AutoMat: automatic differentiation for generalized standard materials on GPUs

We propose a universal method for the evaluation of generalized standard materials that greatly simplifies the material law implementation process. By means of automatic differentiation and a numerical integration scheme, AutoMat reduces the implementation effort to two potential functions. By moving AutoMat to the GPU, we close the performance gap to conventional evaluation routines and demonstrate in detail that the expression level reverse mode of automatic differentiation as well as its extension to second order derivatives can be applied inside CUDA kernels. We underline the effectiveness and the applicability of AutoMat by integrating it into the FFT-based homogenization scheme of Moulinec and Suquet and discuss the benefits of using AutoMat with respect to runtime and solution accuracy for an elasto-viscoplastic example.

[1]  Ernst Hairer,et al.  Solving Ordinary Differential Equations I: Nonstiff Problems , 2009 .

[2]  Andrea Walther,et al.  Getting Started with ADOL-C , 2009, Combinatorial Scientific Computing.

[3]  M. Diehl,et al.  A spectral method solution to crystal elasto-viscoplasticity at finite strains , 2013 .

[4]  E. Hairer,et al.  Solving Ordinary Differential Equations II: Stiff and Differential-Algebraic Problems , 2010 .

[5]  Takahiro Katagiri,et al.  Implementation and Evaluation of 3D Finite Element Method Application for CUDA , 2012, VECPAR.

[6]  Ralph Müller,et al.  A scalable multi‐level preconditioner for matrix‐free µ‐finite element analysis of human bone structures , 2008 .

[7]  Adrian Sandu,et al.  On the discrete adjoints of adaptive time stepping algorithms , 2009, J. Comput. Appl. Math..

[8]  L. Gélébart,et al.  Non-linear extension of FFT-based methods accelerated by conjugate gradients to evaluate the mechanical behavior of composite materials , 2013 .

[9]  P. Y. Wu.  Products of positive semidefinite matrices , 1988 .

[10]  M. Schneider,et al.  Computational homogenization of elasticity on a staggered grid , 2016 .

[11]  W. Voigt,et al.  Lehrbuch der Kristallphysik , 1966 .

[12]  Lars Ruthotto,et al.  Layer-Parallel Training of Deep Residual Neural Networks , 2018, SIAM J. Math. Data Sci..

[13]  Ralph Müller,et al.  A scalable multi‐level preconditioner for matrix‐free µ‐finite element analysis of human bone structures , 2008 .

[14]  M. Schneider.  On the Barzilai‐Borwein basic scheme in FFT‐based computational homogenization , 2019, International Journal for Numerical Methods in Engineering.

[15]  Sachin S. Gautam,et al.  GPU-warp based finite element matrices generation and assembly using coloring method , 2019, J. Comput. Des. Eng..

[16]  I. Cormeau,et al.  Numerical stability in quasi‐static elasto/visco‐plasticity , 1975 .

[17]  U. Naumann,et al.  Meta Adjoint Programming in C + + , 2017 .

[18]  F. Krogh,et al.  Solving Ordinary Differential Equations , 2019, Programming for Computations - Python.

[19]  C. Sauder,et al.  Analysis of the damage initiation in a SiC/SiC composite tube from a direct comparison between large-scale numerical simulation and synchrotron X-ray micro-computed tomography , 2019, International Journal of Solids and Structures.

[20]  Jan Novák,et al.  Accelerating a FFT-based solver for numerical homogenization of periodic media by conjugate gradients , 2010, J. Comput. Phys..

[21]  P. Alam.  ‘G’ , 2021, Composites Engineering: An A–Z Guide.

[22]  W. Marsden.  I and J , 2012 .

[23]  M. Schneider,et al.  The composite voxel technique for inelastic problems , 2017 .

[24]  Andrea Walther,et al.  Automatic differentiation of explicit Runge-Kutta methods for optimal control , 2007, Comput. Optim. Appl..

[25]  Martin Diehl,et al.  Numerically robust spectral methods for crystal plasticity simulations of heterogeneous materials , 2013 .

[26]  Andrea Walther,et al.  Automatic Differentiation of an Entire Design Chain for Aerodynamic Shape Optimization , 2007 .

[27]  Adrian Sandu,et al.  Reverse Automatic Differentiation of Linear Multistep Methods , 2008 .

[28]  M. Schneider,et al.  On Quasi‐Newton methods in fast Fourier transform‐based micromechanics , 2019, International Journal for Numerical Methods in Engineering.

[29]  V. K. Arya,et al.  On the numerical integration of viscoplastic models , 1986 .

[30]  Krzysztof Banas,et al.  3D finite element numerical integration on GPUs , 2010, ICCS.

[31]  Quoc Son Nguyen,et al.  Sur les matériaux standard généralisés , 1975 .

[32]  Hervé Moulinec,et al.  A numerical method for computing the overall response of nonlinear composites with complex microstructure , 1998, ArXiv.

[33]  Thomas Böhlke,et al.  Reduced basis homogenization of viscoelastic composites , 2013 .

[34]  M. Schneider.  A dynamical view of nonlinear conjugate gradient methods with applications to FFT-based computational micromechanics , 2020, Computational Mechanics.

