Unpublished conference/Abstract (Scientific congresses and symposiums)
Dimension Reduction and Breakdown of Recurrent Neural Networks in the context of Multiscale Analyses
Wu, Ling; Noels, Ludovic
2023International Conference on Plasticity, Damage, and Fracture 2023 (ICPDF 2023)
 

Files


Full Text
2023_Plasticity_RNN.pdf
Author postprint (3.9 MB)
Download

All documents in ORBi are protected by a user license.

Send to



Details



Keywords :
neural network; computational mechanics; multiscale
Abstract :
[en] Once properly trained, artificial Neural Networks can serve as surrogate model of non-linear history-dependent micro-scale boundary value problems in multi-scale methods. This paradigm has been essentially used as a mapping between the macro-stress and macro-strain tensors of the micro-scale boundary value problem and the micro-structure information could not be recovered in a so-called localization step. However such information can be of interest in order to assess failure or fatigue, e.g. In this paper, we present a Recurrent Neural Networks (RNNs)-based surrogate of the micro-scale boundary value problem, while being able to recover the evolution of the local micro-structure fields. In order to address the curse of dimensionality arising because of the large amount of internal state variables defining the micro-structure, PCA dimension reduction and dimension break down, i.e. the use of several RNNs instead of a single one, are combined to reduce the size of the different RNNs that have to be trained. We show that RNNs can be used to conduct multi-scale simulations of heterogeneous elasto-plastic structures while micro-scale information can be recovered during a post-simulation localization step.
Research center :
A&M - Aérospatiale et Mécanique - ULiège
Disciplines :
Mechanical engineering
Materials science & 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 :
Dimension Reduction and Breakdown of Recurrent Neural Networks in the context of Multiscale Analyses
Publication date :
2023
Event name :
International Conference on Plasticity, Damage, and Fracture 2023 (ICPDF 2023)
Event place :
Punta Cana, Dominican Republic
Event date :
3-9 January 2023
By request :
Yes
Audience :
International
European Projects :
H2020 - 862015 - MOAMMM - Multi-scale Optimisation for Additive Manufacturing of fatigue resistant shock-absorbing MetaMaterials
Name of the research project :
Multiscale Optimisation for Additive Manufacturing of fatigue resistant shock-absorbing MetaMaterials (MOAMMM)
Funders :
EC - European Commission [BE]
UE - Union Européenne [BE]
Funding number :
862015
Funding text :
This project has received funding from the European Union´s Horizon 2020 research and innovation programme under grant agreement No. 862015 for the project “Multiscale Optimisation for Additive Manufacturing of fatigue resistant shock-absorbing MetaMaterials (MOAMMM)” of the H2020-EU.1.2.1. - FET Open Programme. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union. Neither the European Union nor the granting authority can be held responsible for them.
Commentary :
Keynote lecture
Available on ORBi :
since 12 January 2023

Statistics


Number of views
81 (3 by ULiège)
Number of downloads
33 (1 by ULiège)

Bibliography


Similar publications



Contact ORBi