SO(3)-invariance of informed-graph-based deep neural network for anisotropic elastoplastic materials
Abstract
This work examines the frame-invariance (and the lack thereof) exhibited in simulated anisotropic elasto-plastic responses generated from supervised machine learning of classical multi-layer and informed-graph-based neural networks, and proposes different remedies to fix this drawback. The inherent hierarchical relations among physical quantities and state variables in an elasto-plasticity model are first represented as informed, directed graphs, where three variations of the graph are tested. While feed-forward neural networks are used to train path-independent constitutive relations (e.g., elasticity), recurrent neural networks are used to replicate responses that depends on the deformation history, i.e. or path dependent. In dealing with the objectivity deficiency, we use the spectral form to represent tensors and, subsequently, three metrics, the Euclidean distance between the Euler Angles, the distance from the identity matrix, and geodesic on the unit sphere in Lie algebra, can be employed to constitute objective functions for the supervised machine learning. In this, the aim is to minimize the measured distance between the true and the predicted 3D rotation entities. Following this, we conduct numerical experiments on how these metrics, which are theoretically equivalent, may lead to differences in the efficiency of the supervised machine learning as well as the accuracy and robustness ofmore »
- Authors:
-
- Columbia Univ., New York, NY (United States); RWTH Aachen Univ. (Germany)
- Columbia Univ., New York, NY (United States)
- Publication Date:
- Research Org.:
- Columbia Univ., New York, NY (United States)
- Sponsoring Org.:
- USDOE Office of Nuclear Energy (NE); National Science Foundation (NSF); US Air Force Office of Scientific Research (AFOSR)
- OSTI Identifier:
- 1801241
- Alternate Identifier(s):
- OSTI ID: 1599468
- Grant/Contract Number:
- NE0008534; CMMI-1846875; FA9550-17-1-0169; FA9550-19-1-0318
- Resource Type:
- Accepted Manuscript
- Journal Name:
- Computer Methods in Applied Mechanics and Engineering
- Additional Journal Information:
- Journal Volume: 363; Journal Issue: C; Journal ID: ISSN 0045-7825
- Publisher:
- Elsevier
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 42 ENGINEERING; engineering; mathematics; mechanics; recurrent neural networks; lie algebra; anisotropic materials; machine learning; crystal plasticity; special orthogonal group
Citation Formats
Heider, Yousef, Wang, Kun, and Sun, WaiChing. SO(3)-invariance of informed-graph-based deep neural network for anisotropic elastoplastic materials. United States: N. p., 2020.
Web. doi:10.1016/j.cma.2020.112875.
Heider, Yousef, Wang, Kun, & Sun, WaiChing. SO(3)-invariance of informed-graph-based deep neural network for anisotropic elastoplastic materials. United States. https://doi.org/10.1016/j.cma.2020.112875
Heider, Yousef, Wang, Kun, and Sun, WaiChing. Thu .
"SO(3)-invariance of informed-graph-based deep neural network for anisotropic elastoplastic materials". United States. https://doi.org/10.1016/j.cma.2020.112875. https://www.osti.gov/servlets/purl/1801241.
@article{osti_1801241,
title = {SO(3)-invariance of informed-graph-based deep neural network for anisotropic elastoplastic materials},
author = {Heider, Yousef and Wang, Kun and Sun, WaiChing},
abstractNote = {This work examines the frame-invariance (and the lack thereof) exhibited in simulated anisotropic elasto-plastic responses generated from supervised machine learning of classical multi-layer and informed-graph-based neural networks, and proposes different remedies to fix this drawback. The inherent hierarchical relations among physical quantities and state variables in an elasto-plasticity model are first represented as informed, directed graphs, where three variations of the graph are tested. While feed-forward neural networks are used to train path-independent constitutive relations (e.g., elasticity), recurrent neural networks are used to replicate responses that depends on the deformation history, i.e. or path dependent. In dealing with the objectivity deficiency, we use the spectral form to represent tensors and, subsequently, three metrics, the Euclidean distance between the Euler Angles, the distance from the identity matrix, and geodesic on the unit sphere in Lie algebra, can be employed to constitute objective functions for the supervised machine learning. In this, the aim is to minimize the measured distance between the true and the predicted 3D rotation entities. Following this, we conduct numerical experiments on how these metrics, which are theoretically equivalent, may lead to differences in the efficiency of the supervised machine learning as well as the accuracy and robustness of the resultant models. Neural network models trained with tensors represented in component form for a given Cartesian coordinate system are used as a benchmark. Our numerical tests show that, even given the same amount of information and data, the quality of the anisotropic elasto-plasticity model is highly sensitive to the way tensors are represented and measured. The results reveal that using a loss function based on geodesic on the unit sphere in Lie algebra together with an informed, directed graph yield significantly more accurate rotation prediction than the other tested approaches.},
doi = {10.1016/j.cma.2020.112875},
journal = {Computer Methods in Applied Mechanics and Engineering},
number = C,
volume = 363,
place = {United States},
year = {Thu Feb 13 00:00:00 EST 2020},
month = {Thu Feb 13 00:00:00 EST 2020}
}
Web of Science
Works referenced in this record:
A multiscale multi-permeability poroplasticity model linked by recursive homogenizations and deep learning
journal, June 2018
- Wang, Kun; Sun, WaiChing
- Computer Methods in Applied Mechanics and Engineering, Vol. 334
A plasticity theory for porous solids
journal, April 1972
- Green, R. J.
