Weld seam defect inspection method, apparatus and system, and medium

Abstract

A weld seam defect inspection method, a weld seam defect inspection apparatus, a weld seam defect inspection system, and a computer-readable storage medium. The weld seam defect inspection method comprises: acquiring a defect picture, which includes laser marks, of a weld seam defect to be inspected; and on the basis of the defect picture including the laser marks, using a first model to determine the type and size of said weld seam defect.

Description

Weld defect detection method, device, system and medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This disclosure is based on and claims priority to the Chinese patent application with application number 202411571516.1 and application date November 6, 2024. The disclosure of the Chinese patent application is hereby introduced as a whole into this application. Technical Field

[0003] The present disclosure relates to the technical field of weld appearance detection, and in particular to a weld defect detection method, a weld defect detection device, a weld defect detection system, and a computer-readable storage medium. Background Art

[0004] Companies involved in steel structure welding and fabrication must prioritize weld appearance inspection technology to provide the most fundamental guarantee for the overall performance and safety of welded structures. Within the realm of welding inspection technology, four primary welding inspection methods exist: UT (ultrasonic testing), RT (radiographic testing), MT (magnetic particle testing), and VT (visual testing). VT boasts the widest application, most direct application, and the best results. Establishing a welding quality system based on ISO 3834-2 generally requires at least four certified VT inspectors, double the number required for other inspection methods.

[0005] The existing methods for visual weld defect detection in the related art face numerous challenges in practical application. For example, traditional weld defect detection relies primarily on visual inspection by inspectors, which can be susceptible to subjective factors. Different inspectors may discrepancies in identifying defects in the same weld due to differences in experience, vision, and attention span, leading to missed or false detections. Manual weld defect detection is time-consuming, especially in large-scale production or welding projects, where the sheer number of welds to be inspected makes manual inspection far less efficient than meeting production schedules. Inspectors performing weld defect inspections require specialized training, including but not limited to specialized knowledge in welding processes, materials science, defect identification, and the operation of inspection tools. This training is both time-consuming and costly. Furthermore, due to a shortage of qualified personnel, companies face significant challenges in recruiting and retaining certified inspectors. These issues limit the efficiency and effectiveness of weld quality inspections. Summary of the Invention

[0006] In view of this, the embodiments of the present disclosure provide a weld defect detection method, a weld defect detection device, a weld defect detection system and a computer-readable storage medium, which help to improve the efficiency of weld appearance detection.

[0007] In one aspect of the present disclosure, a method for detecting weld defects is provided, comprising:

[0008] Obtain a laser-marked defect image of the weld defect to be detected;

[0009] Based on the laser-marked defect image, the first model is used to determine the type and size of the weld defect to be detected.

[0010] In some embodiments, the weld defect detection method further includes:

[0011] Obtaining a first image sample, a type of weld defect in the first image sample, and reference information for calculating a size of the weld defect in the first image sample;

[0012] Calculating reference information using the first image sample, the type of weld defect in the first image sample, and the size of the weld defect in the first image sample to train a first model;

[0013] The first image sample includes a weld defect image with laser marking.

[0014] In some embodiments, obtaining the first image sample, the type of weld defect in the first image sample, and reference information for calculating the size of the weld defect in the first image sample includes:

[0015] Obtain a second image sample;

[0016] Adding a plurality of parallel and equidistant laser scale lines to the second image sample to form a first image sample;

[0017] The endpoint of each laser scale line, the type of weld defect in the first image sample, and the endpoint of the weld defect are marked in the first image sample.

[0018] In some embodiments, obtaining the first image sample, the type of weld defect in the first image sample, and reference information for calculating the size of the weld defect in the first image sample includes:

[0019] Obtaining the pixel spacing between two adjacent laser scale lines and the pixel spacing between two end points of each weld defect in the first image sample;

[0020] The actual size of the weld defect is obtained based on the numerical relationship between the pixel spacing between two adjacent laser scale lines in the first image sample and the pixel spacing between the two end points of each weld defect and the actual spacing between the two adjacent laser scale lines.

