Model Development Using Accelerated Simulations of Hypersonic Flow Features

This code enables more accurate models for computational fluid dynamics applications.

Currently, the two main computational tools used by the aerothermodynamics community to model hypersonic flows are Computational Fluid Dynamics (CFD), and the direct simulation Monte Carlo (DSMC) particle method. Both use essentially the same physical models for rotational-vibrational excitation and dissociation phenomenon, which are based on a limited number of near-equilibrium experiments performed at low temperatures.

A three-dimensional solution for Hypersonic Flow of argon over a capsule geometry. This simulation required only 128 cores for 12 hours using the MGDS code.

The purpose of this work was to develop a new modeling capability, based on computational chemistry, to provide a more fundamental understanding and develop more accurate thermochemical models for CFD and DSMC. A parallel DSMC code called the Molecular Gas Dynamic Simulator (MGDS) code, was developed that uses an embedded three-level Cartesian flow grid with automated adaptive mesh refinement (AMR). The refinement is arbitrary (non-binary) and enables accurate and efficient simulations with little user input. In addition, MGDS contains a robust “cut-cell” subroutine that cuts complex 3D geometry from the background Cartesian grid and exactly computes the volumes of all cut cells. Combined within a DSMC solver, these two features enable molecular-level physics to be applied to real engineering problems.

In addition to the practicality for complex 3D flows, the DSMC code acts as a bridge between computational chemistry modeling and continuum fluid mechanics. With existing parallel computer clusters, DSMC is able to perform near-continuum simulations using only molecular physics models. Pure computational chemistry of simulation of shock waves and shock layers is computationally feasible with modest parallel computing resources (100 core processors for a few days). Such simulations, which are referred to as “all-atom molecular dynamics” (MD) simulations, require only a potential energy surface (PES) that dictates the interaction forces between individual atoms as the model input. Thus, unlike DSMC or CFD, no models for viscosity, diffusion, thermal conductivity, internal energy relaxation, chemical reactions, cross-sections, or state-resolved probabilities/ rates are required. Rather, every real atom in the system is simulated.

The methodology was validated with experimental data for shock structure in mixtures of noble gases and diatomic nitrogen. All-atom MD simulations of nitrogen discovered clear “direction-dependence” of translational-rotational energy transfer that is not captured by existing models. Compressing flows involve fast rotational excitation, whereas expanding flows involve slower relaxation for the same equilibrium temperature. A new model for both DSMC and CFD reproduces experimental data and is consistent among MD, DSMC, and CFD.

A combined MD-DSMC technique replaces the collision model in DSMC with MD trajectories. The method reproduces exactly pure MD results. This is a significant advancement that enables simulation of axisymmetric and 3D flows where a potential energy surface is the only model input. Thus, the accuracy of pure MD is maintained for full flow fields, directly linking computational chemistry with aerothermodynamics.

This work was done by Thomas E. Schwartzentruber of the University of Minnesota for the Air Force Office of Scientific Research. AFRL-0226

This Brief includes a Technical Support Package (TSP).
Model Development Using Accelerated Simulations of Hypersonic Flow Features

(reference AFRL-0226) is currently available for download from the TSP library.

Don't have an account? Sign up here.