Assembly process parameter optimization method and system based on improved moss growth algorithm

Abstract

The invention discloses an improved moss growth algorithm-based assembly process parameter optimization method. The method comprises the following steps of (1) selecting a plurality of assembly precision indexes; (2) analyzing and evaluating the influence degree of each influence factor on the assembly precision index, selecting the influence factor with relatively large influence, inputting the influence factor into the improved support vector machine precision prediction model, and outputting an assembly precision prediction value; formulating a process optimization objective function by using the assembly precision predicted value; (3) selecting assembly process parameters needing to be optimized, formulating assembly process parameter optimization constraint conditions, and constructing a process parameter optimization model in combination with the process optimization objective function; (4) solving optimal assembly process parameters by the process parameter optimization model by using an improved moss growth algorithm; and (5) outputting the optimal assembly process parameters to form an assembly process adjustment scheme. According to the method, the convergence speed and optimization capacity of an existing algorithm are improved, the assembly precision and quality are improved, and the repair cost caused by trial and error and correction is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical assembly process optimization, and in particular to an assembly process parameter optimization method and system based on an improved moss growth algorithm, which are used to improve the assembly accuracy of complex mechanical products. Background Art

[0002] Assembly is a critical production step in which components are assembled into finished products. Assembly process parameters significantly impact assembly precision and quality. Optimizing assembly process parameters is a crucial means of ensuring product quality and improving assembly efficiency. Reasonable process parameters can significantly reduce assembly errors and improve product consistency. Traditional process parameter optimization relies primarily on empirical trial-and-error methods or simulation optimization based on physical models, which suffer from low efficiency, high cost, and poor adaptability.

[0003] The existing technology includes "A method for optimising assembly processes based on maximum entropy theory" (CN107609227B), which evaluates the uniformity of stress distribution by calculating the entropy value of the assembly system and performs multi-objective optimization in combination with assembly precision requirements. This method quantifies the long-term stability of the assembly using entropy theory, addressing the shortcomings of traditional optimization methods in terms of stress distribution uniformity. However, this method relies on third-party mechanical simulation software and is mainly targeted at single-piece, small-batch precision assembly, without performing targeted optimization based on the characteristics of the assembly process. In addition, this method does not combine intelligent algorithms to achieve rapid parameter inversion, making it difficult to meet the efficient optimization requirements of large-scale composite assembly and resulting in poor adaptability.

[0004] The moss growth optimization algorithm is a novel metaheuristic algorithm that simulates the reproductive characteristics and cryptobiosis of mosses. It achieves global optimization through a spore propagation mechanism, combines dual reproductive search with local exploitation, and dynamically adjusts the search direction using a wind direction determination mechanism to avoid local optima. Its unique cryptobiosis mechanism enhances global search capabilities by retaining historical information and recovering optimal individuals, achieving a good balance between exploration and exploitation. However, the standard moss growth algorithm still has some drawbacks, such as using a completely random population initialization method, resulting in low initial solution quality; linearly decreasing search step size, making it difficult to adapt to the search requirements of different optimization stages; and lacking an effective mutation mechanism, which can lead to premature loss of population diversity.

[0005] The existing technology has the following problems:

[0006] (1) The method is highly dependent and has poor domain adaptability. It relies on external simulation tools or specific scenarios and lacks deep embedding of domain characteristics, which limits the versatility and large-scale application potential of the method.

[0007] (2) Existing algorithms have inherent defects in global search capabilities, convergence efficiency, and computing resource consumption, making it difficult to balance the contradiction between exploration and development, resulting in limited optimization performance.

[0008] (3) Algorithm design is rigid, population initialization randomness, static search strategy and mutation mechanism are insufficient, dynamic adjustment and diversity maintenance mechanism are missing, and the ability to dynamically adjust parameters according to the optimization process is lacking, which is difficult to meet the multi-stage requirements of complex problems. SUMMARY

[0009] The purpose of the application is to provide an assembly process parameter optimization method based on improved moss growth algorithm, to expand the assembly process optimization method, and to reduce the dependence on third-party tools and scenarios. This method solves the problems of slow convergence speed and easy to fall into local optimum in the prior art through algorithm innovation, improves the convergence speed and optimization ability of the algorithm. This method not only improves the assembly precision and quality, but also reduces the rework cost caused by trial and error and correction.

