A method for inspecting welding quality of vehicle body panels

Abstract

A method for inspecting the welding quality of vehicle body panels involves installing corresponding sensor devices on a welding robot, performing multiple welding operations, capturing images of the welding points using a camera during welding, and detecting the current and voltage at the weld points using current and voltage detection devices. The welding robot's action data, the current, voltage, and resistance of the weld points, and the welding texture, gaps, and curvature obtained from the images are used as inputs, with the quality score of the welding result as the output, forming multiple first sample data sets. Principal component analysis (PCA) is applied to these first sample data sets, using data corresponding to the selected input feature types as inputs, and a new quality score of the welding result formed by the final weights as the output, forming multiple second sample data sets. The second sample data are used as training data to train a model, resulting in a vehicle body panel welding quality inspection model for subsequent vehicle body panel welding quality inspection.

Description

Technical Field

[0001] This invention relates to the field of welding quality inspection, and more particularly to a method for welding quality inspection based on body panels after welding of body panels in automobile manufacturing processes. Background Technology

[0002] In the automotive manufacturing industry, the standard process flow includes stamping, welding, painting, and final assembly workshops. From a single sheet of steel, a car is produced through these four core workshops, undergoing a series of quality inspections before successfully rolling off the production line. However, in this process, the welding quality of the car body is a crucial factor in the overall manufacturing quality and reliability of the vehicle. The main indicators of welding quality are the dimensional accuracy, area, thickness, and weld strength of the welded body panels, all of which determine whether the car body meets national standards for strength and hardness. Understanding the factors related to the welding dimensions and strength of thin metal sheets, grasping their patterns, and optimizing their control are key to improving the welding quality of the car body. Body welding quality inspection methods and systems are crucial indicators during body manufacturing welding. Due to the numerous and complex factors influencing body welding quality—from welding dimensions, welding precision, and welding angles—influencing factors include the selection of body and component positioning points, dimensional chain design, fixture design, tolerance distribution, fixture and tooling maintenance, and component and assembly dimensional monitoring, all significantly impact welding quality. From the perspective of body welding strength, influencing factors include welding time, current, voltage, and pressure. From the perspective of welded structural fatigue reliability, influencing factors include material properties, component structure, weld point distribution, and material thickness. Therefore, it is necessary to provide a reliable method and system for inspecting body sheet metal welding quality to improve body welding quality, promote overall vehicle quality improvement, increase manufacturing efficiency, reduce manufacturing costs, and shorten development and verification cycles. Summary of the Invention

[0003] To solve the above-mentioned technical problems, this invention provides a method for inspecting the welding quality of vehicle body panels, comprising the following steps:

[0004] S1. Install corresponding sensor devices, including cameras, current detection devices and voltage detection devices, on the welding robot used for welding body panels.

[0005] S2. Perform multiple welding operations. During welding, a camera takes pictures of the welding point, and a current detection device and a voltage detection device detect the current and voltage of the welding point during welding.

[0006] S3. Using the welding action data of the welding machine, the current, voltage and resistance of the weld joint, and the welding texture, gap and curvature of the weld obtained based on the image as input, and the quality score of the welding result as output, multiple first sample data are formed; wherein, the quality score is obtained according to the following method: preset standard data for various types of data in the input, set a preset tolerance zone for each of them, determine the score of each type of data according to the degree of deviation from the tolerance zone, and then weight all the scores according to preset weights to obtain the final quality score;

[0007] S4. Perform principal component analysis on the multiple first sample data, use the first sample data as test data for testing, continuously adjust the tolerance zone and weight according to the accuracy of quality inspection, and finally select the input feature type with a larger weight from the input as the data type with a higher correlation to welding quality. Use the data corresponding to the selected input feature type as input, and use the quality score of the new welding result formed by the final weight as output to form multiple second sample data.

[0008] S5. Use the second sample data as training data to train the model and obtain the body panel welding quality inspection model for subsequent body panel welding quality inspection.

[0009] Furthermore, in the method for detecting welding quality of vehicle body panels of the present invention, each of the sensor devices is electrically connected and controlled by the corresponding welding robot for data acquisition. Each of the welding robots is connected to the server cluster through a switch device. The images, current and voltage collected in step S2 are cleaned by an edge computing box before being connected to the server cluster, which then processes them for subsequent steps.

[0010] Furthermore, in the method for inspecting the welding quality of vehicle body panels according to the present invention, the data cleaning operation includes removing invalid or missing data and classifying the data.

