Intelligent radiator welding method and system based on welding residual stress control

Abstract

According to the intelligent radiator welding method and system based on welding residual stress control, temperature field data of a welding area are obtained through a sensor array, a temperature gradient mapping image is generated, residual stress distribution is determined through heat-structure coupling finite element analysis, and a thermal expansion coefficient and a welding speed parameter are extracted; when the deviation exceeds a deviation threshold value, the heat input power and the welding speed are adjusted, parameter combination is optimized, heat input distribution is redistributed, new temperature field distribution is generated by adopting a power segmentation adjustment algorithm, residual stress relief data are extracted, and if the stress reduction rate is insufficient, the welding speed is adjusted; if yes, cooling parameters are adjusted for thermal stress release, through finite element iteration fed back in real time, a process correction scheme is generated, full-process simulation verification is conducted, and the fatigue life and the thermal conductivity index are obtained. Accurate control over the residual stress in the welding process is achieved, and the intelligent welding quality and the structural stability are improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent processing technology, and particularly to an intelligent welding method and system for radiators based on the control of welding residual stress. Background Art

[0002] As an indispensable key area in modern manufacturing, the welding technology of aluminum alloy radiators is directly related to the performance stability, service life and production efficiency of radiators. It is of decisive significance especially in high-precision industries such as aerospace, electronic equipment and new energy vehicles.

[0003] With the increasing demand for lightweight design, aluminum alloy is widely used due to its excellent thermal conductivity and low density. However, the technical complexity of its welding process has also become prominent, becoming a current research hotspot. A reasonable welding method not only affects the structural integrity of the radiator, but also plays a crucial role in the overall efficiency of the thermal management system.

[0004] Currently, traditional welding methods for aluminum alloy radiators, such as fusion welding or conventional friction stir welding, have shown significant limitations in practical applications. These methods often struggle to effectively balance the heat input during welding, resulting in uneven temperature distribution inside the workpiece, which in turn leads to significant residual stress and thermal deformation problems. Especially for radiators with large areas and thin-walled structures, the existing technologies are unable to ensure flatness and dimensional accuracy, and often require additional machining and correction after welding, increasing production costs and process complexity. In this field, the core challenges mainly focus on two key technical factors: the control of residual stress and the compensation of thermal deformation. Due to the non-uniformity of welding heat input, a large temperature gradient is easily formed inside the workpiece, leading to the accumulation of residual stress and inducing deformation, which not only affects the geometric accuracy of the radiator, but also may weaken its long-term service performance. For thin-walled structures, this problem is particularly prominent because the material rigidity is insufficient to resist the action of thermal stress, further exacerbating the uncontrollability of deformation. In addition, the lack of optimization of welding sequence and cooling strategy makes it difficult to precisely control the heat distribution, forming a unique technical problem.

[0005] Therefore, how to achieve effective control of residual stress and precise compensation of thermal deformation by optimizing the heat input distribution and process parameter design during welding has become the key issue for improving the welding quality and production efficiency of aluminum alloy intelligent radiators. Solving this problem requires a systematic breakthrough from the macroscopic welding strategy to the microscopic heat management to meet the stringent requirements of high-precision intelligent manufacturing. Summary of the Invention

[0006] To solve the technical problems raised in the above background art, in the first aspect of the present invention, an intelligent welding method for radiators based on the control of welding residual stress is provided, including:

[0007] S1, Obtain the temperature field data of the welding area and generate the first temperature gradient mapping image using a sensor array;

[0008] S2, From the first temperature gradient mapping image, determine the first residual stress distribution data through thermo-structural coupled finite element analysis;

[0009] S3, Extract the thermal expansion coefficient and welding speed parameters according to the first residual stress distribution data and generate the first deformation offset vector;

[0010] S4, If the first deformation offset vector exceeds the preset geometric accuracy threshold, adjust the heat input power and welding speed parameters to generate the first optimized parameter combination;

[0011] S5, Redistribute the heat input distribution according to the first optimized parameter combination and generate the second temperature field distribution data using the power segmented regulation algorithm;

[0012] S6, Extract the residual stress relief data from the second temperature field distribution data and adjust the cooling parameters to generate the first cooling configuration when the stress reduction rate is lower than the reduction rate threshold;

[0013] S7, Perform thermal stress release treatment using the first cooling configuration to generate the first flatness data;

[0014] S8, If the first flatness data does not reach the geometric accuracy threshold, generate the first process correction plan through real-time feedback finite element iteration;

[0015] S9, Conduct full-process simulation verification on the first process correction plan to obtain the fatigue life and thermal conductivity index data;

[0016] S10, Judge whether the long-term stability meets the preset cycle number standard according to the fatigue life index data.

[0017] Optionally, step S1, obtain the temperature field data of the welding area and generate the first temperature gradient mapping image using a sensor array, includes:

[0018] Step S11, Collect the temperature field data of the welding area through a sensor array to generate an initial temperature distribution image;

[0019] Step S12, Process the initial temperature distribution image using the VGG16 model in the convolutional neural network to extract the temperature gradient features in the image;

[0020] Step S13, If the extracted temperature gradient features exceed the temperature threshold, use the bilinear interpolation algorithm to optimize the gradient mapping accuracy and generate the second temperature gradient image;

[0021] Step S14, extract the change trend of the temperature field according to the second temperature gradient image, and judge the hot spot distribution characteristics;

[0022] Step S15, analyze the spatial consistency of the temperature field through the hot spot distribution characteristics, and use the region growing algorithm to obtain the abnormal point map of the welding area;

[0023] Step S16, process the abnormal point map using the K-means clustering algorithm to determine the boundary range of the temperature abnormal area;

[0024] Step S17, for the temperature field data within the boundary range, use the gradient descent method to generate the final gradient mapping image to complete the temperature field analysis.

[0025] Optionally, in step S2, from the first temperature gradient mapping image, determine the first residual stress distribution data through thermo-structural coupled finite element analysis, including:

[0026] Step S21, obtain the temperature gradient distribution from the first temperature gradient mapping image, and determine the first temperature field data through thermo-structural coupled analysis;

[0027] Step S22, according to the first temperature field data, calculate the heat conduction effect using the finite element method to obtain the preliminary distribution of the first stress field;

[0028] Step S23, through the preliminary distribution of the first stress field, combined with the temperature gradient change, judge the thermo-structural interaction effect to determine the second stress field data;

[0029] Step S24, obtain the residual stress distribution characteristics from the second stress field data, and optimize through coupled analysis to obtain the first residual stress distribution data;

[0030] Step S25, for the first residual stress distribution data, use the support vector machine algorithm to analyze the boundary characteristics of the distribution data to determine the third stress field data;

[0031] Step S26, through the third stress field data, obtain the mapping relationship between the stress field and the temperature field, judge the data consistency, and obtain the final residual stress distribution data;

[0032] Step S27, extract the key distribution characteristics from the final residual stress distribution data, generate a mapping image through a visualization tool, and determine the analysis result.

