Multi-objective optimization method for steel / aluminum dissimilar metal laser welding

Abstract

The invention discloses a multi-objective optimization method for steel / aluminum dissimilar metal laser welding, which comprises the following steps of: constructing an RIME-CatBoost model for optimizing CatBoost by using RIME by taking the buckling deformation of a welding part, the maximum tensile force which can be borne by the welding part and the welding cost as optimization objectives, then searching a Pareto leading edge in various laser welding parameters by using an MOAVOA algorithm on the basis of the model, and finally, optimizing the Pareto leading edge by using the MOAVOA algorithm. And finally, an optimal laser welding parameter combination is determined through CRITIC comprehensive evaluation. According to the method, on the premise that the welding cost is reduced, the buckling deformation of the welding part after steel / aluminum laser welding is reduced, the bearable maximum tension of the steel / aluminum laser welding part is improved, and then the reliability of steel / aluminum laser welding is improved with low cost; and a theoretical basis and data support can be provided for obtaining the optimal technological parameter combination of steel / aluminum laser welding.

Description

Technical Field

[0001] The invention relates to a laser welding multi-objective optimization method, in particular to a Catboost-MOAVOA-based multi-objective optimization method for steel / aluminum dissimilar metal laser welding parameters and CeO2 addition amount, belonging to the technical field of laser welding. Background Art

[0002] Automobile lightweighting means reducing the curb weight of the automobile as much as possible while ensuring the strength and safety performance of the automobile, thereby improving the power of the automobile, reducing fuel consumption, and reducing exhaust pollution. The steel / aluminum heterogeneous metal composite structure has the characteristics of high strength and light weight. While ensuring safety, it can effectively reduce the weight of the transportation vehicle. Therefore, using steel / aluminum welded structures instead of pure steel welded structures is currently recognized as one of the most effective methods for automobile lightweighting.

[0003] However, due to the differences in chemical and physical properties between steel and aluminum, hard and brittle intermetallic compounds (IMCs) will inevitably be produced during the welding process. The presence of IMCs will seriously reduce the mechanical properties of the welded joint. Therefore, how to inhibit the generation of IMCs has become the core issue in obtaining high-quality steel / aluminum welded joints.

[0004] Laser welding (LBW) is a common method for welding dissimilar metals such as steel and aluminum. By adjusting the process parameters such as laser power, welding speed and defocus, the size of the heat-affected zone in the welding area can be controlled and the generation of Fe-Al IMCs can be reduced. However, there is a nonlinear relationship between laser welding parameters and welding results. The traditional method of optimizing laser welding process parameters relies on experiments and experience, which requires a lot of human and material resources and has high process costs.

[0005] 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. Summary of the invention

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

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

[0008] Step 1, take 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, take the laser power x1, welding speed x2, defocus amount x3 and CeO2 addition amount x4 as the optimization design variables, determine the value range of the optimization design variables, design orthogonal experiments, perform laser welding and obtain the test results;

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

[0010]

[0011] 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 CeO2 addition amount, unit is g;

[0012] Step 2, based on the test results, the laser power x1, welding speed x2, defocus x3 and CeO2 addition x4 are taken as input, and the weld warpage y1 and the maximum tensile force y2 that the weld can withstand are taken as output to construct a RIME-CatBoost model that uses RIME to optimize CatBoost.

[0013] Step 3, use the MOAVOA algorithm to perform multi-objective optimization on the RIME-CatBoost model 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 evaluation method.

[0015] Furthermore, in Step 2, the inverse of the maximum tensile force that the weldment can withstand, the warpage deformation of the weldment, and the welding cost are used as the objective function of the optimization algorithm. The mathematical model of welding process optimization based on the RIME-CatBoost model is expressed as follows:

[0016]

[0017] Where: x1 is the laser power, x2 is the welding speed, x3 is the defocus amount, x4 is the CeO2 addition amount, y1 is the warping deformation of the welded part, y2 is the maximum tensile force that the welded part can withstand, and y3 is the welding cost.

[0018] Furthermore, in Step 2, the specific steps of using RIME to optimize CatBoost are as follows:

[0019] ① Take the learning rate and maximum depth of the CatBoost model as the hyperparameters that need to be optimized;

[0020] ② Set the value range of hyperparameters and the initial parameters of the RIME algorithm;

[0021] ③ Train the CatBoost model by updating the combination of hyperparameters in each iteration of the RIME algorithm;

[0022] ④ Calculate the fitness value of the corresponding hyperparameter combination according to the model performance index, and select the hyperparameter combination with the best fitness.

