Skip to main content

Table 3 Architecture of the 1-D CNN

From: A dynamic-model-based fault diagnosis method for a wind turbine planetary gearbox using a deep learning network

Layer

Parameters

Stride

Output size

Input

/

/

2000 × 1

C1

64 × 1 conv

1

16-[2000 × 1]

P1

2 × 1 max-pool

2

16-[1000 × 1]

C2

16 × 1 conv

1

32-[1000 × 1]

P2

2 × 1 max-pool

2

32-[500 × 1]

C3

16 × 1 conv

1

32-[500 × 1]

P3

2 × 1 max-pool

2

32-[250 × 1]

C4

5 × 1 conv

1

64-[250 × 1]

P4

2 × 1 max-pool

2

64-[125 × 1]

C5

5 × 1 conv

1

64-[125 × 1]

P5

2 × 1 max-pool

2

64-[62 × 1]

FC1

Flatten

/

3968 × 1

FC2

Fully-connected

/

256 × 1

FO

Softmax

/

4 × 1