A welding trajectory analysis method for electric welding machine

Abstract

The present invention relates to the field of welding automation control technology, and specifically to a welding trajectory analysis method for an electric welder, comprising the following steps: S1: capturing visual data and physical parameters in the welding process; S2: applying digital filtering technology to process the collected raw data; S3: extracting welding trajectory data from the processed visual data; S4: constructing a three-dimensional dynamic model of the welding trajectory to simulate the heat distribution and material flow in the welding process; S5: predicting the welding quality and automatically adjusting the welding parameters according to the prediction results; S6: adjusting the motion parameters of the welding head in real time according to the feedback of the machine learning model. The present invention realizes the real-time optimization and automatic adjustment of welding parameters by integrating machine learning technology and a closed-loop control system, significantly improves the welding quality and production efficiency, and at the same time enhances the data management and decision support functions, effectively improving the automation level and economic benefits of the entire welding process.

Description

Technical Field

[0001] The invention relates to the technical field of welding automation control, and in particular to a welding trajectory analysis method for an electric welding machine. Background Art

[0002] In industrial manufacturing, electric welding is a common material connection technology. Its welding quality is directly related to the structural strength and durability of the final product. The guarantee of welding quality requires precise control of multiple welding parameters, such as current, voltage, speed and the precise path of the welding head. Traditional welding technology relies on the skills and experience of operators to adjust parameters. This method often seems inadequate when faced with complex welding tasks, especially on production lines that require high consistency and repeatability. In addition, traditional methods make it difficult to monitor and adjust welding parameters in real time during the welding process, making the timely detection and correction of welding defects a major challenge.

[0003] In response to the above problems, although automated welding technology and some sensor monitoring systems have been introduced in the existing technology, these technologies usually lack a comprehensive real-time data processing and feedback adjustment mechanism, and cannot dynamically optimize welding parameters during the welding process, making it difficult to ensure the optimal quality of the welding process. In addition, the existing system still has deficiencies in data integration, processing and quality assessment, which limits its effectiveness in complex welding applications.

[0004] Therefore, it is particularly important and necessary to develop a system that can monitor the welding process in real time, automatically analyze the welding quality and adjust the welding parameters according to the analysis results. Summary of the invention

[0005] Based on the above purpose, the present invention provides a welding trajectory analysis method for an electric welding machine.

[0006] A welding trajectory analysis method for an electric welding machine comprises the following steps:

[0007] S1: Install high-resolution cameras and multi-dimensional sensors on the welding machine to capture visual data and physical parameters during the welding process, including welding speed, current, voltage, temperature, and weld morphology;

[0008] S2: Apply digital filtering technology to process the collected raw data, remove noise and irrelevant information, and format the data;

[0009] S3: extract welding trajectory data from the processed visual data using a preset image recognition algorithm, and fuse the extracted trajectory data with the physical parameter data obtained from the multi-dimensional sensor to form a comprehensive data set;

[0010] S4: Based on the comprehensive data set, a three-dimensional dynamic model of the welding trajectory is constructed to simulate the heat distribution and material flow during the welding process;

[0011] S5: Use machine learning models to analyze the output of the 3D dynamic model, predict welding quality, and automatically adjust welding parameters including current, speed, and welding path based on the prediction results to optimize the welding process;

[0012] S6: Through closed-loop control technology, the motion parameters of the welding head are adjusted in real time according to the feedback of the machine learning model to ensure that the welding quality meets the preset standards;

[0013] S7: Record welding process data and adjustment results, generate welding quality report to provide operation suggestions and parameter adjustment feedback.

[0014] Furthermore, the S1 specifically includes:

[0015] S11: Choose a high-resolution camera with at least 1080p resolution and high dynamic range to clearly record the shape of the weld under extreme lighting conditions and a multi-dimensional sensor including thermocouples, Hall effect sensors and accelerometers to measure temperature, current, voltage and motion data of the welding head;

[0016] S12: Fix the camera on the top or side of the welding gun of the welding machine to ensure that the camera's viewing angle can fully cover the weld area. The sensors are installed on the power supply line and near the welding gun of the welding machine according to the measurement purpose. Use the debugging software to ensure that the data output of all devices is synchronized with the welding machine control system, and adjust the position of the camera and sensor until a clear image and accurate measurement data can be stably captured;

[0017] S13: Equip the camera and the sensor with an anti-interference data transmission line, connect the line to the main control unit of the welding machine, and set the data acquisition frequency so that the camera can capture continuous video at a rate of at least 30 frames per second, and the sensor can collect physical data at a frequency of at least 1kHz;

[0018] S14: Install and configure the necessary software in the welding machine control system to receive, parse and store data received from the camera and sensors. The software is used to automatically identify and record events during the welding process, including welding starts, stops, speed changes and temperature peaks.

[0019] Furthermore, the S2 specifically includes:

[0020] S21: Bandpass filter and Gaussian filter technology are used to perform preliminary processing on the captured data. The bandpass filter is used to filter out signals within a predetermined frequency range and can be used for current and voltage data to remove noise generated by electromagnetic interference. The Gaussian filter is applied to image data to smooth random noise in the welding video and retain the details of the weld morphology.

[0021] S22: Apply sharpening algorithm to enhance edge information of filtered image data so as to more accurately extract weld morphology and features. For physical parameter data, differential filtering technology is used to extract mutation features, including rapid changes in current and temperature.

[0022] S23: Time synchronization of all filtered data to ensure that the image and physical parameter data are aligned on the time axis, and add a timestamp to each data point so that it can be accurately traced back to the specific moment in the welding process in subsequent analysis;

[0023] S24: Convert the processed data into a standard format, specifically converting image data into a uniform resolution and color depth, and standardizing physical parameter data into a uniform measurement unit and range, to ensure that the data can be seamlessly integrated in subsequent machine learning and model building.