[0023] Furthermore, in Step 3, the RIME-CatBoost model is optimized by the MOAVOA algorithm, as follows:

[0024] ① Initialize MOAVOA parameters: parameters include the maximum number of iterations, population size, maximum archive capacity, and the number of objective functions;

[0025] ② Determine whether the current number of iterations has reached the maximum number of iterations: The algorithm starts from the initial number of iterations until the preset maximum number of iterations is reached. If the current number of iterations is less than the maximum number of iterations, the algorithm continues to iterate, otherwise it outputs the Pareto frontier and ends;

[0026] ③Fitness calculation: Each vulture represents a set of solutions, and the fitness of each vulture is obtained according to the RIME-CatBoost model and the welding cost calculation formula;

[0027] ④ Select the best vulture: In each iteration, select the vulture with the best performance as the leader according to the fitness value and store it in the archive;

[0028] ⑤ Update vulture positions: Update the positions of all vultures based on the leader vulture’s behavior and current strategy;

[0029] ⑥ Determine the file size: Check whether the current solution set size exceeds the file setting size. If the solution set size is smaller than the file size, continue to iterate. If the solution set size is larger than the file size, delete the solutions in the dense area through the grid mechanism to retain the diversity of solutions.

[0030] ⑦ Update archive: Update archive according to the criteria of Pareto frontier and retain non-dominated solutions;

[0031] ⑧Perform the next iteration: increase the number of iterations by 1 and go to step ② to iterate again.

[0032] Furthermore, in Step 4, when determining the optimal combination of process parameters in the Pareto frontier of the obtained optimization target by the CRITIC evaluation method, the obtained Pareto frontier is weighted and aggregated by using the weighted aggregation method to assign weights to each Pareto frontier. The expression of the comprehensive evaluation model is as follows:

[0033]

[0034] Where: Z is the comprehensive evaluation value of each experimental design, W j is the weight value of the jth indicator, y ij is the j evaluation index value of the ith solution, y i1 is the warpage of the welded parts in each test design, y i2 is the maximum tensile force that the weldment can withstand in each test design, y i3 is the welding cost in each set of experimental design.

[0035] Compared with the existing technology, the multi-objective optimization method of Benshan / aluminum dissimilar metal laser welding first takes laser power, welding speed, defocus amount, and CeO2 addition as input, and takes the warping deformation of the welded part and the maximum tensile force that the welded part can withstand as output, designs an orthogonal experiment and conducts a welding test to obtain the test results; then, based on the test results and the data measured by the welding samples, a RIME-CatBoost model that uses RIME to optimize CatBoost is constructed; then, the warping deformation of the welded part, the maximum tensile force that the welded part can withstand, and the calculated welding cost are taken as optimization targets, and the RIME-CatBoost model is subjected to multi-objective optimization through the MOAVOA algorithm; finally, the CRITIC comprehensive evaluation method is used to determine the optimal combination of welding parameters. 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 values ​​is controlled within 5%, the model prediction accuracy is good, and the warping deformation and welding cost of the optimized welded parts are significantly reduced compared with the results before optimization. The maximum tensile force that the optimized welded parts can withstand is significantly improved compared with the results before optimization. Welding with optimized laser welding parameters can significantly reduce welding costs and improve product quality, thereby achieving low-cost improvement in the reliability of steel / aluminum laser welding, and can provide a theoretical basis and data support for obtaining the optimal process parameter combination for steel / aluminum laser welding. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 1 is a comparison diagram of the effects before and after the optimization of the model of the present invention, wherein (a) is a comparison diagram of the predicted value and the actual value of the deformation of the weldment before and after the optimization, and (b) is a comparison diagram of the predicted value and the actual value of the maximum tensile force that the weldment can withstand before and after the optimization;

[0037] Figure 2 It is a Pareto frontier graph obtained by performing MOAVOA multi-objective optimization on the RIME-CatBoost model in the present invention;

[0038] Figure 3 is a welding sample diagram after optimization of the present invention;

[0039] Figure 4 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;

