Description
A computational study is conducted on aerospike nozzle flow to determine minimum computational domain and grid requirements for rapid prototyping and in preparation for high-fidelity flow analysis of a truncated aerospike. Using Reynolds- Averaged-Navier-Stokes (RANS) model simulations, the effect of the computational domain size on the averaged flow development is investigated for two representative aerospike geometries, one representative for an air-breathing propulsion system and another for a rocket propulsion system. RANS simulations for the representative air-breathing system are conducted on a large domain which includes the pressure chamber, converging nozzle, aerospike ramp, and sufficient free space downstream of the engine to capture the jet flow. These results are validated against literature. In preparation for Large-Eddy Simulation (LES) of the instantaneous, unsteady flow on a reduced domain, the accuracy of RANS is assessed on a truncated domain using extracted pressure and velocity profiles as boundary conditions. The reduced domain focuses on the base of the truncated aerospike and its near wake. If the domain inflow location is closer to the truncated nozzle base, the wake size and magnitude of the Mach number inside the wake reduces. Wake closure is observed for both a large domain and a reduced domain at the same ratio of combustor stagnation pressure to the ambient pressure. The location of shock-expansion structures and Mach profiles in other regions are within 10% for different domain sizes even if the computational domain encompasses merely the trailing end of spike and three base heights in each direction for flow development. In the case of the rocket aerospike flow, the shock-expansion cells for a baseline case are shown to be dissipated for reduced domain sizes. In addition to a changing wake size, the Mach number contours show up to 20% differences throughout the domain. As compared to the air-breathing aerospike, a larger domain of six base heights downstream of the truncated spike is required for stability of the simulation. In future work, we aim to use the velocity profiles determined by RANS as input for LES. We plan to validate higher fidelity LES against RANS.
