Weld joint quality detection method and system based on welding robot vision

Abstract

The invention relates to the technical field of welding, in particular to a weld joint quality detection system based on welding robot vision. Dry ice particles are stored in the dry ice storage device, and high-pressure gas generated by the gas compressor pushes the dry ice particles to be sprayed out of the spraying nozzle so as to clean a welding seam; the two steering engines are arranged on the two sides of the light source, two mirrors are installed on the two steering engines, and the steering engines drive the mirrors to rotate; and the data processing module is used for controlling the gas compressor to start to generate high-pressure gas before the weld seam image is acquired. The invention further relates to a welding robot vision-based welding seam quality detection method which is characterized by comprising the following steps: S1, the welding robot moves to a welding area to weld a to-be-welded position, so that the problem of inaccurate welding seam quality evaluation caused by the fact that a welding seam is blocked by slag can be solved.

Description

Technical Field

[0001] The present invention relates to the field of welding technology, and in particular to a welding seam quality detection method and system based on welding robot vision. Background Art

[0002] Welding robot technology is a key component in the field of industrial automation. It integrates advanced sensors, visual systems and intelligent control algorithms to realize the automation and intelligence of the welding process, thereby improving welding accuracy, efficiency and safety. It is widely used in the automotive, aerospace, steel structure and other industries, and has promoted the modernization of the manufacturing industry. Welding robots can replace manual labor and operate in many working conditions. Due to the lack of human participation, it is impossible to obtain data for monitoring weld quality based on experience.

[0003] The invention patent with application number 202410968341.1 discloses a weld profile visual detection method and system, which relates to the technical field of weld profile detection, and solves the technical problem that the prior art does not consider the detection of the fusion ratio of the weld and the concavity and excess height of the weld, and the detection accuracy of the weld is not high due to the single detection data; the present invention obtains image data of the weld profile; preprocesses the image data to obtain feature data of the weld profile; calculates the weld forming coefficient based on the feature data, and judges whether the weld quality is qualified based on the weld forming coefficient and a preset forming coefficient range; and, calculates the weld quality evaluation coefficient based on the feature data, and judges whether the weld quality is qualified based on the weld quality evaluation coefficient and a preset quality evaluation threshold, which solves the technical problem that the prior art does not detect the fusion ratio of the weld and the concavity and excess height of the weld, resulting in low weld detection accuracy.

[0004] The above technology can obtain and identify the image of the weld and make a judgment on the quality of the weld. However, during the welding process, welding slag and spatter are likely to adhere to the weld or the welded object, which not only easily affects the subsequent welding, but also easily causes errors in the uploaded image, making it impossible to make a true judgment on the welding quality through the image. Therefore, a technology is needed that can clean the weld before weld inspection to ensure the accuracy of image acquisition, thereby ensuring accurate judgment of the weld quality. Summary of the invention

[0005] The present invention provides a weld quality detection method and system based on welding robot vision, which can solve the problem that the weld is blocked by slag, resulting in inaccurate weld quality assessment.

[0006] In order to solve the above technical problems, the present application provides the following technical solutions: a welding seam quality detection system based on welding robot vision, comprising: An LED light source is installed on the welding robot and is used to illuminate the welding area; the LED moves with the welding gun; Image acquisition module: used to acquire images of the welding point and weld seam after welding; Dry ice blasting module: installed on the welding robot, including dry ice storage, gas compressor, and spray nozzle. The dry ice storage stores dry ice particles. The high-pressure gas generated by the gas compressor pushes the dry ice particles out of the spray nozzle to clean the weld. The spray nozzle is directed toward the weld. Two steering gears: Two steering gears are arranged on both sides of the light source, and two mirrors are installed on the two steering gears, and the steering gears drive the mirrors to rotate; Data processing module: Before the weld image is acquired, the data processing module controls the gas compressor to start, generates high-pressure gas, and drives the dry ice particles to flush the weld. After the flushing is completed, the data processing module controls the two steering gears to rotate, converge the light source into a line and project it on the weld. Then the image acquisition module acquires the weld image, and the data processing module processes the welding image to obtain the weld and the partially enlarged edge image of the weld. The weld and the locally enlarged edge images are then converted into grayscale images, and the weld features are extracted after denoising and enhancement. The features include at least edge, texture and grayscale distribution. The weld defects are identified through the differences and mutations in grayscale distribution to judge the weld quality.

[0007] The basic principle and beneficial effects of this solution: This solution uses the visual system of the welding robot combined with dry ice blasting technology to perform real-time cleaning and image acquisition of the weld to ensure the clarity and accuracy of the weld image, thereby improving the accuracy of weld quality detection.

