and the internal weld joint internal image feature images are classified and labeled for defects, and the g weld joint internal image feature images are processed by feature augmentation methods to obtain h new weld joint internal image feature images; the h new weld joint internal image feature images and the g weld joint internal image feature images are combined to obtain an internal weld joint defect training set, and the YOLOv5 model is trained using the internal weld joint defect training set to obtain the weld joint internal quality inspection model.

[0048] The present invention has the following advantages:

[0049] 1. Through the accurate extraction of the volume characteristics of the weld three-dimensional model, the material usage layer of the present invention can accurately output the corresponding brazing material usage. This accuracy helps reduce material waste and ensures an economical and efficient welding process. The parameter output layer determines the initial welding parameters based on the weld volume characteristic values and the brazing material usage, enabling the welding process to be more precisely adjusted according to the specific welding parts and requirements. This not only improves the welding quality but also makes the welding process more controllable. By precise material selection and parameter setting, the quality of the weld joint can be improved, the probability of defects and failures can be reduced, and the reliability of welding can be enhanced. By adjusting the welding parameters in real time according to the environmental temperature and the brazing contact angle characteristic values, it helps modify the welding strategy when the temperature reaches the warning range and reduces the potential damage of excessive environmental temperature to the equipment and welding results. By adjusting the parameters to achieve the preset wetting degree, the welding quality can be improved. Through the max pooling of the first parameter adjustment strategy and the second parameter adjustment strategy, the most comprehensive and optimal welding parameter adjustment strategy can be obtained, realizing the comprehensive monitoring and optimization of the welding environment.

[0050] 2. By performing quality inspections on the external and internal images of the weld joint, the present invention can immediately detect potential problems during the welding process and thus correct them in a timely manner. By learning from the error situations and failure experiences, the welding quality can be continuously improved. Record the actual amount of brazing material used and the welding parameter adjustment table. These data can drive the algorithm to optimize the brazing material selection model, and for the welding parameters that have an important impact on measuring the welding quality, fine-tuning can be carried out with the help of these data, thereby optimizing the welding process. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic flowchart of a welding control method for copper and aluminum base materials adopted in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0053] Embodiment:

[0054] The copper and aluminum base material welding method usually adopts brazing technology, and the brazing material is usually silver-based filler metal. Taking the welding control method under the brazing technology using silver-based filler metal as an example, a copper and aluminum base material welding control method is shown in Figure 1 as follows, including:

[0055] S1. Weld image acquisition and processing:

[0056] Obtain the weld images of the copper substrate and the aluminum substrate to get the copper weld image to be processed and the aluminum weld image to be processed; perform edge detection on the copper weld image to be processed and the aluminum weld image to be processed to obtain the copper weld edge image and the aluminum weld edge image; determine the amount of brazing material and the initial welding parameters based on the copper weld edge image, the aluminum weld edge image and the brazing material selection model; perform the welding work of the copper and aluminum substrates according to the initial welding parameters; the welding parameters include laser power, focal position, laser beam diameter, shielding gas flow rate, welding speed and welding path;

[0057] A camera is installed in the welding furnace, and the camera is used to capture the welding process; the edge detection algorithm is the Canny edge detection algorithm. The Canny edge detection algorithm first uses a Gaussian filter to reduce image noise, then uses a multi-directional gradient operator to determine the intensity and direction of the edges, and finally applies non-maximum suppression and hysteresis thresholding techniques to determine the positions of the edges. Using the Canny edge detection algorithm can effectively eliminate the influence of other irrelevant factors in the welding furnace on image recognition;

