Aluminum / copper dissimilar metal laser welding parameter multi-objective optimization method

Abstract

The invention discloses a multi-objective optimization method for aluminum / copper dissimilar metal laser welding parameters, which comprises the following steps of: establishing a BKA-GBRT model for optimizing GBRT by using BKA by taking buckling deformation, bearable maximum tension and welding cost of an aluminum / copper dissimilar metal laser welding part as optimization objectives, then searching a Pareto leading edge in various laser welding parameters by using an MOSPO algorithm based on the model, and finally, establishing a BKA-GBRT model for optimizing GBRT by using BKA; and finally, the optimal laser welding parameter combination is determined through CRITIC-TOPSIS comprehensive evaluation. According to the method, on the premise that the welding cost is reduced, the buckling deformation amount of the welding piece after aluminum / copper dissimilar metal laser welding is reduced, the bearable maximum pulling force of the aluminum / copper dissimilar metal laser welding piece is improved, and then the reliability of aluminum / copper dissimilar metal laser welding is improved with low cost; the method is particularly suitable for welding aluminum / copper dissimilar metal in an electrode of a lithium battery of an electric vehicle.

Description

Technical Field

[0001] The invention relates to a laser welding parameter multi-objective optimization method, in particular to an aluminum / copper dissimilar metal laser welding parameter multi-objective optimization method based on BKA-GBRT and MOSPO, belonging to the technical field of laser welding. Background Art

[0002] Compared with traditional fuel vehicles, the use of electric vehicles will greatly reduce carbon emissions. How to improve the reliability of electric vehicles is a key issue in the vigorous development of electric vehicles. Batteries are the core component of electric vehicles, and improving the connection technology of electric vehicle batteries is one of the important ways to improve the reliability of electric vehicles. At present, electric vehicle batteries on the market are composed of a large number of lithium-ion batteries connected in series or in parallel. Therefore, improving the aluminum / copper dissimilar metal welding technology in lithium-ion battery electrodes can increase battery performance and improve the reliability of electric vehicles.

[0003] However, due to the differences in chemical and physical properties between aluminum and copper, defects such as pores, cracks and intermetallic compounds (IMCs) will appear in the joints during the welding process, which greatly affects the performance of the welded joints. Laser welding can accurately control the heat input, and because of its low heat input, it can better control the generation of IMCs. It can reduce welding defects by optimizing laser power, welding speed, defocusing amount and filling materials to improve the performance of aluminum / copper dissimilar metal welded joints. Therefore, laser welding process has become one of the key technologies in battery module production.

[0004] The existing laser welding process parameter optimization methods generally include two parts: laser welding process parameter prediction and parameter optimization. As a method to establish the nonlinear mapping relationship between laser welding process parameters and welding quality assessment parameters, the prediction accuracy of the laser welding parameter prediction model directly affects the optimization effect of laser welding process parameters. Commonly used welding parameter prediction methods include empirical calculation, experimental summary, numerical simulation and other methods. Empirical calculation and experimental summary are often based on experience, and the calculation results may have a large error with the actual situation. Numerical simulation requires professionals to spend a lot of time and energy for simulation and consumes more computing resources. Response surface method and machine learning method are widely used because they can quickly and accurately predict welding deformation. However, when dealing with complex problems, polynomial response surface method requires a large number of experimental samples and has high cost. Machine learning method is gradually promoted because of its advantages of small sample size and high prediction accuracy. However, due to the different performance of different machine learning models in different situations, for a specific data set, the existing welding process parameter prediction method using a single prediction model may not necessarily achieve the optimal prediction effect.

[0005] In addition, in the research on adding filler materials, industry researchers have tried to add filler materials such as nickel foil and tin-based fillers. Although certain results have been achieved, there are still many problems with the welds and the welding effect is not ideal. Summary of the invention

[0006] In view of the above problems, the present invention provides a multi-objective optimization method for aluminum / copper dissimilar metal laser welding parameters, which can reduce the warping deformation of the welded parts after aluminum / copper dissimilar metal laser welding and increase the maximum tensile force that the aluminum / copper dissimilar metal laser welded parts can withstand while reducing the welding cost, thereby improving the reliability of aluminum / copper dissimilar metal laser welding at a low cost. It can provide a theoretical basis and data support for obtaining the optimal process parameter combination for aluminum / copper dissimilar metal laser welding, and is particularly suitable for aluminum / copper dissimilar metal welding in lithium-ion battery electrodes of electric vehicles.

