AI-based welding quality implementation feedback system

Abstract

The present invention discloses an AI-based welding quality implementation feedback system, which relates to the technical field of intelligent detection, including an image recognition module, an image matching module and a welding feedback module, which obtains the corresponding welding feature images in each welding sub-area, obtains a reference welding feature image that matches the welding feature image, sets the reference welding feature image as a predetermined feature image, calculates the similarity between the welding feature image and the predetermined feature image, divides the welding quality grade of the welding feature image according to the similarity, feeds back the welding quality grade to the user end, and generates decision information for visual display. The present invention detects the welding parameters that affect the welding quality, increases the comprehensiveness of quality detection, generates decision information corresponding to the quality status, helps the staff to avoid defects in time, and improves the foresight and scientificity of production.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent detection, and in particular to an AI-based welding quality implementation feedback system. Background Art

[0002] In recent years, AI technology has continued to develop and play an increasingly important role in improving welding quality, reducing production costs and realizing industrial intelligence. AI vision can accurately obtain key parameters in the welding process and improve the sensitivity and accuracy of monitoring.

[0003] At present, in a Chinese invention patent with publication number CN 115684169 A, a welding quality detection method and a welding quality detection device are disclosed. The processing device calculates a dyeing area percentage of the area of ​​the welding area in the detection image that is dyed by the dyeing ink to the area of ​​the welding area, and judges whether the dyeing area percentage is greater than a predetermined dyeing percentage; if it is determined that the dyeing area percentage is not greater than the predetermined dyeing percentage, it is determined that the welding quality of the position to be detected is good, and the detection result information is generated; if it is determined that the dyeing area percentage is greater than the predetermined dyeing percentage, it is determined that the welding quality of the position to be detected is poor, and the detection result information is generated. However, the related art does not detect welding parameters that affect welding quality, lacks comprehensiveness of quality detection, and does not generate decision information corresponding to the quality status, which is not conducive to the staff to avoid defects in time, and lacks foresight and scientificity of production. Summary of the invention

[0004] The technical problem solved by the present invention is that the related technology does not detect the welding parameters that affect the welding quality, lacks the comprehensiveness of quality detection, does not generate decision-making information corresponding to the quality status, is not conducive to the staff to avoid defects in time, and lacks the foresight and scientificity of production.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, a welding quality implementation feedback system based on AI includes an image recognition module, an image matching module and a welding feedback module;

[0006] The image recognition module is used to obtain the corresponding welding feature image in each welding sub-area;

[0007] The image matching module is used to obtain a reference welding feature image that matches the welding feature image, set the reference welding feature image as a predetermined feature image, calculate the similarity between the welding feature image and the predetermined feature image, and classify the welding quality level of the welding feature image according to the similarity;

[0008] The welding feedback module is used to feed back the welding quality level to the user end and generate decision information for visual display.

[0009] As a first quality level selection scheme of the AI-based welding quality feedback system of the present invention, the image recognition module is configured with a welding feature image generation strategy, and the welding feature image generation strategy includes:

[0010] The welding area is divided into a plurality of welding sub-areas, and an image acquisition device is used to acquire images of finished welding products in each welding sub-area, and the images of finished welding products in each welding sub-area are subjected to denoising and enhancement processing to obtain a first image set;

[0011] Collecting welding parameters in the welding sub-area based on the sensing acquisition device to generate a welding parameter set, wherein the welding parameters include welding current, welding speed, welding wire diameter and welding wire extension length;

[0012] The welding parameter set and the first image set are trained by a machine learning model to obtain a reference welding feature vector corresponding to each welding sub-area;

[0013] Analyze the reference welding feature vector to obtain reference defect types, and divide the images in the first image set according to the reference defect types to obtain welding feature images, wherein the reference defect types include cracks, pores, undercuts, and slag inclusions;

[0014] When the welding speed is greater than the first speed threshold or the welding wire extension length is greater than the first length threshold, the corresponding reference defect type is pores;

[0015] When the diameter of the welding wire is greater than the first diameter threshold or the diameter of the welding wire is less than the second diameter threshold, the corresponding reference defect type is undercut;

[0016] When the welding speed is less than a second speed threshold or the welding current is less than a second current threshold, the reference defect type is slag inclusion;

[0017] When the welding current is greater than the first current threshold, the reference defect type is a crack.

