[IvanCFD]

Hi,

I am using the Ansys 13.0 Workbench mesher to mesh a 3D saloon car (quite blunt, without appendices, but with many fillets) in a wind tunnel.

I have tried Patch Conforming and Hex Dominant methods and they take ages and do not seem to finish the meshing process. With Patch Independent it does not seem to capture well the features and with CutCell it cannot cope with an inflation.

The car length is about 1500 mm, the element min. and max. sizes are 1 mm and 250 mm respectively. The growth ratio is set at 1.2 and the global sizing function is based on proximity and curvature.

For instance, with Patch Conforming and leaving everything by default, it does not reach an end.

I hope someone out there can give me a hand. I have been using Gambit, but for less complicated geometries and do not have much experience with Ansys Workbench.

Thanks a lot beforehand.

Regards,

Ivan.

[PSYMN]

Patch Conforming is only slow if you have a very fine mesh on the surface, so perhaps your sizes are not actually set correctly. I would check the little circle indicator on the screen (when you set your sizes) to make sure it is reasonable. Some times users think their model is in meters, but the scale is actually in mm and they have set a much smaller size than they thought. Your sizes seem fine as you describe them, but perhaps try an experiment and set them 10 times larger. What kind of mesh do you get? If you get a coarse mesh, you can always run it again with a finer setting.

The Hexadominant method is not really for CFD at all. Don't use it. It is for Mechanical where they are very interested in having very coarse isotropic hexas orthogonal to the walls and don't care much about the core of the model.

The PI Tetra should be very good, but is subject to the same sizing problems as patch conforming. If it wasn't capturing features, you can adjust some settings. This method is much easier to control in ICEM CFD, so I don't use it much in ANSYS Meshing, but maybe someone else here can give some tips for you.

The new Cutcel method should work well, but I guess I have been using the 14.0 version (due out in about a month) and it has had a lot of work to improve its prism, speed, etc.

I will have to see the model to provide much better help than this.

Have fun...

[IvanCFD]

Hi PSYMN,

thanks a lot for your comments.

I have tried to mesh the car with PI and inflation but for some reason it does not create the inflation layer (smooth transition, max. 7 layers).

For the sizing I use curvature and proximity and as min. element size 0.5 mm, and 250 mm as max.

Attached with this post some feautures fo the car. As you can see it is full with fillets. What feature angle (approx.) would you recommend to use in PI according to those pictures?

Besides, I need to implement the wheels as rotating, and for the inflation layer to work in the vicinity of the intersection between the wheels and the ground I was told to use a fillet (in picture 3.png the rear wheel has a 3 mm fillet whereas the front wheel does not, just for comparison).

Hope you can help me.

Thanks a lot again and looking forward to hearing from you.

Regards,

Ivan.

[IvanCFD]

Hi again Simon,

here you go some pictures of the resulting mesh. It gave errors.

Pictures attached of how the mesh looks like, the errors and the problematic geometries, and the settings I have used.

Any recommendation please?

Thank you so much once more.

Ivan.

[IvanCFD]

Hi Simon,

sorry, there was a mistake with the stage I showed in my previous post. The inflation layer was not generated because, by mistake, I did not choose all the face of the car.

Once I have noticed that and corrected it, an inflation layer has been formed. However, it does not seem to respect the smooth transition behaviour over many areas of the car. It looks like it does not vary much its thickness, and then there is a huge jump in size between the cells in the inflation layer and those outside (picture 9.png). Can you guess why?

Besides it seems that the wheels are problematic (picture 9.png too). I do not understand what that error means.

Making a longitudinal cut, you can have a look close to the rear wheel in picture 10.png. In picture 11.png it seems that the inflation layer has not respected the fillet I created. Although in picture 12.png it seems that the inflation layer is nicely formed over the fillet and wheel.

Hence, where and what is the problem?

Thanks a lot beforehand.

Regards,

Ivan.

[IvanCFD]

Hi Simon,

with Patch Conforming I have been able to have a mesh, although not of a very good quality.

I have created a body inside my domain comprising the wake region of the car (picture 12.png attached). I use that body as a body of influence to make a refinement of the mesh in the wake region. However, when I click on generate the mesh, the entire domain gets meshed but that wake region. Would you know why? Perhaps that little symbol (which I do not know what it means) on the left of "wake" in the project tree could be a hint for you.

Thank you very much beforehand,

Ivan.

[Sushilkumar]

It indicates the meshed bodies in your geometry.

[IvanCFD]

Hi Sushilkumar,

yes, I know it indicated the bodies to be meshed. But what I am asking about is why the body "air_total" has a positive mark symbol on its left and the body "wake" does not.

If you know the reason for it, please let me know. Thanks beforehand.

Regards,

Ivan.

[PSYMN]

the body of influence needs to be a separate body, not subtracted from your original flow domain. Then right click on mesh to insert sizes and change the selection to body of influence, then select your body... I don't have it in front of me, but it is something like that...

[IvanCFD]

Hi Simon,

I have been able to create a mesh (I do not know how decent it is though) with Patch Conforming. It is 5.5M cells (I guess I cannot go beyond 6M as the computer I have for running the simulations has 6GB of RAM).

I have started simulations on this Fluent. My working speed of the free-stream is 25 m/s, but I read it is advisable to start with a lower velocity. So I have started with 5 m/s and a k-e RNG turbulent model. I ran 300 iterations at this speed, and then I increased the speed up to 10 m/s. However, with this new speed, it gives me the warning "turbulent viscosity limited by viscosity ratio..." and the residuals go batty. I decreased the under-relaxation factors to 0.2 for momentum and 0.2 for turbulence. But it does the same.

Attached a picture of the residuals (remember, from 1 to 300 iterations it is at 5 m/s, from 300 onwards it is at 10 m/s)

My working parameters are:

- k-epsilon RNG, enhanced wall treatment, pressure gradient correction
- turbulence intensity at inlet = 0.4%
- except moving ground, the rest of faces are set to symmetry.
- exit as pressure outlet with same turbulence intensity as inlet (0.4%).
- PRESTO! and Second Order for turbulence as discretisation.
-SIMPLE as the pressure-momentum coupling.

What are your suggestions?

Thank you very much beforehand,

Ivan.

[Olds88]

Ivan,

Sorry to revive this old thread I hope you got your stuff figured out. How did you end up meshing the wake region. I'm trying to do something similar with a body of influence and my mesh doesn't get created in the body either. There is a whole in my mesh where the body is supposed to be.