Welding quality detection method, device, electronic equipment and storage medium

Abstract

The present invention provides a welding quality detection method, device, electronic device, and storage medium, relating to the field of computer technology. The method comprises: obtaining multi-source data collected for a welding area during the welding process, the multi-source data including data collected by different types of sensors; inputting the multi-source data into a pre-trained welding quality detection model to obtain the probabilities of multiple welding defect categories; when the maximum probability exceeds a probability threshold, determining that the welding quality detection result of the welding area is abnormal welding quality, and using the welding defect category corresponding to the maximum probability as the defect type of the welding area; and determining the defect location corresponding to the defect type through bounding box regression or feature map coordinate mapping. The present invention uses artificial intelligence to analyze multi-source data, quickly and accurately identifying welding quality in real time, thereby improving welding quality, thereby reducing production costs and product rejection rates to meet large-scale, high-precision production requirements.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a welding quality detection method, device, electronic equipment and storage medium. Background Art

[0002] Currently, welding quality inspections during welding operations primarily rely on manual spot checks or offline testing. Manual spot checks rely on worker experience, are highly subjective, inefficient, and difficult to detect internal defects. Offline testing (such as destructive testing and ultrasonic testing) interrupts the welding process, preventing real-time inspection and potentially damaging the product. These methods result in low welding inspection efficiency, delays in timely error correction, and increased production costs.

[0003] To address the above issues, online welding quality detection methods have been proposed. However, current online welding quality detection methods all use a single sensor for detection, which cannot fully reflect the welding quality and is prone to missed detections. These shortcomings can lead to unstable welding quality, increased production costs and product rejection rates, making it difficult to meet large-scale, high-precision production needs. Summary of the Invention

[0004] Based on the above technical problems, the present invention provides a welding quality detection method, device, electronic device and storage medium, aiming to overcome the above problems or at least partially solve the above problems.

[0005] The present invention provides a welding quality detection method, which comprises:

[0006] Acquire multi-source data collected from the welding area during the welding process, wherein the multi-source data includes data collected by different types of sensors; the multi-source data includes: visual images and temperature distribution data;

[0007] Inputting the multi-source data into a pre-trained welding quality detection model to obtain the probabilities of multiple welding defect categories;

[0008] When the maximum probability exceeds the probability threshold, determining that the welding quality inspection result of the welding area is abnormal welding quality, and taking the welding defect category corresponding to the maximum probability as the defect type of the welding area;

[0009] Determine the defect location corresponding to the defect type through bounding box regression or feature map coordinate mapping;

[0010] Among them, the pre-trained welding quality detection model is trained based on the sample multi-source data collected for the sample welding area during the historical welding process and the real defect information of the sample welding area. The real defect information is the defect type and defect location manually marked for the sample welding area.

[0011] A second aspect of the present invention provides a welding quality detection system, the system comprising at least: different types of sensors and an embedded control system;

[0012] The different types of sensors are used to collect multi-source data for the welding area during the welding process and send the multi-source data to the embedded control system; the multi-source data includes: visual images and temperature distribution data;

[0013] The embedded control system is configured to acquire the multi-source data and input the multi-source data into a pre-trained welding quality detection model to obtain the probabilities of multiple welding defect categories; when the maximum probability exceeds a probability threshold, determine that the welding quality detection result of the welding area is abnormal welding quality, and use the welding defect category corresponding to the maximum probability as the defect type of the welding area; and determine the defect location corresponding to the defect type through bounding box regression or feature map coordinate mapping;

[0014] Among them, the pre-trained welding quality detection model is trained based on the sample multi-source data collected for the sample welding area during the historical welding process and the real defect information of the sample welding area. The real defect information is the defect type and defect location manually marked for the sample welding area.

[0015] The third aspect of the present invention provides an electronic device, which includes: a memory, a processor, and a computer program stored in the memory and runnable on the processor, wherein when the computer program is executed by the processor, the welding quality detection method according to the first aspect of the present invention is implemented.

[0016] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the welding quality detection method of the first aspect of the present invention is implemented.

[0017] In the welding quality detection method provided by the present invention, multi-source data is used to complement each other in multiple dimensions, which can reflect the welding quality in many aspects, avoid the limitations and missed detection of single sensor detection, and improve the accuracy of welding quality detection. In addition, the pre-trained welding quality detection model in the present invention learns a large amount of sample multi-source data and corresponding real defect information, and can quickly and accurately identify the defect type and defect location of the welding area, and then determine the welding inspection quality results of the welding area, thereby improving the welding quality, thereby reducing production costs and product rejection rates to meet large-scale, high-precision production needs. In addition, the present invention can detect welding quality online during the welding process, realize automated and intelligent detection, reduce the workload of manual detection, reduce labor costs, reduce the influence of human factors, and improve the reliability and consistency of welding quality detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0019] Figure 1 This is a flow chart of a welding quality detection method according to an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of signal processing and transmission of a visual sensor according to an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram of signal processing and transmission of an infrared thermal imager according to an embodiment of the present invention;

[0022] Figure 4 is a schematic diagram of an infrared thermal imager assembly and a visual sensor assembly according to an embodiment of the present invention;

[0023] Figure 5 is a schematic diagram of an embedded control system according to an embodiment of the present invention;

[0024] Figure 6 1 is a schematic diagram of a laser welding robot adapter module according to an embodiment of the present invention;

[0025] Figure 7 is a schematic diagram of a parameter adjustment process according to an embodiment of the present invention;

[0026] Figure 8 This is a schematic diagram of signal output between a host computer and an embedded control system according to an embodiment of the present invention;

[0027] Figure 9 is a schematic diagram of an online detection method for welding quality based on laser welding according to an embodiment of the present invention;

[0028] Figure 10 1 is a schematic diagram of an online detection method for welding quality based on laser welding according to an embodiment of the present invention;

[0029] Figure 11 This is a structural block diagram of a welding quality detection system provided by one embodiment of the present invention;

[0030] Figure 12 FIG. 1 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0032] Reference Figure 1 , Figure 1 FIG. 1 is a flow chart of a welding quality detection method according to an embodiment of the present invention. Figure 1 As shown, the welding quality detection method may include the following steps:

[0033] Step S11: Acquire multi-source data collected from the welding area during the welding process, wherein the multi-source data includes data collected by different types of sensors; the multi-source data includes visual images and temperature distribution data.