[0018] As a first quality grade selection scheme of the AI-based welding quality feedback system described in the present invention, the logic for obtaining the reference welding feature vector is as follows: calling the welding database, inputting the welding finished product model, obtaining historical welding parameters and historical welding finished product images, using the historical welding parameters and historical welding finished product images as sample sets, dividing the sample sets into training sets and test sets according to a certain ratio, constructing a machine learning model, inputting the training set into the machine learning model for training, and obtaining a reference welding feature vector, and repeatedly testing the welding feature vector based on the test set until the accuracy reaches a first threshold, and then outputting the reference welding feature vector.

[0019] As a first quality grade selection scheme of the AI-based welding quality implementation feedback system described in the present invention, wherein: the image matching module is configured with a welding quality grade division strategy, and the welding quality grade division strategy includes: acquiring a welding feature image and a reference welding feature image, setting the reference welding image as a predetermined feature image, calculating the similarity between the welding feature image and the predetermined feature image based on a cosine similarity algorithm, and dividing the welding quality grade of the welding feature image according to the similarity, and the welding quality grade includes a first quality grade, a second quality grade and a third quality grade.

[0020] As a first quality grade selection scheme of the AI-based welding quality feedback system of the present invention, the calculation logic of the welding quality grade is as follows: presetting the welding feature threshold of each welding sub-area respectively to generate a welding feature threshold set;

[0021] The welding feature image corresponding to the welding sub-region and the predetermined feature image are calculated by a cosine similarity algorithm to obtain a similarity. If the similarity is less than a welding feature threshold corresponding to the welding sub-region, the similarity between the welding feature image corresponding to the next welding sub-region and the predetermined feature image is calculated. If the similarities between the welding feature images corresponding to each welding sub-region and the predetermined feature image are all less than the welding feature threshold corresponding to each welding sub-region, the welding feature image corresponding to the current welding sub-region is updated to the predetermined feature image.

[0022] If the similarity is greater than or equal to the predetermined characteristic image, the predetermined characteristic image is set as the reference characteristic image of the welding characteristic image, and each reference image is sorted in descending order according to the value of the similarity corresponding to the reference image, and the reference characteristic image corresponding to the greatest similarity is set as the final predetermined characteristic image;

[0023] The similarity between the welding feature image corresponding to each welding sub-region and the final predetermined feature image is cyclically calculated, and the similarity between the welding feature image corresponding to each welding sub-region and the final predetermined feature image is compared with a first similarity threshold and a second similarity threshold respectively, so as to classify the welding quality level of the welding feature image.

[0024] As a first quality grade selection scheme of the AI-based welding quality implementation feedback system described in the present invention, wherein: the image matching module compares the similarity between the welding feature image corresponding to each welding sub-area and the final predetermined feature image with the first similarity threshold and the second similarity threshold respectively, and records the welding quality grade of the welding feature image corresponding to a similarity less than the first similarity threshold as the first quality grade, records the welding quality grade of the welding feature image corresponding to a similarity greater than or equal to the first similarity threshold and less than the second similarity threshold as the second quality grade, and records the welding quality grade of the welding feature image corresponding to a similarity greater than or equal to the first similarity threshold and less than the second similarity threshold as the third quality grade.

[0025] As a first quality grade selection scheme of the AI-based welding quality implementation feedback system described in the present invention, wherein: the welding feedback module includes a decision unit and a display unit, the decision unit generates decision information according to the welding quality grade, and transmits the decision information to the display unit, and the display unit is used to display the welding sub-area number, the welding quality grade corresponding to the welding sub-area number, and the decision information corresponding to the welding sub-area.

[0026] As a first quality level selection scheme of the AI-based welding quality feedback system of the present invention, the generation logic of the decision information is as follows: obtaining the welding quality level, extracting the welding sub-area corresponding to the welding quality level of the third quality level, obtaining the welding feature image, the final predetermined feature image and the welding parameters corresponding to the welding sub-area;

[0027] When the reference defect type corresponding to the welding characteristic image is a crack, the welding current is obtained, and when the welding current is greater than a first current threshold, the decision information is to reduce the welding current until it is less than the first current threshold;

[0028] When the reference defect type corresponding to the welding characteristic image is porosity, the welding current, welding speed and welding wire extension length are obtained; when the welding current is greater than a first current threshold, the decision information is to reduce the welding current until it is less than the first current threshold; when the welding speed is greater than a first speed threshold, the decision information is to reduce the welding speed until it is less than the first speed threshold; when the welding wire extension length is greater than a first length threshold, the decision information is to reduce the welding wire extension length until it is less than the first length threshold;

[0029] When the reference defect type corresponding to the welding feature image is undercut, the welding current and the welding wire diameter are obtained. When the welding current is greater than the first current threshold, the decision information is to reduce the welding current until it is less than the first current threshold. When the welding wire diameter is greater than the first diameter threshold or less than the second diameter threshold, the decision information is to replace the welding wire and select a welding wire with a diameter greater than or equal to the first diameter threshold and less than or equal to the second diameter threshold.