- International Journal of Mechanical Sciences, Vol. 14, Issue 4
Computational thermomechanics of crystalline rock, Part I: A combined multi-phase-field/crystal plasticity approach for single crystal simulations
journal, August 2018
- Na, SeonHong; Sun, WaiChing
- Computer Methods in Applied Mechanics and Engineering, Vol. 338
Some implications of work hardening and ideal plasticity
journal, January 1950
- Drucker, D. C.
- Quarterly of Applied Mathematics, Vol. 7, Issue 4
A Simple Derivation of Representations for Non-Polynomial Constitutive Equations in Some Cases of Anisotropy
journal, January 1979
- Boehler, Jean-Paul
- ZAMM - Zeitschrift für Angewandte Mathematik und Mechanik, Vol. 59, Issue 4
Metrics for 3D Rotations: Comparison and Analysis
journal, June 2009
- Huynh, Du Q.
- Journal of Mathematical Imaging and Vision, Vol. 35, Issue 2
Stress-induced anisotropy in granular materials: fabric, stiffness, and permeability
journal, June 2015
- Kuhn, Matthew R.; Sun, WaiChing; Wang, Qi
- Acta Geotechnica, Vol. 10, Issue 4
Meta-modeling game for deriving theory-consistent, microstructure-based traction–separation laws via deep reinforcement learning
journal, April 2019
- Wang, Kun; Sun, WaiChing
- Computer Methods in Applied Mechanics and Engineering, Vol. 346
Data Structures for Statistical Computing in Python
conference, January 2010
- McKinney, Wes
- Proceedings of the Python in Science Conference
Crystal Plasticity
journal, December 1983
- Asaro, R. J.
- Journal of Applied Mechanics, Vol. 50, Issue 4b
An efficient Monte Carlo strategy for elasto-plastic structures based on recurrent neural networks
journal, July 2019
- Koeppe, Arnd; Bamer, Franz; Markert, Bernd
- Acta Mechanica, Vol. 230, Issue 9
Discrete micromechanics of elastoplastic crystals in the finite deformation range
journal, June 2014
- Borja, Ronaldo I.; Rahmani, Helia
- Computer Methods in Applied Mechanics and Engineering, Vol. 275
Anisotropic Critical State Theory: Role of Fabric
journal, March 2012
- Li, Xiang Song; Dafalias, Yannis F.
- Journal of Engineering Mechanics, Vol. 138, Issue 3
Enhancing Model Predictability for a Scramjet Using Probabilistic Learning on Manifolds
journal, January 2019
- Soize, Christian; Ghanem, Roger; Safta, Cosmin
- AIAA Journal, Vol. 57, Issue 1
A deep material network for multiscale topology learning and accelerated nonlinear modeling of heterogeneous materials
journal, March 2019
- Liu, Zeliang; Wu, C. T.; Koishi, M.
- Computer Methods in Applied Mechanics and Engineering, Vol. 345
Modeling of soil behavior with a recurrent neural network
journal, October 1998
- Zhu, Jian-Hua; Zaman, Musharraf M.; Anderson, Scott A.
- Canadian Geotechnical Journal, Vol. 35, Issue 5
Model-Free Data-Driven inelasticity
journal, June 2019
- Eggersmann, R.; Kirchdoerfer, T.; Reese, S.
- Computer Methods in Applied Mechanics and Engineering, Vol. 350
Reservoir computing approaches to recurrent neural network training
journal, August 2009
- Lukoševičius, Mantas; Jaeger, Herbert
- Computer Science Review, Vol. 3, Issue 3
Artificial neural network as an incremental non-linear constitutive model for a finite element code
journal, July 2003
- Lefik, M.; Schrefler, B. A.
