A Neural Network Constitutive Model and Automatic Stiffness Evaluation for Multiscale Finite Elements
Source abstract
Abstract
A neural network model for a constitutive law in nonlinear structures is proposed in this paper. The artificial neural network (ANN) model is constructed based on a data set of responses from representative volume elements, which was calculated by finite elements and using an open scientific software machine learning platform. The tangential stiffness matrix, which can be used within a multiscale finite element analysis, is calculated via the method of automatic differentiation. Two types of constitutive neural networks are proposed. The first approach involves training a residual model with three respective surrogate models for the stress components, ensuring less computing cost. The second approach considers a separate ANN for each stress component, ensuring a high rate of convergence. The numerical results are compared with the given data set as well as with the results obtained after applying a polynomial regression. The loss function, without and including the Sobolev metrics, is considered. In addition, a physics-informed constitutive neural model, which enforces hyperelasticity principles, is also analyzed. The choice of the hyperparameters is discussed.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Advanced machining processes and optimization · Material Properties and Failure Mechanisms · Topology Optimization in Engineering
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
Institute of Computational Mechanics and Optimization, School of Production Engineering and Management, Technical University of Crete, 731 00 Chania, Greece
Funding
Project Safe-Aorta · European Union-NextGenerationEU
References (42)
Show source-provided references
1. Drosopoulos, G.A., and Stavroulakis, G.E. (2022). Non-Linear Mechanics for Composite, Heterogeneous Structures, CRC Press, Taylor and Francis.10.1201/9781003017240 ↗
2. Urbański, A. (2005). The Unified, Finite Element Formulation of Homogenization of Structural Members with a Periodic Microstructure, Cracow University of Technology.
3. Multi-scale computational homogenization: Trends and challenges10.1016/j.cam.2009.08.077 ↗
4. Neural networks for computing in structural analysis: Methods and prospects of applications10.1002/nme.1620361310 ↗
5. Neural networks for computing in the elastoplastic analysis of structures10.1007/BF00987122 ↗
6. The projective-iterative method and neural network estimation for buckling of elastic plates in nonlinear theory10.1016/j.cnsns.2005.09.001 ↗
7. Stavroulakis, G.E. (2000). Inverse and Identification Problems in Mechanics, Springer/Kluwer Academic.10.1007/978-1-4615-0019-3 ↗
8. Yagawa, G., and Oishi, A. (2021). Computational Mechanics with Neural Networks, Springer.10.1007/978-3-030-66111-3 ↗
9. Neural Networks for Constitutive Modeling: From Universal Function Approximators to Advanced Models and the Integration of Physics10.1007/s11831-023-10009-y ↗
10. Constitutive artificial neural networks: A fast and general approach to predictive data-driven constitutive modeling by deep learning10.1016/j.jcp.2020.110010 ↗
11. A new family of Constitutive Artificial Neural Networks towards automated model discovery10.1016/j.cma.2022.115731 ↗
12. Czarnecki, W.M., Osindero, S., Swirszcz, M.J.G., and Pascanu, R. (2017). Sobolev Training for Neural Networks. arXiv.
13. Nicosia, G., Ojha, V., La Malfa, E., Jansen, G., Sciacca, V., Pardalos, P., Giuffrida, G., and Umeton, R. (2020). Global Convergence of Sobolev Training for Overparameterized Neural Networks. Machine Learning, Optimization, and Data Science, Springer. Lecture Notes in Computer Science, LOD 2020.10.1007/978-3-030-64583-0 ↗
14. Sciann, A keras/tensorflow wrapper for scientific computations and physics-informed deep learning using artificial neural networks10.1016/j.cma.2020.113552 ↗
15. Baydin, A., Pearlmutter, B.A., Radul, A.A., and Siskind, J.M. (2018). Automatic Differentiation in Machine Learning: A Survey. arXiv.
