I am interested in discussing the use of CFD for modeling of breaking water waves. Using Fluent, I have used a Volume of Fluid (VOF) model along with 1st, 2nd, and 3rd order solution schemes. Simulation were performed using laminar and standard k-e models. The smallest grid spacing has been 2cm X 1cm (the domain is approximately 2.5m X 23m).

My main concern is "energy" dissipation. The waves lose energy quickly (numerical dissipation is thought to cause this).

Has anyone had succesful, physical results? What sort of schemes have you used? Have you tried adjusting the constants in the k-e equations?

Thanks in advance for any advice.

Before anyone asks, yes, the solution is transient.

1) your schemes may be good enough provided your mesh is fine enough ...

2) be sure to use a time step that ensures a good temporal resolution

3) just forget the turbulence model, at least at the begining; running it laminar will allow you to identify the problem more easily

regards

why not just leave it laminar throughout and obtain a DNS simulation?

Can't you just turn the energy dissipation down on the model you are using?

Eric: What kind of test case are you running? Is it the Fluent 6.2 or 6.3 version that has been used?

Now here is the deal, Fluent 6.2 uses PLIC based geometric reconstruction schemes while the new 6.3 has front capturing method (CICSAM) completely different than PLIC (front tracking ! methods).

I have coded CICSAM etc and found that front tracking schemes work really nice to get all the breaking collapsing, colaescing of droplets, WAVES etc !!

I have seen lot of dissipation using Fluent's PLIC scheme. The interface is just isnt good enough. Smoothens rapidly !

Now, if you were using Fluent 6.2 I would suggest you play with the pressure velocity coupling. I have seen a huge variation in the interface computation using Fluent just by modifying the Pressure-Velocity Coupling.

Turbulence details, later ! Really, Laminar should give you more breakup and interfacial activity (remember no turbulence..no additional viscous effects !)

Reducing time steps is not a great idea in the sense that you shall be marching slower but without much variation in the interfacial activity. Try different p-v coupling and ofcourse, As I have done sometimes, do adaptive meshing based on gradients of VOF that would work just fine too.

Let me know how it works out for you.

CFDtoy

CFDtoy, I was having trouble getting my adaptive meshing to work. Do you know if having rotationally periodic boundaries should affect the ability in particular of 'dynamic' adaptive meshing?

thanks Phil

I have had some problems combining periodic boundaries with dynamic meshing. Check Fluent manual. I guess I have seen some warning suggesting similar stuff ..Not to use periodic with dynamic meshes.

Thanks

CFDtoy

Thanks for the replies!

I am running Fluent 6.2.

With regards to laminar solutions, I have run many in the past. I notice simiilar dissipation as in the k-e model, albeit not as severe. I am able to create breaking waves in this model, although only gentle spilling waves. My research is focused on the turbulent energy dissipation rate, so I am interested in developing a good turbulent simulation.

I have reduced the time step from 0.002s to 0.001s, and I have set turbulent energy dissipation and turbulent kinetic energy at the wave generator to 0 (although that doesn't seem too physical). The solution will take about a week.

CFDtoy: The simulation P-V coupling is PISO. I have not tried SIMPLE or SIMPLEC. Would PISO give more accuracy given it has velocity and pressure satisfy momentum during the solution process?

I have not had a chance to try grid adaption yet.

Update:

Smaller time discretization does not change the solution.

Can anyone comment on mesh size and numerical dissipation?