[0024] Furthermore, the S22 specifically includes:

[0025] S221: Select the Laplacian sharpening algorithm to enhance the edge information of the image. The Laplacian algorithm emphasizes the high-frequency information in the image through the second-order derivative, thereby enhancing the edge. The formula of the Laplacian algorithm is: Among them, L(x, y) is the Laplacian value of the image at point (x, y), f is the image intensity, and during the sharpening process, the following formula is used for discrete approximation:

[0026] L(x, y) = f(x+1, y) + f(x-1, y) + f(x, y+1) + f(x, y-1) - 4f(x, y), where L(x, y) is the Laplacian value calculated at the position (x, y) of the image; f(x, y) represents the pixel intensity of the image at the position (x, y); f(x+1, y), f(x-1, y), f(x, y+1), and f(x, y-1) represent the intensity of the pixels around the position (x, y), respectively. Finally, the calculated Laplacian value is added to the original image to enhance the edge effect of the image, so as to accurately identify and extract the weld morphology;

[0027] S222: Differential filtering of physical parameter data, specifically using a first-order differential filter to extract the rapid change features in the physical parameter data. The first-order differential filter identifies data mutations by calculating the differences between consecutive data points. The calculation formula is: Δf(t) = f(t) - f(t-1), where Δf(t) represents the change in the physical parameter at time t, which is the output of the differential filter; f(t) represents the actual measured value of the physical parameter at time t; f(t-1) represents the actual measured value of the physical parameter at time t-1; the differential results are then analyzed after calculation to highlight the differential values ​​that exceed the preset threshold, which are used to indicate physical changes, including a sudden increase in current or a rapid rise in temperature during welding.

[0028] Furthermore, the S3 specifically includes:

[0029] S31: a convolutional neural network is selected as an image recognition algorithm for extracting welding tracks from the sharpened visual data; a convolutional neural network is selected to include multiple convolutional layers, a pooling layer and at least one fully connected layer, the convolutional layer is used to extract local features of the image, the pooling layer is used to reduce the spatial size of the features to reduce the amount of calculation, and the fully connected layer is used to output the position information of the welding track in the image; the training process uses a labeled data set with known welding tracks to train the convolutional neural network, and the training goal is to minimize the error between the actual track and the track predicted by the network;

[0030] S32: Synchronous processing of physical parameter data, selecting physical parameters of current, voltage, temperature and speed corresponding to the image data timestamp; ensuring that the data points of the physical parameters completely correspond to the image frames in time, using the timestamp for synchronization, and handling any time deviation or data delay;

[0031] S33: The welding trajectory data extracted from the convolutional neural network is combined with the physical parameter data using a weighted fusion strategy. In the weighted fusion strategy, each data type is assigned a weight, which is adjusted according to the reliability of the data and the importance of the impact on the welding quality. The following formula is used for data fusion: D f =w1·D t +w2·D p , where D f is the final fusion dataset, D t is the trajectory data extracted from CNN, D p is the synchronized physical parameter data, w1 and w2 are the weights of the corresponding data types; the output comprehensive data set D f It contains all visual and physical information for subsequent 3D model construction and welding quality prediction.

[0032] Furthermore, the S4 specifically includes:

[0033] S41: constructing a three-dimensional dynamic model, first receiving the comprehensive data set generated in S3, including the spatial coordinates, time series information and physical parameters of the welding trajectory, including current, voltage and temperature; then selecting a three-dimensional dynamic model construction method based on finite element analysis, dividing the welding area into fine grid elements, each grid element has its own physical properties and initial conditions, and optimizing the size and density of each grid according to the welding trajectory and physical parameters;

[0034] S42: Perform heat distribution simulation. First, define the heat source model and calculate the heat input based on the welding current and voltage data. The size, shape and intensity of the heat source are dynamically adjusted according to the real-time data in the comprehensive data set. Specifically, the heat transfer equation is applied to calculate the heat energy transfer within each grid element and between grid elements through the finite element method. The heat transfer equation includes three mechanisms: conduction, convection and radiation. The specific formula is: Where T is temperature, t is time, ρ is material density, c is specific heat capacity, k is thermal conductivity, and Q is the heat source term;

[0035] S43: Perform material flow simulation, use fluid dynamics equations to simulate the flow of molten metal in the weld, and combine the results of the thermal model to calculate the metal flow caused by heat. The fluid dynamics equations are based on the Navier-Stokes equations. The specific calculation form is: Among them, ρ represents the density of molten metal; u represents the flow rate of molten metal in the weld; Describes the rate of change of the molten metal velocity with time, i.e., acceleration; represents the nonlinear self-interaction of the molten metal flow rate; Indicates the change of pressure at different positions in the weld area; μ indicates the viscosity of the molten metal. The higher the viscosity, the more difficult it is for the metal to flow; represents the second derivative of velocity in space; f represents the external force acting on the molten metal;

[0036] S44: Verify the prediction results of the model through actual welding experiments, compare the temperature distribution and material flow predicted by the model with the measurement results during the actual welding process, and adjust the model parameters, including the intensity and shape of the heat source and the physical properties of the material, based on the results of the experimental verification to improve the accuracy and reliability of the model.

[0037] Furthermore, the S5 specifically includes:

[0038] S51: Select a machine learning model based on long short-term memory network, extract features from the comprehensive data set, including time series data of welding temperature, speed, current, and voltage, as well as heat distribution and material flow data predicted by the three-dimensional model, and use the quality inspection results of the historical welding process as labels to train the machine learning model. During the training process, adjust the model parameters to minimize the prediction error to ensure that the model can accurately identify welding quality factors;

[0039] S52: In the actual welding process, the real-time data of the current welding process is input into the trained machine learning model to predict the quality of the current welding process, including the structural strength of the weld and the quality problems of the defect type;

[0040] S53: Analyze the deviation between the current welding parameters and the ideal welding quality according to the prediction results of the machine learning model, and specifically apply the preset decision logic to determine the welding parameters that need to be adjusted, including current, speed, and welding path;

[0041] S54: According to the decision logic, new welding parameters are automatically calculated, and adjustment commands are issued through the welding control system. The welding equipment adjusts the current, speed and welding path according to the received commands to ensure that the welding quality reaches or maintains the optimal state in real time.

[0042] Furthermore, the S53 specifically includes:

[0043] S531: First, the welding quality prediction results output by the machine learning model are received, including the predicted weld strength, potential defect locations and types; then the deviation between the current welding parameters and the parameters required to achieve the ideal welding quality is calculated, and the deviation is specifically calculated by the following formula:

[0044] Δp=p ideal -p current , where Δp represents the parameter deviation, p ideal is the ideal parameter value calculated based on the quality prediction results, p current is the parameter value used in the current welding process;

[0045] S532: Setting an adjustment threshold for each welding parameter, which defines the minimum acceptable value of the parameter deviation, and adjustment is required if the parameter deviation exceeds the threshold; the following decision rules are specifically applied to determine whether the welding parameter needs to be adjusted:

[0046] When the current deviation Δi is greater than the threshold θ i When , adjust the current;

[0047] When the speed deviation Δv is greater than the threshold θ v When , adjust the speed;

[0048] When the path deviation Δx is greater than the threshold θ x , adjust the path.