[0007] To achieve the above purpose, the multi-objective optimization method of aluminum / copper dissimilar metal laser welding parameters specifically includes the following steps:

[0008] Step 1: The warping deformation of the welded part y 1 , Maximum tensile force that the weldment can withstand 2 and welding cost y 3 As the optimization target, the laser power x 1 , welding speed x 2 , defocus amount x 3 and the amount of filler material added x 4 As the optimization design variable, determine the value range of the optimization design variable, design an orthogonal test, perform laser welding, and obtain the test results;

[0009] The calculation formula of welding cost is as follows:

[0010]

[0011] Where: y 3 is the welding cost, unit: ￥; x 1 is the laser power, in W; x 2 is the welding speed, in mm / min; x 4 is the amount of filling material added, in g;

[0012] Step 2, based on the test results, the laser power x 1 , welding speed x 2 , defocus amount x 3 and the amount of filler material added x 4 As input, the warpage of the weldment y 1 and the maximum tensile force that the weldment can withstand 2As output, a BKA-GBRT model for optimizing GBRT with BKA is constructed;

[0013] Step 3, perform multi-objective optimization on the BKA-GBRT model using the MOSPO algorithm to obtain the Pareto frontier of the optimization objective;

[0014] Step 4: Determine the optimal combination of process parameters in the Pareto frontier of the obtained optimization target through the CRITIC-TOPSIS evaluation method.

[0015] Furthermore, in Step 2, an optimization strategy is used to achieve a balance between the three optimization objectives in the optimization process to obtain a better optimization result. The optimization strategy includes taking the inverse of the prediction function of the maximum tensile force that the weldment can withstand and introducing a weight coefficient of 10 and introducing a weight coefficient of 2 into the prediction function of the welding cost. The mathematical model of the laser welding parameters of the optimization result is expressed as follows:

[0016]

[0017] Furthermore, in Step 2, the specific steps of using BKA to optimize GBRT are as follows:

[0018] ① Load test data;

[0019] ② Define the hyperparameters of BKA, including the population size, the maximum number of iterations, and the upper and lower bounds of the two hyperparameters to be optimized: NumLearningCycles and LearnRate;

[0020] ③ Randomly sort the data and divide it into training set and test set;

[0021] ④Define an anonymous function fobj as the target function;

[0022] ⑤Call the BKA algorithm to obtain the optimal fitness value, optimal solution and algorithm convergence curve;

[0023] ⑥ Use the optimal solution obtained by the BKA algorithm to train the GBRT model;

[0024] ⑦ Calculate the mean square error, determination coefficient and mean relative error based on the predicted value of the test set and the actual test results, and update the optimal parameters based on these evaluation indicators;

[0025] ⑧Output the optimal result and generate a scatter plot of the actual value and the predicted value.

[0026] Furthermore, in Step 3, the BKA-GBRT model is optimized by the MOSPO algorithm, as follows:

[0027] ① Load the BKA-GBRT model trained in Step 2;

[0028] ② Define MOSPO parameters, including the maximum number of iterations, population size, and maximum archive capacity;

[0029] ③Execute MOSPO multi-objective algorithm:

[0030] First, the population and color combination strategy are initialized, where the Position of each particle is randomly generated within the upper and lower limits of the parameters, and its objective function value Cost is calculated;

[0031] Secondly, the color combination strategy generates new solutions. There are four color combination strategies. Each strategy generates a new particle position and updates its objective function value.

[0032] Finally, the archive is updated and solutions are selected. After each new solution is generated, the non-dominated solution is added to the archive, and the non-dominated solution is re-screened through the function. When the archive exceeds the set maximum capacity, the redundant solutions are deleted.

[0033] ④ Filter the results. If there are too few solution sets, continue to iterate. When the number of iterations reaches the defined maximum number of iterations, stop iterating.

[0034] ⑤Save the Pareto frontier and draw a three-dimensional graph based on the Pareto frontier.