16. Effective properties of composite materials with periodic microstructure: A computational approach10.1016/S0045-7825(98)00227-8 ↗
17. Zohdi, T.I., and Wriggers, P. (2008). An Introduction to Computational Micromechanics, Springer.
18. Non-linear FE2 multiscale simulation of damage, micro and macroscopic strains in polyamide 66-woven composite structures: Analysis and experimental validation10.1016/j.compstruct.2020.112926 ↗
19. Metamodeling-assisted numerical homogenization for masonry and cracked structures10.1061/(ASCE)EM.1943-7889.0001500 ↗
20. Reducing internal variables and improving efficiency in data-driven modelling of anisotropic damage from RVE simulations10.1007/s00466-023-02326-7 ↗
21. Computational homogenization of nonlinear elastic materials using neural networks10.1002/nme.4953 ↗
22. Multi-scale modeling of brick masonry using a numerical homogenization technique and an artificial neural network
23. Eivazi, H., Tröger, J.-A., Wittek, S., Hartmann, S., and Rausch, A. (2023, June 07). FE² Computations with Deep Neural Networks: Algorithmic Structure, Data Generation, and Implementation. Available online: https://ssrn.com/abstract=4485434.10.2139/ssrn.4485434 ↗
24. Data-physics driven reduced order homogenization10.1002/nme.7178 ↗
25. A mechanics-informed artificial neural network approach in data-driven constitutive modeling10.1002/nme.6957 ↗
26. A genetically optimized neural classifier applied to numerical pile integrity tests considering concrete piles10.1016/j.compstruc.2015.08.005 ↗
27. Physics-informed neural networks for elastic plate problems with bending and Winkler-type contact effects10.24874/jsscm.2021.15.02.05 ↗
28. Ensemble of physics-informed neural networks for solving plane elasticity problems with examples10.1007/s00707-024-04053-3 ↗
29. Physics-informed machine learning10.1038/s42254-021-00314-5 ↗
30. A Gentle Introduction to Physics-Informed Neural Networks, with Applications in Static Rod and Beam Problems10.15377/2409-5761.2022.09.8 ↗
31. A data-driven computational homogenization method based on neural networks for the nonlinear anisotropic electrical response of graphene/polymer nanocomposites10.1007/s00466-018-1643-0 ↗
32. Mesoscopic and multiscale modelling in materials10.1038/s41563-020-00913-0 ↗
33. An ABAQUS toolbox for multiscale finite element computation10.1016/j.compositesb.2013.04.028 ↗
34. LS-DYNA Machine Learning–Based Multiscale Method for Nonlinear Modeling of Short Fiber–Reinforced Composites10.1061/JENMDT.EMENG-6945 ↗
35. Multiscale computational solid mechanics: Data and machine learning10.1093/jom/ufac037 ↗
36. Learning composite constitutive laws via coupling Abaqus and deep neural network10.1016/j.compstruct.2021.114137 ↗
37. Stavroulakis, G.E., Giannis, K., Drosopoulos, G.A., and Stavroulaki, M.E. (2013, January 23–25). Non-linear Computational Homogenization Experiments. Proceedings of the COMSOL Conference, Rotterdam, The Netherlands. Available online: http://purl.tuc.gr/dl/dias/9D3019B4-7879-4E1F-9845-E53693BB1717.
38. Approximation and Estimation Bounds for Artificial Neural Networks10.1007/BF00993164 ↗
39. Ruder, S. (2017). An overview of gradient descent optimization algorithms. arXiv.
40. Abadi, M., Barham, P., Chen, J., Chen, Z., Davis, A., Dean, J., Devin, M., Ghemawat, S., Irving, G., and Isard, M. (2016, January 2–4). Tensorflow: A system for large-scale machine learning. Proceedings of the 12th USENIX Symposium on Operating Systems Design and Implementation (OSDI 16), USENIX, Savannah, GA, USA. Available online: https://www.usenix.org/conference/osdi16/technical-sessions/presentation/abadi.
41. Novel Nonlinear Elasticity Approach for Analysis of Nonlinear and Hyperelastic Structures10.1016/j.enganabound.2022.06.015 ↗
42. Comparison of hyperelastic models for rubber-like materials10.5254/1.3547969 ↗