[0058] Use laser welding to weld the silver-based brazing material. Laser welding can very precisely control the size, shape and energy of the solder joints, which makes it very ideal for processing small or precision components. Laser welding can achieve targeted local heating and will not cause excessive thermal effects on the surrounding copper and aluminum substrates, thereby protecting the original properties of the welded substrates; set the initial welding parameters for laser welding. The output power of the laser directly affects the welding speed and the penetration depth; the focal position of the laser on the material surface has a significant impact on the weld quality and the penetration depth; the speed at which the laser head moves, the common unit is mm / s; the welding speed needs to be reasonably set according to the melting point and thermal conductivity of the material to obtain an ideal welding effect; during laser welding, an inert gas is usually used to protect the welding area, prevent oxidation and assist heat transfer. The flow rate of the shielding gas is usually measured in L / min, and the specific value depends on the size and shape of the weld; for complex welding tasks, the planning of the movement of the laser welding head along a specific path is also one of the key parameters; the penetration depth refers to the melting depth caused by the welding heat source to the welded metal during the welding process, and it is an important indicator to measure the strength of the welding penetration;

[0059] The brazing material selection model includes a feature recognition layer, a material consumption layer, a parameter output layer and a result output layer;

[0060] The feature recognition layer is used to use stereo vision technology on the copper weld edge image and the aluminum weld edge image to construct a weld stereo model; extract the volume features of the weld stereo model to obtain the weld volume feature values;

[0061] The material consumption layer is used to output the corresponding amount of brazing material according to the weld volume feature values;

[0062] The parameter output layer is used to match according to the brazing material consumption from the brazing material standard library, and obtain the welding standard parameters corresponding to the brazing material consumption with the highest similarity, which are the initial welding parameters; the brazing material standard library contains the brazing material consumption and the corresponding welding standard parameters.

[0063] The result output layer is used to output the brazing material consumption and the initial welding parameters.

[0064] During the brazing process, using too much brazing material may cause accumulation and waste, while too little may cause poor bonding or holes. Through the accurate extraction of the volume characteristics of the weld three-dimensional model, the material consumption layer can accurately output the corresponding brazing material consumption. This accuracy helps reduce material waste and ensure an economical and efficient welding process; the parameter output layer determines the initial welding parameters based on the weld volume characteristic value and the brazing material consumption, enabling the welding process to be more precisely adjusted according to the specific welding parts and requirements. This not only improves the welding quality but also makes the welding process more controllable.

[0065] The brazing material standard library is established by professional technicians according to the actual historical brazing material consumption and the welding parameter adjustment table.

[0066] The specific steps to construct the brazing material selection model include:

[0067] Obtain several weld edge images with known brazing material consumption, each of which contains a copper weld edge image and an aluminum weld edge image; combine several weld edge images to obtain the brazing material selection training set.

[0068] Select the CNN model as the initial model of the brazing material selection model, input the brazing material selection training set into the initial model of the brazing material selection model for model training with the brazing material consumption, and output the loss function value; calculate the accuracy rate of the loss function value and the true value to obtain the accuracy evaluation result; if the accuracy evaluation result is within the error standard range, output the trained initial model of the brazing material selection model, which is the brazing material selection model; otherwise, continue to perform model training using the brazing material selection training set.

[0069] Through precise material selection and parameter setting, the quality of the weld joint can be improved, the probability of defects and failures can be reduced, and the reliability of welding can be enhanced; through precise material consumption evaluation, the waste of brazing materials can be reduced, and the production cost can be lowered. At the same time, high-quality welding reduces the occurrence of rework and scrap; the consistency of the model output helps standardize the production process, which is the key to improving the consistency of weld joints and reducing operation complexity.

[0070] S2. Dynamic Welding Control:

[0071] Obtain the welding environment temperature and the weld surface image at every time interval f; compare the welding environment temperature with the temperature warning range. If the welding environment temperature reaches the temperature warning range, obtain the first parameter adjustment strategy; otherwise, continue welding; perform feature extraction based on the weld surface image to obtain the brazing contact angle feature value, match the corresponding wetting degree based on the brazing contact angle feature value. If the wetting degree is lower than the set wetting degree, output the corresponding second parameter adjustment strategy in the wetting degree improvement library based on the wetting degree; perform max pooling on the first parameter adjustment strategy and the second parameter adjustment strategy to obtain the welding parameter adjustment strategy, and control the welding process based on the welding parameter adjustment strategy; the first parameter adjustment strategy and the second parameter adjustment strategy include adjustments to various welding parameters.

