Improved Fault Detection by Reducing Bandwidth

Vibration analysis algorithms for shaft and gear faults are predominately based on the time synchronous average (TSA). The TSA resamples the vibration data over one shaft revolution, and averages that data over the number of revolutions associated with the acquisition. This has the effect of improving Fourier analysis for that shaft (no spectral smearing due to changes in shaft speed over the acquisition) and reducing non-synchronous noise. A number of analyses can be performed on the TSA to extract features that are indicative of faults, such as the statistics on the TSA itself, the residual analysis, energy operator and Amplitude/Frequency Modulation (AM, FM) of the TSA Narrowband Analysis. This paper discuses two bandwidth reduction technique that improve signal to noise associated with non-synchronous noise. Reduced noise improves fault detection in the TSA, and all analysis dependent on the TSA, (including residual, energy operator). For the TSA, this analysis is validated by comparing the condition indicators of a faulted pinion with and without bandwidth reduced. This technique has other advantages for embedded and online condition monitoring systems, by significantly reducing the computational load of subsequent analysis based on the TSA. For the AM/FM analysis, an empirical rule for bandwidth selection is made that maximizes seperability between a faulted and nominal pinion.