Intelligent high-performance collaborative welding robot system

Abstract

The invention provides an intelligent high-performance collaborative welding robot system, which belongs to the technical field of welding robots and comprises a multi-degree-of-freedom welding robot body, an intelligent control system, a high-performance welding power supply and auxiliary equipment. A welding process control module firstly obtains three-dimensional information of a workpiece based on visual inspection to construct a welding path; then selecting matched process parameters according to material characteristics and welding seam shapes; establishing a welding error matrix to monitor the welding process in real time; constructing a parameter error adjustment matrix for dynamic adjustment; analyzing parameter fluctuation by adopting a neural network to identify molten pool characteristics; the pre-trained neural network is used for evaluating the welding seam quality; and finally, a multi-robot collaborative welding strategy is integrated to achieve efficient operation, and the technical problem that in the prior art, welding process parameters in a welding robot are difficult to adjust accurately in real time so as to adapt to complex working condition changes is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of welding robots, and in particular, relates to an intelligent high-performance collaborative welding robot system. Background Art

[0002] Welding robots play an important role in modern industrial manufacturing, especially in the fields of automobiles, ships, aerospace, etc. Traditional welding robot systems usually use preset program control, which sets the welding trajectory and process parameters in advance and performs welding operations in combination with simple sensor feedback. Such systems perform well in standardized and repetitive welding tasks, which can improve production efficiency and maintain a certain stability in welding quality.

[0003] However, traditional welding robot systems have obvious shortcomings when dealing with complex and changeable welding environments. First, the preset parameters are difficult to adapt to material deformation, thermal deformation and position deviation during welding; second, the simple feedback control mechanism cannot effectively capture the dynamic characteristics of the welding pool, resulting in fluctuations in welding quality; third, the single robot operation mode limits production efficiency and flexibility. Especially in complex working conditions such as the connection of dissimilar materials and the welding of variable thickness components, the difficulty of welding quality control increases significantly.

[0004] At present, the industry is trying to improve the adaptability of welding robots by introducing more sensors and improving control algorithms, but there is still a lack of intelligent systems that can comprehensively analyze the complex dynamic characteristics of the welding process and accurately adjust the welding parameters in real time. The traditional feedback control method has a lag in response and cannot predict the risk of welding defects, while the simple parameter adjustment strategy is difficult to balance the multi-objective optimization requirements of welding quality and efficiency, resulting in the difficulty of accurately adjusting the welding process parameters in real time to adapt to complex working conditions. In other words, there is a technical problem in the existing technology that the welding process parameters in the welding robot are difficult to accurately adjust in real time to adapt to complex working conditions. Summary of the invention

[0005] In view of this, the present invention provides an intelligent high-performance collaborative welding robot system, which can solve the technical problem in the prior art that the welding process parameters of the welding robot are difficult to accurately adjust in real time to adapt to changes in complex working conditions.

[0006] The present invention is implemented as follows: the present invention provides an intelligent high-performance collaborative welding robot system including a multi-degree-of-freedom welding robot body, an intelligent control system, a high-performance welding power supply and auxiliary equipment; the intelligent control system includes a motion control module, a welding process control module, a visual detection and feedback module, a collaborative operation control module and a human-computer interaction interface; the welding process control module performs the following steps: obtaining three-dimensional information of the welding workpiece to construct a welding path; selecting matching welding process parameters; establishing a welding error matrix for real-time monitoring; constructing a welding parameter error adjustment matrix; using a neural network algorithm to analyze welding parameter fluctuations; introducing a welding molten pool fluid dynamics model; applying a welding quality multi-objective optimization function; and using a welding quality assessment neural network model for real-time assessment.

[0007] Among them, the multi-degree-of-freedom welding robot body has at least 6 degrees of freedom, including an articulated robotic arm and an end effector. The articulated robotic arm is made of high-strength lightweight material, and the end effector is equipped with a welding gun; the high-performance welding power supply and auxiliary equipment include a high-performance welding power supply, a wire feeding mechanism, and a gas protection device.

[0008] Among them, the steps of acquiring three-dimensional information of the welding workpiece and constructing the welding path by the welding process control module are: acquiring three-dimensional information of the welding workpiece based on the visual detection and feedback module, constructing a set of key points of the welding path, and generating a welding motion trajectory plan.

[0009] The step of selecting matching welding process parameters executed by the welding process control module is: selecting matching welding process parameters from a welding process database according to the welding workpiece material characteristics and the weld shape, and generating initial welding parameter settings.

[0010] Among them, the welding process control module executes the steps of establishing a welding error matrix for real-time monitoring: establishing a welding error matrix, real-time monitoring of weld position deviation, molten pool shape change and welding depth during the welding process, and forming welding status feedback data.

[0011] Among them, the welding process control module executes the steps of constructing a welding parameter error adjustment matrix: constructing a welding parameter error adjustment matrix, calculating welding parameter deviations based on the welding state feedback data, and adjusting welding current, welding voltage and welding speed.

[0012] Among them, the welding process control module executes the steps of using a neural network algorithm to analyze welding parameter fluctuations: using a neural network algorithm to analyze welding parameter fluctuations during the welding process, identifying the dynamic characteristics of the welding molten pool, and establishing a welding parameter dynamic compensation model; the welding parameter dynamic compensation model adopts a lightweight deep reinforcement learning model, and the specific structure is a hybrid neural network architecture composed of a multi-layer perceptron network and a long short-term memory network. The input layer receives welding state feedback data and welding parameter history data, the hidden layer contains multiple fully connected layers and long short-term memory network units, and the output layer generates welding parameter compensation values.

[0013] Among them, the welding process control module executes the step of introducing the welding molten pool fluid dynamics model: introducing the welding molten pool fluid dynamics model, analyzing the turbulent state of metal flow inside the molten pool, and predicting potential welding defect risks.

[0014] Among them, the steps of applying the welding quality multi-objective optimization function by the welding process control module are: applying the welding quality multi-objective optimization function to optimize welding parameters, and adjusting the welding parameter combination based on the dynamic characteristics of the welding pool and the welding defect risk assessment results.

[0015] Among them, the step of the welding process control module performing real-time evaluation using the welding quality assessment neural network model is: using a pre-trained welding quality assessment neural network model to perform real-time evaluation of the weld morphology to generate welding quality prediction indicators. The specific structure of the welding quality assessment neural network model is a hybrid network architecture combining a multi-layer convolutional neural network and a Transformer structure, including a weld image feature extraction module, a welding sound signal feature extraction module, and a multimodal information fusion module; the welding quality assessment neural network model also includes an attention mechanism, and the parameters of the attention mechanism are determined according to the material properties of the welding workpiece, the weld shape, and the welding process type.

[0016] Compared with the prior art, the present invention provides an intelligent high-performance collaborative welding robot system. The intelligent high-performance collaborative welding robot system proposed by the present invention integrates multimodal sensing, deep learning and fluid dynamics modeling technology to build a complete intelligent perception and control closed loop of the welding process. The system can obtain the three-dimensional information of the welding workpiece in real time, monitor the position deviation of the weld, the shape change of the molten pool and the welding depth, and analyze the fluctuation of welding parameters based on the neural network algorithm, establish a dynamic compensation model, and enable the welding parameters to be accurately adjusted according to the changes in working conditions.

[0017] Compared with traditional technologies, the present invention significantly improves the real-time and precise adjustment capability of welding parameters. First, by establishing a welding error matrix and a parameter error adjustment matrix, the system can quickly respond to deviations in the welding process; second, by introducing a welding molten pool fluid dynamics model to analyze the turbulent state of metal flow inside the molten pool, predictive control of potential welding defects is achieved; finally, by combining a lightweight deep reinforcement learning model and a multimodal information fusion mechanism, the system can adaptively focus on key moments and parameter changes in the welding process, balancing welding quality and energy efficiency.

[0018] Through the above-mentioned technical means, the present invention solves the technical problem in the prior art that the welding process parameters of the welding robot are difficult to accurately adjust in real time to adapt to changes in complex working conditions, ensuring that stable welding quality can be maintained in complex and changeable welding environments, especially in complex working conditions such as connecting dissimilar materials and welding of components with variable thickness, showing significant advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flow chart of the method of the present invention.

[0020] Figure 2 This is a schematic diagram of the composition of the intelligent high-performance collaborative welding robot system in Example 2. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0022] like Figure 1 FIG. 1 is a flow chart of an intelligent high-performance collaborative welding robot system provided by the present invention. The method comprises the following steps:

[0023] Including multi-degree-of-freedom welding robot body, intelligent control system, high-performance welding power supply and auxiliary equipment;

[0024] The multi-degree-of-freedom welding robot body has at least 6 degrees of freedom, including an articulated mechanical arm and an end effector, the articulated mechanical arm is made of high-strength lightweight materials, and the end effector is equipped with an advanced welding gun;

[0025] The intelligent control system includes a motion control module, a welding process control module, a visual detection and feedback module, a collaborative operation control module and a human-computer interaction interface;

[0026] The high-performance welding power source and auxiliary equipment include a high-performance welding power source, a wire feeding mechanism, and a gas protection device;

[0027] The welding process control module performs the following steps:

[0028] S01, based on the visual detection and feedback module, obtain the three-dimensional information of the welding workpiece, construct a set of key points of the welding path, and generate a welding motion trajectory plan;

[0029] S02, selecting matching welding process parameters from a welding process database according to the welding workpiece material properties and the weld shape, and generating initial welding parameter settings;

[0030] S03. Establish a welding error matrix to monitor the weld position deviation, molten pool shape change and welding depth in real time during the welding process to form welding status feedback data;

[0031] S04, constructing a welding parameter error adjustment matrix, calculating welding parameter deviations based on the welding state feedback data, and adjusting welding current, welding voltage and welding speed;

