Advice needed - laplacian pressure drop in gas-liquid flow
 

Hello everybody,

I am trying to model the laplacian pressure drop in a vertical y-junction mini-channel with gas-liquid flow. Basically I try to verify that I am able to simulate the correct laplacian pressure before moving to a more complex model.

The numbers (pressure [Pa]) which my model generates are far behind the pressure I would expect. If the pressure drop dp is calculated by dp=pressure(inside)-pressure(outside) I observe ~80 Pa instead of 292 Pa (for a 1mm channel diameter; surface tension = 0.073 N/m; contact angle=0°). So there must be something fundamentally wrong in my setup.

The theoretical dp of 292 Pa is calculated by: 2*0.073/0.0005*cos(0)

I tried it with different mesh qualities (~400k tet mesh right now), a hex mesh and a more tightly convergence (1e-5) but my best result is ~135 Pa for dp.

If anyone could give me an advice how to match the theoretical pressure drop or dig in the correct direction I would be very thankful.

Thanks in advance

Simsey

What is going on with the surface mesh on the cicruclar face in the last picture? There is not clearly shown evenly sized surface triangles from the volume tetrahedrons, is there a section plane there?

Hey,
yes there is a section plane to show the mesh quality as I am unsure if my "uncertainty" is connected to the mesh or my setup.

Normally the mesh looks like the attached images.. The 1st image shows the full mesh, the 2nd image shows the mesh with removed inlets. I'm constantly trying different things to improve my results, therefore it looks a bit different now.

I recommend you try to model a drop in free space with surface tension as a basic model to develop a method which can predict laplacian pressures accurately. In my experience, if you want accurate laplacian pressures you must use hex meshes (tets will not be accurate enough) and you must have a maximum aspect ratio less than about 1.2. This is an extremely high quality mesh, and far more restrictive than the mesh requirement for other CFD models.

Secondly: If the contact line is moving on a wall boundary then you will come across the moving contact line problem (http://www.sciencedirect.com/science...95034915303160). You will find that the motion of a moving contact line on a wall boundary and the laplacian pressures it generates do not converge with mesh refinement. This is due to a fundamental paradox in the Navier-Stokes equations. This means no Navier Stokes solver is going to give you good results for this type of flow :)