A Blind De-convolution Technique for Machine Fault Diagnosis
Liu Tingting, Ren Xing-Min · 2009
Blind source separation (BSS) is a powerful signal processing technique, based on the advantage of data mining and handing. The observed signal containing too much redundancy information, the application of BSS in machine vibration signals is a new method for feature extracting. The traditional BSS model usually neglect that machine vibration signal always is wideband signal for it can be viewed as sums of differently convolved source, and non-stationary, which will brings the dissatisfactory effects for machine fault diagnosis. In this paper, the BSS technique based on second-order statistic been expended to blind de-convolution (BD), combined the advantage of time-frequency analysis (TFA), a TFA-BD algorithm is proposed for machine vibration signal. The experimental result of the numerical simulation data and the actually measuring data shows that the method is efficiency. Comparable with traditional BSS, the performance index of separation can be increased by once at least. The method is suitable for machine vibration signal processing and fault diagnosis.