[0049] Furthermore, the S54 specifically includes:

[0050] S541: Based on the decision logic in 553, a specific calculation formula is applied to each parameter that needs to be adjusted to determine a new parameter value. Specifically, the following formula is used to adjust the parameters:

[0051] p new =p current +α×(p ideal -p current ), where p new is the new parameter value after adjustment; p current is the parameter value currently used; p ideal It is the ideal parameter value calculated based on the prediction results and quality requirements; α is an adjustment factor, between 0 and 1, used to control the amplitude and speed of parameter adjustment; the setting of the adjustment factor α is based on the sensitivity and safety requirements of the actual welding operation. Specifically for high-precision welding tasks, the value of α is set to be smaller to avoid excessive parameter changes that lead to unstable welding process;

[0052] S542: Generate specific adjustment commands according to the calculated new parameter values, including adjustment instructions for the current, speed and path of the welding machine;

[0053] S543: After executing the new welding parameters, the quality indicators of the welding process are continuously monitored, including weld morphology, temperature distribution and joint strength; and the effect of parameter adjustment is evaluated based on the monitoring results. When the adjusted parameters fail to achieve the expected welding quality, the decision logic and parameter adjustment calculation are started again to form a closed-loop control.

[0054] Furthermore, the S6 specifically includes:

[0055] S61: Establish a closed-loop control system, which includes a sensor, an actuator, and a control unit. The sensor is used to monitor the parameters in the welding process in real time. The actuator responds to the control command to adjust the movement of the welding head. The control unit processes the input data and generates appropriate output to drive the actuator. The real-time monitoring data from the sensor and the prediction results of the machine learning model are integrated and input to the control unit. The control unit analyzes the data to determine whether the movement parameters of the welding head need to be adjusted.

[0056] S62: A proportional-integral-differential PID controller is used to continuously calculate the error of the control input and make adjustments. The control function of the PID controller is:

[0057] Where u(t) is the control output, that is, the control signal calculated based on the PID algorithm at time t; e(t) is the error, which represents the difference between the target parameter value and the actual parameter value; K p is the proportional term coefficient, which is used to adjust the strength of proportional control; K i is the integral term coefficient, which is used to adjust the strength of integral control; K d is the differential term coefficient, which is used to adjust the intensity of differential control; is the integral term, which is used to calculate the cumulative sum of errors from the beginning to the current moment; is the differential term, and the rate of change of the error over time is calculated; by adjusting the parameters of the PID controller (K p , K i , K d ) to ensure the stability and response speed of the welding process, thereby reducing overshoot and oscillation and optimizing control performance;

[0058] S63: According to the control signal output by the PID controller, the motion parameters of the welding head, including the path, speed and current, are automatically adjusted. The actuator adjusts the position and motion speed of the welding head according to the control signal to ensure that the welding trajectory and parameters are consistent with the preset standards.

[0059] Beneficial effects of the present invention:

[0060] The present invention significantly improves the monitoring and control accuracy of the welding process by introducing machine learning technology and a closed-loop control system. It analyzes the relationship between welding parameters and welding quality in real time and automatically adjusts the current, speed and welding path to ensure that the welding operation is carried out under optimal conditions. This method can effectively reduce welding defects such as welding spatter and joint mismatch, thereby directly improving the structural strength and life of the product.

[0061] The present invention can respond to changing demands on the production line more quickly and improve production efficiency by automatically adjusting welding parameters and performing real-time quality control. In addition, by reducing welding defects caused by human errors, rework and scrap rates are reduced, thereby reducing production costs.

[0062] The present invention provides detailed welding process records and quality assessment reports through integrated data acquisition and processing functions. These data are not only used for real-time monitoring and adjustment, but also provide valuable decision-making support information for production management. The long-term accumulated welding data can be used to further analyze the optimization potential of the welding process, providing a data basis for continuously improving the production process and improving product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0064] Figure 1 A schematic diagram of a welding trajectory analysis method for an electric welding machine according to an embodiment of the present invention;

[0065] Figure 2 It is a schematic diagram of the process of constructing a three-dimensional dynamic model of a welding trajectory according to an embodiment of the present invention. DETAILED DESCRIPTION

[0066] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0067] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0068] like Figure 1-Figure 2 As shown, a welding trajectory analysis method for an electric welding machine comprises the following steps:

[0069] S1: Install high-resolution cameras and multi-dimensional sensors on the welding machine to capture visual data and physical parameters during the welding process, including welding speed, current, voltage, temperature, and weld morphology;

[0070] S2: Apply digital filtering techniques to process the collected raw data, remove noise and irrelevant information, and format the data for subsequent analysis;

[0071] S3: extract welding trajectory data from the processed visual data using a preset image recognition algorithm, and fuse the extracted trajectory data with the physical parameter data obtained from the multi-dimensional sensor to form a comprehensive data set;

[0072] S4: Based on the comprehensive data set, a three-dimensional dynamic model of the welding trajectory is constructed to simulate the heat distribution and material flow during the welding process;

[0073] S5: Use machine learning models to analyze the output of the 3D dynamic model, predict welding quality, and automatically adjust welding parameters including current, speed, and welding path based on the prediction results to optimize the welding process;

[0074] S6: Through closed-loop control technology, the motion parameters of the welding head are adjusted in real time according to the feedback of the machine learning model to ensure that the welding quality meets the preset standards;

[0075] S7: Record welding process data and adjustment results, generate welding quality report to provide operation suggestions and parameter adjustment feedback.

[0076] S1 specifically includes:

[0077] S11: Choose industrial-grade high-resolution cameras and multi-dimensional sensors suitable for high-temperature, high-light and electromagnetic interference environments. The camera should have at least 1080p resolution and high dynamic range (HDR) capabilities to clearly record the morphology of the weld under extreme lighting conditions. The multi-dimensional sensors include thermocouples, Hall effect sensors and accelerometers to measure temperature, current, voltage and motion data of the welding head.