[0034] This embodiment can acquire multi-source data from the weld area during the welding process, where the multi-source data includes data collected by different types of sensors. This embodiment can collect multi-source data from the weld area during the welding process using multiple sensors of different types to monitor different dimensional data of the weld area in real time. The multi-source data in this embodiment includes at least visual images and temperature distribution data.

[0035] Step S12: inputting the multi-source data into a pre-trained welding quality detection model to obtain the probabilities of multiple welding defect categories.

[0036] In this embodiment, the acquired multi-source data can be input into a pre-trained welding quality detection model to obtain the probabilities of multiple welding defect categories corresponding to the welding area output by the pre-trained welding quality detection model. The pre-trained welding quality detection model of this embodiment is trained based on sample multi-source data collected for sample welding areas during historical welding processes and actual defect information of the sample welding areas. The pre-trained welding quality detection model can judge welding defects in the welding area and output the probabilities of multiple welding defect categories, that is, obtain the probability values ​​corresponding to each welding defect category in the multiple welding defect categories. For example, defect categories include, but are not limited to, porosity, cracks, lack of fusion, slag inclusions, undercuts, etc., and for example, the output probabilities of multiple welding defect categories include: a probability of 90% for porosity, a probability of 5% for cracks, and so on.

[0037] Among them, the sample welding area is the welding area targeted during the training process of the welding quality detection model, the sample multi-source data is the multi-source data corresponding to the sample welding area, such as the sample visual image and sample temperature distribution data corresponding to the sample welding area, and the real defect information is the defect type and defect location manually marked for the sample welding area.

[0038] Step S13: When the maximum probability exceeds the probability threshold, the welding quality inspection result of the welding area is determined to be abnormal welding quality, and the welding defect category corresponding to the maximum probability is used as the defect type of the welding area.

[0039] In this embodiment, after obtaining the probabilities of multiple welding defect categories, the pre-trained welding quality inspection model can determine the maximum probability and determine whether the maximum probability exceeds a probability threshold (which can be freely set in advance). If the maximum probability is determined to exceed the probability threshold, the welding quality inspection result for the weld area is determined to be abnormal welding quality, and the welding defect category corresponding to the maximum probability is used as the defect type for the weld area. For example, if the probability of a porosity defect (e.g., 90%) is the highest and exceeds a probability threshold (e.g., 70%), then the weld area is determined to have a porosity defect, the defect type for the weld area is determined to be a porosity defect, and the corresponding welding quality inspection result is output.

[0040] Step S14: Determine the defect position corresponding to the defect type through bounding box regression or feature map coordinate mapping.

[0041] In this embodiment, after determining the defect type of the weld area, the pre-trained welding quality inspection model can determine the defect location corresponding to the defect type in the weld area (e.g., the coordinates of a region of the weld) through bounding box regression (e.g., YOLOv5) or feature map coordinate mapping. In other words, if the pre-trained welding quality inspection model of this embodiment determines that the welding quality inspection result is abnormal, the pre-trained welding quality inspection model outputs a welding quality inspection result including at least the defect type and defect location.

[0042] In this embodiment, an artificial intelligence algorithm is introduced to analyze multi-sensor data fusion (visual images and temperature distribution data) to establish a correlation model (i.e., a pre-trained welding quality detection model) between welding quality and multi-source data (visual images and temperature distribution data). This enables real-time diagnosis of defect types and locations in weld areas. The use of multi-source data complements each other in multiple dimensions, reflecting welding quality in multiple aspects. This avoids the limitations and missed detections of single-sensor detection, thereby improving the accuracy of welding quality detection. Furthermore, the pre-trained welding quality detection model of the present invention learns from a large amount of sample multi-source data and the corresponding real-world defect information. It can quickly and accurately identify the defect type and location in the weld area, and then determine the welding quality detection results for the weld area, thereby improving welding quality, reducing production costs and product rejection rates to meet the needs of large-scale, high-precision production. Furthermore, the present invention enables online welding quality detection during the welding process, achieving automated and intelligent detection, reducing the workload and labor costs of manual inspections, and minimizing the impact of human factors, thereby improving the reliability and consistency of welding quality detection results. In addition, the multi-sensor fusion and artificial intelligence analysis used in this embodiment have strong compatibility and adaptability, and can be applied to different types of welding as well as different welding materials and processes, thereby realizing application in various welding scenarios.

[0043] In combination with the above embodiments, in one embodiment, the present invention further provides a welding quality detection method. In this method, in addition to the above steps, it also includes step S21:

[0044] Step S21: When the welding quality detection result indicates that the welding quality is abnormal, adjusting the welding parameters, and welding the welding area according to the adjusted welding parameters.

[0045] In this embodiment, the welding quality inspection results can be classified as abnormal welding quality or normal welding quality. If the pre-trained welding quality inspection model outputs an abnormal welding quality inspection result, the current welding parameters during the welding process can be adjusted to obtain adjusted welding parameters, and then welding can be performed on the welding area according to the adjusted welding parameters. In an optional embodiment, the welding parameters include at least one or more of the following: laser power, welding speed, and welding path.

[0046] In this embodiment, the welding quality during the welding process can be monitored in real time. Once abnormal welding quality is detected, the welding parameters will be automatically adjusted in real time to form a closed-loop control for welding, further ensuring the welding quality. The real-time monitoring and closed-loop control mechanism of this embodiment can make the welding process more stable and efficient, reducing production pauses and adjustment time caused by welding quality problems, thereby improving production efficiency.

[0047] In combination with the above embodiments, in one implementation, when the welding quality detection result output by the pre-trained welding quality detection model is that the welding quality is normal, welding can continue to be performed on the welding area according to the current welding parameters (ie, the original welding parameters).

[0048] In an optional embodiment, when the pre-trained welding quality inspection model determines that the maximum probability among the output probabilities of multiple welding defect categories does not exceed the probability threshold, the welding quality inspection result of the welding area can be directly determined to be normal welding quality, or, when it is determined that the maximum probability does not exceed the probability threshold, further analysis is performed to determine whether the welding quality inspection result is normal welding quality.

[0049] In combination with the above embodiments, in one embodiment, the present invention further provides a fatigue analysis method. In this method, the above step S12 may specifically include steps S31 to S34:

[0050] Step S31: Analyze the visual image to obtain molten pool morphology information, weld width information, and weld depth information.

