Randomized trial demonstrates effective fault diagnosis in rolling bearings, suggesting strong noise resistance.
Key Points
The aim is to develop a deep learning-based multi-modal framework for robust fault diagnosis in rolling bearings under noisy conditions.
Proposed a multi-modal fusion framework utilizing time-domain, frequency-domain via FFT, and time-frequency-domain via Multi-scale wavelet convolution (MWC).
Employed a parallel CNN-LSTM architecture to extract features from each modality, integrating an attention mechanism for optimal feature fusion.
Performed experiments on the HUST bearing dataset.
Achieved 100% classification accuracy for both single and compound faults.
Demonstrated consistent and stable performance across varying signal-to-noise ratios (SNR).
Confirmed the effectiveness of learned feature representations through T-SNE visualization and cross-dataset validation.