SimScale CAE Forum

Problem with accuracy of results on air flow around a sphere

Hello, I have been trying to simulate air flow around a floorball (a ball with 26 1cm wide holes on it) and a smooth sphere, and to compare their drag coefficients at different Re.

My problem is mainly with the accuracy of the results from the smooth sphere, its Cd being at about half of the expected result. At around Re 0.95e5 its 0.26 while the expected Cd would be near 0.5. At the same Re the floorball has Cd of 0.55, much closer, but still not quite. The trend of smooth sphere having Cd half of the floorball seems to continue at other Re, with Cd increasing at lower Re but difference staying same.

Both simulations have exactly the same quality meshes and simulation setup, but the results are very different. The results don´t need to match real life values exactly, but rather be comparable with each other. There is probably something wrong with my simulation setup, so feel free check the project out.

I am quite new to CFD and SimScale, so any help would be appreciated.

Link to my project: https://www.simscale.com/workbench?publiclink=a4a14d5c-c92d-4871-a8ef-2eaaa2432557

EDIT: New link: https://www.simscale.com/workbench?publiclink=a4fe2133-39b4-47a3-9071-d7b2f4dcac38

Hi @jj2r,

You might first check your geometry scaling. You can check that in the Event log or directly at the scale. If it is not correct you can change the scaling of your geometry by adding a new geometry operation as depicted in the picture below.

Let me know if that information helped you or if you already have the wanted size but still do not know where the erroneous values are coming from.

Cheers,

Jousef

The scaling seems to be right so the problem shouldn’t be there.

Hi again @jj2r,

I can have a look at that but that might take some time. Hope you are still fine with it :+1: :slight_smile:

Cheers and till then!

Jousef

Hey @jousefm
Any progress? Can you also check if the lift coefficients in the rotating zone ones are reasonable?

Hi @jj2r! I am very sorry for the late response!

I will try to set up a new simulation and see what results I can get. Did you make any calculations by hand to confirm your expected values or have you taken the values from papers?

All the best!

Jousef

For smooth sphere I used for example this:

For the floorball there isn’t really anything other than [this] (http://publications.lib.chalmers.se/records/fulltext/241131/241131.pdf), where the Reynolds number Drag Coefficient graphs look somewhat curious (lower than i’d expect). However the drag coefficients for the foorball might already be close to real life.

The smooth sphere drag coefficient should be higher, and comparing the lift and drag between simulations with identical/near identical settings but different models is essential since there isn’t much known values to compare to.

My goal is to find reasonable values for at least lift coefficient and hopefully for drag coefficient, since the purpose of the holes, according to my real life experiments, seems to be to either reducing the lift or the spin.

Hi @jj2r!

Awesome! Thanks a lot for the document. Gives a nice overview of how the drag has to look like. Trying to set up a simulation asap!

Coming back to you soon with some results.

Cheers and all the best!

Jousef

1 Like

@jj2r could you try re-running your simulation now? There was a fix in the latest release that is related to modeling of no-slip walls when using turbulence models. Looking at your simulation setup, it might affect your results as well.

2 Likes

@dzivkovic Just re-ran one, no change in results. Thanks anyways!

I think a little update might be useful. At the moment, the problem has been identified to be in the boundary layer. The y+ values, which should be between 30 and 300, are currently most of the time around 15. The desired boundary layer thickness has been calculated using this. However, most of the time y+ looks something like this:


As one can notice, the boundary layer seems to be pretty well defined on the back side of the ball, but not so much in the front. If somebody has any idea how to fix this, please comment, help is urgently needed!

 Hello,

   I did some similar test to yours some time ago with ANSYS and my conclusions were similar too.  Now I can't remember the exact factor Cd differs from theory but in fact it did. I'll look for the data this weekend.  If you dont find a mistake in the mesh take a look to the turbulence model that can have some influence too.

   Regards
2 Likes

Please note that OpenFOAM uses a cell-centered formulation of finite volume method. That means that the value you calculate using this refers to the center of the first cell (You need to double that to get the cell height). This maybe explains why your yPlus is 15 on the most of the surface.

1 Like

I doubled it and now the average for y+ is 7.5
Anyways, it appears that my sim wasnt able to model the boundary layer of the more laminar side (=front) of the ball. while the turbulent back was modeled correctly. I’m now trying to sim it using LES Spalart Allmaras instead of K omega SST, but since I don’t know what to use for nuTilda and eddy viscosity my progress has been quite slow? Any help with what I should use for those? I have used 0.01 for nuTilda and 1 for eddy viscosity so far.

   Hello,
      I share with you my results with ANSYS and parameters if it could be of help. My values are a factor 4 under the expected ones, but maybe the data could be useful for you

  See you

Diego

DragSphere

1 Like

DragSphere

2 Likes