Extracting Acoustical Impulse Signal of Faulty Bearing Using Blind Deconvolution Method

Machine fault diagnosis, based on acoustic signals, is frequently made difficult by noisy environments at a production site. In this paper, an improved time-domain blind deconvolution algorithm, based on envelope spectrum and normalized kurtosis, was proposed to recover acoustic signals of defective bearings. A newly defined distance measure based on envelope spectrum was employed to improve the classification accuracy of independent components in the cluster analysis process, and a kurtosis-based criterion was applied to select optimum components. With the help of these enhancements, reliable estimated results can be obtained with low computational complexity, even when the time-delay or the reverberation time is sufficiently large. Both numerical and experimental studies were carried out. The results show that this algorithm can be efficiently applied to rolling element bearing defect detection in real-world situations, and is very promising in acoustic-based machine diagnosis.