Towards Prediction of 3-D Separated Flows Using an Unstructured LES Code

In this paper, we document large eddy simulations (LES) performed on a sphere and the Advanced Swimmer Delivery System (ASDS), a small blunt-ended submarine. The objective of this work is to develop a methodology for being able to compute the occurrence and strength of stern separation in order to accurately predict drag, maneuvering and structural loads, and acoustic signatures. We are using an energy-conserving large eddy simulations (LES) code called MPCUGLES that runs with hybrid structured-unstructured meshes. We document LES of flow over a sphere at Reynolds numbers 10,000 to 1.14 million with excellent comparison to experimental data. We also document a preliminary effort for LES of flow over ASDS. Even the simplest of these computations is very CPU-intensive necessitating the large amounts of CPU time available through the HPCMP Challenge Project C3U.