1) flow separation point on the top of the tire is much later
- It is hard to tell where flow separates along the tyre centre-plane with a total pressure coefficient contour, specially on a rainbow colouring.
a) Rainbow is confusing with its green part.
b) You need to plot Cp vs Theta on the tread to find the saddle point according to Fackrell and Harvey (1974).
c) What flow separation angle did you expect - between 285 and 290 degrees reading anti-clockwise with zero degree at the most upstream location?
d) On the tyre centre-plane, use Cp between -1 and 3 normalised by free stream dynamic pressure. The coefficient is above unity in this incompressible flow. It is not common to have Cp > 1 in incompressible flow unless there is work done on the system. In this case the moving ground plane does work on the tyre near its contact patch.
e) You would expect a highest Cp value at least 2.0 depending on how ‘tall’ your contact patch step is.
f) You need a lot of cells around the contact patch to capture this high pressure in order to capture the jetting phenomenon, which also affects the downwash of the flow coming off the top of tyre tread and reflects the tread separation location.
Here, read this paper: Sammy Diasinos, Tracie J. Barber, Graham Doig, The effects of simplifications on isolated wheel aerodynamics, Journal of Wind Engineering and Industrial Aerodynamics, Volume 146, November 2015, Pages 90-101
2) there is no stall region behind the rear wing.
- You front wing looks the same - same kind of wake - and how do you easily tell flow separation? Try to look at velocity contours instead of CpT contours and expect velocity distribution of a positive overhang flap. In all my solves, realizableKE was able to predict the rear wing.
Flow is mostly attached on the rear wing below:


3) The solution would be to use higher order schemes but the computations does not converge when using linearUpwind or vanLeer.
- It is not uncommon. Maybe switch off the limiter in the gradient of div(phi, U) when using 2nd order upwind. I can see that on your 25M mesh there are non-orthogonal cells above 70 degrees and you have used limited scheme on the viscous terms.I believe if a case is set up properly and will converge, then starting with 2nd order upwind will also converge. If it still fails, try limitedLinearV.
In terms of the gradient terms, maybe use leastSquares instead if you take away the limiter on the gradient of the convective term.
Added: my bad here. I didnt know which slice you were referring to. After looking at In-house OF32 with 33M and 5 layers, the ‘not stalled’ rear wing does look like a mesh resolution issue.
your post

I also found the slice for the tyre on In-house OF32 with 33M and 5 layers
your post
The In-house run has a much clearer main and tread vortices and a taller and more skinny front wheel wake, consistent with the wake features associated with a rotating wheel in contact with a moving ground.