AOVMD-ScaleShift-Net: An Integrated Framework for Enhanced Rolling Bearing Fault Diagnosis

To overcome VMD parameter sensitivity and feature extraction challenges in bearing fault diagnosis, this paper introduces the AOVMD-ScaleShift-Net framework. Its core, Adaptive Optimization VMD (AOVMD), uses an Enhanced Crested Porcupine Optimizer (ECPO) to co-optimize Spectral Kurtosis (SK) and Time-Frequency Entropy (TFE) for robust IMF decomposition. ScaleShift-Net, a hybrid CNN-BiLSTM architecture for efficient, linear-complexity feature learning, processes the decomposed modes. Validated on the Paderborn dataset, the framework achieves 99.52% accuracy on complex faults, demonstrating the synergy between adaptive decomposition and deep learning.