Discussion Forum

Hi

it could well be a RAM issue, as fluidics often require dozen of Gb of RAM and on a 32 bit machine you are limited to 2Gb per process. But it could also be from a geoemtry + mesh difficulty, alldepends on the geoemtry shape

--
Good luck
Ivar

Dear Mr.Ivar

Thank you so much, I will try my best.

Thanks a lot and best regards,
Kittatat

Dear Mr.Ivar

I have just tried to simulate my model in 2D instead of 3D, but now it showed this kind of error.

"Invalid boundary layer specification." with Geometric entity level at only one point.

I cannot attach the files to show you my work, I am not sure that all errors came from my the geoemtry shape or not.


Thanks a lot and best regards,
Kittatat

Hi

that is probably coming from your geometrical shape. COMSOL applies by default boundary layer to all no-slip walls, but sometimes these have trouble joining over open corners. One way is to try a manual meshing, and perhaps apply boundary layers (a two step operation, selecting the domain of the fluid (boundary layer side) + the boundaries in the subnode where to apply the boundary layers) in several steps. Sometimes one must leave out a small portion, where one can mesh finer locally to get around the numerical issues.

By the way the meshing algorithms have been improved in the newer versions

--
Good luck
Ivar

Dear Mr.Ivar

I can simulate 2D model right now but my 3D simulation has this kind of error as shown below.

Failed to create boundary layer mesh.
-Detail: Could not shift boundary node at all

It occured when I choose Element size from "Fine", when I chose "finer" it showed

Boundary mesh must consist of triangular only.
-Domain: 1
Try converting to triangular on these faces.
-23, 39 (both are top circular faces from two cylinders)

Then I change from "physics-controlled mesh" to "User-controlled mesh" and used "Swept" mesh from one circular face to another face of both cilinders with prism face meshing, but it said that "Source and destination faces must be specified." and cause an error in that damain. Did I do something wrong?

Sincerely yours
Kittatat Suphakulwattana

Dear Mr.Ivar

If I want to simulate the flow of gold nanoparticles in my channel, which Physics mode and gold nanoparticles properties should be used.

Normally you reply every quastion very fast, but this take too long for you. Are you OK?, I am a little bit worry about you.

Sincerely yours
Kitatat Suphakulwattana

Dear Mr.Ivar

I really worry about you right now, are you still fine? If you still ok please help me.

Sincerely yours
Kittatat S.

Hi there
here is the solution for you:

Prelude:
I to have encountered that same error so many times when i modeled 3D geometries yet almost never when modeling 2D geometries. Why?
COMSOL is trying to patch over the sharp corner diagonally in order to better/more smoothly mesh over the geometry in a way that will enable better representation of flow gradients over a corner.
all that is excellent and another reason why i have so much appreciation for COMSOL developers as they are actual Mathematics/Engineering/Physics professionals.
however, the problem is that doing this sort of "patching" is quite simple in 2D but very difficult in 3D, unless the geometry is very specific and of the kind that will not bring the meshing algorithm into collapse upon itself (due to impossible clustering of such elements).

(remainder: the higher the physics-gradient present at geometrical location, the denser the mesh you will need to "capture" and depict that phenomena using FEM, for reasons related to Element order).

Solution:
all you need to do is:
Ver 4.2a ==> uncheck the split boundary layer check box (see pictures attached)
Ver 4.3B==> choose the trimming option from the drop menu /Engineering

works for me every time :)

Excellent meshing tip.

Excellent meshing tip.

Excelent meshing tip