Optimized Numerics for Missile Aero-Propulsive Flow Modeling on Massive Clustered Computational Resources

High-fidelity flowfield simulations of US Army interest for tactical and hypersonic missile aeropropulsive applications require replacement of simplified modeling approximations with more accurate but complex formulations. These improvements have incurred significant computational cost through use of higher-order numerics, dense computational meshes, and advanced turbulent and thermochemistry models with disparate time-scales that introduce additional non-linear transport equations and numerical stiffness issues. Software-based optimizations are needed to improve simulation throughput for system design parametric studies and evaluation purposes. Use of massively parallel computer clusters can extend simulation capability to 3D flowfields but alone cannot address the performance issue; innovative algorithm improvements are needed to complement available hardware resources. The opportunity presented seeks to boost simulation output significantly, without compromising accuracy, using automated dynamic load balancing techniques for parallel systems that compensate for non-uniform computational work distributions from varying physical processes (chemistry, particulate interactions) across the domain. Additionally, improved convergence algorithms are proposed to facilitate use of advanced thermochemistry and turbulence models in an efficient manner and to exploit differing time-stepping requirements through adaptive implicit algorithm selection.