[0011] Furthermore, in the method for inspecting the welding quality of body panels of the present invention, different welding quality inspection models for body panels are established for different vehicle models and different weld points.

[0012] Furthermore, in the method for detecting welding quality of vehicle body panels of the present invention, the model training includes training using any of the following models: BP neural network, random forest.

[0013] Furthermore, in the method for inspecting the welding quality of vehicle body panels of the present invention, the welding action data includes: the coordinate position of the weld point, the welding time, the welding angle, the fixture used, and the welding material.

[0014] Furthermore, in the method for detecting welding quality of vehicle body panels of the present invention, after the quality of each weld point of the vehicle body panel is detected by the vehicle body panel welding quality detection model, the quality of each weld point of the current overall vehicle body is comprehensively analyzed to finally obtain the overall welding quality of the current vehicle body.

[0015] Furthermore, in the method for inspecting the welding quality of body panels according to the present invention, the current detection device measures the current value of the welding point through a current CT coil, and the voltage detection device uses the voltage detection line electrically connected to the voltage detection chip as the input measurement voltage, and uses Ethernet communication to send the measurement data to the edge box for data collection.

[0016] Furthermore, in the method for inspecting welding quality of vehicle body panels according to the present invention, the camera is equipped with vision inspection system software. The camera hardware is installed on the arm of each welding robot. The vision inspection system of the camera and the welding robot are highly integrated and a communication mechanism between them is established. The camera can receive relevant photo-taking instructions from the welding robot controller, including when to take a photo, what angle to take a photo, the number of photos, welding process video, etc. The camera will respond to the instructions according to different instructions and output the response results to a pre-defined folder for classification and storage.

[0017] The method for inspecting welding quality of vehicle body panels according to the present invention has the following technical effects: The present invention can inspect the welding quality of vehicle body panels. It uses welding machine action data, weld current, voltage, and resistance, as well as welding patterns, gaps, and curvature obtained from the images as input, and the quality score of the welding result as output. Simultaneously, through principal component analysis, it identifies data types with higher correlation to welding quality. While comprehensively identifying the factors affecting welding quality, it also effectively improves the identification speed. Furthermore, through a large amount of test data, it identifies the weights and tolerance bands related to the physical characteristics of the weld points. The system will achieve self-optimization in weld point quality judgment and attempt to find more key physical features of the weld points. The model continuously learns, self-repairs, and continuously revises the precise range tolerances of various features, effectively improving the accuracy of welding quality inspection. Attached Figure Description

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0019] Figure 1 This is a system architecture diagram of an embodiment of the method for inspecting the welding quality of vehicle body panels according to the present invention;

[0020] Figure 2This is a signal diagram of an embodiment of the vehicle body panel welding quality inspection method of the present invention;

[0021] Figure 3 This is a diagram of the welding points on a Dongfeng Honda CRV model. Detailed Implementation

[0022] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0023] This embodiment of a method for inspecting the welding quality of vehicle body panels includes the following steps:

[0024] S1. Install appropriate sensor equipment, including cameras, current detection devices, and voltage detection devices, on the welding robot used for welding body panels.

[0025] The current detection device measures the current value at the welding point using a current CT coil, and the voltage detection device measures the input voltage using a voltage detection line electrically connected to a voltage detection chip. The camera is preferably a high-definition AI camera, and these sensors can be installed on the welding arm of the welding robot.

[0026] S2. Perform multiple welding operations. During welding, a camera is used to take pictures of the welding point, and a current detection device and a voltage detection device are used to detect the current and voltage of the welding point.

[0027] refer to Figure 1 as well as Figure 2 In this embodiment, each sensor device is electrically connected and controlled by its corresponding welding robot for data acquisition. Each welding robot is connected to a server cluster via a switch. The images, current, and voltage data acquired in step S2 can be cleaned via an edge computing box, then connected to the server cluster via Ethernet communication for subsequent processing. Invalid or missing data is optimized via the edge computing box, removing invalid or missing data and further classifying it. The cleaned data ensures accuracy and integrity while reducing the data load on the server cluster.

[0028] The camera also comes with vision inspection system software. The camera hardware is installed on the arm of each welding robot. The camera's vision inspection system and the welding robot are highly integrated and a communication mechanism has been established between them. The camera can receive relevant photo-taking instructions from the welding robot controller, including when to take a photo, what angle to take the photo, the number of photos, the welding process video, etc. The camera will respond to different instructions and output the response results to a pre-defined folder for classification and storage.

