Demonstration-free welding track planning method and welding system

Abstract

The invention relates to the technical field of welding, and particularly discloses a teaching-free welding track planning method and a welding system.The teaching-free welding track planning method comprises the following steps that firstly, visual collection and processing are conducted, an industrial camera with the resolution ratio being 30 million pixels is selected, and a sensor monitoring unit is adopted by the system to monitor welding faults in real time; the fault diagnosis unit adopts an advanced algorithm for rapid diagnosis, and after the fault classification unit performs refined classification, the processing strategy formulation and execution unit performs rapid response processing, the recording analysis unit provides data support, the image acquisition analysis and quality evaluation unit accurately detects the welding quality, and the defective product judgment and rejection unit performs efficient operation. The data recording and counting unit assists in quality control, the target analysis unit can formulate targets for different workpieces, the path generation and optimization unit improves the efficiency and quality, the dynamic adjustment unit responds to changes in time, the storage calling and display unit facilitates management and monitoring, and the teaching-free aspect has the remarkable advantages, and all the units achieve autonomous learning and intelligent adjustment.

Description

Technical Field

[0001] The invention relates to the technical field of welding, and in particular to a teaching-free welding trajectory planning method and a welding system. Background Art

[0002] With the continuous development of industrial production, the requirements for welding quality, efficiency and automation are getting higher and higher. Traditional welding methods often rely on technicians to operate robots on site for manual teaching to determine each welding position. This method is not only inefficient, but also requires high experience and skills of technicians. Therefore, it has become an urgent need for the industry to develop welding methods that do not require manual teaching to improve the degree of welding automation and welding efficiency. The current technical solution meets the needs of welding automation to a certain extent. By setting up a welding template library, the applicability of the method is improved and it can adapt to different types of workpieces. At the same time, the first weld of the workpiece is determined by a digital model, which improves the accuracy and completeness of weld identification and extraction, and makes the matching of welds and welding templates more convenient. However, this technical solution still has shortcomings in the following aspects;

[0003] Although the current existing welding technology solutions have improved the degree of welding automation to a certain extent, there are still many shortcomings. In terms of troubleshooting and processing, there is a lack of effective fault detection and troubleshooting mechanisms. Welding equipment failures, electrical failures, mechanical failures, etc. that occur during the welding process may lead to welding interruptions, affecting production efficiency and even damaging equipment and workpieces. There is a lack of processing measures for common problems such as welding defects (pores, cracks, lack of fusion, etc.) and abnormal welding parameters (current and voltage instability), and it is impossible to adjust parameters in time to ensure welding quality. In terms of defective product detection and rejection, there is no perfect detection mechanism, and unqualified products cannot be discovered in time, increasing production costs and quality risks. At the same time, there is a lack of automatic rejection. Device and method, manual screening and elimination reduces production efficiency and the degree of automation. In terms of trajectory planning, it may not be possible to generate the optimal trajectory for complex-shaped workpieces, resulting in low welding efficiency and unstable quality. It lacks dynamic planning capabilities and cannot adjust the trajectory according to the deformation and position change of the workpiece, affecting accuracy and reliability. In terms of related aspects of free teaching, manual intervention may still be required in actual applications to adjust the welding template or handle abnormal situations, which is particularly obvious when the workpiece differs greatly from the template in the template library, reducing the degree of automation. It also lacks autonomous learning and adaptability, and cannot automatically optimize welding parameters and trajectory planning according to different welding tasks and workpiece characteristics, limiting the applicability and flexibility of the welding robot. Summary of the invention

[0004] The purpose of the present invention is to provide a teaching-free welding trajectory planning method and a welding system to solve the problems raised in the above background technology, namely, the lack of effective mechanism for troubleshooting and processing, and the difficulty in dealing with common problems; the lack of perfect mechanism and automatic device for defective product detection and rejection; the poor trajectory planning for complex workpieces and the lack of dynamic adjustment capability; the need for manual intervention in the teaching-free aspect and the lack of autonomous learning and adaptability.

[0005] To achieve the above object, the present invention provides a teaching-free welding trajectory planning method, which is characterized by comprising the following steps:

[0006] Step 1: Visual acquisition and processing: select an industrial camera with a resolution of 30 megapixels, configure a lens focal length of 80mm-100mm according to the size of the welding workpiece and the working distance, design a uniform lighting system with a light intensity of 1500Lux-2000Lux, use a multi-angle LED light source, set the angle between 40°-50°, set the image acquisition frequency to 80 frames / second-100 frames / second, grayscale the collected image, use a 4x4-5x5 median filter window to remove noise, and then adjust the image contrast to 1.5 times-2.0 times to highlight the characteristics of the welding area. Use the Canny edge detection algorithm to extract the edge features of the welding area, set the low threshold to 80-100, and the high threshold to 180-220. At the same time, use the corner detection algorithm to extract key corners;

[0007] Step 2: Data preparation for the teaching-free module, collect images, sensor data, welding process parameters and other information in various welding scenarios, for example, the integrated force sensor has a measurement accuracy of ±1N, the temperature sensor has a measurement accuracy of ±2℃, and the welding current measurement accuracy is ±1A, the voltage measurement accuracy is ±1V and other parameters are recorded, key features are extracted from the collected data, the feature extraction algorithm calculation time is controlled within 80 milliseconds to 120 milliseconds, the neural network algorithm is used to train the features, and a welding pattern recognition model is established. The training time is about 2 hours to 4 hours, and the model accuracy is required to reach more than 92%. Make intelligent adaptive decisions based on the trained model and real-time data. When the workpiece position changes between ±5mm-8mm, calculate the adjusted welding path and parameters, and the decision time is between 80 milliseconds and 100 milliseconds. Receive feedback information to adjust and optimize the teaching-free system, and the adjustment time is between 120 milliseconds and 150 milliseconds. Provide a human-computer interaction interface, and the interface response time is between 120 milliseconds and 150 milliseconds;

[0008] Step 3: Trajectory planning target analysis, analyze the target requirements of the welding task, the length of the straight weld is between 500mm-1000mm, and the error of the generated straight line trajectory is controlled within ±0.5mm; the curvature radius of the curved weld is between 100mm-200mm, and a smooth curved trajectory is generated, and the curvature change rate is controlled between 0.5 degrees / mm-1 degrees / mm, and the analysis time is between 1.5 seconds and 2.5 seconds. The initial welding path is generated according to the target analysis results and the initial position information of the workpiece. The linear interpolation algorithm is used for simple shape welds, and the generation time is between 1.5 seconds and 2 seconds; the spline curve fitting algorithm is used for complex curved welds, and the generation time is between 3 seconds and 5 seconds. The initial path is optimized, considering the welding speed The optimization time is between 8 seconds and 12 seconds, and the goal is to reduce the path length by 10%-15%. The welding path is dynamically adjusted according to real-time visual feedback and sensor data during welding. When the workpiece deformation is between ±3mm-5mm or welding defects are detected, timely adjustments are made. The adjustment response time is between 0.8 seconds and 1.2 seconds. The optimized path is stored for the next call. The storage capacity can store thousands of welding paths. The call time is between 100 milliseconds and 150 milliseconds. The welding path is visualized using a three-dimensional graphical interface with an interface refresh rate of 40 frames / second to 60 frames / second.