[0032] S05. Use neural network algorithm to analyze welding parameter fluctuations during welding, identify the dynamic characteristics of welding pool, and establish a dynamic compensation model for welding parameters;

[0033] S06. Introduce the welding pool fluid dynamics model to analyze the turbulent state of metal flow inside the molten pool and predict the potential welding defect risks;

[0034] S07, applying a welding quality multi-objective optimization function to optimize welding parameters, and adjusting welding parameter combinations based on the dynamic characteristics of the welding molten pool and the welding defect risk assessment results;

[0035] S08, using a pre-trained welding quality assessment neural network model to perform real-time assessment of the weld morphology to generate a welding quality prediction index, wherein the attention mechanism parameters in the welding quality assessment neural network model are determined according to the material properties of the welding workpiece, the weld shape and the welding process type;

[0036] S09, integrate multi-robot collaborative welding strategy, allocate welding areas based on task decomposition algorithm, and realize multi-robot synchronous welding operation;

[0037] The welding path key point set refers to a discrete point set sampled at preset intervals on the welding trajectory, and each key point contains spatial coordinate information and corresponding welding posture information, which is used to guide the motion trajectory planning of the multi-degree-of-freedom welding robot body;

[0038] The welding error matrix refers to a multidimensional data structure that describes the deviation between the actual weld position and the expected position during welding, including multiple parameters such as lateral deviation, longitudinal deviation, and angle deviation, which are used for welding accuracy control and error compensation;

[0039] The welding parameter error adjustment matrix refers to a conversion matrix that maps the welding error matrix to the welding parameter adjustment amount, and is used to automatically calculate and adjust the welding current, the welding voltage, the welding speed and other parameters according to the detected welding error;

[0040] The welding parameter dynamic compensation model refers to a welding parameter real-time adjustment model constructed based on the neural network algorithm, which can dynamically adjust the welding parameters according to the workpiece state changes and environmental factors during the welding process to improve the stability of welding quality;

[0041] The welding pool fluid dynamics model refers to a mathematical model that describes the flow behavior of metal liquid inside the welding pool, taking into account the influence of factors such as surface tension, buoyancy, and electromagnetic force on the morphology of the molten pool, and is used to predict the quality of weld formation;

[0042] The disordered state refers to the irregular flow state of the metal liquid inside the welding pool. Excessive disorder leads to welding defects such as pores and slag inclusions, which need to be controlled by optimizing welding parameters.

[0043] The multi-objective optimization function of welding quality is used to comprehensively consider the optimization objectives of multiple aspects of welding quality. The input includes the temperature distribution parameters of the welding molten pool, the metal flow velocity vector of the molten pool, the molten pool surface fluctuation amplitude index and the weld formation geometric parameters. The output is the recommended value of the optimal welding parameter combination and the corresponding welding quality prediction score.

[0044] The specific structure of the welding quality assessment neural network model is a hybrid network architecture combining a multi-layer convolutional neural network and a Transformer structure, including a weld image feature extraction module, a welding sound signal feature extraction module and a multimodal information fusion module, wherein the weld image feature extraction module uses a residual network structure to extract weld surface morphology features, the welding sound signal feature extraction module uses a one-dimensional convolutional network to extract the sound time-frequency features during the welding process, and the multimodal information fusion module integrates different modal information through a cross-attention mechanism and generates a final welding quality assessment result. The welding quality assessment neural network model has adaptive learning capabilities and can dynamically adjust network parameter weights according to different welding process types and material properties.

[0045] The step of establishing the training data set of the welding quality assessment neural network model specifically includes collecting a large amount of welding process data under different materials and different process parameters, each set of data includes welding process image sequences, welding sound signals, welding parameter records and corresponding welding quality rating labels, using professional welding process experts to manually rate the welding quality, and combining non-destructive testing technology to quantitatively analyze welding internal defects to form a standardized training data set, and using data enhancement technology to expand the training samples to improve the generalization ability of the model;

[0046] The steps of training the welding quality assessment neural network model specifically include first training on a large-scale general welding data set to learn the general feature representation of welding quality assessment, then fine-tuning training for specific welding processes and materials, using cross-validation methods to evaluate model performance and optimize model hyperparameters, improving model robustness by introducing adversarial training, and finally verifying the model in an actual welding environment, and continuously optimizing and updating the model based on the verification results. The specific welding process and material are determined based on the selection of the trainer.

[0047] The motion control module is used to control the motion of the multi-degree-of-freedom welding robot body, including joint motion control, Cartesian space trajectory control and position posture control, to achieve precise positioning and posture adjustment of the welding gun in three-dimensional space;

[0048] The welding process control module is used to control the process parameters during the welding process, including welding current, welding voltage, welding speed, wire feeding speed and shielding gas flow rate, so as to achieve fine control of the welding process;

[0049] The visual inspection and feedback module is used to obtain the three-dimensional information of the welding workpiece and the molten pool state during the welding process in real time, including information such as weld position, weld shape, molten pool morphology and welding depth, and provide feedback data for welding process control;

[0050] The collaborative operation control module is used to coordinate the collaborative operation of multiple welding robots, realize multi-robot synchronous welding, and improve welding efficiency and quality;

[0051] The human-machine interaction interface is used to realize the interaction between the operator and the welding robot system, including functions such as welding parameter setting, welding path planning, welding process monitoring and welding result display.

[0052] The welding parameter dynamic compensation model adopts a lightweight deep reinforcement learning model. The specific structure is a hybrid neural network architecture composed of a multi-layer perceptron network and a long short-term memory network. The input layer receives welding state feedback data and welding parameter history data. The hidden layer contains multiple fully connected layers and long short-term memory network units. The fully connected layer is used to extract the nonlinear mapping relationship between welding parameters and welding quality. The long short-term memory network unit is used to capture the timing characteristics and long-term dependencies of the welding process. The output layer generates welding parameter compensation values, and combines the compensation values ​​with basic welding parameters through a residual connection mechanism to form the final welding parameter control instructions. At the same time, an attention mechanism is introduced to perform weighted processing on welding state information at different time steps, so that the model can adaptively focus on key moments and key parameter changes in the welding process. The network adopts batch normalization and Dropout technology to improve training stability and model generalization ability. The loss function is designed as a weighted combination of welding quality evaluation indicators and energy consumption. The welding quality and energy efficiency are balanced through a multi-objective optimization method. The specific steps of establishing the training data set are: first First, the welding process data under various welding process conditions are collected, including welding parameter records and corresponding welding quality evaluation results of different materials, different thicknesses and different welding positions. Then, the collected raw data are preprocessed, including data cleaning, outlier detection, data normalization and time series alignment. Then, the sliding window method is used to divide the continuous welding process data into training samples of fixed length. Each sample contains a welding parameter sequence and a corresponding welding quality score. In order to solve the problem of data imbalance, stratified sampling and data enhancement techniques are used to expand the number of samples of rare welding conditions. At the same time, expert knowledge is introduced to guide the data annotation process. Based on the experience of welding experts, the mapping relationship between welding parameters and welding quality is constructed. Finally, the entire data set is divided into training set, validation set and test set for model training, hyperparameter tuning and performance evaluation. During the training process, the cross-validation method is used to evaluate the performance stability of the model under different data partitions, and the model parameters are continuously updated through the online learning mechanism, so that the model can adapt to changes in welding process and aging of equipment, thereby improving the stability and consistency of welding quality.

[0053] The specific implementation of the above steps is described in detail below.

[0054] The system includes a multi-degree-of-freedom welding robot body, an intelligent control system, a high-performance welding power supply and auxiliary equipment. The multi-degree-of-freedom welding robot body adopts a 6-axis articulated mechanical arm structure with a load capacity of 10kg and a repeatability accuracy of ±0.05

[0055] ±0.05mm, the maximum working radius is 1800mm; the joint drive adopts high-precision servo motors, equipped with 20-bit absolute encoders; the end effector is equipped with a water-cooled welding gun with automatic adjustment function. The intelligent control system adopts a real-time operating system platform, the main controller uses an industrial-grade computer, the processor frequency is not less than 3.5GHz, and the memory capacity is not less than 16GB; the visual inspection module includes a high-speed industrial camera and a structured light projection device; the human-computer interaction interface uses a 19-inch touch screen with a resolution of 1920×10801920

[0056] 1920×1080. The high-performance welding power source is an inverter digital control welding power source with a rated output current of 400A and a no-load voltage of 80V. The output characteristics can be switched between constant current and constant voltage modes. The wire feeding mechanism adopts a four-wheel drive structure with a wire feeding speed range of 2 to 20m / min. The gas protection device is equipped with an electronic flow controller with a flow range of 5 to 25L / min.

[0057] The welding process control module performs the following steps:

[0058] The specific implementation of step S01 is the process of obtaining three-dimensional information of the welding workpiece based on the visual detection and feedback module. First, the structured light technology is used to perform three-dimensional scanning on the surface of the welding workpiece to obtain the point cloud data of the workpiece surface; then the point cloud data collected from multiple perspectives are fused into a complete three-dimensional model of the workpiece through the point cloud registration algorithm; then the edge detection algorithm is used to extract features from the workpiece surface and identify the edge contour of the weld; the least squares curve fitting method is further applied to construct the welding path curve equation; then the key point set of the welding path is generated by sampling at intervals of 5 to 10 mm on the curve; finally, based on these key points, a smooth and continuous welding motion trajectory is generated using the cubic spline interpolation algorithm. The purpose of this step is to provide accurate motion path planning for the welding robot to ensure the accuracy of the welding position.

