Research On Optimal Development Module Algorithm Of Complex Products Based On Improved Genetic Algorithm

Complex product refers to a product with many design parts and complex assembly and decomposition relations.Module division of complex product is the most common strategy of complex product development in actual manufacturing enterprises.Module partition to complex product parts according to its function or structure of the distinction between form different modules,through the modular operation can produce a more reasonable product structure,decrease the difficulty of the product development to adapt to the company internal organization structure and development process,and through multiple modules in parallel realization of rapid design of complex product development,production.The DSM model,as a tool for planning and analyzing the product development process,is used to establish a relational matrix for the parts of complex products and analyze the interaction between different parts,to lay a foundation for the subsequent module division.The module partition problem of complex product Design Structure Matrix（DSM）model belongs to NP-hard problem from the theoretical research level,and the common heuristic algorithm for solving this problem is genetic algorithm.When the traditional genetic algorithm solves the DSM module partition problem,the two-dimensional chromosome structure is fixed.If the number of rows representing the number of modules is not equal to the number of parts,it will lead to the immutable defect of the maximum number of modules partition,which will further limit the search space of module partition scheme,so that the optimal solution may not be obtained.If the number of rows is equal to the number of parts,there are defects such as large amount of calculation and slow calculation speed.This paper proposes an improved genetic algorithm for module partition of complex product DSM model.In the process of population initialization,heterogeneous chromosomes with different rows of chromosomes were generated,and the clustering algorithm was used to form multiple niches which evolved independently to maintain the diversity of the population and expand the search range.The variable neighborhood search algorithm is inserted into the traditional genetic algorithm to design a multi-neighborhood structure that changes the number of chromosome rows,and the search results that break the limit of the initial number of chromosome rows are obtained.The traditional crossover mutation operator is improved to the adaptive crossover mutation operator,and the adaptive crossover mutation of heterogeneous chromosomes is realized by the custom rules.Improved genetic algorithm allows the different structure of chromosomes in coevolution and generate new chromosome structure,break the two-dimensional chromosome structure fixed search space constrained problem,can be in all the module partition number search,at the same time,do not need to use part number as the chromosome number of rows,reduced the amount of calculation and improves the operation speed.In view of the defects of genetic algorithm,which is easy to "premature" in the early stage and easy to fall into local optimal in the late stage,crossover and mutation operators with variable probability are designed to realize large range search in the early stage of evolution and preserve population diversity in the late stage of evolution.In the process of evolution,the chromosomes with the highest fitness were selected to carry out frequent pattern mining between different niches at fixed intervals of evolutionary algebra,and frequent patterns were found and used to improve the niches with the lowest fitness,to improve the overall fitness of the population while maintaining the diversity of the population.Based on two real DSM modeling of complex product,this paper USES improved genetic algorithm for module partition experiment,and through the analysis of the experimental process and results,found that the improved genetic algorithm can achieve different chromosome structure search,not only fast convergence rate,and the quality is superior to the traditional genetic algorithm solving,verify the feasibility and effectiveness of the algorithm.