A Hybrid Deep Learning Prediction Method of Remaining Useful Life for Rolling Bearings Using Multiscale Stacking Deep Residual Shrinkage Network

Table 5

Structural parameters of MSDRSN.

Module	Layers	Output size and activation
Convolution layers	Conv 1	(None, 3, 6), Activation: ReLU
	Conv 2	(None, 3, 6), Activation: ReLU
	Conv 3	(None, 3, 6), Activation: ReLU-sigmoid
DRSN layers	Residual building unit 1	(None, 6, 3), Activation: ReLU-sigmoid
	Residual building unit 2	(None, 6, 3), Activation: ReLU-sigmoid
	Residual building unit 3	(None, 6, 3), Activation: ReLU-sigmoid
	Pooling	(None, 6, 1), Activation: ReLU-sigmoid
	Fully connected	(None, 1), Activation: ReLU-sigmoid