Learning macroscopic internal variables and history dependence from microscopic models
Source abstract
Abstract
This paper concerns the study of history dependent phenomena in heterogeneous materials in a two-scale setting where the material is specified at a fine microscopic scale of heterogeneities that is much smaller than the coarse macroscopic scale of application. We specifically study a polycrystalline medium where each grain is governed by crystal plasticity while the solid is subjected to macroscopic dynamic loads. The theory of homogenization allows us to solve the macroscale problem directly with a constitutive relation that is defined implicitly by the solution of the microscale problem. However, the homogenization leads to a highly complex history dependence at the macroscale, one that can be quite different from that at the microscale. In this paper, we examine the use of machine-learning, and especially deep neural networks, to harness data generated by repeatedly solving the finer scale model to: (i) gain insights into the history dependence and the macroscopic internal variables that govern the overall response; and (ii) to create a computationally efficient surrogate of its solution operator, that can directly be used at the coarser scale with no further modeling. We do so by introducing a recurrent neural operator (RNO), and show that: (i) the architecture and the learned internal variables can provide insight into the physics of the macroscopic problem; and (ii) that the RNO can provide multiscale, specifically FE2, accuracy at a cost comparable to a conventional empirical constitutive relation.
Source: openaire · Quality: candidate unverified · Source record ↗
Research context
OpenAlex topics: Composite Material Mechanics · Advanced Mathematical Modeling in Engineering · Machine Learning in Materials Science
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (38)
Show source-provided references
1. Anon · 2014 · 281
2. A mechanics?informed artificial neural network approach in data?driven constitutive modeling
3. Bensoussan · 2011
4. Learning markovian homogenized models in viscoelasticity10.1137/22M1499200 ↗
5. On the importance of self-consistency in recurrent neural network models representing elasto-plastic solids10.1016/j.jmps.2021.104697 ↗
6. On the importance of self-consistency in recurrent neural network models representing elasto-plastic solids10.1016/j.jmps.2021.104697 ↗
7. Overall response of viscoelastic composites and polycrystals: Exact asymptotic relations and approximate estimates10.1016/j.ijsolstr.2013.02.011 ↗
8. A variational constitutive model for slip-twinning interactions in HCP metals: Application to single- and polycrystalline magnesium10.1016/j.ijplas.2015.03.008 ↗
9. Gradient descent finds global minima of deep neural networks
10. Homogenization and mechanical dissipation in thermoviscoelasticity10.1007/BF00251909 ↗
11. Periodic three-dimensional mesh generation for crystalline aggregates based on Voronoi tessellations10.1007/s00466-008-0339-2 ↗
12. Accelerating multiscale finite element simulations of history-dependent materials using a recurrent neural network10.1016/j.cma.2019.112594 ↗
13. Gurtin · 2010
14. Surrogate modeling of elasto-plastic problems via long short-term memory neural networks and proper orthogonal decomposition10.1016/j.cma.2021.114030 ↗
15. Johnson, G., Cook, W., 1983. A constitutive model and data for metals subjected to large strains high strain rates and high temperatures. In: Proceedings of the Seventh International Symposium on Ballistics. pp. 541–548.
16. Deep learning without poor local minima
17. Kingma · 2014
18. Self-normalizing neural networks
19. Kocks · 2000
20. Deep learning10.1038/nature14539 ↗
21. Artificial neural networks in numerical modelling of composites10.1016/j.cma.2008.12.036 ↗
22. Li, Z., Kovachki, N., Azizzadenesheli, K., Liu, B., Bhattacharya, K., Stuart, A.M., Anandkumar, A., 2020. Multipole graph neural operator for parametric partial differential equations. In: Proceedings of the 34th Conference on Neural Information Processing Systems. NeurIPS-20, pp. 1–17.
23. Li, Z., Kovachki, N., Azizzadenesheli, K., Liu, B., Bhattacharya, K., Stuart, A.M., Anandkumar, A., 2021. Fourier neural operator for parametric partial differential equations. In: International Conference on Learning Representations. pp. 1–16.
24. Visualizing the loss landscape of neural nets
25. A learning-based multiscale method and its application to inelastic impact problems10.1016/j.jmps.2021.104668 ↗
26. Smart constitutive laws: Inelastic homogenization through machine learning10.1016/j.cma.2020.113482 ↗
27. Thermodynamics-based artificial neural networks for constitutive modeling10.1016/j.jmps.2020.104277 ↗
28. Deep learning predicts path-dependent plasticity10.1073/pnas.1911815116 ↗
29. Pavliotis · 2008
30. Phillips · 2001
31. Inelastic constitutive relations for solids: An internal-variable theory and its application to metal plasticity10.1016/0022-5096(71)90010-X ↗
32. Trainability and accuracy of artificial neural networks: An interacting particle system approach10.1002/cpa.22074 ↗
33. Spurious local minima are common in two-layer Relu neural networks
34. Sur certains problèmes physiques d’homogénéisation donnant lieu à des phénomènes de relaxation
35. Roadmap on multiscale materials modeling10.1088/1361-651X/ab7150 ↗
36. A recurrent neural network-accelerated multi-scale model for elasto-plastic heterogeneous materials subjected to random cyclic and non-proportional loading paths10.1016/j.cma.2020.113234 ↗
37. Recurrent neural networks (RNNs) with dimensionality reduction and break down in computational mechanics; Application to multi-scale localization step10.1016/j.cma.2021.114476 ↗
38. Accelerated computational micromechanics and its application to polydomain liquid crystal elastomers10.1016/j.jmps.2021.104470 ↗