A method for optimizing the quality of press riveting based on genetic algorithm and grey correlation analysis

Abstract

The present application relates to a kind of based on genetic algorithm and grey correlation analysis's press-in quality optimization method.The based on genetic algorithm and grey correlation analysis's press-in quality optimization method, with optimization quality target as with maximum optimization target quality set Nx as optimization target, with parameter set K as definition parameter to establish genetic algorithm model;Parameter set K is obtained to the optimization target quality Nx, and parameter set F has greater influence to the parameter set F that is obtained to interactive test, and grey correlation degree value is obtained to parameter set F by Taguchi test and grey correlation degree analysis, the fitness value of genetic algorithm model is established according to grey correlation degree value, whether the parameter of individual q of genetic algorithm belongs to set F is judged, and according to the value of grey correlation degree to establish mapping function to adjust the mutation rate of genetic algorithm model, with the mutation rate of adjusted individual q is mutated iteration, control genetic algorithm model to parameter set K is repeatedly calculated, until reach maximum iteration, output optimal solution at this time.

Description

Technical Field

[0001] This invention relates to the field of riveting optimization technology, specifically a riveting quality optimization method based on genetic algorithm and grey relational analysis. Background Technology

[0002] Press riveting is a riveting method that specifically refers to the process of using external pressure to force the riveting parts (such as rivets, nuts, or studs) into the body material during the riveting process. By causing plastic deformation of the body material, the parts enter the special pre-made grooves in the riveting screw or nut structure, thus achieving a reliable connection between two or more parts.

[0003] Riveting strength is an important indicator of riveting quality, mainly including push-out force, twisting force, and ultimate installation torque. Optimizing riveting quality can ensure a stronger connection between riveted parts and reduce problems such as loosening and detachment caused by substandard riveting strength.

[0004] Existing technologies for optimizing riveting quality primarily employ the Taguchi method and grey relational analysis. These methods may rely on traditional experimental design, statistical analysis, or empirical rules. However, these approaches do not consider the influence of parameters on the optimization objective, necessitating a comprehensive evaluation of all parameters. This results in high computational cost and low efficiency when solving complex optimization problems. Furthermore, the optimization process often requires trying different parameter combinations one by one in the solution space, lacking specificity for critical influencing parameters and easily leading to local optima.

[0005] To address the aforementioned problems, this invention provides a riveting quality optimization method based on genetic algorithms and grey relational analysis, which can quickly and accurately identify the optimal riveting quality optimization scheme, thereby significantly improving the riveting quality. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a riveting quality optimization method based on genetic algorithms and grey relational analysis. This method has advantages such as the ability to quickly evaluate the optimal parameter set from a large number of parameters to optimize riveting quality. It solves the problem caused by the failure to screen parameters and the direct evaluation of parameters by algorithms when optimizing riveting in traditional technologies.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a riveting quality optimization method based on genetic algorithm and grey relational analysis, comprising the following steps:

[0008] S1. Determine the target quality Nx for riveting and the parameter set K that affects the riveting quality. The parameter set K includes an adjustable design parameter set Ami and a riveting machine control parameter set Bnj. Establish a genetic algorithm model with maximizing the target quality Nx as the optimization objective and the parameter set K as the defined parameter.

[0009] S2. Encode the parameter set K, and encode all parameter combinations in the parameter set K as X, X=(Ami, Bnj), where m and n are parameter types, and i and j are parameter levels;

[0010] S3. Set the population size to P and randomly generate q individuals;

[0011] S4. Conduct interactive experiments on the adjustable design parameter set Ami and the riveting machine control parameter Bnj to obtain the parameter set F that has a significant impact on the optimization target quality set Nx;

[0012] S5. The parameter set F obtained through the interactive experiment is used to design an orthogonal experiment using the Taguchi method to calculate the signal-to-noise ratio, and the signal-to-noise ratio is then normalized and standardized.

[0013] S6. The normalized and standardized signal-to-noise ratio is evaluated using grey relational analysis to assess the grey relational degree of each parameter in parameter set F on the target quality of riveting quality optimization.