[0051] In this embodiment, the visual image may be data collected from the weld area using a visual sensor (e.g., a high-precision visual sensor, such as a CCD camera), and the temperature distribution data may be data collected from the weld area using an infrared thermal imager. The weld area in this embodiment includes at least the molten pool. In this embodiment, the real-time collected visual image can be analyzed to obtain information about the molten pool topography, weld width, and weld depth.

[0052] Step S32: converting the temperature distribution data to obtain a thermal map.

[0053] In this embodiment, the temperature distribution data may be converted into an image format to obtain a thermal map.

[0054] Step S33: splicing the visual image carrying the molten pool morphology information, the weld width information and the weld depth information with the thermal map representing the temperature distribution data to obtain a spliced ​​image.

[0055] In this embodiment, the visual image carrying the molten pool morphology information, weld width information and weld depth information can be spliced ​​with the thermal map representing the temperature distribution data to obtain a spliced ​​image.

[0056] Step S34: inputting the stitched image into the pre-trained welding quality detection model to obtain the probabilities of the multiple welding defect categories.

[0057] In this embodiment, after the stitched image is obtained, the stitched image can be input into a pre-trained welding quality detection model to obtain the probabilities of multiple welding defect categories corresponding to the welding area output by the pre-trained welding quality detection model.

[0058] In this embodiment, a multi-sensor fusion method of visual sensors and infrared thermal imagers is adopted. The visual sensor can accurately capture the geometric characteristics of the weld, while the infrared thermal imager can accurately monitor the temperature changes in the welding area. The two types of data complement each other, avoiding the limitations of single sensor detection and greatly improving the accuracy of welding quality detection.

[0059] In combination with the above embodiments, the present invention further provides a welding quality detection method, in which the "obtaining multi-source data collected for the welding area during the welding process" in the above step S11 may specifically include step S41, and the "analyzing the visual image to obtain molten pool morphology information and weld depth information" in the above step S31 may specifically include steps S42 to S45:

[0060] Step S41: Projecting a light source onto the welding area to obtain a visual image collected for the welding area.

[0061] In this embodiment, during the welding process, a visual sensor can be combined with a light source, and the structured light principle can be used to project the light source onto the welding area. The visual image of the welding area can be collected by the visual sensor, and the visual image collected for the welding area can be obtained.

[0062] In one optional example, a high-speed camera can be combined with a light source of a specific wavelength to project the light onto the surface of the weld area. The distortion of the reflected light can then be analyzed to obtain relevant information about the weld area. For example, the light source of the specific wavelength can be light of a known structure, such as a laser. Semiconductor lasers are commonly used, and their output wavelengths are generally in the near-infrared band, such as 808nm and 980nm. Lasers in this wavelength band can both meet the requirements of structured light projection and interact well with the surface of the weld area. They also have a high degree of compatibility with the photosensitivity characteristics of the high-speed camera, facilitating the acquisition of clear visual images (i.e., reflected light images).

[0063] Step S42: Analyze the visual image to determine distortion information of the reflected light generated by the light source projected onto the welding area, where the distortion information at least includes geometric distortion information and grayscale distortion information.

[0064] The molten pool topography information of this embodiment includes at least: molten pool concavity and convexity information, depth information, width information, and surface physical information. This embodiment can analyze the acquired visual image to determine distortion information of the reflected light generated by the light source projecting onto the weld area. This reflected light distortion information includes at least: geometric distortion information and grayscale value distortion information.

[0065] Geometric distortion information refers to geometric shape distortion. When structured light is projected onto the surface of the melt pool, the three-dimensional undulations of the melt pool alter the propagation path of the reflected light in space, causing the originally regular structured light pattern (such as stripes or grids) to appear curved and distorted on the imaging plane. For example, the spacing between originally parallel stripes in the reflected light image will decrease in convex areas of the melt pool, resulting in an inward curvature. In concave areas, the spacing between stripes will increase, resulting in an outward curvature.

[0066] Grayscale value distortion information is information about grayscale value distortion. Grayscale value distortion is: physical properties of the molten pool surface (such as uneven temperature distribution, metal vapor, etc.) will affect the intensity of reflected light, thereby causing distortion of the grayscale value of the reflected light.

[0067] Step S43: Determine the three-dimensional morphology information of the molten pool based on the geometric distortion information.

[0068] In this embodiment, the three-dimensional morphology information of the molten pool can be determined based on the geometric distortion information.

[0069] Step S44: Based on the three-dimensional morphology information, determine the weld depth information, the concave-convex information and the depth information of the molten pool.

[0070] After determining the three-dimensional morphology information, this embodiment can determine the concavity and convexity and depth information of the molten pool based on the three-dimensional morphology information, and calculate the weld depth information based on the three-dimensional morphology information and combined with the triangulation principle.

[0071] Step S45: determining the surface physical information and width information of the molten pool based on the grayscale distortion information.

[0072] In this embodiment, the surface physical information and width information of the molten pool can be determined based on the obtained grayscale distortion information. In areas with higher molten pool temperatures, the metal's reflectivity may change, increasing the intensity of the reflected light and correspondingly increasing the grayscale value in the image. In areas obscured by metal vapor, the reflected light intensity decreases, the grayscale value decreases, and dark spots or shadows appear. Thus, the surface physical information of the molten pool can be determined based on the grayscale distortion information. The surface physical information of the molten pool in this embodiment refers to information about the physical properties of the molten pool surface, including but not limited to uneven temperature distribution and metal vapor.

[0073] Regarding the width information of the molten pool, this embodiment can be based on an edge detection algorithm, using the grayscale difference between the molten pool and the surrounding area in the visual image to determine the edge and thus calculate the width information of the molten pool.

[0074] In one embodiment, the photoelectric conversion device within the visual sensor converts the real-time collected optical signals into electrical signals (analog signals). After preliminary processing such as amplification and filtering, the signals are converted to digital signals via an analog-to-digital converter (ADC) for transmission. To address signal failure, in one embodiment, a redundant backup visual sensor can be employed. When the primary visual sensor fails, the system automatically switches to the backup. To address errors, in one embodiment, the visual sensor can be regularly calibrated to calibrate the optical system parameters. Filtering algorithms can also be used during the data processing phase to remove noise and outliers.

