Health assessment of rotary machinery based on integrated feature selection and Gaussian mixed model

Bearing failure is the most common failure mode of all rotary machinery failures, and can interrupt the production in a plant causing unscheduled downtime and production losses. A bearing failure also has the potential to damage machinery causing soaring machinery repair and/or replacement costs. In order to prevent unexpected bearing failure, a health assessment method is proposed in this paper. It employs an integrated feature selection approach and Gaussian mixture model (GMM). Firstly, the integrated feature selection approach, which combines empirical mode decomposition (EMD), singular value decomposition (SVD) and Principal Component Analysis (PCA), processes nonlinear and non-stationary vibration signals of a bearing and extracts features for health assessment. Then, GMM is utilized to evaluate and track the health degradation of the bearing in terms of confidence values (CV). This method, which is notable for bearing health tracking and detect the defect at its incipient stage, can be used without the need for failure datasets in applications. Finally, the feasibility and efficiency of this method was validated by two datasets of different bearing experiments.