[0009] Step 4: Welding execution. Provide a suitable welding power supply with a welding current between 100A-150A and a voltage between 25V-30V according to the welding process and workpiece material. The power supply has stable output characteristics and adjustable parameters. For welding processes that require wire feeding, the welding gun accurately controls the wire feeding speed between 5m / min-10m / min. The welding gun angle and position are adjusted through mechanical devices to ensure that the welding arc acts stably on the weld. The movement of the welding equipment is controlled, including translation, rotation and other actions. The position accuracy can reach ±0.1mm and the angle accuracy can reach ±0.1°.

[0010] Step 5: Troubleshooting and processing. Integrate multiple sensors, such as current, voltage, and temperature sensors. The current sensor has a measurement accuracy of ±1A, the voltage sensor has a measurement accuracy of ±1V, and the temperature sensor has a measurement accuracy of ±2°C. Diagnose possible faults based on sensor data and system operation status. Use algorithms such as fault tree analysis. The diagnosis time is between 1 second and 3 seconds. Classify the diagnosed faults, formulate processing strategies for different types of faults, and the processing strategy formulation time is between 1 second and 2 seconds. Execute the processing strategy, and the processing response time is between 0.5 seconds and 1 second. Record the time, type, and processing measures of the fault for analysis;

[0011] Step 6: Detection and rejection of defective products. After welding is completed, use a high-resolution camera to capture images of the welding area, use image processing algorithms to analyze welding quality features in the image, and evaluate welding quality based on image analysis results and preset quality standards. The accuracy of the evaluation is required to reach more than 90%. Determine whether the welded product is a defective product based on the quality evaluation results. The judgment time is between 500 milliseconds and 1 second. When it is determined to be a defective product, a rejection signal is generated. The signal generation time is between 100 milliseconds and 200 milliseconds. The rejection signal is received to perform the rejection operation. The rejection response time is between 500 milliseconds and 1 second. Record the number, type, appearance time and other information of defective products for statistical analysis.

[0012] Furthermore, a welding system includes a main control system module, a power supply module, a visual acquisition processing module, a teaching-free module, a trajectory planning module, a welding execution module, a troubleshooting processing module and a defective product detection and rejection module, wherein the power supply module is electrically connected to the main control system module, the visual acquisition processing module, the teaching-free module, the trajectory planning module, the welding execution module, the troubleshooting processing module and the defective product detection and rejection module;

[0013] The main control system module includes a main control unit, a human-computer interaction unit and a communication unit. The human-computer interaction unit is connected to the main control unit through signals from the communication unit. The main control unit is connected to the visual acquisition processing module, the teaching-free module, the trajectory planning module, the welding execution module, the troubleshooting processing module and the defective product detection and rejection module through signals from the communication unit.

[0014] Furthermore, the visual acquisition and processing module includes a camera configuration unit, a lighting unit, an image acquisition unit, an image preprocessing unit and a feature analysis unit, and the configuration unit, the lighting unit, the image acquisition unit, the image preprocessing unit and the feature analysis unit are connected via a communication unit and a main control unit signal.

[0015] Furthermore, the teaching-free module includes a data acquisition unit, a feature extraction unit, a model training unit and an intelligent adaptive decision-making unit, and the data acquisition unit, feature extraction unit, model training unit and intelligent adaptive decision-making unit are connected via a communication unit and a main control unit signal.

[0016] Furthermore, the teaching-free module also includes a feedback adjustment unit and a human-computer interaction interface unit, and the feedback adjustment unit and the human-computer interaction interface unit are connected via a communication unit and a main control unit signal.

[0017] Furthermore, the trajectory planning module includes a target analysis unit, an initial path generation unit, a path optimization unit and a dynamic adjustment unit, and the target analysis unit, the initial path generation unit, the path optimization unit and the dynamic adjustment unit are connected via a communication unit and a main control unit signal.

[0018] Furthermore, the trajectory planning module also includes a path storage and calling unit and a visualization display unit, and the path storage and calling unit and the visualization display unit are connected via a communication unit and a main control unit signal.

[0019] Furthermore, the welding execution module includes a welding power supply unit, a welding gun and wire feeding unit and a welding motion unit, and the welding power supply unit, the welding gun and wire feeding unit and the welding motion unit are connected via a communication unit and a main control unit signal.

[0020] Furthermore, the troubleshooting processing module includes a sensor monitoring unit, a fault diagnosis unit, a fault classification unit, a processing strategy formulation unit, a fault processing execution unit and a fault record analysis unit, and the sensor monitoring unit, the fault diagnosis unit, the fault classification unit, the processing strategy formulation unit, the fault processing execution unit and the fault record analysis unit are connected through a communication unit and a main control unit signal.

[0021] Furthermore, the defective product detection and rejection module includes an image acquisition and analysis unit, a quality assessment unit, a monitoring feedback module, a welding parameter monitoring unit, a quality monitoring unit, a feedback control unit, a defective product determination unit, a rejection signal generating unit, a rejection execution unit and a data recording and statistical unit. The image acquisition and analysis unit, the quality assessment unit, the monitoring feedback module, the welding parameter monitoring unit, the quality monitoring unit, the feedback control unit, the defective product determination unit, the rejection signal generating unit, the rejection execution unit and the data recording and statistical unit are connected to the main control unit signal through the communication unit.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] First, in the present invention, the sensor monitoring unit integrated in the system includes sensors such as current, voltage, and temperature, which can monitor various types of faults in the welding process in real time. For example, the current sensor has a measurement accuracy of up to ±1A, and the voltage sensor has a measurement accuracy of up to ±1V, which can detect electrical faults in time. At the same time, the temperature sensor can detect abnormal temperature of the equipment, which helps to prevent mechanical failures. Secondly, the fault diagnosis unit adopts advanced algorithms, such as fault tree analysis, which can quickly and accurately diagnose the type and location of the fault between 1 second and 3 seconds, greatly reducing the troubleshooting time. After the fault classification unit classifies the fault in detail, the processing strategy formulation unit can quickly formulate processing measures for different types of faults. For example, for abnormal welding parameters, the current, voltage and other parameters can be automatically adjusted to ensure the welding quality. Moreover, the fault processing execution unit responds quickly, and the processing response time is between 0.5 seconds and 1 second, which can solve the fault in time and avoid welding interruption and equipment damage. Finally, the fault recording and analysis unit can record the occurrence of the fault, provide data support for subsequent maintenance and improvement, and reduce the probability of the fault recurring.

[0024] Secondly, in the present invention, the system uses an image acquisition and analysis unit and a quality assessment unit to accurately detect welding quality, identify welding defects such as pores, cracks, and lack of fusion, and the accuracy of the assessment is required to reach more than 90%. The defective product judgment unit can quickly determine whether the product is a defective product between 500 milliseconds and 1 second, thereby improving the detection efficiency. The rejection signal generation unit and the rejection execution unit can automatically reject defective products, and the rejection response time is between 500 milliseconds and 1 second. There is no need for manual screening and rejection, which improves production efficiency and automation, and reduces production costs and quality risks. At the same time, the data recording and statistical unit can record the situation of defective products and provide data support for quality control.

