Application of EMD-AR and MTS for hydraulic pump fault diagnosis

A real-time diagnosis of hydraulic pumps is very crucial for the reliable operation of hydraulic systems. The main purpose of this study is to propose a fault diagnosis approach for hydraulic systems based on the empirical mode decomposition (EMD), autoregressive (AR) model, singular value decomposition (SVD), and Mahalanobis-Taguchi system (MTS). The AR model effectively extracts the fault feature of vibration signals. However, it can only be applied to stationary signals; the fault vibration signals of hydraulic pumps are non-stationary. To address this problem, the EMD method is used as a pretreatment step to decompose the non-stationary vibration signals of hydraulic pumps. First, the vibration signals of hydraulic pumps are decomposed into a finite number of stationary intrinsic mode functions (IMF). The AR model of each IMF component is established. The AR parameters and the remnant's variance are regarded as the initial feature vector matrices. Third, the singular values are obtained by applying the SVD to the initial feature vector matrices. Finally, these values serve as the fault feature vectors to be entered to the MTS, thereby classifying the fault pattern of the hydraulic pumps. The Taguchi methods are employed to reduce the redundant features and extract the principal components. Experimental analysis results indicate that this method can effectively accomplish the fault diagnosis of hydraulic pumps.