[0072] The time interval f is set by professionals according to the actual situation.

[0073] The welding environment temperature is obtained by a temperature sensor. The temperature warning range is set based on the welding threshold temperature, and the welding threshold temperature is set by the melting points of the base material and the brazing material. The ambient temperature needs to be lower than the melting points of all base materials. Otherwise, if the ambient temperature is higher than the melting point of the base material, the base material will be melted to generate oxides, and the oxides adhering to the weld will affect the strength of the weld joint, and too high ambient temperature will affect the quality of the weld joint; the temperature warning range is set by professional technicians based on the actual situation.

[0074] The specific steps of step S2 include:

[0075] S21. Obtain the welding environment temperature and the weld surface image at every time interval f, compare the welding environment temperature with the temperature warning range. If the welding environment temperature reaches the temperature warning range, then enter step S22; otherwise, continue welding.

[0076] S22. Calculate the difference between the welding environment temperature and the welding standard temperature to obtain the welding temperature difference ΔT; input the welding temperature difference ΔT into the welding thermodynamics model for digital twin simulation to obtain the first adjustment strategy; the welding standard temperature is the optimal temperature during welding; the first adjustment strategy includes adjustments to various welding parameters, such as strategies to change the laser power, change the welding speed, and increase the shielding gas flow rate.

[0077] The welding thermodynamics model is set by professional technicians. The welding thermodynamics model can describe the parameter relationships such as temperature, material properties, heat transfer, heat consumption, and time. Input the welding temperature difference ΔT into the welding thermodynamics model for digital twin simulation, and calculate with the goal of making the welding temperature difference ΔT infinitely approach 0 to obtain the first adjustment strategy.

[0078] S23. Extract the enhanced features of the weld surface image to obtain the brazing contact angle eigenvalue; perform wetness matching based on the brazing contact angle eigenvalue to obtain the wetting degree; if the wetting degree is lower than the set wetting degree, output the corresponding second parameter adjustment strategy in the wetting degree improvement library based on the wetting degree;

[0079] The wetting degree improvement library contains the current wetting degree and the corresponding second parameter adjustment strategy for improving the wetting degree. The second parameter adjustment strategy is set by professional technicians based on multiple welding experiments; the second parameter adjustment strategy includes adjustments to various welding parameters, such as strategies for changing the focus position, changing the diameter of the laser beam, etc.; the set wetting degree is set by professional technicians based on practical experience;

[0080] The contact angle is the physical angle formed by a liquid droplet contacting a solid surface, specifically the angle between the edge of the liquid droplet formed by the liquid on the solid surface and the solid surface; having a small contact angle means that the liquid wets the solid better, while a large contact angle means a poor wetting degree. In welding, if the contact angle of the filler metal on the base material is small, then the wettability is good, and vice versa, a large contact angle means poor wettability; for liquid solder, a high wetting degree means it can better diffuse and spread on the base material surface to form a uniform weld;

[0081] S24. Perform max pooling on the first parameter adjustment strategy and the second parameter adjustment strategy, and for each parameter, select the one with the largest adjustment value to form the welding parameter adjustment strategy;

[0082] S25. Based on the welding parameter adjustment strategy, perform real-time control on the welding instrument;

[0083] The specific steps for extracting the enhanced features of the weld surface image in step S23 include:

[0084] Denoise the weld surface image to obtain the denoised weld surface image; the Gaussian filtering algorithm can be used to denoise the weld surface image. Denoising is to reduce the irrelevant information or background noise in the image, which can help to more clearly see the true shape of the weld;

[0085] Perform grayscale processing and edge enhancement on the denoised weld surface image to obtain the edge-enhanced weld surface image; common grayscale methods such as the average method, weighted method, etc.; then the weld area can be highlighted by enhancing the edges of the image; converting the color image to a grayscale image reduces the computational complexity, and at the same time, the grayscale image retains most of the important information reflecting the weld edges; enhancing the contrast between the weld and the background makes the recognition of the weld edges more obvious;

[0086] Extract the contour of the surface image of the reinforced edge weld to obtain the brazing contact angle contour image. The findContours function in the OpenCV library can be used to obtain the edge contour of the image, and the obtained contour can be used to generate an accurate representation of the weld form, which is the basis for calculating the weld characteristic values.