[0014] S7. Using the gray relational degree as the fitness value, establish a fitness function, use the fitness function as the evaluation criterion of the genetic algorithm model, and obtain the fitness value of individual q.

[0015] S8. Calculate the selection probability of each individual using the roulette wheel selection method based on the fitness value of individual q. Randomly select multiple individuals q from the population P to form a mating pool based on the selection probability. Randomly select two individuals from the mating pool and generate two new offspring individuals in a single-point crossover manner.

[0016] S9. Set the initial mutation rate to W1, and establish a mapping function based on the parameter set F and the grey relational degree of the parameter set F. According to the mapping function To determine the mutation rate W of the offspring of individual q, we randomly select a parameter of the offspring individual based on the mutation rate W and mutate it. The mutated offspring individuals are then merged into the population P and a new population P1 is generated.

[0017] S10. Using the maximum number of iterations as the termination condition, repeat S7-S9 until the termination condition is met, stop the calculation and generate the final population. Select the individual with the highest fitness from the final population as the optimal solution, and finally output the parameter combination of the optimal individual and the corresponding optimization objective quality Nx value.

[0018] Furthermore, the interactive experiment in S4 includes the following sub-steps:

[0019] S4.1 There are m design parameters A, and each parameter has multiple levels. Control parameters There are n control parameters, each with multiple levels. An orthogonal array is constructed based on the number of factors and levels.

[0020] S4.2. Based on the parameter combination determined by the orthogonal array, set the control parameter B of the riveting machine and the design parameter A of the busbar. Distribute different levels of each parameter to each test and perform the riveting operation. Record the response variable value of each test, i.e., the optimized target quality Nx.

[0021] S4.3 Calculate the mean value of the optimization target quality Nx, and use analysis of variance to evaluate the impact of each parameter and its interaction on the optimization target quality Nx, and establish an analysis of variance model:

[0022]

[0023] in , is the value of the corresponding variable under the combination of the i-th level of the m-th design parameter and the j-th level of the n-th control parameter, which is the optimization target quality. It is the overall mean. and These are the main effects of design parameters and control parameters, respectively. It is an interaction effect. This is the error term.

[0024] Analysis of variance model The output value of p is used to determine whether P satisfies the condition. If so, it means that this parameter combination has a significant impact on the optimization objective quality N. The parameter combinations that satisfy the conditions are used to establish parameter F, where... .

[0025] Furthermore, the Taguchi method in S5 includes the following steps:

[0026] S5.1 Determine the target quality Nx to be optimized, establish an orthogonal array based on the number of parameter types (m+n) in the parameter set F and the number of levels i and j for each parameter, and conduct orthogonal experiments.

[0027] S5.2 Measure and record the data of each quality characteristic Nx under each test condition.

[0028] S5.3 For each quality target Nx, calculate its signal-to-noise ratio according to the formula:

[0029]

[0030] Where n is the number of trials. It is the value of the optimized target mass Nx obtained in the l-th experiment. It is the target value of optimizing the target quality Nx. This refers to the signal-to-noise ratio.

[0031] Furthermore, the signal-to-noise ratio in S5.3 is normalized using the formula:

[0032]

[0033] Where min and max They represent The minimum and maximum values.

[0034] The signal-to-noise ratio in S5.3 is standardized using the formula:

[0035]

[0036] in yes The mean, yes The standard deviation.

[0037] Furthermore, the grey relational analysis in S6 also includes the following sub-steps:

[0038] S6.1 Select X normalized and standardized values. As a reference sequence .

[0039] S6.2. For each parameter combination, use the grey relational analysis formula to calculate its corresponding... Values ​​and reference sequences The correlation coefficient is obtained through the formula:

[0040]

[0041] in It is the k-th element of the reference sequence, and xi(k) is the k-th quality characteristic value under the i-th test condition. It is the resolution coefficient.