[0059] The specific implementation method of step S02 is to select welding process parameters according to the material properties of the welding workpiece and the shape of the weld. First, the physical property parameters of the workpiece material are retrieved from the database, including melting point, thermal conductivity, specific heat capacity, etc.; then the cross-sectional area of ​​the weld is calculated according to the weld shape parameters; then based on the material properties and weld geometry, the fuzzy reasoning system is used to select the most matching welding process parameter template from the welding process database; further combined with the workpiece thickness and welding position, the weight coefficient correction method is applied to optimize and adjust the template parameters, and the initial welding current range is 80~320A, the welding voltage range is 18~32V, and the welding speed range is 5~15mm / s. The function of this step is to provide reasonable initial welding parameters and lay the foundation for subsequent welding process control.

[0060] The specific implementation method of step S03 is to establish a welding error matrix and monitor the welding process in real time. First, define a four-dimensional error matrix including lateral deviation, longitudinal deviation, angle deviation and depth deviation; then use a high-speed camera to capture the welding process image in real time, with an acquisition frequency of 60 to 120 Hz; then use image processing technology to extract the actual position and theoretical position of the weld; further use the Kalman filter algorithm to filter the detection results to eliminate the influence of random noise; at the same time, monitor the temperature distribution of the molten pool through an infrared thermal imager, with a temperature detection range of 800 to 1600 ° C; finally, calculate the deviation values ​​of each dimension based on the image analysis results and temperature distribution data, and update the welding error matrix. The purpose of this step is to obtain real-time status information of the welding process and provide a basis for subsequent parameter adjustments.

[0061] The specific implementation method of step S04 is to construct a welding parameter error adjustment matrix and adjust the welding parameters. First, a linear mapping relationship matrix from the error vector to the welding parameter adjustment amount is established; then, the historical welding data is regressed and analyzed based on the least squares method to determine the value of each element in the matrix; then, the welding parameter deviation is calculated based on the welding state feedback data using an adaptive control algorithm; further, the welding current adjustment step is set to 3 to 8A, the welding voltage adjustment step is set to 0.5 to 1.5V, and the welding speed adjustment step is set to 0.5 to 2mm / s; finally, the adjusted parameters are transmitted to the welding power supply control system in real time through the serial communication interface. The role of this step is to achieve closed-loop control of welding parameters and improve the stability of the welding process.

[0062] The specific implementation method of step S05 is to use a neural network algorithm to analyze welding parameter fluctuations. First, a feedforward neural network with three hidden layers is constructed, with 64, 32, and 16 nodes in each layer respectively; then the time series data of welding current, welding voltage, and welding speed are used as input, and the morphological characteristics of the molten pool are used as output; then the network is trained using a back propagation algorithm, and the learning rate is set to 0.001-0.005; the trained network model is further used to analyze the impact of welding parameter fluctuations on the dynamic characteristics of the molten pool in real time; finally, a welding parameter dynamic compensation model is established based on the analysis results to achieve predictive adjustment of welding parameters. The purpose of this step is to deeply analyze the parameter fluctuation law in the welding process and provide theoretical support for refined welding control.

[0063] The specific implementation method of step S06 is to introduce the welding pool fluid dynamics model for analysis. First, a group of molten pool fluid dynamics equations considering surface tension, buoyancy, and electromagnetic force is established; then the finite element method is used to discretize the molten pool area into grid units with a grid size of 0.1 to 0.5 mm; then the temperature field and velocity field at each unit are calculated according to the welding process parameters; the Reynolds number is further introduced as a judgment indicator of the flow turbulence state, and the critical value is set to 2000; finally, the high turbulence area is identified by calculating the Reynolds number distribution, and the potential risk of defects such as pores and slag inclusions is predicted. The role of this step is to theoretically analyze the internal flow state of the molten pool and predict the generation mechanism of welding defects.

[0064] The specific implementation method of step S07 is to apply the welding quality multi-objective optimization function to optimize parameters. First, a multi-objective optimization function including weld formation quality, penetration consistency, porosity and energy efficiency is constructed; then a particle swarm algorithm is used to optimize parameters, the number of particles is set to 50-100, and the maximum number of iterations is 200; then, optimization constraints are set based on the dynamic characteristics of the welding pool and the defect risk assessment results; further search is performed in the three-dimensional parameter space of welding current, welding voltage, and welding speed to obtain the Pareto optimal solution set; finally, the parameter combination with the best comprehensive performance is selected from the optimal solution set as the recommended value. The purpose of this step is to seek the best balance between multiple welding quality indicators and provide the best welding parameter combination.

[0065] The specific implementation method of step S08 is to evaluate the weld morphology using a pre-trained neural network model. First, the welding quality evaluation neural network model is called; then the weld image is collected in real time through a high-resolution camera with a resolution of not less than 1920×1080 pixels; then the image is preprocessed, including grayscale, contrast enhancement and noise filtering; the processed image is further input into the neural network model; finally, the welding quality score is generated according to the model output result, with a score range of 0 to 100, where 85 points or more are excellent welds, 70 to 85 points are qualified welds, and below 70 points are unqualified welds. The purpose of this step is to evaluate the welding quality in real time and provide feedback for welding process control.

[0066] The specific implementation method of step S09 is to integrate the multi-robot collaborative welding strategy. First, the overall welding task is decomposed into multiple subtasks based on the geometric shape of the welded workpiece; then the task allocation algorithm is used to allocate welding areas to each robot to balance the workload of each robot; then the robot collaborative motion plan is formulated to avoid collisions between robots; further, a communication protocol between robots is established to achieve real-time sharing of status information; finally, the master-slave control architecture is used to coordinate the synchronous operation of multiple robots, and the welding progress of the master robot is used as the synchronization benchmark, and the progress deviation of the slave robot does not exceed 10%. The purpose of this step is to improve welding efficiency and quality and shorten the welding cycle.

[0067] The detailed structure of the dynamic compensation model of welding parameters is a hybrid neural network composed of a multi-layer perceptron network and a long short-term memory network. The input layer of the model receives 16 dimensions of welding state data and historical parameter data; the hidden layer contains 3 fully connected layers and 2 long short-term memory network layers; the number of nodes in the fully connected layer is 128, 64, and 32 respectively, and the activation function uses the ReLU function; the number of units in the long short-term memory network layer is 64, which is used to learn the timing characteristics of welding parameter changes; the output layer generates compensation values ​​for welding current, welding voltage, and welding speed; the model uses a residual connection mechanism to combine the compensation value with the basic parameters; at the same time, the attention mechanism is introduced to give higher weight to the state information at critical moments; the network optimization uses the Adam optimizer, the learning rate is 0.001, and the training batch size is 32. The process of establishing the training data set includes collecting welding process data of five common materials at different thicknesses and welding positions; cleaning, normalizing and time-series aligning the original data; using a sliding window of 10s to split the training samples; expanding rare working condition samples through data enhancement technology; and finally dividing the training set, validation set and test set into a ratio of 7:2:1.

[0068] The detailed structure of the neural network model for welding quality assessment is a hybrid network that combines a multi-layer convolutional neural network with a Transformer structure. The image feature extraction module uses a ResNet-50 residual network structure, which contains 5 residual blocks; the sound signal feature extraction module uses a one-dimensional convolutional network, which contains 4 convolutional layers, and the convolution kernel size decreases from 16 to 4; the multimodal information fusion module uses a 6-layer Transformer encoder structure with 8 attention heads; the final output layer is a fully connected layer with 5 nodes, corresponding to 5 welding quality levels. The training data set establishment process includes collecting 100,000 sets of welding data under different process parameters; each set of data contains a welding process image sequence, sound signal and parameter record; inviting 5 professional welding process experts to rate the welding quality; combining X-ray and ultrasonic detection technology to quantitatively analyze internal defects; using image rotation, scaling, brightness adjustment and other methods to enhance data; and finally forming a standardized training data set containing 150,000 sets of samples.

[0069] Furthermore, the specific implementation of the welding robot is described as follows: the multi-degree-of-freedom welding robot body adopts a 6-axis articulated mechanical arm structure with a load capacity of 10kg, a repeated positioning accuracy of ±0.05mm, and a maximum working radius of 1800mm; the joint drive adopts a permanent magnet synchronous servo motor, equipped with a 20-bit absolute encoder, and the angle resolution can reach 0.00034 degrees; the mechanical arm body is made of high-strength aluminum alloy and titanium alloy composite materials, which reduces weight by 30% while maintaining unchanged rigidity; the maximum rotation speed of each joint can reach 210 degrees / second, and the acceleration can reach 2500 degrees / second2 The end effector is equipped with a water-cooled welding gun with a cooling water flow rate of 6 to 8 L / min. The cooling efficiency can control the welding gun temperature below 65°C. The welding gun adopts a modular design, and the replacement time is less than 30 seconds. It has an automatic angle adjustment function with an adjustment range of ±15 degrees.

[0070] The optional high-performance welding power source is an IGBT inverter digital control welding power source with a rated output current of 400A, a maximum pulse current of 600A, a no-load voltage of 80V, and a response time of less than 1ms; the output characteristics can be switched between constant current, constant voltage and mixed modes; the power efficiency is greater than 88%, and the power factor is greater than 0.93; it supports a variety of welding methods, including MAG, MIG, TIG and pulse MIG welding; the pulse frequency can be adjusted from 30 to 500Hz, and the duty cycle can be adjusted from 10% to 90%. The wire feeding mechanism adopts a four-wheel drive structure, and the driving wheel is made of special alloy steel with a hardness of HRC58 to 62; the wire feeding speed range is 2 to 20m / min, and the speed stability is ±1.5%; it can adapt to various types of welding wires with a diameter of 0.8 to 1.6mm. The gas protection device is equipped with an electronic flow controller with a flow range of 5 to 25L / min and a flow accuracy of ±2%; the gas preheater can heat the gas to 10 to 15°C above normal temperature to reduce welding spatter; the gas circuit system adopts a quick connector design with a pressure loss of less than 5%.

