[en] Multi-scale simulations can be accelerated by substituting the micro-scale problem resolution by a surrogate trained from off-line simulations. In the context of history-dependent materials, recurrent neural networks have widely been considered to act as such a surrogate, e.g. [1], since their hidden variables allow for a memory effect.
However, defining a training dataset which virtually covers all the possible strain-stress state evolution encountered during the online phase remains a daunting task. This is particularly true in the case in which the strain increment size is expected to vary by several orders of magnitude. Self-Consistent recurrent networks were thus introduced in [2] to reinforce the self-consistency of neural network predictions when small strain increments are expected. This new cell was applied to substitute an elasto-plastic material model.
However when considering a representative volume element response in the context of multi-scale simulations, it was found that the Self-Consistent recurrent networks requires a long training process. In this work, we revisit the Self-Consistent recurrent unit to improve the training performance and reduce the number of trainable variables for the neural network to act as a composite surrogate model in multi-scale simulations.
This project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No. 101056682.
REFERENCES
[1] L. Wu, V. D. Nguyen, N. G. Kilingar, L. Noels, A recurrent neural network-accelerated multi-scale model for elasto-plastic heterogeneous materials subjected to random cyclic and non-proportional loading paths, Computer Methods in Applied Mechanics and Engineering 369 (2020) 113234. doi: https://doi.org/10.1016/j.cma.2020.113234
[2] C. Bonatti, D. Mohr, On the importance of self-consistency in recurrent neural network models representing elasto-plastic solids, Journal of the Mechanics and Physics of Solids 158 (2022) 104697. doi: https://doi.org/10.1016/j.jmps.2021.104697
[3] L. Wu and L. Noels. "Self-consistency Reinforced minimal Gated Recurrent Unit for surrogate modeling of history-dependent non-linear problems: Application to history-dependent homogenized response of heterogeneous materials." Computer Methods in Applied Mechanics and Engineering, 424 (01 May 2024): 116881. doi:10.1016/j.cma.2024.116881
Research Center/Unit :
A&M - Aérospatiale et Mécanique - ULiège [BE]
Disciplines :
Mechanical engineering
Author, co-author :
Wu, Ling ; Université de Liège - ULiège > Département d'aérospatiale et mécanique > Computational & Multiscale Mechanics of Materials (CM3)
Noels, Ludovic ; Université de Liège - ULiège > Département d'aérospatiale et mécanique > Computational & Multiscale Mechanics of Materials (CM3)
Language :
English
Title :
A Self-consistent Reinforced minimal Gated Recurrent Unit for surrogate modelling of elasto-plastic multi-scale problems
Publication date :
31 May 2024
Event name :
The 19th European Mechanics of Materials Conferences (EMMC19)
Event place :
Madrid, Spain
Event date :
29-31 May 2024
Event number :
19th
Audience :
International
Peer reviewed :
Editorial reviewed
Development Goals :
9. Industry, innovation and infrastructure
European Projects :
HE - 101056682 - DIDEAROT - Digital Design strategies to certify and mAnufacture Robust cOmposite sTructures
Name of the research project :
DIDEAROT
Funders :
EC - European Commission [BE] Union Européenne [BE]
Funding number :
101056682
Funding text :
This project has received funding from the European Union’s Horizon Europe Framework Programme under grant agreement No. 101056682 for the project ‘‘DIgital DEsign strategies to certify and mAnufacture Robust cOmposite sTructures (DIDEAROT)’’. The contents of this publication are the sole responsibility of ULiege and do not necessarily reflect the opinion of the European Union. Neither the European Union nor the granting authority can be held responsible for them.
