Fault Diagnosis to Nuclear Power Plant System Based on Time-Series Convolution Neural Network

Table 2

CNN network parameters.

Structure name	Network structure
Convolution layer 1	Number of filters:8; Kernel size: Activation function:ReLu
Max-pooling layer 1	Pool size:
Convolution layer 2	Number of filters:16; Kernel size: Activation function:ReLu
Max-pooling layer 2	Pool size:
Convolution layer 3	Number of filters:16; Kernel size: Activation function:ReLu
Max-pooling layer 3	Pool size:
Fully-connected layer	Nodes: 32; activation function: ReLu
Output layer	Nodes:14; activation function: Softmax
Optimization algorithm	Adam algorithm (learning rate: 0.001; beta1: 0.9; beta2: 0.99)
Loss function	Sparse_categorical_crossentropy