An efficient multi-objective optimization method based on the adaptive approximation model of the radial basis function

Considering the high computational cost caused by solving multi-objective optimization (MOO) problems, an efficient multi-objective optimization method based on the adaptive approximation model is developed. Firstly, the Latin hypercube design (LHD) is employed for obtaining the initial sample points. Secondly, initial approximation models of objective functions and constraints are established by using the radial basis function (RBF). For ensuring the accuracy of the approximation models, the reverse shape parameter analysis method (RSPAM) is proposed to obtain improved approximation models. Thirdly, the micro multi-objective genetic algorithm (mu MOGA) is adopted to solve the Pareto optimal set and the local-densifying approximation method is also applied to strengthen the ability of solving accurate Pareto optimal sets. Finally, the effectiveness and practicability of the proposed method is demonstrated by two numerical examples and two engineering examples.

Automated summary

This method utilizes Latin hypercube design to obtain initial sample points, establishes approximation models using radial basis functions, and improves accuracy through reverse shape parameter analysis. It employs micro multi-objective genetic algorithm to solve Pareto optimal sets and enhances the ability to find accurate Pareto optimal sets using local-densifying approximation method.