[0075] For example, photoelectric conversion devices such as complementary metal-oxide-semiconductors (CMOS) or charge-coupled devices (CCDs) can be used. For example, a CMOS image sensor contains a large number of pixel units, each of which has a photodiode. When reflected light strikes the photodiode, the photons interact with the semiconductor material, generating electron-hole pairs. Due to the photovoltaic effect, a potential difference forms across the photodiode's PN junction, converting the light signal into a corresponding charge signal. The amount of charge generated is determined by the intensity of the light.

[0076] like Figure 2 As shown, Figure 2 Figure 2 is a schematic diagram illustrating signal processing and transmission for a visual sensor according to an embodiment of the present invention. In this embodiment, the visual sensor is a CCD camera. During the welding process, the CCD camera collects optical signals from the weld area (such as those indicating the weld pool morphology, weld width, and weld depth). These signals are converted to digital signals through photoelectric conversion, analog signal processing, and ADC conversion. These digital signals are then transmitted to the data processing module of an embedded control system, which is used to monitor weld quality.

[0077] In one embodiment, an infrared thermal imager synchronously monitors the infrared radiation signal of the welding area. The infrared radiation signal is processed to form temperature distribution data (i.e., temperature distribution map) of the welding area. The signal processing process is: infrared radiation signal → analog electrical signal → digital signal → temperature distribution data.

[0078] like Figure 3 As shown, Figure 3 The figure below is a schematic diagram of signal processing and transmission for an infrared thermal imager, illustrating an embodiment of the present invention. The infrared thermal imager uses an infrared detector to convert infrared radiation signals from the weld area into electrical signals (analog signals). After signal processing (such as analog signal conditioning), the signals are converted into digital signals using an ADC and transmitted to the data processing module of an embedded control system. This embedded control system is used to monitor weld quality. The data processing module analyzes the visual images to obtain information about the weld pool morphology, weld width, and weld depth.

[0079] In combination with the above embodiments, the present invention further provides a welding quality detection method. In this method, the above step S34 may specifically include steps S51 to S54:

[0080] Step S51: extracting features of different dimensions from the spliced ​​image through the multi-layer convolutional layer to obtain a first image feature.

[0081] In this embodiment, the pre-trained welding quality inspection model is a convolutional neural network, comprising at least multiple convolutional layers, pooling layers, and fully connected layers. In this embodiment, after the stitched image is input into the pre-trained welding quality inspection model, the pre-trained welding quality inspection model first extracts features of different dimensions from the stitched image using multiple convolutional layers to obtain first image features. During this process, the weld pool morphology, weld width, and weld depth information carried by the visual image can be used as part of the visual features of the visual image. The geometric features extracted through the multiple convolutional layers ultimately produce the first image features output by the multiple convolutional layers.

[0082] Among them, the convolution layer extracts features from the stitched image to obtain image features by sliding and calculating the convolution kernel, and obtains a series of new values, which constitute a new feature map. Different convolution kernels can extract different features. For example, some convolution kernels can extract edge features in the image, and some can extract texture features, etc. This embodiment can extract more and more abstract and complex features from the original data by superimposing multiple layers of convolution layers, thereby obtaining the first image feature. For the stitched image (taking a two-dimensional image as an example), assuming that the size of the stitched image is H*W (height H and width W), and the size of the convolution kernel is h*w (height h and width w). At each slide, the convolution kernel performs a dot multiplication operation with the pixel value at the corresponding position on the stitched image, and then adds all the product results to obtain a new value. The calculation formula for the convolution kernel and its corresponding area can be: , where s represents the new feature value generated at a specific position after the convolution operation between the input image (i.e., the spliced ​​image) and the convolution kernel. It is a refinement and expression of the local features of the data. By calculating a series of s, new feature maps can be constructed. These feature maps retain the characteristics of the input image (melting pool shape, texture, etc.); To stitch images together The pixel value of the position, Is the convolution kernel in The weight value of the position.

[0083] In an optional embodiment, the visual image may be preprocessed to obtain a preprocessed visual image, and then the preprocessed visual image and the heat map may be spliced ​​to obtain a spliced ​​image. The preprocessing in this embodiment includes, but is not limited to, the following operations: denoising, grayscale conversion, rotation, scaling, and noise addition, thereby improving data quality and enhancing the generalization capability of the model.

[0084] Step S52: fusing the temperature distribution data, the molten pool morphology information, the weld width information, and the weld depth information with the first image feature to obtain a second image feature.

[0085] In this embodiment, after extracting the first image feature, the temperature distribution data, molten pool morphology information, weld width information and weld depth information can be fused with the first image feature to obtain the second image feature, thereby improving the sensitivity of the welding quality detection model to temperature-related defects, and judging the temperature abnormality area through the threshold to assist in identifying welding defects (such as pores caused by local overheating).

[0086] Step S53: performing dimensionality reduction processing on the second image features through the pooling layer to obtain welding data features.

[0087] In this embodiment, the second image feature is input into the pooling layer, and the second image feature is subjected to dimensionality reduction processing by the pooling layer to obtain the welding data feature. In an optional example, maximum pooling or average pooling can be used in the pooling layer for pooling processing. For example, average pooling or maximum pooling: let the input feature map (i.e., the second image feature) be F, and the size be , the pooling kernel size is , the step size is s; let the output feature map (i.e. welding data feature) be G, and the size be For the coordinates in the output feature map G The elements of are calculated as follows: ;

[0088] in, This represents the pixel values ​​offset (i, j) within the pooling kernel, starting from coordinates (m*s, n*s) on the input feature map F, with a step size of s. These pixel values ​​are calculated to obtain the element value at the corresponding position (m, n) in the output G. Taking the molten pool image as an example, features such as grayscale variations and texture details at the molten pool location are crucial for identifying welding defects. This offset fully considers the information carried by each pixel within the pooling kernel, such as the grayscale variation characteristics of pixels at the edge of the molten pool. This allows for the calculation of more representative output values, enabling the network to better capture local features of the welding process and improve the accuracy of welding defect identification.

[0089] Step S54: Processing the welding data features through the fully connected layer to obtain the probabilities of the multiple welding defect categories.

[0090] In this embodiment, the feature map generated by multiple convolutional and pooling layers is expanded into a one-dimensional vector, which contains the weld data features after layer-by-layer extraction and dimensionality reduction. This weld data feature is then fed into a fully connected layer, which processes the weld data features to obtain the probabilities of multiple weld defect categories corresponding to the weld area.

