As part of a multi scale modeling effort, I discuss progress on a challenge in continuum-scale modeling: the direct incorporation of complex molecular level processes in the constitutive evaluation.In this initial phase of the research we use a concurrent scale-bridging approach, with a hierarchical multi scale frame work running in parallel to couple a particle-based model (the “lower scale”) to the constitutive response in a finite-element multi-physics simulation (the “upper scale”). In this approach, many orders of magnitude in length scale separate the lower and upper scales, and the lower scale is able to be used in the constitutive model for all elements of the upper scale simulations. Molecular level response provided by the lower scale includes the constitutive equation of state and non-equilibrium chemistry. Response dependent upon lower scale stochastic micro structure and challenges for scale-bridging in time are also discussed. The lower scale simulations of hexahydro-1,3,5-trinitro-s-triazine (RDX)use a force-matched coarse-grain model and dissipative particle dynamics methods, and the upper scale simulates Taylor anvil and plate impact experiments. Results emphasize use of machine learning (via Gaussian process regression, or “kriging”) that accelerates time to solution, and its comparison to fully on-the-flyruns.