[0071] Optionally, the whole system power supply is three-phase 380V AC, with a total power of less than 15kW; the modular base design allows for quick installation and disassembly; the equipment occupies an area of ​​less than 2 square meters; the protection level reaches IP54, which can adapt to dust and moisture in industrial environments; the surface of the robot arm adopts a special anti-splash coating, and the service life can reach more than 3 years; the control cabinet adopts a forced air cooling system, and the internal temperature is controlled below 40°C; all cable connectors use military-grade waterproof connectors to ensure stable and reliable signal transmission.

[0072] Optionally, the specific implementation methods of each module of the intelligent control system are detailed as follows: the motion control module adopts an algorithm architecture based on model predictive control (MPC), with a control period of 1ms, a position control accuracy of less than ±0.03mm, and a speed control accuracy of less than ±1%; the module includes a feedforward compensation unit and a feedback adjustment unit, the feedforward compensation unit calculates the joint torque based on the robot dynamics model, and the feedback adjustment unit corrects the control amount in real time based on the encoder feedback signal; the module has a built-in collision detection function, and when a torque change exceeding a threshold (usually set to 20% of the rated torque) is detected, the movement can be stopped within 10ms; at the same time, a compliance control algorithm is integrated to allow the robot to exhibit controllable compliance characteristics in a preset direction, and the compliance can be adjusted in the range of 0 to 100%.

[0073] Optionally, the visual inspection and feedback module is equipped with two high-speed industrial cameras and a structured light projector. The camera resolution is 2448×2048 pixels and the maximum acquisition frequency is 120 frames per second. The structured light projector uses a blue LED light source, the projection pattern is pseudo-random stripes, and the projection accuracy can reach 0.05mm. The module uses a combination of binocular stereo vision algorithm and phase shift fringe analysis algorithm to achieve three-dimensional reconstruction of welding workpieces. The visual processing unit uses GPU accelerated computing, equipped with 8GB video memory, and the three-dimensional reconstruction speed can reach 15 frames per second. The module also integrates a molten pool monitoring subsystem, including a high-speed camera with narrow-band filtering and an infrared thermal imager, which can monitor the molten pool morphology and temperature distribution in real time.

[0074] Optionally, the collaborative operation control module adopts a hierarchical control architecture, including a task planning layer, a trajectory planning layer and an execution control layer; the task planning layer decomposes the welding task into multiple subtasks based on the workpiece CAD model and welding process requirements, and assigns them to each robot; the trajectory planning layer uses an improved fast mosaic method to generate a collision-free trajectory with a planning cycle of 100ms; the execution control layer is responsible for the synchronous coordination of the movements of each robot, and the synchronization error is controlled within ±5ms; the module adopts a distributed computing architecture, each robot is equipped with a local controller, and communicates through real-time Ethernet, with a communication delay of less than 1ms; the system supports up to 8 robots working together at the same time.

[0075] Optionally, the human-computer interaction interface uses a 19-inch capacitive touch screen with a resolution of 1920×1080 and a touch response time of less than 15ms; the interface is developed based on HTML5 and WebGL technology, and supports real-time rendering of 3D scenes; a graphical welding path editing tool is provided, and the operator can adjust the welding trajectory by dragging and dropping; the system has a built-in welding process knowledge base and provides parameter recommendation function; the interface supports multi-language switching, including simplified Chinese, English and Russian; it has remote monitoring function, and the welding status can be viewed remotely through mobile devices.

[0076] The mathematical model or calculation process involved in the present invention is described in detail below.

[0077] The process of constructing the welding path key point set in step S01 is specifically expressed as follows:

[0078] P = {p1, p2, ..., p n};

[0079] Where P is the set of key points of the welding path; p i is the coordinate and posture information of the i-th key point, p i =[x i ,y i , z i , α i , βi , γ i ] T ;x i ,y i , z i are the position coordinates of the key points in the spatial rectangular coordinate system, in mm; α i , β i , γ i are the posture angles of the welding gun at the key points, in degrees; n is the total number of key points, usually 20 to 100.

[0080] The construction of the welding path curve equation adopts the cubic spline interpolation algorithm, which is expressed as follows:

[0081] S j (t) = a j +b j (tt j )+c j (tt j ) 2 +d j (tt j ) 3 , t∈[t j , t j+1 ];

[0082] In the formula, S j (t) is the j-th segment spline curve function; t is the curve parameter, t∈[0,1]; a j , b j , c j , d j are the spline function coefficients, obtained by solving the following system of equations:

[0083] S j (t j )=p j ;

[0084] S j (t j+1 )=p j+1 ;

[0085] S j ′(t j+1 )=S j+1 ′(t j+1 );

[0086] S j ″(t j+1 )=S j+1 ″(t j+1 );

[0087] Among them, S j ′(t) and S j″(t) represents S j The first and second order derivatives of (t). The coefficient matrix is ​​solved by the pursuit method, and the computational complexity is O(n).

[0088] The welding error matrix in step S03 is defined as follows:

[0089]

[0090] Where, E is the welding error matrix; e x The lateral deviation is in mm, which indicates the offset between the actual weld position and the planned position perpendicular to the welding direction, ranging from -2.0 to 2.0 mm; y The longitudinal deviation is in mm, which indicates the offset between the actual weld position and the planned position in the welding direction, ranging from -3.0 to 3.0 mm; θ is the angle deviation, in degrees, indicating the angle between the actual posture of the welding gun and the planned posture, ranging from -5.0 to 5.0 degrees; d It is the depth deviation in mm, which indicates the difference between the actual melting depth and the target melting depth, ranging from -1.0 to 1.0 mm.

[0091] The calculation of welding error adopts image processing technology, which is specifically expressed as follows:

[0092] e x =x actual -x planned ;

[0093] e y =y actual -y planned ;

[0094] e θ =θ actual -θ planned ;

[0095] e d =d actual -d planned ;

[0096] In the formula, (x actual ,y actual ) is the actual position coordinate of the weld obtained through image processing; (x planned ,y planned ) is the theoretical position coordinate on the welding planning path; θ actual is the actual welding angle; θ planned To plan the welding angle; d actual is the actual melting depth obtained through the analysis of the molten pool morphology; d planned The target melting depth required by the process.

[0097] The extraction of position coordinates uses a method combining edge detection and Hough transform, which is expressed as:

[0098]

[0099] Where G(x, y) is the image gradient amplitude; G x (x, y) and G y (x, y) are the gradients of the image in the x and y directions respectively; θ(x, y) is the gradient direction angle. When G(x, y) is greater than the threshold T (usually T is 1.5 times the average gradient of the image), the point is considered to be an edge point.

[0100] The Kalman filter algorithm is used for filtering welding errors. The state equation and observation equation are as follows:

[0101] X k =AX k-1 +BU k +w k ;

[0102] Z k =HX k +v k ;

[0103] Where, X k is the state vector at the current moment, Contains the error and its rate of change; A is the state transfer matrix; U k is the control input; B is the control matrix; w k is process noise, which obeys a Gaussian distribution with a mean of 0 and a covariance of Q; Z k is the observation vector, The superscript m represents the measurement value; H is the observation matrix; v k is the observation noise, which obeys a Gaussian distribution with a mean of 0 and a covariance of R.

[0104] The welding parameter error adjustment matrix in step S04 is defined as follows:

[0105] ΔP = CE;

[0106] Where ΔP is the welding parameter adjustment, ΔP = [ΔI, ΔU, Δv] T ; ΔI is the welding current adjustment, unit is A; ΔU is the welding voltage adjustment, unit is V; Δv is the welding speed adjustment, unit is mm / s; C is the error adjustment matrix, dimension is 3×4, which represents the mapping relationship from welding error to parameter adjustment.

[0107] The error adjustment matrix C is solved by the least square method based on historical welding data:

[0108] C=(E T E) -1 E T ΔP history ;

[0109] Where E is the error matrix sample in multiple sets of historical welding data; ΔP history The elements of the matrix C have clear physical meanings, such as C 11 Indicates the influence coefficient of lateral deviation on welding current adjustment, C 23 Indicates the influence coefficient of angle deviation on welding voltage adjustment.

[0110] The updating formula of welding parameters is:

[0111] I k+1 =I k +K I ΔI;

[0112] U k+1 =U k +K U ΔU

[0113] v k+1 =v k +k v Δv;

[0114] In the formula, I k , U k 、v k They are the current welding current, voltage and speed respectively; I k+1 , U k+1 、v k+1 are the adjusted welding current, voltage and speed respectively; K I , K U , K v They are the adjustment gain coefficients of current, voltage and speed, usually K I The value is 0.5~0.8, K U The value is 0.6~0.9, K v The value is 0.4 to 0.7. These gain coefficients are obtained through welding experiment optimization, which includes parameter adjustment tests under different workpiece materials and weld types, recording the changes in weld quality before and after adjustment, and determining the optimal gain coefficient based on the quality improvement rate.

[0115] The molten pool fluid dynamics model in step S06 is described by the Navier-Stokes equation:

[0116]

[0117] Where ρ is the density of the metal liquid, in kg / m 3; is the flow velocity vector, The unit is m / s; t is time, the unit is s; p is pressure, the unit is Pa; μ is dynamic viscosity, the unit is Pa·s; is the body force, including buoyancy, electromagnetic force, etc., For buoyancy, Where β is the coefficient of thermal expansion, g is the acceleration of gravity, T is the local temperature, and T0 is the reference temperature; is the electromagnetic force, in is the current density, is the magnetic induction intensity; is the Marangoni force caused by the surface tension gradient, Where γ is the surface tension coefficient.

