Vyssion wrote: ↑10 Nov 2017, 21:07
RicerDude wrote: ↑10 Nov 2017, 20:39
The fact that teams still have correlation problems between simulation software and performance on the track proves to me that blowing smoke over a car doesn't give a clear enough picture of what's going on.
Gotta be careful here mate; because the issues that we currently have with correlation issues between CFD vs. Real-World Testing is because at a fundamental level, we still don't know how turbulence works when applied to complex geometry and different surfaces. Solving your typical RANS equations has your Reynolds Stress Tensor that is the main issue there of which there are tens or even hundreds of nth equation models etc aimed to predict this, and even when you get up to the LES type simulations which require stupid computational power even today, you still have to use a Sub-Grid Scale (SGS) model which makes the assumption that at a particular point in a turbulent eddy's dissipation, it will dissipate in a "univerally-same" way - which we don't know whether that is true or not. Obviously, predicting the turbulent flow over aerofoil profiles and body panels is kind of well backed up nowadays, however, most of the complexity for CFD, at least in my experience, of modelling aerodynamic phenomena comes from the four rotating tyres and the six vortices they shed each...
As far as correlation between Wind Tunnel Testing vs. Real Track Testing, it comes mostly down to the shear number of "outside factors" which real track testing can throw at you, i.e. altitude, gusting winds, humidity, dust particulates, track roughness, inclination in track surface, blah blah, etc etc... You just can't cover all of that in a wind tunnel especially with the strict number of hours you're allowed each week (I think they can still trade in their terraflops for more hours, but I think it peaks at like 20hrs or something - might be wrong here) Couple that with similarity parameters affecting result translation due to matching Mach Number
and Reynolds Number being very difficult across all facets of a model (teams cannot run 100% scale models anymore) and its a bit of a nightmare for correlation!
I'm going to come from a different perspective as Vyssion as a more wind tunnel focussed engineer vs CFD. We rarely 'blow smoke over the car'. If you see smoke in a wind tunnel it's normally for a marketing/head office visualization, it's same the reason CFD was initially dubbed 'colourful for directors'. That said it can be useful for a quick check if you have some unexpected result and don't have a PIV system in situe, but it's never in the test program.
The correlation issues, certainly for Red Bull this year, were blockage based with the wider cars, don't know what their solution was but Force India have curved walls in their tunnel (I know they now use TMG) to combat the same issue. A lot of the teams are using bigger models than their tunnels were designed/built for. Other correlation issues as Vyssion mentions, and it's the same in the road sector for fuel targets, are atmospheric turbulence, i.e. crosswinds and gusting, vibrations/ground clearance, dynamic similarity is a big one - scale and speed restrictions is equivalent to a medium-ish speed corner...etc. Even with pneumatic tyres getting the correct contact patch, side-wall, and top tread deformation is difficult, and I still don't think has been perfectly achieved. To visualise the flow field on track also requires 'intrusive' devises, i.e. the aero rakes affect the flow field around the car.
Where I don't see this technique as being useful, and it's something I bang on about, is giving value. For something to be useful in an engineering application it has to add something. PIV gives velocity vectors, so you see where flow is going and how fast. Pressure probes, 5,7,13-hole, will give pressures but also velocity and directionality. Obviously force and surface pressures are measured...etc.
:
.