[0025] Thirdly, in the present invention, the system adopts a target analysis unit to formulate appropriate trajectory planning targets according to workpieces of different shapes, and can accurately analyze even workpieces of complex shapes. The initial path generation unit and the path optimization unit can generate the optimal trajectory to improve welding efficiency and quality stability. For example, for curved welds, a spline curve fitting algorithm can be used to generate a smooth curved trajectory, and the curvature change rate can be controlled within a reasonable range. The dynamic adjustment unit can adjust the trajectory in time according to the deformation and position change of the workpiece, and the adjustment response time is between 0.8 seconds and 1.2 seconds, thereby improving the accuracy and reliability of welding. In addition, the path storage and calling unit and the visual display unit facilitate the management and monitoring of the trajectory and improve work efficiency.

[0026] Fourthly, in the present invention, the system has many advantages in terms of no teaching. The data acquisition unit, feature extraction unit and model training unit can realize autonomous learning, and establish a welding pattern recognition model according to different welding tasks and workpiece characteristics, thereby improving the applicability and flexibility of the system. The intelligent adaptive decision unit can automatically adjust the welding template according to real-time data, and can also quickly adapt when the workpiece and the template in the template library are very different, thereby reducing manual intervention and improving the degree of automation. The feedback adjustment unit can continuously optimize the model and decision according to feedback information, thereby improving the stability and accuracy of the system. The human-computer interaction interface unit is convenient for operators to monitor and adjust, thereby improving the ease of use of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of the system module structure of the present invention;

[0028] Figure 2 This is a schematic diagram of the main control system module structure in the present invention;

[0029] Figure 3 This is a schematic diagram of the structure of the visual acquisition processing module in the present invention;

[0030] Figure 4 It is a structural schematic diagram of the teaching-free module in the present invention;

[0031] Figure 5 This is a schematic diagram of the trajectory planning module structure in the present invention;

[0032] Figure 6 This is a schematic diagram of the structure of the welding execution module in the present invention;

[0033] Figure 7 This is a schematic diagram of the structure of the troubleshooting processing module in the present invention;

[0034] Figure 8 It is a schematic diagram of the structure of the defective product detection and rejection module in the present invention. DETAILED DESCRIPTION

[0035] Example

[0036] See also Figure 1-Figure 8 In an embodiment of the present invention, a teaching-free welding trajectory planning method includes the following steps: Step 1: visual acquisition and processing, selecting an industrial camera with a resolution of 30 million pixels, configuring a lens focal length of 80mm-100mm according to the size of the welding workpiece and the working distance, designing a uniform lighting system with a light intensity of 1500Lux-2000Lux, using a multi-angle LED light source, the angle is set between 40°-50°, setting the image acquisition frequency to 80 frames / second-100 frames / second, graying the collected image, and then using a 4x4-5x5 median filter window to remove noise, and then adjusting the image contrast to 1.5 times-2.0 times to highlight the characteristics of the welding area, using the Canny edge detection algorithm to extract the edge features of the welding area, setting the low threshold to 80-100 and the high threshold to 180-220, and using the corner detection algorithm to extract key corner points;

[0037] Step 2: Data preparation for the teaching-free module, collect images, sensor data, welding process parameters and other information in various welding scenarios, for example, the integrated force sensor has a measurement accuracy of ±1N, the temperature sensor has a measurement accuracy of ±2℃, and the welding current measurement accuracy is ±1A, the voltage measurement accuracy is ±1V and other parameters are recorded, key features are extracted from the collected data, the feature extraction algorithm calculation time is controlled within 80 milliseconds to 120 milliseconds, the neural network algorithm is used to train the features, and a welding pattern recognition model is established. The training time is about 2 hours to 4 hours, and the model accuracy is required to reach more than 92%. Make intelligent adaptive decisions based on the trained model and real-time data. When the workpiece position changes between ±5mm-8mm, calculate the adjusted welding path and parameters, and the decision time is between 80 milliseconds and 100 milliseconds. Receive feedback information to adjust and optimize the teaching-free system, and the adjustment time is between 120 milliseconds and 150 milliseconds. Provide a human-computer interaction interface, and the interface response time is between 120 milliseconds and 150 milliseconds;

[0038] Step 3: Trajectory planning target analysis, analyze the target requirements of the welding task, the length of the straight weld is between 500mm-1000mm, and the error of the generated straight line trajectory is controlled within ±0.5mm; the curvature radius of the curved weld is between 100mm-200mm, and a smooth curved trajectory is generated, and the curvature change rate is controlled between 0.5 degrees / mm-1 degrees / mm, and the analysis time is between 1.5 seconds and 2.5 seconds. The initial welding path is generated according to the target analysis results and the initial position information of the workpiece. The linear interpolation algorithm is used for simple shape welds, and the generation time is between 1.5 seconds and 2 seconds; the spline curve fitting algorithm is used for complex curved welds, and the generation time is between 3 seconds and 5 seconds. The initial path is optimized, considering the welding speed The optimization time is between 8 seconds and 12 seconds, and the goal is to reduce the path length by 10%-15%. The welding path is dynamically adjusted according to real-time visual feedback and sensor data during welding. When the workpiece deformation is between ±3mm-5mm or welding defects are detected, timely adjustments are made. The adjustment response time is between 0.8 seconds and 1.2 seconds. The optimized path is stored for the next call. The storage capacity can store thousands of welding paths. The call time is between 100 milliseconds and 150 milliseconds. The welding path is visualized using a three-dimensional graphical interface with an interface refresh rate of 40 frames / second to 60 frames / second.

[0039] Step 4: Welding execution. Provide a suitable welding power supply with a welding current between 100A-150A and a voltage between 25V-30V according to the welding process and workpiece material. The power supply has stable output characteristics and adjustable parameters. For welding processes that require wire feeding, the welding gun accurately controls the wire feeding speed between 5m / min-10m / min. The welding gun angle and position are adjusted through mechanical devices to ensure that the welding arc acts stably on the weld. The movement of the welding equipment is controlled, including translation, rotation and other actions. The position accuracy can reach ±0.1mm and the angle accuracy can reach ±0.1°.

[0040] Step 5: Troubleshooting and processing. Integrate multiple sensors, such as current, voltage, and temperature sensors. The current sensor has a measurement accuracy of ±1A, the voltage sensor has a measurement accuracy of ±1V, and the temperature sensor has a measurement accuracy of ±2°C. Diagnose possible faults based on sensor data and system operation status. Use algorithms such as fault tree analysis. The diagnosis time is between 1 second and 3 seconds. Classify the diagnosed faults, formulate processing strategies for different types of faults, and the processing strategy formulation time is between 1 second and 2 seconds. Execute the processing strategy, and the processing response time is between 0.5 seconds and 1 second. Record the time, type, and processing measures of the fault for analysis;

[0041] Step 6: Detection and rejection of defective products. After welding is completed, use a high-resolution camera to capture images of the welding area, use image processing algorithms to analyze welding quality features in the image, and evaluate welding quality based on image analysis results and preset quality standards. The accuracy of the evaluation is required to reach more than 90%. Determine whether the welded product is a defective product based on the quality evaluation results. The judgment time is between 500 milliseconds and 1 second. When it is determined to be a defective product, a rejection signal is generated. The signal generation time is between 100 milliseconds and 200 milliseconds. The rejection signal is received to perform the rejection operation. The rejection response time is between 500 milliseconds and 1 second. Record the number, type, appearance time and other information of defective products for statistical analysis.

