基于改进MPE与KELM的滚动轴承故障诊断

Fault Diagnosis of Rolling Bearings Based on Improved MPE and KELM

针对滚动轴承发生故障时非线性信号特征难以提取导致诊断效率较低的难题,提出一种基于参数优化改进的多尺度排列熵(Multi-scale permutation entropy,MPE)与核极限学习机(Kernel extreme learning machine,KELM)相结合的故障诊断方法。首先,使用关联积分法(C-C算法)对MPE的嵌入维数和时间延迟进行优化;其次,计算滚动轴承振动信号在选定的经验参数与优化参数下各尺度的排列熵值并以此构建特征向量;最后,利用KELM对滚动轴承进行故障分类。结果表明,参数优化后的MPE结合KELM的故障诊断方法能够有效地提取出故障特征进而很好地实现故障诊断。

In view of the difficulty for extracting the non-linear signal features and low efficiency of fault diagnosis of rolling bearings,a fault diagnosis method combining multi-scale permutation entropy(MPE)with parameter optimization and kernel extreme learning machine(KELM)was proposed.Firstly,the correlation integration method(C-C method)was used to optimize the embedding dimensions and time delay of the MPE.Secondly,the permutation entropy of the rolling bearing vibration signal was calculated using the selected empirical parameters and the optimized parameters,and then the feature vector was constructed.Finally,the KELM method was used for fault classification of rolling bearings.The results show that the MPE and KELM fault diagnosis methods based on parameter optimization can extract fault features effectively for realizing the fault diagnosis of rolling bearings.