[0118] The formula for calculating the Reynolds number is:

[0119]

[0120] In the formula, Re is the Reynolds number, dimensionless; ρ is the metal liquid density; v is the characteristic flow velocity, usually the maximum flow velocity in the molten pool; L is the characteristic length, usually the diameter of the molten pool; μ is the dynamic viscosity. When Re>2000, it is considered that the flow enters a turbulent state and welding defects may occur.

[0121] The welding quality multi-objective optimization function in step S07 is defined as follows:

[0122] F(I,U,v)=w1Q f (I, U, v)+w2Q p (I, U, v)+w3[1-P d (I, U, v)]+w4E e (I, U, v);

[0123] Where, F(I, U, v) is the comprehensive optimization objective function; I, U, v are welding current, voltage and speed respectively; Q f Q is the weld quality score, ranging from 0 to 1; p P is the penetration consistency score, ranging from 0 to 1; d is the defect rate, ranging from 0 to 1; E eis the energy efficiency score, ranging from 0 to 1; w1, w2, w3, and w4 are weight coefficients, satisfying w1+w2+w3+w4=1, usually w1=0.35, w2=0.25, w3=0.3, and w4=0.1. These weight coefficients can be adjusted according to specific application requirements. When the weld appearance requirements are high, w1 should be increased; when the welding strength requirements are high, w2 should be increased; when the reliability requirements are high, w3 should be increased; and when the energy cost is sensitive, w4 should be increased.

[0124] The calculation of each sub-objective function is as follows:

[0125]

[0126] P d (I, U, v) = α1P pore (I, U, v)+α2P crack (I, U, v)+α3P slag (I, U, v);

[0127]

[0128] In the formula, w a is the weld width, w a，target is the target width, w a，max is the maximum allowable width; h r is the residual height, h r，target is the target residual height, h r，max is the maximum allowable residual height; σ d is the standard deviation of penetration depth, d mean is the average penetration depth; P pore , P crack and P slag are the probabilities of pores, cracks and slag inclusions, respectively, predicted by the modified turbulence function, α1, α2, α3 are weight coefficients, satisfying α1+α2+α3=1; η melt is the actual melting efficiency, η max is the theoretical maximum melting efficiency.

[0129] Optionally, the feedforward neural network model of the neural network analyzing the welding parameter fluctuation in step S05 is expressed as follows:

[0130] H1=σ(W1X+b1);

[0131] H2=σ(W2H1+b2);

[0132] H3=σ(W3H2+b3);

[0133] Y=W4H3+b4;

[0134] Where X is the input vector, which contains the time series data of welding current, welding voltage and welding speed, X = [I1, I2, ..., I n , U1, U2, ..., U n , v1, v2, ..., v n ] T ; H1, H2, H3 are the output vectors of the three hidden layers, with dimensions of 64, 32, 16 respectively; Y is the output vector, representing the morphological characteristics of the melt pool; W1, W2, W3, W4 are weight matrices; b1, b2, b3, b4 are bias terms; σ is the activation function, using the ReLU function, σ(x) = max(0, x).

[0135] Optionally, the neural network is trained using the back-propagation algorithm, and the loss function is defined as:

[0136]

[0137] Where L is the loss function value; m is the number of training samples; Y i is the actual output of the i-th sample; is the predicted output of the i-th sample; λ is the regularization coefficient, which is 0.001; ||W j ||2 is the L2 norm of the weight matrix, which is used to prevent overfitting.

[0138] Optionally, the task allocation algorithm for multi-robot collaborative welding in step S09 adopts the following mathematical model:

[0139] minZ=max i＝1，2，...，k T i ;

[0140]

[0141] x ij ∈{0, 1}, i=1, 2,...,k; j=1, 2,...,n;

[0142] Where Z is the optimization target, i.e., the maximum completion time; k is the number of robots; n is the number of welding tasks; T i is the total working time of the ith robot; t j is the execution time of the jth welding task; x ij is the decision variable. When the jth task is assigned to the i-th robot, x ij =1, otherwise x ij = 0. This optimization problem is an NP-hard problem, and a greedy algorithm is used to find an approximate optimal solution.

[0143] Optionally, the robot collaborative motion planning adopts an artificial potential field-based method, and the potential field function for avoiding collision is defined as:

[0144]

[0145] Where U total is the total potential field energy; U rep (d ij ) is the repulsive potential field between the ith robot and the jth robot; d ij is the minimum distance between the two robots; d0 is the safety distance threshold, usually 1000mm; η is the proportional coefficient, which is 10 6 The robot's motion planning needs to ensure U while minimizing the working time. total Less than the safety threshold.

[0146] Specifically, the principle of the present invention is:

[0147] The technical principle of the present invention to solve the problem of real-time and precise adjustment of welding process parameters is based on the closed-loop intelligent control framework of "perception-analysis-prediction-optimization-execution". Under this framework, the system first obtains the three-dimensional information of the welding workpiece and the state of the molten pool during the welding process through the visual detection and feedback module, constructs the welding error matrix, and realizes accurate perception of the welding process; then, based on the neural network algorithm, the welding parameter fluctuations are analyzed, the dynamic characteristics of the welding molten pool are identified, and the turbulent state of the metal flow inside the molten pool is analyzed in combination with the welding molten pool fluid dynamics model, so as to realize in-depth analysis of the welding process.

[0148] The prediction link is the key breakthrough of the present invention. By establishing a dynamic compensation model for welding parameters, the system can predict the welding quality performance under different welding parameter combinations. The model adopts a lightweight deep reinforcement learning architecture, integrating multi-layer perceptron and long short-term memory network, and can capture the nonlinear mapping relationship and timing characteristics between welding parameters and welding quality. The introduction of the attention mechanism enables the model to adaptively focus on changes in key parameters, overcoming the lag of traditional feedback control.

[0149] In the optimization phase, the system applies a multi-objective optimization function for welding quality, comprehensively considers the temperature distribution of the welding pool, the metal flow rate, the surface fluctuation amplitude, and the geometric parameters of the weld formation, and generates the optimal welding parameter combination. This optimization process balances welding quality and energy efficiency, and meets the multi-dimensional requirements of actual industrial applications. Finally, the optimized parameters are executed through the welding process control module to achieve precise control of welding current, voltage, speed and other parameters.

[0150] The scientificity and logic of the technical solution of the present invention lie in: first, it is based on the scientific understanding of the physical process of welding, combined with the principles of fluid dynamics and materials science to establish a welding pool behavior model; second, it uses modern artificial intelligence technology to deal with high-dimensional nonlinear problems in the welding process, and improves the system robustness by combining data-driven and model-driven methods; finally, it constructs a complete multi-level control architecture, from key point path planning to parameter fine-tuning, to form a coordinated control strategy to ensure that the system can effectively respond to various welding challenges.

[0151] A specific embodiment 1 of the present invention is provided below, and the specific implementation method of each step in this embodiment 1 is described in detail as follows.

[0152] The specific implementation method of step S01 is the process of obtaining three-dimensional information of the welding workpiece based on the visual detection and feedback module. First, the structured light technology is used to perform a three-dimensional scan on the surface of the welding workpiece to obtain the point cloud data of the workpiece surface; then, the point cloud data collected from multiple perspectives are fused into a complete three-dimensional model of the workpiece through the point cloud registration algorithm; then, the edge detection algorithm is used to extract features from the workpiece surface and identify the edge contour of the weld; the least squares curve fitting method is further applied to construct the welding path curve equation; then, the curve is sampled at intervals of 5 to 10 mm to generate a set of key points of the welding path; finally, based on these key points, a cubic spline interpolation algorithm is used to generate a smooth and continuous welding motion trajectory. The welding path key point set is specifically expressed as follows: P = {p1, p2, ..., p n}; where P is the set of key points of the welding path; p i is the coordinate and posture information of the i-th key point, p i =[x i ,y i , z i , α i , β i , γ i ] T ;x i ,y i , z i are the position coordinates of the key points in the spatial rectangular coordinate system, in mm; α i , β i , γ i are the posture angles of the welding gun at the key points, in degrees; n is the total number of key points, usually 20 to 100. The construction of the welding path curve equation adopts the cubic spline interpolation algorithm, which is expressed as follows: S j (t) = a j +b j (tt j )+c j (tt j ) 2 +dj (yt j ) 3 , t∈[t j , t j+1 ]; where S j (t) is the j-th segment spline curve function; t is the curve parameter, t∈[0,1]; a j , b j , c j , d j are the spline function coefficients, obtained by solving the following system of equations: j (t j )=p j ; S j (t j+1 )=p j+1 ; S j ′(t j+1 )=S j+1 ′(t j+1 );S j ″(t j+1 )=S j+1 ″(t j+1 ), where S j ′(t) and S j ″(t) represents S j The first and second order derivatives of (t). The coefficient matrix is ​​solved by the pursuit method, and the computational complexity is O(n). The purpose of this step is to provide accurate motion path planning for the welding robot to ensure the accuracy of the welding position.

[0153] The specific implementation method of step S02 is to select welding process parameters according to the material properties of the welding workpiece and the shape of the weld. First, the physical property parameters of the workpiece material are retrieved from the database, including melting point, thermal conductivity, specific heat capacity, etc.; then the cross-sectional area of ​​the weld is calculated according to the weld shape parameters; then based on the material properties and weld geometry, the fuzzy reasoning system is used to select the most matching welding process parameter template from the welding process database; further combined with the workpiece thickness and welding position, the weight coefficient correction method is applied to optimize and adjust the template parameters, and the initial welding current range is 80~320A, the welding voltage range is 18~32V, and the welding speed range is 5~15mm / s. The function of this step is to provide reasonable initial welding parameters and lay the foundation for subsequent welding process control.