[35]  Andrea Walther,et al.  Differentiating Fixed Point Iterations with ADOL-C: Gradient Calculation for Fluid Dynamics , 2006, HPSC.

[36]  Adrian Sandu,et al.  Forward and adjoint sensitivity analysis with continuous explicit Runge-Kutta schemes , 2009, Appl. Math. Comput..

[37]  J. C. Simo,et al.  Consistent tangent operators for rate-independent elastoplasticity☆ , 1985 .

[38]  Adrian Sandu,et al.  Forward, Tangent Linear, and Adjoint Runge-Kutta Methods in KPP-2.2 , 2006, International Conference on Computational Science.

[39]  Yang Chen,et al.  A FFT solver for variational phase-field modeling of brittle fracture , 2019, Computer Methods in Applied Mechanics and Engineering.

[40]  L. Shampine,et al.  A 3(2) pair of Runge - Kutta formulas , 1989 .

[41]  Adrian Sandu,et al.  On the Properties of Runge-Kutta Discrete Adjoints , 2006, International Conference on Computational Science.

[42]  Nicolas R. Gauger,et al.  High-Performance Derivative Computations using CoDiPack , 2017, ACM Trans. Math. Softw..

[43]  Didier Auroux,et al.  Optimal parameters identification and sensitivity study for abrasive waterjet milling model , 2016, 1605.08583.

[44]  Cyril Flaig,et al.  Bone structure analysis on multiple GPGPUs , 2014, J. Parallel Distributed Comput..

[45]  M. Schneider,et al.  Mixed boundary conditions for FFT-based homogenization at finite strains , 2016 .

[46]  Nicolas R. Gauger,et al.  Expression templates for primal value taping in the reverse mode of algorithmic differentiation , 2018, Optim. Methods Softw..

[47]  Mohamed Rouainia,et al.  Explicit Runge–Kutta methods for the integration of rate-type constitutive equations , 2008 .

[48]  Mathias Wilke,et al.  Gewöhnliche Differentialgleichungen und dynamische Systeme , 2011 .

[49]  M. Schneider.  Convergence of FFT‐based homogenization for strongly heterogeneous media , 2015 .

[50]  S. Hartmann,et al.  Automatic differentiation for stress and consistent tangent computation , 2014, Archive of Applied Mechanics.

[51]  Robin J. Hogan,et al.  Fast Reverse-Mode Automatic Differentiation using Expression Templates in C++ , 2014, ACM Trans. Math. Softw..

[52]  Luc Dormieux,et al.  FFT-based methods for the mechanics of composites: A general variational framework , 2010 .

[53]  E. Hairer,et al.  Solving Ordinary Differential Equations II: Stiff and Differential-Algebraic Problems , 2010 .

[54]  L. Dormieux,et al.  Combining Galerkin approximation techniques with the principle of Hashin and Shtrikman to derive a new FFT-based numerical method for the homogenization of composites , 2012 .

[55]  Roger P. Pawlowski,et al.  Efficient Expression Templates for Operator Overloading-based Automatic Differentiation , 2012, ArXiv.

[56]  C. Farhat.  International Journal for Numerical Methods in Engineering , 2019 .

[57]  F. Willot,et al.  Fourier-based schemes for computing the mechanical response of composites with accurate local fields , 2014, 1412.8398.

[58]  Javier Segurado,et al.  On the accuracy of spectral solvers for micromechanics based fatigue modeling , 2018, Computational Mechanics.

[59]  M. Kabel,et al.  Runtime optimization of a memory efficient CG solver for FFT-based homogenization: implementation details and scaling results for linear elasticity , 2019, Computational Mechanics.

[60]  Andreas Griewank,et al.  Evaluating derivatives - principles and techniques of algorithmic differentiation, Second Edition , 2000, Frontiers in applied mathematics.

[61]  M. Schneider,et al.  Efficient fixed point and Newton–Krylov solvers for FFT-based homogenization of elasticity at large deformations , 2014 .

[62]  Felix Fritzen,et al.  Fourier-Accelerated Nodal Solvers (FANS) for homogenization problems , 2018 .

[63]  H. Moulinec,et al.  A fast numerical method for computing the linear and nonlinear mechanical properties of composites , 1994 .

[64]  M. Schneider,et al.  FFT‐based homogenization for microstructures discretized by linear hexahedral elements , 2017 .

[65]  Pierre Suquet,et al.  A model-reduction approach in micromechanics of materials preserving the variational structure of constitutive relations , 2016 .

[66]  Lawrence F. Shampine,et al.  The MATLAB ODE Suite , 1997, SIAM J. Sci. Comput..

[67]  Peter Eberhard,et al.  Automatic differentiation of numerical integration algorithms , 1999, Math. Comput..

[68]  Jean Utke,et al.  OpenAD/F: A Modular Open-Source Tool for Automatic Differentiation of Fortran Codes , 2008, TOMS.

[69]  R. Ma,et al.  FFT-based homogenization of hypoelastic plasticity at finite strains , 2019, Computer Methods in Applied Mechanics and Engineering.

[70]  Hervé Moulinec,et al.  A computational scheme for linear and non‐linear composites with arbitrary phase contrast , 2001 .