[0040] Figure 5 It is a flowchart of the present invention performing multi-objective optimization on the RIME-CatBoost model through the MOAVOA algorithm. DETAILED DESCRIPTION

[0041] In the process of laser welding of steel / aluminum dissimilar metals, the metallurgical reaction of the molten pool can be changed by adding elements such as Ti, Cu, Zn, Ni, and rare earth elements to form an intermediate layer. Since the use of the intermediate layer can not only reduce the mixing of Fe and Al elements in the molten pool and reduce the thickness of brittle Fe-Al IMCs, but also the elemental substances added in the intermediate layer can combine with Fe or Al elements to form new binary or ternary IMCs with good elasticity, which can improve the mechanical properties of the steel / aluminum welded joints, the introduction of the intermediate layer in the process of laser welding of steel / aluminum dissimilar metals for alloying metallurgical regulation can well inhibit the generation of Fe-Al IMCs. The present invention is based on the formation of an intermediate layer by adding rare earth element CeO2 to perform alloying metallurgical regulation and achieve the inhibition of IMCs in the process of laser welding of steel / aluminum dissimilar metals, and the purpose of obtaining a high-quality steel / aluminum welded joint is achieved by multi-objective optimization of CeO2 addition and laser welding parameters.

[0042] The multi-objective optimization method of steel / aluminum dissimilar metal laser welding first takes laser power, welding speed, defocus, and CeO2 addition as input, and takes the warping deformation of the welded part and the maximum tensile force that the welded part can withstand as output, and conducts welding tests; then, a CatBoost mathematical model is established based on the measured data of the welding sample, and the robust iterative multiview embedding (RIME) algorithm is used to optimize its hyperparameters; then, the warping deformation of the welded part, the maximum tensile force that the welded part can withstand, and the calculated welding cost are taken as optimization targets, and multi-objective optimization is performed based on the multi-objective african vultures optimization algorithm (MOAVOA); finally, the CRITIC comprehensive evaluation method is used to determine the optimal welding parameters.

[0043] The present invention is specifically described below by taking laser welding of DP780 dual-phase steel and 6016 aluminum alloy, which are widely used in the automotive industry, as an example.

[0044] 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 according to the value ranges of each process parameter, and perform laser welding to obtain experimental results.

[0045] A laser lap welding test scheme with an intermediate layer of CeO2 was designed. The particle size of CeO2 powder was 20 nm. The size of the DP780 dual-phase steel and 6016 aluminum alloy specimens was 40 mm × 10 mm × 1 mm. The chemical composition of the specimens is shown in Table 1. Acetone was used to clean the surface oil and oxide film before the welding test.

[0046] Table 1 Chemical composition of specimens (wt.%)

[0047]

[0048] The test selected four parameters that have a significant impact on welding quality: laser power (x1, unit w), welding speed (x2, unit mm / min), defocus (x3, unit mm) and CeO2 addition (x4, unit g) as influencing factors. According to the level range recommended for laser welding, the value range of the influencing factors was adjusted and defined by trial welding. Within the adjusted range, there were no obvious defects in the joint. The laser power (x1), welding speed (x2) and defocus (x3) were divided into 6 levels, and the CeO2 addition (x4) was divided into 3 levels, as shown in Table 2 below.

[0049] Table 2 Experimental factors and level parameters

[0050]

[0051] Using the four factors shown in Table 2, 36 different combinations of welding process parameters were generated according to the orthogonal experimental design principle. The laser welding test of steel / aluminum dissimilar metal specimens was carried out using a JHM-1GX-1000F fber laser welding machine. Argon was used as the shielding gas during the welding process at a rate of 10L / min.

[0052] The deformation of welded parts after laser welding of steel / aluminum dissimilar metals has a significant impact on the installation accuracy and appearance quality of welded components. At the same time, the maximum tensile force that welded parts can withstand is an important mechanical property index of metal materials and an important factor affecting the overall mechanical properties of automobiles. Therefore, the warpage deformation of welded parts (y1, unit mm) and the maximum tensile force that welded parts can withstand (y2, unit N) are taken as optimization targets.

[0053] In addition, welding cost (y3, unit ￥) is one of the most important factors in industrial production. In order to achieve low-cost welding, welding cost (y3, unit ￥) is also taken as an optimization target. The total welding cost is the product of the hourly welding cost and the welding time (assuming the utilization rate is 0.8) plus the cost of consuming the intermediate layer CeO2. The calculation formula of welding cost (y3) is as follows:

[0054]

[0055] In the formula: x1 is welding power, x2 is welding speed, and x4 is the amount of CeO2 added.

