Project Overview

Objective

The overall objective of the Center is to simulate with quantified uncertainty, from grain-to-continuum-length-scales, a class of problems involving large deformations, fracture, and fragmentation of unbonded and plastically-bonded particulate materials. The overarching problem is to predict with quantified uncertainty the processing and mechanical behavior of pressed mock High Explosive (HE) material subjected to quasi-static and high-strain-rate confined and unconfined compression with in-situ X-ray computed tomography (CT) and Digital Image Correlation (DIC).

../_images/length_time_plot_large_text_DNS_micro.png

Description

To accomplish the objective, a micromorphic multiscale computational framework is being developed, verified, and validated with quantified uncertainty, and executed on Exascale computing platforms. Machine Learning (ML) algorithms are being applied to fill the gaps in multiscale constitutive modeling via coordinated grain-scale experiments and Direct Numerical Simulations (DNS). An extensive, integrated, experimental program at quasi-static, dynamic, and high-strain rates, ranging from grain-to-specimen-length-scales, is being conducted to validate heterogeneous grain-to-continuum-length-scale computational models, calibrate model parameters, and validate the overall computational framework. Exascale computing is needed to simulate these more sophisticated micromorphic bridged-DNS simulations, with offline ML training of micromorphic constitutive relations to DNS. Furthermore, for Validation and Uncertainty Quantification (UQ) requiring multiple instances of these simulations over statistical distributions of inputs (such as grain-size distribution), with high and low fidelity, Exascale computing is a necessity.

Anticipated Outcomes and Benefits of Research

The Center research will attempt to develop higher fidelity multiscale computation through large deformation micromorphic continuum field theories informed by DNS through the latest ML techniques calibrated and validated against a rich experimental data set. Applying advances in V&V/UQ, Exascale computing, and Integration/Workflows will make the applicability of such an approach to reduce uncertainty in continuum-scale computations based on statistical distributions of materials information at the grain-scale of plastically-bonded particulate materials a reality, which has significant influence on the success of the Stockpile Stewardship Program (e.g., High Explosive (HE) materials) and beyond.