Impact of pulse time uncertainty on synchronous average: statistical analysis and relevance to rotating machinery diagnosis

Time synchronous averaging for the extraction of periodic waveforms is a rather common processing method for rotating machinery diagnosis. By synchronizing the signal to the rotational angle of the component of interest, e.g. by using a keyphasor reference signal, it is possible to perform the averaging in the angular domain, thus obtaining an angle-synchronous signal. Jittering of the reference signal affects the quality of the synchronous averaging process, resulting in attenuation and uncertain estimation of the extracted synchronous signal, especially in the high frequency band. In this paper, the effects of random uncertainty in the pulse arrival times of the reference signal on the synchronous averaging method are studied, with the objective of assessing the relevance of such a jitter error to the extracted waveform and the indicators derived for monitoring purposes. First, a unified framework for the computed order tracking method is presented, and then a model linking the statistics of the random jitter to the statistics of the waveform extracted through synchronous averaging in angle domain is developed. The theoretical model connects the random jitter distribution, the number of averaged periods and the ratio of the period of interest to the reference trigger period, to the distribution of the amplitudes of the synchronous frequency components in the synchronously averaged signal. Approximate analytical solutions are derived for cases of interest, allowing the prediction of the attenuation bias and variability of the extracted components amplitudes. The model is first verified against numerical simulations in order to assess consistency, and then parametric studies are presented. Experimental validation is performed on both an experimental and an operational data sets involving respectively a helicopter gearbox and a helicopter fleet.