A Fault Identification Method for Electric Submersible Pumps Based on DAE-SVM

Table 5

.Model training process final parameter.

Layer	Name	Parameter for layer	Other parameters
1	Input layer	Nodes = 14
2	Hidden layer (1)	Nodes = 10	Activation = ReLU
3	Hidden layer (2)	Nodes = 8	Optimizer = Adam
4	Hidden layer (3)	Nodes = 6	Learning rate = 0.03
5	Hidden layer (4)	Nodes = 8	Batch size = 64
6	Hidden layer (5)	Nodes = 10	Denoising parameter = 0.3
7	Output layer	Nodes = 14