[0021] In some embodiments, the operation of obtaining the second image sample specifically includes:

[0022] Obtain an initial image sample of weld defects;

[0023] Generate a second image sample using the second model based on the initial image sample;

[0024] The number of samples of the initial image samples of weld defects is less than the number of samples of the second image samples.

[0025] In another aspect of the present disclosure, a weld defect detection device is provided, comprising:

[0026] Memory; and

[0027] A processor coupled to the memory executes the weld defect detection method of any one of the above embodiments.

[0028] In another aspect of the present disclosure, a weld defect detection system is provided, comprising:

[0029] Mounting body;

[0030] A detection component is provided on the mounting body, the detection component comprising: a laser marking emitter and an image acquisition unit, the laser marking emitter being configured to emit a laser marking around the weld defect, and the image acquisition unit being configured to acquire an image of the weld defect with the laser marking;

[0031] As in the weld defect detection device of the above embodiment, the processor is signal-connected to the image acquisition unit and the laser marker emitter;

[0032] The laser marking includes a plurality of parallel and equidistant laser scale lines.

[0033] In some embodiments, the detection assembly is rotatable relative to the mounting body;

[0034] The detection components also include:

[0035] Laser rangefinders are symmetrically arranged on both sides of the laser marking transmitter and are configured to respectively obtain distances to weld defects;

[0036] The processor is connected to the laser rangefinder signal and is configured to adjust the angle of the detection component relative to the mounting body according to the distance information respectively obtained by the two laser rangefinders, so that the laser marking emitter emits a vertical laser mark around the weld defect.

[0037] In another aspect of the present disclosure, a weld defect detection device is provided, comprising:

[0038] An image acquisition module is configured to acquire an image of a defect with a laser mark on the weld to be inspected;

[0039] The defect detection module is configured to determine the type and size of the weld to be detected using the first model based on the defect image with the laser mark.

[0040] In another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the weld defect detection method of any of the above embodiments is implemented.

[0041] Therefore, according to the embodiment of the present disclosure, the first model can identify information such as the type and size of weld appearance defects based on photos of weld defects, thereby improving the efficiency of weld appearance detection, realizing the automation and intelligence of the entire process of weld appearance detection, and helping to enhance the accuracy of weld appearance detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0043] FIG1 is a flow chart of some embodiments of a weld defect detection method according to the present disclosure;

[0044] FIG2 is a schematic diagram of a defect image with laser marking according to some embodiments of the weld defect detection method disclosed herein;

[0045] FIG3 is a schematic diagram of obtaining the type and size of a weld defect based on a weld defect image according to some embodiments of the weld defect detection method disclosed herein;

[0046] FIG4 is a schematic structural diagram of some embodiments of a weld defect detection system according to the present disclosure;

[0047] FIG5 is a schematic diagram showing the connection relationship of some embodiments of the weld defect detection system according to the present disclosure.

[0048] The reference numerals represent: 1. processor; 2. mounting body; 3. detection component; 31. laser marker transmitter; 32. image acquisition unit; 33. laser rangefinder. DETAILED DESCRIPTION

[0049] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. The description of the exemplary embodiments is merely illustrative and is in no way intended to limit the present disclosure, its application, or use. The present disclosure can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to make the present disclosure thorough and complete and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that unless otherwise specifically stated, the relative arrangement of parts and steps, the composition of materials, numerical expressions, and numerical values ​​set forth in these embodiments should be interpreted as being merely exemplary and not as limiting.

