[MarkT]

Dear fellow Fluent users,

for my graduation project i am simulating a DrivAer generic car model as found in several german SAE papers. I started my research using an Ahmed model and did some variations. Now with the more complex DrivAer model i am running into some problems regarding convergence etc. As far as i know continuity doesn't solely tell if a solution is converged but the parameter i am solely interested in is the Cd coefficient of the vehicle body, varies more as after for instance 500 iterations than it does by variating the vehicle's shape. 100 more iterations can give a 10% higher value and 100 iterations ahead again 10% lower.
Depicted below is the continuity on the x-axis, and the Cd value on the y-axis, the big jump is when switching from first order to 2nd order to aid convergence and stability.

I have done a lot of variations on the simulations to find out where the problem originates but can not conclude yet. Can anyone give me some tips?

some more facts:
ansys fluent 15.0
4-cores i5 3570k 24gb ram
turbulence model K-Omega SST, Coupled.
4M+ elements, Polyhedral and hexahedral, refined near the vehicle body, and in the wake region of the body.
Y+ max of 150-200 at very few spots, the rest should be around max 100.
first 100 iterations with first order (pressure, momentum, turb kin energy, specific dissipation rate) and turbulence viscosity underrelaxation at 0.8, then second order (momentum, turb kin energy, specific dissipation rate) excluding pressure with turb viscosity underrelaxation 0.95.
(am i missing something?)

many thanks!
Mark