Table 1 Comparison of local stress distributions.

From: Deep material network via a quilting strategy: visualization for explainability and recursive training for improved accuracy

 

Matrix - mean

Matrix - std.

Inclusion - mean

Inclusion - std.

Vol. frac.

Macro

 

[MPa]

[MPa]

[MPa]

[MPa]

[%]

[MPa]

DNS (FFT)

56.69

7.80

82.32

15.90

0.29

64.18

N = 3 Random

50.97 ± 7.65

13.98 ± 10.65

179.54 ± 70.99

183.17 ± 172.83

0.29 ± 0.00

88.48 ± 19.38

N = 5 Random

55.42 ± 1.30

10.28 ± 2.14

96.98 ± 10.93

64.83 ± 58.21

0.29 ± 0.00

67.57 ± 2.77

N = 7 Random

54.70 ± 0.63

11.08 ± 0.93

93.37 ± 4.07

48.13 ± 17.55

0.29 ± 0.00

66.00 ± 0.91

N = 3 Recursive

58.71

1.84

81.24

40.60

0.29

65.29

N = 5 Recursive

58.47

4.65

87.93

57.24

0.29

67.06

N = 7 Recursive

58.58

4.57

79.90

31.50

0.29

64.83

  1. DNS results show the mean and standard deviation of the stress local distribution for the matrix and inclusion, respectively. Distribution of local stresses obtained directly from the nodes of the networks are listed for DMN randomly initialized (Random) or via the recursive training strategy (Recursive). In the case of randomly initialized networks, results are averaged over 30 different networks which were initialized differently.
Back to article page