[0056] After welding, the deformation of 36 groups of welding samples was measured by a 3D laser scanner, and the maximum tensile force of the samples was tested by the room temperature tensile module of the dynamic thermal simulation test machine. The measured deformation, maximum tensile force and calculated welding cost are shown in Table 3.

[0057] Table 3 Test results

[0058]

[0059] Step 2, based on the experimental results, establish a CatBoost model optimized by the robust iterative multiview embedding (RIME) algorithm.

[0060] CatBoost is a machine learning algorithm based on gradient boosting decision trees, which is specifically designed to optimize categorical feature processing and prevent overfitting to achieve efficient and robust model performance. Compared with traditional models, CatBoost models have excellent performance, and through strategies such as ranking learning and goal-oriented coding, they can effectively process categorical features and improve model performance.

[0061] Two CatBoost models were constructed with laser power (x1), welding speed (x2), defocus (x3) and CeO2 addition (x4) as inputs, and weld warpage (y1) and the maximum tensile force (y2) that the weld can withstand as outputs. In order to prove the prediction accuracy and effectiveness of the model, 80% of the data from 36 orthogonal tests were randomly selected as the training set of the CatBoost model, and the remaining 20% ​​were used as the test set. The trained CatBoost model was then used for prediction analysis.

[0062] Although the CatBoost model has excellent performance, its hyperparameters have a great impact on the model performance. If the hyperparameters are not set appropriately, it is easy to cause overfitting. Therefore, it is necessary to optimize the hyperparameters of the CatBoost model.

[0063] The robust iterative multiview embedding (RIME) algorithm is an optimization algorithm based on physical phenomenon simulation. It is inspired by the process of frost and ice crystals gradually forming on a cooling surface. The algorithm draws on the growth characteristics of soft frost and uses the strong randomness and coverage of frost particles to enable the algorithm to quickly cover the entire search space and find the optimal solution to complex optimization problems.

[0064] The RIME algorithm is used to optimize the CatBoost hyperparameters. The specific steps are as follows:

[0065] ① Take the learning rate (learning_rate) and maximum depth (depth) of the CatBoost model as the hyperparameters that need to be optimized;

[0066] ② Set the value range of hyperparameters and the initial parameters of the RIME algorithm;

[0067] ③ Train the CatBoost model by updating the combination of hyperparameters in each iteration of the RIME algorithm;

[0068] ④ Calculate the fitness value of the corresponding hyperparameter combination according to the model performance index, and select the hyperparameter combination with the best fitness.

[0069] The RIME algorithm performs global search and local fine-tuning in the parameter space in an iterative manner, and uses a greedy selection mechanism to continuously update individuals and optimal solutions, so that the depth and learning rate of the CatBoost model reach the optimal configuration, thereby improving model performance and prediction accuracy.

[0070] The prediction error values ​​of the Catboost model for the warpage deformation of welded parts and the maximum tensile force that welded parts can withstand before and after optimization are shown in Tables 4 and 5.

[0071] Table 4 Prediction error of welded part warpage

[0072]

[0073] Table 5 Prediction error of the maximum tensile force that the weldment can withstand

[0074]

[0075] It can be seen from Tables 4 and 5 that the RIME-CatBoost model is more accurate in predicting the warpage deformation of welded parts and the maximum tensile force that welded parts can withstand. The predicted data results before and after optimization are compared with the actual welding test results. Figure 1 As shown, Figure 1 (a) shows the comparison between the warpage deformation of welded parts predicted by CatBoost model and the actual value before and after optimization. Figure 1 (b) shows the comparison between the maximum tensile force that the weldment can withstand predicted by the CatBoost model before and after optimization and the actual value.

[0076] Step 3, use the multi-objective African vultures optimization algorithm (MOAVOA) to perform multi-objective optimization and obtain the Pareto frontier of the optimization target.

