A Hybrid Neural Network Based on Optimized Variational Mode Decomposition Denoising for Wind Turbine Rolling Bearing Fault Diagnosis

The health of the wind turbine bearing has a decisive impact on the continuity and safety of wind power generation. However, diagnosing faults in rolling bearings presents challenges due to their complex operational environment and the non-linear characteristics of the original vibration data generated. To address this issue, this paper proposes a rolling bearing fault diagnosis method, based on optimized Variational Mode Decomposition (VMD) combined with a Convolutional Neural Network (CNN)Transformer architecture. First, by optimizing the parameters of VMD, WOA effectively solves the defects of VMD in parameter selection and decomposition accuracy, decomposes the key information modes in the signal more accurately, and improves the quality of feature extraction, which significantly improves the overall performance and accuracy of fault diagnosis. A CNN-transformer fusion model is proposed for existing fault diagnosis algorithms that suffer from defects such as incomplete feature extraction, insufficient dynamic feature capture, and limited generalization capability. Leveraging the hybrid model to fully explore and utilize the complex patterns and dynamic characteristics within the vibration data for precise identification of wind turbine bearing fault conditions. Finally, experimental results demonstrate the superior performance of the proposed method in wind turbine rolling bearing fault diagnosis.