[0035] Furthermore, in Step 4, the optimal combination of process parameters is determined in the Pareto frontier of the obtained optimization target by the CRITIC-TOPSIS evaluation method, as follows:

[0036] ① Load the Pareto frontier obtained in Step 3;

[0037] ②Standardize the data;

[0038] ③ Calculate the standard deviation and correlation coefficient matrix;

[0039] ④ Calculate the CRITIC weights and standardize the weights by dividing the amount of information for each indicator by the sum of all information;

[0040] ⑤ Apply TOPSIS method:

[0041] First, the weighted normalized decision matrix is ​​calculated, and the weighted normalized matrix is ​​obtained by multiplying the normalized data matrix by the corresponding weights;

[0042] Secondly, identify the positive ideal solution and the negative ideal solution;

[0043] Finally, the distance between each solution and the positive ideal solution and the negative ideal solution is calculated;

[0044] ⑥ Calculate the relative proximity and sort the relative proximity in descending order to obtain the ranking of the schemes;

[0045] ⑦ Find the best group according to the output ranking, and then find the corresponding process parameter combination according to the Pareto frontier.

[0046] Furthermore, the filling material is high entropy alloy powder.

[0047] Compared with the prior art, the multi-objective optimization method of aluminum / copper dissimilar metal laser welding parameters takes the warpage deformation, the maximum tensile force that can be tolerated and the welding cost of aluminum / copper dissimilar metal laser welded parts as optimization targets, establishes a BKA-GBRT model that uses BKA to optimize GBRT, and then uses the MOSPO algorithm based on the model to search the Pareto frontier in various laser welding parameters, and finally determines the optimal laser welding parameter combination through CRITIC-TOPSIS comprehensive evaluation. The test results show that the error between the welding test results predicted by the optimization method used in the present invention and the actual experimental value is controlled within 5%, the model prediction accuracy is good, and the warpage deformation and welding cost of the optimized welded parts are significantly reduced compared with the results before optimization, and the maximum tensile force that can be tolerated by the optimized welded parts is significantly improved compared with the results before optimization. Welding with the optimized laser welding parameters can significantly reduce the welding cost and improve the product quality, which can provide a theoretical basis and data support for obtaining the optimal process parameter combination of aluminum / copper dissimilar metal laser welding, and is particularly suitable for aluminum / copper dissimilar metal welding in lithium-ion battery electrodes of electric vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 1 is a comparison diagram of the predicted value and the experimental value of the welded part by the BKA-GBRT model of the present invention, wherein (a) is a comparison diagram of the warpage deformation of the welded part, and (b) is a comparison diagram of the maximum tensile force that the welded part can withstand;

[0049] Figure 2 It is a Pareto frontier diagram obtained by the present invention through MOSPO multi-objective optimization based on the BKA-GBRT model;

[0050] Figure 3 The figures are the tensile fracture micromorphology images of the sample before and after optimization of the present invention, wherein (a) is the tensile fracture micromorphology image of the sample before optimization, and (b) is the tensile fracture micromorphology image of the sample after optimization. DETAILED DESCRIPTION

[0051] The multi-objective optimization method for aluminum / copper dissimilar metal laser welding parameters takes the warping deformation, the maximum tensile force and the welding cost of aluminum / copper dissimilar metal laser welded parts as optimization objectives, and establishes a gradient boosting regression tree (GBRT) model optimized by the black-winged kite algorithm (BKA). Then, based on the model, a multi-objective stochastic paint optimizer (MOSPO) is used to search the Pareto frontier among various laser welding parameters. Finally, the optimal combination of laser welding parameters is determined through the CRITIC-TOPSIS comprehensive evaluation.

[0052] The present invention is specifically described below by taking T2 copper and 6061 aluminum alloy (40 mm×10 mm×1 mm) as base materials, FeCoNiCrTi high entropy alloy powder as filler material, and an aluminum / copper dissimilar metal laser welding test in an overlapping manner as an example.

[0053] Step 1. Determine the optimization target, select laser welding process parameters and their value ranges, use the value ranges of each process parameter as the experimental design space, design orthogonal experiments based on the value ranges of each process parameter, and perform laser welding to obtain experimental results.

[0054] The experiment selected four parameters that have a significant impact on welding quality: laser power (x 1 , unit: W), welding speed (x 2 , unit: mm / min), defocus (x 3 , unit: mm) and the amount of filler material added (x 4 , unit: g) as the design variable. The test parameters were preliminarily adjusted through trial welding. No obvious welding defects were found within the adjusted parameter range. The specific test factors and their level parameters are detailed in Table 1 below.

[0055] Table 1 Experimental factors and level parameters

[0056]

[0057] As shown in Table 1, the laser power (x 1 ), welding speed (x 2 ) and defocus (x 3 ) These three factors are divided into six levels, the amount of filler material added (x 4) were divided into three levels, and 36 different combinations of laser welding parameters were formed according to the principle of orthogonal experimental design. The experiment used a JHM-1GX-1000F fiber laser welding machine for laser welding. In order to protect the weld from oxidation during the welding process, argon was used as the shielding gas, and its flow rate was set to 10L / min.