[0008] In this solution, the LED light source moves with the welding gun to provide lighting for the welding area. The dry ice blasting module cleans the weld before and after welding, removes slag and spatter, and ensures that the weld image is not blocked. The image acquisition module acquires the image of the weld and the weld image after welding. The two servos drive the mirror to rotate, converge the LED light source into a line and project it on the weld to obtain a clearer weld image. The data processing module processes the acquired image, including converting the image into a grayscale image, denoising, enhancing, and extracting the features of the weld. The features of the weld edge, texture, and grayscale distribution are extracted. The weld defects are identified and the weld quality is judged by analyzing the differences and mutations in the grayscale distribution. This solution improves the detection accuracy: by cleaning the weld with dry ice blasting, the influence of slag blocking on the weld quality assessment is avoided, and the accuracy of the detection results is improved. The system is integrated on the welding robot, realizing the automation of weld cleaning, image acquisition, and quality inspection, and reducing manual intervention. Real-time acquisition and processing of weld images can timely detect problems in the welding process and improve welding quality. Fast cleaning and image acquisition reduce downtime during welding and improve production efficiency. Reduce rework and scrap caused by welding defects, and reduce production costs. The LED light source and dry ice blasting module can work in various welding environments, which improves the adaptability of the system. The system can record weld images and quality inspection results to facilitate subsequent data analysis and quality traceability. Through accurate evaluation of weld quality, the welding process can be continuously optimized and the overall welding level can be improved.

[0009] This solution converges the light source through the rotation of the servo, so that the influence of lighting factors can be eliminated when the image of the weld is acquired, so that the intensity of the illuminated light comes from the same beam of light, reducing the influence of brightness difference; compared with the method of irradiating from multiple directions, this solution can reduce the generation of lighting shadows, that is, the acquired image is more accurate, and the synchronous data processing module performs grayscale processing on the acquired image. The type of weld defect can be determined through the grayscale gap and mutation, so as to judge the quality. Before the image is acquired, the weld area is cleaned by the dry ice spray module, which effectively prevents welding slag and molten material from covering the weld surface, resulting in the obstruction of the weld acquired in the image, and normal analysis cannot be performed, or the analyzed quality does not match the actual quality.

[0010] In summary, the present invention provides an effective weld quality inspection solution by combining dry ice blasting technology and a visual inspection system, which can improve welding quality, reduce costs, and improve production efficiency.

[0011] Furthermore, the weld defects include pores, cracks and lack of fusion.

[0012] Furthermore, the data processing module identifies the grayscale image of the weld image, marks the area where the grayscale value and coverage area exceed the threshold, and issues a re-brushing instruction to control the dry ice blasting module to blast the area a second time; then re-acquires the weld image information of the area to overwrite the original weld image information.

[0013] Beneficial effect: A second dry ice flushing is performed in areas that are found to be obviously inconsistent with the grayscale distribution, further reducing the probability of error in the image sample and improving the accuracy of the sample.

[0014] Furthermore, several grayscale images of previously defective welds and their defect types were collected and imported into a convolutional neural network model for training to obtain a weld defect judgment model. The weld image information returned by the image acquisition module on the welding robot was processed and input into the weld defect judgment model to determine whether the weld has defects and the defect type.

[0015] Beneficial effects: Using models for intelligent judgment can reduce the error of manual judgment and avoid the increase of judgment error caused by fatigue caused by long-term manual judgment. It has the advantages of high efficiency and high accuracy.

[0016] Furthermore, the jet nozzle is arranged on the steering gear, and when the steering gear rotates, the light source is converged into a line and projected onto the weld. At this time, the jet nozzle is directly opposite to the weld area where the image acquisition module acquires the image.

[0017] Beneficial effect: When impurities are found in the image, the dry ice blasting module can be directly started for cleaning without adjusting the position of the blasting nozzle.

[0018] Further, the welding process of the welding robot and the dry ice blasting process are performed alternately.

[0019] Beneficial effects: The high temperature generated during welding can cause thermal stress in the welding material. If it is not cooled in time, it may cause thermal cracks in the weld. Cooling the weld can help reduce thermal stress and reduce the risk of thermal cracks. The cooling rate of the weld is related to the formation of its microstructure. Appropriate cooling of the weld can promote rapid solidification of the weld, form a dense grain structure, and improve the strength and hardness of the weld.