[0010] Technical scheme: The assembly process parameter optimization method based on improved moss growth algorithm comprises the following steps:

[0011] (1) Selecting several assembly precision indexes;

[0012] (2) Analyzing the influencing factors of the assembly precision indexes, evaluating the influence degree of each influencing factor on the assembly precision indexes, selecting the influencing factors with greater influence to input the constructed improved support vector machine precision prediction model, and outputting the assembly precision prediction value; using the assembly precision prediction value to formulate a process optimization objective function;

[0013] (3) Selecting the assembly process parameters to be optimized, formulating the assembly process parameter optimization constraint conditions, combining the process optimization objective function, and constructing a process parameter optimization model;

[0014] (4) The process parameter optimization model uses the improved moss growth algorithm to solve the optimal assembly process parameters;

[0015] (5) Outputting the optimal assembly process parameters to form an assembly process adjustment scheme.

[0016] Further, the assembly precision indexes include assembly deformation, change of key component size, and change of assembly gap.

[0017] Further, the construction and training of the improved support vector machine precision prediction model comprises: firstly, normalizing the multi-dimensional error data of the assembly process; secondly, setting the initial search range of the SVM hyperparameters based on the maximum standard deviation of the input data; then, using a coarse and fine granularity combined grid search strategy to optimize the hyperparameters, first performing coarse granularity search at a larger step to locate the potential optimal interval, then narrowing the range based on the coarse search result to perform fine granularity refinement search, and finally determining the optimal hyperparameter combination; finally, training the SVM model using the optimal parameters, and verifying the prediction performance through cross-validation and independent test set.

[0018] Furthermore, the process optimization objective function is:

[0019]

[0020] Where x is the process parameter vector, ε is the assembly error, n is the number of assembly accuracy indicators, and f i (x) is the assembly accuracy prediction value output by the improved support vector machine accuracy prediction model, [y min_i ,y max_i ] is the accuracy range required by the process, i=1,2,……n.

[0021] Furthermore, the influencing factors include assembly matching dimensions, assembly process parameters, and environmental data.

[0022] Furthermore, the influencing factors of the assembly accuracy index are analyzed and the influence degree of each influencing factor on the assembly accuracy index is evaluated using a grey relational analysis or a principal component analysis method.

[0023] Furthermore, the assembly process parameter optimization constraint condition refers to setting an adjustment range of the assembly process parameter to be optimized in combination with the current production status and production conditions.

[0024] Furthermore, the improved moss growth algorithm is used in step (4) to solve the optimal assembly process parameters, and the specific steps are as follows:

[0025] (4.1) Initialize algorithm related parameters;

[0026] (4.2) Improve the random population initialization to use Latin hypercube sampling to initialize the position of the moss population. Divide the interval of each dimension into several equal parts, and randomly select points in each part to ensure uniform distribution in each dimension.

[0027] (4.3) After obtaining the initial moss population, calculate the fitness of the initial population in sequence, obtain the optimal individual, calculate the average distance between the individual and the optimal individual, and smooth the path of the individual approaching the optimal individual;

[0028] (4.4) The dynamic Levy flight strategy is introduced to dynamically adjust the search step size to update the position. The formula is as follows:

[0029]

[0030] Where: i is the index of the moss individual currently being updated, is the population position determined by the original moss growing algorithm, The population location determined by the present invention, levy step(FEs) is the dynamic Levy flight step length introduced in the present invention, and R is a uniform random number independent of each dimension;

[0031] The dynamic Levy flight step length introduced by the present invention is shown in the following formula:

[0032]

[0033] Where: FEs is the current number of iterations, levy step (FEs) is its flight step size, MaxFEs is the maximum number of iterations; β is the step size distribution parameter, u and v are random variables that obey a specific distribution;

[0034] (4.5) The improved Gaussian mutation is introduced to perform population individual mutation. The Gaussian mutation formula is as follows:

[0035]

[0036] Where: Δx i is the variation of the i-th variable, N(0,σ 2 ) represents a model with mean 0 and variance σ 2 is a normally distributed random variable; γ is the variation intensity adjustment coefficient, which is used to adjust the variation range of σ and thus control the variation range of Gaussian variation;

[0037] (4.6) When a better individual is found in the population, the better individual is used to replace the current best individual to achieve population evolution;

[0038] (4.7) When the maximum number of recorded generations of cryptobiotic search is reached, the moss cryptobiotic mechanism is activated and the current individual is replaced by the best individual recorded;

[0039] (4.8) Record the current number of iterations and determine whether the maximum number of iterations has been reached. When the number of iterations has not reached the maximum number of iterations, the algorithm repeats steps (4.2) to (4.7) until the maximum number of iterations is reached, and outputs the optimized assembly process parameters.

