Automatic gear and bearing fault localization using vibration and acoustic signals

The vibration and the acoustic responses from a rotary system in a given situation are not identical. The commencement of signal bursts in presence of defect has been investigated to design a robust system of filters, which can perform adequately for both the class of signals captured from geared and bearing systems for a wide range of faults. The proposed system of filters has several stages of signal processing such as denoising, time–frequency analysis, extraction of smooth envelope signal (SES) followed by a robust peak detection technique. First, the strength of wavelet packet transform (WPT) has been exploited along with a proposed algorithm to identify the denoised signal for further processing. In the second stage, the SES has been generated by integrating the enhanced spectrogram coefficients in time domain. The corresponding enhanced time–frequency spectrogram has been generated by adopting complex Morlet wavelet transform (CMWT) followed by a thresholding routine. As the objective is to localize faults in time domain signals, in the last stage, a robust peak detection technique has been integrated in the proposed system of filters. In all the aforementioned stages of filter design a two stage validation process has been followed. This involves a performance analysis with a synthetic signal followed by an experimental investigation. The strength of a strong signal preconditioning, which helps in identifying an appropriate mother wavelet function for different systems and for a wide range of faults, has been highlighted.