So, even the most advanced usable methods are still using tricks to do these calculations? That almost seems unreasonable at a F1 level with modern hardware partners.n_anirudh wrote: ↑18 Mar 2024, 23:05One could setup an unsteady sim with moving mesh and look at the vortical structures that change/evolve as the car pitches/rolls etc - but these are highly expensive simulations - maybe tens to 100 x the times for a RANS. They would use DDES or hybrid methods for these.
One may (very cautiously!) perform RANS at the extremities/few intermediate points along any characteristic to obtain performance curves.
If you look at a car at a very discreet period of time, say 0.05s, the car seems almost stationary, so that's a good approximation. You can than measure the forces, the air is acting upon the car, do this over and over again in different scenarios (geometries) and create a database of forces in relation to the geometry. You can also use a bit more resources to get some simplified (e.g. rear wing only, as you can use formerly generated satic data as boundary conditions) aeroelasticity simulations and add it to the database too.
Are there any softwares using point cloud data instead of meshing for this process? I thought the Euclideon Geometry breakthrough from 2005 had already unlocked that tech...G-raph wrote: ↑19 Mar 2024, 01:56The technology for this absolutely exists in theory. You need to couple a CFD solver with a FEA solver (which is called FSI : Fluid Structure Interaction) and run a transient simulation with movable mesh (either using a mesh morpher or a geometry morpher and remesh at each time step).
It would be incredibly expensive (way more than 100 times a standard RANS simulation), as already mentioned, especially if you are looking at an entire F1 car.
But the real show stopper is that each tiny modification of the mesh (= each time step) would count as a new run, and as you know these are now restricted by the FIA (in both CFD and Wind Tunnel). I don't think F1 teams would like to spend an entire monthly CFD development allocation to run such a complex simulation.
And the results would probably be wrong anyway.
I agree with this.SharkY wrote: ↑19 Mar 2024, 15:11The problem with complex transient simulations is that they require extreme computing power and you can never be sure if the results are true as small errors can build up and throw off the flow. So why bother, when you can simplify it to get the close enough results and get the corrections from the wind tunnel and track data?
Yes. This is exactly what I refer as "morpher" in my original post. They are quite common and work very well.