Objective
Objectives and content
The aim of this project is to add a major step in the
accuracy and efficiency of unstructured grid methods for
simulating high Reynolds number flows in aeronautical
industry. This will be achieved by adopting a new class
of superior inherently multidimensional discretisation
schemes combined with innovative multigrid acceleration,
both designed for unstructured hybrid grids.
Industrial benefits are expected in allowing fast and
accurate simulations of viscous drag in design cycles of
aeronautical components. Such simulations are at present
not possible on a routine basis: structured grid solvers
are accurate but require excessively large human effort
in multi-block grid generation, while unstructured grid
finite volume solvers allow automated grid generation but
fail due to insufficient accuracy.
More specifically, the objectives are:
to obtain a conservative Euler solver, which for
subsonic flow has an accuracy comparable to that of
standard potential flow solvers. For viscous flow
simulation, this will lead to a reduction of the error
margin on drag by a factor 3 compared to standard finite
volume Navier-Stokes solvers. Such gains are possible by
applying new genuinely Multi-Dimensional upwind High
Resolution (MDHR) monotone shock capturing
discretisations, which combine the advantages of Godunov
finite volume schemes (monotonicity, excellent shock
capturing and robustness) with those of Finite Element
methods (proven second order accuracy on arbitrary grids,
generality, multidimensionality, compactness of the
computational support limited to nearest neighbours).
to reduce the cost compared to standard dimensionally
split finite volume solvers by a factor of 10: allow a
smaller number of iterations to obtain the steady
solution due to improved and more robust iterative
solvers, based on a combination of Krylov subspace
acceleration and new unnested semicoarsening multigrid
algorithms. Allow smaller computational effort per
meshpoint due to the compact support and because fewer
Riemann problems are to be solved (one per cell instead
of one per cell-face). Allow reduced communication
overhead for parallel computing due to reduced message
passing, resulting from the compactness of the MDHR
discretisation, and demonstrate robust solution adaptive
computations in industrial context.
to demonstrate the new algorithms for industrially
relevant high Reynolds number turbulent flow testcases
(complex 3D aerodynamic configurations), both steady and
unsteady, and prepare implementation in industrial
environment. Focus will be on solution quality,
robustness, efficiency and computational cost.
Due to its fundamental and theoretical aspects, the
proposed project is basic research upstream of industrial
research. Brite Euram III Area covered by the proposal
is 3A.3.5.; the technology is also applicable to 3A.3.9.
Fields of science
Not validated
Not validated
Topic(s)
Call for proposal
Data not availableFunding Scheme
CSC - Cost-sharing contractsCoordinator
1640 Rhode-Saint-Genése
Belgium