Fault Diagnosis of Wind Turbine Bearing Based on Parameter-Optimized VMD and Multi-feature Fusion

Deep learning techniques have been widely applied to wind turbine bearing fault diagnosis in recent years. However, current methods often face challenges in feature extraction, which may lead to diagnostic failures when working with limited samples. To address this challenge, this paper proposes a novel fault diagnosis method based on parameter-optimized variational mode decomposition (VMD) and multi-feature fusion. Firstly, an improved sparrow search algorithm is proposed to optimize the decomposition parameters of VMD, achieving more accurate feature extraction. Next, ten prior fault features derived from our experience in fault diagnosis are introduced, which effectively assist deep learning techniques in learning fault patterns. Subsequently, a lightweight one-dimensional convolution neural network (1D-CNN) is employed to extract adaptive features from both the raw signal and VMD modal components. Finally, these adaptive features and the prior features are input into a fully connected layer for fault diagnosis. Through the fusion of numerous features, the proposed method attains exceptional performance and robustness in two experiments, particularly in scenarios with limited samples.