[0077] The optimization of the optimization target of the present invention is to minimize the warpage deformation of the welded part (y1) and the welding cost (y3), and at the same time maximize the maximum tensile force (y2) that the welded part can withstand. Considering the significant differences in the magnitude and optimization direction of these three optimization targets and their equal importance, the inverse of the maximum tensile force that the welded part can withstand, the warpage deformation of the welded part and the welding cost are used as the objective function of the optimization algorithm. In order to balance the magnitude of the various factors of the objective function and facilitate comprehensive evaluation, the weight coefficient of the warpage deformation of the welded part is set to 1, the inverse coefficient of the maximum tensile force that the welded part can withstand is set to 1000, and the weight coefficient of the welding cost is set to 20. Then, the mathematical model of welding process optimization based on the RIME-CatBoost model can be summarized as follows:

[0078]

[0079] Where: x1 is the laser power, x2 is the welding speed, x3 is the defocus, x4 is the CeO2 addition, y1 is the warpage of the welded part, y2 is the maximum tensile force that the welded part can withstand, and y3 is the welding cost. y1 and y2 are predicted outputs of the RIME-CatBoost model, and y3 is calculated by the aforementioned calculation formula.

[0080] The African vultures optimization algorithm (AVOA) simulates the search behavior of African vultures in the process of foraging to find the optimal solution to the optimization problem. The multi-objective African vulture optimization algorithm (MOAVOA) adds three key mechanisms: archive mechanism, grid mechanism and leader selection mechanism on the basis of the single-objective artificial vulture optimization algorithm (AVOA) to enhance its performance in multi-objective optimization. MOAVOA finds a uniformly distributed Pareto frontier in the search space through the processes of initializing the population, calculating fitness, calculating hunger level, exploration phase, development phase, updating archives, iterative updating, etc., so as to effectively solve the multi-objective optimization problem.

[0081] The aforementioned mathematical model of welding process optimization based on RIME-CatBoost model is used as the objective function of MOAVOA, such as Figure 5 As shown in the figure, the MOAVOA algorithm is used to perform multi-objective optimization on the RIME-CatBoost model, as follows:

[0082] ① Initialize MOAVOA parameters: Initialize the parameters of the algorithm, including the maximum number of iterations, population size, maximum archive capacity, and the number of objective functions.

[0083] ② Determine whether the current number of iterations has reached the maximum number of iterations: The algorithm starts from the initial generation until the preset maximum number of iterations is reached. If the current number of iterations is less than the maximum number of iterations, the algorithm continues to iterate, otherwise it outputs the Pareto frontier and ends.

[0084] ③ Calculate the fitness of each vulture: Each vulture represents a set of solutions. The fitness of each vulture (i.e., solution) is obtained based on the RIME-CatBoost model and the welding cost calculation formula.

[0085] ④ Select the best vulture: In each iteration, the vulture with the best performance is selected as the leader (i.e., non-dominated solution) according to the fitness value and stored in the archive.

[0086] ⑤Update vulture positions: Update the positions of all vultures based on the leader vulture’s behavior and current strategy.

[0087] ⑥ Determine the file size: Check whether the current solution set size exceeds the file size. If the solution set size is smaller than the file size, continue to iterate. If the solution set size is larger than the file size, delete the solutions in the dense area through the grid mechanism to retain the diversity of solutions.

[0088] ⑦ Update archive: Update archive according to the Pareto frontier criterion and retain non-dominated solutions.

[0089] ⑧Perform the next iteration: increase the number of iterations by 1 and go to step ② to iterate again.

[0090] The Pareto frontier of the final output of CatBoost-MOAVOA is as follows Figure 2 As shown, a total of 23 Pareto frontiers are obtained.

[0091] Step 4: Determine the optimal combination of process parameters through the CRITIC evaluation method.

[0092] Figure 2 A total of 23 Pareto frontiers are obtained through MOAVOA optimization, and a comprehensive evaluation method is needed to select the optimal solution. The 23 Pareto frontiers obtained are weighted aggregated using a weighted aggregation method to assign appropriate weights to each Pareto frontier. The expression of the comprehensive evaluation model is as follows:

[0093]

[0094] Where: Z is the comprehensive evaluation value of each experimental design, W j is the weight value of the jth indicator, y ij is the j evaluation index value of the ith solution, y i1 is the warpage of the welded parts in each test design, yi2 is the maximum tensile force that the weldment can withstand in each test design, y i3 is the welding cost in each set of experimental design.

