关键词: 广义精细复合多尺度样本熵, 判别式扩散映射分析, 故障诊断, 流形学习, 滚动轴承

Abstract: Aiming at the difficulty of extracting fault features of rolling bearings, a feature extraction method was proposed based on GRCMSE and manifold learning. GRCMSE was utilized to extract the features of rolling bearings.DDMA method was employed to reduce the dimension of the high-dimensional feature sets. The low-dimensional fault features were input into particle swarm optimization support vector machine(PSO-SVM) multi-fault classifier for fault identification. The experimental results of rolling bearing fault diagnosis show that the features extraction effectiveness of GRCMSE is better than that of MSE, RCMSE and GMSE,the dimensionality reduction effectiveness of DDMA is preferable to Isomap and local tangent space alignment(LTSA),the fault recognition accuracy of rolling bearings reaches 100% by combining GRCMSE and DDMA.

Key words: generalized refined composite multiscale sample entropy(GRCMSE), discriminant diffusion maps analysis(DDMA), fault diagnosis, manifold learning, rolling bearing