[0033] Optionally, in step S3, extract the thermal expansion coefficient and welding speed parameters according to the first residual stress distribution data to generate the first deformation offset vector, including:

[0034] Step S31, obtain the first residual stress distribution data, and extract the thermal expansion coefficient and welding speed parameters through analysis to obtain the first parameter set;

[0035] Step S32: Calculate the deformation offset by using the finite element analysis method with the first parameter set to obtain the first offset result;

[0036] Step S33: Extract the offset vector feature from the first offset result to generate the first deformation offset vector;

[0037] Step S34: Use the support vector machine algorithm to judge the vector consistency for the first deformation offset vector to obtain the consistency index;

[0038] Step S35: If the consistency index is lower than the consistency threshold, recalculate the offset vector by adjusting the thermal expansion coefficient to obtain the second offset vector;

[0039] Step S36: Predict the final deformation trend by using the linear regression algorithm according to the second offset vector to obtain the trend data;

[0040] Step S37: Generate the description of the final deformation offset vector through the trend data.

[0041] Optionally, in step S4, if the first deformation offset vector exceeds the preset geometric accuracy threshold, adjust the heat input power and welding speed parameters to generate the first optimization parameter combination, including:

[0042] Step S41: If the first deformation offset vector exceeds the preset geometric accuracy threshold, obtain the offset data through vector analysis to determine the degree of the exceeded range;

[0043] Step S42: Calculate the adjustment ratio of the heat input power and welding speed by using the linear regression algorithm according to the offset data to obtain the preliminary adjustment value;

[0044] Step S43: Update the parameters of the heat input power and welding speed through the preliminary adjustment value to generate the first adjustment parameter set;

[0045] Step S44: Apply the first adjustment parameter set to the welding process to obtain the new deformation offset vector;

[0046] Step S45: If the new deformation offset vector still exceeds the geometric accuracy threshold, obtain the new offset data and judge whether further optimization is required;

[0047] Step S46: According to the judgment result, use the gradient descent algorithm to iteratively adjust the power parameter and speed parameter, initialize the parameter as the first adjustment parameter set, and set the learning rate to 01 to obtain the second optimization parameter combination;

[0048] Step S47: Apply the second optimized parameter combination to the welding process and analyze the change trend of the deformation offset vector. If the change amplitude of the deformation offset vector is less than 1 mm for three consecutive times, it is determined that the change trend is stable.

[0049] Step S48: If the change trend is stable and the offset vector meets the geometric accuracy threshold, generate the first optimized parameter combination through the final parameter adjustment plan.

[0050] Optionally, in step S5, according to the first optimized parameter combination, redistribute the heat input distribution and generate the second temperature field distribution data by using the power segmented adjustment algorithm, including:

[0051] Step S51: Obtain the heat input data from the heat sensor through the first optimized parameter combination and determine the initial input distribution.

[0052] Step S52: Calculate the power segmentation result by using the linear segmentation method according to the initial input distribution.

[0053] Step S53: Run the linear regression algorithm for the power segmentation result to obtain the adjusted distribution data.

[0054] Step S54: Extract the temperature characteristics from the adjusted distribution data to generate the preliminary temperature field distribution.

[0055] Step S55: If the preliminary temperature field distribution exceeds the temperature threshold, optimize the parameter combination by using the gradient descent method and calculate the new input distribution.

[0056] Step S56: Perform data calculation by using the least squares method according to the new input distribution to determine the second temperature field distribution data.

[0057] Step S57: Compare the second temperature field distribution data with the temperature threshold to verify the field distribution consistency and output the final result.

[0058] Optionally, in step S6, extract the residual stress relief data from the second temperature field distribution data. When the stress reduction rate is lower than the reduction rate threshold, adjust the cooling parameters to generate the first cooling configuration, including:

[0059] Step S61: Extract the residual stress relief data from the second temperature field distribution data and obtain the stress change characteristics by using the data extraction technology.

[0060] Step S62: Calculate the stress reduction rate through the stress change characteristics to obtain the reduction rate value.

[0061] Step S63: If the reduction rate value is lower than the reduction rate threshold, determine the adjusted cooling parameters by using the genetic algorithm.

[0062] Step S64: Generate a first cooling configuration based on the adjusted cooling parameters to obtain a preliminary cooling plan;

[0063] Step S65: Extract the temperature field distribution characteristics from the preliminary cooling plan, and use principal component analysis to judge the distribution consistency;

[0064] Step S66: Adjust the second temperature field distribution data according to the distribution consistency result to obtain optimized temperature field data;

[0065] Step S67: Re-extract the residual stress relief data according to the optimized temperature field data, and judge whether it meets the preset conditions.

[0066] Optionally, in step S7, perform a thermal stress release treatment using the first cooling configuration to generate first flatness data, including:

[0067] Step S71: Perform a thermal stress release treatment using the first cooling configuration to generate first flatness data;

[0068] Step S72: Obtain the first flatness data through a sensor to determine the initial flatness value;

[0069] Step S73: If the initial flatness value exceeds the flatness threshold, adjust the cooling configuration parameters to obtain optimized flatness data;

[0070] Step S74: According to the optimized flatness data, use the support vector machine algorithm to judge the thermal stress release effect and determine the effect index;

[0071] Step S75: For the effect index, obtain the temperature distribution data of the heat treatment process to obtain the stress release trend;

[0072] Step S76: Through the stress release trend, use the linear regression algorithm to predict the change of the flatness value and obtain the prediction result;

[0073] Step S77: According to the prediction result, adjust the release treatment process to generate the final flatness data.

[0074] Optionally, in step S10, judge whether the long-term stability meets the preset cycle number standard according to the fatigue life index data, including:

[0075] Step S101: Obtain the index data through the fatigue life data, and use data cleaning technology to process the original data to obtain the cleaned data set;

[0076] Step S102: Extract the cycle number characteristics from the cleaned data set, use statistical analysis methods to calculate the characteristic values, and determine the cycle number distribution;

[0077] Step S103: Generate a stability judgment basis according to the cycle number distribution. If the eigenvalue exceeds the preset cycle threshold, mark it as abnormal data;

[0078] Step S104: Conduct data analysis on the abnormal data, identify the abnormal pattern through a support vector machine, and obtain the classification result;

[0079] Step S105: Obtain the analysis result based on the classification result, group the results using K-means clustering, and determine the stability trend after grouping;

[0080] Step S106: Judge the long-term stability through the stability trend after grouping, compare with the number standard using a comparison method, and obtain the final judgment result;

[0081] Step S107: Extract the standard judgment conclusion from the final judgment result, verify the consistency through logic verification technology, and obtain the verified judgment data.

