An Integrated Bearing Fault Diagnosis Method Based on Multibranch SKNet and Enhanced Inception-ResNet-v2

Table 1

Description of SIR-CNN model parameters.

Modules	Blocks	Operations and parameters
Stem	—	As in Figure 1
Depthwise separable convolution with NewSKNet	Branch_0	, , P = “same”; , , P = “valid”
	Branch_1	C =  − 3, S = 1 × 1, P = “same”; C = , S = , P = “valid”
	Branch_2	, , P = “same”; , , P = “valid”
	NewSKNet	As in Figure 3
Reduction A	Branch_0	, , P = “valid”
	Branch_1	; ; , , P = “valid”
	Branch_2	Max , , = “valid”
Depthwise separable convolution with NewSKNet	Branch_0	, , P = “same”; , , P = “valid”
	Branch_1	, ,  = “same”; , , = “valid”
	Branch_2	, ,  = “same”; , , = “valid”
	NewSKNet	As in Figure 3
Reduction B	Branch_0	; , , = “valid”
	Branch_1	; , , = “valid”
	Branch_2	; ; , , = “valid”
	Branch_3	Max , , = “valid”
Depthwise separable convolution with NewSKNet	Branch_0	, ,  = “same”; , , = “valid”
	Branch_1	, ,  = “same”; , , = “valid”
	Branch_2	, ,  = “same”; , , = “valid”
	NewSKNet	As in Figure 3
Average pooling	—	512 × 1
Dropout	—	Ratios 0.8
SoftMax	—	n_classes × 1