An improvement of the spatial accuracy of an upwind biased finite volume scheme adopting a pre-processing approach is investigated. In the pre-processing approach, the cell face values of the dependent variables are estimated with a higher order spatial accuracy at first, usually based on the Taylor expansion in the computational space, then numerical fluxes are determined by an appropriate method such as upwind scheme. A sufficient grid smoothness in the physical space is implicitly assumed when the Taylor expansion is applied to find the average slopes in the computational space. This can lead to a possible generation of numerical errors on the skewed grid system. In the present study, the distribution of conservative variables is fitted, rather in the physical space with the accuracy up to a third order, then the averaged slopes in each cell were found. Numerical experiments comparing various methods of finding the relevant averaged slopes including the present one are carried out.
Aircraft Computational Aerodynamics
- Pub Date:
- November 1987
- Computational Fluid Dynamics;
- Computational Grids;
- Finite Volume Method;
- Fluid Mechanics and Heat Transfer