[SaufiSangi]

Good morning everybody,
I am working on the simulation of droplets evaporation using the VOF method, modifying the already existing "interFOAM" solver. I started in absence of gravity, thus no surface tension was necessary in my calculations.

Now I am trying to activate gravity. My droplets have a diameter of 0.5-1 mm, and surface tension effects are definitely dominating. As you (who work on surface tension dominating flows in OpenFOAM) know, spurious currents arise and destroy the whole droplet after few time steps.

Now, I read ALL the post and threads concerning this problem in interFOAM, from 2009 to 2016. I found very interesting ideas, which I tried to implement, but none of them is really satisfactory for such small scales. The best one is the one proposed by Ali Q. Raeini et. al. [2012], but it still gives me some problems.

All of the "definitely solved" problems in VOF method concerning parasitic currents are based on the reconstruction of the interface (e.g VOF-PLIC), which I think it's what Fluent is using, or using marker functions ecc.
From what I understood these methods are quite difficult to implement in OpenFOAM, which already uses a compressive scheme for the interface reconstruction. I would be really scared if I had to implement geometrical methods in OpenFOAM, it does not seem to be easy at all, even for skilled programmers.

So, what's the situation now in 2017 about spurious currents in interFOAM? Are there any improvements/news about them? Is there anybody currently working on this major problem of interFOAM? As many people here, my work is not related to this specific problem, but I really need to solve them in order to proceed with my research, that's why I'm asking for your help

Thanks guys for you attention, any suggestion would be highly appreciated!!
Sangi

[liquidspoon]

Hi Sangi,

I find myself in the same boat, and would be very happy to know if there are any robust solutions to this issue.

As a note, I have tried implementing the Raeini surface tension model (not the whole solver, just replacing the surface tension force calculation). It seemed to help reduce spurious currents, but it also seemed to damp out some real interface dynamics.

I have been meaning to try out the isoAdvectorFoam solver on some test cases. It was developed to address other issues, but may help out with the surface tension force calculations as a side effect (no proof of this, just an idea).

I think one part of the issue is the diffuse interface in interFoam. The surface tension forces are distributed across the interface weighted by grad(alpha) (basically a smoothed heaviside function). As a result, a large component of the force is applied to the low-inertia gas side. Typically you see the highest velocities on the gas side. This is still not the complete picture, because large currents can also appear on the liquid side if the surface tension effects are large enough.

[SaufiSangi]

Hi Alex, thank you for your answer
What are you working on exactly?

From what I know isoAdvector does not solve the problem of spurious currents, there is a thread about that (don't remember where though).
Yes, the highest velocities are in the gas-side, but even the density correction proposed by Brackbill doesn't seem to be enough. This is a major problem for me, I am working with 1 mm liquid droplets, which are extremely difficult to stabilize. I need a smart way to rapidly solve this problem, since this is not the main topic of my work, I don't want to loose too much time on this.
So far I implemented the Raeini model and I am trying to couple it with a Level-Set method, this can be a way.

Keep in touch!

Sangi

[SaufiSangi]

Here it is by the way: the thread about isoAdvector

IsoAdvector release

[liquidspoon]

Hi Sangi,
I am also working on capillary scale flows (small droplets, bubbles/Taylor flows in minichannels).

I tried out this coupled LS-VOF solver a while back: https://bitbucket.org/nunuma/public . It was very nice for the developer to release it, but I had mixed experiences. Maybe it could be worth a try for you?

Depending on what you are doing, you may be able to use the extend code interTrackFoam (https://github.com/Unofficial-Extend...interTrackFoam). This uses a "body-fitted mesh" approach. It may not work if you are interested in large deformations, or coalescence, etc.

To be honest, I don't have any great suggestions at this time. I ended up in this thread by asking the same questions you are.

[SaufiSangi]

Yeah I am also having not the best experience with LSM-VOF coupling, but I try to go deeper in that.

It's the first time I see this "interTrackFoam". Is it still a VOF method or it's something like a Front-Tracking method? Did you try it?

Thank you very much anyway

Sangi

[liquidspoon]

I have not tried interTrackFoam, but it is interesting. It updates the cell vertex positions so that the mesh is coincident with the interface. It seems like this would be a good thing because the pressure jump could be imposed directly across a cell face.

[eric]

It is very difficult to avoid spurious currents on general, unstructured meshes. If you are only interested in the drops/bubbles themselves, I would recommend using Cartesian grids, which makes it much easier to remove the spurious currents. See for instance the level set method with the ghost-fluid method to handle the discontinuities.

[SaufiSangi]

Thank you Eric for your answer,

Yeah I noticed that. It would be very nice to use an Adaptive Mesh Refinement at the interface, but most of the times OF gives unstructured meshes, which of course enhance spurious currents.

I've seen some papers about the ghost-fluid method, but I cannot understand it properly, especially in OF environment. Do you have some nice articles about it?


Sangi

[santiagomarquezd]

Hello, I know three projects on VOF-PLIC in FOAM, the ones from Maric, Cifani and myself. The whole thing with spurious currents is the calculation of curvature. Even if you initialize your VOF field analitically the representation with averaged quantities (alpha field) leads to a loss of information regarding the shape of the interface and is worst for first derivative (normal) and second derivative (curvature).

The computation of normals and curvature in FOAM is pretty basic and not mesh convergent, can't be used in surface tension dominated problems. Today the best method for curvature calculation is Height Functions (see Popinet's review in ARFM), two good codes from him are Gerris and Basilisk, both with structured meshes and AMR.

Curvature with HF was extended to 2D unstructured meshes by Ito. I implemented a version of Ito's method for slab 3D meshes and showed to be convergent in normals and curvature for the proposed tests (static) but had regular performance for dynamic cases,

https://youtu.be/3uyJYuKx5qc

After that I extended the code to 3D unstructured parallel cases but it is still experimental and with very poor performance.

Doing VOF en general cases lacks of closed formulas for interface positioning except for hexahedrons and tetrahedrons, the same for intersections (flux computing) and curvature. Without closed formulas or using split VOF the solution is very expensive. We use here narrow band techniques (written by Cesar Pairetti) to improve the performance.

At this point we're considering returning to structured meshes after publishing some ideas on VOF method errors.

You have alpha smoothing techniques, Raeini for example, but are limited to moderate surface tension effects.

Regards!