A novel intelligent fault diagnosis model for planetary gearboxes using wavelet-based synchroextracting transform-based deep learning networks

Planetary gearboxes are critical components in rotating machinery where timely fault diagnosis is essential for operational safety and reliability. However, diagnosing faults from vibration signals generated by these complex systems remains challenging due to their non-stationary nature and inherent complexity. To address the challenges, this paper proposes a novel intelligent fault diagnosis model integrating wavelet-based synchroextracting transform (WSET) with a convolutional neural network-bidirectional long short-term memory-black winged kite optimization algorithm-least squares support vector machine (CNN-BiLSTM-BKA-LSSVM) framework. First, the WSET is employed to preprocess raw vibration signals, leveraging the multi-resolution analysis capability of wavelets to suppress noise and extract time-frequency synchronized features, thereby enhancing the separability of fault patterns. Subsequently, a CNN architecture is designed to automatically learn high-level features from the preprocessed signals, capturing local spatial correlations and hierarchical fault signatures. To further improve classification performance, the LSSVM classifier is integrated with the CNN’s feature extraction layer, and its hyperparameters (e.g., kernel parameter and penalty factor) are optimized via the BKA optimization algorithm to enhance the diagnostic performance. Experimental validation is conducted using a planetary gearbox fault dataset under healthy and multiple fault conditions. Comparative results with some existing deep learning methods demonstrate that the proposed intelligent fault model achieves high diagnostic accuracy across all fault types.