[0078] S12: Fix the camera on the top or side of the welding gun of the welding machine to ensure that the camera's viewing angle can fully cover the weld area. The sensors are installed on the power supply line and near the welding gun of the welding machine according to the measurement purpose. Use the debugging software to ensure that the data output of all devices is synchronized with the welding machine control system, and adjust the position of the camera and sensor until a clear image and accurate measurement data can be stably captured;

[0079] S13: Equip the camera and sensor with an anti-interference data transmission line and connect the line to the main control unit of the welding machine. Set the data acquisition frequency so that the camera can capture continuous video at a rate of at least 30 frames per second, and the sensor can collect physical data at a rate of at least 1kHz. Ensure that all data is encrypted during transmission and processed by a real-time operating system to reduce latency.

[0080] S14: Install and configure the necessary software in the welding machine control system to receive, parse and store the data received from the camera and sensors. The software is used to automatically identify and record events during the welding process, including welding start, stop, speed change and temperature peak. Through the above steps, high-quality capture of key visual and physical data during the welding process is ensured, providing reliable input for subsequent data processing and welding quality analysis.

[0081] S2 specifically includes:

[0082] S21: Bandpass filter and Gaussian filter technology are used to perform preliminary processing on the captured data. The bandpass filter is used to filter out signals within a predetermined frequency range and can be used for current and voltage data to remove noise generated by electromagnetic interference. The Gaussian filter is applied to image data to smooth random noise in the welding video and retain the details of the weld morphology.

[0083] S22: Apply sharpening algorithms to enhance edge information of filtered image data in order to more accurately extract weld morphology and features. For physical parameter data, differential filtering techniques are used to extract mutation features, including rapid changes in current and temperature, which are often associated with welding defects.

[0084] S23: Time synchronization of all filtered data to ensure that the image and physical parameter data are aligned on the time axis, and add a timestamp to each data point so that it can be accurately traced back to the specific moment in the welding process in subsequent analysis;

[0085] S24: Convert the processed data into a standard format, specifically convert the image data into a uniform resolution and color depth, and standardize the physical parameter data into a uniform measurement unit and range, to ensure that the data can be seamlessly integrated in subsequent machine learning and model building; through the above steps, not only the noise and irrelevant information in the welding process data are effectively removed, but also the accuracy and applicability of the data are guaranteed, providing high-quality input for subsequent analysis and model building.

[0086] S22 specifically includes:

[0087] S221: Select the Laplacian sharpening algorithm to enhance the edge information of the image. The Laplacian algorithm emphasizes the high-frequency information in the image through the second-order derivative, thereby enhancing the edge. The formula of the Laplacian algorithm is: Among them, L(x, y) is the Laplacian value of the image at point (x, y), f is the image intensity, and during the sharpening process, the following formula is used for discrete approximation:

[0088] L(x, y) = f(x+1, y) + f(x-1, y) + f(x, y+1) + f(x, y-1) - 4f(x, y), where L(x, y) is the Laplacian value calculated at the position (x, y) of the image, representing the second-order derivative of the point, and is used to measure the amplitude of the image intensity change at the point. A high value indicates that this is an edge area; f(x, y) represents the pixel intensity of the image at the position (x, y); f(x+1, y), f(x-1, y), f(x, y+1), and f(x, y-1) represent the intensity of the pixels around the position (x, y), respectively. These values ​​are used to calculate the intensity difference with the central pixel, and then determine whether the central pixel is located at the edge. Finally, the calculated Laplacian value is added to the original image to enhance the edge effect of the image, so as to accurately identify and extract the weld morphology;

[0089] S222: Differential filtering of physical parameter data, specifically using a first-order differential filter to extract the rapid change characteristics in physical parameter (such as current and temperature) data. The first-order differential filter identifies data mutations by calculating the difference between consecutive data points. The calculation formula is: Δf(t) = f(t)-f(t-1), where Δf(t) represents the change in the physical parameter at time t, which is the output of the differential filter. This value is used to capture and mark Rapid changes in the welding process, such as sudden changes in current or temperature; f(t) represents the measured value of the physical parameter at time t, such as current or temperature; f(t-1) represents the measured value of the physical parameter at time t-1 (i.e., the measurement point before t); then the differential result is analyzed after calculation to highlight the differential value exceeding the preset threshold value, which is used to indicate the physical change, including the sudden increase in current or the rapid increase in temperature during the welding process; through the application of the above steps, the edge recognition ability of the weld in the image data and the detection ability of the key mutation features in the physical parameter data can be effectively improved; in practical applications, for the Laplacian sharpening processing of the image data, by enhancing the edge features in the image, the morphology of the weld is more obvious, which is convenient for subsequent image analysis and feature extraction. For the differential filtering of the physical parameters, by identifying the mutation points in the data, possible welding defects or anomalies can be immediately discovered, thereby adjusting or alarming the welding process. The combination of these two technologies provides a powerful data processing capability for the welding trajectory analysis method of the present invention, ensuring the quality and safety of the welding process.

[0090] S3 specifically includes:

[0091] S31: Select a convolutional neural network (CNN) as an image recognition algorithm to extract welding trajectories from sharpened visual data. CNN is particularly suitable for identifying and classifying features in images and can effectively identify the morphology and features of weld trajectories. Select a convolutional neural network including multiple convolutional layers, a pooling layer and at least one fully connected layer. The convolutional layer is used to extract local features of the image. The pooling layer is used to reduce the spatial size of the features to reduce the amount of calculation. The fully connected layer is used to output the location information of the welding trajectory in the image. The training process uses a labeled data set with known welding trajectories to train the convolutional neural network. The training goal is to minimize the error between the actual trajectory and the network predicted trajectory.

[0092] S32: Synchronous processing of physical parameter data, selecting physical parameters of current, voltage, temperature and speed corresponding to the image data timestamp; ensuring that the data points of the physical parameters completely correspond to the image frames in time, using the timestamp for synchronization, and handling any time deviation or data delay;

[0093] S33: The welding trajectory data extracted from the convolutional neural network is combined with the physical parameter data using a weighted fusion strategy. In the weighted fusion strategy, each data type is assigned a weight, which is adjusted according to the reliability of the data and the importance of the impact on the welding quality. The following formula is used for data fusion: D f =w1·D t +w2·D p , where D f is the final fusion dataset, D t is the trajectory data extracted from CNN, D p is the synchronized physical parameter data, w1 and w2 are the weights of the corresponding data types; the output comprehensive data set D f Contains all visual and physical information for subsequent 3D model construction and welding quality prediction;

[0094] Through the above-mentioned S31-S33, advanced image processing technology and real-time physical parameter monitoring are effectively combined to form a comprehensive integrated data set, which provides strong data support for high-quality welding operations. The implementation of this technology ensures the optimization and automated control of the welding process, improving welding quality and efficiency.

