Early fault detection in rolling contact bearings

Abstract

[EN] This thesis studies the improvement of the detection of failures in rolling contact bearings using signal processing techniques. Rolling contact bearings are fundamental components in rotating machinery. An improvement of the failure detection in bearings leads to an increase of the safety and the economic profitability of all the applications that use this kind of machines. The first step of the study was to review some of the available methodologies used in condition monitoring of rotating machinery. A deeper review was made for two of these methodologies, vibration diagnosis and acoustic emissions. The next step was to investigate some signal processing techniques that can be applied in order to improve the detection of failures in rolling contact bearings. Finally, one of these techniques, the Kurtogram in conjunction with envelope analysis, was applied to the vibration data of a secondary flight control and to the acoustic emissions data of a test involving varying defect sizes in RC bearings. The main conclusion extracted from the study was that the Kurtogram is a very powerful tool to detect frequency bands with high impulsiveness and therefore containing very useful information for the detection of failures. For the vibration data the Kurtogram was able to detect the defect two days earlier than the conventional frequency spectrum analysis. The acoustic emission proved to be a very powerful tool to detect incipient defects by itself and left slight space for an improvement of the fail detection using the Kurtogram. However, the Kurtogram was useful to extract a frequency band where the signal to noise