[0087] Conduct feature analysis on the brazing contact angle contour image to obtain the brazing contact angle characteristic values. Through fitting methods such as the least squares method, fit the contact angle and calculate the characteristic values of the brazing contact angle.

[0088] S3. Weld quality inspection and feedback:

[0089] Obtain the external image and internal image of the weld joint. Conduct quality inspection on the weld joint based on the external image of the weld joint, the internal image of the weld joint, and the weld joint quality inspection model to obtain the weld joint quality inspection result. If the weld joint quality inspection result is defect-free, record the actual consumption of brazing materials and the welding parameter adjustment table, and optimize the brazing material selection model based on the actual consumption of brazing materials and the welding parameter adjustment table. Otherwise, judge the welding problem based on the external image of the weld joint, the internal image of the weld joint, and the weld joint quality inspection result, and adjust the welding control method based on the welding problem.

[0090] The weld joint is the area that connects the copper and aluminum base materials through the welding process. By conducting quality inspection on the external and internal images of the weld joint, problems that may occur during the welding process can be immediately detected, and thus corrected in a timely manner. By learning from the wrong situations and failed experiences, the welding quality can be continuously improved. Record the actual amount of brazing materials used and the welding parameter adjustment table. These data can drive the algorithm to optimize the brazing material selection model. For the welding parameters that have an important impact on measuring the welding quality, fine-tuning can be carried out with the help of these data, thereby optimizing the welding process.

[0091] The specific steps for conducting quality inspection on the weld joint in step S3 include:

[0092] T1. Obtain the external image and internal image of the weld joint. The external image of the weld joint is taken by a high-definition camera, and the internal image of the weld joint is obtained using X-ray imaging.

[0093] T2. Input the external image and internal image of the weld joint into the weld joint quality inspection model for quality inspection to obtain the weld joint quality inspection result.

[0094] T3. If the weld joint quality inspection result is qualified, record the actual consumption of brazing materials and the welding parameter adjustment table, and enter step T4; otherwise, enter step T5. The welding parameter adjustment table contains the records of each adjustment of welding parameters during the welding process.

[0095] T4. Take the average value of each parameter value in the welding parameter adjustment table to obtain the welding standard parameters, and input the actual usage of the brazing material and the corresponding welding standard parameters into the brazing material standard library;

[0096] T5. Based on the weld joint quality inspection results output by the weld joint quality inspection model, judge the welding problems, and adjust the welding control method according to the welding problems; the weld joint quality inspection results include the defect causes of the weld. According to the weld defect causes, professional technicians can judge the welding problems and thus adjust the welding control method;

[0097] Common defect causes of weld joints include porosity, slag inclusion, incomplete penetration, lack of fusion, crack, cold lap, hardening in the heat affected zone, burn-through, slag splashing, sun scar and walkway, etc.;

[0098] The weld joint quality inspection model includes an image classification layer, an external image detection layer, an internal image detection layer and a result output layer;

[0099] The image classification layer is used to classify the input images, input the external images of the weld joints into the external image detection layer, and input the internal images of the weld joints into the internal image detection layer;

[0100] The external image detection layer is used to detect defects in the external images of the weld joints and output the external defect detection results;

[0101] The internal image detection layer is used to detect defects in the internal images of the weld joints and output the internal defect detection results;

[0102] The result output layer is used to output the weld joint quality inspection results. If both the external defect detection results and the internal defect detection results are defect-free, the weld joint quality inspection results are qualified; otherwise, the weld joint quality inspection results are unqualified, and the reasons for the weld joint quality inspection are output;

[0103] The specific steps for defect detection of the external images and internal images of the weld joints include:

