Abstract:A critical issue to ensuring the effectiveness of wavelet transform in machine condition monitoring and health diagnosis is the choice of the most suited base wavelet for signal decomposition and feature extraction. This paper addresses this issue by introducing a quantitative measure to select an appropriate base wavelet for analyzing vibration signals measured on rotary mechanical systems. Specifically, the measure based on energy-to-Shannon entropy ratio has been investigated. Both the simulated Gaussian-modulated sinusoidal signal and an actual ball bearing vibration signal have been used to evaluate the effectiveness of the developed measure on base wavelet selection. Experimental results demonstrate that the wavelet selected using the developed measure is better suited than other wavelets in diagnosing structural defects in the bearing. The method developed provides systematic guidance in wavelet selection.

Base Wavelet Selection for Bearing Vibration Signal Analysis