Site menu:
News:
Nov 12, 2007:
First version of our new group webpage is released!
Group wiki »
(restricted access)
Projects
New Schemes and Solvers
In order to make high-order schemes on unstructured meshes competitive for real-world problems, new discretization schemes and solvers are required. We have made several significant steps in this direction, including the development of the Compact Discontinuous Galerkin (CDG) method. This is a stable and optimally accurate scheme for discretization of viscous problems, which is easy to implement and requires lower computational cost that any other known scheme.
Another development is our combined block-ILU and two-grid preconditioner for Newton-GMRES iterative solvers. By using ILU as a smoother in a simplified multigrid step, we have obtained a very efficient low-cost preconditioner which applies to a wide range of problems. The performance is highly dependent on the element ordering, and our minimum discarded fill algorithm optimizes the order automatically using a matrix-based approach, which does not need any specific knowledge about the problem.
Flapping Flight
One of our primary application areas is the accurate modeling of flapping-flight. As part of a major effort in Computational Flapping Flight, we are modeling flows around pitching and heaving airfoils, and coupling the fluid with structural models such as a membrane.
The ultimate goal of our simulations is to model the flow around a flying bat, which is
a highly complex problem involving large deformations, fluid-structure interaction,
and transitional flow.
Aeroacoustics
We are applying our high-order solvers to problems in aeroacoustics. By solving the compressible Navier-Stokes equations accurately, the sound waves are taken into account automatically and no modeling is required. Currently we are working on noise prediction for flow over a wedge-box geometry.
Transonic and Turbulent Flows
One of the major challenges for high order solvers is the robust treatment
of discontinuities and under-resolved flow features. These appear naturally
for problems with shocks, and to some extent for RANS flows with certain
turbulence models. We have developed a general purpose approach for solving
these issues, which is based on classical artificial viscosity and a highly
selective element-wise sensor.
