Bearing Defect Classification Algorithm Based on Autoencoder Neural Network

Table 2

Parameters of convolutional fully connected network structure.

Network layer	Operation	Kernel size	Slip step	Number of neurons	Edge fill type
1	C	1 × 1	1 × 1	1	—
2	C	5 × 5	2 × 2	6	SAME
3	C	5 × 5	2 × 2	16	SAME
4	F	—	—	1024	—
5	F	—	—	3	—

C denotes the convolutional layer and F denotes the fully-connected layer.