A Newton-Krylov solver with a loosely-coupled turbulence model for aerodynamic flows

Blanco, Max (2006) A Newton-Krylov solver with a loosely-coupled turbulence model for aerodynamic flows. University of Toronto, Department of Aerospace Sciences, Doctoral Thesis, 160pp.

Abstract

ex Computational solutions of the Navier-Stokes equations have proven to be a useful tool in the design of aircraft. A Newton-Krylov flow solver for unstructured grids is developed in order to demonstrate that a formulation in which the mean-flow and turbulence mechanism equations are loosely coupled can be more economical than a similar fully-coupled formulation.
The Favre-averaged Navier-Stokes equations are derived for steady two-dimensional flows, and the turbulence mechanism is described. These equations constitute a model of the physics of aerodynamic flows. The model is validated against experimental data. The objective of this thesis is to examine a means to improve the iterative process by which the solutions are generated. The Newton-Krylov iteration is selected in order to refine the solution, and its features examined. The authors of current Newton-Krylov techniques have fully coupled the turbulence mechanism to the Navier-Stokes equations. A contrast and comparison study made here between the fully-coupled formulation and a loosely-coupled alternative favours the latter. An 'equivalent function evaluation' metric is selected for comparison purposes, and is assessed by means of diverse computers. Published results which use the metric are located, and the present loosely-coupled formulation for unstructured grids is found to be significantly faster in this metric than similar fully-coupled formulations. The advantages of the loosely-coupled formulation with respect to the fully-coupled formulation are stated and future avenues for exploitation of the proposed technology are examined. Appendices consist of: a formalism for the Favre average and consequences of its derivation; a short tract on Taylor series; and an essay on the Fréchet differential.

This record has no associated files available for download.

Contributors

Download statistics