Dual-Enhanced Sparse Decomposition for Wind Turbine Gearbox Fault Diagnosis

Abstract

The gearbox is one of the most important components in a wind turbine (WT) system, and fault diagnosis of WT gearbox for maintenance cost reduction is of paramount importance. However, fault feature identification is a primary challenge in gearbox fault diagnosis because weak fault features are always obscured by heavy background noise and multiple harmonic interferences. In this paper, a dual-enhanced sparse decomposition (DESD) method is proposed to address the feature enhancement and identification for gearbox fault vibration signal. Within the proposed method, the nonconvex generalized minimax-concave (GMC) penalty is used to construct the sparse-regularized cost function, the convexity of which can be maintained and the cost function can be minimized using convex optimization algorithms to obtain its global minimum. Furthermore, an adaptive regularization parameter selection scheme is proposed for the proposed DESD method in signal decomposition and feature extraction. Simulation studies and a real case study validate that the proposed method can better preserve the feature components of interest and can significantly improve the estimation accuracy. The comparison studies also show that the proposed method outperforms those methods with L1 norm regularization and spectral kurtosis.