[0050] The terms "first", "second" and similar terms used in this disclosure do not indicate any order, quantity or importance, but are only used to distinguish different parts. The terms "include" or "comprises" and similar terms mean that the elements before the term include the elements listed after the term, and do not exclude the possibility of also including other elements. The terms "upper", "lower", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0051] In the present disclosure, when a specific device is described as being located between a first device and a second device, an intervening device may or may not be present between the specific device and the first device or the second device. When a specific device is described as being connected to another device, the specific device may be directly connected to the other device without an intervening device, or may be not directly connected to the other device but with an intervening device.

[0052] All terms (including technical or scientific terms) used in this disclosure have the same meaning as those understood by one of ordinary skill in the art to which this disclosure belongs, unless otherwise specifically defined. It should also be understood that terms defined in, for example, general dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an idealized or highly formal sense, unless explicitly defined herein.

[0053] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0054] Companies involved in steel structure welding and fabrication must prioritize weld appearance inspection technology to provide the most fundamental guarantee for the overall performance and safety of welded structures. Within the realm of welding inspection technology, four primary welding inspection methods exist: UT (ultrasonic testing), RT (radiographic testing), MT (magnetic particle testing), and VT (visual testing). VT boasts the widest application, most direct application, and the best results. Establishing a welding quality system based on ISO 3834-2 generally requires at least four certified VT inspectors, double the number required for other inspection methods.

[0055] The weld appearance defect detection methods in related technologies have many problems in practical applications, such as:

[0056] Subjectivity of test results: Traditional weld appearance defect detection relies primarily on visual inspection by the inspector, and the results of this method are easily affected by the inspector's subjectivity. Different inspectors may have different defect recognition results for the same weld due to differences in experience, vision, attention, and other factors, resulting in missed detections or false detections.

[0057] Low inspection efficiency: Manual inspection of weld appearance defects is time-consuming, especially in large-scale production or large-scale welding projects, where a large number of welds need to be inspected. The efficiency of manual inspection is far from meeting the requirements of production schedules.

[0058] High training costs: Inspectors performing weld visual defect inspections require specialized training, including but not limited to specialized knowledge of welding processes, materials science, defect identification, and the operation of inspection tools. This training is both time-consuming and costly. Furthermore, due to a shortage of specialized talent, companies face significant challenges in recruiting and retaining certified inspectors.

[0059] These problems limit the efficiency and effectiveness of weld quality inspection. In recent years, many automated appearance defect detection methods have emerged, mainly including contour detection methods based on line lasers and target detection methods based on pure vision.

[0060] Practical applications have found that line laser scanning systems usually require more complex hardware configurations, including laser marker transmitters, optical lenses, CCD cameras, etc. The detection range is relatively narrow and is not suitable for large-size weld detection of large structural parts. It is not effective for detecting certain types of defects, such as it is difficult to detect small cracks.

[0061] Therefore, the contour detection method based on line laser has the problems of "large equipment size, high cost, and narrow detection range", which results in the equipment being large in size and relatively high in cost and cannot be applied to the large-size weld detection of large structural parts.

[0062] The pure vision-based target detection method still requires manual intervention in terms of accurate dimensional measurement and quality grade evaluation, which limits the improvement of its automation level and cannot effectively improve the efficiency of weld appearance inspection.

[0063] In view of this, in one aspect of an embodiment of the present disclosure, a weld defect detection method is provided to help improve the efficiency of weld appearance detection.

[0064] Figure 1 is a flow chart of some embodiments of the weld defect detection method disclosed herein, Figure 2 is a schematic diagram of a defect picture with a laser mark according to some embodiments of the weld defect detection method disclosed herein, and Figure 3 is a schematic diagram of obtaining the type and size of weld defects based on the weld defect picture according to some embodiments of the weld defect detection method disclosed herein. A in Figure 2 is a weld defect, and B in Figure 2 is a laser mark. Referring to Figures 1 to 3, the weld defect detection method includes steps S1 to S2.

[0065] In step S1, a laser-marked image of the weld defect to be detected is obtained. This includes, but is not limited to, emitting a laser mark B around the weld defect A using a laser mark emitter 31, and taking an on-site image of the weld so that the image includes both the weld defect A and the surrounding laser mark B.

