| Types | Operation | parameter settings |
| Improved pulse codingmodule | Spilking Separable Convs_1 Spilking Separable Convs_2 |
|
| Spilking Separable Convs_1 | Depthwise Convs | Conv2d, out_channels = 3, kernel_size = (3,3), padding = 1, groups = 3 |
| Pointwise Convs | Conv2d, out_channels = 128, kernel_size = (1,1), padding = 0, groups = 1 | |
| BatchNorm2d | out_channels = 128 | |
| neuron.IFNode | out_channels = 128 | |
| Spilking Separable Convs_2 | Depthwise Convs | Conv2d, out_channels = 128, kernel_size = (3,3), padding = 1, groups = 3 |
| Pointwise Convs | Conv2d, out_channels = 128, kernel_size = (1,1), padding = 0, groups = 1 | |
| BatchNorm2d | out_channels = 128 | |
| neuron.IFNode | out_channels = 128 | |
|
| ECABlock | out_channels = 128 |
| PulseMaxpool2d | out_channels = 128, kernel_size = (4,1), padding = 0 | |
| RUL prediction module | Transformer-Encoder MLP |
|
| Transformer-Encoder | Transformer-Encoder_1 |
|
| Transformer-Encoder_2 |
| |
| Transformer-Encoder_1 | Multi-head Attention | Input_dim = 3, d_model = 128, heads = 4, dropout = 0.1 |
| Add & Norm | d_model = 128, Num_layers = 2 | |
| Feed Forward | dim_feedforward = 4* input_dim | |
| Add & Norm | d_model = 128, Num_layers = 2 | |
| Transformer-Encoder_2 | Multi-head Attention | Input_dim = 3, d_model = 128, heads = 4, dropout = 0.1 |
| Add & Norm | d_model=128, Num_layers = 2 | |
| Feed Forward | dim_feedforward = 4* input_dim | |
| Add & Norm | d_model = 128, Num_layers = 2 | |
| PulseMaxpool2d | out_channels = 128, kernel_size = (8,1), padding = 0 | |
| MLP | LINEAR_1 | Input = 32 * n; Ouput = 1 |
| Training | Epoch = 1000 | batch_size = 32; Optimizers = SGD; lr = 0.001 |