A novel fault diagnosis model synergizing weighted health indicator with feature-enhanced Transformer for wind turbine bearing

Fault signals of wind turbine bearings have nonlinear and non-stationary characteristics, and the fault features are weak under actual noise interference conditions, leading to difficulties in fault feature extraction and identification. This paper proposes a novel fault diagnosis model synergizing weighted health indicator with feature-enhanced Transformer for wind turbine bearing to improve the fault recognition ability. Firstly, a lens opposition-based learning strategy is introduced to enhance the electric eel foraging optimization (EEFO) algorithm. The improved EEFO (IEEFO) is then used to optimize multivariate variational mode decomposition (MVMD), enabling efficient signal decomposition. The signal is reconstructed by constructing a new health indicator to amplify fault feature distinctiveness. Then, the continuous wavelet transform (CWT) is applied to convert the reconstructed signal into a 2D image, preserving its time-frequency characteristics. Finally, an improved Swin Transformer model incorporating linear deformable convolution (LDConv) is designed to further enhance feature extraction and classification performance. The proposed method is validated using bearing data from Case Western Reserve University (CWRU) and a wind turbine bearing simulation test bench. Comparative experiments demonstrate that this approach achieves higher diagnostic accuracy for wind turbine bearings compared to other methods.