An FE-DMN method for the multiscale analysis of fiber reinforced plastic components

Abstract

In this work, we propose a fully coupled multiscale strategy for components made from short fiber reinforced composites, where each Gauss point of the macroscopic finite element model is equipped with a deep material network (DMN) which covers the different fiber orientation states varying within the component. These DMNs need to be identified by linear elastic precomputations on representative volume elements, and serve as high-fidelity surrogates for full-field simulations on microstructures with inelastic constituents. We discuss how to extend direct DMNs to account for varying fiber orientation, and propose a simplified sampling strategy which significantly speeds up the training process. To enable concurrent multiscale simulations, evaluating the DMNs efficiently is crucial. We discuss dedicated techniques for exploiting sparsity and high-performance linear algebra modules, and demonstrate the power of the proposed approach on an industrial-scale three-dimensional component. Indeed, the DMN is capable of accelerating two-scale simulations significantly, providing possible speed-ups of several magnitudes.

arXiv commentSubmitted to "Computer Methods in Applied Mechanics and Engineering"

References in corpus (13)

• Adam: A Method for Stochastic Optimization
• On the Convergence of Adam and Beyond
• A numerical method for computing the overall response of nonlinear composites with complex microstructure
• A deep material network for multiscale topology learning and accelerated nonlinear modeling of heterogeneous materials
• Fourier-based schemes for computing the mechanical response of composites with accurate local fields
• Accelerating a FFT-based solver for numerical homogenization of periodic media by conjugate gradients
• Exploring the 3D architectures of deep material network in data-driven multiscale mechanics
• Learning Constitutive Relations using Symmetric Positive Definite Neural Networks
• Fourier-based schemes with modified Green operator for computing the electrical response of heterogeneous media with accurate local fields
• Exact tensor closures for the three dimensional Jeffery's equation
• Deep material network with cohesive layers: Multi-stage training and interfacial failure analysis
• On-the-fly adaptivity for nonlinear twoscale simulations using artificial neural networks and reduced order modeling
• Intelligent multiscale simulation based on process-guided composite database

Cited by in corpus (4)

• LS-DYNA Machine Learning-based Multiscale Method for Nonlinear Modeling of Short Fiber-Reinforced Composites
• A probabilistic virtual process chain to quantify process-induced uncertainties in Sheet Molding Compounds
• Micromechanics-Informed Parametric Deep Material Network for Physics Behavior Prediction of Heterogeneous Materials with a Varying Morphology
• Computational homogenization for aerogel-like polydisperse open-porous materials using neural network--based surrogate models on the microscale