Rotating machinery failure detection using different vibration analysis techniques

This paper deals with the application of some vibration analysis techniques used for the diagnosis of machine condition for the discrimination of the different types of defects in rotating machines from the data contained in the vibration spectra. Cluster analysis and discriminant analysis are the main statistical techniques used. These techniques have been extensively used in various scientific fields for the discrimination of data, but rarely for the diagnosis of machinery faults using vibration analysis. The application of these techniques was carried out in the detection of failures in a set of rolling element bearings with four different known bearing conditions (new bearing, outer race defect bearing, inner race defect bearing and ball defect bearing). The different elements (rolling element bearings) are grouped into clusters as homogenous as possible according to the observed variables (vibration amplitude at different frequencies). The results provide a useful criterion for the discrimination among the different types of failures in the bearings.