[0104] In the external image detection layer, there is a weld joint external quality inspection model deployed; input the external images of the weld joints into the weld joint external quality inspection model for image analysis to obtain the external defect detection results;

[0105] Using the YOLOv5 model as the initial model for the external quality inspection model of weld joints, m external weld joint defect images and external weld joint defect normal images are obtained, and the weld joint defect images are classified and labeled for defects, while the external weld joint defect normal images are labeled as defect-free; q external weld joint normal images are processed using data augmentation methods to obtain q new external weld joint defect images; r external weld joint defect images are processed using data augmentation methods to obtain p new external weld joint defect images; the p + q new external weld joint defect images are combined with the m external weld joint defect images and external weld joint defect normal images to obtain an external weld joint defect training set, and the YOLOv5 model is trained using the external weld joint defect training set to obtain an external quality inspection model for weld joints;

[0106] In the internal image detection layer, an internal quality inspection model for weld joints is deployed. Weld joint internal images are obtained, image annotation processing is performed on the weld joint internal images, and the images in the annotation boxes are extracted to obtain weld joint internal image feature images; the image annotation target is the darker or brighter areas in the weld joint internal images; the weld joint internal image feature images are input into the internal quality inspection model for weld joints for image analysis to obtain internal defect detection results;

[0107] Using the YOLOv5 model as the initial model for the internal quality inspection model of weld joints, g weld joint internal image feature images are obtained, and the internal weld joint internal image feature images are classified and labeled for defects. The g weld joint internal image feature images are processed using feature augmentation methods to obtain h new weld joint internal image feature images; the h new weld joint internal image feature images and the g weld joint internal image feature images are combined to obtain an internal weld joint defect training set, and the YOLOv5 model is trained using the internal weld joint defect training set to obtain an internal quality inspection model for weld joints;

[0108] In the object detection model, a large number of image samples are required for model training. However, the number of normal images is much larger than that of defective images. Therefore, data augmentation methods need to be used to amplify the collected dataset and increase the diversity of weld defects. In the external quality inspection model of weld joints, the Copy-Pasting data augmentation method and the Mosaic+Mixup data augmentation method can be used to increase the defective images of external weld joints. The Copy-Pasting data augmentation method pastes the defective part with the annotation box into a normal image to artificially create a defective image. The Mosaic+Mixup data augmentation method first uses the Mosaic data augmentation method to segment and recombine a defective image and three additional defective images to obtain a new defective image. The Mixup data augmentation method adds two images output by the Mosaic data augmentation method in proportion to obtain a new defective image. In the internal quality inspection model of weld joints, contrast enhancement algorithms, edge enhancement algorithms, denoising algorithms, and sharpening algorithms can be used to enhance the feature details of the internal image features of weld joints. Through different data augmentation methods, the number of weld defect samples in the training set has been greatly increased, and the accuracy of object detection model training has been improved.

[0109] It should be understood that those of ordinary skill in the art can make improvements or transformations according to the above description, and all such improvements and transformations shall fall within the protection scope of the appended claims of the present invention. The parts not detailedly described in this specification belong to the prior art well-known to those of ordinary skill in the art.

