Structural Vibration Data Anomaly Detection Based on Multiple Feature Information Using CNN-LSTM Model

Table 2

Description of parameters of the LSTM-CNN model.

Network layer	Parameter setting	Activation function	Output size
Input layer	—	—	(None, 300, 2)
1D-convolutional layer	Size of convolution kernel: 3 × 2; Number of channels: 16; Stride: 1	ReLU	(None, 298, 16)
Max pooling layer	Pooling window: 2 × 1; Stride: 2	—	(None, 149, 16)
Reshape layer	—	—	(None, 2384)
LSTM layer	Number of nerve cell: 100	Tanh	(None, 100)
Fully connected layer	—	—	(None, 100)
Output layer	—	Softmax	(None, 5)