Advances, Systems and Applications
Table 2 Parameter setup of ISNN
From: Bearing fault diagnosis method based on improved Siamese neural network with small sample
Network name | Layer type | Main parameters | Output size |
|---|---|---|---|
Feature extraction network | Input | \ | 40 \(\times\) 100 |
LSTM1 | (40,10) | 10 \(\times\) 100 | |
LSTM2 | (40,10) | 10 \(\times\) 100 | |
1D CNN | (1,16,9,3,1) | 16 \(\times\) 332 | |
Max pooling | (5,3) | 16 \(\times\) 110 | |
Relationship measurement network | Input | \ | 32 \(\times\) 110 |
1D CNN | (32,32,5,2,1) | 32 \(\times\) 54 | |
1D CNN | (32,16,5,2,1) | 16 \(\times\) 26 | |
Global average pooling | \ | 16 | |
Full connection | (16,1) | 1 | |
Fault classification network | Input | \ | 16 \(\times\) 110 |
1D CNN | (16,32,5,2,1) | 32 \(\times\) 54 | |
1D CNN | (32,16,5,2,1) | 16 \(\times\) 26 | |
Global average pooling | \ | 16 | |
Fully connected | (16,5) | 5 |