[0091] In combination with the above embodiments, the present invention further provides a welding quality detection method, in which the above step S54 may specifically include step S61:

[0092] Step S61: Mapping the welding data features to different welding defect categories through the fully connected layer to obtain the probabilities of the multiple welding defect categories.

[0093] In this embodiment, the fully connected layer can map the extracted features to welding defect categories. Specifically, each neuron in this embodiment's fully connected layer is connected to all neurons in the previous layer. These features are linearly transformed using a weight matrix, a bias term is added, and then processed using an activation function (such as the Softmax function). For the welding defect judgment task in this embodiment, the Softmax function can map the welding data features to different welding defect categories and output the probabilities of multiple welding defect categories, thereby obtaining the corresponding probability values ​​for each of the multiple welding defect categories.

[0094] In conjunction with the above embodiments, in an optional embodiment, the multi-source data includes visual images and temperature distribution data. When training a welding quality inspection model, the training samples are molten pool images captured during actual welding production (e.g., including normal welds and defective welds such as pores and cracks), temperature distribution data collected simultaneously with an infrared thermal imager, and manually annotated defect types (e.g., pores, cracks) and locations (e.g., weld region coordinates). Abnormal temperature regions can be annotated using threshold analysis. The base network used to train the welding quality inspection model can be ResNet-50 or YOLOv5, or other lightweight and improved classic convolutional neural networks. After obtaining the output of the base model, the output is compared with the manually annotated defect types and / or defect locations to perform a loss calculation. If the defect type is multi-classified (e.g., pores, cracks, lack of fusion, etc.), a cross-entropy loss function can be used. If the defect location is to be output, a mean squared error (MSE) loss (regression coordinates) can be combined. Finally, the classification and localization losses can be optimized simultaneously, for example, using a weighted combination loss function.

[0095] The trained welding quality inspection model takes as input the melt pool image (including geometric and grayscale distortion information). The temperature distribution data is converted into an image format (such as a heat map) and then concatenated with the melt pool image before being input into the welding quality inspection model. The trained welding quality inspection model outputs quality judgment, defect classification, and defect location. Defect classification uses a softmax function to output defect type probabilities (e.g., 90% probability of porosity, 5% probability of cracks, etc.). Defect localization uses bounding box regression (e.g., YOLOv5) or feature map coordinate mapping to output defect locations (e.g., coordinates of a specific weld region). Quality judgment uses a probability threshold to determine whether a defect exists and, if so, outputs its specific type and location.

[0096] In conjunction with the above embodiments, one implementation proposes an online weld quality detection method based on laser welding. This method utilizes a visual sensor (e.g., a high-precision visual sensor) and an infrared thermal imager to collect data on the weld pool morphology, weld width and depth, and temperature distribution in the weld area during the welding process, enabling multi-source data fusion monitoring. Artificial intelligence (i.e., a pre-trained welding quality detection model) is introduced to analyze the collected data, establishing a correlation model between weld quality and this data, enabling real-time diagnosis of welding defects. Based on the weld quality analysis results, an embedded control system adjusts the laser power, welding speed, or welding path in real time, forming a closed-loop control system to ensure weld quality.

[0097] In this embodiment, before performing online welding quality detection, the following preparations can be made:

[0098] Securely mount the visual sensor and infrared thermal imager on the end of the laser welding robot using a suitable mounting bracket, ensuring that the measurement field of view completely covers the welding area and adjusting the sensor's mounting angle and height so that it can accurately collect the required data. Use a data transmission line to connect the visual sensor and infrared thermal imager to the embedded control system to ensure stable data transmission. At the same time, establish communication connections between the embedded control system and the laser welding robot adapter module and the host computer. Also, set parameters on the visual sensor and infrared thermal imager, such as the resolution and frame rate of the visual sensor, and the temperature measurement range of the infrared thermal imager. In the embedded control system, set the relevant parameters of the welding quality detection model and the welding quality judgment threshold. In addition, the embedded control system has corresponding feedback panels for the visual sensor and infrared thermal imager.

[0099] like Figure 4 As shown, Figure 4 FIG is a schematic diagram of an infrared thermal imager assembly and a visual sensor assembly according to an embodiment of the present invention. Figure 4 In the figure, the upper part shows the infrared thermal imager assembly, which includes the infrared imager, mounting hardware, and calibration points. The lower part shows the high-precision vision sensor assembly, which includes the vision sensor, mounting bracket, and data transmission line, which connects to the embedded control system.

[0100] During welding, visual sensors and infrared thermal imagers can collect multi-source data in the welding area in real time and send the multi-source data to the embedded control system. Figure 5 As shown, Figure 5 FIG is a schematic diagram of an embedded control system according to an embodiment of the present invention. Figure 5In the embedded control system, the following components are included: a data processing module, an AI analysis module, a control instruction generation module, and a communication module. After the multi-source data is transmitted to the embedded control system, the visual image can be analyzed by the data processing module to obtain information such as the molten pool morphology, weld width and depth, and then the visual image and temperature distribution data carrying the molten pool morphology, weld width and depth, etc. are input into the AI ​​analysis module in the embedded control system. The AI ​​analysis module performs welding quality inspection through a pre-deployed welding quality inspection model to obtain welding quality inspection results. When the AI ​​analysis module determines that there is an abnormality in the welding quality, it sends a signal to the control instruction generation module. The control instruction generation module generates a control instruction based on the defect type and severity, and sends it to the laser welding robot adaptation module through the communication module.

[0101] like Figure 6 As shown, Figure 6 FIG. 1 is a schematic diagram of a laser welding robot adapter module according to an embodiment of the present invention. Figure 6 The laser welding robot adaptation module includes: an interface conversion device and a parameter adjustment actuator. The interface conversion device can convert the digital signal of the control instruction into a signal format suitable for the laser welding robot control system, and the parameter adjustment actuator adjusts the laser power, welding speed or path of the laser welding robot according to the adapted instruction. Figure 7 As shown, Figure 7 FIG. 1 is a flow chart showing a parameter adjustment process according to an embodiment of the present invention. Figure 7 In the embedded control system, after the control instruction generation module generates the control instruction, the control instruction is sent to the interface conversion device in the laser welding robot adaptation module through the communication module for signal format conversion to obtain the adapted control signal and send it to the parameter adjustment actuator in the laser welding robot adaptation module. The parameter adjustment actuator adjusts the laser power, welding speed or path of the laser welding robot according to the adapted control signal.

