Fig 1.
CNN structure.
Fig 2.
GADF conversion.
Fig 3.
Self-calibrated convolution.
Fig 4.
Cooperative compensation attention mechanism.
Fig 5.
DSCNN model structure.
Table 1.
DSCNN model hyperparameters.
Table 2.
Implementation process of DSCNN.
Fig 6.
HUST bearing fault physical map.
Fig 7.
HIT aero-engine test rig.
Fig 8.
HIT bearing fault physical map.
Table 3.
Aero-engine bearing data.
Fig 9.
Comparison of GADF plots before and after adding noise.
(a)-(e) denotes no noisy. (a’)-(e’) denotes adding noisy.
Fig 10.
HUST classification results.
Fig 11.
t-SNE visualization and confusion matrix.
Fig 12.
HIT classification results of different datasets.
Fig 13.
Data visualization results on the HIT dataset.
(a)-(f) represent the t-SNE visualization result. (a’)-(f’) represent probability density estimation curves and histograms for different models.