[0082] In the second aspect of the present invention, an intelligent welding system for a radiator based on controlling welding residual stress is provided. The radiator is intelligently welded using the method described above. The system further includes:

[0083] A temperature field data acquisition module, configured to acquire the temperature field data of the welding area and generate a first temperature gradient mapping image using a sensor array;

[0084] A first residual stress analysis module, configured to determine the first residual stress distribution data from the first temperature gradient mapping image through thermo-structural coupled finite element analysis;

[0085] A deformation offset vector generation module, configured to extract the thermal expansion coefficient and welding speed parameters from the first residual stress distribution data and generate a first deformation offset vector;

[0086] A parameter optimization module, configured to adjust the heat input power and welding speed parameters to generate a first optimized parameter combination if the first deformation offset vector exceeds the geometric accuracy threshold;

[0087] A temperature field distribution generation module, configured to reallocate the heat input distribution according to the first optimized parameter combination and generate second temperature field distribution data using a power segmented adjustment algorithm;

[0088] A cooling configuration generation module, configured to extract residual stress relief data from the second temperature field distribution data and adjust the cooling parameters to generate a first cooling configuration when the stress reduction rate is lower than the reduction rate threshold;

[0089] A flatness data generation module, configured to perform thermal stress release treatment using the first cooling configuration and generate first flatness data;

[0090] A process correction module, which is used to generate a first process correction plan through finite element iteration with real-time feedback if the first flatness data fails to reach the geometric accuracy threshold;

[0091] A simulation verification module, which is used to perform full-process simulation verification on the first process correction plan to obtain fatigue life and thermal conductivity index data;

[0092] A stability judgment module, which is used to judge whether the long-term stability meets the preset cycle number standard according to the fatigue life index data.

[0093] An intelligent welding method and system for a radiator based on welding residual stress control provided by the present invention obtain temperature field data of a welding area through a sensor array, generate a temperature gradient mapping image, and use thermo-structural coupling finite element analysis to determine the residual stress distribution. Based on this, the coefficient of thermal expansion and welding speed parameters are extracted to generate a deformation offset vector. When the offset exceeds the offset threshold, the present invention adjusts the heat input power and welding speed, optimizes the parameter combination and redistributes the heat input distribution, adopts a power segmented adjustment algorithm to generate a new temperature field distribution, extracts residual stress relief data, and if the stress reduction rate is insufficient, adjusts the cooling parameters for thermal stress release. Through finite element iteration with real-time feedback, a process correction plan is generated and full-process simulation verification is carried out to obtain fatigue life and thermal conductivity indexes. The present invention realizes precise control of residual stress during the welding process and improves the quality and structural stability of intelligent welding. Description of the Drawings

[0094] Figure 1 It is a flowchart of an intelligent welding method for a radiator based on welding residual stress control of the present invention.

[0095] Figure 2 It is a structural schematic diagram of an intelligent welding system for a radiator based on welding residual stress control of the present invention. Detailed Embodiments

[0096] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments in this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this specification.

[0097] As Figure 1 shown, in the first aspect of the present invention, an intelligent welding method for a radiator based on welding residual stress control is provided, including:

[0098] S1. Obtain the temperature field data of the welding area and generate the first temperature gradient mapping image using a sensor array.

[0099] Optionally, this step further includes:

[0100] Step S11. Collect the temperature field data of the welding area through the sensor array and generate an initial temperature distribution image.

[0101] Step S12. Use the VGG16 model in the convolutional neural network to process the initial temperature distribution image and extract the temperature gradient features in the image.

[0102] Step S13. If the extracted temperature gradient features exceed the temperature threshold, use the bilinear interpolation algorithm to optimize the gradient mapping accuracy and generate the second temperature gradient image.

[0103] Step S14. Extract the change trend of the temperature field according to the second temperature gradient image and judge the hot spot distribution characteristics.

[0104] Step S15. Analyze the spatial consistency of the temperature field through the hot spot distribution characteristics and use the region growing algorithm to obtain the abnormal point map of the welding area.

[0105] Step S16. Use the K-means clustering algorithm to process the abnormal point map and determine the boundary range of the temperature abnormal area.

[0106] Step S17. For the temperature field data within the boundary range, use the gradient descent method to generate the final gradient mapping image and complete the temperature field analysis.

[0107] Specifically, when obtaining the temperature field data of the welding area, first collect the temperature data through the sensor array arranged around the welding area. The sensor array consists of 16 thermocouples with a spacing of 10 mm and a sampling frequency of 100 times per second. The collected temperature data is processed for noise using the Kalman filter algorithm, and the filter parameters are set as the process noise covariance of 1 and the measurement noise covariance of 5 to ensure the accuracy of the data. The processed temperature data is used to generate a continuous temperature field distribution through the two-dimensional interpolation algorithm. The interpolation algorithm uses bilinear interpolation with an interpolation step of 1 mm to obtain high-resolution temperature field data. Based on the temperature field data, use the gradient calculation algorithm to generate the first temperature gradient mapping image. The gradient calculation uses the Sobel operator to calculate the temperature gradient value of each pixel point, and the gradient value range is from 0 to 50 degrees Celsius per millimeter. Finally, use image processing technology to enhance the gradient mapping image, using the histogram equalization algorithm with an enhancement coefficient of 5 to improve the contrast and clarity of the image for subsequent analysis and application.

[0108] S2. Determine the first residual stress distribution data from the first temperature gradient mapping image through thermo-structural coupled finite element analysis.

[0109] Optionally, this step further includes:

[0110] Step S21, obtaining the temperature gradient distribution from the first temperature gradient mapping image, and determining the first temperature field data through thermal-structural coupling analysis.

[0111] Step S22, calculating the heat conduction effect by using the finite element method according to the first temperature field data, and obtaining the preliminary distribution of the first stress field.

[0112] Step S23, judging the thermal-structural interaction influence through the preliminary distribution of the first stress field and combining with the temperature gradient change, and determining the second stress field data.

[0113] Step S24, obtaining the residual stress distribution characteristics from the second stress field data, and optimizing through coupling analysis to obtain the first residual stress distribution data.

[0114] Step S25, for the first residual stress distribution data, analyzing the boundary characteristics of the distribution data by using the support vector machine algorithm, and determining the third stress field data.

[0115] Step S26, obtaining the mapping relationship between the stress field and the temperature field through the third stress field data, judging the data consistency, and obtaining the final residual stress distribution data.

[0116] Step S27, extracting the key distribution characteristics from the final residual stress distribution data, generating a mapping image through a visualization tool, and determining the analysis result.

[0117] Specifically, to determine the first residual stress distribution data from the first temperature gradient mapping image through thermal-structural coupled finite element analysis, it is first necessary to convert the temperature gradient mapping image into the node temperature data in the finite element model.