[0058] In the process of laser welding of aluminum / copper dissimilar metals, the deformation degree of the weldment has a crucial influence on the installation accuracy and surface quality of the weldment; the strength of the weldment is directly related to the reliability of the battery electrode of the electric vehicle; at the same time, the low process cost is considered. In view of this, the warpage deformation amount (y 1 , unit: mm), maximum tensile force that can be sustained (y 2 , unit: N) and welding cost (y 3 , unit: ￥) These three key parameters are used as optimization targets, among which welding cost (y 3 ) is the product of the hourly welding cost and the welding time (assuming a utilization factor of 0.8) plus the cost of consumed filler material, calculated as follows:

[0059]

[0060] In order to measure the warpage of welded parts, 36 welded parts were accurately measured using a 3D laser scanner. The room temperature tensile test module equipped with a dynamic thermal simulation test machine was then used to evaluate the maximum tensile force that these welded parts can withstand. 1 ), the maximum tensile force that can be sustained (y 2 ) and welding cost (y 3 ) The results are shown in Table 2 below.

[0061] Table 2 Test results

[0062]

[0063]

[0064] Step 2. Based on the experimental results, a gradient boosting regression tree (GBRT) model optimized by the black-winged kite algorithm (BKA) was established.

[0065] As a prediction model in ensemble learning, the GBRT model is essentially an additive model with decision trees as basic learners. It was first proposed by Friedmanl and has the advantages of high prediction accuracy, fast calculation speed, strong robustness to outliers, and not prone to overfitting. Although the GBRT model performs well in many practical applications, its hyperparameters have a great impact on model performance. If the hyperparameters are not set properly, overfitting is likely to occur. Therefore, it is necessary to optimize the hyperparameters of the GBRT model.

[0066] BKA is an efficient global optimization algorithm, mainly used to solve complex optimization problems without calculating derivatives. It simulates the migration behavior of black-winged kites (the algorithm simulates the leader strategy in the group and searches for the optimal solution through dynamic selection. This process ensures that the algorithm achieves a good balance between the global and local and avoids falling into the local optimum too early) and predation behavior (in each iteration, the algorithm fine-tunes the current solution to jump out of the local optimum and improve the efficiency of global search), generates random solutions in the initial population, and adds these solutions to the historical sample set to update the current optimal solution. The global search capability is increased through the Cauchy mutation strategy, and the leadership strategy is used to balance exploration and development, so that the optimal solution can be found in a shorter time and the algorithm can be ensured to converge to the optimal solution.

[0067] The laser power (x 1 ), welding speed (x 2 ), defocus amount (x 3 ) and the amount of filler material added (x 4 ) as input, and the warpage deformation of the weldment (y 1 ) and the maximum tensile force that the weldment can withstand (y 2 ) as output, and two BKA-GBRT models are constructed. The specific steps of using BKA to optimize GBRT are as follows:

[0068] ① Loading data: The test data is loaded from the Excel file (Table 2) through the xlsread function. The goal is to improve the model performance by optimizing the relationship between these test parameters and the test results.

[0069] ②BKA parameter setting: define the hyperparameters of BKA, including the population size (pop), the maximum number of iterations (T), and the upper and lower bounds of the two hyperparameters to be optimized: NumLearningCycles (number of iterations) and LearnRate (learning rate).

[0070] ③Data preprocessing: The data were randomly sorted and divided into training and test sets. In order to verify the prediction accuracy and effectiveness of the model, 80% of the data were randomly selected from the 36 groups of experiments as the training set of the BKA-GBRT model, and the remaining 20% ​​were used as the test set.

[0071] ④Define the objective function: Define an anonymous function fobj as the objective function, which adjusts the hyperparameters (NumLearningCycles and LearnRate) of the GBRT model through BKA to minimize the mean square error (MSE).

[0072] ⑤Execute BKA optimization: call the BKA algorithm, and the input parameters include the population size (pop), the maximum number of iterations (T), the upper and lower limits of the test parameters (lb and ub), and the objective function (fobj); BKA returns the results including the optimal fitness value (Best_Fitness_BKA, that is, the optimal objective function value), the optimal solution (Best_Pos_BKA, that is, the optimal combination of NumLearningCycles and LearnRate), and the algorithm convergence curve (Convergence_curve, recording the best fitness of each iteration).

