Stéphane Amant

Technical comments on the results (to go with the slides)

Slide 2) : RANS solver, Spalart-Allmaras model, central scheme, implicit LU-SGS method for the time marching, multigrid.

Slide 4) : the colormaps are exactly the same ! On the Airbus France grid, you can notice the separation at the trailing edge of the root of the wing, which is a feature that many other participants have also captured.

Slide 6) : there is an obvious correlation between the zones where the skin friction is overpredicted and the regions where the grid is inclined !

Slide 7) : this is a front view of the leading edge of the wing at mid span, for both pictures. The lack of orthogonality of the DPW provided grid is clearly demonstrated.

Slide 9) : the various physical components of drag are represented here. They are computed and spatially distributed thanks to a dedicated post-processing program.

Slide 10) : the advantages brought by the far-field method are obvious. It is more accurate (removal of the spurious drag contribution) and it provides a very useful drag breakdown for the designers and the aerodynamic database models (performance and handling qualities) as well.
The relative importance of each component is associated to the low Reynolds number and the low transonic Mach number : small value of wave drag, huge values of viscous drag.
Estimating to 10 drag counts the effect of the forced transition, compared to a fully turbulent calculation, and assuming that the Spalart-Allmaras model naturally triggers the transition at the leading edge of the wing (which is indeed the case, a posteriori), you can notice that the far-field approach yields results very close to those of the experiment (around 298-10=288 d.c.)
 

My other feelings on the AIAA CFD Drag Prediction Workshop

The structured grid provided by the committe was obviously inadapted to the topic. Calculating and analysing the drag force requires high quality grids, especially with the Navier-Stokes modelization, which was not the case with the proposed mesh. If the purpose was to highlight the sensitivity of each code to the mesh quality, then I think it was irrelevant in the framework of a DRAG Workshop. Either you test the robustness of a code, or you test its accuracy (with a given parameter : here, the drag), but you can't conciliate both at the same time, in my opinion. I mean that, as well in an industrial context as in a research context, significant efforts are undertaken to build appropriate grids when it deals with the prediction of drag. This should be the basic priority for a Drag Worshop.
If I remember well, among those who have used the the structured multiblock grid, 4 participants out of 10 have not been able to provide satisfactory results : this is the proof that a Workshop focused on the prediction of drag MUST propose an adequate mesh, for the relevance and the validity of its own conclusions.

This was the negative aspect of my analysis. For the remainder, this Drag Workshop was a great initiative since it has allowed worldwide searchers and engineers to talk to each other, and to suggest various interpretations on everybody's problems. Moreover, each participant seemed to have been as honest as possible, without cheating or advertising too much on the results. This state of mind was definitely one of the keys of the success of the Workshop.
In addition, such a meeting is a very efficient mean of building oneself an overview of the state of the art in CFD, nowadays, either in industry or in the research centers.

I regret I can't provide the committee with my own grid, for obvious reasons of confidentiality.

Feedback on Langley Products and Services
Accessibility