[0066] In step S2, the first model is used to determine the type and size of the weld defect to be detected based on the laser-marked defect image. Based on the captured laser-marked defect image, the first model can be used to determine the type and size of the defect. The first model includes, but is not limited to, a weld appearance defect recognition model.

[0067] The size of a weld defect includes, but is not limited to, the distance between the two ends of a weld defect or the distance between the starting and ending points of a weld defect. Weld defects include, but are not limited to, appearance defects such as cracks, pores, and slag inclusions.

[0068] The first model can identify information such as the type and size of weld appearance defects based on photos of weld defects, thereby improving the efficiency of weld appearance inspection, realizing the automation and intelligence of the entire weld appearance inspection process, and helping to enhance the accuracy of weld appearance inspection.

[0069] 1 to 3 , in some embodiments, the weld defect detection method further includes: step S3 to step S4 .

[0070] In step S3, a first image sample, the type of weld defect in the first image sample, and reference information for calculating the size of the weld defect in the first image sample are obtained. The source of the first image sample includes, but is not limited to, photos of weld appearance defects collected online, photos of weld appearance defects taken on-site, and photos and / or videos of weld appearance defects previously acquired through a weld defect detection system. If the video is a video, frames are extracted at fixed intervals.

[0071] In step S4, the first model is trained using the first image sample, the type of weld defect in the first image sample, and the size of the weld defect in the first image sample. Methods for training the first model include, but are not limited to, deep learning training using a convolutional neural network as a foundation.

[0072] The first image sample includes an image of a weld defect with a laser mark, such as the photograph shown in FIG2 . The laser mark on the first image sample simulates an image projected onto the workpiece surface by laser marker emitter 31. Reference information for calculating the weld defect type and size in the first image sample includes, but is not limited to, information added to the first image sample in response to user input.

[0073] The first model is trained by calculating reference information based on the first image sample, the type of weld defect in the first image sample, and the size of the weld defect in the first image sample, so that the first model can identify the size and type of defects in the image based on the image of the weld appearance defect, which helps to achieve accurate and efficient weld appearance detection.

[0074] Referring to Figures 2 and 3, in some embodiments, obtaining a first image sample, a weld defect type in the first image sample, and reference information for calculating the size of the weld defect in the first image sample includes: obtaining a second image sample; adding multiple parallel and equidistant laser scale lines in the second image sample to form a first image sample; marking the endpoints of each laser scale line, the weld defect type in the first image sample, and the endpoints of the weld defect in the first image sample, and the endpoints of the weld defect serve as reference points for calculating the weld size.

[0075] In response to user input, laser markings are added to the second image sample. The laser markings include, but are not limited to, multiple parallel and equally spaced laser scale lines. The laser markings are used to compare with the weld defect size in the first image sample to obtain reference information for defect size calculation. The actual size of the weld defect is then determined based on the defect size calculation reference information and the actual distance between the laser scale lines.

[0076] The endpoints of the laser scale lines include but are not limited to those added as C in FIG3 , i.e., the green “+” symbol in FIG3 . The types of weld defects include but are not limited to those marked as D in FIG3 , i.e., the word “Crack” in red and the word “Hole” in blue in FIG3 .

[0077] The endpoints of weld defects include, but are not limited to, those marked as E in Figure 3, namely, the red and blue "+" symbols in Figure 3. The two red "+" symbols are placed at the endpoints of a crack marked "Crack." The endpoints of a crack include, but are not limited to, the starting and ending points of the crack along its extension direction. The two blue "+" symbols are placed at the endpoints of a hole, including, but not limited to, the edges of the hole along its longest diameter.

