支持向量机的参数优化及其在故障诊断中的应用  

Optimizing Support Vector Machine Parameters and Application to Fault Diagnosis

作　　者：林辉[1] 王德成[1] 

机构地区：[1]西北工业大学自动化学院,西安710072

出　　处：《计算机科学》2010年第4期255-257,284,共4页Computer Science

基　　金：航空科学基金项目(2007ZC53036)资助

摘　　要：针对支持向量机分类器的参数优化问题,提出了一种基于混沌遗传算法的参数选择方法。采用轨道均匀分布的反三角函数Logistic映射产生优化变量,等概率搜索优化区间,克服了Logistic映射优化算法对优化区间边缘进行大概率搜索的缺陷;利用混沌的遍历性产生初始群体,对部分适应度较差的个体进行混沌寻优,解决了遗传算法的早熟和收敛问题。将该方法应用于无刷直流电机功率变换器开关管开路故障分类器中,实现了分类器参数优化。结果表明,该算法是可行、有效的。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 anti-trigonometric 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 convergence problem of genetic algorithm. This method was applied to fault classifier parameters optimization of openswitch damage in brushless dc motor power converter. Experimental results assess effectiveness and feasibility of the proposed approach.

关 键 词：混沌优化 遗传算法 支持向量机 高斯核函数 故障诊断 

分 类 号：TM307[电气工程—电机]