Claims

1. A welding control method for copper and aluminum substrates, characterized in that Including: S1. Weld image acquisition and processing: Acquire the weld images of the copper substrate and the aluminum substrate to obtain the to-be-processed copper weld image and the to-be-processed aluminum weld image; perform edge detection on the to-be-processed copper weld image and the to-be-processed aluminum weld image to obtain the copper weld edge image and the aluminum weld edge image; determine the brazing material dosage and the initial welding parameters based on the copper weld edge image, the aluminum weld edge image and the brazing material selection model; perform the welding work of the copper and aluminum substrates according to the initial welding parameters; the welding parameters include laser power, focus position, laser beam diameter, shielding gas flow rate, welding speed and welding path. S2. Dynamic welding control: Obtain the welding environment temperature and the weld surface image at every time interval f; compare the welding environment temperature with the temperature warning range. If the welding environment temperature reaches the temperature warning range, obtain the first parameter adjustment strategy; otherwise, continue welding; perform feature extraction based on the weld surface image to obtain the brazing contact angle feature value, match the corresponding wetting degree based on the brazing contact angle feature value. If the wetting degree is lower than the set wetting degree, output the corresponding second parameter adjustment strategy in the wetting degree improvement library based on the wetting degree; perform max pooling on the first parameter adjustment strategy and the second parameter adjustment strategy to obtain the welding parameter adjustment strategy, and control the welding process based on the welding parameter adjustment strategy; the first parameter adjustment strategy and the second parameter adjustment strategy include the adjustment of each welding parameter. S3. Weld quality inspection and feedback: Obtain the external image and the internal image of the weld joint; perform quality inspection on the weld joint based on the external image of the weld joint, the internal image of the weld joint and the weld joint quality inspection model to obtain the weld joint quality inspection result; if the weld joint quality inspection result is defect-free, record the actual brazing material dosage and the welding parameter adjustment table, and optimize the brazing material selection model based on the actual brazing material dosage and the welding parameter adjustment table; otherwise, judge the welding problem based on the external image of the weld joint, the internal image of the weld joint and the weld joint quality inspection result, and adjust the welding control method based on the welding problem.

2. The welding control method for copper and aluminum substrates according to claim 1, characterized in that The brazing material selection model includes a feature recognition layer, a material dosage layer, a parameter output layer and a result output layer. The feature recognition layer is used to use stereo vision technology for the copper weld edge image and the aluminum weld edge image to construct a weld stereo model; perform volume feature extraction on the weld stereo model to obtain the weld volume feature value. The material dosage layer is used to output the corresponding brazing material dosage according to the weld volume feature value. The parameter output layer is used to match from the brazing material standard library according to the brazing material dosage to obtain the welding standard parameters corresponding to the brazing material dosage with the highest similarity, which are the initial welding parameters; the brazing material standard library contains the brazing material dosage and the corresponding welding standard parameters. The result output layer is used to output the brazing material dosage and the initial welding parameters.

3. A method for controlling the welding of copper and aluminum substrates according to claim 2, characterized in that, The specific steps for constructing the brazing material selection model include: Obtain several weld edge images with known brazing material dosages, and each weld edge image contains a copper weld edge image and an aluminum weld edge image; combine the several weld edge images to obtain the brazing material selection training set. Select the CNN model as the initial model of the brazing material selection model. Input the brazing material selection training set into the initial model of the brazing material selection model to train the model with the brazing material usage, and output the loss function value. Calculate the accuracy rate of the loss function value and the true value to obtain the accuracy evaluation result. If the accuracy evaluation result is within the error standard range, output the trained initial model of the brazing material selection model, which is the brazing material selection model; otherwise, continue to train the model using the brazing material selection training set.

4. A method for controlling the welding of copper and aluminum substrates according to claim 1, characterized in that, The specific steps of step S2 include: S21. Obtain the welding environment temperature and the weld surface image at every time interval f. Compare the welding environment temperature with the temperature warning range. If the welding environment temperature reaches the temperature warning range, go to step S22; otherwise, continue welding. S22. Calculate the difference between the welding environment temperature and the welding standard temperature to obtain the welding temperature difference ΔT. Input the welding temperature difference ΔT into the welding thermodynamics model for digital twin simulation to obtain the first adjustment strategy. The welding standard temperature is the optimal temperature during welding. S23. Perform enhanced feature extraction on the weld surface image to obtain the brazing contact angle feature value. Based on the brazing contact angle feature value, perform wetting degree matching to obtain the wetting degree. If the wetting degree is lower than the set wetting degree, output the corresponding second parameter adjustment strategy in the wetting degree improvement library based on the wetting degree. S24. Perform max pooling on the first parameter adjustment strategy and the second parameter adjustment strategy. For each parameter, select the one with the largest adjustment value to form the welding parameter adjustment strategy. S25. Based on the welding parameter adjustment strategy, perform real-time control on the welding instrument.