[0042] S6.3 Calculate the grey relational degree using the formula:

[0043]

[0044] Where X is the quantity of the target quality to be optimized. It is the grey relational degree between the i-th parameter combination and the optimal value of the optimization objective.

[0045] Furthermore, the output value of grey relational analysis As the fitness value, a fitness function is established. The fitness function is: ;

[0046] Furthermore, the roulette wheel selection formula in S8 is as follows:

[0047]

[0048] in Represents an individual The probability of being selected. This represents the fitness value of an individual. This represents the sum of the fitness values ​​of all individuals in the population, where p is the population size. Indicates the first in the population Individual.

[0049] Furthermore, the formula for establishing the mutation rate W(q) in S9 based on the parameter set F and the grey relational degree of the parameter set F is as follows:

[0050]

[0051] Where w1 is the initial mutation rate.

[0052] Compared with the prior art, the technical solution of this application has the following beneficial effects:

[0053] 1. The riveting quality optimization method based on genetic algorithm and grey relational analysis, through interactive experiments on the original evaluation parameters, enables it to derive the parameter set F that has the greatest impact on riveting quality from among many evaluation parameters, thereby reducing the blindness of the genetic algorithm in the search space, greatly narrowing the search space in the subsequent optimization process, and thus improving the optimization efficiency of the genetic algorithm.

[0054] 2. This riveting quality optimization method based on genetic algorithm and grey relational analysis uses the grey relational degree obtained from the parameter set F through Taguchi method and grey relational analysis as the fitness function and mutation rate adjustment basis of the genetic algorithm. This enables the genetic algorithm to more intelligently adjust the mutation rate and selection strategy of different parameters during the evolution process, better reflect the actual impact of the parameters, and thus improve the optimization effect. This method effectively balances efficiency and accuracy by first screening parameters with significant influence and then conducting in-depth optimization. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the optimization process for riveting quality in this invention.

[0056] Figure 2 This is a flowchart of the interactive experiment of the present invention;

[0057] Figure 3 This is a flowchart of the Taguchi experiment and grey relational analysis of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Please see Figure 1 The riveting quality optimization method based on genetic algorithm and grey relational analysis in this embodiment includes the following steps:

[0060] S1. Determine the target quality Nx for riveting and the parameter set K that affects the riveting quality. The parameter set K includes an adjustable design parameter set Ami and a riveting machine control parameter set Bnj. Establish a genetic algorithm model with maximizing the target quality Nx as the optimization objective and the parameter set K as the defined parameter.

[0061] In the actual evaluation process, workers can use the rivet's push-out force, twisting force, and ultimate installation torque after riveting as the optimization target quality Nx for riveting. Adjustable design parameters Ami, in practical use, include the design parameters of the busbar, such as hole diameter, material, and thickness, as well as the rivet's design parameters, such as diameter, length, and material. The riveting press's control parameters include the pressing force, pressing speed, pressing holding time, and the distance from the hole centerline to the edge. Using these design and control parameters as the parameter set K, and the optimization target quality Nx as the optimization objective, a genetic algorithm model is established. The genetic algorithm calculates the optimal combination from these parameters to achieve the optimal riveting quality. Compared to traditional algorithms, it has a powerful global optimization capability, capable of searching for the globally optimal solution or a near-global optimal solution in a complex parameter space. This means it can not only find local optima but also explore possible solution spaces on a larger scale, thereby increasing the chance of finding the truly optimal solution.

[0062] S2. Encode the parameter set K, encoding all parameter combinations within K as X, where X = (Ami, Bnj), and m and n represent parameter types, while i and j represent parameter levels. m and n include design parameters for the busbar, such as hole diameter, material, and thickness, as well as design parameters for the rivet, such as rivet diameter, length, and material. The control parameters for the riveting machine include the pressing force, pressing speed, pressing holding time, and the distance from the hole centerline to the edge. i and j can represent the levels of different parameters, such as different materials or different lengths.