[0102] In one embodiment, the embedded control system can also transmit detection data and analysis results to a host computer in real time, allowing operators to monitor welding quality in real time through the host computer's monitoring and management software. The host computer stores and analyzes historical data, providing data support for subsequent welding process optimization. Furthermore, the operator can manually intervene in the welding process by sending control commands to the embedded control system through the host computer. During the welding process, the system continuously collects, analyzes, and adjusts data, forming a closed-loop control loop until welding is complete. Upon completion, the system stops and saves the relevant data from the welding session.

[0103] like Figure 8 As shown, Figure 8This is a schematic diagram of signal output between a host computer and an embedded control system according to an embodiment of the present invention. Figure 8 In the embedded control system, digital signals are transmitted between the communication module and the host computer.

[0104] In one embodiment, an online detection system is designed, which at least includes: a high-precision visual sensor component, an infrared thermal imager component, a host computer, an embedded control system, a laser welding robot adapter module and a laser welding robot. It can monitor the welding quality in real time during the laser welding process, and accurately identify welding defects (such as pores, cracks, etc.) even under complex welding conditions, thereby improving the stability and consistency of welding quality. Through AI algorithms and real-time monitoring mechanisms, the parameter adjustment during the welding process is optimized to ensure that laser welding can complete the welding task efficiently and with high quality under diverse conditions such as different welding materials and welding speeds. Figure 9 As shown, Figure 9 FIG is a schematic diagram of an online detection method for welding quality based on laser welding according to an embodiment of the present invention. Figure 9 During the laser welding process, the high-precision visual sensor component and the infrared thermal imager component can collect multi-source data for the welding area and send the multi-source data to the embedded control system. The embedded control system analyzes the multi-source data to obtain the welding quality detection results. If the welding quality detection results are abnormal, it sends control instructions to the laser welding robot adapter module to control the laser welding robot to adjust the welding parameters. The laser welding robot adapter module adjusts the laser power, welding speed or path of the laser welding robot based on the control instructions, and controls the laser welding robot to perform subsequent welding in the welding area according to the adjusted laser power, welding speed or path. The embedded control system can also transmit the detection data and analysis results to the host computer in real time, and the operator can grasp the welding quality in real time through the monitoring and management software of the host computer.

[0105] In combination with the above embodiments, an online detection method for welding quality based on laser welding is proposed in one embodiment, aiming to solve the problems that the existing welding quality detection relies on manual sampling or offline detection, resulting in low efficiency, inability to correct errors in time, and increased production costs. This embodiment combines multi-sensor fusion with AI algorithms to achieve real-time diagnosis and closed-loop control of welding quality, significantly improving the efficiency and accuracy of welding quality detection. Figure 10 As shown, Figure 10 FIG is a schematic diagram of an online detection method for welding quality based on laser welding according to an embodiment of the present invention. Figure 10 The method can be performed as follows:

[0106] Step S1: Start the system: turn on the laser welding robot and the welding quality online detection system to put all equipment into working state.

[0107] Step S2: Sensor data collection: The high-precision visual sensor collects the molten pool morphology, weld width and depth data, and the infrared thermal imager collects the temperature distribution data of the welding area.

[0108] Step S3: Data transmission to the embedded control system: The collected multi-source data is transmitted to the embedded control system via the data transmission line.

[0109] Step S4: Data preprocessing: The data processing module of the embedded control system performs preprocessing operations such as denoising and grayscale conversion on the data.

[0110] Step S5: AI algorithm analyzes data: The AI ​​analysis module uses a convolutional neural network algorithm (such as a pre-trained welding quality detection model) to analyze the pre-processed data to determine whether there are welding defects.

[0111] Step S6: Whether there are welding defects: Determine whether there are welding defects based on the AI ​​analysis results.

[0112] Step S7: Generate control instructions: If there is a welding defect, the control instruction generation module generates a control instruction.

[0113] Step S8: Adjust welding parameters: The laser welding robot adaptation module adjusts the laser power, welding speed or path according to the control instructions.

[0114] Step S9: Continue normal welding: If no welding defects are detected, the laser welding robot continues normal welding according to the original parameters.

[0115] Step S10: Continue welding and monitoring: continuously collect, analyze and adjust data during the welding process.

[0116] Step S11: Whether welding is completed: Determine whether welding is completed.

[0117] Step S12: End: After welding is completed, the system stops running.

[0118] The online welding quality detection method provided by this embodiment has at least the following advantages:

[0119] 1. Improve accuracy:

[0120] Multi-sensor fusion data collection: High-precision visual sensors and infrared thermal imagers are integrated at the end of the laser welding robot to synchronously collect multi-source data: Using a multi-sensor fusion method of high-precision visual sensors and infrared thermal imagers, the visual sensor can accurately capture the geometric features of the weld, while the infrared thermal imager can accurately monitor the temperature changes in the welding area. The two types of data complement each other to achieve multi-dimensional information acquisition of the welding process, avoiding the limitations of single sensor detection and greatly improving the accuracy of welding quality detection.

[0121] AI Algorithm Analysis: AI algorithms (such as convolutional neural networks) are used to deeply analyze multi-source data and establish a correlation model between welding quality and this data. AI algorithms can learn from large amounts of welding data and corresponding welding quality results, accurately identifying the type and location of welding defects (such as pores and cracks).

[0122] 2. Enhanced security:

[0123] Real-time monitoring and closed-loop control: The system can monitor the quality of the welding process in real time. Once abnormal welding quality is detected, the embedded control system will immediately and automatically adjust the laser power, welding speed or path of the laser welding robot to form a closed-loop control.

[0124] Improved protection mechanism: Through real-time monitoring of the welding process, the system can promptly detect abnormal conditions such as overvoltage and undervoltage, and take corresponding protective measures. In one optional embodiment, by analyzing the collected visual images and data, abnormal conditions such as overvoltage and undervoltage can be indirectly inferred. For example, the AI ​​algorithm indirectly identifies voltage fluctuations through abnormalities in the molten pool morphology and temperature data (for example, overvoltage may cause the laser power to be too high, causing excessive evaporation or splashing of the molten pool, and the visual sensor can detect increased distortion of the molten pool morphology; undervoltage may cause insufficient laser power and abnormal temperature distribution in the molten pool (such as temperature below the threshold), and the abnormal temperature area can be captured by the infrared thermal imager). In another optional embodiment, it is also possible to monitor the voltage in real time through the built-in voltage sensor of the embedded control system or to interact with the control system of the laser welding robot in real time to promptly detect abnormal conditions such as overvoltage and undervoltage.

