Table 4 Configuration details of the IPM-PCNN model.

From: Damage identification based on the inner product matrix and parallel convolution neural network for frame structure

IPM-PCNN

1D branch

2D branch

Input data

Time series (8108 × 600)

IPM (8108 × 12 × 12)

Feature extraction

Layer

kernel

Shape

Layer

kernel

Shape

1DConv1

7

(8108 × 594 × 64)

2DConv1

3 × 3

(8108 × 10 × 10 × 64)

Max pooling

2

(8108 × 297 × 64)

Max pooling

2 × 2

(8108 × 5 × 5 × 64)

1DConv2

5

(8108 × 293 × 128)

2DConv2

2 × 2

(8108 × 3 × 3 × 128)

Global max-pooling

–

(8108 × 128)

Global max-pooling

–

(8108 × 128)

Feature fusion

Fully connected layers

Layer

Neuron numbers

Concatenate features

256

FC layer 1

150

FC layer2

100

Damaged classification

Dropout

0.5

Fully connection 1ayer

4

Back to article page