Application of non-stationary signal characteristics using wavelet packet transformation

Although Fourier-based methods have been standard methods for frequency analysis, they are not well suited for the analysis of nonlinear or non-stationary systems due to their time-varying natures. Thus, in this paper, a wavelet packet-based technique, which calculates time-varying coherence functions for input/output relationships, is developed. The developed method uses the Coiflet wavelet that has been widely used in signal processing. It is applied to obtain the time-varying coherence function, and to detect the impulse signal from the impulse-embedded signal such as an automobile sound/vibration signal with an external impact caused by a collision or passing over rough terrain. Some characteristics of non-stationary behavior such as the wavelet packet coefficients, maximum phase plane (MPP) analysis and fault detection are also demonstrated. The method gives promising results of non-stationary input-output systems, and so may be used as an effective tool for condition monitoring or fault detection area. This paper was recommended for publication in revised form by Associate Editor Yeon June Kang Jae-Eung Oh received his B.S. degree of mechanical engineering at Hanyang University in 1975 and his M.S. degree of Safety Engineering at Yokohama National University in 1980. He then went on to receive his doctorate degree in environmental engineering from Tokyo Institute of Technology in 1983. He is currently the Vice President of KSNVE.