[0078] By marking the endpoints of the weld in the first image sample to obtain the pixel size of the weld defect in the first image sample, and marking the endpoints of each laser scale line to obtain the pixel size between the laser scale lines, the size information of the weld defect in the image can be quantified, thereby enabling quantitative detection of weld defects. Introducing the weld defect size reference points and laser scale lines into the dataset annotation and model training process, it is possible to simultaneously identify the defect and the laser marked feature points and the starting and ending points of the defect, thereby obtaining the true size of the defect through pixel comparison.

[0079] Referring to Figures 2 and 3, in some embodiments, obtaining the first image sample, the type of weld defect in the first image sample, and reference information for calculating the size of the weld defect in the first image sample includes: obtaining the pixel spacing between two adjacent laser scale lines in the first image sample and the pixel spacing between the two endpoints of each weld defect; obtaining the actual size of the weld defect based on the numerical relationship between the pixel spacing between two adjacent laser scale lines in the first image sample and the pixel spacing between the two endpoints of each weld defect and the actual spacing between the two adjacent laser scale lines.

[0080] The pixel spacing between two adjacent laser scale lines includes, but is not limited to, calculating the pixels between the ends of all two adjacent scale lines in the laser scale mark and taking an average value to obtain more accurate pixel spacing data of the laser scale lines.

[0081] Since the spacing of the scale lines emitted by the laser marking emitter around the weld defect is a constant and will not change due to the distance between the laser marking emitter and the weld defect, the actual size of the weld defect can be obtained through the numerical relationship between the pixel size of the weld defect in the first image sample and the pixel spacing of the laser scale lines and the actual spacing of the laser scale lines.

[0082] In some embodiments, the operation of obtaining the second image sample specifically includes: obtaining an initial image sample of the weld defect; generating a second image sample based on the initial image sample using a second model; wherein the number of samples of the initial image sample of the weld defect is less than the number of samples of the second image sample.

[0083] The initial image samples of weld defects include but are not limited to photos of weld appearance defects collected online, photos of weld appearance defects taken on site, photos and / or videos of weld appearance defects obtained in the early stage through the weld defect detection system, etc. If it is a video, frames are extracted at fixed intervals.

[0084] The second model includes but is not limited to a weld appearance defect model, which is obtained by fine-tuning training based on diffusion models, generative adversarial networks (GANs), or self-attention models (transformers).

[0085] The second model can expand the initial image sample with appropriate prompt words to capture more images of weld defects. The second image sample can contain multiple images of the same weld defect with different morphologies and structures. For example, the dataset can be expanded by inputting words such as carbon steel, crater crack, and dispersed, and / or words such as aluminum alloy, surface, and diffuse porosity.

[0086] Based on the initial image samples, the second model can expand the types of weld defects in the samples so that the first model can be trained based on richer samples, thereby being able to identify more types of weld defects and obtain more accurate recognition results.

[0087] 1 to 3 , in some embodiments, the method for establishing the first model specifically includes:

[0088] Collect photos of weld appearance defects, including online images, on-site defect photos, and photos and videos taken by the previous weld defect detection system. If it is a video, extract frames at fixed intervals;

[0089] Fine-tuning the diffusion model as a foundation for training yields a second model specifically tailored to the welding vertical domain, augmenting photos of weld appearance defects with appropriate prompt words.

[0090] All images are reviewed for rationality, and if qualified, a mark is added to the image. The added mark is the image formed by the laser marking transmitter 31 projected onto the workpiece surface;

[0091] Manually annotate the image. The content to be annotated includes: weld defect type, defect size calculation reference point, and end point of each scale line on the laser mark, which are packaged into a piece of data.

[0092] Based on the convolutional neural network, deep learning training was performed to obtain the first model.

[0093] Furthermore, the recognition results of the first model are processed to calculate the defect size. The pixels between the reference points are calculated, and the pixels between all two adjacent end points within the mark are calculated and averaged. After comparing the two, the actual size of the defect is obtained based on the actual spacing of the laser scale lines. The first model and its recognition result processing algorithm are packaged into a defect recognition measurement tool, which can also be combined with real measurement data to obtain the confidence level of defect recognition.