[0095] The comprehensive evaluation model fully considers the importance of each evaluation criterion by weighted aggregation of each evaluation criterion. This method can generate comprehensive evaluation values ​​covering each Pareto frontier, so as to comprehensively evaluate and rank different solutions. For the 23 groups of Pareto frontiers obtained by the MOAVOA optimization method, the CRITIC method was used to objectively determine the weight of each target. The target weights of the maximum tensile force that the welded parts can withstand, the warping deformation of the welded parts, and the welding cost are 0.58, 0.29, and 0.13, respectively. The calculated target weights are introduced into the above formula to obtain the comprehensive evaluation score of the Pareto frontier. Among them Figure 2 The circled solution evaluation score is the minimum value of the 23 Pareto frontiers, and the corresponding maximum tensile force that the welded part can withstand is 751.2N, the warpage of the welded part is 0.676mm, and the welding cost is 0.046￥. The corresponding laser welding parameter combination is: laser power is 593.20W, welding speed is 816.78mm / min, defocus is 11.62mm, and CeO2 addition is 0.0202g.

[0096] According to the actual accuracy of the JHM-1GX-1000F fber fiber laser welding machine, the optimal laser welding parameters were finally determined to be: laser power 593W, welding speed 816.8mm / min, defocus 11.60mm, CeO2 addition 0.02g. Using these determined optimal laser welding parameters, laser welding of steel / aluminum dissimilar metals was performed on the JHM-1GX-1000F fber fiber laser welding machine. Argon was used as the shielding gas during the welding process, and the flow rate was 10L / min. In order to verify the accuracy of the optimization results, three laser welding tests were carried out using the optimized laser welding parameters. The samples after welding are as follows Figure 3 The deformation of the sample was measured using a 3D laser scanner, and the maximum tensile force of the sample after welding was measured using the room temperature tensile module of the dynamic thermal simulation test machine. The test values ​​of the three welded samples were averaged to obtain the optimized test results. The comparison of the results before and after optimization is shown in Table 6 below.

[0097] Table 6 Comparison of the results before and after optimization and model prediction

[0098]

[0099] It can be seen from Table 6 that the errors between the experimental values ​​and the predicted values ​​after CatBoost-MOAVOA optimization are 3.7% and 4.8%, respectively, both less than 5%, which shows that the optimization method has high accuracy. The warpage deformation of the optimized welded part is 0.650mm, the maximum tensile force that the welded part can withstand is 715N, and the welding cost is 0.046￥. Compared with the experimental values ​​before optimization, the warpage deformation of the welded part of the optimized sample is reduced by 37.7%, the maximum tensile force that the welded part can withstand is increased by 20.6%, and the welding cost is reduced by 18%. This shows that using the optimized parameters for welding can not only significantly improve the performance of the welded joint, but also reduce the processing cost, verifying the feasibility of this optimization scheme.

[0100] Scanning electron microscopy (SEM) was used to compare and analyze the tensile fracture surfaces of the samples before and after the welding parameters were optimized. Figure 4 As shown. Figure 4 As shown in (a), a large number of cracks appeared on the fracture interface of the sample welded using the parameters before optimization. These cracks not only increased the brittleness of the material, but also led to local concentration of stress, which would significantly reduce the overall reliability and load-bearing capacity of the welded joint. Figure 4 As shown in (b), the welding test was carried out with the optimized laser welding parameters. It can be observed that the tensile fracture surface of the sample has been significantly improved. The optimized parameters reduced the number of cracks and a tear ridge appeared on the fracture surface, which indicates that the toughness and fracture mode of the welded joint have been improved. This improvement can reduce the stress concentration effect, thereby improving the overall strength and durability of the welded joint and reducing the risk of sudden fracture.

[0101] The reasons why optimized laser welding parameters can improve fracture performance are analyzed as follows: ① By optimizing laser welding parameters, the size of the heat affected zone in the welding area can be controlled, stress concentration and thermal stress can be reduced, thereby reducing the generation of Fe-Al IMCs. ② The addition of CeO2 can promote the formation and flow of the molten pool, thereby reducing the accumulation of thermal stress and reducing the hot cracking sensitivity of the welded joint. ③ The addition of CeO2 can effectively reduce the formation of intermetallic compounds (IMCs) during the welding process, and improve the quality and corrosion resistance of the welded joint by promoting alloying reactions, inhibiting the growth rate of IMCs, and reducing the presence of oxide inclusions.