[0125] 3. Improve reliability:

[0126] Reduce interference from human factors: Traditional manual sampling or offline testing methods are easily interfered with by human factors, while this patented online testing system realizes automated and intelligent testing, reduces the impact of human factors, and improves the reliability and consistency of test results.

[0127] Data Storage and Analysis: The host computer stores and analyzes test data and analysis results in real time, providing strong data support for subsequent welding process optimization. By analyzing historical data, potential problems and patterns in the welding process can be discovered, and welding process parameters can be adjusted in a timely manner to further improve the reliability of welding quality.

[0128] 4. Improve real-time performance:

[0129] Real-time data acquisition and processing: High-precision visual sensors and infrared thermal imagers collect real-time data from the welding process and quickly transmit it to the embedded control system via high-speed data transmission lines. The embedded control system processes and analyzes the data in real time, quickly determining any abnormalities in welding quality.

[0130] Real-time adjustment of welding parameters: When abnormal welding quality is detected, the embedded control system can generate control instructions in real time and adjust the parameters of the laser welding robot in real time through the laser welding robot adapter module to ensure stable and reliable welding quality.

[0131] 5. Cost savings:

[0132] Reduced manual inspection costs: Traditional manual spot checks or offline testing require significant manpower and time, resulting in high costs. This patented online inspection system, however, automates inspection, reducing the workload and labor costs. Furthermore, the system can detect and correct weld defects in real time, minimizing product scrap and rework costs.

[0133] Improve production efficiency: Real-time monitoring and closed-loop control mechanisms make the welding process more stable and efficient, reducing production stoppages and adjustment time caused by welding quality issues.

[0134] 6. Strong compatibility and adaptability:

[0135] Multi-scenario application: The multi-sensor fusion and AI algorithm technology used in the system has strong compatibility and adaptability, and can be applied to different types of welding as well as different welding materials and processes.

[0136] Scalability: The system's architectural design is highly scalable, allowing for easy addition of new sensors or optimization of AI algorithms to adapt to evolving welding technology and quality inspection requirements.

[0137] It should be noted that for the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0138] Based on the same inventive concept, an embodiment of the present invention provides a welding quality detection system. Figure 11 , Figure 11 FIG. 1 is a structural block diagram of a welding quality detection system provided by an embodiment of the present invention. Figure 11 As shown, the system includes at least: different types of sensors and embedded control systems;

[0139] The different types of sensors are used to collect multi-source data for the welding area during the welding process and send the multi-source data to the embedded control system; the multi-source data includes: visual images and temperature distribution data;

[0140] The embedded control system is configured to acquire the multi-source data and input the multi-source data into a pre-trained welding quality detection model to obtain the probabilities of multiple welding defect categories; when the maximum probability exceeds a probability threshold, determine that the welding quality detection result of the welding area is abnormal welding quality, and use the welding defect category corresponding to the maximum probability as the defect type of the welding area; and determine the defect location corresponding to the defect type through bounding box regression or feature map coordinate mapping;

[0141] Among them, the pre-trained welding quality detection model is trained based on the sample multi-source data collected for the sample welding area during the historical welding process and the real defect information of the sample welding area. The real defect information is the defect type and defect location manually marked for the sample welding area.

[0142] Optionally, the system further comprises: a welding robot adaptation module and a welding robot;

[0143] The embedded control system is further configured to generate a control instruction when the welding quality detection result indicates that the welding quality is abnormal, and send the control instruction to the welding robot adaptation module;

[0144] The welding robot adaptation module is configured to adjust welding parameters based on the control instructions and send the adjusted welding parameters to the welding robot;

[0145] The welding robot is used to weld the welding area according to the adjusted welding parameters.

[0146] Optionally, the embedded control system is specifically used to:

[0147] Analyzing the visual image to obtain molten pool morphology information, weld width information, and weld depth information;

[0148] converting the temperature distribution data to obtain a thermal map;

[0149] splicing the visual image carrying the molten pool morphology information, the weld width information, and the weld depth information with the thermal map representing the temperature distribution data to obtain a spliced ​​image;

[0150] The stitched image is input into the pre-trained welding quality detection model to obtain the probabilities of the multiple welding defect categories.

[0151] Optionally, the embedded control system is specifically used to:

[0152] Projecting a light source onto the welding area to obtain a visual image collected for the welding area; the molten pool morphology information includes at least: concave-convex information, depth information, width information and surface physical information of the molten pool;

[0153] Analyzing the visual image to determine distortion information of reflected light generated by the light source projected onto the welding area, wherein the distortion information of the reflected light includes at least geometric distortion information and grayscale distortion information;

[0154] Determining three-dimensional morphology information of the molten pool based on the geometric distortion information;

[0155] Determining the weld depth information, the concave-convex information, and the depth information of the molten pool based on the three-dimensional topography information;

[0156] Based on the grayscale value distortion information, surface physical information and width information of the molten pool are determined.

[0157] Optionally, the pre-trained welding quality detection model includes at least: multiple convolutional layers, pooling layers, and fully connected layers; the embedded control system is specifically used to:

[0158] Extracting features of different dimensions from the spliced ​​image through the multi-layer convolutional layer to obtain a first image feature;

[0159] fusing the temperature distribution data, the molten pool morphology information, the weld width information, and the weld depth information with the first image feature to obtain a second image feature;

[0160] Performing dimensionality reduction processing on the second image feature through the pooling layer to obtain welding data features;

[0161] The welding data features are processed by the fully connected layer to obtain the probabilities of the multiple welding defect categories.

[0162] Optionally, the embedded control system is specifically used to:

[0163] The welding data features are mapped to different welding defect categories through the fully connected layer to obtain the probabilities of the multiple welding defect categories.

[0164] Based on the same inventive concept, another embodiment of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps in the welding quality detection method described in any of the above embodiments of the present invention.

[0165] Based on the same inventive concept, another embodiment of the present invention provides an electronic device, such as Figure 12 shown. Figure 12 This is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps of the welding quality detection method according to any of the above embodiments of the present invention.

[0166] As for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0167] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0168] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, embodiments of the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0169] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the process in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0170] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0171] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0172] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0173] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.