[0063] S3. Set the population size to P and randomly generate q individuals;

[0064] Please see Figure 2 S4. Conduct interactive experiments on the adjustable design parameter set A and the riveting machine control parameters B to derive the parameter set F that has a significant impact on the optimization target quality set Nx; S4.1. There are m design parameters A, and each parameter has multiple levels. Control parameters There are n control parameters, each with multiple levels. An orthogonal array is constructed based on the number of factors and levels.

[0065] S4.2. Based on the parameter combination determined by the orthogonal array, set the control parameter B and design parameter A of the riveting machine, allocate different levels of each parameter to each experiment and perform the riveting operation, and record the response variable value of each experiment, i.e., the optimized target quality Nx.

[0066] S4.3 Calculate the mean value of the optimization target quality Nx, and use analysis of variance to evaluate the impact of each parameter and its interaction on the optimization target quality Nx, and establish an analysis of variance model:

[0067]

[0068] in , is the value of the corresponding variable under the combination of the i-th level of the m-th design parameter and the j-th level of the n-th control parameter, which is the optimization target quality. It is the overall mean. and These are the main effects of design parameters and control parameters, respectively. It is an interaction effect. This is the error term.

[0069] Analysis of variance model The output value of p is used to determine whether P satisfies the condition. If so, it means that this parameter combination has a significant impact on the optimization objective quality Nx. The parameter combinations that satisfy the conditions are used to establish parameter F, where... .

[0070] By establishing an orthogonal array of the above parameters and obtaining the parameter set F for optimizing the target quality through analysis of variance, we can initially screen out the parameters and interactions that have a significant impact on system performance. This will reduce unnecessary test combinations and lower test costs and time in subsequent Taguchi experiments.

[0071] The results of interactive experiments can guide how to adjust parameter levels in Taguchi experiments to maximize system performance and improve its robustness. For example, if a parameter is found to have a significant impact on system performance and an optimal level exists, then in Taguchi experiments, this parameter can be kept at that level while optimizing other parameters.

[0072] It is worth noting that the riveting operation in the above experiment can be replaced by a simulation model, which can solve the experimental cost problem and increase work efficiency.

[0073] By employing analysis of variance to derive the parameter set F that has a significant impact on the target quality Nx, compared to other methods, it can more comprehensively consider the influence of all parameters and their interactions, thus yielding more accurate conclusions.

[0074] Set the threshold for the p-value in the ANOVA results to [value]. It enables statistical significance assessment. The p-value represents the probability of observing the current data or more extreme data if the null hypothesis (i.e., the parameter has no significant effect on the optimization objective) is true. Setting the p-value threshold to 0.05 means that if the p-value is less than 0.05, we have a 95% confidence level to reject the null hypothesis, that is, we consider the parameter to have a significant effect on the optimization objective. This provides a clear statistical standard for determining which parameters are important.

[0075] Please see Figure 3 S5. The parameter set F obtained from the interactive experiment is used to design an orthogonal experiment using the Taguchi method to calculate the signal-to-noise ratio, and the signal-to-noise ratio is then normalized and standardized.

[0076] The Taguchi method in S5 includes the following steps:

[0077] S5.1 Determine the target quality Nx to be optimized, establish an orthogonal array based on the number of parameter types (m+n) in the parameter set F and the number of levels i and j for each parameter, and conduct orthogonal experiments.

[0078] S5.2 Measure and record the data of each target mass Nx under each test condition.

[0079] S5.3 For each quality target Nx, calculate its signal-to-noise ratio according to the formula:

[0080]

[0081] Where n is the number of trials. It is the value of the optimized target mass Nx obtained in the l-th experiment. It is the target value of optimizing the target quality Nx. This refers to the signal-to-noise ratio.

[0082] The signal-to-noise ratio in S5.3 is normalized using the formula:

[0083]

[0084] Where min and max They represent The minimum and maximum values.

[0085] The signal-to-noise ratio in S5.3 is standardized using the formula:

[0086]

[0087] in yes The mean, yes The standard deviation.

[0088] By employing the Taguchi method to establish orthogonal experimental tables, the number of experiments can be systematically reduced while maintaining a certain ability to evaluate the interactions between parameters. This means that with limited resources, more parameter combinations can be covered more quickly, thereby accelerating the optimization process.

