[mcfdma]

Hi there.

I know this question has been asked before but would be grateful if we all including myself can get some new opinion regarding which is better Gerris or OpenFOAM since the last time the question was asked (back in 2012).

The link to the previous forum
(Gerris or Openfoam?)

I am working on simulating electrohydrodynamic atomization processes (electrospray)and my area of interest is Taylor cone and jet breakup.
I wanted to know the following:

1. I know both methods employ VoF method, which one is computationally faster?

2. I have read articles on CLSVOF (Coupled Level Set and Volume of Fluid) using OpenFOAM. Can it be used for GERRIS?

3. I know OpenFOAM can work with unstructured grids, can GERRIS do?

4. I read both methods have adaptive mesh refinement, is it true?

5. I am a novice in this area, which is easier to learn? I am familiar with Linux commands.

6. Which support is highly reliable and quicker in responding?

Thank you in advance.

[ssyadav]

Dear mcfdma

I know its little late, you might already have gained experience with EHD and with Gerris flow solver and OpenFOAM. Still, I think its good if we share our experience. Here is mine:

1) I do not know about the computational speed as it depends on lot of other factors like Octree vs normal meshes, Parallel algorithm, Data structure etc.

Gerris uses geometric VoF whereas OpenFOAM uses algebraic VoF. Accuracy wise, Gerris is still the best one can get for incompressible two phase flows (Only its cousin Basilisk flow solver can match it!). Spurious velocity currents are small. OpenFOAM uses MULES scheme for the advection of volume fraction (its algebraic VoF). It computationally better but accuracy wise not so good. There is isoAdvector scheme recently developed for OpenFOAM, see here: https://github.com/isoAdvector/isoAdvector

It is a geometric VoF based scheme for OpenFOAM.

2) CLSVOF is not available by default in Gerris, one has to code it. But the source code of Gerris has a steep learning curve. Prof. Popinet has implemented sort of classes in native C language for object orientation and high execution speed.

CLSVOF will loose its core capability with algebraic VoF, i.e., mass conservation + better accuracy for surface normal, curvature and hence accurate surface tension forces. So CLSVOF with algebraic VoF in OpenFOAM will do little to improve accuracy.

Yes, CLSVOF implemented with isoAdvector in OpenFOAM will be much better.

3) Gerris works with Octrees and Quadtrees respectively in 3D and 2D domains. Local mesh refinement with automatic load balancing is really a good strength of Gerris but parallel running efficiency and speedup is not so good when large number of processor cores are used. OpenFOAM works with unstructured meshes but external libraries based on Octree and Quadtree meshes have been made available, see here
https://github.com/ElsevierSoftwareX/SOFTX_2018_143

4) Answered in 3

5) Its a tricky question! If accuracy is what you want then go for Gerris or Basilisk.
A lot depends on how much OpenFOAM source code you know. OpenFOAM has a wider scope but Gerris is really good at simulating incompressible two phase flows.

6) OpenFOAM has a much wider user community compared to Gerris, so chances of getting help are much higher. Although project Gerris seems stalled with no recent developments but you can consider using Basilisk. It is currently under active development.

I may be wrong / outdated in some of the above points but I hope these will be useful.

Regards