Περίληψη: | This study explores the potential for automated detection of incipient faults in rotating machinery under different operating speeds based on few vibration signals and advanced Statistical Time Series (STS) methods. The considered faults cause no obvious effects on the time domain signals and their characteristics (e.g., rms, variance), while their effects on the vibration signals power spectral density are completely masked by the effects due to the different operating speeds, thus making their detection highly challenging. Despite the importance to the machinery maintenance organization and reduced cost, as well as the increased safety, such problems have received limited attention in the pertinent literature. To this end, three powerful STS methods are employed in this study, the Functional Model Based Method (FMBM), a Multiple Model (MM) based method, and a PCA based one, while just a single accelerometer is used on the considered machinery, which consists of two electric motors coupled via a claw clutch (jaw coupling). The methods’ experimental performance assessment is based on thousands of inspection (unknown) test cases with the healthy machinery, as well as with three incipient faults, the first and second corresponding to the reduction of the tightening torque on one of the machinery’s mounting bolts, and the third to slight wear in the claw clutch's elastic element (spider), while the machinery operates under 21 distinct speeds. The results indicate remarkable detection performance via the FMBM and the MM based methods, with the FMBM also offering an accurate estimate of the machinery rotation speed.
|