5. A method for controlling the welding of copper and aluminum substrates according to claim 4, characterized in that, The specific steps of performing enhanced feature extraction on the weld surface image in step S23 include: Denoise the weld surface image to obtain the denoised weld surface image. Perform grayscale processing and edge enhancement on the denoised weld surface image to obtain the weld surface image with enhanced edges. Extract the contour of the weld surface image with enhanced edges to obtain the brazing contact angle contour image. Perform feature analysis on the brazing contact angle contour image to obtain the brazing contact angle feature value.

6. A copper and aluminum substrate welding control method according to claim 1, characterized in that, The specific steps of inspecting the weld joint in step S3 include: T1. Obtain the external image and the internal image of the weld joint. T2. Input the external image and the internal image of the weld joint into the weld joint quality inspection model for quality inspection to obtain the weld joint quality inspection result. T3. If the weld joint quality inspection result is qualified, record the actual usage of the brazing material and the welding parameter adjustment table, and go to step T4; otherwise, go to step T5. T4. Take the average value of each parameter value in the welding parameter adjustment table to obtain the welding standard parameters. Input the actual usage of the brazing material and the corresponding welding standard parameters into the brazing material standard library. T5. Based on the weld joint quality inspection result output by the weld joint quality inspection model, judge the welding problem and adjust the welding control method according to the welding problem.

7. A method for controlling the welding of copper and aluminum substrates according to claim 6, characterized in that, The weld joint quality inspection model includes an image classification layer, an external image detection layer, an internal image detection layer, and a result output layer. The image classification layer is used to classify the input image, input the external image of the weld joint into the external image detection layer, and input the internal image of the weld joint into the internal image detection layer; The external image detection layer is used to detect defects in the external image of the weld joint and output the external defect detection result; The internal image detection layer is used to detect defects in the internal image of the weld joint and output the internal defect detection result; The result output layer is used to output the quality inspection result of the weld joint. If both the external defect detection result and the internal defect detection result are defect-free, the quality inspection result of the weld joint is qualified; otherwise, the quality inspection result of the weld joint is unqualified, and the reason for the quality inspection of the weld joint is output.

8. A welding control method for copper and aluminum substrates according to claim 7, characterized in that, The specific steps for defect detection of the external image and the internal image of the weld joint include: In the external image detection layer, an external quality inspection model for the weld joint is deployed; the external image of the weld joint is input into the external quality inspection model for the weld joint for image analysis to obtain the external defect detection result; Using the YOLOv5 model as the initial model of the external quality inspection model for the weld joint, m external weld joint defect images and external weld joint defect normal images are obtained, and the defect classification of the weld joint defect images is labeled, and the external weld joint defect normal images are labeled as defect-free; data augmentation methods are used for q external weld joint normal images to obtain q new external weld joint defect images; data augmentation methods are used for r external weld joint defect images to obtain p new external weld joint defect images; the p+q new external weld joint defect images are combined with the m external weld joint defect images and external weld joint defect normal images to obtain the external weld joint defect training set, and the YOLOv5 model is trained using the external weld joint defect training set to obtain the external quality inspection model for the weld joint; In the internal image detection layer, an internal quality inspection model for the weld joint is deployed, the internal image of the weld joint is obtained, the internal image of the weld joint is processed for image annotation, and the image in the annotation box is extracted to obtain the internal image feature image of the weld joint; the image annotation target is the darker or brighter area in the internal image of the weld joint; the internal image feature image of the weld joint is input into the internal quality inspection model for the weld joint for image analysis to obtain the internal defect detection result; Using the YOLOv5 model as the initial model of the internal quality inspection model for the weld joint, g internal image feature images of the weld joint are obtained, and the defect classification of the internal image feature images of the internal weld joint is labeled. Feature augmentation methods are used for the g internal image feature images of the weld joint to obtain h new internal image feature images of the weld joint; the h new internal image feature images of the weld joint and the g internal image feature images of the weld joint are combined to obtain the internal weld joint defect training set, and the YOLOv5 model is trained using the internal weld joint defect training set to obtain the internal quality inspection model for the weld joint.

Images