SimScale CAE Forum

How do I set Level 0 cell size in Hex-dominant (only for CFD) meshing nowadays? OR where did Base Mesh Box Discretization go to ❓

#1

Posts will appear below soon…

#2

I need to specifically set Level 0 cell size, how do I do this :question:

It appears to me that this is not possible to do any more, it looks like level 0 size is set automatically to 1/4 of the minimum range of the three sizes of the geometry in the three primary axis direction.

This is very bad for CFD verification projects that use a small span airfoil in a large non rectangular BMB (hence the need to use this algorithm because HEX-parametric does not do internal meshes :frowning: )

The mesh sizes are ridiculous for these cases… since the far field cell size can not be greater than 1/4 the span.

I want to make a HEX mesh with cell density gradient like this (it is a non rectangular BMB so HEXparametric is not an option…):
BMBsize

And all I can get is this (this view is not taken from one of the 3 axis directions… notice how thin it is in the Y direction, hence the VERY fine farfield cells, where I do not want it fine…):
HEXbad

So, please give use the ability to specifically set Level 0 cells sizes in Hex-dominant (only for CFD) meshes again…

#3

Hi Dale,

we have a feature request for that in our product backlog where I can link your request. Let me add some insights though:

  • The hex-dominant mesher did also not all allow to modify the basemesh resolution in the past. It was intended to be an easy to use tool specifically for unexperienced users, hence the level of automation of refinements etc.
  • The idea was always to improve/extend this algorithm to in the end make the parametric version of the mesher obsolete and let the user decide on the level of details that he wants to provide, essentially working his way down from full automation to more advanced properties. The first features on this roadmap would be the basemesh resolution as you mentioned and an opt-out of the automatic thin gap refinements that we apply.
  • Currently we are focusing our main efforts in meshing onto the new mesh preview, which will become our standard mesher, which leads to these improvements being a bit pushed back
  • The parametric mesher can also create internal meshes, you just need to put the point inside of the region of interest. We detect automatically in which region the point is placed and map the entities of CAD and mesh, so it should be even possible to use the CAD-based simulation setup in that case.

As you mentioned that the parametric mesher doesn’t help in your case, could you explain that a bit more or share your project. Maybe I misunderstood sth.

Best Alex

#4

Sorry, I just assumed from the current help docs that you were previously able to set Level 0 cell size…
.

Aha, if the HEXparametric will do internal meshes, then I am good to go… I guess I should have known that but I wanted to try the simpler algorithm…

I still think the ‘HEX-dominant (only CFD)’ algorithm should let you set Level 0 cell size…

I just figured out a VERY simple way to do that, just have the the code use the ‘Sizing>Manual mesh sizing>Maximum edge length’ as Level 0 cell size, what do you think :question:

#5

Hi Dale,

that works well if you are fine with equilateral basemesh sizing. You can’t adjust x, y, z separately, not sure if you need that in your 2D case.

Best Alex

#6

But equilateral sizing is preferred in almost all cases, is it not? (I would say ALL, but maybe there is an exception)

And anyways, it has already been decided that equilateral sizing will be used for this simple mesh algo. So the equilateral issue is not relevant as to using Maximum Edge Length as Level 0 cell size …

With the current inability to make oblong cells, using Maximum Edge Length as Level 0 cell size is way better than the non-logical/undocumented approach that is used now…

#7

I am not wanting a 2D mesh, just a thin 3D because I want to verify with the 3D TETforCFD mesh.

I am also trying to use a HEX 3D as another comparison that does verify better than TETforCFD…

#8

I see! Would be also interesting if you want to give our new mesher a try or - in case not - why?

All the best for your validation project!

Best Alex

#9

LOL, I am using it, I have coined it TETforCFD meshing because ‘new mesh preview’ just does not work for me :wink:

#10

I think we do not understand each other because I am having trouble trying to figure out how to use the Hex parametric algorithm on a geometry that already has a curved surface BMB as part of it (see post 1 images, there is NACA0012 airfoil running through that curved BMB, towards the center where the mesh is very fine… ).

Here is my attempt that only generates a 1 cell mesh ;(

I have placed the material point inside the space which the air will flow through.

I assume that the rectangular BMB box would have to at least totally enclose my complete curved BMB geometry, which it does.

Does ANYONE have any idea why I only get a 1 cell mesh (with x0.25m, y0.5m, z0.25m faces, x and z are their Level 0 size and y is 2 times yLevel 0 size) ?

I hope I am not being stupid from lack of sleep here :wink:

OR, is what I want to do even possible with HEXpara (my simple google searching did not help).

Yaw Simulation for a fsae car
Yaw Simulation for a fsae car
#11

So you don’t have to go to the project and see the problem, here is the geometry with odd shape BMB as part of the CAD file, designed for internal meshing:

image
image.png545x828 17.4 KB

.
.
And in this mesh attempt, this is the mesh that is created:

image
image.png1659x926 126 KB

.
.
.

Which is not the ‘eye’ shape expected :frowning:

What to do if geometry does not appear in your mesh - try the Feature Trick
#12

@DaleKramer: by curiosity I did set HEX mesh for parametric simulation on my ‘eye’ shape.

image
image.png1100x1092 336 KB

I verified first the Material Point (inside the domain)
I did edit the BMB resolution:

image

I DID NOT touch the BMB size, as it was automatically sized, but I do not think it is critical.

So perhaps your 999 faces of airfoil are causing havoc?

Example is here: EyeSimulation_HEX

#13

Thanks, I will give that theory a try.

I have discovered since that Step file was exported from Rhino, that If I want faces to be combined in the Step file, they have to be extruded from lines that are more than first order (1st order are straight lines)… I have lately been exporting higher order surfaces as Step and I would only have 1 face instead of 999… (but in theory it should do 999 faces too :wink: )…

I have also found a way to nicely group my faces in the step file buy selecting the faces I want in a Step file group and making a Rhino block out of them before the Step is exported…

I will test that soon and report…

#14

Although the answer to to post topic was in Post 3 (the answer is ‘You can’t set Level 0 size of a Hex-dominant mesh, although it is on the feature request list’)

I was able to get my mesh to work using a Hex parametric mesh to appear as you see below:

I will leave it to the coders and mathematicians to figure out why this Feature Trick made my ‘eye’ shaped BMB and geometry appear, but it did…