[0089] Signal-to-noise ratio (SNR) is a quantitative metric used to assess the impact of each parameter or combination of parameters on the optimization objective. A high SNR value indicates that the parameter or combination has a significant and stable impact on the optimization objective, while a low SNR value may indicate a smaller or unstable impact.

[0090] In order to facilitate subsequent grey relational analysis of the signal-to-noise ratio (SNR) value, it is necessary to process the SNR value.

[0091] S6. The normalized and standardized signal-to-noise ratio is evaluated using grey relational analysis to assess the grey relational degree of each parameter in parameter set F on the target quality of riveting quality optimization.

[0092] The grey relational analysis in S6 also includes the following sub-steps:

[0093] S6.1 Select X normalized and standardized values. As a reference sequence .

[0094] S6.2. For each parameter combination, use the grey relational analysis formula to calculate its corresponding... Values ​​and reference sequences The correlation coefficient is obtained through the formula:

[0095]

[0096] in It is the k-th element of the reference sequence, and xi(k) is the k-th quality characteristic value under the i-th test condition. It is the resolution coefficient.

[0097] S6.3 Calculate the grey relational degree using the formula:

[0098]

[0099] Where X is the quantity of the target quality to be optimized. It represents the grey relational degree between the i-th parameter combination and the optimal value of the optimization objective. Where the grey relational degree... The range of values ​​is

[0100] Normalization and standardization of the signal-to-noise ratio allow results from different parameters or experimental conditions to be compared on the same scale. This helps identify which parameters are critical, which have negligible influence, and their relative importance.

[0101] For this solution, in multi-objective optimization problems such as simultaneously optimizing the push-out force, twisting force, and ultimate installation torque, the normalized and standardized signal-to-noise ratio can be used as a weight or priority indicator for each objective function. This helps to find a balance among multiple optimization objectives, thereby obtaining a riveting solution with better overall performance.

[0102] Output value of grey relational degree As the fitness value, a fitness function is established. The fitness function is: ;

[0103] Using the grey relational degree as the fitness function of a genetic algorithm can effectively optimize the algorithm. Genetic algorithms are search algorithms based on natural selection and genetic mechanisms, possessing global search capabilities. Using grey relational analysis as the fitness function can further guide the algorithm to search more broadly in the solution space, because grey relational analysis considers the degree of association between multiple factors and the target, helping to discover potential global optima. Simultaneously, it can reduce the impact of these parameters on algorithm performance to some extent, as grey relational analysis itself has a certain degree of anti-interference capability and robustness, and can more stably evaluate the quality of individuals.

[0104] S7. Using the gray relational degree as the fitness value, establish a fitness function, use the fitness function as the evaluation criterion of the genetic algorithm model, and obtain the fitness value of individual q.

[0105] S8. Calculate the selection probability of each individual using the roulette wheel selection method based on the fitness value of individual q. Randomly select multiple individuals q from the population P to form a mating pool based on the selection probability. Randomly select two individuals from the mating pool and generate two new offspring individuals in a single-point crossover manner.

[0106] The roulette wheel selection formula in S8 is as follows:

[0107]

[0108] in Represents an individual The probability of being selected. This represents the fitness value of an individual. This represents the sum of the fitness values ​​of all individuals in the population, where p is the population size. Indicates the first in the population Individual.

[0109] The roulette wheel selection method directly evaluates and selects individuals based on their fitness, which in this scheme is represented by grey relational analysis. Since grey relational analysis has been normalized or standardized, data consistency and comparability are ensured, making fitness evaluation more accurate and efficient.

[0110] Furthermore, although the roulette wheel selection method itself is random, its impact on algorithm performance can be reduced through certain strategies. For example, in this scheme, an elite retention strategy can be incorporated, where the best individual from each generation is directly retained to the next generation. Alternatively, the roulette wheel selection method can be appropriately improved, such as by introducing nonlinear transformations, to better adapt to the needs of specific problems.