[0040] The assembly process parameter optimization system based on the improved moss growth algorithm includes an indicator input module, a data analysis module, a process parameter optimization model module, and a solution output module;

[0041] The index input module selects several assembly accuracy indicators;

[0042] The data analysis module analyzes the factors affecting the assembly accuracy index, evaluates the degree of influence of each factor on the assembly accuracy index, selects the factors with greater influence, inputs them into the constructed improved support vector machine accuracy prediction model, and outputs the assembly accuracy prediction value; and uses the assembly accuracy prediction value to formulate the process optimization objective function;

[0043] The process parameter optimization model module selects the assembly process parameters to be optimized, formulates the assembly process parameter optimization constraints, and constructs the process parameter optimization model in combination with the process optimization objective function; the process parameter optimization model uses the improved moss growth algorithm to solve the optimal assembly process parameters;

[0044] The solution output module outputs the optimal assembly process parameters to form an assembly process adjustment solution.

[0045] Beneficial effects: 1. Based on the improved support vector machine model, accurate prediction of assembly accuracy under the coupling of multi-source errors is achieved, overcoming the defect that traditional mathematical modeling is difficult to handle nonlinear relationships.

[0046] 2. Innovative methods such as Latin hypercube sampling initialization, dynamic Lévy flight, and dynamic Gaussian mutation address key issues of the original moss growth algorithm, including low initial population quality and insufficient population diversity. The improved algorithm excels in solution space coverage, resistance to premature convergence, and cross-generational optimization capabilities, making it particularly suitable for process parameter optimization scenarios with high-dimensional and complex constraints.

[0047] 3. A constraint system based on the principles of controllability, feasibility, and conflict balance ensures that optimization results closely match actual production conditions. Through an end-to-end intelligent optimization process, a closed-loop control system from measured assembly parameters to assembly accuracy is achieved, significantly reducing assembly adjustment costs and improving the consistency and stability of complex product assembly quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Optimize the overall flow chart for assembly process parameters;

[0049] Figure 2 Iterative comparison chart of different optimization algorithms. DETAILED DESCRIPTION

[0050] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0051] Example 1:

[0052] like Figure 1 As shown, the assembly process parameter optimization method based on the improved moss growth algorithm includes the following steps:

[0053] (1) Select assembly accuracy indicators;

[0054] The assembly accuracy index described in step (1) refers to an index that can objectively reflect or evaluate assembly accuracy and assembly quality, including but not limited to assembly deformation, changes in key dimensions, and changes in assembly clearance. When selecting an accuracy index, priority is given to the index required to be achieved in the assembly accuracy requirements. In addition, more than one accuracy index can be selected.

[0055] (2) Analyze the factors affecting assembly accuracy, evaluate the degree of influence of each factor on assembly accuracy, select the factors with greater influence and input them into the constructed improved support vector machine accuracy prediction model, and output the assembly accuracy prediction value; use the assembly accuracy prediction value to formulate the process optimization objective function;

[0056] Based on the specific assembly process and scenario, analyze the factors affecting assembly accuracy, including but not limited to assembly dimensions, assembly process parameters, and environmental data. Use methods such as grey correlation analysis and principal component analysis to assess the impact of each factor on assembly accuracy indicators and select the factors with the greatest impact.

[0057] In view of the fact that assembly accuracy is affected by the complex coupling of multi-dimensional data, traditional mathematical modeling methods and empirical formulas are difficult to achieve high-precision predictive analysis. In order to solve this technical problem, the present invention adopts an intelligent modeling method based on machine learning theory, and establishes an accuracy prediction model that integrates multi-dimensional error factors by constructing an end-to-end improved support vector machine model. Using the improved support vector machine model, the distribution interval of the hyperparameter is determined based on the maximum standard deviation of the input sample, and based on the two-stage grid search algorithm, the model hyperparameter optimization is completed, the optimal hyperparameter is assigned to the prediction model, and the prediction model training is completed on the established accuracy prediction data set. The trained model is used to construct the assembly process parameter optimization objective function. When there are multiple accuracy indicators, the weight coefficient method can be used to formulate a unified optimization objective function. This technical solution can effectively capture the nonlinear correlation between multi-source errors and achieve accurate mapping from input data to accuracy prediction.