[0042] See also Figure 1-Figure 8 A welding system includes a main control system module, a power supply module, a visual acquisition processing module, a teaching-free module, a trajectory planning module, a welding execution module, a troubleshooting processing module and a defective product detection and rejection module, and the power supply module is electrically connected to the main control system module, the visual acquisition processing module, the teaching-free module, the trajectory planning module, the welding execution module, the troubleshooting processing module and the defective product detection and rejection module;

[0043] The main control system module includes the main control unit, the human-machine interaction unit and the communication unit. The human-machine interaction unit is connected to the main control unit through the communication unit. The main control unit is connected to the visual acquisition processing module, the teaching-free module, the trajectory planning module, the welding execution module, the troubleshooting processing module and the defective product detection and rejection module through the communication unit. The main control system module plays a core control role. The main control unit can receive and process data from each module, coordinate the work of each module, and ensure the orderly operation of the entire welding system. The human-machine interaction unit is connected to the main control unit through the communication unit, so that the operator can easily input instructions and view the system The communication unit ensures the fast and accurate transmission of signals between modules, and guarantees the real-time and stability of the system. The power supply module provides stable power support for the whole system, and the electrical connection with each module ensures the normal operation of the system is not affected by power problems. The visual acquisition and processing module collects images of the welding area through high-resolution equipment, and performs a series of processing, including graying, median filtering denoising, contrast adjustment, edge detection and corner detection, etc., to provide accurate image information for the subsequent welding process, which helps to improve the accuracy and quality of welding and ensure the accuracy of welding position and the quality of welds.

[0044] The teaching-free module can collect a variety of data, perform feature extraction and model training, establish a welding pattern recognition model, make intelligent adaptive decisions based on real-time data, reduce manual intervention, and improve the degree of automation. When the workpiece position changes or differs greatly from the template, it can quickly adjust the welding path and parameters to adapt to different welding tasks. This can improve production efficiency and reduce dependence on operator experience. The trajectory planning module first analyzes the welding task target requirements, uses different algorithms to generate the initial path for welds of different shapes, and then considers multiple factors for optimization. It can dynamically adjust the path according to real-time feedback during the welding process, and can also store and call the path and perform visual display to facilitate operator monitoring and management. This helps to improve welding efficiency and quality stability and ensure the optimality of the welding path.

[0045] The welding execution module provides appropriate power supply and controls the wire feeding speed and angle position of the welding gun according to the welding process and workpiece material to ensure that the welding arc acts stably on the weld, which can improve the quality and efficiency of welding and ensure the stability and reliability of the welding process. The troubleshooting processing module integrates multiple sensors and uses advanced algorithms to diagnose faults, classify faults and formulate processing strategies. It quickly executes processing strategies and records and analyzes fault information, which can promptly discover and resolve faults in the system, avoid welding interruptions and equipment damage, and improve the reliability and stability of the system. The defective product detection and rejection module collects image analysis quality after welding is completed, determines whether it is a defective product and automatically rejects it, and records statistical related information at the same time. This can improve product quality, reduce costs and quality risks, and provide data support for quality control.

[0046] See also Figure 1-Figure 4 The visual acquisition and processing module includes a camera configuration unit, a lighting unit, an image acquisition unit, an image preprocessing unit and a feature analysis unit. The configuration unit, the lighting unit, the image acquisition unit, the image preprocessing unit and the feature analysis unit are connected to the main control unit through the communication unit. The teaching-free module includes a data acquisition unit, a feature extraction unit, a model training unit and an intelligent adaptive decision-making unit. The data acquisition unit, the feature extraction unit, the model training unit and the intelligent adaptive decision-making unit are connected to the main control unit through the communication unit. In the visual acquisition and processing module, the camera configuration unit carefully selects a high-resolution camera and configures a lens with a suitable focal length according to different welding scenes and workpiece requirements. It can clearly capture the details of the welding area and provide a high-quality image source for subsequent processing.

[0047] The lighting unit has a reasonably designed lighting system and adopts a multi-angle light source arrangement to eliminate shadows and reflections to ensure the quality of image acquisition. The image acquisition unit controls the camera to acquire images at a specific frequency to capture dynamic changes in the welding process in real time. The acquisition frequency is flexibly adjusted according to the welding speed and the movement of the workpiece to ensure that key moments are not missed. The image preprocessing unit performs grayscale, median filtering denoising and contrast adjustment on the acquired images to reduce noise interference in the image, enhance the difference between the welding area and the background, highlight the image features, and provide a good foundation for subsequent feature analysis.

[0048] The feature analysis unit uses advanced edge detection and corner detection algorithms to accurately extract the edge features and key corners of the welding area, determine the position and shape of the weld, and provide important reference information for welding trajectory planning and quality inspection. Through the collaborative work of these units, the visual acquisition and processing module can provide accurate image information for the entire welding system and improve the accuracy and quality of welding. In the teaching-free module, the data acquisition unit widely collects various welding-related data, including image data, sensor data, and welding process parameters, etc., to provide a rich information basis for establishing an accurate welding pattern recognition model. The feature extraction unit extracts key features from a large amount of data and converts high-dimensional data into representative low-dimensional feature vectors, providing concise and effective data input for model training and intelligent decision-making. The model training unit uses advanced machine learning algorithms to train the extracted features and establish a model that can accurately identify different welding modes. Through a large amount of training data and optimized training algorithms, the accuracy and generalization ability of the model are continuously improved.

[0049] The intelligent adaptive decision-making unit makes intelligent adaptive decisions based on real-time collected data and trained models. When the workpiece position changes or there is a difference with the template, it can quickly calculate the adjusted welding path and parameters, reduce manual intervention, and improve the automation and adaptability of welding. Through the collaboration of these units, the teaching-free module can achieve autonomous learning and intelligent adaptation, thereby improving the automation level and flexibility of the welding system.

[0050] See also Figure 1-Figure 5The teaching-free module also includes a feedback adjustment unit and a human-computer interaction interface unit, which are connected to the main control unit through the communication unit. The trajectory planning module includes a target analysis unit, an initial path generation unit, a path optimization unit and a dynamic adjustment unit, which are connected to the main control unit through the communication unit. The trajectory planning module also includes a path storage and calling unit and a visualization display unit, which are connected to the main control unit through the communication unit. In the teaching-free module, the feedback adjustment unit can receive various feedback information during the welding process, such as welding quality evaluation results, sensor data changes, etc. According to these feedback information, the teaching-free system is adjusted and optimized, and the model and decision-making strategy are continuously improved, which helps to improve the stability and accuracy of the system, so that it can better adapt to different welding scenarios and workpiece changes.

