Which CFD package fastest and ideal for moving bodies in water
 

Hi guys,

I'm trying to simulate 2 or more fully submerged bodies moving in 3D. I need some advice on what best CFD package/solver/code that offers the quickest solution within reasonable accuracy. Speed is of higher importance rather than accuracy.

I've tried ANSYS-CFX using a adaptive remeshing approach but it's taking too long due to the remeshing; the ratio of computational time to real time about 1000:1. I hope to apply the same approach in OpenFoam but I think the reduction in time won't be much of a difference.

I've read about the immersed boundary method which looks promising but I can't find any info on its speed (is remeshing needed for this?). I also have been suggested to use Flow3D but I'm not sure if it will solve faster than the ones mentioned above.

Is your calculation single phase calculation, that is no VOF etc??

Hi Arjun,

The calculation will be single phase (fresh water); no free surface

then if you can live with windows version, then i can share my immersed boundary solver.
I am still in middle of testing and improving it so it is kinda beta version.

Short discription is here: http://inavier.com/
My ramblings about IBM are here: http://inavier.com/currentissues.html

The solver is reasonably fast, and should be faster than CFX you were using.

Dear Arjun,


That's great to know! I'm happy with a windows version. My email's 'leongzq@gmail.com' if it not meant to be released into the public domain yet.

I'll be more than happy too in providing feedback and results of my simulation for your development of the solver.

Best wishes,
Darren

can you give me any other email than gmail. The thing is gmal does not allow exe files in attachment. It does not accept emails if it has exe files in it.

Further i need to know what type of translation motion your bodies will be doing. I support rotation at tthe moment but this weekend will try to add translation too.

Then you can try it , we can fine tune it then.

If accuracy is not an issue, could you please explain what the objective of your simulation is; i.e., what are you trying to get out of the simulation?

Also, what problem are you actually trying to solve - geometry, Reynolds number, etc? Are the objects moving in prescribed motion or are they responding to the flow (i.e., fluid-structure interaction)?

adrin

@Arjun:
Thanks, Arjun. You can try cc to 'darren_ll@hotmail.com' and 'darren_ll@mailcity.com' and see which works. The translation/motion will be prescribed eventually in 6DOF, e.g.
-For sphere it would in X,Y,Z to simulate motion towards the aft or side proximity of another sphere similar to a docking simulation
-For eclipse as a possible next geometry, this would include rotation to simulate directional heading
-Other geometries will be evaluated once the capabilities of the setup are stable.

@Adrin:
The motions are prescribed as mentioned above. The data I'm interested is the hydrodynamic coefficients of the bodies. I am aware that attraction and repulsion forces occur in close proximity. The objective at the moment is to evaluate the trade-off in speed and accuracy of different approaches available for such simulations. It is hoped that accuracy is within 15-30% compared to experimental results.

Okey I created a page for ibm version.

Anyone one can now download the test version. Please let me know if you can run the solver , because sometimes due to some visual studio things it might not work.

Here is the link http://inavier.com/ibm-inavier.html
Page contains the test case for which images are posted.


Further, i forgot to mention one thing before, that for ibm the meshes needs to be very fine. So in case you are looking for speed, ibm may not work out. But it would be worth a try.

Thanks, Arjun =)!

It might still prove faster than remeshing. Will keep you posted when I get the solver running successfully.

Edit: It's running well with no errors =) I'll wait for the results and get back to you.

Might want to look at: XFlow

I have added the translation also. I will put it online as i get time. Remember you will need small time step for body movement. Not due to courant issues but because of accuracy purposes.


Thanks for the link. I will try to look what they are during. Feel like a moving particle method.

Don't have any experience in it though, but they come from a cg background with their RealFlow product so I think it might be in line with your speed over raw accuracy. Let's just say it might be great for trend data (ok maybe all cfd is like that)

Okey I have updated the solver on IBM page. It now supports translation too.
Added a description of solid input file and how to specify rotation and translation.

Note: Today evening i will upload the version of Dravvya that has solid marking model. that you can use to set up calculations.



Yes it may be faster than normal cfd solver. Or it may be slow. Nothing specific could be said until it is tried.

Thanks, Arjun.

I've restart the previous rotational sim to run on a better pc yesterday and has yet to complete.

I appreciate the translation model! Will get back to you when i get the results out.


Kevin, I'll contact XFlow for more info but it doesn't look cheap :D

I forgot to mention one thing. When the solid marking is done , solver reports pressure forces. You should change the sign for correct forces experience by the solids. (this is due to sign of surface faces).

so if solver says force = -15N then force experienced by body is 15N.

As with most CFD codes the graphics are splendid!
Unfortunately there are no extensive comparisons of predictions with experimental measurements on their page.

Has Xflow shown its capabilities in proper, calibrated studies such as those which were submitted to Gothenburg 2010?

CFD: Color For Dollars.

I had same doubt about XFlow, if we can get some validation of standard problems for CFD than we can have more confidence, anyone knows someone working for X-Flow, I will like to communicate with them.

There are several existing "standard" problems in ship hydrodynamics. Many CFD suppliers do not present results for these cases because their codes give very bad predictions, or they have been massaged to give good agreement for advertising purposes and can't be reproduced by independent users.
Shame on them for bringing CFD into disrepute!

I uploaded the dravvya.exe that contains 'immersed-boundary-model'. Please download it from IBM page.

You can set up your calculation as normal. But when in the end when exporting solver input file please select 'immersed-boundary-model' from list of general models and include that.

And you have to create the solid marking input file. (whose name should be same as i use in example).



Note: The list of general model shows vof model too, but it is not available in inavier.exe that is uploaded. I am still trying to implement it so it will not work even if selected.