Fault Diagnosis of Planetary Gearbox Based on Motor Current Signal Analysis

Table 2

CNN structure.

Input signal data (2560 × 1)
Convolutional layer 1	Input channel: 1, output channel: 4, kernel size: 3
	ReLU
	Max pooling, filter size: 5
	Output size (512 × 4)
Convolutional layer 2	Input channel: 4, output channel: 8, kernel size: 3
	ReLU
	Max pooling, filter size: 4
	Output size (128 × 8)
Convolutional layer 3	Input channel: 8, output channel: 16, kernel size: 3
	ReLU
	Max pooling, filter size: 4
	Output size (32 × 16)
Convolutional layer 4	Input channel: 16, output channel: 32, kernel size: 3
	ReLU
	Max pooling, filter size: 4
	Output size (8 × 32)
Flatten	Feature size (256 × 1)
Fully connected layer	Input size: 256, output size: 2560
	ReLU
	Input size: 2560, output size: 8
Final output (classified by softmax)