Fault diagnosis and health assessment for bearings using the Mahalanobis–Taguchi system based on EMD-SVD

Bearing failure is the main cause of breakdown in rotating machinery. This paper proposes a new method for diagnosing faults and assessing the health of bearings using the Mahalanobis–Taguchi System (MTS). Our approach utilizes empirical mode decomposition and singular value decomposition to process the non-linear and non-stationary vibration signal of a bearing. In this method, the vibration signal is first decomposed to a number of intrinsic mode functions and a residue to form a feature matrix. Singular values of this feature matrix are obtained by SVD, at which point MTS is employed. MTS provides: 1) a computational scheme based on the Mahalanobis distance for fault clustering; and 2) Taguchi methods to extract the key features. In addition, we formulate a new assessment method that obtains the health index of a bearing. This method is based on a normal condition dataset, without the need for failure data, which is a notable indicator for bearing health tracking and defect detection at the incipient stage. Finally, the feasibility and efficiency of this method is validated by two different bearing experiments.