Wavelet Based Spectral Kurtosis and Kurtogram: A Smart and Sparse Characterization of Impulsive Transient Vibration

Mechanical signature analysis is of vital importance to the structural health monitoring of mechanical equipment. However, the fast development of mechanical signature analysis tool always requires a rich and deep understanding of state-of-the-art technologies, which is often lacked by the on-site staff. In this chapter, we introduce an effective methodology that ensure automatic detection of impulsive transient vibrations occurring during machinery fault events. This methodology is originally derived from the concept of spectral kurtosis, whose advent has a close relation with the early development of wavelet theory, and acquired a fast computation implementation named fast kurtogram. The essential originality of this methodology lies in its unique way of combining multi-scale analysis and scalar indicator based characterization of impulsive transient components. As a result, this methodology emerges as a single-input-single-output system for both theoretical researchers and on-site engineers. In the presented materials, basics and fundamentals of this fast developing methodology are introduced. The recent improvements mainly focus on the construction of new multi-scale signal decomposition frames and the invention of new scalar-valued indicators. All the efforts are motivated to obtain a satisfactory sparse characterization of impulsive transient components induced by machinery faults. A range of construction examples of wavelet-based spectral kurtosis with their engineering applications are presented to demonstrate the developments.