Learning Constitutive Relations using Symmetric Positive Definite Neural Networks

We present the Cholesky-factored symmetric positive definite neural network (SPD-NN) for modeling constitutive relations in dynamical equations. Instead of directly predicting the stress, the SPD-NN trains a neural network to predict the Cholesky factor of a tangent stiffness matrix, based on which the stress is calculated in the incremental form. As a result of the special structure, SPD-NN weakly imposes convexity on the strain energy function, satisfies time consistency for path-dependent materials, and therefore improves numerical stability, especially when the SPD-NN is used in finite element simulations. Depending on the types of available data, we propose two training methods, namely direct training for strain and stress pairs and indirect training for loads and displacement pairs. We demonstrate the effectiveness of SPD-NN on hyperelastic, elasto-plastic, and multiscale fiber-reinforced plate problems from solid mechanics. The generality and robustness of the SPD-NN make it a promising tool for a wide range of constitutive modeling applications.

[1]  Wenbin Yu,et al.  A neural network enhanced system for learning nonlinear constitutive relation of fiber reinforced composites , 2020 .

[2]  Kun Wang,et al.  A multiscale multi-permeability poroplasticity model linked by recursive homogenizations and deep learning , 2018, Computer Methods in Applied Mechanics and Engineering.

[3]  Roi Livni,et al.  On the Computational Efficiency of Training Neural Networks , 2014, NIPS.

[4]  M Mozaffar,et al.  Deep learning predicts path-dependent plasticity , 2019, Proceedings of the National Academy of Sciences.

[5]  WaiChing Sun,et al.  SO(3)-invariance of informed-graph-based deep neural network for anisotropic elastoplastic materials , 2020, Computer Methods in Applied Mechanics and Engineering.

[6]  Jintai Chung,et al.  A Time Integration Algorithm for Structural Dynamics With Improved Numerical Dissipation: The Generalized-α Method , 1993 .

[7]  Zeliang Liu,et al.  Exploring the 3D architectures of deep material network in data-driven multiscale mechanics , 2019, Journal of the Mechanics and Physics of Solids.

[8]  Leslie Pack Kaelbling,et al.  Effect of Depth and Width on Local Minima in Deep Learning , 2018, Neural Computation.

[9]  Charbel Farhat,et al.  Predictive Modeling with Learned Constitutive Laws from Indirect Observations , 2019, 1905.12530.

[10]  Rui Zhao,et al.  Stress-Strain Modeling of Sands Using Artificial Neural Networks , 1995 .

[11]  Ted Belytschko,et al.  Continuum Theory for Strain‐Softening , 1984 .

[12]  Julia Ling,et al.  Machine learning strategies for systems with invariance properties , 2016, J. Comput. Phys..

[13]  B. Schrefler,et al.  Multiscale Methods for Composites: A Review , 2009 .

[14]  Jacob Fish,et al.  Toward realization of computational homogenization in practice , 2008 .

[15]  M. Bonnet,et al.  Overview of Identification Methods of Mechanical Parameters Based on Full-field Measurements , 2008 .

[16]  Eric R. Ziegel,et al.  The Elements of Statistical Learning , 2003, Technometrics.

[17]  Razvan Pascanu,et al.  On the difficulty of training recurrent neural networks , 2012, ICML.

[18]  Jorge Nocedal,et al.  A Limited Memory Algorithm for Bound Constrained Optimization , 1995, SIAM J. Sci. Comput..

[19]  Larry R. Oliver,et al.  Finite element analysis of V-ribbed belts using neural network based hyperelastic material model , 2005 .

[20]  P. Withers,et al.  Quantitative X-ray tomography , 2014 .

[21]  F. Ghavamian,et al.  Accelerating multiscale finite element simulations of history-dependent materials using a recurrent neural network , 2019 .

[22]  Grace X. Gu,et al.  Generative Deep Neural Networks for Inverse Materials Design Using Backpropagation and Active Learning , 2020, Advanced science.

