Health Monitoring Based on the Higher Order Time-Frequency Analysis

초록

In machine condition monitoring many fault-related vibration signals are highly non-stationary. Time-frequency representations comprise the natural signal-processing tool traditionally used for representing such non-stationary signals. The methods used in this analysis are based on the second order statistics of a signal. This paper discusses the use of higher order time frequency methods in the context of a condition monitoring application. The paper outlines the two classes of higher order representations: the L-Wigner distribution and the sliced Wigner higher order distribution. The use of higher order distributions introduces the possibility of non-oscillating cross-terms, a problem not encountered in conventional bilinear time-frequency methods. Techniques for reducing these cross-terms are presented. The paper compares the performance of bilinear and higher order time-frequency methods using synthetic data and vibration signals measured on the engine block of the passenger car for identification of the dynamic characteristics of the engine valve train system.