[0095] S4 specifically includes:

[0096] S41: constructing a three-dimensional dynamic model, first receiving the comprehensive data set generated in S3, including the spatial coordinates, time series information and physical parameters of the welding trajectory, including current, voltage and temperature; then selecting a three-dimensional dynamic model construction method based on finite element analysis (FEA), which can handle complex geometric shapes and the interaction of multiple physical phenomena, suitable for simulating the welding process, dividing the welding area into fine grid elements, each of which has its own physical properties and initial conditions, and optimizing the size and density of each grid according to the welding trajectory and physical parameters to ensure the accuracy and computational efficiency of the model;

[0097] S42: Perform heat distribution simulation. First, define the heat source model and calculate the heat input based on the welding current and voltage data. The size, shape and intensity of the heat source are dynamically adjusted according to the real-time data in the comprehensive data set. Specifically, the heat transfer equation is applied to calculate the heat energy transfer within each grid element and between grid elements through the finite element method. The heat transfer equation includes three mechanisms: conduction, convection and radiation. The specific formula is: Where T is temperature, t is time, ρ is material density, c is specific heat capacity, k is thermal conductivity, and Q is the heat source term;

[0098] S43: Material flow simulation is performed. The fluid dynamics equation is used to simulate the flow of molten metal in the weld. Considering the non-Newtonian characteristics of metal flow, a flow model suitable for high-temperature molten metal is used. Combined with the results of the thermal model, the metal flow caused by heat is calculated. The fluid dynamics equation is based on the Navier-Stokes equation. The specific calculation form is: Among them, ρ (density) represents the density of molten metal. During the welding process, the metal changes from solid to liquid, and the density may change. This parameter is very important for flow dynamics and heat transfer; u (velocity vector) represents the flow rate of molten metal in the weld. This vector contains the velocity components of the fluid in all directions in three-dimensional space and is a key variable to describe the motion state of the fluid; (Time derivative of velocity vector) describes the rate of change of molten metal velocity with time, i.e., acceleration. During welding, the velocity of molten metal may change when the heat source moves or welding parameters change; (Convection term) represents the nonlinear self-interaction of the molten metal flow velocity, that is, the effect of the fluid velocity on its own migration; (pressure gradient) indicates the change of pressure at different positions in the weld area. Pressure gradient is one of the main forces driving the flow of molten metal; μ (dynamic viscosity) indicates the viscosity of molten metal. The higher the viscosity, the more difficult it is for the metal to flow. This parameter is crucial to simulate the flow characteristics of molten metal during welding. (Laplacian of velocity) represents the second-order derivative of velocity in space, which is related to the viscous dissipation of the fluid. This term is particularly important when dealing with viscous fluids; f (volume force vector) represents the external forces acting on the molten metal, such as gravity, electromagnetic force, etc. In the welding process, it may include the electromagnetic force generated by the welding current, etc.;

[0099] S44: The prediction results of the model are verified through actual welding experiments, and the temperature distribution and material flow conditions predicted by the model are compared with the measurement results in the actual welding process. According to the results of the experimental verification, the model parameters are adjusted, including the intensity and shape of the heat source and the physical properties of the material, to improve the accuracy and reliability of the model. The above steps provide a high-precision three-dimensional dynamic model to simulate the welding process, making the monitoring and control of the welding process more accurate and scientific, and providing strong technical support for improving the welding quality.

[0100] S5 specifically includes:

[0101] S51: Select a machine learning model based on long short-term memory network, extract features from the comprehensive data set, including time series data of welding temperature, speed, current, and voltage, as well as heat distribution and material flow data predicted by the three-dimensional model, and use the quality inspection results of the historical welding process as labels to train the machine learning model. During the training process, adjust the model parameters to minimize the prediction error to ensure that the model can accurately identify welding quality factors;

[0102] S52: In the actual welding process, the real-time data of the current welding process is input into the trained machine learning model to predict the quality of the current welding process, including the structural strength of the weld and the quality problems of the defect type;

[0103] S53: Analyze the deviation between the current welding parameters and the ideal welding quality based on the prediction results of the machine learning model, and specifically apply the preset decision logic to determine the welding parameters that need to be adjusted, including current, speed and welding path. The decision logic is based on the relationship between the model prediction results and the preset quality standards, as well as the understanding of the impact on the welding process;

[0104] S54: According to the decision logic, new welding parameters are automatically calculated, and adjustment commands are issued through the welding control system. The welding equipment adjusts the current, speed and welding path according to the received commands to ensure that the welding quality reaches or is maintained in the optimal state in real time. Through the above steps, real-time quality monitoring and automatic adjustment of the welding process can be achieved, which can significantly improve the welding quality and efficiency. The introduction of machine learning models not only enhances the intelligence level of the welding process, but also makes the optimization of welding parameters more accurate and timely, thereby improving the overall performance and reliability of the welding operation.

[0105] S53 specifically includes:

[0106] S531: First, the welding quality prediction results output by the machine learning model are received, including the predicted weld strength, potential defect locations and types; then the deviation between the current welding parameters (current, speed and path) and the parameters required to achieve the ideal welding quality is calculated, and the deviation is specifically calculated by the following formula:

[0107] Δp=p ideal -p current , where Δp represents the parameter deviation, p ideal is the ideal parameter value calculated based on the quality prediction results, p current is the parameter value used in the current welding process;

[0108] S532: Setting an adjustment threshold for each welding parameter, which defines the minimum acceptable value of parameter deviation, and adjustment is required if the parameter deviation exceeds the threshold. For example, the current deviation threshold may be set to 5% of the current value. Specifically, the following decision rules are applied to determine whether the welding parameter needs to be adjusted:

[0109] When the current deviation Δi is greater than the threshold θ i When , adjust the current;

[0110] When the speed deviation Δv is greater than the threshold θ v When , adjust the speed;

[0111] When the path deviation Δx is greater than the threshold θ x When , adjust the path;

[0112] Through the above steps, the real-time optimization of welding parameters is ensured, and adjustments are made directly based on the predicted results of welding quality and the actual measured deviations, thereby significantly improving the automation level of the welding process and the quality of the final product.