[0118] For example, assuming that the image resolution is 1024×768, and the temperature range corresponding to each pixel is from 0°C to 500°C, the pixel temperature values are mapped to the finite element mesh nodes through the bilinear interpolation algorithm to ensure the spatial continuity of the temperature field. Next, the thermal-structural coupled finite element analysis method is used, with the temperature field as the input, to calculate the thermal expansion effect of the material. Assuming that the material is 316L stainless steel, its thermal expansion coefficient is 15×10^-6 / °C, the elastic modulus is 193GPa, and the Poisson's ratio is 3. By solving the thermoelastic constitutive equation, the thermal strain of each node is calculated, and combined with the mechanical property parameters of the material, the stress field is further solved.

[0119] For example, at a certain node, the temperature change is 200°C, the thermal strain is 0.0033, and the thermal stress calculated by Hooke's law is 635 MPa. Finally, the thermal stress field is superimposed on the initial stress field to obtain the first residual stress distribution data. Assuming that the initial stress field is uniformly distributed with a magnitude of 100 MPa, the final residual stress is 735 MPa. Through the above process, the residual stress distribution inside the material can be accurately determined, providing data support for subsequent process optimization and performance evaluation.

[0120] S3. Extract the coefficient of thermal expansion and welding speed parameters according to the first residual stress distribution data to generate a first deformation offset vector.

[0121] Optionally, this step further includes:

[0122] Step S31. Obtain the first residual stress distribution data, and analyze and extract the coefficient of thermal expansion and welding speed parameters to obtain a first parameter set.

[0123] Step S32. Calculate the deformation offset by using the finite element analysis method through the first parameter set to obtain a first offset result.

[0124] Step S33. Extract the offset vector features from the first offset result to generate a first deformation offset vector.

[0125] Step S34. For the first deformation offset vector, use the support vector machine algorithm to judge the vector consistency to obtain a consistency index.

[0126] Step S35. If the consistency index is lower than the consistency threshold, recalculate the offset vector by adjusting the coefficient of thermal expansion to obtain a second offset vector.

[0127] Step S36. According to the second offset vector, use the linear regression algorithm to predict the final deformation trend to obtain trend data.

[0128] Step S37. Generate a description of the final deformation offset vector through the trend data.

[0129] Specifically, based on the first residual stress distribution data, the coefficient of thermal expansion and the welding speed parameters are first extracted through finite element analysis. Assuming that the maximum stress value in the residual stress distribution data is 300 MPa and the minimum stress value is 50 MPa, the coefficient of thermal expansion is calculated to be 12×10^-6 / °C through the stress-strain relationship. The welding speed parameter is calculated through the welding heat input formula Q = ηUI / v, where η is the welding efficiency taken as 0.8, U is the welding voltage taken as 24 V, I is the welding current taken as 200 A, and v is the welding speed taken as 5 mm / s, and the welding heat input Q = 768 J / mm is obtained. Then, using the thermo-elastoplastic theory, combined with the extracted coefficient of thermal expansion and welding speed parameters, numerical simulation is carried out using finite element software to generate the first deformation offset vector. Assuming that the simulation results show that the maximum deformation is 5 mm and the minimum deformation is 1 mm, and the deformation direction is mainly along the welding direction. By analyzing the deformation offset vector, the welding process parameters can be further optimized, such as adjusting the welding speed to 6 mm / s, recalculating the welding heat input to 640 J / mm, and performing simulation verification again to check whether the deformation amount is within the allowable range. The whole process is automatically processed by information technology to ensure the accuracy of data and the reliability of the process.

[0130] S4. If the first deformation offset vector exceeds the geometric accuracy threshold, adjust the heat input power and welding speed parameters to generate the first optimized parameter combination.

[0131] Optionally, this step further includes:

[0132] Step S41. If the first deformation offset vector exceeds the geometric accuracy threshold, obtain the offset data through vector analysis to determine the extent of the exceeded range.

[0133] Step S42. According to the offset data, use the linear regression algorithm to calculate the adjustment ratios of the heat input power and the welding speed to obtain the preliminary adjustment values.

[0134] Step S43. Through the preliminary adjustment values, update the parameters for the heat input power and the welding speed to generate the first set of adjusted parameters.

[0135] Step S44. Apply the first set of adjusted parameters to the welding process to obtain a new deformation offset vector.

[0136] Step S45. If the new deformation offset vector still exceeds the geometric accuracy threshold, obtain the new offset data and judge whether further optimization is required.

[0137] Step S46. According to the judgment result, use the gradient descent algorithm to iteratively adjust the power parameter and the speed parameter, initialize the parameters as the first set of adjusted parameters, and set the learning rate to 01 to obtain the second optimized parameter combination.

[0138] Step S47: Apply the second optimized parameter combination to the welding process and analyze the change trend of the deformation offset vector. If the change amplitude of the deformation offset vector is less than 1 mm for three consecutive times, it is determined that the change trend is stable.

[0139] Step S48: If the change trend is stable and the offset vector meets the geometric accuracy threshold, generate the first optimized parameter combination through the final parameter adjustment plan.

[0140] Specifically, during the welding process, if the first deformation offset vector detected by the real-time monitoring system exceeds the preset geometric accuracy threshold (e.g., 0.5 mm), the system will automatically trigger the optimization algorithm for adjustment. First, based on the finite element analysis model, the system will calculate the influence degree of the current heat input power and welding speed on the deformation. Assuming the current heat input power is 2000 W and the welding speed is 10 mm / s, the system finds through iterative calculation that reducing the heat input power to 1800 W and increasing the welding speed to 12 mm / s can effectively reduce the deformation offset vector to less than 0.3 mm. Then, the system will call the machine learning model to predict the feasibility of this optimized parameter combination based on historical welding data and verify its stability through simulating the welding process. If the simulation results meet the expectations, the system will automatically update the welding parameters and store the optimized parameter combination (heat input power 1800 W, welding speed 12 mm / s) in the database for subsequent welding tasks to call. During this process, the system will continuously monitor the welding quality to ensure that the deformation offset vector always remains within the offset threshold range, thereby achieving high-precision welding control.

[0141] S5: Reallocate the heat input distribution according to the first optimized parameter combination and generate the second temperature field distribution data by using the power segmented regulation algorithm.

[0142] Optionally, this step further includes:

[0143] Step S51: Obtain the heat input data from the heat sensor through the first optimized parameter combination and determine the initial input distribution.

[0144] Step S52: Calculate the power segmentation result by using the linear segmentation method according to the initial input distribution.

[0145] Step S53: Run the linear regression algorithm for the power segmentation result to obtain the adjusted distribution data.

[0146] Step S54: Extract the temperature features from the adjusted distribution data to generate the preliminary temperature field distribution.

[0147] Step S55: If the preliminary temperature field distribution exceeds the temperature threshold, optimize the parameter combination through the gradient descent method and calculate the new input distribution.