[0094] The workflow is compiled, including obtaining laser-marked defect images of the weld defects to be detected from the image acquisition unit 32, using defect recognition and measurement tools to perform recognition and measurement, summarizing the detection results, and compiling a detection report, which can eventually form a complete weld appearance defect detection intelligent agent application.

[0095] The weld defect detection method can be implemented using an agent application. The image acquisition unit 32 scans and captures weld defects while moving, and transmits the video information stream to the agent application deployed on a computer. The agent application primarily includes weld appearance defect identification and measurement tools and workflow orchestration. The agent summarizes the measurement results and compiles inspection reports, significantly improving the efficiency and accuracy of weld appearance inspections and effectively preventing the adverse effects of defective structural components on subsequent production and use.

[0096] In another aspect of an embodiment of the present disclosure, a weld defect detection device is provided, comprising: a memory and a processor 1 coupled to the memory, wherein the processor 1 executes a weld defect detection method as described in any of the above embodiments.

[0097] The processor 1 described herein may be implemented or executed by a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. The general-purpose processor 1 may be a microprocessor, but in an alternative embodiment, the processor 1 may be any conventional processor, controller, microcontroller, or state machine. The processor 1 may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0098] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by the processor 1, or in a combination of the two. The software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0099] Figure 4 is a structural schematic diagram of some embodiments of the weld defect detection system according to the present disclosure, and Figure 5 is a connection relationship schematic diagram of some embodiments of the weld defect detection system according to the present disclosure. Referring to Figures 4 and 5, in another aspect of the embodiments of the present disclosure, a weld defect detection system is provided, including a mounting body 2, a detection component 3 and a weld defect detection device as in the above-mentioned embodiments.

[0100] Detection assembly 3 is mounted on mounting body 2, which is connectable to an end effector. Detection assembly 3 includes a laser marker emitter 31, which is configured to emit a laser marker around the weld defect, and an image acquisition unit 32, which is configured to capture an image of the weld defect with the laser marker.

[0101] The processor 1 is signal-connected to both the image acquisition unit 32 and the laser marker emitter 31. The laser marker includes multiple parallel, equidistant laser scale lines, the spacing of which can be adjusted via an internal lens. There are at least 20 laser scale lines. The image acquisition unit 32 includes, but is not limited to, a camera.

[0102] The weld defect detection system emits equally spaced, parallel laser beams on the surface of the workpiece through the laser marking emitter 31, and takes pictures of the weld appearance defects with laser markings through the image acquisition unit 32. Based on the pictures of the weld appearance defects with laser markings, the weld defect detection device can obtain information such as the type and size of the weld appearance defects, thereby realizing quantitative detection of weld appearance defects, effectively improving detection efficiency and detection accuracy, and realizing efficient and automated detection of weld appearance defects.

[0103] 4 , in some embodiments, the detection assembly 3 can rotate relative to the mounting body 2. The detection assembly 3 further includes a laser rangefinder 33 symmetrically disposed on both sides of the laser marking emitter 31 and configured to respectively obtain the distance to the weld defect.

[0104] The processor 1 is connected to the laser rangefinder 33 signal and is configured to adjust the angle of the detection component 3 relative to the mounting body 2 according to the distance information respectively obtained by the two laser rangefinders 33 so that the laser marking emitter 31 emits a vertical laser mark around the weld defect.

[0105] The mounting body 2 includes, but is not limited to, a cubic shape, and the detection assembly 3 includes, but is not limited to, a cylindrical shape. The detection assembly 3 can rotate about its own axis relative to the mounting body 2. The laser rangefinder 33, the image acquisition unit 32, and the laser marking emitter 31 are disposed on the end surface of the detection assembly 3.

