Optimal wavelet analysis and enhanced independent component analysis for isolated and combined mechanical faults diagnosis

In this paper, a new approach is suggested for isolated and combined mechanical faults diagnosis. The suggested approach consists of two main steps: vibration signal denoising and characteristic frequency extracting. Firstly, an optimal wavelet multi-resolution analysis is employed for reducing noise from vibration signals. Secondly, the enhanced independent component analysis (EICA) algorithm which overcomes the shortcoming of the ICA algorithm and allows selecting the reliable independent components is adopted for source separation. Therefore, simple and comprehensible spectra will be obtained. Finally, the suggested method is tested using real vibration signals. Compared with other approaches, it has been revealed that the suggested method can efficiently be employed to diagnose both isolated and combined mechanical faults.