Feng Y., Schlindwein F.

A five stage 'Roots and Claw' dry vacuum pump is a typical kind of quasi-steady state rotating machine. The research using the novel Acoustic Emission-based data acquisition system aims to develop advanced fault prediction methods for dry vacuum pumps to prevent pumps’ faults or failure. Traditional signal-model-based fault prediction methods, including parametric or non-parametric spectral estimation, assume that signal is stationary. While traditional frequency spectrum is dominated by low frequency contents and has no information regarding the time, wavelet transform is more suitable to analyse the transient Acoustic Emission signal. It well locates Acoustic Emission signal at both time and scale. In order to meet the requirements of fast computation implementation, Discrete Wavelet Transform is needed and signals can be analyzed at different scales. In this paper, we introduce the Acoustic Emission-based data acquisition system and then study the thresholding problem of Discrete Wavelet Transform. The investigation includes the prevalent thresholding estimators: SURE, Heuristic SURE, SquareTwoLog, Minimax, Penalization strategy. We propose the use of threshold estimators depends on the sample size N, wavelet basis and resolution scale j. The study convinces us the use of Acoustic Emission and Discrete Wavelet Transform is a promising way for fault prediction on quasi-steady rotating machines.