[0058] The present invention adopts an improved support vector machine (SVM) model for assembly accuracy prediction, and realizes the optimization construction and training of the model through the following steps: first, the multidimensional error data of the assembly process is normalized and preprocessed to eliminate the dimensional effect; second, the initial search range of the SVM hyperparameters (including the penalty coefficient C and the kernel function parameter γ) is adaptively set based on the maximum standard deviation of the input data to ensure the rationality of the parameter combination; then a grid search strategy combining coarse and fine granularity is adopted to optimize the hyperparameters, firstly a coarse-grained search is performed at a large step size to locate the potential optimal interval, and then a fine-grained refined search is performed based on the coarse search results to finally determine the optimal hyperparameter combination; finally, the SVM model is trained using the optimal parameters, and its prediction performance is verified by cross-validation and an independent test set. This improved method significantly improves the efficiency of hyperparameter optimization and model prediction accuracy through a data-driven parameter initialization strategy and a staged optimization algorithm, effectively solves the problem of high computational cost of traditional methods, and realizes accurate prediction of assembly accuracy under high-dimensional coupling error conditions.

[0059] Based on the constructed improved support vector machine accuracy prediction model, the present invention realizes the quantitative evaluation of the influence of process parameters on assembly accuracy by establishing a nonlinear mapping relationship between assembly process parameters and prediction accuracy. On this basis, with the improvement of assembly accuracy as the goal, based on the constructed improved support vector machine accuracy prediction model, an objective function is established as shown in Formula 1:

[0060]

[0061] Where x is the process parameter vector, ε is the assembly error, n is the number of assembly accuracy indicators, and f i (x) is the assembly accuracy value output by the prediction model, [y min_i ,y max_i ] is the required accuracy range for the process. This objective function directly reflects the nonlinear mapping relationship between process parameters and assembly accuracy. By optimizing the moss growth algorithm to find the minimum value of this function, the process parameter combination that achieves the optimal assembly accuracy can be obtained. During the construction process, the support vector machine model's ability to accurately model the coupling of multidimensional process parameters was emphasized to ensure that the objective function accurately represents the actual impact of process parameter changes on assembly accuracy.

[0062] (3) Select the assembly process parameters that need to be optimized, formulate the assembly process parameter optimization constraints, combine the process optimization objective function, and build a process parameter optimization model;

[0063] The optimization object described in step (3) refers to the most suitable assembly process parameters for adjustment selected after comprehensive consideration of multiple factors such as parameter controllability, assembly feasibility, practical feasibility, ease of adjustment, and conflict balance. In combination with the current production status and production conditions, the adjustment range of the assembly process parameters to be optimized is set. The objective function constructed in step (2) is combined with the constraints set in this step to complete the construction of the assembly process parameter optimization model. The present invention optimizes assembly process parameters based on assembly accuracy prediction.

[0064] In the accuracy prediction stage, the present invention comprehensively considers factors such as size and environment that affect assembly accuracy, and obtains the assembly accuracy prediction value based on the improved support vector machine model; in the parameter optimization stage, the present invention takes into account the non-variability of size and environmental factors, and focuses on optimizing assembly process parameters.

[0065] The main factors and principles to be considered when selecting the optimization object and formulating the process parameter constraints include: the controllability principle of assembly parameters, the assembly feasibility principle, the realistic feasibility principle, the feasibility principle, the conflict balance principle, etc.

[0066] 1. Controllability Principle. During assembly process parameter optimization, the controllability principle requires that the selected optimization object must be a process variable that can be actively adjusted and stably controlled. Controllability ensures that the optimized parameters can be accurately executed and maintained consistently in actual production.

[0067] 2. Assembly feasibility principle. The assembly feasibility principle emphasizes that optimized parameters must meet the physical and functional requirements of the assembly process. This principle requires that parameter optimization must not undermine the basic functionality of the assembly and must also consider practical constraints such as production line layout and tool accessibility. If the optimization results significantly increase the difficulty of assembly, feasibility must be reassessed.

[0068] 3. Principle of Realistic Feasibility. The principle of realistic feasibility requires that the optimization goal must match existing production conditions. This means fully considering the objective limitations imposed by current assembly conditions on the adjustment range of process parameters. The core of this principle is to avoid solutions that are "theoretically optimal but impractical in practice." This principle requires integration with production line research and data verification.

[0069] 4. Feasibility Principle. The selected assembly process parameters to be optimized should be easy to modify and adjust. The core of this principle is to minimize the impact on the assembly line or assembly plan while improving assembly accuracy.

[0070] 5. Conflict balance principle: When multiple optimization objectives conflict with each other, a compromise or priority division is required to achieve the global optimum.

[0071] Taking the above principles into comprehensive consideration, the optimization object and its variation range are determined to form the constraint conditions of the process parameter optimization model.