[0102] The above results verify the significant influence of the optimization parameters on the fracture properties, and can provide a reliable theoretical basis and data support for the application of steel / aluminum dissimilar metal laser welding.

Claims

1. A multi-objective optimization method for steel / aluminum dissimilar metal laser welding, characterized in that: The specific steps include: Step 1, take 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, take the laser power x1, welding speed x2, defocus amount x3 and CeO2 addition amount x4 as the optimization design variables, determine the value range of the optimization design variables, design orthogonal experiments, perform laser welding and obtain 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 CeO2 addition amount, unit is g; Step 2, based on the test results, the laser power x1, welding speed x2, defocus x3 and CeO2 addition x4 are taken as input, and the weld warpage y1 and the maximum tensile force y2 that the weld can withstand are taken as output to construct a RIME-CatBoost model that uses RIME to optimize CatBoost. Step 3, use the MOAVOA algorithm to perform multi-objective optimization on the RIME-CatBoost model 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 evaluation method.

2. The multi-objective optimization method for steel / aluminum dissimilar metal laser welding according to claim 1, characterized in that: In Step 2, the inverse of the maximum tensile force that the weldment can withstand, the warpage deformation of the weldment and the welding cost are used as the objective function of the optimization algorithm. The mathematical model of welding process optimization based on the RIME-CatBoost model is expressed as follows: Where: x1 is the laser power, x2 is the welding speed, x3 is the defocus amount, x4 is the CeO2 addition amount, y1 is the warping deformation of the welded part, y2 is the maximum tensile force that the welded part can withstand, and y3 is the welding cost.

3. The multi-objective optimization method for steel / aluminum dissimilar metal laser welding according to claim 1, characterized in that: In Step 2, the specific steps of using RIME to optimize CatBoost are as follows: ① Take the learning rate and maximum depth of the CatBoost model as the hyperparameters that need to be optimized; ② Set the value range of hyperparameters and the initial parameters of the RIME algorithm; ③ Train the CatBoost model by updating the combination of hyperparameters in each iteration of the RIME algorithm; ④ Calculate the fitness value of the corresponding hyperparameter combination according to the model performance index, and select the hyperparameter combination with the best fitness.

4. The multi-objective optimization method for steel / aluminum dissimilar metal laser welding according to claim 1, characterized in that: In Step 3, the MOAVOA algorithm is used to perform multi-objective optimization on the RIME-CatBoost model, as follows: ① Initialize MOAVOA parameters: parameters include the maximum number of iterations, population size, maximum archive capacity, and the number of objective functions; ② Determine whether the current number of iterations has reached the maximum number of iterations: The algorithm starts from the initial number of iterations until the preset maximum number of iterations is reached. If the current number of iterations is less than the maximum number of iterations, the algorithm continues to iterate, otherwise it outputs the Pareto frontier and ends; ③Fitness calculation: Each vulture represents a set of solutions, and the fitness of each vulture is obtained according to the RIME-CatBoost model and the welding cost calculation formula; ④ Select the best vulture: In each iteration, select the vulture with the best performance as the leader according to the fitness value and store it in the archive; ⑤ Update vulture positions: Update the positions of all vultures based on the leader vulture’s behavior and current strategy; ⑥ Determine the file size: Check whether the current solution set size exceeds the file setting size. If the solution set size is smaller than the file size, continue to iterate. If the solution set size is larger than the file size, delete the solutions in the dense area through the grid mechanism to retain the diversity of solutions. ⑦ Update archive: Update archive according to the criteria of Pareto frontier and retain non-dominated solutions; ⑧Perform the next iteration: increase the number of iterations by 1 and go to step ② to iterate again.

5. The multi-objective optimization method for steel / aluminum dissimilar metal laser welding according to claim 1, characterized in that: In Step 4, when the optimal combination of process parameters is determined in the Pareto frontier of the obtained optimization target by the CRITIC evaluation method, the obtained Pareto frontier is weighted and aggregated to assign weights to each Pareto frontier. The expression of the comprehensive evaluation model is as follows: Where: Z is the comprehensive evaluation value of each experimental design, W j is the weight value of the jth indicator, y ij is the j evaluation index value of the ith solution, y i1 is the warpage of the welded parts in each test design, y i2 is the maximum tensile force that the weldment can withstand in each test design, y i3 is the welding cost in each set of experimental design.

Images