[0020] A welding seam quality detection method based on welding robot vision comprises the following steps: S1: The welding robot goes to the welding area to weld the position to be welded; S2: Dry ice flushing and cleaning of welding positions; S3: Obtain image information of the weld; S4: grayscale processing of image information; S5: Grayscale differences and mutations are identified for the grayscale processed images to obtain weld defects and then determine weld quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of a weld quality inspection system based on welding robot vision. DETAILED DESCRIPTION

[0022] The following is further described in detail through specific implementation methods: Embodiment 1 is as attached Figure 1 As shown, A welding seam quality detection system based on welding robot vision, comprising: The LED light source is installed on the welding robot to illuminate the welding area; the LED moves with the welding gun; the LED light source in this solution is an adjustable light source, which can select the appropriate light intensity for lighting according to the actual needs of the construction area. This solution only uses one light source to minimize the problem of overlapping shadows caused by multiple light sources, which causes inaccurate image accuracy in subsequent sampling.

[0023] Image acquisition module: used to acquire images of the welding point and the weld after welding; this solution uses a CCD camera.

[0024] Dry ice blasting module: installed on the welding robot, including a dry ice storage, a gas compressor, and a spray nozzle. The dry ice storage stores dry ice particles, and the high-pressure gas generated by the gas compressor pushes the dry ice particles to be sprayed out from the spray nozzle to clean the weld; the spray nozzle is facing the weld; The welding process of the welding robot and the dry ice blasting process are carried out alternately.

[0025] Two steering gears: Two steering gears are arranged on both sides of the light source, and two mirrors are installed on the two steering gears, and the steering gears drive the mirrors to rotate; the injection nozzle is arranged on one of the steering gears or two injection nozzles are arranged, and one is installed on each of the two steering gears. When the steering gear rotates, the light source is converged into a line and projected on the weld. At this time, the injection nozzle is facing the weld area where the image acquisition module acquires the image.

[0026] Data processing module: Before the weld image is acquired, the data processing module controls the gas compressor to start, generates high-pressure gas, and drives the dry ice particles to flush the weld. After the flushing is completed, the data processing module controls the two steering gears to rotate, converge the light source into a line and project it on the weld. Then the image acquisition module acquires the weld image, and the data processing module processes the welding image to obtain the weld and the partially enlarged edge image of the weld. The weld and the locally enlarged edge images are then converted into grayscale images, and the weld features are extracted after denoising and enhancement. The features include at least edges, textures, and grayscale distribution. The weld defects are identified through the differences and mutations in grayscale distribution to judge the weld quality. The data processing module identifies the grayscale image of the weld image, marks the area where the grayscale value and coverage area exceed the threshold, and issues a re-brush instruction to control the dry ice spray module to perform a second spray on the area. Then the weld image information of the area is re-acquired to overwrite the original weld image information.

[0027] Several grayscale images of previously defective welds and their defect types are collected and imported into a convolutional neural network model for training to obtain a weld defect judgment model. The weld image information returned by the image acquisition module on the welding robot is processed and input into the weld defect judgment model to determine whether the weld has defects and the defect type.

[0028] 10,000 weld image information was collected, and their defects were annotated and randomly divided into 3 groups, 1 / 3 of which was used as a sample set, 1 / 3 as an experimental set, and 1 / 3 as a verification set. The convolutional neural network model was imported for training, and the obtained model could be put into use when the verification result reached more than 95%.

[0029] The weld defects include porosity, cracks and lack of fusion.

[0030] It also relates to a weld quality detection method based on welding robot vision, which is applicable to the above system: The following steps are involved: S1: The welding robot goes to the welding area to weld the position to be welded; S2: Dry ice flushing and cleaning of welding positions; S3: Obtain image information of the weld; S4: grayscale processing of image information; S5: Grayscale differences and mutations are identified for the grayscale processed images to obtain weld defects and then determine weld quality.

[0031] This solution uses the welding robot's vision system combined with dry ice blasting technology to perform real-time cleaning and image acquisition of the weld to ensure the clarity and accuracy of the weld image, thereby improving the accuracy of weld quality inspection.

[0032] In this solution, the LED light source moves with the welding gun to provide lighting for the welding area. The dry ice blasting module cleans the weld before and after welding, removes slag and spatter, and ensures that the weld image is not blocked. The image acquisition module acquires the image of the weld and the weld image after welding. The two servos drive the mirror to rotate, converge the LED light source into a line and project it on the weld to obtain a clearer weld image. The data processing module processes the acquired image, including converting the image into a grayscale image, denoising, enhancing, and extracting the features of the weld. The features of the weld edge, texture, and grayscale distribution are extracted. The weld defects are identified and the weld quality is judged by analyzing the differences and mutations in the grayscale distribution. This solution improves the detection accuracy: by cleaning the weld with dry ice blasting, the influence of slag blocking on the weld quality assessment is avoided, and the accuracy of the detection results is improved. The system is integrated on the welding robot, realizing the automation of weld cleaning, image acquisition, and quality inspection, and reducing manual intervention. Real-time acquisition and processing of weld images can timely detect problems in the welding process and improve welding quality. Fast cleaning and image acquisition reduce downtime during welding and improve production efficiency. Reduce rework and scrap caused by welding defects, and reduce production costs. The LED light source and dry ice blasting module can work in various welding environments, which improves the adaptability of the system. The system can record weld images and quality inspection results to facilitate subsequent data analysis and quality traceability. Through accurate evaluation of weld quality, the welding process can be continuously optimized and the overall welding level can be improved.