[0030] When the reference defect type corresponding to the welding feature image is slag inclusion, the welding speed and welding current are obtained. When the welding current is less than the second current threshold, the decision information is to increase the welding current until the welding current is greater than or equal to the second current threshold and less than or equal to the first current threshold. When the welding speed is less than the second speed threshold, the decision information is to increase the welding speed until the welding speed is greater than or equal to the second speed threshold and less than or equal to the first speed threshold.

[0031] In a second aspect, the present invention provides an electronic device comprising a memory, a processor and storage in the memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps in any of the methods described above are executed.

[0032] In a third aspect, the present invention provides a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps in any of the methods described above are executed.

[0033] The beneficial effects of the present invention are as follows: by partitioning the welding area and obtaining the corresponding welding feature images in each welding sub-area, the welding quality detection is made clearer; by obtaining a reference welding feature image that matches the welding feature image, the similarity between the welding feature image and the reference welding feature image is calculated and the welding quality grade is divided; by comprehensively judging the welding parameters, the reference defect type and the similarity, the welding quality grade is obtained, so that the quality detection is more comprehensive; by feeding back the welding quality grade to the user end and generating decision information, the quality detection is made more practical, which is convenient for the staff to avoid erroneous production in advance, is beneficial to the efficiency and economy of the welding work, and is beneficial to energy saving and environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A basic flow chart of an AI-based welding quality feedback system provided for one embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to make the above-mentioned objects, features and first quality grade points of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings in the specification. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of them.

[0036] Example 1, reference Figure 1 , as an embodiment of the present invention, provides an AI-based welding quality implementation feedback system, including an image recognition module, an image matching module and a welding feedback module;

[0037] The image recognition module is used to obtain the corresponding welding feature image in each welding sub-area;

[0038] The image matching module is used to obtain a reference welding feature image that matches the welding feature image, set the reference welding feature image as a predetermined feature image, calculate the similarity between the welding feature image and the predetermined feature image, and classify the welding quality level of the welding feature image according to the similarity;

[0039] The welding feedback module is used to feed back the welding quality level to the user end and generate decision information for visual display.

[0040] The present invention partitions the welding area and obtains the corresponding welding feature images in each welding sub-area, so that the welding quality detection is clearer, and by obtaining a reference welding feature image that matches the welding feature image, the similarity between the welding feature image and the reference welding feature image is calculated and the welding quality grade is divided, and the welding quality grade is obtained through comprehensive judgment of welding parameters, reference defect types and similarities, so that the quality detection is more comprehensive, and by feeding back the welding quality grade to the user end and generating decision information, the quality detection is more practical, which is convenient for staff to avoid erroneous production in advance, is beneficial to the efficiency and economy of welding work, and is beneficial to energy saving and environmental protection.

[0041] The image recognition module is configured with a welding feature image generation strategy, and the welding feature image generation strategy includes:

[0042] The welding area is divided into a plurality of welding sub-areas, and an image acquisition device is used to acquire images of finished welding products in each welding sub-area, and the images of finished welding products in each welding sub-area are subjected to denoising and enhancement processing to obtain a first image set;

[0043] Collecting welding parameters in the welding sub-area based on the sensing acquisition device to generate a welding parameter set, wherein the welding parameters include welding current, welding speed, welding wire diameter and welding wire extension length;

[0044] The welding parameter set and the first image set are trained by a machine learning model to obtain a reference welding feature vector corresponding to each welding sub-area;

[0045] Analyze the reference welding feature vector to obtain reference defect types, and divide the images in the first image set according to the reference defect types to obtain welding feature images, wherein the reference defect types include cracks, pores, undercuts, and slag inclusions;

[0046] When the welding speed is greater than the first speed threshold or the welding wire extension length is greater than the first length threshold, the corresponding reference defect type is pores;

[0047] When the diameter of the welding wire is greater than the first diameter threshold or the diameter of the welding wire is less than the second diameter threshold, the corresponding reference defect type is undercut;

[0048] When the welding speed is less than a second speed threshold or the welding current is less than a second current threshold, the reference defect type is slag inclusion;

[0049] When the welding current is greater than the first current threshold, the reference defect type is a crack.

