Table 1 Model structures, reactor transients, computational costs, MSEs, and relative errors.

From: Physics-informed neural network with transfer learning (TL-PINN) based on domain similarity measure for prediction of nuclear reactor transients

Experiment

Model

Pre-train

Predicted transient

Iterations

Learning rate

Time [s]

Train loss (MSE)

Test loss (MSE)

Test metric (L2)

1

PINN

–

RT-1

95,000

0.0003

62.94

6.58·10–5

8.08·10–5

1.19·10–4

2

TL-PINN

RT-2

RT-1

4000

0.001

2.65

2.36·10–4

2.86·10–4

1.62·10–4

3

TL-PINN

RT-3

RT-1

7000

0.001

4.14

1.52·10–4

1.27·10–4

1.66·10–4

4

TL-PINN

RT-4

RT-1

9000

0.002

5.06

3.63·10–4

3.72·10–4

1.98·10–4

5

TL-PINN

RT-5

RT-1

10,000

0.001

5.48

2.91·10–4

2.83·10–4

2.64·10–4

6

PINN

–

RT-2

95,000

0.0003

63.37

8.15·10–5

8.81·10–5

2.50·10–4

7

TL-PINN

RT-1

RT-2

5000

0.0003

3.08

2.31·10–4

2.06·10–4

4.69·10–4

8

TL-PINN

RT-3

RT-2

3000

0.001

2.01

2.37·10–4

2.27·10–4

3.77·10–4

9

TL-PINN

RT-4

RT-2

4000

0.001

2.56

2.39·10–4

2.11·10–4

2.65·10–4

10

TL-PINN

RT-5

RT-2

6000

0.0003

3.63

1.89·10–4

2.02·10–4

4.36·10–4

11

PINN

–

RT-3

93,000

0.0003

68.22

8.04·10–5

8.76·10–5

3.22·10–4

12

TL-PINN

RT-1

RT-3

12,000

0.002

6.62

2.57·10–4

2.55·10–4

4.24·10–4

13

TL-PINN

RT-2

RT-3

9000

0.0001

5.37

3.96·10–4

4.42·10–4

4.52·10–4

14

TL-PINN

RT-4

RT-3

3000

0.0006

2.03

6.45·10–5

7.45·10–5

2.74·10–4

15

TL-PINN

RT-5

RT-3

3000

0.001

2.06

4.89·10–4

6.01·10–4

2.82·10–4

16

PINN

–

RT-4

95,000

0.0003

68.36

5.47·10–5

8.89·10–5

2.76·10–4

17

TL-PINN

RT-1

RT-4

18,000

0.001

9.83

2.05·10–4

2.07·10–4

4.74·10–4

18

TL-PINN

RT-2

RT-4

13,000

0.0001

7.48

3.55·10–4

4.85·10–4

5.11·10–4

19

TL-PINN

RT-3

RT-4

0

–

0.19

8.04·10–5

8.76·10–5

3.22·10–4

20

TL-PINN

RT-5

RT-4

6000

0.0003

3.69

3.39·10–4

4.32·10–4

2.94·10–4

21

PINN

–

RT-5

105,000

0.0003

72.63

1.04·10–4

1.11·10–4

4.64·10–4

22

TL-PINN

RT-1

RT-5

14,000

0.0003

7.84

1.29·10–4

1.40·10–4

4.86·10–4

23

TL-PINN

RT-2

RT-5

15,000

0.0001

8.53

2.06·10–4

2.49·10–4

4.99·10–4

24

TL-PINN

RT-3

RT-5

15,000

0.0001

8.38

5.84·10–5

6.04·10–5

5.01·10–4

25

TL-PINN

RT-4

RT-5

7000

0.0003

4.03

8.21·10–5

7.95·10–5

4.12·10–4

  1. The total number of iterations in an operational sequence can be obtained by adding the corresponding number of iterations for PINN and TL-PINN models.
Back to article page