[0072] (4) The process parameter optimization model uses the improved moss growth algorithm to solve the optimal assembly process parameters;

[0073] Step (4) uses the improved moss growth algorithm to obtain the optimal assembly process parameters, and the specific steps are as follows:

[0074] (4.1) Initialize algorithm related parameters;

[0075] During the algorithm initialization phase, algorithm initialization parameters are set, such as the initial population size, maximum number of iterations, maximum population update generations, wind intensity coefficient, spore diffusion boundary parameter, moss fragment diffusion boundary parameter, and maximum number of recorded generations for cryptobiotic search. In this embodiment, by analyzing the impact of these different parameters on the algorithm, it was found that the algorithm had better convergence and optimization capabilities when the initial population size was 40, the maximum population update generations was 100, and the wind intensity coefficient, spore diffusion boundary parameter, moss fragment diffusion boundary parameter, and maximum number of recorded generations for cryptobiotic search were 5, 0.01, 0.8, and 10, respectively.

[0076] (4.2) Population initialization;

[0077] The original moss growth algorithm uses random initialization to initialize the population. The uncontrollable randomness of this initialization method causes the initial population to exhibit non-uniform distribution characteristics in the solution space, which may cause a large number of individuals to concentrate in non-optimal solution areas, significantly affecting the algorithm's convergence efficiency and increasing the risk of local optimality. In addition, the incomplete coverage of the solution space causes the coexistence of search blind spots and redundant searches, seriously weakening the algorithm's global exploration capabilities. Moreover, the population homogeneity caused by the random mechanism further weakens the population diversity and exacerbates the algorithm's tendency to converge locally. When the initial population is significantly away from the global optimal solution area, the algorithm requires additional iterations to enter the effective search phase. This characteristic is particularly disadvantageous in real-time application scenarios. These technical shortcomings jointly restrict the actual performance of the algorithm in complex optimization problems.

[0078] To address these shortcomings, the present invention improves the algorithm's initialization method, using Latin hypercube sampling to initialize the moss population's positions. Each dimension's interval is divided into several equal parts, and points are randomly selected within each part to ensure uniform distribution within each dimension, allowing the algorithm to more comprehensively explore the entire search space. The specific process is as follows:

[0079] First, the interval length is divided. Based on the upper and lower limits of each dimension and the number of populations, the search range of each dimension is evenly divided into multiple small intervals. Secondly, an offset α is generated in [0,1]. i,j , ensure that the individuals in the population after position mapping fall within different intervals; finally, use formula (2) to perform position mapping and map them to the upper and lower limit intervals of each dimension.

[0080] Position i,j =lb i +[α i,j ×(ub i -lb i )] (2)

[0081] Where: Position i,j is the position of the jth particle in the i-th dimension, lb i is the lower limit of the i-th dimension, ub i is the upper limit of the i-th dimension.

[0082] Latin hypercube sampling is used to replace the random initialization of the traditional algorithm to ensure uniform distribution in each dimension, thereby improving the diversity of the initial population. Compared with random population initialization, it greatly improves the quality of the initial solution, thereby improving the optimization effect and efficiency of the optimization algorithm.

[0083] (4.3) Wind direction determination;

[0084] After obtaining the initial moss population, the fitness of the initial population is calculated sequentially to obtain the optimal individual. To make the population evolve towards the optimal individual, the moss growth algorithm performs wind direction measurement. This wind direction measurement determines the direction of population evolution. The algorithm performs wind direction measurement according to formula (3).

[0085]

[0086] Where: k is the current iteration number, is the wind direction measurement value, M best is the optimal individual, M k is the kth individual.

[0087] Wind direction determination calculates the average distance between individuals and the optimal individual, smoothes the path of individuals approaching the optimal individual, and improves the quality of the solution and the optimization ability of the algorithm.

[0088] (4.4) Population individual update;

[0089] The original moss growth algorithm sets two moss individual update modes: sexual reproduction and asexual reproduction, and provides search step formulas for each. However, the search step of the original algorithm gradually decreases, lacks jumps, and is prone to falling into local optimal solutions. To this end, the present invention introduces a dynamic Levy flight strategy based on the position update method of the original moss growth algorithm, dynamically adjusts the algorithm search step, and helps to jump out of the local optimal solution. The position update formula of the present invention is shown in formula (4):

[0090]

[0091] Where: i is the index of the moss individual currently being updated, is the population position determined by the original moss growing algorithm, The population location determined by the present invention, levy step (FEs) is the dynamic Lévy flight step length introduced in this paper, and R is an independent uniform random number in each dimension. By combining the dynamic Lévy step length with a symmetric random direction (R-0.5), the algorithm retains the global exploration capability of Lévy flights while avoiding the directional bias that may be caused by heavy-tailed distributions, thereby balancing the contradiction between exploration and exploitation.

