|
Executive summaries of Action Groups of Group of Responsables in Aerodynamics
The main purpose of the Action Group is to validate Reynolds-Averaged Navier-Stokes flow computation methods for the prediction of transonic flow physics on transport aircraft wings in design and off-design conditions. Both comparison with experiments and code-to-code comparisons have been carried out. The partners involved in this Action Group are Aerospatiale (AS), BAE SYSTEMS, EADS (DASA), NLR, ONERA, and Saab. The present Action Group can be considered as the follow-up of GARTEUR Group AG(AD)17, which verified computational methods for three-dimensional, subsonic and transonic, inviscid flow around complex configurations (Ref. 6, 1995). The activities of the Action Group are closely related to the activities of the Action Group AG(AD)-28, ‘Transonic wing/body code validation experiment.’ This Action Group aims at obtaining accurate and detailed pressure distributions, boundary layer traverses, and wake traverses for assessing and validating CFD codes. A design and off-design condition are simulated for the AS28G wing/body geometry of AS. The design condition is defined by a Mach number of 0.8, an angle of attack of 2.2 degrees for a Reynolds number of 10.35 million. The off-design condition is given by a Mach number of 0.8, an angle of attack of 2.57 degrees. At this off-design condition the transonic flow shows incipient shock induced flow separation on the wing upper surface. The multi-block structured grid has been generated by BAE SYSTEMS, and consists of 96 blocks with a total of 2,518,528 cells. Code-to-code comparisons are executed on the basis of pressure distributions, skin frictions, and boundary layer profiles. The pressure distributions are also compared with experimental data obtained in the ONERA S1-Modane wind tunnel (Ref. 5). Comparison of the computed and measured pressure distribution is generally good, except for the shock position. The overall correlation between the CFD results of the partners indicates that the wind tunnel model deforms under aerodynamic load, resulting in a little torsion with respect to the computational geometry. Agreement with experiment is excellent for the lower side of the wing for all models. The computed lift and drag coefficients on the common grid of 2.5 million cells are within the following range: condition DCL range DCD range design(a=2.2°) exp+7.1% exp+2.4% exp-2.7% exp-12.0% off-design(a=2.57°) exp+6.0% exp+1.2% exp-2.5% exp-9.7% Based on a methods’ comparison the following conclusions can be drawn. There is a significant variation in the computed aerodynamic coefficients, for which no grouping of results can be presented. The Baldwin-Lomax model predicts shock-induced boundary-layer separation for both cases. The models applying two-equation turbulence models fall into four groups: BAE(k- g) and Saab(k-w); NLR(k-w) and DASA(k-w); Saab(k-e); and ONERA(k-w). Even though the numerical schemes of the methods are comparable, specific choices in turbulence model parameters and/or implementation have a significant effect on the prediction of the shock position and shock-induced separation. It is recommended for future work to continue a detailed validation exercise as done in the present study with the following modifications/extensions: · replace the ‘common grid’ strategy with a ‘grid convergence’ strategy. This grid convergence strategy will involve a sequence of computational grids to be used by every partner and resulting in ‘grid converged’ solutions of each individual CFD method. This new grid convergence strategy will allow a separation between discretisation errors and errors arising from turbulence models, and from forcing laminar/turbulent transition. It can safely be asserted that the discretisation errors are not negligibly small on the current grid of 2.5 million cells, and that these discretisation errors are different for different CFD methods. · more detailed experimental data as planned in AG(AD)-28, · clarification of wind tunnel model deformation effects, so the simulations can be carried out for the proper flow conditions and correct geometry, · perform a range of flow condition test cases rather than two isolated cases, investigate the effect of the forced laminar/turbulent transition locations. The overall conclusion is that the common grid strategy should be replaced with a grid convergence strategy to enable further CFD validation.
|
