Fault Diagnosis of Aero-Engine Bearings Using Deep Learning and Feature Fusion

This study proposes a novel fault diagnosis method for aero-engine bearings, combining FFT (Fast Fourier Transform), VMD (Variational Mode Decomposition), and a deep learning model (CNN-BiLSTM-CrossAttention) to enhance diagnosis accuracy and real-time performance. The method extracts multi-scale time-frequency features using FFT and VMD, spatial features with CNN (Convolutional Neural Networks), temporal dependencies with BiLSTM (Bidirectional Long Short-Term Memory), and fuses them through a cross-attention mechanism to improve classification accuracy. Experimental results demonstrate 100% accuracy on both the CWRU dataset and the Harbin Institute of Technology aero-engine bearing dataset, showing strong robustness and generalization. In addition, The CNN+BiLSTM model also maintains high accuracy under noise, indicating the stability and adaptability of the method in complex environments. This study provides a reliable technical framework for aero-engine bearing fault diagnosis with significant application value.