Applicability of Artificial Neural Networks (ANN) for Equilibrium State Prediction in Tunnel Excavation - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Applicability of Artificial Neural Networks (ANN) for Equilibrium State Prediction in Tunnel Excavation

Résumé

Simplified 2D analysis of ground support interaction using the convergence-confinement method, although useful in preliminary stages of tunnel design, may be inadequate for complex situations which necessitate 3D numerical simulations with high computational efforts and costs. An alternative approach based on machine learning is proposed here in order to evaluate the stresses and displacements at equilibrium in the lining of a supported tunnel. Based on data previously obtained by numerical simulations, the Bagging Method applied to Artificial Neural Networks (ANNs) is used. We consider a circular tunnel excavated in a Mohr-Coulomb elastoplastic medium. The analysis accounts for a large range of ground conditions, support characteristics, lay distances and tunnel radius. The results show that ANNs models perform well with the small dataset used here and can be considered as a useful alternative to complex 3D numerical simulations.
Fichier non déposé

Dates et versions

hal-04376928 , version 1 (07-01-2024)

Identifiants

  • HAL Id : hal-04376928 , version 1

Citer

Alec Tristani, Lina-María Guayacán-Carrillo, Jean Sulem, Sebastián Ariel Donzis. Applicability of Artificial Neural Networks (ANN) for Equilibrium State Prediction in Tunnel Excavation. 15th ISRM Congress, Oct 2023, Salzbourg, Austria. ⟨hal-04376928⟩
19 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More