Experimental Investigation on the Performance of Signal Processing Tools for the Analysis of Mechanical Vibrations in Rotor Rubbing

Abstract

Rubbing is an important problem in machinery industry which occurs when a rotating element hits a stationary part. This rotor-to-stator rub may result in the catastrophic breakdown of the machine. In this work, the phenomenon of rotor rubbing is analyzed from the perspective that the signal analysis tools that are in use today to detect this defect emphasize or highlight particular aspects of the studied phenomenon. So, sometimes it is necessary to use more than one tool to deepen the understanding of the problem. For this purpose, laboratory tests were performed on a rotor system with a rubbing mechanism, while mechanical vibrations were measured with an accelerometer and a data acquisition system. Experiments were carried out for fixed rotor speed, and for run-up and run-down rotor speed conditions. The analysis approach included various processing tools to study their capabilities in rubbing detection: Root Mean Square (RMS), Fourier transform, Wavelet transform and Hurst exponent. Fixed rubbing conditions show similar results for RMS and Hurst exponent on the information obtained. For variable run-up and run-down rotor speed conditions, the Hurst exponent shows predictability, a fact that can be used for rub detection. However, the Wavelet and Fourier Transforms operated in a very distinct way. Although both transforms give frequency information, Fourier transform results in a more detailed frequency analysis, while the Wavelet transform can give time localization of the rubbing phenomenon.