[0050] In a specific implementation, the welding area is divided into zones and the corresponding welding feature images in each welding sub-area are obtained, so that the welding quality detection is clearer.

[0051] The logic for obtaining the reference welding feature vector is as follows: calling a welding database, inputting a welding product model, obtaining historical welding parameters and historical welding product images, taking the historical welding parameters and historical welding product images as a sample set, dividing the sample set into a training set and a test set according to a certain ratio, building a machine learning model, inputting the training set into the machine learning model for training, and obtaining a reference welding feature vector, and repeatedly testing the welding feature vector based on the test set until the accuracy reaches a first threshold, and then outputting the reference welding feature vector.

[0052] The image matching module is configured with a welding quality grade classification strategy, which includes: acquiring a welding feature image and a reference welding feature image, setting the reference welding image as a predetermined feature image, calculating the similarity between the welding feature image and the predetermined feature image based on a cosine similarity algorithm, and dividing the welding quality grades of the welding feature image according to the similarity, wherein the welding quality grades include a first quality grade, a second quality grade, and a third quality grade.

[0053] The calculation logic of the welding quality grade is: presetting the welding feature threshold of each welding sub-area respectively to generate a welding feature threshold set;

[0054] The welding feature image corresponding to the welding sub-region and the predetermined feature image are calculated by a cosine similarity algorithm to obtain a similarity. If the similarity is less than a welding feature threshold corresponding to the welding sub-region, the similarity between the welding feature image corresponding to the next welding sub-region and the predetermined feature image is calculated. If the similarities between the welding feature images corresponding to each welding sub-region and the predetermined feature image are all less than the welding feature threshold corresponding to each welding sub-region, the welding feature image corresponding to the current welding sub-region is updated to the predetermined feature image.

[0055] In the specific implementation, the calculation expression of similarity is:

[0056]

[0057] in, is the similarity between the welding feature image corresponding to the welding sub-region and the predetermined feature image, is the number of feature vectors in the welding feature image, is the number of the feature vector in the welding feature image and the predetermined feature image, is the first feature vector in the welding feature image feature vectors, is the first feature vectors.

[0058] If the similarity is greater than or equal to the predetermined characteristic image, the predetermined characteristic image is set as the reference characteristic image of the welding characteristic image, and each reference image is sorted in descending order according to the value of the similarity corresponding to the reference image, and the reference characteristic image corresponding to the greatest similarity is set as the final predetermined characteristic image;

[0059] The similarity between the welding feature image corresponding to each welding sub-region and the final predetermined feature image is cyclically calculated, and the similarity between the welding feature image corresponding to each welding sub-region and the final predetermined feature image is compared with a first similarity threshold and a second similarity threshold respectively, so as to classify the welding quality level of the welding feature image.

[0060] The image matching module compares the similarity between the welding feature images corresponding to each welding sub-area and the final predetermined feature image with a first similarity threshold and a second similarity threshold, respectively, and records the welding quality level of the welding feature image corresponding to a similarity less than the first similarity threshold as a first quality level, records the welding quality level of the welding feature image corresponding to a similarity greater than or equal to the first similarity threshold and less than the second similarity threshold as a second quality level, and records the welding quality level of the welding feature image corresponding to a similarity greater than or equal to the second similarity threshold as a third quality level.

[0061] In the specific implementation, by obtaining a reference welding feature image that matches the welding feature image, calculating the similarity between the welding feature image and the reference welding feature image and dividing the welding quality grade, the welding quality grade is obtained through a comprehensive judgment of the welding parameters, reference defect type and similarity, making the quality inspection more comprehensive.

[0062] The welding feedback module includes a decision unit and a display unit. The decision unit generates decision information according to the welding quality level and transmits the decision information to the display unit. The display unit is used to display the welding sub-area number, the welding quality level corresponding to the welding sub-area number, and the decision information corresponding to the welding sub-area.