[23]  Jamshid Ghaboussi,et al.  Autoprogressive training of neural network constitutive models , 1998 .

[24]  T. Sadowski,et al.  Sensitivity analysis of crack propagation in pavement bituminous layered structures using a hybrid system integrating Artificial Neural Networks and Finite Element Method , 2014 .

[25]  Arnulf Jentzen,et al.  Solving high-dimensional partial differential equations using deep learning , 2017, Proceedings of the National Academy of Sciences.

[26]  James H. Garrett,et al.  Knowledge-Based Modeling of Material Behavior with Neural Networks , 1992 .

[27]  Frederick R. Eirich,et al.  Rheology : theory and applications , 1956 .

[28]  Xia-Ting Feng,et al.  Genetic evolution of nonlinear material constitutive models , 2001 .

[29]  Timon Rabczuk,et al.  Transfer learning enhanced physics informed neural network for phase-field modeling of fracture , 2019, Theoretical and Applied Fracture Mechanics.

[30]  Cv Clemens Verhoosel,et al.  Non-Linear Finite Element Analysis of Solids and Structures , 1991 .

[31]  Z. Hashin.  Analysis of Composite Materials—A Survey , 1983 .

[32]  François Hild,et al.  Identification of elastic parameters by displacement field measurement , 2002 .

[33]  David J. Thuente,et al.  Line search algorithms with guaranteed sufficient decrease , 1994, TOMS.

[34]  Genki Yagawa,et al.  Implicit constitutive modelling for viscoplasticity using neural networks , 1998 .

[35]  Charbel Farhat,et al.  Learning constitutive relations from indirect observations using deep neural networks , 2020, J. Comput. Phys..

[36]  Wei Chen,et al.  A framework for data-driven analysis of materials under uncertainty: Countering the curse of dimensionality , 2017 .

[37]  Wing Kam Liu,et al.  Nonlinear Finite Elements for Continua and Structures , 2000 .

[38]  Guanghui Liang,et al.  Neural network based constitutive model for elastomeric foams , 2008 .

[39]  F. Darve,et al.  Second-order work criterion: from material point to boundary value problems , 2017 .

[40]  Panagiotis G. Asteris,et al.  Anisotropic masonry failure criterion using artificial neural networks , 2017, Neural Computing and Applications.

[41]  C. Farhat,et al.  A multilevel projection‐based model order reduction framework for nonlinear dynamic multiscale problems in structural and solid mechanics , 2017 .

[42]  J. Chaboche,et al.  FE2 multiscale approach for modelling the elastoviscoplastic behaviour of long fibre SiC/Ti composite materials , 2000 .

[43]  B. A. Le,et al.  Computational homogenization of nonlinear elastic materials using neural networks , 2015 .

[44]  Zeliang Liu,et al.  A Deep Material Network for Multiscale Topology Learning and Accelerated Nonlinear Modeling of Heterogeneous Materials , 2018, Computer Methods in Applied Mechanics and Engineering.

[45]  D. Jeulin,et al.  Determination of the size of the representative volume element for random composites: statistical and numerical approach , 2003 .

[46]  Yves Surrel,et al.  Moire and grid methods: a signal-processing approach , 1994, Other Conferences.

[47]  Miguel A. Bessa,et al.  Self-consistent clustering analysis: An efficient multi-scale scheme for inelastic heterogeneous materials , 2016 .

[48]  C. Sun,et al.  Prediction of composite properties from a representative volume element , 1996 .

[49]  Stéphane Avril,et al.  The Virtual Fields Method for Extracting Constitutive Parameters From Full‐Field Measurements: a Review , 2006 .

[50]  Rui Xu,et al.  Structural-Genome-Driven computing for composite structures , 2019, Composite Structures.

[51]  J. Pascon.  Large deformation analysis of plane-stress hyperelastic problems via triangular membrane finite elements , 2019, International Journal of Advanced Structural Engineering.