There are so many solvers, and so many different turbulence models that it's too difficult to say.raymondu999 wrote:Hey folks. Just wondering. As we know from miserable ventures such as Wirth and Virgin, CFD is not bulletproof. As teams such as Ferrari have shown us too, you can have an almost "faulty" CFD with bad correlation. I assume that plugging in numbers from the cars' sensors would allow the CFD to assimilate these numbers, and improve its correlation. So what "baseline" is open source/amateur CFD working from, and how accurate is it really?
CFD cannot fully model fluid flow (it's just too complex for even the fastest supercomputers), so we make assumptions about the flow, especially regarding turbulence . These are dealt with in smaller 'sub models' or just specifired values. When those assumptions turn out to be correct, we get an accurate simulation. When they are wrong, you get garbage out.
Then you get into situations where one turbulence model may be valid for high speed on a hot and humid day, but be wrong for a cold dry day. Or that it's valid at higher ride heights, but fails at low ride heights due to the inteaction with the ground.
The problem of not validating a model vs windtunnel and realworld testing is that you don't have a bloody clue why the model is spitting out rubbish. Even then you may get something that doesn't fully translate to reality, it's why CFD is a bit of a black art.
CFD is also --- for stuff like testing ride hieghts (and similar), where a windtunnel is fantastic. Each change in ride height, requires a seperate simulation, but in the tunnel it's a simple 5 minute job.
It's why the Virgin idea of CFD only was always doomed to fail.
