Bearing Health Assessment Using Gaussian Mixed Model Based on a Hybrid Feature Extraction Schemer

Bearings, as important components, are widely used in almost all forms of rotary machines. Bearing failure is one of the foremost causes of breakdown in rotating machinery. Such failure can be catastrophic and often results in lengthy industrial downtime that has economic consequence. In order to prevent unexpected bearing failure, this paper presents a health assessment method using Gaussian mixture model (GMM) based on a hybrid feature extraction method. This hybrid feature extraction method combines Empirical Mode Decomposition (EMD) and Singular Value Decomposition (SVD) to process the nonlinear and non-stationary vibration signal of bearing. Then, the health condition of bearing can be assessed and tracked in terms of confidence values (CVs) obtained by GMM. This method can be employed only using normal condition datasets without the need of failure data, which is a notable indicator for bearing health tracking and defect detection at the incipient stage. Its performance and effectiveness has also been validated via a bearing test-bed.