[0111] S9. Set the initial mutation rate to W1, and establish a mapping function based on the parameter set F and the grey relational degree of the parameter set F. According to the mapping function To determine the mutation rate W of the offspring of individual q, we randomly select a parameter of the offspring individual based on the mutation rate W and mutate it. The mutated offspring individuals are then merged into the population P and a new population P1 is generated.

[0112] The formula for establishing the mutation rate W(F) in S9 based on the parameter set F and the grey relational degree of the parameter set F is as follows:

[0113]

[0114] Where w1 is the initial mutation rate and W is the mutation rate during the genetic algorithm calculation.

[0115] In practical applications, the initial mutation rate W1 can be set between 0.01 and 0.05. A lower initial mutation rate helps maintain population stability in the early stages of the algorithm. At the beginning of the genetic algorithm, individuals in the population are often generated randomly or based on some heuristic method, and their quality may vary considerably. A lower mutation rate can reduce the destruction of superior individuals in the early stages of the algorithm, thus helping the algorithm converge to a better solution more quickly. The specific value of the initial mutation rate W1 can be determined by the user based on the actual application.

[0116] By incorporating the correlation analysis between the parameters of individual q and the set F into the adjustment of the mutation rate, the algorithm can more intelligently select the direction and intensity of mutation. When the correlation between the individual parameters and the target is high, the mutation rate can be appropriately reduced to maintain the stability of the solution; while when the correlation is low, the mutation rate is increased to explore a wider solution space. This strategy helps the algorithm avoid getting trapped in local optima while maintaining search efficiency.

[0117] S10. Using the maximum number of iterations as the termination condition, repeat S7-S9 until the termination condition is met, stop the calculation and generate the final population. Select the individual with the highest fitness from the final population as the optimal solution, and finally output the parameter combination of the optimal individual and the corresponding optimization objective quality Nx value.

[0118] In practical evaluation, through repeated computation, genetic algorithms can more comprehensively search the solution space, increasing the likelihood of finding the global optimum. Crossover and mutation operations in genetic algorithms help escape local optima and explore a wider solution space. Multiple repeated computations reduce the uncertainty of single computation results, improving the algorithm's stability and reliability. By statistically analyzing the results of multiple computations, the algorithm's performance and effectiveness can be evaluated. After reaching the maximum number of iterations, the algorithm can output the optimal parameter combination after multiple iterations, which typically exhibits better performance.

[0119] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing the quality of a press-in rivet based on a genetic algorithm and grey relational analysis, characterized in that: It comprises the following steps: S1, determine the optimal target quality Nx of the rivet and the parameter set K affecting the rivet quality, the parameter set K includes the adjustable design parameter set Ami and the rivet press control parameter set Bnj; the genetic algorithm model is established with the parameter set K as the definition parameter and the maximum optimization target quality set Nx as the optimization target; S2, encode the parameter set K, encode all parameter combinations in the parameter set K as X, X= (Ami, Bnj), where m and n are parameter types, and i and j are parameter levels; S3, set the population size to P and randomly generate q individuals; S4, conduct an interactive test on the adjustable design parameter set Ami and the rivet press control parameter Bnj to obtain the parameter set F that has a significant impact on the optimization target quality set Nx; S5, the parameter set F obtained through the interactive test is designed by orthogonal test through the Taguchi method, the signal-to-noise ratio is calculated, and the signal-to-noise ratio is normalized and standardized; S6, the signal-to-noise ratio after normalization and standardization is evaluated by the grey correlation analysis method to evaluate the grey correlation degree of each parameter in the parameter set F to the rivet quality optimization target quality; The grey correlation analysis in S6 further comprises the following sub-steps: S6.1, select X normalized and standardized as a reference sequence ; S6.2, the grey correlation analysis formula is used to calculate the corresponding correlation coefficient of each parameter combination with the reference sequence , by the formula: ​ ； wherein is the kth element of the reference sequence, xi(k) is the kth mass property value under the ith trial condition, is the resolution coefficient; S6.3 Calculate the grey correlation degree by the formula: ； wherein X is the number of optimization target qualities, is the grey correlation degree between the ith parameter combination and the optimal value of the optimization target; S7, take the value of the grey correlation degree as the fitness value, establish the fitness function, take the fitness function as the evaluation standard of the genetic algorithm model, and obtain the fitness value of the individual q; S8. According to the fitness value of the individual q, the selection probability of each individual is calculated by the roulette selection method, and a plurality of individuals q are randomly selected from the population P based on the selection probability to form a mating pool, and two new offspring individuals are generated by randomly selecting two individuals from the mating pool and in the form of single-point crossover; S9, set the initial mutation rate as W1, and establish a mapping function according to the parameter set F and the grey correlation degree of the parameter set F , judging the individual q according to the mapping function , determining the mutation rate W of the offspring individual of the individual q, selecting a parameter of the offspring individual to be mutated at the mutation rate W, merging the mutated offspring individual into the population P, and generating a new population P1; The formula of the mutation rate W(q) is established according to the parameter set F and the grey correlation degree of the parameter set F in S9 as follows: ； Wherein w1 is the initial mutation rate; S10, take the maximum iteration number as the termination condition, repeat S7-S9 until the termination condition is met, stop calculation and generate the final population, select the individual with the highest fitness value from the final population as the optimal solution, and finally output the parameter combination of the optimal individual and the corresponding optimization target quality Nx value.