[0154] The specific implementation method of step S03 is to establish a welding error matrix and monitor the welding process in real time. First, define a four-dimensional error matrix including lateral deviation, longitudinal deviation, angle deviation and depth deviation; then use a high-speed camera to capture the welding process image in real time, with an acquisition frequency of 60 to 120 Hz; then use image processing technology to extract the actual position and theoretical position of the weld; further use the Kalman filter algorithm to filter the detection results to eliminate the influence of random noise; at the same time, monitor the temperature distribution of the molten pool through an infrared thermal imager, and the temperature detection range is 800 to 1600 ° C; finally, calculate the deviation values ​​of each dimension based on the image analysis results and temperature distribution data, and update the welding error matrix. The welding error matrix is ​​defined as follows: Where, E is the welding error matrix; e x The lateral deviation is in mm, which indicates the offset between the actual weld position and the planned position perpendicular to the welding direction, ranging from -2.0 to 2.0 mm; y The longitudinal deviation is in mm, which indicates the offset between the actual weld position and the planned position in the welding direction, ranging from -3.0 to 3.0 mm; θ is the angle deviation, in degrees, indicating the angle between the actual posture of the welding gun and the planned posture, ranging from -5.0 to 5.0 degrees; d is the depth deviation, in mm, indicating the difference between the actual penetration depth and the target penetration depth, ranging from -1.0 to 1.0 mm. The calculation of welding error adopts image processing technology, which is specifically expressed as follows: x =x actual -x planned ;e y =y actual -y planned ;e θ =θ actual -θ planned ;e d =d actual -d planned ; In the formula, (x actual ,y actual ) is the actual position coordinate of the weld obtained through image processing; (x planned ,y planned ) is the theoretical position coordinate on the welding planning path; θ actual is the actual welding angle; θ planned To plan the welding angle; d actual is the actual melting depth obtained through the analysis of the molten pool morphology; d planned is the target melting depth required by the process. The position coordinates are extracted by combining edge detection and Hough transform, which is expressed as: Where G(x, y) is the image gradient amplitude; G x (x, y) and Gy (x, y) are the gradients of the image in the x and y directions respectively; θ(x, y) is the gradient direction angle. When G(x, y) is greater than the threshold T (usually T is 1.5 times the average gradient of the image), the point is considered to be an edge point. The Kalman filter algorithm is used for filtering welding errors. The state equation and observation equation are as follows: X k =AX k-1 +BU k +w k ; Z k =HX k +v k Where X k is the state vector at the current moment, Contains the error and its rate of change; A is the state transfer matrix; U k is the control input; B is the control matrix; w k is process noise, which obeys a Gaussian distribution with a mean of 0 and a covariance of Q; Z k is the observation vector, The superscript m represents the measurement value; H is the observation matrix; v k For observation noise, it obeys a Gaussian distribution with a mean of 0 and a covariance of R. The purpose of this step is to obtain real-time status information of the welding process and provide a basis for subsequent parameter adjustment.

[0155] The specific implementation method of step S04 is to construct a welding parameter error adjustment matrix and adjust the welding parameters. First, a linear mapping relationship matrix from the error vector to the welding parameter adjustment amount is established; then, a regression analysis is performed on the historical welding data based on the least squares method to determine the value of each element in the matrix; then, an adaptive control algorithm is used to calculate the welding parameter deviation based on the welding state feedback data; further, the welding current adjustment step is set to 3~8A, the welding voltage adjustment step is set to 0.5~1.5V, and the welding speed adjustment step is set to 0.5~2mm / s; finally, the adjusted parameters are transmitted to the welding power supply control system in real time through the serial communication interface. The welding parameter error adjustment matrix is ​​defined as follows: ΔP=CE; where ΔP is the welding parameter adjustment amount, ΔP=[ΔI, ΔU, Δv] T ; ΔI is the welding current adjustment, in A; ΔU is the welding voltage adjustment, in V; Δv is the welding speed adjustment, in mm / s; C is the error adjustment matrix, with a dimension of 3×4, which represents the mapping relationship from welding error to parameter adjustment. The error adjustment matrix C is solved by the least squares method based on historical welding data: C=(E T E) -1 E T ΔP history ; Where E is the error matrix sample in multiple sets of historical welding data; ΔP historyThe elements of the matrix C have clear physical meanings, such as C 11 Indicates the influence coefficient of lateral deviation on welding current adjustment, C 23 Indicates the influence coefficient of angle deviation on welding voltage adjustment. The update formula of welding parameters is: k+1 =I k +K I ΔI; U k+1 =U k +K U ·ΔU;v k+1 =v k +K v ·Δv; where I k , U k 、v k They are the current welding current, voltage and speed respectively; I k+1 , U k+1 、v k+1 are the adjusted welding current, voltage and speed respectively; K I , K U , K v They are the adjustment gain coefficients of current, voltage and speed, usually K I The value is 0.5~0.8, K U The value is 0.6~0.9, K v The value is 0.4 to 0.7. These gain coefficients are obtained through welding experiment optimization, which includes parameter adjustment tests under different workpiece materials and weld types, recording the changes in weld quality before and after adjustment, and determining the optimal gain coefficient based on the quality improvement rate. The role of this step is to achieve closed-loop control of welding parameters and improve the stability of the welding process.

[0156] The specific implementation method of step S05 is to use a neural network algorithm to analyze welding parameter fluctuations. First, a feedforward neural network with three hidden layers is constructed, and the number of nodes in each layer is 64, 32, and 16 respectively; then the time series data of welding current, welding voltage, and welding speed are used as input, and the morphological characteristics of the molten pool are used as output; then the network is trained using a back propagation algorithm, and the learning rate is set to 0.001-0.005; the trained network model is further used to analyze the influence of welding parameter fluctuations on the dynamic characteristics of the molten pool in real time; finally, a welding parameter dynamic compensation model is established based on the analysis results to achieve predictive adjustment of welding parameters. The feedforward neural network model for analyzing welding parameter fluctuations by neural network is expressed as follows: H1 = σ(W1X+b1); H2 = σ(W2H1+b2); H3 = σ(W3H2+b3); Y = W4H3+b4; where X is the input vector, including the time series data of welding current, welding voltage, and welding speed, and X = [I1, I2, ..., I n , U1, U2, ..., Un , v1, v2, ..., v n ] T ; H1, H2, H3 are the output vectors of the three hidden layers, with dimensions of 64, 32, and 16 respectively; Y is the output vector, representing the morphological characteristics of the melt pool; W1, W2, W3, and W4 are weight matrices; b1, b2, b3, and b4 are bias terms; σ is the activation function, using the ReLU function, σ(x) = max(0, x). The neural network training uses the back propagation algorithm, and the loss function is defined as: Where L is the loss function value; m is the number of training samples; Y i is the actual output of the i-th sample; is the predicted output of the i-th sample; λ is the regularization coefficient, which is 0.001; ||W j ||2 is the L2 norm of the weight matrix, which is used to prevent overfitting. The purpose of this step is to deeply analyze the parameter fluctuation law in the welding process and provide theoretical support for refined welding control.

[0157] The specific implementation method of step S06 is to introduce the welding pool fluid dynamics model for analysis. First, a group of molten pool fluid dynamics equations considering surface tension, buoyancy, and electromagnetic force is established; then the finite element method is used to discretize the molten pool area into grid units with a grid size of 0.1 to 0.5 mm; then the temperature field and velocity field at each unit are calculated according to the welding process parameters; the Reynolds number is further introduced as a judgment indicator of the flow turbulence state, and the critical value is set to 2000; finally, the high turbulence area is identified by calculating the Reynolds number distribution, and the potential risk of defects such as pores and slag inclusions is predicted. The molten pool fluid dynamics model is described by the Navier-Stokes equation: Where ρ is the density of the metal liquid, in kg / m 3 ; is the flow velocity vector, The unit is m / s; t is time, the unit is s; p is pressure, the unit is Pa; μ is dynamic viscosity, the unit is Pa·s; is the body force, including buoyancy, electromagnetic force, etc., For buoyancy, Where β is the coefficient of thermal expansion, g is the acceleration of gravity, T is the local temperature, and T0 is the reference temperature; is the electromagnetic force, in is the current density, is the magnetic induction intensity; is the Marangoni force caused by the surface tension gradient, Where γ is the surface tension coefficient. The Reynolds number is calculated as: Where Re is the Reynolds number, dimensionless; ρ is the metal liquid density; v is the characteristic flow velocity, usually the maximum flow velocity in the molten pool; L is the characteristic length, usually the diameter of the molten pool; μ is the dynamic viscosity. When Re>2000, it is considered that the flow enters a turbulent state and welding defects may occur. The purpose of this step is to theoretically analyze the flow state inside the molten pool and predict the mechanism of welding defects.

