Abstract: Aiming at parameters optimization problem of classifier based on support vector machine, one kind of parameters selection algorithm was presented based on chaos genetic algorithm. The antirigonometric function Logistic map that has uniform track distribution was used to carry out chaos optimization. Therefore, it can search whole optimization interval by equal probability, overcoming disadvantage that Logistic map chaos optimization searches optimization interval edge by greater probability. Start population was produced by using chaos ergodicity. Chaos disturbance was added to chromosome that has bad fitness, in order to carry out chaos optimization. It solved premature problem and convergenee problem of genetic algorithm.This method was applied to fault classifier parameters optimization of openswitch damage in brushless do motor power converter. Experimental results assess effectiveness and feasibility of the proposed approach.

Key words: Chaos optimization, Genetic algorithm, Support vector machine, Gauss kernel, Fault diagnosis