基于VMD-MPE-KPCA特征提取与MRVM相混合的滚动轴承故障诊断方法

针对滚动轴承振动信号在强噪声环境下出现非线性、非平稳、强干扰特性,进而导致故障特征难以提取及故障诊断准确率低的问题,提出变分模态分解(VMD)-多尺度排列熵(MPE)-核主元分析(KPCA)特征提取与多分类相关向量机(MRVM)相混合的滚动轴承故障诊断方法.该方法首先通过VMD-MPE进行滚动轴承振动信号的高维故障特征提取,其次对提取的故障特征进行KPCA可视化降维,最后将降维后的故障特征输入可实现不同样本概率输出的MRVM进行滚动轴承故障诊断.通过美国西储大学的滚动轴承故障数据集对该方法的有效性进行验证,结果表明提出的VMD-MPE-KPCA特征提取与MRVM相混合的滚动轴承故障诊断方法能够有效提取和识别滚动轴承故障特征,所提出的混合智能故障诊断方法与相关文献报道的故障诊断方法相比较,故障识别准确率达到了99.18%.

Fault diagnosis method of rolling bearing based on VMD-MPE -KPCA feature extraction mixed with MRVM

In view of the nonlinear, non-stationary and strong interference characteristics of rolling bearing vibration signals in strong noise environment, which leads to difficulty of fault feature extraction and low accuracy of fault diagnosis, this paper proposes a rolling bearing fault diagnosis method on the basis of combination of variational mode decomposition (VMD)-multi-scale permutation entropy (MPE)- kernel principal component analysis (KPCA) and multi-class relevant vector machine (MRVM). In this method, high-dimensional fault features of rolling bearing vibration signals are extracted first by VMD-MPE, and then the extracted fault features are visually reduced dimensionally by KPCA. Finally, the dimensionally-reduced fault features are input into MRVM which can realize different sample probability output for rolling bearing fault diagnosis. The effectiveness of the proposed method is verified by a set of rolling bearing fault data published by Western Reserve University in the United States. The results show that the proposed rolling bearing fault diagnosis method based on “VMD-MPE-KPCA” feature extraction and MRVM may extract and identify rolling bearing fault features effectively. If compared with those fault diagnosis methods reported in related literature, the fault identification accuracy of the proposed hybrid intelligent fault diagnosis method reaches up to 99.18%.