Hi Everyone,

I am a new Fluent user and am trying to simulate airflow over a car in a wind tunnel. The wind enters the tunnel at 16.667 m/s. I am trying to get grid independence using a steady solver. I am using the k-e tubulence model. HOWEVER, I CONSTANTLY HAVE TROUBLE TRYING TO GET THE CONTINUITY RESIDUAL TO REACH THE CONVERGENCE CONDITION OF .0001 error reduction. I have tried reducing the under relaxation condition for pressure but it doesn't seem to work. I have played around with the solution controls also. I would be very grateful is anybody could make some other suggestions.

Thanks alot,

Jim.

Hi Jim,

This could be for a number of reasons. Ones I can think of are: 1. not enough grid points 2. maybe it's not steady state 3. upstream and downstream boundary locations - are they far enough away?

Try checking your y+ values (adapt--> y+), and the mass flow through the domain (report-->fluxes).

Trac.

Hi Tracie, thanks a lot for your speedy and very helpful reply. I tried you suggestion of refining my mesh by checking the y+ values and I found this did indeed enable me to get first order convergence on grids where I couldn't achieve it before. I am still having trouble getting second order convergence. I hope this isn't a problem to do with he number of nodes as I have about 200000 cells already for each wind tunnel/car and I thought this should be enough. I also checked my mass flow rate and the pressure in and pressure out values are very close. I'll let you know when I resolve my problems. Thanks again for your help and if you have any further advice please don't hesitate to put it forward.

Cheers, Jim.

Jim,

200,000 hex cells sounds pretty coarse for a wind tunnel simulation. If you are using tets, then it is really coarse.

What level of detail do you have for the car?

Is it just a bluff body representation?

Are you modeling half of the vehicle or all of it?

Chinor

Hi Jim, If your car is a blunt body, which most are, you will probably need hex elements near the surface to pick up the boundary layer accurately. If there is lots of separation happening, you may also need to switch to a better wall model, however this will make your grid size increase lots. Are your residuals going steadily down then flattening off or oscillating a bit? How are the lift and drag values looking or are you going for more qualitative results? Trac.

Hi Chinor,

Yes I am using tets, so from what you are saying I guess I should refine the mesh. I am modelling the whole car as I am interested in air flow for different yaw angles so I am rotating the car, thinking it would be easier than working with symmetries and mirroring etc. I have just modelled the outside of the car (it is just a very simple model). The car is about 2 metres long and I initially meshed its faces using 50 mm spacing between nodes. I have since refined it in Fluent using the ADAPT -> REGION selection to get more mesh points behind the car and on each side. I then smoothed the mesh using ADAPT -> VOLUME (CHANGE). This made the triangles on the face of the car smaller especially aroun the edges. Thanks a lot for your advice, I will try even finer meshes. Perhaps I will make the intial meshes finer (they were originally about 80000 cells).

Again, Thanks for your help, I'll keep you posted.

Cheers,

Jim.

Hi Trac

With regard to my residuals,everything falls nicely except continuity which often gets down to about 2.3e-3 and then starts oscillating at slightly higher values. With respect for the drag and lift forces, I haven't considered them until I read your email. I have since run one of my cases which doesn't converge because of the continutiy residual and it seems the drag stays constant at about the value of 168 while the lift oscillates between 4 and 5. I admit I don't really know the significance of these results.

I am looking for qualitative results - trying to simulate the creation of vortices on the A-pillar region of the car, near where the side mirrors would be. With respect to the boundary layer on the car, when I made the mesh using gambit, I had the mesh on the faces of the car 4 times smaller than the mesh I used to mesh the volume of the wind tunnel, thus hoping that the tetrahedral cells around the car would gradually get bigger as one moves away from the car. The origianal mesh contained 80000 tetrahedral cells and the wind tunnel was 3000mm x 2000 mm x 11000mm. The car is about 2000 mm long. The mesh on the car was meshed using node intervals of 50 mm. I have since refined the mesh even more using fluent's ADAPT -> REGIONS and ADAPT -> VOLUME CHANGE to try to get a finer, smooth mesh, especially in areas I expect a lot of turbulence. Is it still necessary to use Hex cells on the boundary layer? Perhaps my mesh may still not be fine enough.

Again, Thank you very much for the time you have taken to consider my problem. I appreciate your help alot. I am new to CFD and Fluent as you can probably tell and am appreciative of the help of others with more experience.

Cheers, jim.

Hey Jim,

I just noticed your email address! Are you part of VPAC in Melbourne? If so, then this is just the weirdest coincidence in the world, because I am actually your Fluent support person. Ask Steve and give me a call if you like.

Trac.