[0063] The generation logic of the decision information is: obtaining the welding quality level, extracting the welding sub-area corresponding to the welding quality level of the third quality level, and obtaining the welding characteristic image, the final predetermined characteristic image and the welding parameters corresponding to the welding sub-area;

[0064] When the reference defect type corresponding to the welding characteristic image is a crack, the welding current is obtained, and when the welding current is greater than a first current threshold, the decision information is to reduce the welding current until it is less than the first current threshold;

[0065] When the reference defect type corresponding to the welding characteristic image is porosity, the welding current, welding speed and welding wire extension length are obtained; when the welding current is greater than a first current threshold, the decision information is to reduce the welding current until it is less than the first current threshold; when the welding speed is greater than a first speed threshold, the decision information is to reduce the welding speed until it is less than the first speed threshold; when the welding wire extension length is greater than a first length threshold, the decision information is to reduce the welding wire extension length until it is less than the first length threshold;

[0066] When the reference defect type corresponding to the welding feature image is undercut, the welding current and the welding wire diameter are obtained. When the welding current is greater than the first current threshold, the decision information is to reduce the welding current until it is less than the first current threshold. When the welding wire diameter is greater than the first diameter threshold or less than the second diameter threshold, the decision information is to replace the welding wire and select a welding wire with a diameter greater than or equal to the first diameter threshold and less than or equal to the second diameter threshold.

[0067] When the reference defect type corresponding to the welding feature image is slag inclusion, the welding speed and welding current are obtained. When the welding current is less than the second current threshold, the decision information is to increase the welding current until the welding current is greater than or equal to the second current threshold and less than or equal to the first current threshold. When the welding speed is less than the second speed threshold, the decision information is to increase the welding speed until the welding speed is greater than or equal to the second speed threshold and less than or equal to the first speed threshold.

[0068] In specific implementation, by feeding back the welding quality level to the user end and generating decision information, quality inspection is made more practical, which is convenient for staff to avoid erroneous production in advance, is beneficial to the efficiency and economy of welding work, and is beneficial to energy conservation and environmental protection.