[0106] The two laser rangefinders 33 are symmetrically distributed relative to a vertical axis passing through the center of the detection component 3 , and the image acquisition unit 32 and the laser marking emitter 31 are arranged on the vertical axis passing through the center of the detection component 3 .

[0107] During image capture of weld defects, two laser rangefinders 33 simultaneously measure the distance to the workpiece. If the difference between the two distances is greater than a preset value, it indicates that the mounting body 2 is currently tilted. The processor 1 rotates the mounting body 2 until the difference between the distances measured by the two laser rangefinders 33 to the workpiece is less than a preset value, so that the laser mark emitted by the laser marker emitter 31 around the weld defect is vertical. The preset value can be selected as 2 mm.

[0108] The automated inspection process based on the weld defect detection system requires no human intervention. The compact and simple detection hardware improves inspection speed, effectively shortening inspection cycles and improving production efficiency, especially in large-scale production. By using laser marking, the system can accurately measure defect dimensions, providing important data support for subsequent defect analysis and treatment.

[0109] 2 to 5 , in some embodiments, the steps of detecting weld appearance defects using a weld defect detection system include:

[0110] First, the mounting body 2 is mounted on the end axis of the robot, and the detection component 3 communicates with the processor 1;

[0111] Place a ruler with a scale approximately 20 cm from the end face of the image acquisition unit 32. Adjust the spacing of the laser scale lines emitted by the laser marking emitter 31 to ensure that all laser scale lines are within the image captured by the image acquisition unit 32 and that the laser scale lines completely overlap with the scale lines on the ruler. This ensures that the spacing between the laser scale lines is consistent with the actual value to improve the accuracy of the laser marking. The spacing between adjacent laser scale lines can be selected according to the actual situation and can be 1 mm or 5 mm.

[0112] Move the inspection assembly 3 to a distance of 10 to 40 cm from the weld surface to prevent the laser mark from becoming thicker due to excessive distance, which would affect accuracy. The laser marker emitter 31, image acquisition unit 32, and laser rangefinder 33 are facing the weld. After the inspection begins, the two laser rangefinders 33 simultaneously measure the distance to the workpiece. When the difference between the two distances is greater than a preset value, it indicates that the mounting body 2 is currently skewed. The mounting body 2 is rotated until the difference between the distances measured by the two laser rangefinders 33 to the workpiece is less than a preset value. This ensures that the laser mark emitted by the laser marker emitter 31 toward the weld defect is vertical, thereby calibrating the accuracy.

[0113] When the detection component 3 moves along the direction of the weld, the obtained image information flow will directly call the weld defect detection device, and then the obtained detection results will be summarized and the information will be deduplicated. When all welds are inspected, the large language model will be called to compile the inspection report based on the final measurement results.

[0114] In another aspect of the present disclosure, a weld defect detection device is provided, comprising an image acquisition module and a defect detection module. The image acquisition module is configured to acquire an image of a laser-marked defect of a weld to be inspected, and the defect detection module is configured to determine the type and size of the weld to be inspected based on the laser-marked defect image using a first model.

[0115] The defect detection module can identify the type and size of weld appearance defects based on the photos of weld defects obtained by the image acquisition module, thereby improving the efficiency of weld appearance inspection, realizing the automation and intelligence of the entire weld appearance inspection process, and helping to enhance the accuracy of weld appearance inspection.

[0116] In another aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by the processor 1, the weld defect detection method of any of the above embodiments is implemented.

[0117] In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functions may be stored on or transmitted via a computer-readable medium as one or more instructions or codes. Computer-readable media include both computer storage media and communication media, including any media that facilitates the transfer of a computer program from one location to another. A storage medium may be any available medium that can be accessed by a computer. By way of example and not limitation, such computer-readable media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Any connection is also properly referred to as a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwaves, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwaves are included in the definition of medium. As used herein, disk and disc include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc, where disks typically reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0118] Thus far, various embodiments of the present disclosure have been described in detail. To avoid obscuring the concept of the present disclosure, some details known in the art have not been described. Based on the above description, those skilled in the art can fully understand how to implement the technical solutions disclosed herein.

