Singular Spectrum Analysis for Bearing Defect Detection

Abstract

In this work, singular spectrum analysis is employed to process vibration signals resulting from rolling bearings. A monitoring indicator is defined from the fact that the structure of a signal recorded from a bearing becomes more complex when the bearing becomes defective. This can be explained by the nonstationarity and the nonlinearity induced by the defect. The effects of operating parameters such as load and speed on the indicator are studied. Results demonstrate that the indicator defined in this paper is less sensitive to these parameters than the rms value, a traditional vibratory indicator.