Unified time–scale–frequency analysis for machine defect signature extraction: Theoretical framework

The effectiveness of signal processing plays a critical role in machine condition monitoring and health diagnosis, especially under the presence of noise contamination. This paper presents a new approach to unifying techniques in the time, scale, and frequency domains. Specifically, spectral post-processing is performed on the data set extracted by wavelet transforms to enhance the effectiveness of defect feature extraction. The theoretical framework for such a generalized signal transformation platform is introduced, and boundary conditions for implementing the new technique are discussed. Comparison with enveloping technique based on band-pass filtering and wavelet transform has shown that the new technique is more effective in identifying structural defects in bearings, and computationally more efficient, thus providing a good alternative to envelope analysis for defect signature extraction in machine condition monitoring.