Application of Variational Mode Decomposition to Feature Isolation and Diagnosis in a Wind Turbine

Abstract

In general, when the misalignment fault occurs in a wind turbine, the vibration signals present the non-stationary and non-linear characteristic nature. The early misalignment fault signal is easily overwhelmed by the strong background signals and noise, making it difficult to detect reliable fault feature. This work focuses on the signal processing-based feature Isolation and Diagnosis for misalignment faults. In this paper, a novel variational mode decomposition (VMD) is introduced to address the issue instead of other common adaptive decomposition algorithms such as empirical mode decomposition (EMD) and wavelet transform. VMD is capable of decomposing the fault vibration signal into several stable components and realize the separation of misalignment fault component from background signals. Both the numerical simulation and a case study using the fault data from our test rig demonstrate the effectiveness of this method. The characteristic 2X frequency can be extracted from the stable components obtained by VMD efficiently. On the contrary, the fault feature of the components decomposed by the comparative methods is relatively unconspicuous due to the mode mixing and frequency aliasing.