[0073] ⑥ Use the optimal parameters to train the final model: Use the optimal solution (Best_Pos_BKA) obtained by the BKA algorithm to train the GBRT model, where the hyperparameter NumLearningCycles is rounded to an integer and the hyperparameter LearnRate retains decimals.

[0074] ⑦ Prediction and evaluation: Calculate the mean square error (MSE) and determination coefficient (R) based on the predicted value of the test set and the actual test results. 2 ) and mean relative error (MRE), and update the optimal parameters based on these evaluation indicators.

[0075] ⑧Save the optimal result and visualize it: Output the optimal result and generate a scatter plot of the true value and the predicted value to visualize the model effect.

[0076] Use the trained BKA-GBRT model to perform prediction analysis and compare the predicted data results with the actual test results. Figure 1 As shown, Figure 1 (a) is the comparison between the predicted value of the warpage deformation of the welded part by the BKA-GBRT model and the experimental value. Figure 1 (b) is the comparison between the predicted value of the maximum tensile force that the weldment can withstand by the BKA-GBRT model and the experimental value. According to the prediction results of the BKA-GBRT model, the model error can be calculated as shown in Table 3 below.

[0077] Table 3 Model training results

[0078]

[0079] It can be seen from Table 3 that the determination coefficients R between the predicted values ​​of the two BKA-GBRT models and the experimental results are2 All of them are greater than 0.9, which shows that the model has good learning ability, and the mean relative error value MRE is less than 5%, which further verifies the accurate description of the model on the nonlinear relationship between laser welding parameters and post-weld performance.

[0080] Step 3: Use the multi-objective stochastic paint optimizer (MOSPO) to perform multi-objective optimization to obtain the Pareto frontier of the optimization target.

[0081] The optimization target of the present invention is to minimize the warping deformation of the welded part (y 1 ) and welding cost (y 3 ), while maximizing the maximum tensile force that the weldment can withstand (y 2 ), considering that the three optimization objectives are significantly different in magnitude and optimization direction and are equally important, an optimization strategy is introduced. The optimization strategy includes taking the inverse of the prediction function of the maximum tensile force that the weldment can withstand and introducing a weight coefficient of 10 and introducing a weight coefficient of 2 for the prediction function of the welding cost. The purpose is to achieve a balance between the three objectives in the optimization process. By appropriately amplifying the influence of the objectives, the influence of the three objectives is fully considered, so as to more effectively handle the relationship between the three, and finally obtain a better optimization result. The mathematical model of the laser welding parameters of the optimization result is expressed as follows:

[0082]

[0083] Through this method, the optimization of deformation, maximum tensile force and welding cost can be taken into account in the multi-objective optimization process to ensure the effectiveness and feasibility of the final welding process.

[0084] Stochastic paint optimizer (SPO) is a meta-heuristic algorithm proposed by Ali et al. in 2020. The SPO algorithm is inspired by the color aesthetics in painting art, has efficient exploration capabilities and the ability to generate new solutions with multiple strategies.

[0085] Nima et al. proposed a multi-objective version of SPO, called multi-objective stochastic paint optimizer (MOSPO). MOSPO solves multi-objective optimization problems by finding a set of approximate optimal solutions, so that it can handle multiple conflicting objective function problems. MOSPO introduces an archive to maintain the non-dominated solution set and store multiple Pareto frontiers. The archive is updated and maintained based on the individual non-dominance and grid mechanism to retain the optimal solution set. Compared with commonly used algorithms such as multi-objective particle swarm optimization and multi-objective ant lion optimization, MOSPO can find a more accurate and evenly distributed Pareto frontier within a limited number of iterations.

[0086] The BKA-GBRT model is optimized by multi-objective optimization using the MOSPO algorithm, as follows:

[0087] ① Load model: Load the BKA-GBRT model trained in Step 2.

[0088] ②MOSPO parameter settings: define parameters such as the maximum number of iterations (max_iter), population size (Pop), and maximum archive capacity (ArchiveMaxSize).