[0148] Step S56: Perform data calculation using the least squares method according to the new input distribution to determine the second temperature field distribution data.

[0149] Step S57: Verify the field distribution consistency by comparing the second temperature field distribution data with the temperature threshold, and output the final result.

[0150] Specifically, according to the first optimization parameter combination, the system first redistributes the heat input distribution and generates the second temperature field distribution data using the power segmented regulation algorithm. In specific implementation, the system divides the heat input power into three intervals: low power interval (0 - 500W), medium power interval (500 - 1000W), and high power interval (1000 - 1500W). By analyzing the first temperature field distribution data, the system identifies that the high temperature area is mainly concentrated in the central position. Therefore, the high power interval is applied to the central area, the medium power interval is applied to the transition area, and the low power interval is applied to the edge area. The system uses the linear interpolation algorithm to ensure smooth transition of power within the interval and avoid temperature mutation. When generating the second temperature field distribution data, the system simulates the heat conduction process through finite element analysis (FEA), calculates the temperature value of each node, and generates the temperature field distribution map. By comparing the first and second temperature field distribution data, the system finds that the second temperature field distribution is more uniform. The temperature in the central area drops from 1200°C to 1100°C, and the temperature in the edge area rises from 800°C to 900°C. The overall temperature gradient decreases, and the thermal stress distribution is more reasonable. The system further optimizes the heat input distribution, improving the uniformity and stability of the temperature field.

[0151] S6: Extract the residual stress relief data from the second temperature field distribution data, and adjust the cooling parameters to generate the first cooling configuration when the stress reduction rate is lower than the reduction rate threshold.

[0152] Optionally, this step further includes:

[0153] Step S61: Extract the residual stress relief data from the second temperature field distribution data and obtain the stress change characteristics using data extraction techniques.

[0154] Step S62: Calculate the stress reduction rate through the stress change characteristics to obtain the reduction rate value.

[0155] Step S63: If the reduction rate value is lower than the reduction rate threshold, determine the adjusted cooling parameters through the genetic algorithm.

[0156] Step S64: Generate the first cooling configuration according to the adjusted cooling parameters to obtain the preliminary cooling plan.

[0157] Step S65: Extract the temperature field distribution characteristics from the preliminary cooling plan and judge the distribution consistency using principal component analysis.

[0158] Step S66: Adjust the second temperature field distribution data based on the distribution consistency result to obtain the optimized temperature field data.

[0159] Step S67: Re-extract the residual stress relief data according to the optimized temperature field data and determine whether it meets the preset conditions.

[0160] Specifically, when extracting the residual stress relief data from the second temperature field distribution data, first process the temperature field data through finite element analysis software and calculate the residual stress distribution using a thermo-elastoplastic model.

[0161] For example, during the cooling process of a certain aluminum alloy component, the temperature field data shows that when the cooling rate is too fast, the residual stress is concentrated at the edge of the component, and the maximum stress value reaches 250 MPa. Through the stress reduction rate calculation formula, that is, stress reduction rate = (initial stress - current stress) / initial stress × 100%, it is found that the stress reduction rate is only 15%, which is lower than the reduction rate threshold of 30%. At this time, the system automatically adjusts the cooling parameters, reduces the cooling rate from 10 degrees Celsius per minute to 5 degrees Celsius per minute, and adopts a gradient cooling strategy, that is, maintaining a higher cooling rate at the initial stage of cooling and gradually decreasing it later. Through finite element simulation verification, the adjusted cooling configuration increases the residual stress reduction rate to 35%, and the maximum stress value drops to 160 MPa, effectively alleviating the problem of residual stress concentration. At the same time, the system further optimizes the cooling path according to the temperature field distribution data to ensure cooling uniformity and avoid local overheating or overcooling phenomena, thereby improving the overall performance of the component.

[0162] S7: Perform thermal stress release treatment using the first cooling configuration to generate the first flatness data.

[0163] Optionally, this step further includes:

[0164] Step S71: Perform thermal stress release treatment using the first cooling configuration to generate the first flatness data.

[0165] Step S72: Obtain the first flatness data through a sensor to determine the initial flatness value.

[0166] Step S73: If the initial flatness value exceeds the flatness threshold, adjust the cooling configuration parameters to obtain the optimized flatness data.

[0167] Step S74: According to the optimized flatness data, use the support vector machine algorithm to judge the thermal stress release effect and determine the effect index.

[0168] Optionally, use the following algorithm to evaluate the thermal stress release effect:

[0169]

[0170] Among them, γ represents the evaluation index of thermal stress release effect, m represents the number of samples, and y i represents the actual stress value, represents the predicted stress value, and σ y represents the stress standard deviation.

[0171] Step S75: For the effect index, obtain the temperature distribution data of the heat treatment process to get the stress release trend.

[0172] Step S76: Through the stress release trend, use the linear regression algorithm to predict the change of flatness value and obtain the prediction result.

[0173] Step S77: According to the prediction result, adjust the release treatment process to generate the final flatness data.

[0174] Specifically, when performing thermal stress release treatment with the first cooling configuration, first collect the temperature distribution data on the surface of the workpiece through a thermal imager. The temperature range is from 200°C to 500°C, and the accuracy is ±1°C. Subsequently, use finite element analysis software to model the temperature data, and perform numerical simulation using the heat conduction equation and thermoelasticity equation to calculate the thermal stress distribution inside the workpiece. The simulation results show that the maximum thermal stress appears in the central region of the workpiece, reaching 350 MPa. To optimize the cooling effect, the system automatically adjusts the cooling rate, reducing the cooling rate from the initial 10°C / min to 5°C / min to reduce thermal stress concentration. Then, perform a three-dimensional scan of the workpiece surface through a laser scanner to generate the first flatness data, and the scanning accuracy is 0.01 mm. Data analysis shows that the maximum unevenness on the surface of the workpiece is 15 mm, mainly concentrated in the edge region. Based on these data, the system further optimizes the cooling path, adopts a segmented cooling strategy, and adjusts the cooling rate in the edge region to 3°C / min, while maintaining 5°C / min in the central region. Finally, through re-scanning verification, the maximum unevenness on the surface of the workpiece is reduced to 0.08 mm, meeting the process requirements.

[0175] S8: If the first flatness data does not reach the geometric accuracy threshold, generate the first process correction plan through real-time feedback of finite element iteration.

[0176] Optionally, this step further includes:

[0177] Step S81: If the first flatness data does not reach the geometric accuracy threshold, perform iterative calculation through the finite element method to generate the initial correction parameters.

[0178] Step S82: Obtain the calculation result of the initial correction parameters through real-time feedback to determine whether it meets the geometric accuracy requirements.

[0179] Step S83, if the calculation result still fails to reach the set geometric accuracy threshold, then analyze the deviation distribution characteristics by data analysis to obtain the adjustment direction.