[0158] The specific implementation method of step S07 is to apply the welding quality multi-objective optimization function to optimize parameters. First, a multi-objective optimization function including weld formation quality, penetration consistency, porosity and energy efficiency is constructed; then a particle swarm algorithm is used to optimize parameters, the number of particles is set to 50-100, and the maximum number of iterations is 200; then, optimization constraints are set based on the dynamic characteristics of the welding pool and the defect risk assessment results; further search is performed in the three-dimensional parameter space of welding current, welding voltage, and welding speed to obtain the Pareto optimal solution set; finally, the parameter combination with the best comprehensive performance is selected from the optimal solution set as the recommended value. The welding quality multi-objective optimization function is defined as follows: F(I, U, v) = w1Q f (I, U, v)+w2Q p (I, U, v)+w3[1-P d (I, U, v)]+w4E e (I, U, v); where F(I, U, v) is the comprehensive optimization objective function; I, U, v are welding current, voltage and speed respectively; Q f Q is the weld quality score, ranging from 0 to 1; p P is the penetration consistency score, ranging from 0 to 1; d is the defect rate, ranging from 0 to 1; E e is the energy efficiency score, ranging from 0 to 1; w1, w2, w3, and w4 are weight coefficients, satisfying w1+w2+w3+w4=1, usually w1=0.35, w2=0.25, w3=0.3, and w4=0.1. These weight coefficients can be adjusted according to specific application requirements. When the weld appearance requirements are high, w1 is increased; when the welding strength requirements are high, w2 is increased; when the reliability requirements are high, w3 is increased; when the energy cost is sensitive, w4 is increased. The calculation of each sub-objective function is as follows: P d (I, U, v) = α1P pore (I, U, v)+α2P crack (I, U, v)+α3P slag (I, U, v); In the formula, w a is the weld width, w a，target is the target width, w a，max is the maximum allowable width; h r is the residual height, hr，target is the target residual height, h r，max is the maximum allowable residual height; σ d is the standard deviation of penetration depth, d mean is the average penetration depth; P pore , P crack and P slag are the probabilities of pores, cracks and slag inclusions, respectively, predicted by the modified turbulence function, α1, α2, α3 are weight coefficients, satisfying α1+α2+α3=1; η melt is the actual melting efficiency, η max The purpose of this step is to find the best balance among multiple welding quality indicators and provide the best welding parameter combination.

[0159] The specific implementation method of step S08 is to evaluate the weld morphology using a pre-trained neural network model. First, the welding quality evaluation neural network model is called; then the weld image is collected in real time through a high-resolution camera with a resolution of not less than 1920×1080 pixels; then the image is preprocessed, including grayscale, contrast enhancement and noise filtering; the processed image is further input into the neural network model; finally, the welding quality score is generated according to the model output result, with a score range of 0 to 100, where 85 points or more are excellent welds, 70 to 85 points are qualified welds, and below 70 points are unqualified welds. The purpose of this step is to evaluate the welding quality in real time and provide feedback for welding process control.

[0160] The specific implementation method of step S09 is to integrate the multi-robot collaborative welding strategy. First, the overall welding task is decomposed into multiple subtasks based on the geometric shape of the welding workpiece; then the task allocation algorithm is used to allocate welding areas to each robot to balance the workload of each robot; then the robot collaborative motion plan is formulated to avoid collisions between robots; further establish a communication protocol between robots to achieve real-time sharing of status information; finally, the master-slave control architecture is used to coordinate the synchronous operation of multiple robots, and the welding progress of the master robot is used as the synchronization benchmark. The progress deviation of the slave robot does not exceed 10%. The task allocation algorithm for multi-robot collaborative welding adopts the following mathematical model: minZ=max i＝1，2，...，k T i ; x ij ∈{0, 1}, i = 1, 2, ..., k; j = 1, 2, ..., n; where Z is the optimization target, i.e., the maximum completion time; k is the number of robots; n is the number of welding tasks; T i is the total working time of the ith robot; t j is the execution time of the jth welding task; x ij is the decision variable. When the jth task is assigned to the i-th robot, xij =1, otherwise x ij =0. This optimization problem is an NP-hard problem, and a greedy algorithm is used to find the approximate optimal solution. The robot collaborative motion planning adopts a method based on artificial potential field, and the potential field function to avoid collision is defined as: Where U total is the total potential field energy; U rep (d ij ) is the repulsive potential field between the ith robot and the jth robot; d ij is the minimum distance between the two robots; d0 is the safety distance threshold, usually 1000mm; η is the proportional coefficient, which is 10 6 The robot's motion planning needs to ensure U while minimizing the working time. total The purpose of this step is to improve welding efficiency and quality and shorten the welding cycle.

[0161] The detailed structure of the dynamic compensation model of welding parameters is a hybrid neural network composed of a multi-layer perceptron network and a long short-term memory network. The input layer of the model receives 16 dimensions of welding state data and historical parameter data; the hidden layer contains 3 fully connected layers and 2 long short-term memory network layers; the number of nodes in the fully connected layer is 128, 64, and 32 respectively, and the activation function uses the ReLU function; the number of units in the long short-term memory network layer is 64, which is used to learn the timing characteristics of welding parameter changes; the output layer generates compensation values ​​for welding current, welding voltage, and welding speed; the model uses a residual connection mechanism to combine the compensation value with the basic parameters; at the same time, the attention mechanism is introduced to give higher weight to the state information at critical moments; the network optimization uses the Adam optimizer, the learning rate is 0.001, and the training batch size is 32. The process of establishing the training data set includes collecting welding process data of five common materials at different thicknesses and welding positions; cleaning, normalizing and time-series aligning the original data; using a sliding window of 10s to split the training samples; expanding rare working condition samples through data enhancement technology; and finally dividing the training set, validation set and test set into a ratio of 7:2:1.

[0162] The detailed structure of the neural network model for welding quality assessment is a hybrid network that combines a multi-layer convolutional neural network with a Transformer structure. The image feature extraction module uses a ResNet-50 residual network structure, which contains 5 residual blocks; the sound signal feature extraction module uses a one-dimensional convolutional network, which contains 4 convolutional layers, and the convolution kernel size decreases from 16 to 4; the multimodal information fusion module uses a 6-layer Transformer encoder structure with 8 attention heads; the final output layer is a fully connected layer with 5 nodes, corresponding to 5 welding quality levels. The training data set establishment process includes collecting 100,000 sets of welding data under different process parameters; each set of data contains a welding process image sequence, sound signal and parameter record; inviting 5 welding process experts to rate the welding quality; combining X-ray and ultrasonic detection technology to quantitatively analyze internal defects; using image rotation, scaling, brightness adjustment and other methods to enhance data; and finally forming a standardized training data set containing 150,000 sets of samples.

[0163] In order to better understand and implement the present invention, Example 2 of a specific application scenario of the present invention is provided below: In the production process of large-scale aviation parts, researchers applied the present invention to the welding process of high-strength aluminum alloy structural components. This component is a connection structure between the wing truss and the skin of a certain type of aircraft. It is made of 7075 aluminum alloy material, with a length of 3.6m, a complex structure, a total weld length of more than 12m, and includes a variety of welding positions and weld forms. In view of the difficulties in the welding process of this component, the researchers built an intelligent high-performance collaborative welding robot system to achieve high-quality and efficient welding production.

[0164] The welding robot system in this embodiment is composed of Figure 2 As shown in the figure, the system hardware configuration uses three ABB IRB4600 6-axis articulated robots, each with a load of 20kg and a repeatability of ±0.05mm. The robot arm is made of carbon fiber composite material, which reduces the weight by 28% and has a maximum working radius of 2.55m. The end effector is equipped with a water-cooled TPS5000MIG / MAG welding gun with a cooling water flow of 7.2L / min. The welding gun adopts a modular design and has a ±12-degree automatic angle adjustment function. The robot control cabinet uses IRC5 multi-machine model, integrates high-precision robot control software, and has a minimum interpolation cycle of 0.4ms.

[0165] The welding power source adopts Fronius TPS 5000 CMT digital pulse MIG welding power source, with a rated output current of 500A, a maximum pulse current of 680A, a no-load voltage of 70V, and a response time of less than 0.8ms. The power source supports the CMT (cold metal transition) process, which can effectively reduce heat input, deformation and spatter. The wire feeding mechanism adopts a four-wheel drive structure, with a wire feeding speed range of 2 to 25m / min and a speed stability of ±1.2%, which is suitable for 1.2mm ER5356 aluminum alloy welding wire. The gas protection device adopts a high-precision electronic flow controller with a flow range of 8 to 22L / min and a flow accuracy of ±1.8%. High-purity argon is used as the shielding gas.

[0166] The visual inspection system is equipped with three Basler acA2440-35um industrial cameras with a resolution of 2448×2048 pixels and a maximum acquisition frequency of 35 frames per second. It is equipped with the Cognex PatMax algorithm to achieve sub-pixel precision weld recognition. It is also equipped with a FLIR A655sc infrared thermal imager with a temperature measurement range of -40 to 650°C and a thermal sensitivity of 0.03°C, which can monitor the temperature distribution of the molten pool in real time. The structured light projection system uses a blue LED light source with a projection accuracy of up to 0.03mm.

[0167] In terms of computing equipment, an industrial control computer equipped with an Intel Xeon E5-2697 v4 processor is used as the main control unit, with 18 cores, 36 threads, a main frequency of 3.6GHz, and a memory capacity of 64GB. The graphics processing unit uses NVIDIA QuadroP5000 with 8GB of video memory to accelerate visual processing and neural network reasoning calculations. The data storage uses an NVMe SSD array with a capacity of 4TB, and the sequential read and write speeds reach 3500MB / s and 2700MB / s respectively.

[0168] In this embodiment, the welding process control module performs the following steps to control the welding process:

[0169] First, the 3D point cloud data of the wing truss and skin is obtained through the structured light scanning system, with a point cloud density of 0.5 points / mm 2 , point cloud data collected from 8 different perspectives are fused based on the point cloud registration algorithm. The weld contour is extracted by the edge detection algorithm, and the accuracy of the extracted weld reaches ±0.12mm. Based on the extracted weld contour, the cubic spline interpolation algorithm is used to generate a smooth and continuous welding trajectory. The trajectory is sampled at 8mm intervals to generate a set of key points of the welding path, totaling 1520 key points.

