12 March 2018
I have couple of suggestions:
1. You are using laminar model. Investigate your Reynolds number to understand what kind of flow it is. Naturally your flow should be turbulent (velocity of 28 m/s) so use turbulent model (I could get started with k-e).
2. With above in mind, I could converge your solution below 1e-3 by initialising from inlet instead of hybrid initialisation and running it for 700 iterations. Flow reversal is not a problem here. Flow direction can change due to flow separation and recalculation effects.
3. Focus on the value of Y+ plus your mesh is carrying. After performing certain iterations, Fluent could estimate Y+ value over a car body. Currently it is 5600, you need to bring it down between 30-300 by refining your mesh near the wall if possible. (ANSYS Student version may not able to produce such a fine mesh but research version can.)
4. BCs: if this is a simple flow over a car just for an illustration purpose then its ok to have this demo model. In this case, your results (cd & Cl) will be highly sensitive to wind tunnel wall effects, as walls are placed very close to the car. To avoid this, you should allow sufficient space around the car so that computed flow variables should not be wall sensitive. Here is the tutorial that shows how to do this.