[0029] The resistance value can be obtained by calculating the ratio of voltage to resistance.

[0030] S3. Using the welding action data of the welding machine, the current, voltage and resistance of the weld joint, and the welding texture, gap and curvature of the weld obtained based on the image as input, and the quality score of the welding result as output, multiple first sample data are formed; wherein, the quality score is obtained according to the following method: preset standard data for various types of data in the input, set a preset tolerance zone for each of them, determine the score of each type of data according to the degree of deviation from the tolerance zone, and then weight all the scores according to the preset weight to obtain the final quality score.

[0031] Welding action data includes: the coordinate position of the weld point, the welding time, the welding angle, the fixture used, and the welding material.

[0032] Different welding quality inspection models for body panels were established for different vehicle models and different welding points.

[0033] Taking the models produced by Dongfeng Honda's Plant 3 as an example, their Plant 3 currently produces three main types of vehicles: CRV, XRV, and UR-V. Based on different configurations within these main categories, many different models are derived. Therefore, a model library needs to be built as standard data for all models produced by Dongfeng Honda's Plant 3. This model library mainly includes a database of weld point image models generated by cameras, a database of weld point current, voltage, and resistance models connected by sensors, and a database of welding actions. Simultaneously, tolerance range values ​​conforming to standards are calculated, and each measurement point has a database of acceptable standards. Similarly, it is also necessary to collect samples with various problems that lead to weld point defects, such as bubbles, impurities, cracks, etc., and build a database of acceptable standards for each measurement point.

[0034] S4. Perform principal component analysis on the multiple first sample data, use the first sample data as test data for testing, continuously adjust the tolerance zone and weight according to the accuracy of quality inspection, and finally select the input feature type with a larger weight from the input as the data type with a higher correlation to welding quality. Use the data corresponding to the selected input feature type as input, and use the quality score of the new welding result formed by the final weight as output to form multiple second sample data.

[0035] like Figure 3 As shown, the current CRV has 38 measurement points, each with a weld point image model library. Through offline qualified welding samples, a large number of image libraries are collected, including qualified sample libraries collected from different angles based on the model and weld defective sample libraries classified by different defects. In order to continuously improve the accuracy of the image model, the image sample library is continuously improved, enriched, and enhanced from the massive amount of images collected later. The model is revised, self-learned, and self-revised, so the image detection model will become more and more accurate.

[0036] S5. Using the second sample data as training data, train the model to obtain a vehicle body panel welding quality inspection model for subsequent vehicle body panel welding quality inspection. After detecting the quality of each weld point on the vehicle body panel using the vehicle body panel welding quality inspection model, the quality of all weld points on the current overall vehicle body is aggregated and analyzed to finally determine the overall welding quality of the current vehicle body. Model training can be performed using any of the following models: BP neural network, random forest, or other commonly used models or their improvements.

[0037] When using the detection model of this invention for quality inspection, a virtual simulation system can be constructed. This system maps all on-site vehicle data, welding machine action data, weld point current, voltage, and resistance, as well as images of the welded points and welding quality results onto the virtual simulation system. This ensures that every action of the on-site robot, the coordinate position of each weld point (X-axis, Y-axis, Z-axis), welding angle, welding time, pause time, and welding material are synchronously mapped into the simulation system. Simultaneously, real-time images of the welded points captured by a camera can be virtualized and synchronously displayed in the simulation system using image processing techniques. The weld point's forming effect, color, and surface texture are all mapped into the model based on the actual image data. Through this virtual simulation system, we can continuously learn, revise, and improve the welding quality inspection system. Furthermore, when the model's data volume reaches a certain level, we can use specific algorithms not only to quickly detect welding quality but also to predict the weld point's quality in advance—whether it is qualified and meets quality inspection standards—based on the current welding process, angle, time, and other indicators. Being able to predict the outcome in advance allows the model to correct errors such as improper welding that could lead to substandard welding quality. By pre-judging these errors, the model can revise subsequent welding processes to remedy previous mistakes and non-standard welding. Ultimately, the model can quickly detect the welding quality of each weld point and predict the outcome, allowing for reverse revision of the welding process to ensure that the weld quality meets extremely high standards.

