基于贝叶斯推断LSSVM的滚动轴承故障诊断

Bayesian inference LSSVM based fault diagnosis method for rolling bearing

针对传统最小二乘支持向量机分类器的参数选择具有随意性和不确定性等不足,采用贝叶斯推断方法通过三级分层推断优化确定最小二乘支持向量机的各参数,有效提高了最小二乘支持向量机的建模效率.将基于贝叶斯推断最小二乘支持向量机分类方法应用于滚动轴承故障诊断中,实验仿真结果表明该方法能有效地识别滚动轴承的故障,且训练时间和测试时间均小于传统最小二乘支持向量机方法。

In order to remedy the randomicity and uncertainty in parameters selection,the least squares support vector machines（LSSVM） classifier＇s parameters are optimally selected by the Bayesian inference with three levels hierarchy,and the modeling efficiency is availably improved.Then,the Bayesian inference LSSVM classification method is applied to the fault diagnosis of rolling bearing.The experiment simulation results show that the proposed approach can identify availably the faults and has shorter training and testing time than traditional LSSVM.