Table 1 Details of proposed model architecture by layer
From: Deep Spatiotemporal Convolutional-Neural-Network-Based Remaining Useful Life Estimation of Bearings
Layer | Details |
|---|---|
Input | Size: N × L × W × H × C; channel: L × C |
Convolution | 3D kernel: 7 × 7 × 2; channel: 32; stride: 2 × 2 × 1 |
Convolution | 3D kernel: 3 × 3 × 2; channel: 64; stride: 2 × 2 × 1 |
Convolution | 3D kernel: 3 × 3 × 2; channel: 128; stride: 2 × 2 × 1 |
Dropout | Rate: 0.25 |
Pooling | Average_3Dpooling: 12×12×L |
Fully-connected | Channel: 128 |
Fully-connected | Channel: 50 |
Output | Channel: 1 |