[0038] The method for inspecting welding quality of vehicle body panels according to the present invention has the following technical effects: The present invention can inspect the welding quality of vehicle bodies by using welding machine action data, weld current, voltage, and resistance, as well as welding patterns, gaps, and curvature obtained from the images as input, and the quality score of the welding result as output. Simultaneously, through principal component analysis, it identifies data types with higher correlation to welding quality. While comprehensively identifying the factors affecting welding quality, it can effectively improve the identification speed. Furthermore, through a large amount of test data, it identifies the weights and tolerance bands related to the physical characteristics of the weld points, and achieves self-optimization of the system's weld point quality judgment, attempting to find more key physical features of the weld points. The model continuously learns, self-repairs, and continuously revises the precise range tolerances of various features, effectively improving the accuracy of welding quality inspection.

[0039] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for inspecting the welding quality of vehicle body panels, characterized in that, It includes the following steps: S1. Install corresponding sensor devices, including cameras, current detection devices and voltage detection devices, on the welding robot used for welding body panels. S2. Perform multiple welding operations. During welding, a camera takes pictures of the welding point, and a current detection device and a voltage detection device detect the current and voltage of the welding point during welding. S3. Using the welding action data of the welding robot, the current, voltage, and resistance of the weld point, and the welding texture, gap, and curvature of the weld obtained based on the image as input, and the quality score of the welding result as output, multiple first sample data are formed; wherein, the welding action data includes: the coordinate position of the weld point, the welding time, the welding angle, the fixture used, the welding material, and the quality score is obtained according to the following method: preset standard data for various types of data in the input, set a preset tolerance zone for each of them, determine the score of each type of data according to the degree of deviation from the tolerance zone, and then weight all the scores according to preset weights to obtain the final quality score; S4. Perform principal component analysis on the multiple first sample data, use the first sample data as test data for testing, continuously adjust the tolerance zone and weight according to the accuracy of quality inspection, and finally select the input feature type with a larger weight from the input as the data type with a higher correlation to welding quality. Use the data corresponding to the selected input feature type as input, and use the quality score of the new welding result formed by the final weight as output to form multiple second sample data. S5. Use the second sample data as training data to train the model and obtain the body panel welding quality inspection model for subsequent body panel welding quality inspection. S6. Construct a virtual simulation system to map on-site vehicle data, welding action data, weld current, voltage and resistance, weld images, and welding quality result data into the virtual simulation system. Continuously revise and improve the welding quality inspection model through the virtual simulation system, and predict the quality of the weld in advance based on the current welding process, angle, and time indicators.

2. The method for inspecting the welding quality of vehicle body panels according to claim 1, characterized in that, Each of the sensor devices is electrically connected to and controlled by the corresponding welding robot for data acquisition. Each of the welding robots is connected to the server cluster through a switch device. The images, current and voltage collected in step S2 are cleaned by the edge computing box before being connected to the server cluster for subsequent processing.

3. The method for inspecting the welding quality of vehicle body panels according to claim 2, characterized in that, The data cleaning operation includes removing invalid or missing data and classifying the data.

4. The method for inspecting the welding quality of vehicle body panels according to claim 1, characterized in that, Different welding quality inspection models for body panels were established for different vehicle models and different welding points.

5. The method for inspecting the welding quality of vehicle body panels according to claim 1, characterized in that, The model training includes training using any of the following models: BP neural network, random forest.

6. The method for inspecting the welding quality of vehicle body panels according to claim 1, characterized in that, After using a body panel welding quality inspection model to detect the quality of each weld point on the body panel, the quality of each weld point on the current overall body is comprehensively analyzed to finally obtain the overall welding quality of the current vehicle body.

7. The method for inspecting the welding quality of vehicle body panels according to claim 1, characterized in that, The current detection device measures the current value of the welding point through a current CT coil, and the voltage detection device measures the input voltage through the voltage detection line electrically connected to the voltage detection chip, and sends the measured data to the edge box for data collection using Ethernet communication.

8. The method for inspecting the welding quality of vehicle body panels according to claim 1, characterized in that, The camera is equipped with visual inspection system software. The camera hardware is installed on the arm of each welding robot. The camera's visual inspection system and the welding robot are highly integrated and a communication mechanism has been established between them. The camera can receive relevant photo-taking instructions from the welding robot controller, including when to take a photo, what angle to take a photo, the number of photos, and the welding process video. The camera will respond to different instructions and output the response results to a pre-defined folder for classification and storage.

Images