CABARET on Jet Flap Noise and Quasigeostrophic Ocean Circulation Models
Numerical Algorithms Group Ltd,
Wilkinson House, Jordan Hill Road, Oxford, OX2 8DR, UK,
Date: October 1, 2012
This dCSE project will concern two very different CFD applications, which both share a common method for resolving advection, forcing and dissipation - the Compact Accurately Boundary Adjusting high-REsolution Technique (CABARET). The CABARET method is a low dissipative and low dispersive scheme that constitutes a substantial upgrade of the second-order upwind leapfrog method. The algorithm is very suitable for distributed HPC since it has a very local computational stencil that for scalar advection constitutes only one cell in space and time.
For geophysical fluid dynamics, CABARET is used to model the geostrophic, turbulent midlatitude ocean circulation, in terms of the quasigeostrophic potential-vorticity dynamics. This code is called PEQUOD (Parallel Quasi-Geostrophic Model) and is primarily intended for the study of mesoscale eddy processes on a multi-layer structured rectangular domain. In aeronautics, CABARET is applied to a turbulence model for jet-flap-noise interaction (Cfoam-CABARET). Modelling the unsteady component of the acoustic model for the free-stream flow/jet interaction with the wing-flap, as well as for the study of hot/complex geometry jet physics, is very challenging. The computational model uses a Monotonically Integrated LES (MILES) approach that is explicit Sub-Grid-Scale (SGS) turbulence-model free.
Cfoam-CABARET uses an unstructured domain decomposition which is suitable for complex geometries, but this gives rise to issues related to I-O performance on large HPC systems, whereas PEQUOD does not experience this problem, due to the use of a structured multi-dimensional domain decomposition. However, unlike Cfoam-CABARET, PEQUOD requires an additional global inversion scheme which is used to calculate the potential vorticity (stream function). This project will implement MPI-IO in Cfoam-CABARET to improve the I-O performance, and optimise the parallel Swartztrauber elliptic solver for the potential vorticity in PEQUOD. This work will help both applications: in aeronautics, to investigate the effect of fine-scale-flow structures on far-field noise in the audible range of frequencies; in ocean modelling, a whole set of very interesting and important results are waiting to be revisited and upgraded, with a completely new level of dynamical realism that can only be achieved with a scalable
implementation of the existing code.
This project was approved for 6 months effort in December 2010 and was completed August 2012. The work was supervised by Dr Sergey Karabasov of the School of Materials Science and Engineering at Queen Mary University of London and Dr Pavel Berloff of the Department of Mathematics at Imperial College London.