APPLICATION OF SPECTRAL CORRELATION TECHNIQUES ON MINING MACHINE SIGNALS: EXTRACTION OF FAULT SIGNATURES

摘要

The early detection of faults in crucial elements of mechanical systems like the bearings is of great interest to industrialists to perform preventive maintenance. However, the extraction of the time waveform linked to a fault is a hard task due to the presence of temporally and spectrally overlapping interferences stemming from the machine vibrating parts. A solution to this problem resides in exploiting the property of cyclostationarity characterising rotating machine signals. In this work, a novel cyclostationarity-based blind signal extraction (BSE) technique is applied to mining machine signals using the prior information obtained from the spectral correlation analysis. Experiments carried on complex gearboxes show that, while simple band pass filtering is not efficient due to the mentioned overlapping effect, the application of BSE achieved a significant improvement in the quality of the separation.