PSI - Issue 5

Piotr Nazarko et al. / Procedia Structural Integrity 5 (2017) 460–467 Nazarko and Ziemianski / Structural Integrity Procedia 00 (2017) 000 – 000

466

7

a)

b)

Fig. 8. (a) Signals received at the initial state; (b) projection of two first principal components.

First, it was assumed that data from all bolts (2, 4, 5) would be used for the purpose of ANNs training, assuming a constant distribution of patterns for testing and validation. The obtained results are very promising, what is shown in Fig. 9. These results are also very accurate if presented with respect to the force changes during the test (Fig. 10). However, if patterns from the bolts no. 3 were provided to the ANN trained, its generalization ability can not be accepted. It was expected that the predicted force values would be on the same level like the others, while the obtained results are quite unpredictable in the initial phase (see “Rp - ” path in Fig. 10) .

b)

c)

a)

Fig. 9. Results of ANNs training based on the signals recorded in bolts of the flange connection (constant patterns distribution): (a) learning error, (b) testing error, (c) validation error.

Fig. 10. The axial force prediction in flange connection based on the ANN training results.