[0069] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. Among them, the storage medium may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, magnetic disk or optical disk. These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0070] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. AI-based welding quality implementation feedback system, characterized by: It includes an image recognition module, an image matching module and a welding feedback module; The image recognition module is used to obtain the corresponding welding feature image in each welding sub-area; The image matching module is used to obtain a reference welding feature image that matches the welding feature image, set the reference welding feature image as a predetermined feature image, calculate the similarity between the welding feature image and the predetermined feature image, and classify the welding quality level of the welding feature image according to the similarity; The welding feedback module is used to feed back the welding quality level to the user end and generate decision information for visual display; The image recognition module is configured with a welding feature image generation strategy, and the welding feature image generation strategy includes: The welding area is divided into a plurality of welding sub-areas, and an image acquisition device is used to acquire images of finished welding products in each welding sub-area, and the images of finished welding products in each welding sub-area are subjected to denoising and enhancement processing to obtain a first image set; Collecting welding parameters in the welding sub-area based on the sensing acquisition device to generate a welding parameter set, wherein the welding parameters include welding current, welding speed, welding wire diameter and welding wire extension length; The welding parameter set and the first image set are trained by a machine learning model to obtain a reference welding feature vector corresponding to each welding sub-area; Analyze the reference welding feature vector to obtain reference defect types, and divide the images in the first image set according to the reference defect types to obtain welding feature images, wherein the reference defect types include cracks, pores, undercuts, and slag inclusions; When the welding speed is greater than the first speed threshold or the welding wire extension length is greater than the first length threshold, the corresponding reference defect type is pores; When the diameter of the welding wire is greater than the first diameter threshold or the diameter of the welding wire is less than the second diameter threshold, the corresponding reference defect type is undercut; When the welding speed is less than a second speed threshold or the welding current is less than a second current threshold, the reference defect type is slag inclusion; When the welding current is greater than a first current threshold, the reference defect type is a crack; The image matching module is configured with a welding quality grade classification strategy, which includes: acquiring a welding feature image and a reference welding feature image, setting the reference welding image as a predetermined feature image, calculating the similarity between the welding feature image and the predetermined feature image based on a cosine similarity algorithm, and classifying the welding quality grades of the welding feature image according to the similarity, wherein the welding quality grades include a first quality grade, a second quality grade, and a third quality grade; The calculation logic of the welding quality grade is: presetting the welding feature threshold of each welding sub-area respectively to generate a welding feature threshold set; The welding feature image corresponding to the welding sub-region and the predetermined feature image are calculated by a cosine similarity algorithm to obtain a similarity. If the similarity is less than a welding feature threshold corresponding to the welding sub-region, the similarity between the welding feature image corresponding to the next welding sub-region and the predetermined feature image is calculated. If the similarities between the welding feature images corresponding to each welding sub-region and the predetermined feature image are all less than the welding feature threshold corresponding to each welding sub-region, the welding feature image corresponding to the current welding sub-region is updated to the predetermined feature image. If the similarity is greater than or equal to the predetermined characteristic image, the predetermined characteristic image is set as the reference characteristic image of the welding characteristic image, and each reference image is sorted in descending order according to the value of the similarity corresponding to the reference image, and the reference characteristic image corresponding to the greatest similarity is set as the final predetermined characteristic image; Circularly calculating the similarity between the welding feature image corresponding to each welding sub-region and the final predetermined feature image, comparing the similarity between the welding feature image corresponding to each welding sub-region and the final predetermined feature image with a first similarity threshold and a second similarity threshold, respectively, and classifying the welding quality level of the welding feature image; The generation logic of the decision information is: obtaining the welding quality level, extracting the welding sub-area corresponding to the welding quality level of the third quality level, and obtaining the welding characteristic image, the final predetermined characteristic image and the welding parameters corresponding to the welding sub-area; When the reference defect type corresponding to the welding characteristic image is a crack, the welding current is obtained, and when the welding current is greater than a first current threshold, the decision information is to reduce the welding current until it is less than the first current threshold; When the reference defect type corresponding to the welding characteristic image is porosity, the welding current, welding speed and welding wire extension length are obtained; when the welding current is greater than a first current threshold, the decision information is to reduce the welding current until it is less than the first current threshold; when the welding speed is greater than a first speed threshold, the decision information is to reduce the welding speed until it is less than the first speed threshold; when the welding wire extension length is greater than a first length threshold, the decision information is to reduce the welding wire extension length until it is less than the first length threshold; When the reference defect type corresponding to the welding feature image is undercut, the welding current and the welding wire diameter are obtained. When the welding current is greater than the first current threshold, the decision information is to reduce the welding current until it is less than the first current threshold. When the welding wire diameter is greater than the first diameter threshold or less than the second diameter threshold, the decision information is to replace the welding wire and select a welding wire with a diameter greater than or equal to the first diameter threshold and less than or equal to the second diameter threshold. When the reference defect type corresponding to the welding feature image is slag inclusion, the welding speed and welding current are obtained. When the welding current is less than the second current threshold, the decision information is to increase the welding current until the welding current is greater than or equal to the second current threshold and less than or equal to the first current threshold. When the welding speed is less than the second speed threshold, the decision information is to increase the welding speed until the welding speed is greater than or equal to the second speed threshold and less than or equal to the first speed threshold.

2. The AI-based welding quality feedback system according to claim 1, characterized in that: The logic for obtaining the reference welding feature vector is as follows: calling a welding database, inputting a welding product model, obtaining historical welding parameters and historical welding product images, taking the historical welding parameters and historical welding product images as a sample set, dividing the sample set into a training set and a test set according to a certain ratio, building a machine learning model, inputting the training set into the machine learning model for training, and obtaining a reference welding feature vector, and repeatedly testing the welding feature vector based on the test set until the accuracy reaches a first threshold, and then outputting the reference welding feature vector.

3. The AI-based welding quality feedback system according to claim 1, characterized in that: The image matching module compares the similarity between the welding feature images corresponding to each welding sub-area and the final predetermined feature image with a first similarity threshold and a second similarity threshold, respectively, and records the welding quality level of the welding feature image corresponding to a similarity less than the first similarity threshold as a first quality level, records the welding quality level of the welding feature image corresponding to a similarity greater than or equal to the first similarity threshold and less than the second similarity threshold as a second quality level, and records the welding quality level of the welding feature image corresponding to a similarity greater than or equal to the second similarity threshold as a third quality level.

4. The AI-based welding quality feedback system according to claim 1, characterized in that: The welding feedback module includes a decision unit and a display unit. The decision unit generates decision information according to the welding quality level and transmits the decision information to the display unit. The display unit is used to display the welding sub-area number, the welding quality level corresponding to the welding sub-area number, and the decision information corresponding to the welding sub-area.

5. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores a computer program that can be called by the processor, and the processor executes the AI-based welding quality implementation feedback system as described in any one of claims 1 to 4 by calling the computer program stored in the memory.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the computer executes the AI-based welding quality feedback system as described in any one of claims 1-4.

Images