[0092] In order to balance the global search capability and convergence speed of the algorithm, the present invention improves the Levy flight strategy so that the flight step size is dynamically adjusted according to the progress of the algorithm, as shown in formula (5).

[0093]

[0094] Where: FEs is the current number of iterations, levy step (FEs) is the flight step length, MaxFEs is the maximum number of iterations, It is the progress factor, which indicates the progress of the algorithm. The flight step size of Levy flight is dynamically adjusted according to this progress factor.

[0095] In the early stages of the algorithm, the flight step size is large, which facilitates a wide-scale global search; in the later stages, the flight step size gradually decreases, thereby accelerating the convergence of the algorithm. This improvement not only ensures a strong global search capability, but also effectively improves the convergence speed of the algorithm. This improvement not only retains the diversity advantage of the original sexual / asexual reproduction, but also significantly improves the algorithm's robustness, convergence accuracy, and ability to resist local optimality in complex optimization problems through the randomness and dynamic adaptability of Lévy flights. In this embodiment, multiple experiments showed that the algorithm performs better when the step size distribution parameter β = 1.5.

[0096] (4.5) Individual variation in a population;

[0097] The original moss growth algorithm lacks population mutation operations, which may lead to premature convergence of the algorithm, decreased population diversity, and falling into local optimality. To overcome this shortcoming, Gaussian mutation is introduced. The original Gaussian mutation has a fixed mutation amplitude, which is difficult to adapt to the needs of different search stages of the algorithm. To control the mutation amplitude, the present invention uses a progress factor to control the mutation amplitude. The Gaussian mutation formula of the present invention is shown in formula (6):

[0098]

[0099] Where: Δx iis the variation of the i-th variable, γ is the variation intensity adjustment coefficient, which is used to adjust the variation range of σ and thus control the variation range of Gaussian variation, FEs is the current number of iterations, MaxFEs is the maximum number of iterations, is the progress factor, which indicates the progress of the algorithm.

[0100] In the early stages, the amplitude of the mutation is large, enabling the algorithm to explore a wide range of solutions and avoid being restricted by local solutions. In the later stages, the amplitude of the mutation decreases, allowing the algorithm to perform more detailed local optimization. This adjustment mechanism allows the algorithm to perform different types of searches at different stages, enabling both extensive exploration and fine-grained optimization, improving the algorithm's overall performance. In this embodiment, the Gaussian mutation intensity adjustment coefficient γ is set to 0.1.

[0101] (4.6) Optimal individual update;

[0102] When a better individual is found in the population, the better individual is used to replace the current best individual to achieve the evolution of the population.

[0103] (4.7) Activation of cryptobiotic mechanisms;

[0104] When the maximum number of recorded generations of cryptogenetic search is reached, the algorithm activates the cryptogenetic mechanism, replacing the current individual with the best recorded individual. This mechanism mimics the cryptogenetic phenomenon of mosses, enabling rapid recovery through intergenerational gene transfer and resisting premature convergence through an "evolutionary memory bank." This helps preserve population diversity, balance exploration and exploitation, and improve the quality of the algorithm's solutions.

[0105] (4.8) Iterative optimization;

[0106] Record the current number of iterations and determine whether the maximum number of iterations has been reached. When the number of iterations has not reached the maximum number of iterations, the algorithm repeats steps (4.2) to (4.7) until the maximum number of iterations is reached, and outputs the optimized assembly process parameters.

[0107] (5) Output optimal assembly process parameters;

[0108] The optimized assembly process parameters are output to form an assembly process adjustment plan, ultimately achieving improvements in assembly accuracy and quality.

[0109] The Slime Mould Algorithm (SMA), the moss growth algorithm and the improved moss growth algorithm of this embodiment are compared when the population size is 40 and the update generation number is 100. Figure 2As shown in the figure, the improved moss growth algorithm approaches the minimum fitness value at the 50th generation. Compared with other optimization algorithms, the average computation time is significantly reduced, and the convergence speed and optimization ability are improved. The improved moss growth algorithm shows good convergence and requires fewer iterations to reach the optimal solution.

