基于双谱和PSO-SVM的滚动轴承故障诊断

Rolling bearing fault diagnosis based on bispectrum and PSO-SVM

针对故障轴承信号的非线性、非高斯性,提出了一种基于双谱和纠错编码支持向量机(error-correcting output codes support vector machine,ECOC-SVM)的滚动轴承故障诊断方法。使用故障轴承振动信号双谱特征构造特征向量,在SVM的训练过程中,使用微粒群算法(particleswarm optimization,PSO)对支持向量机的参数进行优化。实验结果表明该方法能获得较高分类准确率。

Vibration signals of fault rolling bearings are of non-linear and non-Gaussian.In this study,a new bearing diagnosis method based on bispectrum and error-correcting output codes support vector machine is presented.Several features are extracted from the bispectrum for SVM inputs.The particle swarm optimization algorithm is used for searching the optimum SVM parameters.Experimental results show that the algorithm result in good classification purpose.