Research Article
An Intelligent Fault Diagnosis Based on Adversarial Generating Module and Semi-supervised Convolutional Neural Network
Table 3
The architecture of the proposed framework.
| Purpose | Layers | Number of filters | Kernel size | Output size |
| Feature extraction | Convolutional layer1 | 16 | 116 | 102416 | Down-sampling layer1 | 16 | | 51216 | Convolutional layer2 | 32 | 16 | 51232 | Down-sampling layer2 | 32 | | 25632 | Convolutional layer3 | 64 | 16 | 25664 | Down-sampling layer3 | 64 | | 12864 |
| Fault classification | Fully connected layer1 | 1 | 1024 | 1024 | Fully connected layer2 | 1 | 300 | 300 | Fully connected layer3 | 1 | 30 | 30 | Softmax layer | 1 | | N/A |
|
|