[0113] S54 specifically includes:

[0114] S541: Based on the decision logic in 553, a specific calculation formula is applied to each parameter that needs to be adjusted to determine a new parameter value. Specifically, the following formula is used to adjust the parameters:

[0115] p new =p current +α×(p ideal -p current ), where p new is the new parameter value after adjustment; p current is the parameter value currently used; p idealIt is the ideal parameter value calculated based on the prediction results and quality requirements; α is an adjustment factor, between 0 and 1, which is used to control the amplitude and speed of parameter adjustment to ensure the stability and adaptability of the welding process; the setting of the adjustment factor α is based on the sensitivity and safety requirements of the actual welding operation. Specifically for high-precision welding tasks, the value of α is set to a smaller value to avoid excessively fast parameter changes that lead to unstable welding process;

[0116] S542: Generate specific adjustment commands according to the calculated new parameter values, including adjustment instructions for the current, speed and path of the welding machine;

[0117] S543: After executing the new welding parameters, continuously monitor the quality indicators of the welding process, including weld morphology, temperature distribution and joint strength; and evaluate the effect of parameter adjustment based on the monitoring results. When the adjusted parameters fail to achieve the expected welding quality, start the decision logic and parameter adjustment calculation again to form a closed-loop control; through the above steps, not only the real-time optimization and automatic adjustment of welding parameters are achieved, but also the closed-loop feedback mechanism ensures the continuous improvement of welding quality, greatly improving the accuracy and efficiency of welding operations.

[0118] S6 specifically includes:

[0119] S61: Establish a closed-loop control system, which includes sensors, actuators, and control units. The sensors are used to monitor parameters (such as current, speed, temperature, and weld shape) during the welding process in real time. The actuators respond to control commands to adjust the movement of the welding head. The control unit processes input data and generates appropriate outputs to drive the actuators. The real-time monitoring data from the sensors and the prediction results of the machine learning model are integrated and input to the control unit. The control unit analyzes the data to determine whether the movement parameters of the welding head need to be adjusted.

[0120] S62: Proportional-integral-differential PID controller is used. This is a commonly used feedback control algorithm for continuously calculating the error of the control input and making adjustments. The control function of the PID controller is:

[0121] Among them, u(t) is the control output, that is, the control signal calculated based on the PID algorithm at time t. This signal is used to adjust the parameters such as current, speed and path of the welding machine to ensure that the welding quality meets the predetermined standards; e(t) is the error, which represents the difference between the target parameter value and the actual parameter value. In welding, this can be the difference between the preset welding speed, current or path and the actual measured value; K p K is the proportional term coefficient, which is used to adjust the intensity of proportional control. The proportional control part directly adjusts the control output according to the error size. The larger the error, the stronger the adjustment. iis the integral term coefficient, which is used to adjust the strength of the integral control. The integral control takes into account the sum of the accumulated errors over time and helps to eliminate the steady-state error of the system;

[0122] K d It is the differential term coefficient, which is used to adjust the intensity of differential control. The differential control takes into account the error change rate, responds quickly to sudden changes in the error, and enhances the stability and response speed of the system. is the integral term, which is used to calculate the cumulative sum of the errors from the beginning to the current moment. This term helps to correct the persistent deviations; is the differential term, which calculates the rate of change of the error over time. This term helps to predict the future trend of the error and make timely adjustments to the rapidly changing error. By adjusting the parameters of the PID controller (K p , K i , K d ) to ensure the stability and response speed of the welding process, thereby reducing overshoot and oscillation and optimizing control performance;

[0123] S63: According to the control signal output by the PID controller, the motion parameters of the welding head, including the path, speed and current, are automatically adjusted, and the actuator adjusts the position and motion speed of the welding head according to the control signal to ensure that the welding trajectory and parameters are consistent with the preset standards;

[0124] Through the above steps, the closed-loop control system can achieve precise adjustment and real-time quality control during the welding process, greatly improve the automation level of welding operations and the quality of the final product, and ensure the efficiency and reliability of the welding process.

[0125] S7 specifically includes:

[0126] S71: Establish a data recording mechanism. First, determine the key data collection points during the welding process, including current, voltage, speed, temperature, and the position and trajectory of the welding head. These data points are crucial to understanding and controlling the welding process. Use sensors installed on the welding machine to collect the above parameters in real time. Ensure that the data collection system has high speed and high precision to accurately record every change in the welding process.

[0127] S72: Synchronize the data collected from different sources and sensors to ensure that all data are aligned with the corresponding timestamps, which is the basis for accurate analysis and subsequent report generation; then perform necessary processing on the collected raw data, including filtering, denoising and normalization, to facilitate analysis and report preparation;

[0128] S73: Determine the key contents that the welding quality report should include, such as statistical analysis of welding parameters, historical records of parameter adjustments, any welding defects detected and their possible causes; use specialized software tools to automatically generate welding quality reports from the integrated data set. These tools should be able to provide graphical data displays, such as parameter change trend charts, weld quality heat maps, etc., to intuitively display the quality control of the welding process;

[0129] S74: Generate specific operational recommendations based on the analysis results in the welding quality report. These recommendations may include adjustments to welding parameters, improvements to welding techniques, or maintenance and calibration of equipment. A feedback mechanism is established to provide the generated operational recommendations and reports to welding operators and engineers to ensure that these feedbacks are incorporated into future welding operations.

