Cepstrum editing (liftering) to remove discrete frequency signals-leaving a signal dominated by structural response effects- and enhance fault detection in rolling element bearings

摘要

This paper proposes a new approach to completely remove all discrete frequencies (periodic signals) from the spectrum by setting the high quefrency part of the real cepstrum of the original signal to zero (keeping only the low quefrency part which is dominated by transfer path effects (structural resonances)). Normally, to edit the cepstrum and return to the time domain, it is necessary to use the complex cepstrum, but the latter requires the phase signal to be unwrapped. This is not possible for response signals containing discrete frequencies and noise, where the phase is not continuous. The procedure proposed in this paper uses the real cepstrum to edit the log amplitude of the original signal, setting the high quefrency part of the real cepstrum to zero, and then combines the edited (smoothed) amplitude with the original phase spectrum to return to the time domain. Two cases are presented in this paper to illustrate the benefits and advantages of the proposed approach. The first is taken from a helicopter gearbox with defective planetary bearing to show the effectiveness of the approach in enhancing fault detection by removing gear signals. The second case is taken from a jet engine (Larzac) that has two spools. The speeds of the two shafts have slight variations, and the high pressure (HP) spool tacho signal was taken from a geared shaft with limited information about the exact gear ratio. The proposed Cepstrum approach illustrates the removal of the two sets of harmonics of the spools in one operation, and compares this to an earlier approach using the tachometer signals and separate time synchronous averaging (TSA) for each shaft. The results are very well comparable. The proposed approach provides a very beneficial way of separating discrete frequency and random signals in a very effective and quick manner. There is no need for the removal of speed fluctuations, as required in other approaches such as TSA and Discrete Random Separation (DRS). The proposed approach has a very attractive potential application in operational modal analysis and work is being undertaken to extend this concept further.