[0110] Example 2:

[0111] The assembly process parameter optimization system based on the improved moss growth algorithm includes an indicator input module, a data analysis module, a process parameter optimization model module, and a solution output module;

[0112] The index input module selects several assembly accuracy indicators;

[0113] The data analysis module analyzes the factors affecting the assembly accuracy index, evaluates the degree of influence of each factor on the assembly accuracy index, selects the factors with greater influence, inputs them into the constructed improved support vector machine accuracy prediction model, and outputs the assembly accuracy prediction value; and uses the assembly accuracy prediction value to formulate the process optimization objective function;

[0114] The process parameter optimization model module selects the assembly process parameters to be optimized, formulates the assembly process parameter optimization constraints, and constructs the process parameter optimization model in combination with the process optimization objective function; the process parameter optimization model uses the improved moss growth algorithm to solve the optimal assembly process parameters;

[0115] The solution output module outputs the optimal assembly process parameters to form an assembly process adjustment solution.

Claims

1. An assembly process parameter optimization method based on an improved moss growth algorithm, characterized in that: The following steps are involved: (1) Select several assembly accuracy indicators; (2) Analyze the influencing factors of assembly accuracy indicators, evaluate the degree of influence of each influencing factor on the assembly accuracy indicators, select the influencing factors with greater influence, input them into the constructed improved support vector machine accuracy prediction model, and output the assembly accuracy prediction value; The objective function of process optimization is formulated using the assembly accuracy prediction value; (3) Select the assembly process parameters that need to be optimized, formulate the assembly process parameter optimization constraints, combine the process optimization objective function, and build a process parameter optimization model; (4) The process parameter optimization model uses the improved moss growth algorithm to solve the optimal assembly process parameters; (5) Output the optimal assembly process parameters to form an assembly process adjustment plan.

2. The assembly process parameter optimization method based on the improved moss growth algorithm according to claim 1, characterized in that: The assembly accuracy index includes the assembly deformation, the change in the size of key components and the change in the assembly gap.

3. The assembly process parameter optimization method based on the improved moss growth algorithm according to claim 1, characterized in that: The construction and training of the improved support vector machine accuracy prediction model include: first, normalizing and preprocessing the multidimensional error data of the assembly process; second, adaptively setting the initial search range of the SVM hyperparameters based on the maximum standard deviation of the input data; then using a grid search strategy that combines coarse and fine granularity to optimize the hyperparameters, first performing a coarse-grained search at a larger step size to locate the potential optimal interval, and then narrowing the range based on the coarse search results to perform a fine-grained and refined search, ultimately determining the optimal hyperparameter combination; finally, using the optimal parameters to train the SVM model, and verifying its prediction performance through cross-validation and an independent test set.

4. The assembly process parameter optimization method based on the improved moss growth algorithm according to claim 1, characterized in that: The process optimization objective function is: Where x is the process parameter vector, ε is the assembly error, n is the number of assembly accuracy indicators, and f i (x) is the assembly accuracy prediction value output by the improved support vector machine accuracy prediction model, [y min_i ,y max_i ] is the accuracy range required by the process, i=1,2,……n.

5. The assembly process parameter optimization method based on the improved moss growth algorithm according to claim 1, characterized in that: The influencing factors include assembly matching dimensions, assembly process parameters, and environmental data.

6. The assembly process parameter optimization method based on the improved moss growth algorithm according to claim 1, characterized in that: The influencing factors of the assembly accuracy index are analyzed and the influence degree of each influencing factor on the assembly accuracy index is evaluated using a grey relational analysis or a principal component analysis method.

7. The assembly process parameter optimization method based on the improved moss growth algorithm according to claim 1, characterized in that: The assembly process parameter optimization constraint condition refers to setting the adjustment range of the assembly process parameter to be optimized in combination with the current production status and production conditions.