[0051] The human-machine interaction interface unit provides an interactive channel between the operator and the teaching-free module. The operator can view the system's operating status, adjust parameters, intervene in decisions, etc. through this interface. This increases the flexibility and operability of the system and enables the operator to monitor and adjust the system when necessary. In the trajectory planning module, the target analysis unit deeply analyzes the target requirements of the welding task, including the shape, size, welding process requirements, etc. of the weld, clarifies the specific parameters and quality standards, and provides a clear direction for subsequent path generation and optimization. The initial path generation unit uses different algorithms to generate the initial welding path based on the results of the target analysis and the initial position information of the workpiece. For simple shapes Welds and complex curve welds use efficient linear interpolation algorithms and spline curve fitting algorithms respectively to ensure that the generated initial path meets the requirements of the welding task. The path optimization unit comprehensively considers multiple factors such as welding speed, robot kinematic constraints, energy consumption, etc., optimizes the initial path, and uses advanced optimization algorithms to reduce path length and improve welding efficiency. The path parameters are continuously adjusted to make the path more optimized. During the welding process, the dynamic adjustment unit monitors the deformation and position changes of the workpiece and various feedback information during the welding process in real time. According to these real-time feedback, the welding path is dynamically adjusted to ensure that welding is always carried out along the correct trajectory, thereby improving the accuracy and reliability of welding.

[0052] The path storage and calling unit stores the optimized welding path and establishes a path database. When encountering similar welding tasks, the stored path can be quickly called to improve work efficiency and reduce the time and workload of repeated planning. The visualization display unit intuitively displays the shape, position and movement process of the welding path through a three-dimensional graphical interface, so that the operator can clearly understand the planning of the welding path and facilitate monitoring and adjustment.

[0053] See also Figure 6-Figure 8 The welding execution module includes a welding power supply unit, a welding gun and a wire feeding unit, and a welding motion unit. The welding power supply unit, the welding gun and the wire feeding unit, and the welding motion unit are connected to the main control unit through a communication unit. The fault elimination processing module includes a sensor monitoring unit, a fault diagnosis unit, a fault classification unit, a processing strategy formulation unit, a fault processing execution unit, and a fault record analysis unit. The sensor monitoring unit, the fault diagnosis unit, the fault classification unit, the processing strategy formulation unit, the fault processing execution unit, and the fault record analysis unit are connected to the main control unit through a communication unit. The defective product detection and rejection module includes an image acquisition and analysis unit, a quality assessment unit, a monitoring feedback module, a welding parameter monitoring unit, a quality monitoring unit, a feedback control unit, a defective product judgment unit, a rejection signal generation unit, a rejection execution unit, and a data record The recording and statistical unit, the image acquisition and analysis unit, the quality assessment unit, the monitoring and feedback module, the welding parameter monitoring unit, the quality monitoring unit, the feedback control unit, the defective product determination unit, the rejection signal generation unit, the rejection execution unit and the data recording and statistical unit are connected to the main control unit through the communication unit. In the welding execution module, the welding power supply unit provides appropriate welding current and voltage according to the welding process and workpiece material requirements to ensure that the welding power supply output is stable and reliable to meet the needs of different welding tasks. The stability and consistency of the welding quality are ensured by accurately controlling the current and voltage. For the welding process that requires wire feeding, the welding gun and the wire feeding unit accurately control the wire feeding speed to match the welding wire supply with the welding process requirements. At the same time, the welding gun angle and position are adjusted through a mechanical device to allow the welding arc to act stably on the weld, thereby improving the welding efficiency and quality.

[0054] The welding motion unit controls the movement of the welding equipment, including translation, rotation and other actions, to ensure that the welding equipment moves accurately according to the planned path, realize the automation of the welding process, and ensure the accuracy of the welding position and the stability of the welding quality through high-precision motion control. In the troubleshooting processing module, the sensor monitoring unit integrates a variety of sensors, such as current sensors, voltage sensors, temperature sensors, etc., to monitor various parameters in the welding process in real time, and promptly discover potential problems such as welding equipment failures, electrical failures and mechanical failures, and provide data support for fault diagnosis and processing. The fault diagnosis unit adopts advanced algorithms, such as fault tree analysis, etc. According to the sensor monitoring data and the system operation status, it can quickly and accurately diagnose the type and location of the fault, reduce the troubleshooting time, and improve production efficiency. The fault classification unit classifies the diagnosed faults, formulates corresponding processing strategies for different types of faults, and improves the pertinence and efficiency of fault processing. The processing strategy formulation unit formulates specific processing strategies for different types of faults, including adjusting welding parameters, suspending welding for maintenance, and issuing alarms. The fault processing execution unit is responsible for executing the formulated processing strategies, actually processing the faults, solving the fault problems in a timely manner, and restoring the normal operation of the welding process.

[0055] The fault record analysis unit records the time, type, handling measures and other information of the fault, and analyzes it to find out the fault pattern and cause, provide a basis for system optimization and improvement, and reduce the probability of fault recurrence. In the defective product detection and rejection module, the image acquisition and analysis unit uses a high-resolution camera to capture the image of the welding area after welding is completed, and extracts the welding quality characteristics through image processing algorithm analysis image, providing data support for subsequent quality assessment and defective product determination. The quality assessment unit evaluates the welding quality according to the image analysis results and preset quality standards, determines whether the welding product meets the quality requirements, and provides a basis for defective product determination. The monitoring and feedback module includes a welding parameter monitoring unit, a quality monitoring unit and a feedback control unit. The welding parameter monitoring unit monitors the parameters of the welding process in real time. The quality monitoring unit detects the welding quality through various means. The feedback control unit adjusts the welding parameters and controls the welding process according to the monitoring results to ensure that the welding quality is stable and reliable. The defective product judgment unit determines whether the welding product is a defective product according to the quality assessment results, and promptly discovers unqualified products to prevent defective products from flowing into the next process. The rejection signal generation unit generates a rejection signal when it is judged to be a defective product, and provides a trigger signal for subsequent rejection operations. The rejection execution unit receives the rejection signal and executes the rejection operation to remove the defective products from the production line, thereby improving production efficiency and product quality. The data recording and statistical unit records the quantity, type, occurrence time and other information of defective products, and performs statistical analysis to provide data support for quality control, help enterprises find problems in the production process, and take corresponding improvement measures.

[0056] Visual acquisition and processing module

[0057] Image preprocessing unit-median filter algorithm (taking 5x5 window as an example)

[0058] For each pixel point (x, y), take the pixel value pij (i from x-2 to x+2, j from y-2 to y+2) within a 5×5 window centered on it.

[0059] Sort these 25 pixel values. Assume that the sorted array is A and the median is A

[12] (index starts from 0).

[0060] The filtered pixel value I(x,y)=A

[12] .

[0061] Feature analysis unit-Canny edge detection algorithm (simplified gradient calculation part)

[0062] The Sobel operator Gx in the horizontal direction is the matrix [-1, 0, 1; -2, 0, 2; -1, 0, 1], and the matrix Gy in the vertical direction is [-1, -2, -1; 0, 0, 0; 1, 2, 1].

[0063] For the pixel point (x, y) in image I, the gradient magnitude M(x, y) = Sqrt((GxI(x, y))^2 + (GyI(x, y))^2).

[0064] The gradient direction theta(x, y) = Atan((GyI(x, y)) / (GxI(x, y))).

[0065] Double - threshold processing: Let the low threshold be Tl and the high threshold be Th. If M(x, y) > Th, the pixel point is a strong edge point; if M(x, y) < Tl, it is a non - edge point; if Tl <= M(x, y) && M(x, y) <= Th, and it is adjacent to a strong edge point, then it is an edge point.

