An intelligent design optimization framework significantly impacting the propulsion system design cycle utilizes an evolutionary algorithm that searches the global design landscape in an efficient manner, a hybrid unstructured CFD solver that provides high fidelity turbulent reacting flow solutions and a feature identification and knowledge extraction utility that extracts meaningful patterns of physical phenomena and provides useful correlations as the optimization process unfolds. The design optimization and extraction procedure is especially attractive for propulsive elements such as injectors, combustors, supersonic inlets, turbopumps and nozzles. This is primarily because these elements are replete with complex physics such as shocks, secondary flows, turbulence, and reacting zones that can be resolved by the high fidelity solver, accounted for by the evolutionary design algorithm in searching for novel designs and tracked by the feature extraction toolkit. Incorporating the information about important correlations between physical flow features and design variables can lead to both robust and novel designs and invaluable knowledge and insights for the design engineer. Since multi-element unstructured methodology provides the flexibility of generating efficient grids for very complex shapes, it facilitates the use of Navier-Stokes solvers in the design optimization loop and is at the core of our design optimization framework. To this extent the economy and success of using Reynolds Averaged Navier-Stokes (RANS) solvers with an evolutionary algorithm is directly attributed to the multi-element unstructured approach. We illustrate the multi-element grid generation procedure on a representative supersonic ramp injector (see below) that is typically used in scramjet combustors.
Hybrid Unstructured Grid
for Supersonic Injector Ramp
