Research shows improved fault diagnosis accuracy in equipment manufacturing using deep learning and sensor data fusion, indicating potential for better maintenance.
Key Points
The aim is to improve fault diagnosis accuracy in equipment manufacturing using deep learning combined with multi-source sensor data.
Constructed a multi-source sensing data cube with time-space synchronization.
Utilized linear interpolation and Z-score standardization for heterogeneous data alignment.
Designed a parallel CNN-LSTM hybrid architecture for spatial and temporal feature extraction.
Introduced a cross-modal attention mechanism for adaptive weighting of multi-modal features.
Achieved an accuracy of 94.7% in fault identification.
Obtained a macro-average F1 score of 0.932.
Demonstrated better robustness against noise and class imbalance compared to traditional and single modal models.