[0066] Teaching - free module

[0067] Feature extraction unit - Principal Component Analysis (PCA) algorithm (detailed steps)

[0068] Let the data set X = [x1, x2,..., xm], where xi is an n - dimensional vector and m is the number of samples.

[0069] Calculate the mean vector mu = (1 / m) * Sum(xi from 1 to m).

[0070] Decentralization: Xcentered = X - mu.

[0071] Calculate the covariance matrix C = (1 / (m - 1)) * (Xcentered transpose * Xcentered).

[0072] Perform eigenvalue decomposition on C to obtain eigenvalues lambda1 >= lambda2 >=... >= lambdan and corresponding eigenvectors v1, v2,..., vn.

[0073] Select the first k eigenvectors (k < n) to form the projection matrix P = [v1, v2,..., vk].

[0074] The projected feature Y = Xcentered * P.

[0075] Model training unit - Neural network algorithm (taking a simple single - hidden - layer as an example)

[0076] Let the input vector be x = (x1, x2,..., xn), the weight matrix from the input layer to the hidden layer W1(n×h, h is the number of neurons in the hidden layer), the bias vector of the hidden layer b1 = (b11, b12,..., b1h), and the activation function of the hidden layer be f1.

[0077] The hidden layer outputs h = (h1, h2, ..., hh), where hj = f1 (Sum (W1ij*xi from 1 to n) + b1j).

[0078] The weight matrix from the hidden layer to the output layer is W2 (h×m, m is the number of outputs), the output layer bias vector is b2=(b21, b22, ..., b2m), and the output layer activation function is f2.

[0079] Output vector y = (y1, y2, ..., ym), where yk = f2(Sum(W2jk*hj from 1 to h) + b2k).

[0080] Error calculation: Assume the true output is t = (t1, t2, ..., tm), mean square error E = (1 / m) * Sum ((yk-tk)^2 from 1 to m).

[0081] Backward Propagation:

[0082] For the output layer, delta2k = (yk-tk)*f2 derivative (Sum(W2jk*hj from 1 to h)+b2k).

[0083] For the hidden layer, delta1j = (Sum(delta2k*W2jk from 1 to m))*f1 derivative(Sum(W1 ij*xi from 1 to n)+b1j).

[0084] Weight update:

[0085] W1ij=W1ij-alpha*delta1j*xi.

[0086] W2jk=W2jk-alpha*delta2k*hj, alpha is the learning rate.

[0087] Trajectory planning module

[0088] Initial path generation unit - linear interpolation algorithm (three-dimensional space)

[0089] Assume the starting point coordinates are (xs, ys, zs), the end point coordinates are (xe, ye, ze), and the number of interpolation points is N.

[0090] For the i-th interpolation point (i from 1 to N):

[0091] xi=xs+(i / N)*(xe-xs).

[0092] yi=ys+(i / N)*(ye-ys).

[0093] zi=zs+(i / N)*(ze-zs).

[0094] Path Optimization Unit-Genetic Algorithm (Basic Steps)

[0095] Coding: Assuming that the path is encoded with real numbers, the path point coordinate sequence is P = (p1, p2, ..., pm), pi = (xi, yi, zi), then the chromosome can be represented as C = [x1, y1, z1, x2, y2, z2, ..., xm, ym, zm].

[0096] Population initialization: Randomly generate N chromosomes as the initial population P0 = {C1, C2, ..., CN}.

[0097] Fitness evaluation:

[0098] Suppose the fitness function f(C), for example f(C) = 1 / (L(C) + omega1*V(C) + omega2*K(C) + omega3*E(C)), where L(C) is the length of path C, V(C) is the degree of speed constraint violation, K(C) is the degree of kinematic constraint violation, E(C) is the energy consumption, and omega1, omega2, and omega3 are weights.

[0099] Path length L(C)=Sum(Sqrt((xi+1-xi)^2+(yi+1-yi)^2+(zi+1-zi)^2)from 1 to m-1).

[0100] Selection operation: Roulette wheel selection is used, and the selection probability pi of individual Ci = f(Ci) / Sum(f(Cj) from 1 to N).

[0101] Crossover operation: crossover is performed with probability pc, for example, single-point crossover is used, a crossover point k is randomly selected, and the second half of the two parent chromosomes Ca and Cb are exchanged to obtain the offspring chromosomes Ca' and Cb'.

[0102] Mutation operation: mutate with probability pm, for example, randomly perturb a gene (coordinate value) in the chromosome, xi'=xi+Deltax, where Deltax is a small random value.

[0103] Generation of a new generation of population: Generate a new generation of population Pt+1 through selection, crossover and mutation operations, and repeat the above steps until the stopping condition is met.

[0104] Troubleshooting Processing Module

[0105] Fault Diagnosis Unit - Fault Tree Analysis Algorithm (Probability Calculation Example)

[0106] For a simple fault tree, the occurrence probabilities of the bottom events A, B, and C are assumed to be P(A), P(B), and P(C), respectively.

[0107] If the top event T is connected by A and B through an AND gate and then connected with C through an OR gate to cause a fault, then the probability of the AND gate outputting event D is P(D) = P(A)*P(B).

[0108] The probability of the top event T is P(T)=1-(1-P(D))*(1-P(C)).

[0109] The working principle of the present invention is: the technical solution is mainly composed of visual acquisition and processing, a teaching-free module, a trajectory planning module, a welding execution module, a troubleshooting and processing module, and a defective product detection and rejection module. The modules work together to achieve an efficient and accurate welding process.

[0110] The visual acquisition and processing module collects images of the welding area through high-resolution industrial cameras, specific focal length lenses and optimized lighting systems, and then performs grayscale, median filter denoising, contrast adjustment, edge detection and corner detection to provide accurate image information for subsequent modules. For example, the camera configuration unit carefully selects high-resolution cameras and configures lenses with appropriate focal lengths to capture welding details; the lighting unit designs uniform and stable multi-angle LED light sources to eliminate shadow reflections; the image acquisition unit controls the camera to collect images at a specific frequency; the image preprocessing unit performs preliminary processing to highlight image features; and the feature analysis unit uses edge detection and corner detection algorithms to determine the position and shape of the weld.

[0111] The teaching-free module collects a variety of data for feature extraction and model training, establishes a welding pattern recognition model, and can make intelligent adaptive decisions based on real-time data, reduce manual intervention, and continuously optimize the system through feedback adjustment and human-computer interaction interface. The data acquisition unit widely collects data such as images, sensors, and process parameters; the feature extraction unit extracts key features from the data; the model training unit uses algorithms such as neural networks to train models to improve accuracy and generalization ability; the intelligent adaptive decision-making unit quickly adjusts welding path parameters based on real-time data and models; the feedback adjustment unit optimizes the system based on feedback information; the human-computer interaction interface unit facilitates operators to monitor and adjust;

[0112] The trajectory planning module first analyzes the welding task target requirements, uses different algorithms to generate the initial path for welds of different shapes, and then considers multiple factors for optimization. It can dynamically adjust the path based on real-time feedback, and can also store and call the path and perform visual display. The target analysis unit clarifies the weld parameters and quality standards; the initial path generation unit generates the initial path based on the target and workpiece information; the path optimization unit comprehensively considers multiple factors to optimize the path; the dynamic adjustment unit monitors and adjusts the path in real time; the path storage and call unit stores the path for easy call; and the visual display unit intuitively displays the path for easy monitoring and adjustment.

