Damage identification in multifield materials using neural networks

Figures & data

Fig. 1 Scheme of the a back-propagation NN.

Table 1 Material properties for MEE composite with .

Properties
(GPa)
(GPa)
(GPa)
(GPa)
(N/A m)
(N/A m)
(N/A m)

Table 2 Crack parameters range.

min
max

Fig. 2 Scheme of the direct problem.

Fig. 3 Reference cracks. (a) Large crack, and (b) Small crack.

Table 3 Crack identification of the reference cracks for SNNs.

Noise	Input
Large crack – Figure 3(a)
Real crack
4*
4*
4*
Small crack – Figure 3(b)
Real crack
4*
4*
4*

Fig. 4 Damage identification with SNNs. Input: eletric and magnetic potentials .

Fig. 5 Training set division.

Fig. 6 NN training parameters.

Fig. 7 Damage identification for partitioned training set. Input: eletric and magnetic potentials .

Table 4 Detail of the crack identification of the reference cracks. Input: eletric and magnetic potentials . Noiseless case.

Division
Large crack - Figure 3(a)
Real crack
NN1
NN2
NN3
NN4
NN5
Small crack - Figure 3(b)
Real crack
NN1
NN2
NN3
NN4
NN5

Table 5 NN crack identification of the reference cracks. TSD.

Noise	Input
Large crack – Figure 3(a)
Real crack
4*
4*
4*
Small crack – Figure 3(b)
Real crack
4*
4*
4*

Fig. 8 parameter histograms with VE lower than 0.5. TSD. Input: eletric and magnetic potentials .

Fig. 9 parameter for several noise levels. Input: eletric and magnetic potentials .

Fig. 10 parameter histograms for several noise levels. Input: eletric and magnetic potentials .