[0180] Step S84, update the boundary conditions of the finite element method according to the adjustment direction to generate the second process correction plan.

[0181] Step S85, through the feedback mechanism, compare the matching degree between the second process correction plan and the first flatness data to judge the improvement amplitude.

[0182] Step S86, if the improvement amplitude does not meet the expectation, then optimize the second process correction plan through iterative calculation to obtain the final correction parameters.

[0183] Step S87, update the process correction process with the final correction parameters to generate a plan that meets the accuracy judgment.

[0184] Specifically, when it is detected that the first flatness data does not reach the geometric accuracy threshold, the system first processes the data through the finite element analysis algorithm. Assume that the initial flatness is 5 mm, while the geometric accuracy threshold is 3 mm. The system uses the iterative method to perform stress analysis on the workpiece through the finite element model and calculates that under the existing process conditions, the deformation of the workpiece is 2 mm. Based on this, the system generates the first process correction plan and adjusts the processing parameters, such as reducing the cutting speed from 100 m per minute to 80 m and reducing the feed rate from 1 mm per revolution to 0.08 mm to reduce the thermal deformation and mechanical stress during processing. The system further verifies the corrected process parameters through finite element simulation and predicts that the flatness will be improved to 25 mm, approaching the geometric accuracy threshold. At the same time, the system combines the material property database to optimize the tool path to ensure the stability during processing. Through real-time monitoring and data analysis, the system continuously adjusts the process parameters until the flatness data is stable within 3 mm, meeting the geometric accuracy requirements.

[0185] S9. Conduct a full-process simulation verification on the first process correction plan to obtain the fatigue life and thermal conductivity index data.

[0186] Optionally, this step further includes:

[0187] Step S91, conduct a full-process simulation verification on the process plan through a preset simulation tool to obtain the initial simulation data.

[0188] Step S92, extract the relevant characteristics of the fatigue life and thermal conductivity from the initial simulation data to determine the characteristic data set.

[0189] Step S93, use the support vector machine algorithm to classify the characteristic data set to judge the distribution interval of the fatigue life.

[0190] Step S94, analyze the stability of the process plan throughout the process through the distribution interval to obtain the stability parameter.

[0191] Step S95, if the stability parameter exceeds the stability threshold, adjust the parameters of the correction plan to obtain the adjusted plan data.

[0192] Step S96, re-run the simulation verification according to the adjusted plan data to obtain the updated life index and thermal conductivity index.

[0193] Step S97, compare the updated life index and thermal conductivity index with the initial simulation data to determine the final index data.

[0194] Specifically, in the full-process simulation verification of the first process correction plan, first use finite element analysis software to model the corrected process, specifically including the definition of material properties, the setting of boundary conditions, and the application of loads.

[0195] For example, set the Young's modulus of the material to 210 GPa, the Poisson's ratio to 3, and apply a cyclic fatigue load with a maximum stress of 500 MPa, a minimum stress of 50 MPa, and a frequency of 10 Hz. Then, calculate the fatigue life of the material through a fatigue life prediction algorithm, such as the Miner linear cumulative damage theory. During the simulation, use the Paris formula to calculate the crack growth rate, where the material constants C and m are taken as 0×10^-10 and 0 respectively. Through these calculations, the fatigue life after the corrected process is obtained as 2×10^6 cycles. Subsequently, use the heat conduction analysis module to calculate the thermal conductivity of the material. In the simulation, set the thermal conductivity of the material to 200 W / m·K, and solve the heat conduction equation by the finite element method to obtain the thermal conductivity after the corrected process as 190 W / m·K. Finally, compare the simulation results with the initial process to analyze the influence of the correction plan on the fatigue life and thermal conductivity.

[0196] For example, the fatigue life of the initial process is 0×10^6 cycles, and the thermal conductivity is 180 W / m·K. After the corrected process, the fatigue life increases by 20%, and the thermal conductivity increases by 6%. Through this series of analyses, the effectiveness of the corrected process is verified, and data support is provided for further optimization.

[0197] S10, judge whether the long-term stability meets the preset cycle number standard according to the fatigue life index data.

[0198] Optionally, this step further includes:

[0199] Step S101, obtain the index data through the fatigue life data, and process the original data using data cleaning technology to obtain the cleaned data set.

[0200] Step S102: Extract the cycle number feature from the cleaned dataset, calculate the feature value using statistical analysis methods, and determine the cycle number distribution.

[0201] Step S103: Generate a stability judgment basis according to the cycle number distribution. If the feature value exceeds the preset cycle threshold, it is marked as abnormal data.

[0202] Step S104: Conduct data analysis on the abnormal data, identify the abnormal pattern through a support vector machine, and obtain the classification result.

[0203] Step S105: Obtain the classification result to generate an analysis result, use K-means clustering to group the results, and determine the stability trend after grouping.

[0204] Step S106: Judge the long-term stability through the stability trend after grouping, use a comparison method to compare with the number standard, and obtain the final judgment result.

[0205] Step S107: Extract the standard judgment conclusion from the final judgment result, verify the consistency through logical verification technology, and obtain the verified judgment data.

[0206] Specifically, in the analysis of fatigue life index data, first, experimental data is collected. For example, when the cycle number is 10,000 times, the fatigue strength of the material is 200 MPa. Combining with the stress-life curve of the material, the Miner linear cumulative damage theory is used for fatigue life prediction. According to the Miner theory, when the cumulative damage value reaches 1, the material will undergo fatigue failure. Assuming the damage value per cycle is 0.0001, the predicted fatigue life is 10,000 cycles. Further, through simulation calculation using finite element analysis software, the stress distribution of the material under actual working conditions is simulated, and the stress concentration coefficient at the key part is obtained as 5. Applying the stress concentration coefficient to the fatigue life calculation, the corrected fatigue life is 4,000 cycles. To evaluate the long-term stability, the preset cycle number standard is set as 5,000 times. Comparing the corrected fatigue life with the standard, it is found that 4,000 cycles is lower than the standard of 5,000 times. Therefore, it is judged that the long-term stability of the material does not meet the requirements. On this basis, an optimization algorithm is used to improve the material structure. By adjusting the material thickness and shape, the stress concentration coefficient is reduced to 8, and the recalculated fatigue life is 5,556 cycles, which meets the preset standard at this time. The whole process is completed through automated data processing and simulation calculation to ensure the accuracy and reliability of the analysis results.