[0113] The welding execution module provides appropriate power according to the welding process and workpiece material, controls the wire feeding speed and angle position of the welding gun to ensure that the welding arc acts stably on the weld. The welding power supply unit provides stable and reliable current and voltage; the welding gun and wire feeding unit accurately control the wire feeding speed and adjust the angle position of the welding gun; the welding motion unit controls the movement of the equipment to achieve automated welding.

[0114] The troubleshooting and processing module integrates multiple sensors, uses advanced algorithms to diagnose and classify faults and formulate processing strategies, quickly executes processing strategies and records and analyzes fault information. The sensor monitoring unit monitors parameters in real time to detect potential faults; the fault diagnosis unit uses advanced algorithms to quickly and accurately diagnose the type and location of faults; the fault classification unit classifies faults to improve processing efficiency; the processing strategy formulation unit formulates strategies for different faults; the fault processing execution unit executes strategies to solve faults; and the fault recording and analysis unit records and analyzes faults to provide a basis for system optimization.

[0115] After welding is completed, the defective product detection and rejection module collects image analysis quality, determines whether it is a defective product and automatically rejects it, and records and statistics related information. The image acquisition and analysis unit collects images to extract quality features; the quality assessment unit evaluates the quality based on the image and standard; the monitoring feedback module includes parameter monitoring, quality monitoring and feedback control units to ensure stable welding quality; the defective product judgment unit judges the defective products based on the evaluation results; the rejection signal generation unit generates a rejection signal after judgment; the rejection execution unit executes the rejection operation; the data recording and statistics unit records and analyzes the defective product information to provide support for quality control.

[0116] This technical solution has many advantages. In terms of high-precision visual acquisition and processing, high-resolution cameras, specific focal length lenses combined with optimized lighting systems can clearly capture the details of the welding area, and multiple image processing algorithms highlight features to improve the accuracy of feature extraction. The powerful teaching-free function realizes autonomous learning and intelligent adaptation by collecting data training models, reducing manual intervention and improving the degree of automation. Feedback adjustment and human-computer interaction interface facilitate system optimization and operation monitoring. Efficient trajectory planning sets target requirements for welds of different shapes, uses different algorithms to generate and optimize the initial path, considers multiple factors to improve welding efficiency and quality stability, and dynamic adjustment, storage call and visual display functions improve accuracy, reliability and work efficiency. Comprehensive troubleshooting and processing as well as perfect defective product detection and rejection functions provide comprehensive protection for the welding process. Integrated sensors monitor faults in real time, advanced diagnostic algorithms quickly and accurately diagnose faults, classification processing strategies quickly handle faults to avoid interruption and damage, high-resolution cameras and algorithms accurately identify welding defects, and automated defective product judgment and rejection functions improve production efficiency and automation, reduce costs and quality risks, and data recording and statistics provide support for quality control.

[0117] The camera configuration unit can use Hikvision MV-CA050-10GC industrial camera, Basler acA4112-8gm industrial camera, the lighting unit can use Osram OslonSSLLED light source, the image acquisition unit can use NIPCIe-1433 image acquisition card, the image preprocessing unit can use OpenCV software library, the feature analysis unit can use OpenCV software library, the data acquisition unit can use Omron E2E series proximity sensor, Honeywell MLH series temperature sensor, Yokogawa WT3000 high-precision power analyzer, the feature extraction unit can use TensorFlow or PyTorch software library, the model training unit can use T The ensorFlow or PyTorch software library can be used, the intelligent adaptive decision-making unit can use rule-based custom decision-making system software or reinforcement learning algorithm software, the feedback adjustment unit can use custom gradient descent method-based software, the human-computer interaction interface unit can use Advantech TPC-1561H touch screen display, the target analysis unit can use AutoCAD (software), SolidWorks (software), the initial path generation unit can use ABB RobotStudio (software), MATLAB (software), the path optimization unit can use PyGMO software library, and the dynamic adjustment unit can use proportional-integral-derivative (PID) control algorithm software or model predictive control algorithm software.

[0118] The path storage calling unit can adopt a solid state drive (SSD) or a network storage device (NAS) (there is no specific model of the storage device), the visualization display unit can adopt Unity3D or UnrealEngine (software), the welding power supply unit can adopt Miller Electric Dynasty series welding power supply, the welding gun and wire feeding unit can adopt Lincoln Electric Magnum PRO welding gun, the welding motion unit can adopt ABBIRB series industrial robot, the sensor monitoring unit can adopt LEMIT series Hall current sensor, Texas Instruments INA219 voltage sensor, platinum resistance temperature sensor (PT100), and the fault diagnosis unit can use MATLAB or Python data analysis library.

[0119] The fault classification unit may use machine learning algorithm software such as support vector machine, decision tree, etc., and the processing strategy formulation unit may use customized fault processing strategy library software.

[0120] The fault handling execution unit can use control devices such as relays and solenoid valves, the fault record analysis unit can use MySQL, Oracle database management system, the image acquisition and analysis unit can use Basler acA4112-8gm industrial camera, the quality assessment unit can use support vector machine, random forest and other machine learning algorithm software, the monitoring feedback module can use PanasonicYM-350AII welding parameter monitor, ultrasonic flaw detector, eddy current flaw detector, PLC (programmable logic controller) or industrial computer, the defective product judgment unit can use custom judgment algorithm software, the rejection signal generation unit can use PLC or industrial computer, the rejection execution unit can use SMC pneumatic actuator, and the data recording and statistics unit can use MySQL, Oracle database management system.

