Planetary Gearbox Fault Diagnosis Based On a Multi-Convolutional Neural Network with SVMD and Feature Fusion Under Variable Speed Conditions

Strong background noise interference and weak fault characteristics of wind power gearbox under variable speed are difficult to effectively identify, which hinders the implementation of intelligent diagnosis. To solve the above problems, an intelligent fault diagnosis method based on SVMD and multi-channel convolutional neural networks under time-varying speed is proposed. Firstly, the fault vibration signal of the variable speed gearbox is transformed into an angular domain signal by using computational order tracking. Then, the angular domain signal is adaptively decomposed by SVMD, and the Gini index of each modal signal is calculated for angle domain signal reconstruction, thereby highlighting the weak fault feature components in the angle domain signal. Subsequently, the reconstructed angle domain signal is input into the constructed network for feature fusion and learning, and the network parameters are continuously updated. Finally, the fully trained model is used for wind power gearboxes under time-varying conditions to realize intelligent diagnosis of health status. To verify the feasibility and effectiveness of the proposed method, the wind power gearbox signals under time-varying conditions were collected for verification. The experiment indicates that the diagnostic accuracy of the proposed model method is over 99%. Compared with other typical methods, it improves the characterization ability and diagnostic accuracy of wind power gearbox fault characteristics under time-varying conditions.