[0130] The present invention is intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A welding trajectory analysis method for an electric welding machine, characterized in that: The following steps are involved: S1: Install high-resolution cameras and multi-dimensional sensors on the welding machine to capture visual data and physical parameters during the welding process, including welding speed, current, voltage, temperature, and weld morphology; S2: Apply digital filtering technology to process the collected raw data, remove noise and irrelevant information, and format the data; S3: extract welding trajectory data from the processed visual data using a preset image recognition algorithm, and fuse the extracted trajectory data with the physical parameter data obtained from the multi-dimensional sensor to form a comprehensive data set; The S3 specifically includes: S31: a convolutional neural network is selected as an image recognition algorithm for extracting welding tracks from the sharpened visual data; a convolutional neural network is selected to include multiple convolutional layers, a pooling layer and at least one fully connected layer, the convolutional layer is used to extract local features of the image, the pooling layer is used to reduce the spatial size of the features to reduce the amount of calculation, and the fully connected layer is used to output the position information of the welding track in the image; the training process uses a labeled data set with known welding tracks to train the convolutional neural network, and the training goal is to minimize the error between the actual track and the track predicted by the network; S32: Synchronous processing of physical parameter data, selecting physical parameters of current, voltage, temperature and speed corresponding to the image data timestamp; ensuring that the data points of the physical parameters completely correspond to the image frames in time, using the timestamp for synchronization, and handling any time deviation or data delay; S33: The welding trajectory data extracted from the convolutional neural network is combined with the physical parameter data using a weighted fusion strategy. In the weighted fusion strategy, each data type is assigned a weight, which is adjusted according to the reliability of the data and the importance of the impact on the welding quality. The following formula is used for data fusion: D f =w1·D t +w2·D p , where D f is the final fusion dataset, D t is the trajectory data extracted from CNN, D p is the synchronized physical parameter data, w1 and w2 are the weights of the corresponding data types; the output comprehensive data set D f Contains all visual and physical information for subsequent 3D model construction and welding quality prediction; S4: Based on the comprehensive data set, a three-dimensional dynamic model of the welding trajectory is constructed to simulate the heat distribution and material flow during the welding process; The S4 specifically includes: S41: constructing a three-dimensional dynamic model, first receiving the comprehensive data set generated in S3, including the spatial coordinates, time series information and physical parameters of the welding trajectory, including current, voltage and temperature; then selecting a three-dimensional dynamic model construction method based on finite element analysis, dividing the welding area into fine grid elements, each grid element has its own physical properties and initial conditions, and optimizing the size and density of each grid according to the welding trajectory and physical parameters; S42: Perform heat distribution simulation. First, define the heat source model and calculate the heat input based on the welding current and voltage data. The size, shape and intensity of the heat source are dynamically adjusted according to the real-time data in the comprehensive data set. Specifically, the heat transfer equation is applied to calculate the heat energy transfer within each grid element and between grid elements through the finite element method. The heat transfer equation includes three mechanisms: conduction, convection and radiation. The specific formula is: Where T is temperature, t is time, ρ is material density, c is specific heat capacity, k is thermal conductivity, and Q is the heat source term; S43: Perform material flow simulation, use fluid dynamics equations to simulate the flow of molten metal in the weld, and combine the results of the thermal model to calculate the metal flow caused by heat. The fluid dynamics equations are based on the Navier-Stokes equations. The specific calculation form is: Among them, ρ represents the density of molten metal; u represents the flow rate of molten metal in the weld; Describes the rate of change of the molten metal velocity with time, i.e., acceleration; represents the nonlinear self-interaction of the molten metal flow rate; Indicates the change of pressure at different positions in the weld area; μ indicates the viscosity of the molten metal. The higher the viscosity, the more difficult it is for the metal to flow; represents the second derivative of velocity in space; f represents the external force acting on the molten metal; S44: Verify the prediction results of the model through actual welding experiments, compare the temperature distribution and material flow predicted by the model with the measurement results during the actual welding process, and adjust the model parameters, including the intensity and shape of the heat source and the physical properties of the material, according to the results of the experimental verification to improve the accuracy and reliability of the model; S5: Use machine learning models to analyze the output of the 3D dynamic model, predict welding quality, and automatically adjust welding parameters including current, speed, and welding path based on the prediction results to optimize the welding process; The S5 specifically includes: S51: Select a machine learning model based on long short-term memory network, extract features from the comprehensive data set, including time series data of welding temperature, speed, current, and voltage, as well as heat distribution and material flow data predicted by the three-dimensional model, and use the quality inspection results of the historical welding process as labels to train the machine learning model. During the training process, adjust the model parameters to minimize the prediction error to ensure that the model can accurately identify welding quality factors; S52: In the actual welding process, the real-time data of the current welding process is input into the trained machine learning model to predict the quality of the current welding process, including the structural strength of the weld and the quality problems of the defect type; S53: Analyze the deviation between the current welding parameters and the ideal welding quality according to the prediction results of the machine learning model, and specifically apply the preset decision logic to determine the welding parameters that need to be adjusted, including current, speed, and welding path; S54: According to the decision logic, new welding parameters are automatically calculated, and adjustment commands are issued through the welding control system. The welding equipment adjusts the current, speed and welding path according to the received commands to ensure that the welding quality reaches or is maintained at the optimal state in real time; S6: Through closed-loop control technology, the motion parameters of the welding head are adjusted in real time according to the feedback of the machine learning model to ensure that the welding quality meets the preset standards; S7: Record welding process data and adjustment results, generate welding quality report to provide operation suggestions and parameter adjustment feedback.

2. A welding trajectory analysis method for an electric welding machine according to claim 1, characterized in that: The S1 specifically includes: S11: Choose a high-resolution camera with at least 1080p resolution and high dynamic range to clearly record the shape of the weld under extreme lighting conditions and a multi-dimensional sensor including thermocouples, Hall effect sensors and accelerometers to measure temperature, current, voltage and motion data of the welding head; S12: Fix the camera on the top or side of the welding gun of the welding machine to ensure that the camera's viewing angle can fully cover the weld area. The sensors are installed on the power supply line and near the welding gun of the welding machine according to the measurement purpose. Use the debugging software to ensure that the data output of all devices is synchronized with the welding machine control system, and adjust the position of the camera and sensor until a clear image and accurate measurement data can be stably captured; S13: Equip the camera and the sensor with an anti-interference data transmission line, connect the line to the main control unit of the welding machine, and set the data acquisition frequency so that the camera can capture continuous video at a rate of at least 30 frames per second, and the sensor can collect physical data at a frequency of at least 1kHz; S14: Install and configure the necessary software in the welding machine control system to receive, parse and store data received from the camera and sensors. The software is used to automatically identify and record events during the welding process, including welding starts, stops, speed changes and temperature peaks.