8. The assembly process parameter optimization method based on the improved moss growth algorithm according to claim 1, characterized in that: The improved moss growth algorithm is used in step (4) to solve the optimal assembly process parameters. The specific steps are: (4.1) Initialize algorithm related parameters; (4.2) Improve the random population initialization to use Latin hypercube sampling to initialize the position of the moss population. Divide the interval of each dimension into several equal parts, and randomly select points in each part to ensure uniform distribution in each dimension. (4.3) After obtaining the initial moss population, calculate the fitness of the initial population in sequence, obtain the optimal individual, calculate the average distance between the individual and the optimal individual, and smooth the path of the individual approaching the optimal individual; (4.4) The dynamic Levy flight strategy is introduced to dynamically adjust the search step size to update the position. The formula is as follows: Where: i is the index of the moss individual currently being updated, is the population position determined by the original moss growing algorithm, The population location determined by the present invention, levy step (FEs) is the dynamic Levy flight step length introduced in the present invention, and R is a uniform random number independent of each dimension; The dynamic Levy flight step length introduced by the present invention is shown in the following formula: Where: FEs is the current number of iterations, levy step (FEs) is its flight step size, MaxFEs is the maximum number of iterations; β is the step size distribution parameter, u and v are random variables that obey a specific distribution; (4.5) The improved Gaussian mutation is introduced to perform population individual mutation. The Gaussian mutation formula is as follows: Where: Δx i is the variation of the i-th variable, N(0,σ 2 ) represents a model with mean 0 and variance σ 2 is a normally distributed random variable; γ is the variation intensity adjustment coefficient, which is used to adjust the variation range of σ and thus control the variation range of Gaussian variation; (4.6) When a better individual is found in the population, the better individual is used to replace the current best individual to achieve population evolution; (4.7) When the maximum number of recorded generations of cryptobiotic search is reached, the moss cryptobiotic mechanism is activated and the current individual is replaced by the best individual recorded; (4.8) Record the current number of iterations and determine whether the maximum number of iterations has been reached. When the number of iterations has not reached the maximum number of iterations, the algorithm repeats steps (4.2) to (4.7) until the maximum number of iterations is reached, and outputs the optimized assembly process parameters.

9. The assembly process parameter optimization system based on the improved moss growth algorithm is characterized by: It includes indicator input module, data analysis module, process parameter optimization model module and solution output module; The index input module selects several assembly accuracy indicators; The data analysis module analyzes the factors affecting the assembly accuracy index, evaluates the degree of influence of each factor on the assembly accuracy index, selects the factors with greater influence, inputs them into the constructed improved support vector machine accuracy prediction model, and outputs the assembly accuracy prediction value; The objective function of process optimization is formulated using the assembly accuracy prediction value; The process parameter optimization model module selects the assembly process parameters to be optimized, formulates the assembly process parameter optimization constraints, and constructs the process parameter optimization model in combination with the process optimization objective function; the process parameter optimization model uses the improved moss growth algorithm to solve the optimal assembly process parameters; The solution output module outputs the optimal assembly process parameters to form an assembly process adjustment solution.

10. The assembly process parameter optimization system based on the improved moss growth algorithm according to claim 9, characterized in that: The improved moss growth algorithm is used to solve the optimal assembly process parameters, and the specific steps are as follows: (4.1) Initialize algorithm related parameters; (4.2) Improve the random population initialization to use Latin hypercube sampling to initialize the position of the moss population. Divide the interval of each dimension into several equal parts, and randomly select points in each part to ensure uniform distribution in each dimension. (4.3) After obtaining the initial moss population, calculate the fitness of the initial population in sequence, obtain the optimal individual, calculate the average distance between the individual and the optimal individual, and smooth the path of the individual approaching the optimal individual; (4.4) The dynamic Levy flight strategy is introduced to dynamically adjust the search step size to update the position. The formula is as follows: Where: i is the index of the moss individual currently being updated, is the population position determined by the original moss growing algorithm, The population location determined by the present invention, levy step (FEs) is the dynamic Levy flight step length introduced in the present invention, and R is a uniform random number independent of each dimension; The dynamic Levy flight step length introduced by the present invention is shown in the following formula: Where: FEs is the current number of iterations, levy step (FEs) is its flight step size, MaxFEs is the maximum number of iterations; β is the step size distribution parameter, u and v are random variables that obey a specific distribution; (4.5) The improved Gaussian mutation is introduced to perform population individual mutation. The Gaussian mutation formula is as follows: Where: Δx i is the variation of the i-th variable, N(0,σ 2 ) represents a model with mean 0 and variance σ 2 is a normally distributed random variable; γ is the variation intensity adjustment coefficient, which is used to adjust the variation range of σ and thus control the variation range of Gaussian variation; (4.6) When a better individual is found in the population, the better individual is used to replace the current best individual to achieve population evolution; (4.7) When the maximum number of recorded generations of cryptobiotic search is reached, the moss cryptobiotic mechanism is activated and the current individual is replaced by the best individual recorded; (4.8) Record the current number of iterations and determine whether the maximum number of iterations has been reached. When the number of iterations has not reached the maximum number of iterations, the algorithm repeats steps (4.2) to (4.7) until the maximum number of iterations is reached, and outputs the optimized assembly process parameters.

Images