2. The rivet quality optimization method based on genetic algorithm and grey correlation analysis according to claim 1, characterized in that: The interactive test in S4 comprises the following sub-steps: S4.1 There are m design parameters A, and each parameter has multiple levels. Control parameters There are n control parameters, each with multiple levels. An orthogonal array is constructed based on the number of factors and levels. S4.2, according to the parameter combination determined by the orthogonal table, set the control parameters B of the rivet press and the design parameters A of the busbar, allocate different levels of each parameter to each test and perform rivet operation, and record the response variable value of each test, i.e. the optimization target quality Nx; S4.3, calculate the average value of the optimization target quality Nx, use variance analysis to evaluate the influence of each parameter and its interaction on the optimization target quality Nx, and establish a variance analysis model: ； wherein is the corresponding response variable value under the combination of the ith level of the mth design parameter and the jth level of the nth control parameter, i.e. the optimization objective quality, is the overall mean, and are the main effects of the design parameters and the control parameters, respectively, is the interaction effect, is the error term; Opposite variance analysis model The p value in the output value is judged to determine whether P satisfies If yes, it indicates that this parameter combination has a significant influence on the optimization target quality N, and the parameter combination satisfying the condition is established as the parameter F, wherein .

3. The method of claim 1, wherein the method is characterized by: The Taguchi method in S5 comprises the following steps: S5.1, determine the target quality Nx to be optimized, establish an orthogonal table according to the number of parameter types (m+n) in the parameter set F and the number of levels i and j of each parameter, and perform orthogonal test; S5.2, measure and record the data of each quality characteristic Nx under each test condition; S5.3, for each quality target Nx, calculate its signal-to-noise ratio respectively, according to the formula: ； where n is the number of trials, is the value of the optimized target mass Nx measured in the 1th trial, is the target value of the optimized target mass Nx, is the signal-to-noise ratio.

4. The method of claim 3, wherein the method is characterized by: The signal-to-noise ratio in S5.3 is normalized by the formula: ； where min and max represent the minimum and maximum values of respectively; The signal-to-noise ratio in S5.3 is standardized by the formula: ； wherein is the mean of is the standard deviation of 5. The method of claim 1, wherein the method is characterized by: The output value of the grey correlation degree as a fitness value, and a fitness function is established where the fitness function is: .

6. The method of claim 1, wherein the method is characterized by: The roulette selection formula in S8 is: ； wherein represents an individual the probability of being selected, represents the fitness value of an individual, represents the sum of the fitness values of all individuals in the population, wherein p is the population size, represents the fitness value of the th individual in the population.

Images