Improved Of Multi-objective Genetic Algorithm And Its Applied In Gearbox Parameters Optimization Design

The solution of multi-objective optimization problems has become the focus ofengineering researches. Genetic algorithm, which has superior ability of globaloptimization, is a powerful tool in dealing with multi-objective optimization problems.Recently, multi-objective genetic algorithm based on Pareto has been focused byscholars and applied in engineering cases.This paper first introduces previous researches on multi-objective genetic algorithmwith its applications in engineering. Then related concepts based on Pareto andmathematical model of multi-objective problems are given. Some classicmulti-objective genetic algorithms are introduced, such as NSGA-II, NPGA algorithms.The advantages and disadvantages of these algorithms are further analyzed.Due to the high computational complexity of non-dominated sorting in NSGA-II anddifficulty of shared parameter determination in NPGA, this paper introduces a minimumcriterion based non-dominated sorting algorithm and verifies that the algorithm has highefficiency in non-dominated sorting from the viewpoint of both theory and practice. Onthe other hand, in the evolution process of genetic algorithm, populations are generatedby chaotic method, and the sparse area of individual distribution in evolutionarypopulation space is filled in, which improves the global searching ability of geneticalgorithm. Finally, in the terms of convergence of multi-objective genetic algorithm, aconvergence criterion is set that the difference between the distances of externalnon-dominated individual storage sets of two adjacent generations should be less thanthe set threshold. Through the examination of DTLZ1-5, the feasibility and validity ofthe improved algorithm is verified and the distribution of improved algorithm is provedto be more uniform than NSGA-II in getting Pareto-optimal solutions.Finally, for the optimal design problems of parallel gearbox gear planetary gear train,the establishment process of the mathematical model is discussed and the gearbox geargear parameter is optimized by improved multi-objective genetic algorithm. Throughsimulating experiments, the distribution curve of Pareto-optimal solution is obtained.The distribution is uniform and it is then compared with the model parameter obtainedby actual design. The comparison result verifies that the improved algorithm can dealwith gearbox gear gear optimization problems efficiently.