Method of Optimal Design with SVR-PSO for Ultrasonic Cutter Assembly

摘要

The optimal design parameters of the ultrasonic cutter assembly can be determined by means of the finite element method (FEM). However it is time-consuming because the combinations of design parameters are varied and the optimization of acoustic performance indexes is multi-objective. In order to achieve the optimal combination quickly a new method is proposed which combines support vector regression (SVR) and particle swarm optimization (PSO) into the design for ultrasonic cutter assembly. When the ranges of the design parameters are set, some samples are produced easily with the aid of the parameterized software of the finite element method. The approximate model between the design parameters and the acoustic performance indexes is established with support vector regression. Based on the approximate model a large number of samples are produced quickly as initial particles, and the optimal design parameters of the ultrasonic cutter assembly are defined efficiently with particle swarm optimization. The result demonstrates that the optimization method with SVR-PSO not only reduces the computation time, but also has good accuracy compared with the finite element method.