[0207] Please refer to Figure 2 , the second aspect of the present invention provides an intelligent welding system for radiators based on the control of welding residual stress, and uses the method described above to perform intelligent welding on the radiators. The system includes:

[0208] A temperature field data acquisition module, which is used to acquire the temperature field data of the welding area and generate a first temperature gradient mapping image by using a sensor array;

[0209] A first residual stress analysis module, which is used to determine the first residual stress distribution data from the first temperature gradient mapping image through thermo-structural coupled finite element analysis;

[0210] A deformation offset vector generation module, which is used to extract the thermal expansion coefficient and welding speed parameters from the first residual stress distribution data and generate a first deformation offset vector;

[0211] A parameter optimization module, which is used to adjust the heat input power and welding speed parameters to generate a first optimized parameter combination if the first deformation offset vector exceeds a preset geometric accuracy threshold;

[0212] A temperature field distribution generation module, which is used to redistribute the heat input distribution according to the first optimized parameter combination and generate second temperature field distribution data by using a power segmented regulation algorithm;

[0213] A cooling configuration generation module, which is used to extract residual stress mitigation data from the second temperature field distribution data and adjust the cooling parameters to generate a first cooling configuration when the stress reduction rate is lower than the reduction rate threshold;

[0214] A flatness data generation module, which is used to perform thermal stress release treatment by using the first cooling configuration and generate first flatness data;

[0215] A process correction module, which is used to generate a first process correction scheme through finite element iteration with real-time feedback if the first flatness data does not reach the geometric accuracy threshold;

[0216] A simulation verification module, which is used to perform full-process simulation verification on the first process correction scheme and obtain fatigue life and thermal conductivity index data;

[0217] A stability judgment module, which is used to judge whether the long-term stability meets the preset cycle number standard according to the fatigue life index data.

[0218] An intelligent welding method and system for radiators based on controlling welding residual stress provided by the present invention obtain temperature field data of the welding area through a sensor array, generate a temperature gradient mapping image, and determine the residual stress distribution by using thermo-structural coupled finite element analysis. Based on this, the coefficient of thermal expansion and welding speed parameters are extracted to generate a deformation offset vector. When the offset exceeds the offset threshold, the present invention adjusts the heat input power and welding speed, optimizes the parameter combination and redistributes the heat input distribution, uses a power segmented adjustment algorithm to generate a new temperature field distribution, extracts residual stress relief data, and if the stress reduction rate is insufficient, adjusts the cooling parameters for thermal stress release. Through real-time feedback finite element iteration, a process correction scheme is generated and full-process simulation verification is carried out to obtain fatigue life and thermal conductivity indexes. The present invention realizes the precise control of residual stress during the welding process and improves the quality and structural stability of intelligent welding.

[0219] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the present invention to the specific implementation manners. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principle and practical application of the present invention, so that those skilled in the relevant technical fields can understand and utilize the present invention well. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A radiator intelligent welding method based on welding residual stress control, characterized in that: The method comprises: S1, acquiring temperature field data of the welding area, and generating a first temperature gradient mapping image using a sensor array; S2, determining first residual stress distribution data from the first temperature gradient mapping image by thermal-structural coupling finite element analysis; S3, extracting thermal expansion coefficient and welding speed parameters according to the first residual stress distribution data, and generating a first deformation offset vector; S4, if the first deformation offset vector exceeds the geometric accuracy threshold, adjusting the heat input power and welding speed parameters to generate a first optimized parameter combination; S5, reallocating heat input distribution according to the first optimization parameter combination, and generating second temperature field distribution data using a power segmentation adjustment algorithm; S6, extracting residual stress relief data from the second temperature field distribution data, and adjusting cooling parameters to generate a first cooling configuration when the stress reduction rate is lower than a reduction rate threshold; S7, performing thermal stress release processing using the first cooling configuration to generate first flatness data; S8, if the first flatness data does not reach the geometric accuracy threshold, generating a first process correction plan through finite element iteration with real-time feedback; S9, performing full-process simulation verification on the first process correction scheme to obtain fatigue life and thermal conductivity index data; S10, judging whether the long-term stability meets a preset cycle number standard according to the fatigue life index data.

2. The method according to claim 1, characterized in that The step S1, obtaining the temperature field data of the welding area and using the sensor array to generate a first temperature gradient mapping image, includes: Step S11, collecting temperature field data of the welding area through a sensor array to generate an initial temperature distribution image; Step S12, using the VGG16 model in the convolutional neural network to process the initial temperature distribution image and extract the temperature gradient features in the image; Step S13, if the extracted temperature gradient feature exceeds the temperature threshold, a bilinear interpolation algorithm is used to optimize the gradient mapping accuracy to generate a second temperature gradient image; Step S14, extracting the change trend of the temperature field according to the second temperature gradient image, and determining the hot spot distribution characteristics; Step S15, analyzing the spatial consistency of the temperature field through the hot spot distribution characteristics, and using the region growing algorithm to obtain an abnormal point map of the welding area; Step S16, using K-means clustering algorithm to process the abnormal point map to determine the boundary range of the temperature abnormal area; Step S17, for the temperature field data within the boundary range, a gradient descent method is used to generate a final gradient mapping image to complete the temperature field analysis.

3. The method according to claim 1, characterized in that: The step S2, determining first residual stress distribution data from the first temperature gradient mapping image by thermal-structural coupling finite element analysis, comprises: Step S21, acquiring temperature gradient distribution from the first temperature gradient mapping image, and determining first temperature field data through thermal-structural coupling analysis; Step S22, calculating the heat conduction effect using the finite element method according to the first temperature field data to obtain a preliminary distribution of the first stress field; Step S23, judging the thermal-structural interaction through the preliminary distribution of the first stress field and combining the temperature gradient change, and determining the second stress field data; Step S24, obtaining residual stress distribution characteristics from the second stress field data, and obtaining first residual stress distribution data through coupling analysis optimization; Step S25, for the first residual stress distribution data, using a support vector machine algorithm to analyze the boundary characteristics of the distribution data to determine the third stress field data; Step S26, obtaining a mapping relationship between the stress field and the temperature field through the third stress field data, judging the data consistency, and obtaining the final residual stress distribution data; Step S27, extracting key distribution features from the final residual stress distribution data, generating a mapping image through a visualization tool, and determining the analysis results.

4. The method according to claim 1, characterized in that The step S3, extracting the thermal expansion coefficient and the welding speed parameter according to the first residual stress distribution data to generate a first deformation offset vector, includes: Step S31, obtaining first residual stress distribution data, and obtaining a first parameter set by analyzing and extracting thermal expansion coefficient and welding speed parameters; Step S32, using the first parameter set, using a finite element analysis method to calculate the deformation offset to obtain a first offset result; Step S33, extracting an offset vector feature from the first offset result to generate a first deformation offset vector; Step S34, for the first deformation offset vector, using a support vector machine algorithm to determine the vector consistency and obtain a consistency index; Step S35, if the consistency index is lower than the consistency threshold, recalculating the offset vector by adjusting the thermal expansion coefficient to obtain a second offset vector; Step S36, predicting the final deformation trend using a linear regression algorithm according to the second offset vector to obtain trend data; Step S37, generating a final deformation offset vector description through trend data.

