A Deep Domain-Adversarial Transfer Fault Diagnosis Method for Rolling Bearing Based on Ensemble Empirical Mode Decomposition

Table 2

Specific parameters of RNB.

Type	Layer	Kernel/stride/channels	Input size	Output size
Input		/	(32,32,1)	(32,32,1)
Conv1	Conv2d	(3,3)/1/16	(32,32,1)	(32,32,16)
	InstanceNorm2d	/	(32,32,16)	(32,32,16)
	ReLU	/	(32,32,16)	(32,32,16)
Block1	Conv2d	(3,3)/1/16	(32,32,16)	(32,32,16)
	InstanceNorm2d	/	(32,32,16)	(32,32,16)
	ReLU	/	(32,32,16)	(32,32,16)
	Conv2d	(3,3)/1/16	(32,32,16)	(32,32,16)
	InstanceNorm2d	/	(32,32,16)	(32,32,16)
Block2	Conv2d	(3,3)/1/32	(32,32,16)	(32,32,32)
	InstanceNorm2d	/	(32,32,32)	(32,32,32)
	ReLU	/	(32,32,32)	(32,32,32)
	Conv2d	(3,3)/2/32	(32,32,32)	(16,16,32)
	InstanceNorm2d	/	(16,16,32)	(16,16,32)
Block3	Conv2d	(3,3)/1/32	(16,16,32)	(16,16,32)
	InstanceNorm2d	/	(16,16,32)	(16,16,32)
	ReLU	/	(16,16,32)	(16,16,32)
	Conv2d	(3,3)/2/32	(16,16,32)	(8,8,32)
	InstanceNorm2d	/	(8,8,32)	(8,8,32)
Pooling	AvgPool	/	(8,8,32)	(1,1,32)