[0033] This solution converges the light source through the rotation of the servo, so that the influence of lighting factors can be eliminated when the image of the weld is acquired, so that the intensity of the illuminated light comes from the same beam of light, reducing the influence of brightness difference; compared with the method of irradiating from multiple directions, this solution can reduce the generation of lighting shadows, that is, the acquired image is more accurate, and the synchronous data processing module performs grayscale processing on the acquired image. The type of weld defect can be determined through the grayscale gap and mutation, so as to judge the quality. Before the image is acquired, the weld area is cleaned by the dry ice spray module, which effectively prevents welding slag and molten material from covering the weld surface, resulting in the obstruction of the weld acquired in the image, and normal analysis cannot be performed, or the analyzed quality does not match the actual quality.

[0034] The above are only embodiments of the present invention. The invention is not limited to the fields involved in this implementation case. The common sense such as the known specific structures and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field to which the invention belongs before the application date or the priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The protection scope required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to explain the content of the claims.

Claims

1. A welding seam quality inspection system based on welding robot vision, characterized in that: include: An LED light source is installed on the welding robot and is used to illuminate the welding area; the LED moves with the welding gun; Image acquisition module: used to acquire images of the welding point and weld seam after welding; Dry ice blasting module: installed on the welding robot, including dry ice storage, gas compressor, and spray nozzle. The dry ice storage stores dry ice particles. The high-pressure gas generated by the gas compressor pushes the dry ice particles out of the spray nozzle to clean the weld. The spray nozzle is directed toward the weld. Two steering gears: Two steering gears are arranged on both sides of the light source, and two mirrors are installed on the two steering gears, and the steering gears drive the mirrors to rotate; Data processing module: Before the weld image is acquired, the data processing module controls the gas compressor to start, generates high-pressure gas, and drives the dry ice particles to flush the weld. After the flushing is completed, the data processing module controls the two steering gears to rotate, converge the light source into a line and project it on the weld. Then the image acquisition module acquires the weld image, and the data processing module processes the welding image to obtain the weld and the partially enlarged edge image of the weld. The weld and the locally enlarged edge images are then converted into grayscale images, and the weld features are extracted after denoising and enhancement. The features include at least edge, texture and grayscale distribution. The weld defects are identified through the differences and mutations in grayscale distribution to judge the weld quality.

2. The weld quality inspection system based on welding robot vision according to claim 1 is characterized in that: The weld defects include porosity, cracks and lack of fusion.

3. The weld quality inspection system based on welding robot vision according to claim 1 is characterized in that: The data processing module identifies the grayscale image of the weld image, marks the area where the grayscale value and coverage area exceed the threshold, and issues a re-brushing instruction to control the dry ice blasting module to perform a second blasting on the area; then re-acquires the weld image information of the area to overwrite the original weld image information.

4. The weld quality inspection system based on welding robot vision according to claim 3 is characterized in that: Several grayscale images of previously defective welds and their defect types are collected and imported into a convolutional neural network model for training to obtain a weld defect judgment model. The weld image information returned by the image acquisition module on the welding robot is processed and input into the weld defect judgment model to determine whether the weld has defects and the defect type.

5. The weld quality inspection system based on welding robot vision according to claim 4 is characterized in that: The injection nozzle is arranged on the steering gear, and when the steering gear rotates, the light source is converged into a line and projected onto the weld. At this time, the injection nozzle is facing the weld area where the image acquisition module acquires the image.

6. The weld quality inspection system based on welding robot vision according to claim 5 is characterized in that: The welding process of the welding robot and the dry ice blasting process are carried out alternately.

7. A weld quality detection method based on welding robot vision, characterized in that: The following steps are involved: S1: The welding robot goes to the welding area to weld the position to be welded; S2: Dry ice flushing and cleaning of welding positions; S3: Obtain image information of the weld; S4: grayscale processing of image information; S5: Grayscale differences and mutations are identified for the grayscale processed images to obtain weld defects and then determine weld quality.

Images