3. A welding trajectory analysis method for an electric welder according to claim 1, characterized in that: The S2 specifically includes: S21: Bandpass filter and Gaussian filter technology are used to perform preliminary processing on the captured data. The bandpass filter is used to filter out signals within a predetermined frequency range and can be used for current and voltage data to remove noise generated by electromagnetic interference. The Gaussian filter is applied to image data to smooth random noise in the welding video and retain the details of the weld morphology. S22: Apply sharpening algorithm to enhance edge information of filtered image data so as to more accurately extract weld morphology and features. For physical parameter data, differential filtering technology is used to extract mutation features, including rapid changes in current and temperature. S23: Time synchronization of all filtered data to ensure that the image and physical parameter data are aligned on the time axis, and add a timestamp to each data point so that it can be accurately traced back to the specific moment in the welding process in subsequent analysis; S24: Convert the processed data into a standard format, specifically converting image data into a uniform resolution and color depth, and standardizing physical parameter data into a uniform measurement unit and range, to ensure that the data can be seamlessly integrated in subsequent machine learning and model building.

4. A welding trajectory analysis method for an electric welder according to claim 3, characterized in that: The S22 specifically includes: S221: Select the Laplacian sharpening algorithm to enhance the edge information of the image. The Laplacian algorithm emphasizes the high-frequency information in the image through the second-order derivative, thereby enhancing the edge. The formula of the Laplacian algorithm is: Among them, L(x, y) is the Laplacian value of the image at point (x, y), f is the image intensity, and during the sharpening process, the following formula is used for discrete approximation: L(x, y) = f(x+1, y) + f(x-1, y) + f(x, y+1) + f(x, y-1) - 4f(x, y), where L(x, y) is the Laplacian value calculated at the position (x, y) of the image; f(x, y) represents the pixel intensity of the image at the position (x, y); f(x+1, y), f(x-1, y), f(x, y+1), and f(x, y-1) represent the intensity of the pixels around the position (x, y), respectively. Finally, the calculated Laplacian value is added to the original image to enhance the edge effect of the image, so as to accurately identify and extract the weld morphology; S222: Differential filtering of physical parameter data, specifically using a first-order differential filter to extract the rapid change features in the physical parameter data. The first-order differential filter identifies data mutations by calculating the differences between consecutive data points. The calculation formula is: Δf(t) = f(t) - f(t-1), where Δf(t) represents the change in the physical parameter at time t, which is the output of the differential filter; f(t) represents the actual measured value of the physical parameter at time t; f(t-1) represents the actual measured value of the physical parameter at time t-1; the differential results are then analyzed after calculation to highlight the differential values ​​that exceed the preset threshold, which are used to indicate physical changes, including a sudden increase in current or a rapid rise in temperature during welding.

5. A welding trajectory analysis method for an electric welder according to claim 1, characterized in that: The S53 specifically includes: S531: First, the welding quality prediction results output by the machine learning model are received, including the predicted weld strength, potential defect locations and types; then the deviation between the current welding parameters and the parameters required to achieve the ideal welding quality is calculated, and the deviation is specifically calculated by the following formula: Δp=p ideal -p current , where Δp represents the parameter deviation, p ideal is the ideal parameter value calculated based on the quality prediction results, p current is the parameter value used in the current welding process; S532: Setting an adjustment threshold for each welding parameter, which defines the minimum acceptable value of the parameter deviation, and adjustment is required if the parameter deviation exceeds the threshold; the following decision rules are specifically applied to determine whether the welding parameter needs to be adjusted: When the current deviation Δi is greater than the threshold θ i When , adjust the current; When the speed deviation Δv is greater than the threshold θ v When , adjust the speed; When the path deviation Δx is greater than the threshold θ x , adjust the path.

6. A welding trajectory analysis method for an electric welding machine according to claim 5, characterized in that: The S54 specifically includes: S541: Based on the decision logic in 553, a specific calculation formula is applied to each parameter that needs to be adjusted to determine a new parameter value. Specifically, the following formula is used to adjust the parameters: p new =p current +α×(p ideal -p current ), where p new is the new parameter value after adjustment; p current is the parameter value currently used; p ideal It is the ideal parameter value calculated based on the prediction results and quality requirements; α is an adjustment factor, between 0 and 1, used to control the amplitude and speed of parameter adjustment; the setting of the adjustment factor α is based on the sensitivity and safety requirements of the actual welding operation. Specifically for high-precision welding tasks, the value of α is set to be smaller to avoid excessive parameter changes that lead to unstable welding process; S542: Generate specific adjustment commands according to the calculated new parameter values, including adjustment instructions for the current, speed and path of the welding machine; S543: After executing the new welding parameters, the quality indicators of the welding process are continuously monitored, including weld morphology, temperature distribution and joint strength; and the effect of parameter adjustment is evaluated based on the monitoring results. When the adjusted parameters fail to achieve the expected welding quality, the decision logic and parameter adjustment calculation are started again to form a closed-loop control.

7. A welding trajectory analysis method for an electric welder according to claim 1, characterized in that: The S6 specifically includes: S61: Establish a closed-loop control system, which includes a sensor, an actuator, and a control unit. The sensor is used to monitor the parameters in the welding process in real time. The actuator responds to the control command to adjust the movement of the welding head. The control unit processes the input data and generates appropriate output to drive the actuator. The real-time monitoring data from the sensor and the prediction results of the machine learning model are integrated and input to the control unit. The control unit analyzes the data to determine whether the movement parameters of the welding head need to be adjusted. S62: A proportional-integral-differential PID controller is used to continuously calculate the error of the control input and make adjustments. The control function of the PID controller is: Where u(t) is the control output, that is, the control signal calculated based on the PID algorithm at time t; e(t) is the error, which represents the difference between the target parameter value and the actual parameter value; K p is the proportional term coefficient, which is used to adjust the strength of proportional control; K i is the integral term coefficient, which is used to adjust the strength of integral control; K d is the differential term coefficient, which is used to adjust the intensity of differential control; is the integral term, which is used to calculate the cumulative sum of errors from the beginning to the current moment; is the differential term, and the rate of change of the error over time is calculated; by adjusting the parameters of the PID controller (K p , K i , K d ) to ensure the stability and response speed of the welding process, thereby reducing overshoot and oscillation and optimizing control performance; S63: According to the control signal output by the PID controller, the motion parameters of the welding head, including the path, speed and current, are automatically adjusted. The actuator adjusts the position and motion speed of the welding head according to the control signal to ensure that the welding trajectory and parameters are consistent with the preset standards.

Images