Claims

1. A teaching-free welding trajectory planning method, characterized in that: The steps include: Step 1: Visual acquisition and processing: select an industrial camera with a resolution of 30 megapixels, configure a lens focal length of 80mm-100mm according to the size of the welding workpiece and the working distance, design a uniform lighting system with a light intensity of 1500Lux-2000Lux, use a multi-angle LED light source, set the angle between 40°-50°, set the image acquisition frequency to 80 frames / second-100 frames / second, grayscale the collected image, use a 4x4-5x5 median filter window to remove noise, and then adjust the image contrast to 1.5 times-2.0 times to highlight the characteristics of the welding area. Use the Canny edge detection algorithm to extract the edge features of the welding area, set the low threshold to 80-100, and the high threshold to 180-220. At the same time, use the corner detection algorithm to extract key corners; Step 2: Data preparation for the teaching-free module, collect images, sensor data, welding process parameters and other information in various welding scenarios, for example, the integrated force sensor has a measurement accuracy of ±1N, the temperature sensor has a measurement accuracy of ±2℃, and the welding current measurement accuracy is ±1A, the voltage measurement accuracy is ±1V and other parameters are recorded, key features are extracted from the collected data, the feature extraction algorithm calculation time is controlled within 80 milliseconds to 120 milliseconds, the neural network algorithm is used to train the features, and a welding pattern recognition model is established. The training time is about 2 hours to 4 hours, and the model accuracy is required to reach more than 92%. Make intelligent adaptive decisions based on the trained model and real-time data. When the workpiece position changes between ±5mm-8mm, calculate the adjusted welding path and parameters, and the decision time is between 80 milliseconds and 100 milliseconds. Receive feedback information to adjust and optimize the teaching-free system, and the adjustment time is between 120 milliseconds and 150 milliseconds. Provide a human-computer interaction interface, and the interface response time is between 120 milliseconds and 150 milliseconds; Step 3: Trajectory planning target analysis, analyze the target requirements of the welding task, the length of the straight weld is between 500mm-1000mm, and the error of the generated straight line trajectory is controlled within ±0.5mm; the curvature radius of the curved weld is between 100mm-200mm, and a smooth curved trajectory is generated, and the curvature change rate is controlled between 0.5 degrees / mm-1 degrees / mm, and the analysis time is between 1.5 seconds and 2.5 seconds. The initial welding path is generated according to the target analysis results and the initial position information of the workpiece. The linear interpolation algorithm is used for simple shape welds, and the generation time is between 1.5 seconds and 2 seconds; the spline curve fitting algorithm is used for complex curved welds, and the generation time is between 3 seconds and 5 seconds. The initial path is optimized, considering the welding speed The optimization time is between 8 seconds and 12 seconds, and the goal is to reduce the path length by 10%-15%. The welding path is dynamically adjusted according to real-time visual feedback and sensor data during welding. When the workpiece deformation is between ±3mm-5mm or welding defects are detected, timely adjustments are made. The adjustment response time is between 0.8 seconds and 1.2 seconds. The optimized path is stored for the next call. The storage capacity can store thousands of welding paths. The call time is between 100 milliseconds and 150 milliseconds. The welding path is visualized using a three-dimensional graphical interface with an interface refresh rate of 40 frames / second to 60 frames / second. Step 4: Welding execution. Provide a suitable welding power supply with a welding current between 100A-150A and a voltage between 25V-30V according to the welding process and workpiece material. The power supply has stable output characteristics and adjustable parameters. For welding processes that require wire feeding, the welding gun accurately controls the wire feeding speed between 5m / min-10m / min. The welding gun angle and position are adjusted through mechanical devices to ensure that the welding arc acts stably on the weld. The movement of the welding equipment is controlled, including translation, rotation and other actions. The position accuracy can reach ±0.1mm and the angle accuracy can reach ±0.1°. Step 5: Troubleshooting and processing. Integrate multiple sensors, such as current, voltage, and temperature sensors. The current sensor has a measurement accuracy of ±1A, the voltage sensor has a measurement accuracy of ±1V, and the temperature sensor has a measurement accuracy of ±2°C. Diagnose possible faults based on sensor data and system operation status. Use algorithms such as fault tree analysis. The diagnosis time is between 1 second and 3 seconds. Classify the diagnosed faults, formulate processing strategies for different types of faults, and the processing strategy formulation time is between 1 second and 2 seconds. Execute the processing strategy, and the processing response time is between 0.5 seconds and 1 second. Record the time, type, and processing measures of the fault for analysis; Step 6: Detection and rejection of defective products. After welding is completed, use a high-resolution camera to capture images of the welding area, use image processing algorithms to analyze welding quality features in the image, and evaluate welding quality based on image analysis results and preset quality standards. The accuracy of the evaluation is required to reach more than 90%. Determine whether the welded product is a defective product based on the quality evaluation results. The judgment time is between 500 milliseconds and 1 second. When it is determined to be a defective product, a rejection signal is generated. The signal generation time is between 100 milliseconds and 200 milliseconds. The rejection signal is received to perform the rejection operation. The rejection response time is between 500 milliseconds and 1 second. Record the number, type, appearance time and other information of defective products for statistical analysis.

2. A welding system, characterized in that: A teaching-free welding trajectory planning method according to claim 1 is adopted, characterized in that it includes a main control system module, a power supply module, a visual acquisition processing module, a teaching-free module, a trajectory planning module, a welding execution module, a troubleshooting processing module and a defective product detection and rejection module, and the power supply module is electrically connected to the main control system module, the visual acquisition processing module, the teaching-free module, the trajectory planning module, the welding execution module, the troubleshooting processing module and the defective product detection and rejection module; The main control system module includes a main control unit, a human-computer interaction unit and a communication unit. The human-computer interaction unit is connected to the main control unit through signals from the communication unit. The main control unit is connected to the visual acquisition processing module, the teaching-free module, the trajectory planning module, the welding execution module, the troubleshooting processing module and the defective product detection and rejection module through signals from the communication unit.

3. A welding system according to claim 1, characterized in that: The visual acquisition processing module includes a camera configuration unit, an illumination unit, an image acquisition unit, an image preprocessing unit and a feature analysis unit, and the configuration unit, the illumination unit, the image acquisition unit, the image preprocessing unit and the feature analysis unit are connected via a communication unit and a main control unit signal.

4. A welding system according to claim 1, characterized in that: The teaching-free module includes a data acquisition unit, a feature extraction unit, a model training unit and an intelligent adaptive decision unit, and the data acquisition unit, the feature extraction unit, the model training unit and the intelligent adaptive decision unit are connected through a communication unit and a main control unit signal.

5. A welding system according to claim 4, characterized in that: The teaching-free module also includes a feedback adjustment unit and a human-computer interaction interface unit, and the feedback adjustment unit and the human-computer interaction interface unit are connected via a communication unit and a main control unit signal.

6. A welding system according to claim 1, characterized in that: The trajectory planning module includes a target analysis unit, an initial path generation unit, a path optimization unit and a dynamic adjustment unit, and the target analysis unit, the initial path generation unit, the path optimization unit and the dynamic adjustment unit are connected through a communication unit and a main control unit signal.

7. A welding system according to claim 6, characterized in that: The trajectory planning module also includes a path storage and calling unit and a visualization display unit, and the path storage and calling unit and the visualization display unit are connected via a communication unit and a main control unit signal.

8. A welding system according to claim 1, characterized in that: The welding execution module includes a welding power supply unit, a welding gun and a wire feeding unit, and a welding motion unit, and the welding power supply unit, the welding gun and the wire feeding unit, and the welding motion unit are connected through a communication unit and a main control unit signal.

9. A welding system according to claim 1, characterized in that: The troubleshooting processing module includes a sensor monitoring unit, a fault diagnosis unit, a fault classification unit, a processing strategy formulation unit, a fault processing execution unit and a fault record analysis unit. The sensor monitoring unit, the fault diagnosis unit, the fault classification unit, the processing strategy formulation unit, the fault processing execution unit and the fault record analysis unit are connected through a communication unit and a main control unit signal.

10. A welding system according to claim 1, characterized in that: The defective product detection and rejection module includes an image acquisition and analysis unit, a quality assessment unit, a monitoring feedback module, a welding parameter monitoring unit, a quality monitoring unit, a feedback control unit, a defective product determination unit, a rejection signal generation unit, a rejection execution unit and a data recording and statistics unit. The image acquisition and analysis unit, the quality assessment unit, the monitoring feedback module, the welding parameter monitoring unit, the quality monitoring unit, the feedback control unit, the defective product determination unit, the rejection signal generation unit, the rejection execution unit and the data recording and statistics unit are connected to the main control unit through a communication unit signal.

Images