[0089] ③Execute MOSPO multi-objective algorithm:

[0090] First, the population and color combination strategy are initialized, where the 'Position' of each particle is randomly generated within the upper and lower limits, and its objective function value 'Cost' is calculated. The 'Position' of a particle represents the position of the particle in the solution space of the optimization problem. Specifically, the position of each particle is a multidimensional vector, and the value of each dimension corresponds to a parameter in the optimization problem. This experiment has three parameters, so when the algorithm is initialized, the three dimensions of the particle's 'Position' will be randomly assigned a value, which is between the upper and lower limits of the parameters. The 'Cost' of a particle represents the performance of the particle under the objective function (i.e., the quality of the solution). In the optimization problem, the objective function is used to evaluate the quality of a given solution, and 'Cost' is the value calculated by the 'Position' of the particle under the objective function. In this experiment, after the optimization strategy of Step 2, the optimization problem of the three parameters has become a minimization problem, so the three smaller 'Cost's represent a better combination solution. Each particle has a historical best solution 'Best', which is similar to the initial color in the color combination strategy.

[0091] Secondly, the color combination strategy generates new solutions. There are four color combination strategies, each of which generates a new particle position and updates its objective function value. These strategies use color combinations between different groups to ensure the diversity of solutions and enhance the global and local refinement capabilities of the search.

[0092] Finally, the archive is updated and solutions are selected. Each time a new solution is generated, the program will add the non-dominated solution to the archive and re-screen the non-dominated solution through the function. When the archive exceeds the set maximum capacity, the 'DeleteFromRep' function is used to delete the redundant solutions to ensure that the solutions in the archive remain diverse.

[0093] ④ Result screening and continued iteration: MOSPO relies on non-dominated solution sets to generate the Pareto frontier. Therefore, if the solution set is too small, it means that the algorithm’s search in the current iteration is insufficient and needs to continue iterating. When the number of iterations reaches the defined maximum number of iterations, the iteration is stopped.

[0094] ⑤Save and visualize the Pareto front: Draw a three-dimensional graph based on the Pareto front to show the Pareto fronts of the three optimization objectives.

[0095] The Pareto frontier obtained by MOSPO multi-objective optimization of the gradient boosting regression tree model optimized by the black kite optimization algorithm is as follows: Figure 2 shown.

[0096] Step 4: Determine the optimal combination of process parameters through the CRITIC-TOPSIS evaluation method.

[0097] Figure 2 There are 38 Pareto frontiers in China, and a comprehensive evaluation method is needed to select the optimal solution. The CRITIC-TOPSIS comprehensive evaluation method combines the objective weight calculation of the CRITIC method and the multi-criteria ranking of the TOPSIS method. By considering the correlation and variability between indicators, it provides a more comprehensive and scientific basis for decision-making. It can process data of different dimensions and scales and is suitable for various complex decision-making scenarios.

[0098] The optimal combination of process parameters is determined in the Pareto frontier of the obtained optimization target by the CRITIC-TOPSIS evaluation method, as follows:

[0099] ① Load data: Load the Pareto frontier obtained in Step 3.

[0100] ②Data preprocessing: standardize the data.

[0101] ③ Calculate the standard deviation and correlation coefficient matrix: The standard deviation measures the degree of dispersion of the data. Indicators with a large degree of dispersion have a larger amount of information; the correlation coefficient is used to measure the correlation between indicators. The information provided by indicators with high correlation may be redundant, while indicators with low correlation provide more independent information.

[0102] ④ Calculate the information content and CRITIC weight: The weight reflects the importance of each indicator. The greater the information content, the greater the weight of the indicator. The weight is standardized by dividing the information content of each indicator by the sum of all information content, so that the sum of all weights is equal to 1.

[0103] ⑤ Apply TOPSIS method:

[0104] First, the weighted standardized decision matrix is ​​calculated. The weighted standardized matrix is ​​obtained by multiplying the standardized data matrix by the corresponding weights. This step multiplies the standardized value of each indicator by its weight to reflect the relative importance of each indicator in the decision-making process.

[0105] Next, positive and negative ideal solutions are identified. The positive ideal solution is the best value (maximum value) for each metric, while the negative ideal solution is the worst value (minimum value). These solutions represent the best and worst solutions for each metric, respectively.

[0106] Finally, the distance between each solution and the positive and negative ideal solutions is calculated. By comparing the distance between the solution and the ideal solution and the negative ideal solution, the quality of the solution can be measured.

[0107] ⑥ Calculate relative proximity and sort: Relative proximity reflects the degree to which each solution is close to the negative ideal solution. The larger the value, the closer the solution is to the positive ideal solution (optimal solution). By sorting the relative proximity in descending order, the final ranking of the solutions is obtained.

[0108] ⑦ Select the optimal combination of process parameters: find the best combination based on the output ranking, and then find the corresponding combination of process parameters based on the Pareto frontier.

