A Study on Parameter Selection Using Health Index in Gabor Wavelet for Fault Diagnosis

Abstract

Prognostics and Health Management (PHM) aims monitoring condition of equipment life trend to detect possible future failure earlier. This kind of applications are fields of Condition Monitoring Maintenance. Therefore the vibration (signal) data is used to diagnose the fault. Signals are generally are including noise. In recent years, several good methods have been developed for the time-frequency signal analysis to reduce the external noise from the PHM data. One of the method of time-frequency domain method is Gabor Wavelet Transform. Since Gabor Wavelet Transform uses Gaussian window as a band-pass filter, it gives the best time-frequency localization. Because of that, the good de-noising process depends on good band-pass filters performance. With that parameter selection for band-pass filtering is the critical stage of de-noising process. In this study, we used Squared Modulus of Gabor Wavelet coefficients-based Crest Factor for selection of the optimal parameter of band-pass filter.