- Computer Methods in Applied Mechanics and Engineering, Vol. 192, Issue 28-30
Adversarial uncertainty quantification in physics-informed neural networks
journal, October 2019
- Yang, Yibo; Perdikaris, Paris
- Journal of Computational Physics, Vol. 394
LSTM: A Search Space Odyssey
journal, October 2017
- Greff, Klaus; Srivastava, Rupesh K.; Koutnik, Jan
- IEEE Transactions on Neural Networks and Learning Systems, Vol. 28, Issue 10
A unified method to predict diffuse and localized instabilities in sands
journal, June 2013
- Sun, WaiChing
- Geomechanics and Geoengineering, Vol. 8, Issue 2
Smooth invariant interpolation of rotations
journal, July 1997
- Park, F. C.; Ravani, Bahram
- ACM Transactions on Graphics, Vol. 16, Issue 3
Computational thermo-hydro-mechanics for multiphase freezing and thawing porous media in the finite deformation range
journal, May 2017
- Na, SeonHong; Sun, WaiChing
- Computer Methods in Applied Mechanics and Engineering, Vol. 318
LIBSVM: A library for support vector machines
journal, April 2011
- Chang, Chih-Chung; Lin, Chih-Jen
- ACM Transactions on Intelligent Systems and Technology, Vol. 2, Issue 3
A micromorphically regularized Cam-clay model for capturing size-dependent anisotropy of geomaterials
journal, September 2019
- Bryant, Eric C.; Sun, WaiChing
- Computer Methods in Applied Mechanics and Engineering, Vol. 354
An updated Lagrangian LBM–DEM–FEM coupling model for dual-permeability fissured porous media with embedded discontinuities
journal, February 2019
- Wang, Kun; Sun, WaiChing
- Computer Methods in Applied Mechanics and Engineering, Vol. 344
Aspects of computational rate-independent crystal plasticity
journal, December 1997
- Schröder, J.; Miehe, C.
- Computational Materials Science, Vol. 9, Issue 1-2
Exploring the 3D architectures of deep material network in data-driven multiscale mechanics
journal, June 2019
- Liu, Zeliang; Wu, C. T.
- Journal of the Mechanics and Physics of Solids, Vol. 127
Neural network based constitutive modeling of nonlinear viscoplastic structural response
journal, January 2019
- Stoffel, Marcus; Bamer, Franz; Markert, Bernd
- Mechanics Research Communications, Vol. 95
A Distance Metric for Finite Sets of Rigid-Body Displacements via the Polar Decomposition
journal, July 2006
- Larochelle, Pierre M.; Murray, Andrew P.; Angeles, Jorge
- Journal of Mechanical Design, Vol. 129, Issue 8
Lie-group interpolation and variational recovery for internal variables
journal, June 2013
- Mota, Alejandro; Sun, WaiChing; Ostien, Jakob T.
- Computational Mechanics, Vol. 52, Issue 6
A graph theoretic framework for representation, exploration and analysis on computed states of physical systems
journal, July 2019
- Banerjee, R.; Sagiyama, K.; Teichert, G. H.
- Computer Methods in Applied Mechanics and Engineering, Vol. 351
Long Short-Term Memory
journal, November 1997
- Hochreiter, Sepp; Schmidhuber, Jürgen
- Neural Computation, Vol. 9, Issue 8
Invariant formulation of hyperelastic transverse isotropy based on polyconvex free energy functions
journal, January 2003
- Schröder, Jörg; Neff, Patrizio
- International Journal of Solids and Structures, Vol. 40, Issue 2
Autoprogressive training of neural network constitutive models
journal, May 1998
- Ghaboussi, Jamshid; Pecknold, David A.; Zhang, Mingfu
- International Journal for Numerical Methods in Engineering, Vol. 42, Issue 1
The computation of the exponential and logarithmic mappings and their first and second linearizations
journal, December 2001
- Ortiz, M.; Radovitzky, R. A.; Repetto, E. A.
- International Journal for Numerical Methods in Engineering, Vol. 52, Issue 12
Knowledge‐Based Modeling of Material Behavior with Neural Networks
journal, January 1991
- Ghaboussi, J.; Garrett, J. H.; Wu, X.
- Journal of Engineering Mechanics, Vol. 117, Issue 1
Implicit constitutive modelling for viscoplasticity using neural networks
journal, September 1998
- Furukawa, Tomonari; Yagawa, Genki
- International Journal for Numerical Methods in Engineering, Vol. 43, Issue 2
Data-driven computational mechanics
journal, June 2016
- Kirchdoerfer, T.; Ortiz, M.
- Computer Methods in Applied Mechanics and Engineering, Vol. 304
Learning to forget: continual prediction with LSTM
conference, January 1999
- Gers, F. A.
- 9th International Conference on Artificial Neural Networks: ICANN '99
A semi-implicit discrete-continuum coupling method for porous media based on the effective stress principle at finite strain
journal, June 2016
- Wang, Kun; Sun, WaiChing
- Computer Methods in Applied Mechanics and Engineering, Vol. 304