Experimental Evaluation of a Unified Time-Scale-Frequency Technique for Bearing Defect Feature Extraction

Abstract

A systematic experimental study is presented in this paper on evaluating the effectiveness of a unified, multidomain algorithm for defect feature extraction in bearing condition monitoring and health diagnosis. The algorithm decomposes vibration signals measured on bearings by discrete wavelet transform and subsequently performs the Fourier transform on the wavelet coefficients. The effectiveness of such a unified technique is demonstrated through experimental case studies, which confirmed its advantage over the wavelet or Fourier transform techniques employed alone. Also, the unified technique has shown to be computationally more efficient than the enveloping technique based on continuous wavelet transform, thus providing a good signal processing tool for bearing defect diagnosis.