Spectral negentropy and kurtogram performance comparison for bearing fault diagnosis

While investigating rotating faults in vibration signals, one of the typical symptoms is the presence of repetitive transients, which are characterized by impulsive and cyclostationarity signature. The approach quite popular nowadays in the industry for their detection is time-frequency techniques. Those techniques are mainly analysis tools as opposed to processing tools, and in any case, are unable to offer a versatile methodology that applies to all mechanical signals in all circumstances. The paper is motivated by ideas borrowed from thermodynamics, where transients are seen as departures from a state of equilibrium; it is proposed to measure the negentropy of the squared envelope (SE) and the squared envelope spectrum (SES) of the signal. The main objective of the work is to investigate connections in both negentropy and kurtogram approaches to capture the signature of this repetitive behavior. The methodology used in this paper proposes to display spectral negentropy as images. The impulsive events are then detected and localized in frequency by high values of the squared envelope spectrum (SES) infogram in some frequency bands. In order to analyze the signal in the frequency domain, the Short-Time Fourier Transform (STFT) can then be used. The STFT is suggested in this study due to its simplicity and high flexibility. On the other hand, STFT is used for the analysis of kurtosis of temporal signals; this is well known as Kurtogram. For fault, such as bearings, Kurtogram was demonstrated to be efficient.