[0109] The 38 Pareto frontiers obtained by MOSPO are evaluated based on the CRITIC-TOPSIS comprehensive evaluation of the scores of each solution group. Figure 2 The first group of solutions circled in the figure has the best comprehensive evaluation among the 38 groups, and the corresponding solution is: the warpage deformation of the weldment (y 1 ) is 0.576mm, and the maximum tensile force (y 2 ) is 800.42N, welding cost (y 3 ) is 0.139￥. At this time, the corresponding laser welding parameter combination is: laser power (x 1 ) is 677.234W, welding speed (x 2) is 639.265mm / min, and the defocusing amount (x 3 ) is 2.745 mm, and the amount of high entropy alloy filler material added (x 4 ) is 0.05g.

[0110] According to the actual accuracy of JHM-1GX-1000F fiber laser welding machine, the optimal laser welding parameter combination is finally determined as follows: laser power (x 1 ) is 677.2W, welding speed (x 2 ) is 639.3mm / min, and the defocusing amount (x 3 ) is 2.75 mm, and the amount of high entropy alloy filler material added (x 4 ) is 0.05g. These parameters were used to perform laser welding of aluminum / copper dissimilar metals on a JHM-1GX-1000F fiber laser welding machine. Argon was used as the shielding gas during welding, and the flow rate was 10L / min. In order to verify the accuracy of the optimization results, three laser welding tests were carried out on each set of laser welding parameters before and after optimization, and the warping deformation of the welded parts was measured using a three-dimensional laser scanner, and the maximum tensile force that the welded parts could withstand was determined using the room temperature tensile method. The experimental values ​​of the three welding specimens were averaged to obtain the optimized experimental results. The comparison of the results before and after optimization is shown in Table 4 below.

[0111] Table 4 Experimental results after optimization

[0112]

[0113] It can be seen from Table 4 that the errors between the welding test results predicted by the optimization method used in the present invention and the actual experimental values ​​are 1.21% and 1.57%, respectively, both within the acceptable range of 5%, which indicates that the optimization method has high accuracy. The experimental value of the warping deformation of the optimized welded part is 0.583mm, the maximum tensile force that the welded part can withstand is 813.00N, and the welding cost is 0.139￥. When the parameters before optimization are used, the experimental values ​​of the warping deformation of the welded part, the maximum tensile force that the welded part can withstand, and the welding cost are 0.730mm, 543.00N, and 0.156￥, respectively. Compared with the parameters before optimization, the warping deformation of the welded part is reduced by 20.14%, the maximum tensile force that the welded part can withstand is increased by 49.72%, and the welding cost is reduced by 10.90%. The results show that welding with optimized laser welding parameters can significantly reduce welding costs and improve product quality, verifying the feasibility of the optimization method.

[0114] Scanning electron microscopy (SEM) was used to compare and analyze the tensile fracture morphology of the samples before and after the optimization of laser welding parameters. Figure 3 As shown. Figure 3As shown in (a), under the laser welding parameters before optimization, the fracture surface of the tensile fracture has more and larger cracks and pores. The presence of cracks and pores can easily cause stress concentration and cause brittle fracture of the welded part, thereby reducing the maximum tensile force that the welded part can withstand. Figure 3 As shown in (b), by optimizing the laser welding parameters, the tensile fracture morphology of the sample has been significantly improved. The optimized parameters reduce the number of cracks and pores and form a large number of deep dimples. The presence of dimples enables the welded part to undergo more plastic deformation and absorb more energy before fracture, thereby increasing the maximum tensile force that the welded part can withstand. The above results verify the significant influence of the optimized parameters on the fracture properties, and can provide a reliable theoretical basis and data support for the application of aluminum / copper dissimilar metal laser welding.

Claims

1. A multi-objective optimization method for aluminum / copper dissimilar metal laser welding parameters, characterized in that: The specific steps include: Step 1, taking the warpage deformation y1 of the welded part, the maximum tensile force y2 that the welded part can withstand and the welding cost y3 as the optimization targets, taking the laser power x1, welding speed x2, defocusing amount x3 and filling material addition amount x4 as the optimization design variables, determining the value range of the optimization design variables, designing orthogonal experiments, performing laser welding and obtaining the test results; The calculation formula of welding cost is as follows: In the formula: y3 is the welding cost, unit is ￥; x1 is the laser power, unit is W; x2 is the welding speed, unit is mm / min; x4 is the amount of filler material added, unit is g; Step 2, based on the test results, the laser power x1, welding speed x2, defocus amount x3 and filler material addition amount x4 are taken as input, and the warpage deformation y1 of the welded part and the maximum tensile force y2 that the welded part can withstand are taken as output to construct a BKA-GBRT model for optimizing GBRT using BKA; Step 3, perform multi-objective optimization on the BKA-GBRT model using the MOSPO algorithm to obtain the Pareto frontier of the optimization objective; Step 4: Determine the optimal combination of process parameters in the Pareto frontier of the obtained optimization target through the CRITIC-TOPSIS evaluation method.

