Novel enhancement techniques for noisy vibration signals applied
to local damage detection in rotating machines

Jakub Obuchowski¹, Agnieszka Wyłomanska² and Radosław Zimroz¹

¹Diagnostics and Vibro-Acoustics Science Laboratory - Wroclaw University of Technology
Na Grobli 15, 50-421 Wroclaw, Poland

²Hugo Steinhaus Center, Institute of Mathematics and Computer Science
Wroclaw University of Technology
Janiszewskiego 14 a, 50-370 Wroclaw, Poland

jakub.obuchowski@pwr.wroc.pl, agnieszka.wylomanska@pwr.wroc.pl, radoslaw.zimroz@pwr.wroc.pl

Abstract

In the paper new enhancement techniques for noisy vibration signals are presented in application to local damage detection in heavy rotating machinery. The methodology is consisted of four steps and utilizes time-frequency analysis, statistical analysis and, finally, filtering of a vibration signal. One of the steps is based on inverse procedure to pre-whitening and allows to indicate transients related to local damage without reference to a signal from a healthy machine of the same type. The last step is an extension of the spectral kurtosis (SK) based filter. It is shown that the novel techniques lead to better local damage detection including rejecting false positives, i.e. incidental impulsive signals that might appear during normal machine operation under industrial conditions. The methodology is illustrated by analyses of both real and simulated data representing vibrations of a heavy rotating machinery operating in mining industry.

Please contact the authors in order to get the full paper