Abstract:Modeling nonlinear materials with arbitrary microstructures and loading paths is crucial in structural analyses with heterogeneous materials with uncertainty. However, it is computationally prohibitive because (1) fine meshes are required to resolve the microstructures for nonlinear simulations, (2) enormous model evaluations are expected to simulate stochastic microstructures and the loading variabilities, and (3) nonlinear material models are generally expensive to solve. In this paper, we present the microstructure-guided deep material network (MgDMN), a deep-learning-based approach to model multi-phase materials and efficiently predict their mechanical response under virtually any loading paths and microstructure definitions considering plasticity. This work is built upon the earlier development of the deep material network (DMN) method for fast predictions of nonlinear homogenized responses of multi-phase materials using linear elastic training data for a predefined microstructure. We extend DMN to MgDMN for any microstructure definition by parameterizing the microstructure space, mapping a small set of DMNs to their respective microstructure parameters, and interpolating the DMNs to generate new models in the microstructure space. The proposed method achieves superb efficiency and accuracy since the interpolated DMNs do not need additional training to predict nonlinear behavior of any new microstructures. We demonstrate our approach via modeling and uncertainty quantification of short fiber reinforced polymer (SFRP) composites. We show that for this dual-phase material, only four base DMNs are needed to predict the nonlinear stress responses for any loading path given any fiber volume fraction, orientation states, and constituent properties as long as the combination is within the microstructure space enclosed by the base models. The accelerated modeling process enables complex structural simulations of heterogeneous materials and nonlinear stochastic analyses with uncertain microstructures, which are essential in computational materials engineering and multiscale modeling. Our research also draws insights into the impact of microstructure uncertainties on linear and nonlinear material behaviors.

Graph-enhanced deep material network: multiscale materials modeling with microstructural informatics

Microstructure-guided deep material network for rapid nonlinear material modeling and uncertainty quantification

Exploring the 3D architectures of deep material network in data-driven multiscale mechanics

A deep material network for multiscale topology learning and accelerated nonlinear modeling of heterogeneous materials

A Microstructure-based Graph Neural Network for Accelerating Multiscale Simulations

Graph Neural Networks for an Accurate and Interpretable Prediction of the Properties of Polycrystalline Materials

On the micromechanics of deep material networks

Micromechanics-informed parametric deep material network for physics behavior prediction of heterogeneous materials with a varying morphology

Mesh-based graph convolutional neural networks for modeling materials with microstructure

Deep learning and multi-level featurization of graph representations of microstructural data

Transfer learning of deep material network for seamless structure–property predictions

Microstructure Representation and Reconstruction of Heterogeneous Materials Via Deep Belief Network for Computational Material Design

Multiscale modeling of woven composites by deep learning neural networks and its application in design optimization

A Data-Driven Framework for Designing Microstructure of Multifunctional Composites with Deep-Learned Diffusion-Based Generative Models

Statistical characterization and reconstruction of heterogeneous microstructures using deep neural network

Self-supervised Representations and Node Embedding Graph Neural Networks for Accurate and Multi-scale Analysis of Materials

Microstructural Materials Design Via Deep Adversarial Learning Methodology

LS-DYNA Machine Learning-based Multiscale Method for Nonlinear Modeling of Short Fiber-Reinforced Composites

Accelerated multiscale mechanics modeling in a deep learning framework

A machine-learning-enabled approach for bridging multiscale simulations of CNTs/PDMS composites