[0119] Although some specific embodiments of the present disclosure have been described in detail through examples, those skilled in the art will understand that the above examples are for illustration only and are not intended to limit the scope of the present disclosure. Those skilled in the art will understand that the above embodiments may be modified or some technical features may be replaced with equivalents without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A weld defect detection method, comprising: Obtain a laser-marked defect image of the weld defect to be detected; According to the defect picture with laser marking, the type and size of the weld defect to be detected are determined using the first model.

2. The weld defect detection method according to claim 1, further comprising: Acquire a first image sample, a type of weld defect in the first image sample, and reference information for calculating a size of the weld defect in the first image sample; Calculating reference information using the first image sample, the type of weld defect in the first image sample, and the size of the weld defect in the first image sample to train the first model; The first image sample includes a weld defect image with laser marking.

3. The weld defect detection method according to claim 2, wherein obtaining the first image sample, the weld defect type in the first image sample, and the size calculation reference information of the weld defect in the first image sample comprises: Obtain a second image sample; Adding a plurality of parallel and equidistantly arranged laser scale lines to the second image sample to form the first image sample; The endpoints of each of the laser graticule lines, the weld defect type in the first image sample, and the endpoints of the weld defects are marked in the first image sample.

4. The weld defect detection method according to claim 3, wherein obtaining the first image sample, the weld defect type in the first image sample, and the size calculation reference information of the weld defect in the first image sample comprises: Obtaining the pixel spacing between two adjacent laser graticule lines and the pixel spacing between two end points of each weld defect in the first image sample; The actual size of the weld defect is obtained according to the numerical relationship between the pixel spacing between two adjacent laser graticule lines in the first image sample and the pixel spacing between the two end points of each weld defect and the actual spacing between two adjacent laser graticule lines.

5. The weld defect detection method according to claim 3 or 4, wherein the operation of obtaining the second image sample specifically comprises: Get an initial sample of images of weld defects; Generate the second picture sample using the second model according to the initial picture sample; The number of samples of the initial image samples of the weld defects is less than the number of samples of the second image samples.

6. A weld defect detection device, comprising: Memory; and A processor (1) coupled to the memory, wherein the processor (1) executes the weld defect detection method according to any one of claims 1 to 5.

7. A weld defect detection system, comprising: Mounting body (2); A detection component (3) is arranged on the mounting body (2), the detection component (3) comprising: a laser marking emitter (31) and an image acquisition unit (32), the laser marking emitter (31) being configured to emit a laser mark around the weld defect, and the image acquisition unit (32) being configured to acquire an image of the weld defect with the laser mark; The weld defect detection device according to claim 6, wherein the processor (1) is signal-connected to both the image acquisition unit (32) and the laser marking transmitter (31); Wherein, the laser mark includes a plurality of parallel and equidistant laser scale lines.

8. The weld defect detection system according to claim 7, wherein the detection component (3) is rotatable relative to the mounting body (2); in, The detection component (3) further comprises: Laser rangefinders (33) are symmetrically arranged on both sides of the laser marking transmitter (31) and are configured to respectively obtain the distances to the weld defects; The processor (1) is connected to the laser rangefinder (33) by signal and is configured to adjust the angle of the detection component (3) relative to the mounting body (2) according to the distance information respectively acquired by the two laser rangefinders (33), so that the laser marking emitter (31) emits a vertical laser mark around the weld defect.

9. A weld defect detection device, comprising: An image acquisition module is configured to acquire a defect image with a laser mark of the weld to be inspected; The defect detection module is configured to determine the type and size of the weld to be detected by using a first model based on the defect image with the laser mark.

10. A computer-readable storage medium having a computer program stored thereon, wherein: When the program is executed by a processor (1), the weld defect detection method described in any one of claims 1 to 5 is implemented.

Images