Neural Networks NARX for Durability Bonded Joint Prediction
Résumé
From the experimental data of the asymetric cleavage test at the bonded joint we studied the crack length propagation as a function of the fracture energy thanks an artificial neural networks. We used the Nonlinear Autoregressive Exogenous (NARX) neural network. The model predicts the crack length with a good agreement with the experimental findings. However, training of the stabilization crack on too long time can cause over-training.
| Origine | Fichiers éditeurs autorisés sur une archive ouverte |
|---|