Ogami musashi wrote: ↑
Wed May 17, 2017 2:08 pm
I was thinking about the setup of general wind tunnel and CFD experiments. If i'm not mistaken, it is always as if the two cars follow in straight line.
In reality, at racing separation distances, there's always a lateral offset between the cars.
I did lateral offsets in the wind tunnel and in CFD. I think I saw a picture from the OWG study with a lateral offset too, I'm really annoyed I didn't take a screen grab it at the time. Even small lateral offsets do allow the forces to recover towards the 'clean air' condition, downforce faster than drag. I think this is because the wheels of the following car (large % of total drag) are more aligned to the centerline of the lead car, so low drag, while the downforce generating surfaces operate in a higher dynamic pressure flow, with maybe small effect from down-wash helping to increase effective incidence of wings. I would argue that in corners the lateral offset is negligible as there is really only one racing line, and drivers will drive this consistently to within a few centimeters.
The real difficulties for wind tunnel studies are dynamic similarity, or how accurately the simulation matches flow state of the on track conditions, and representative separations. Both are something my study could have improved on but was not possible to within the limitations of our wind tunnel. We got around the separation issue to some extent by using a short bodied wake generator, which compared well to CFD with a full upstream car both aligned and with an offset, but still only managed a one vehicle length separation. The OWG group study had the same issues, and might I add with a lot more money to spend on the study. For impartiality reasons they used the Fondtech/fondmetal wind tunnel with 40% models, I estimate from the belt length and average length of a circa 2008 car that, even with the following car at the very rear of the rolling road, the maximum streamwise separation was only ~0.8 of a car length. Both TMG and Sauber wind tunnels can accommodate full scale cars so maybe a more representative separation could be achieved in these tunnels - though perhaps the only reason for that wind tunnel study was to validate one case from CFD over a large range of attitudes.
With regard to cornering, that is a tricky topic as it's not something I studied. There are CFD studies of vehicles in cornering, TotalSim https://www.totalsimulation.co.uk/corne ... -openfoam/
did some work on this with an F3 car. You can see that the wake follows the path of the car, you can also see this in wet races, but there is increased asymmetry in the wake (i.e. if you straightened the domain along the arc of the car centreline) from the effect of the yaw on surface pressures. So for the most part I would say the effect of following in the wheel tracks in a corner, as the drivers have to do if there is only one racing line, will be the same or very similar to following in the wheel tracks in a straight line.
But when you start to increase the complexity of simulations by adding cornering there's always something else to consider e.g. the effect of atmospheric conditions. I mentioned earlier in the thread that wind tunnels are designed for very low freestream turbulence (FST), this is to improve repeatability. The higher the FST the longer you have to average forces to reduce the size of your error bars. F1 teams run constant motion studies in the wind tunnel to reduce their 'wind on' time per case, and low turbulence is required so that error is not introduced, i.e. when you stick on a new part the delta you see is because of the part not because of the jet. The issue is how representative this smooth airflow is to on road conditions, so you have to start considering how clean 'clean air' is. There's a big push in the automotive sector to improve the correlation of studies to on road conditions for more accurate prediction of fuel consumption. On road/track you have cross-winds and head-winds which are interrupted by track side trees, buildings, spectators etc, so 'clean air' actually includes a baseline turbulence along a spectrum of length scales. You can model this in CFD but it increases the computational cost, normal transient simulations will only simulate 1-1.5s of time, to include atmospheric effects that would increase to 16-30s. This turbulence may alter the wake, the cross-winds change the way the wake comes off the car and how is propagates downstream... etc. Ultimately the most accurate way to test the wake effect will always be on-track but this is very expensive to implement, it's just not feasible especially if you're changing the geometry during testing, so we have to accept (to some extent) that wind tunnel and CFD are what they are, which is simulations.