5. The method according to claim 1, characterized in that The step S4, if the first deformation offset vector exceeds the geometric accuracy threshold, adjusting the heat input power and welding speed parameters to generate a first optimization parameter combination, includes: Step S41, if the first deformation offset vector exceeds the geometric accuracy threshold, then obtaining offset data through vector analysis to determine the degree of exceeding the range; Step S42, according to the offset data, a linear regression algorithm is used to calculate the adjustment ratio of the heat input power and the welding speed to obtain a preliminary adjustment value; Step S43, updating the parameters of heat input power and welding speed by using the preliminary adjustment values ​​to generate a first adjustment parameter set; Step S44, applying the first adjustment parameter set to the welding process to obtain a new deformation offset vector; Step S45, if the new deformation offset vector still exceeds the geometric accuracy threshold, new offset data is obtained to determine whether further optimization is required; Step S46, according to the judgment result, the power parameter and the speed parameter are iteratively adjusted using a gradient descent algorithm, the initialization parameters are the first adjustment parameter set, the learning rate is set to 01, and a second optimization parameter combination is obtained; Step S47, applying the second optimized parameter combination to the welding process, analyzing the change trend of the deformation offset vector, and if the change amplitude of the deformation offset vector is less than 1 mm for three consecutive times, it is determined that the change trend is stable; Step S48: If the change trend is stable and the offset vector meets the preset geometric accuracy threshold, a first optimized parameter combination is generated through a final parameter adjustment scheme.

6. The method according to claim 1, characterized in that The step S5, reallocating the heat input distribution according to the first optimization parameter combination and using a power segmentation adjustment algorithm to generate second temperature field distribution data, includes: Step S51, obtaining heat input data from a thermal sensor through a first optimization parameter combination to determine an initial input distribution; Step S52, calculating the power segmentation result by using a linear segmentation method according to the initial input distribution; Step S53, running a linear regression algorithm for the power segmentation results to obtain adjusted distribution data; Step S54, extracting temperature features from the adjusted distribution data to generate a preliminary temperature field distribution; Step S55, if the preliminary temperature field distribution exceeds the temperature threshold, the parameter combination is optimized by the gradient descent method to calculate a new input distribution; Step S56, performing data calculation using the least square method according to the new input distribution to determine the second temperature field distribution data; Step S57, comparing the second temperature field distribution data with the temperature threshold, verifying the field distribution consistency, and outputting the final result.

7. The method according to claim 1, characterized in that The step S6, extracting residual stress relief data from the second temperature field distribution data, and adjusting cooling parameters to generate a first cooling configuration when the stress reduction rate is lower than a reduction rate threshold, includes: Step S61, extracting residual stress relief data from the second temperature field distribution data, and obtaining stress change characteristics using data extraction technology; Step S62, calculating the stress reduction rate through the stress change characteristics to obtain the reduction rate value; Step S63, if the reduction rate value is lower than the reduction rate threshold, determining the adjusted cooling parameter by using a genetic algorithm; Step S64, generating a first cooling configuration according to the adjusted cooling parameters to obtain a preliminary cooling solution; Step S65, extracting temperature field distribution characteristics from the preliminary cooling scheme, and using principal component analysis to determine distribution consistency; Step S66, adjusting the second temperature field distribution data according to the distribution consistency result to obtain optimized temperature field data; Step S67, re-extracting the residual stress relief data according to the optimized temperature field data to determine whether the preset conditions are met.

8. The method according to claim 1, characterized in that The step S7, using the first cooling configuration to perform thermal stress release processing to generate first flatness data, includes: Step S71, performing a thermal stress release process using a first cooling configuration to generate first flatness data; Step S72, obtaining first flatness data through a sensor to determine an initial flatness value; Step S73, if the initial flatness value exceeds the flatness threshold, the cooling configuration parameters are adjusted to obtain optimized flatness data; Step S74, judging the thermal stress release effect by using a support vector machine algorithm according to the optimized flatness data, and determining the effect index; Step S75, obtaining temperature distribution data of the heat treatment process according to the effect index to obtain the stress release trend; Step S76, predicting the change of the flatness value by using a linear regression algorithm based on the stress release trend to obtain a prediction result; Step S77, according to the prediction results, adjust the release processing flow to generate the final flatness data.

9. The method according to claim 1, characterized in that: The step S10, judging whether the long-term stability meets the preset cycle number standard according to the fatigue life index data, includes: Step S101, obtaining index data through fatigue life data, and processing the original data using data cleaning technology to obtain a cleaned data set; Step S102, extracting cycle number features from the cleaned data set, calculating feature values ​​using a statistical analysis method, and determining cycle number distribution; Step S103, generating a stability judgment basis according to the cycle number distribution, and if the characteristic value exceeds the preset cycle threshold, it is marked as abnormal data; Step S104, performing data analysis on the abnormal data, identifying abnormal patterns through a support vector machine, and obtaining classification results; Step S105, obtaining the classification results to generate analysis results, grouping the results using K-means clustering, and determining the stability trend after grouping; Step S106, judging long-term stability by the stability trend after grouping, using a comparison method to compare with the frequency standard to obtain a final judgment result; Step S107, extracting the standard judgment conclusion from the final judgment result, verifying the consistency through logic verification technology, and obtaining the verified judgment data.

10. A radiator intelligent welding system based on welding residual stress control, characterized in that: The method according to any one of claims 1 to 9 is used to perform intelligent welding on the radiator, and the system comprises: A temperature field data acquisition module is used to acquire temperature field data of the welding area and generate a first temperature gradient mapping image using a sensor array; A first residual stress analysis module, used to determine first residual stress distribution data from the first temperature gradient mapping image through thermal-structural coupling finite element analysis; A deformation offset vector generating module, used for extracting the thermal expansion coefficient and the welding speed parameter according to the first residual stress distribution data, and generating a first deformation offset vector; a parameter optimization module, configured to adjust heat input power and welding speed parameters to generate a first optimized parameter combination if the first deformation offset vector exceeds a geometric accuracy threshold; a temperature field distribution generating module, configured to reallocate heat input distribution according to the first optimization parameter combination and generate second temperature field distribution data by adopting a power segmentation adjustment algorithm; A cooling configuration generating module, configured to extract residual stress relief data from the second temperature field distribution data, and to adjust cooling parameters to generate a first cooling configuration when the stress reduction rate is lower than a reduction rate threshold; a flatness data generating module, configured to perform a thermal stress release process using the first cooling configuration to generate first flatness data; A process correction module, configured to generate a first process correction scheme through finite element iteration with real-time feedback if the first flatness data does not reach a geometric accuracy threshold; A simulation verification module, used to perform full-process simulation verification on the first process correction scheme to obtain fatigue life and thermal conductivity index data; The stability judgment module is used to judge whether the long-term stability meets the preset cycle number standard according to the fatigue life index data.

Images