[0170] According to the material properties of 7075 aluminum alloy and the shape of T-weld, the most matching parameter template is selected from the welding process database. The melting point of the material is 635℃, the thermal conductivity is 130W / (m·K), and the specific heat capacity is 960J / (kg·K). Welding is performed according to the parameters shown in Table 1 of the weld:

[0171] Table 1 7075 aluminum alloy T-weld welding process parameters

[0172]

[0173] During the welding process, the weld position and molten pool morphology were monitored in real time by a high-speed camera with an acquisition frequency of 80 Hz and an image resolution of 1280 × 1024 pixels. The edge detection algorithm was used to extract the actual weld position from the image, and the edge detection threshold was set to 1.6 times the average gradient. At the same time, the temperature distribution of the molten pool was monitored by an infrared thermal imager with an acquisition frequency of 60 Hz. The welding error matrix was calculated based on the detection results, and the measured data are shown in Table 2:

[0174] Table 2 Error data statistics during welding

[0175] Error type Maximum Minimum average value Standard Deviation Lateral deviation(mm) 1.86 -1.65 0.08 0.42 Longitudinal deviation(mm) 2.21 -2.35 0.12 0.56 Angle deviation(°) 4.35 -4.12 0.28 0.98 Depth deviation(mm) 0.86 -0.92 0.05 0.24

[0176] Based on the error data, the error adjustment matrix C is constructed, and the historical welding data is regressed by the least squares method to obtain the error adjustment matrix as shown in Table 3:

[0177] Table 3 Welding parameter error adjustment matrix

[0178] coefficient Lateral deviation Longitudinal deviation Angle deviation Depth Bias Current adjustment (A / mm) 6.8 4.2 2.5 12.6 Voltage adjustment (V / mm) 1.2 0.8 0.6 2.4 Speed ​​adjustment (mm / s·mm) 0.5 0.8 0.3 1.6

[0179] The influence of welding parameter fluctuation on molten pool characteristics is analyzed by feedforward neural network. The network contains three hidden layers with 64, 32 and 16 nodes respectively. The network is trained by back propagation algorithm, the learning rate is set to 0.003 and the number of training samples is 25000. The parameter fluctuation law is analyzed by the trained network model, the dynamic characteristics of the molten pool are identified, and the dynamic compensation model of welding parameters is established.

[0180] A fluid dynamics model was introduced to analyze the metal flow state inside the molten pool. The finite element method was used to discretize the molten pool area into 0.2mm grid units. The flow field inside the molten pool was calculated based on the Navier-Stokes equation, with an average calculation time of 5ms / frame. The potential turbulent areas inside the molten pool were identified through Reynolds number analysis. The measured Reynolds number distribution range was 850 to 2200. When the Reynolds number exceeded 2000, it was determined to be turbulent flow, and the defect risk of the corresponding area increased by 46%.

[0181] The welding parameters were optimized by multi-objective optimization function, and the weight coefficients were set to w1 = 0.32, w2 = 0.28, w3 = 0.27, and w4 = 0.13. The particle swarm algorithm was used for parameter optimization, the number of particles was set to 80, and the maximum number of iterations was 180. The optimized welding parameters are shown in Table 4:

[0182] Table 4 Optimized welding process parameters

[0183]

[0184] Real-time evaluation of weld quality, using a hybrid neural network model based on ResNet-50 and Transformer structure, the image feature extraction module contains 5 residual blocks, the sound signal feature extraction module contains 4 convolutional layers, and the multimodal information fusion module uses a 6-layer Transformer encoder structure. The model inference speed reaches 25 frames / second, and the quality assessment accuracy reaches 93.2%.

[0185] A multi-robot collaborative welding strategy is adopted, and the welding task is assigned to three robots based on the task decomposition algorithm. The assignment results are shown in Table 5:

[0186] Table 5 Multi-robot welding task allocation

[0187] Robot Number Welding area Weld length (m) Estimated time (min) Workload ratio (%) Robot-1 Wing root area 3.8 22.5 33.8 Robot-2 Wing area 4.2 21.8 32.8 Robot-3 Wing tip area 4.0 22.2 33.4

[0188] The master-slave control architecture coordinates the synchronous operation of the three robots, and the welding progress deviation is controlled within ±5.8%, effectively avoiding interference and collision between robots. The total welding cycle is shortened by 65.2% compared with traditional single-robot operation.

[0189] Traditional welding process control mainly relies on manual experience or simple closed-loop control, which has problems such as insufficient parameter optimization, large fluctuations in welding quality, and low production efficiency. Traditional methods usually use preset parameter tables for welding, and cannot dynamically adjust parameters according to changes in the welding process, resulting in unstable welding quality. Especially in the welding of complex aluminum alloy structural parts, it is difficult to control the molten pool, which is prone to defects such as pores and cracks, resulting in low pass rate and high rework rate, which seriously affects production efficiency and product quality.

[0190] The present invention realizes intelligent control and optimization of the whole welding process by introducing an intelligent high-performance collaborative welding robot system. Compared with the traditional method, the system has the following advantages: the welding defect rate is reduced by 28.5%, and the welding quality stability is improved by 26.8%. Especially in the field of welding of complex aluminum alloy structural parts, it solves the difficult problem of molten pool control that is difficult to overcome by traditional methods, and significantly improves product quality and production efficiency.

[0191] It should be noted that the variables involved in the present invention are explained in detail as shown in Tables 6, 7 and 8 below.

[0192] Table 6 Variable explanation table (Part I)

[0193]

[0194]

[0195] Table 7 Variable explanation table (Part II)

[0196]

[0197]

[0198] Table 8 Variable explanation table (Part 3)

[0199]

[0200] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An intelligent high-performance collaborative welding robot system, characterized in that: It includes a multi-degree-of-freedom welding robot body, an intelligent control system, a high-performance welding power supply and auxiliary equipment; the intelligent control system includes a motion control module, a welding process control module, a visual detection and feedback module, a collaborative operation control module and a human-computer interaction interface; the welding process control module performs the following steps: obtaining three-dimensional information of the welding workpiece to construct a welding path; selecting matching welding process parameters; establishing a welding error matrix for real-time monitoring; constructing a welding parameter error adjustment matrix; using a neural network algorithm to analyze welding parameter fluctuations; introducing a welding molten pool fluid dynamics model; applying a welding quality multi-objective optimization function; and using a welding quality assessment neural network model for real-time assessment.

2. The intelligent high-performance collaborative welding robot system according to claim 1 is characterized in that: The multi-degree-of-freedom welding robot body has at least 6 degrees of freedom, including an articulated robotic arm and an end effector. The articulated robotic arm is made of high-strength lightweight material, and the end effector is equipped with a welding gun; the high-performance welding power supply and auxiliary equipment include a high-performance welding power supply, a wire feeding mechanism, and a gas protection device.

3. The intelligent high-performance collaborative welding robot system according to claim 2 is characterized in that: The steps of acquiring three-dimensional information of the welding workpiece and constructing a welding path by the welding process control module are: acquiring three-dimensional information of the welding workpiece based on the visual detection and feedback module, constructing a set of key points of the welding path, and generating a welding motion trajectory plan.

4. The intelligent high-performance collaborative welding robot system according to claim 3 is characterized in that: The step of selecting matching welding process parameters by the welding process control module is: selecting matching welding process parameters from a welding process database according to the welding workpiece material properties and the weld shape, and generating initial welding parameter settings.

5. The intelligent high-performance collaborative welding robot system according to claim 4 is characterized in that: The welding process control module performs the steps of establishing a welding error matrix for real-time monitoring as follows: establishing a welding error matrix, monitoring the weld position deviation, molten pool shape change and welding depth in real time during the welding process, and forming welding status feedback data.

6. The intelligent high-performance collaborative welding robot system according to claim 5, characterized in that: The welding process control module executes the steps of constructing a welding parameter error adjustment matrix: constructing a welding parameter error adjustment matrix, calculating welding parameter deviations based on the welding state feedback data, and adjusting welding current, welding voltage and welding speed.

7. The intelligent high-performance collaborative welding robot system according to claim 6, characterized in that: The welding process control module executes the steps of using a neural network algorithm to analyze welding parameter fluctuations: using a neural network algorithm to analyze welding parameter fluctuations during the welding process, identifying the dynamic characteristics of the welding molten pool, and establishing a welding parameter dynamic compensation model; the welding parameter dynamic compensation model adopts a lightweight deep reinforcement learning model, and its specific structure is a hybrid neural network architecture composed of a multi-layer perceptron network and a long short-term memory network, the input layer receives welding state feedback data and welding parameter history data, the hidden layer includes multiple fully connected layers and long short-term memory network units, and the output layer generates welding parameter compensation values.

8. The intelligent high-performance collaborative welding robot system according to claim 7, characterized in that: The welding process control module executes the steps of introducing the welding molten pool fluid dynamics model: introducing the welding molten pool fluid dynamics model, analyzing the turbulent state of metal flow inside the molten pool, and predicting potential welding defect risks.

9. The intelligent high-performance collaborative welding robot system according to claim 8, characterized in that: The steps of applying the welding quality multi-objective optimization function by the welding process control module are: applying the welding quality multi-objective optimization function to optimize welding parameters, and adjusting the welding parameter combination based on the dynamic characteristics of the welding pool and the welding defect risk assessment result.

10. The intelligent high-performance collaborative welding robot system according to claim 9, characterized in that: The step of the welding process control module performing real-time evaluation using the welding quality assessment neural network model is: using a pre-trained welding quality assessment neural network model to perform real-time evaluation of the weld morphology to generate welding quality prediction indicators. The specific structure of the welding quality assessment neural network model is a hybrid network architecture combining a multi-layer convolutional neural network and a Transformer structure, including a weld image feature extraction module, a welding sound signal feature extraction module, and a multimodal information fusion module; the welding quality assessment neural network model also includes an attention mechanism, and the parameters of the attention mechanism are determined according to the material properties of the welding workpiece, the weld shape, and the welding process type.

Images