[0174] The above is a detailed introduction to a welding quality detection method, device, electronic device and storage medium provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A welding quality detection method, characterized in that: The method comprises: Acquire multi-source data collected from the welding area during the welding process, wherein the multi-source data includes data collected by different types of sensors; the multi-source data includes: visual images and temperature distribution data; Analyzing the visual image to determine distortion information of reflected light generated by the light source projected onto the welding area, the distortion information of the reflected light including at least geometric distortion and grayscale distortion information, wherein the grayscale distortion is caused by uneven temperature distribution on the surface of the molten pool or metal vapor affecting the intensity of the reflected light, thereby causing distortion of the grayscale value of the reflected light; Based on the grayscale distortion information, surface physical information and width information of the molten pool are determined, wherein the surface physical information of the molten pool includes at least: uneven temperature distribution and metal vapor; wherein in areas with higher molten pool temperatures, the intensity of reflected light increases, and the grayscale value in the corresponding image increases; and in areas obscured by metal vapor, the intensity of reflected light decreases, and the grayscale value decreases; Inputting the multi-source data into a pre-trained welding quality detection model to obtain the probabilities of multiple welding defect categories; When the maximum probability exceeds the probability threshold, determining that the welding quality inspection result of the welding area is abnormal welding quality, and taking the welding defect category corresponding to the maximum probability as the defect type of the welding area; Determine the defect location corresponding to the defect type through bounding box regression or feature map coordinate mapping; Among them, the pre-trained welding quality detection model is trained based on the sample multi-source data collected for the sample welding area during the historical welding process and the real defect information of the sample welding area. The real defect information is the defect type and defect location manually marked for the sample welding area.

2. The welding quality detection method according to claim 1, characterized in that: The method further comprises: When the welding quality detection result shows that the welding quality is abnormal, the welding parameters are adjusted, and the welding area is welded according to the adjusted welding parameters.

3. The welding quality detection method according to claim 1 or 2, characterized in that: The multi-source data is input into a pre-trained welding quality inspection model to obtain the probabilities of multiple welding defect categories, including: Analyzing the visual image to obtain molten pool morphology information, weld width information, and weld depth information; converting the temperature distribution data to obtain a thermal map; splicing the visual image carrying the molten pool morphology information, the weld width information, and the weld depth information with the thermal map representing the temperature distribution data to obtain a spliced ​​image; The stitched image is input into the pre-trained welding quality detection model to obtain the probabilities of the multiple welding defect categories.

4. The welding quality detection method according to claim 3, characterized in that: Acquire multi-source data collected on the weld area during the welding process, including: Projecting a light source onto the welding area to obtain a visual image collected for the welding area; The molten pool morphology information includes at least: convex-concave information, depth information, width information and surface physical information of the molten pool; analyzing the visual image to obtain the molten pool morphology information and weld depth information includes: Analyzing the visual image to determine distortion information of reflected light generated by the light source projected onto the welding area, wherein the distortion information of the reflected light includes at least geometric distortion information and grayscale distortion information; Determining three-dimensional morphology information of the molten pool based on the geometric distortion information; Determining the weld depth information, the concave-convex information, and the depth information of the molten pool based on the three-dimensional topography information; Based on the grayscale value distortion information, surface physical information and width information of the molten pool are determined.

5. The welding quality detection method according to claim 3, characterized in that: The pre-trained welding quality detection model includes at least: multiple convolutional layers, pooling layers, and fully connected layers; the spliced ​​image is input into the pre-trained welding quality detection model to obtain the probabilities of the multiple welding defect categories, including: Extracting features of different dimensions from the spliced ​​image through the multi-layer convolutional layer to obtain a first image feature; fusing the temperature distribution data, the molten pool morphology information, the weld width information, and the weld depth information with the first image feature to obtain a second image feature; Performing dimensionality reduction processing on the second image feature through the pooling layer to obtain welding data features; The welding data features are processed by the fully connected layer to obtain the probabilities of the multiple welding defect categories.

6. The welding quality detection method according to claim 5, characterized in that: Processing the welding data features through the fully connected layer to obtain the probabilities of the multiple welding defect categories includes: The welding data features are mapped to different welding defect categories through the fully connected layer to obtain the probabilities of the multiple welding defect categories.

7. A welding quality detection system, characterized in that: The system includes at least: different types of sensors and an embedded control system; The different types of sensors are used to collect multi-source data for the welding area during the welding process and send the multi-source data to the embedded control system; the multi-source data includes: visual images and temperature distribution data; The embedded control system is used to analyze the visual image and determine distortion information of the reflected light generated by the light source projected onto the welding area, wherein the distortion information of the reflected light includes at least information of geometric distortion and grayscale value distortion, wherein the grayscale value distortion is caused by uneven temperature distribution on the surface of the molten pool or metal vapor affecting the intensity of the reflected light, thereby causing distortion of the grayscale value of the reflected light; Based on the grayscale distortion information, surface physical information and width information of the molten pool are determined, wherein the surface physical information of the molten pool includes at least: uneven temperature distribution and metal vapor, wherein in areas with higher molten pool temperatures, the intensity of reflected light increases, and the grayscale value in the corresponding image increases; in areas obscured by metal vapor, the intensity of reflected light decreases, and the grayscale value decreases; Acquire the multi-source data, and input the multi-source data into a pre-trained welding quality detection model to obtain the probabilities of multiple welding defect categories; When the maximum probability exceeds the probability threshold, determining that the welding quality inspection result of the welding area is abnormal welding quality, and taking the welding defect category corresponding to the maximum probability as the defect type of the welding area; Determine the defect location corresponding to the defect type through bounding box regression or feature map coordinate mapping; Among them, the pre-trained welding quality detection model is trained based on the sample multi-source data collected for the sample welding area during the historical welding process and the real defect information of the sample welding area. The real defect information is the defect type and defect location manually marked for the sample welding area.

8. The welding quality inspection system according to claim 7, characterized in that: The system further comprises: a welding robot adaptation module and a welding robot; The embedded control system is further configured to generate a control instruction when the welding quality detection result indicates that the welding quality is abnormal, and send the control instruction to the welding robot adaptation module; The welding robot adaptation module is configured to adjust welding parameters based on the control instructions and send the adjusted welding parameters to the welding robot; The welding robot is used to weld the welding area according to the adjusted welding parameters.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the computer program is executed by the processor, the welding quality detection method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the welding quality detection method according to any one of claims 1 to 6 is implemented.

Images