A Feature Selection Framework-Based Multiscale Morphological Analysis Algorithm for Fault Diagnosis of Rolling Element Bearing

Multiscale morphological analysis (MMA) is considered as a prevalent and efficient approach of mathematical morphology (MM), which has received a lot of attention in fault diagnosis. However, traditional MMA mainly focuses on the selection of structuring element (SE) scale and fault feature extraction based on single index, which is not easy to get comprehensive and rich fault information. Consequently, for the purpose of solving the issue of losing local fault information of integrating traditional MMA with single index and improving fault feature extraction accuracy, a new algorithm named feature selection framework-based multiscale morphological analysis (FS-MMA) is proposed in this paper. Within this algorithm, the weighted MMA is firstly formulated through the incorporation of three operations (i.e. combination morphological filter-hat transform (CMFH), multicale SE and weighted arithmetic), which can cover fault symptoms at different scales. Subsequently, multi-domain features of the raw vibration signal are calculated and entropy weight method (EWM) is adopted to select several typical sensitive features. Finally, grey correlation analysis (GCA) is conducted to determine the optimal SE scale of MMA and achieve fault feature extraction of rolling element bearing. The effectiveness and feasibility of the presented algorithm are validated by analyzing the simulated and experimental bearing fault data. The analysis results show that FS-MMA has better performance in bearing fault feature extraction and diagnosis accuracy compared with traditional MMA with single index.