2. The multi-objective optimization method for aluminum / copper dissimilar metal laser welding parameters according to claim 1 is characterized in that: In Step 2, an optimization strategy is used to achieve a balance between the three optimization objectives in the optimization process to obtain a better optimization result. The optimization strategy includes taking the inverse of the prediction function of the maximum tensile force that the weldment can withstand and introducing a weight coefficient of 10 and introducing a weight coefficient of 2 for the prediction function of the welding cost. The mathematical model of the laser welding parameters of the optimization result is expressed as follows:

3. The multi-objective optimization method for aluminum / copper dissimilar metal laser welding parameters according to claim 1 is characterized in that: In Step 2, the specific steps of using BKA to optimize GBRT are as follows: ① Load test data; ② Define the hyperparameters of BKA, including the population size, the maximum number of iterations, and the upper and lower bounds of the two hyperparameters to be optimized: NumLearningCycles and LearnRate; ③ Randomly sort the data and divide it into training set and test set; ④Define an anonymous function fobj as the target function; ⑤Call the BKA algorithm to obtain the optimal fitness value, optimal solution and algorithm convergence curve; ⑥ Use the optimal solution obtained by the BKA algorithm to train the GBRT model; ⑦ Calculate the mean square error, determination coefficient and mean relative error based on the predicted value of the test set and the actual test results, and update the optimal parameters based on these evaluation indicators; ⑧Output the optimal result and generate a scatter plot of the actual value and the predicted value.

4. The multi-objective optimization method for aluminum / copper dissimilar metal laser welding parameters according to claim 1 is characterized in that: In Step 3, the BKA-GBRT model is optimized by the MOSPO algorithm, as follows: ① Load the BKA-GBRT model trained in Step 2; ② Define MOSPO parameters, including the maximum number of iterations, population size, and maximum archive capacity; ③Execute MOSPO multi-objective algorithm: First, the population and color combination strategy are initialized, where the Position of each particle is randomly generated within the upper and lower limits of the parameters, and its objective function value Cost is calculated; Secondly, the color combination strategy generates new solutions. There are four color combination strategies. Each strategy generates a new particle position and updates its objective function value. Finally, the archive is updated and solutions are selected. After each new solution is generated, the non-dominated solution is added to the archive, and the non-dominated solution is re-screened through the function. When the archive exceeds the set maximum capacity, the redundant solutions are deleted. ④ Filter the results. If there are too few solution sets, continue to iterate until the number of iterations reaches the defined maximum number of iterations. ⑤Save the Pareto frontier and draw a three-dimensional graph based on the Pareto frontier.

5. The multi-objective optimization method for aluminum / copper dissimilar metal laser welding parameters according to claim 1 is characterized in that: In Step 4, the optimal combination of process parameters is determined in the Pareto frontier of the obtained optimization target by the CRITIC-TOPSIS evaluation method, as follows: ① Load the Pareto frontier obtained in Step 3; ②Standardize the data; ③ Calculate the standard deviation and correlation coefficient matrix; ④ Calculate the CRITIC weights and standardize the weights by dividing the amount of information for each indicator by the sum of all information; ⑤ Apply TOPSIS method: First, the weighted normalized decision matrix is ​​calculated, and the weighted normalized matrix is ​​obtained by multiplying the normalized data matrix by the corresponding weights; Secondly, identify the positive ideal solution and the negative ideal solution; Finally, the distance between each solution and the positive ideal solution and the negative ideal solution is calculated; ⑥ Calculate the relative proximity and sort the relative proximity in descending order to obtain the ranking of the schemes; ⑦ Find the best group according to the output ranking, and then find the corresponding process parameter combination according to the Pareto frontier.

6. The multi-objective optimization method for aluminum / copper dissimilar metal laser welding parameters according to claim 1 is characterized in that: The filling material is high entropy alloy powder.

Images