[aerospain]

Hi everyone,

I'm trying to mesh a semi-sphere. Thus, there is a sharp angle at the trailing edge of my geometry which I need to capture nicely in order to replicate some experiments. The phenomena I'm studying is heavily dependent on geometry definition.

I have tried with several parameters in the snappyHexMeshDict file, and what you see in the files attached is the best I could get.

Any hints, please?

[HakikiCanakkaleli]

Hi,

I had a very quick look:

1. Could you please put a snapshot of the extracted 'features'?

2. Increase

nFeatureSnapIter   5;

several orders of magnitude? Say 30?

3. Do you really need `features level 11`?

Sorry for very quick observations. I will try to have a look more thoroughly when I have more time.

[aerospain]

Thanks for the reply and questions Hakiki,

I've answered your questions below. I am trying other approaches and will post soon my progress.

cheers!

1. Could you please put a snapshot of the extracted 'features'?
The first snapshot, where I show the geometry defined by the STL file, includes the extracted features in red, i.e. the circumference around the base edge.

2. Increase nFeatureSnapIter 5; several orders of magnitude? Say 30?
It doesn't help, I still get a "snatched" edge.

3. Do you really need `features level 11`?
I just went "buzzinga" with this level trying to capture the edge perfectly. My objective was to tweak other variables in that Dict file after that was achieved.

[HakikiCanakkaleli]

Hi,

Could you please turn off 'addLayers' and give a shot?

In addition, please reduce
- Explicit feature edge refinement level (considerably)
- Refinement region and surface levels
- nCellsBetweenLevels

and increase several orders:
- maxGlobalCells
- maxLocalCells

If possible, it would be helpful if you share the output streams from the commands.

[aerospain]

Hello Hakiki,

I have followed your steps. Actually, I had done so already with some of them before you proposed them ;-).
Still no luck!

addLayers has no effect since this command appears after snapping, and it is at this point when we want the edges to be reproduced correctly.

I have reduced and increased the values you mentioned with no effect on the mesh.

I am already looking in using some external software to create the mesh and import it afterwards. Such a 'simple' operation shouldn't be that hard, specially when any relevant geometry, even in the most basic experiments in academia, has sharp edges.

thanks for your time
C.
ps: I'm attaching the logs from each step, i.e. blockMesh - surfaceFeatureExtract - snappyHexMesh - checkMesh

[HakikiCanakkaleli]

Hi,

Could you please surfaceCheck your STL file and qualitatively inspect this? Could you please also check wheter STL file and feature's coordinates coincide?

It definitely shouldnt be a problem for snappyHex!

[aerospain]

Sending zip file with logs from surfaceCheck and surfaceFeatureConvert in order to plot both, the stl and the eMesh files in paraview.

And, two zoomed in views of the geometry, gray surface and blue surface mesh, with the Feature, in white, showing matching at edge.

The look of the base of my geomtry is the one I get from CAD program when exporting to STL file. But, the log file shows the geometry is closed. I remember I did some fixing a couple of weeks ago, I don't remember how now, sorry.

[aerospain]

Dear Kutalmis,

I came back to snappy and trusted your faith in the algorithm Actually, if people were able to do it, why wasn't I able? So, I've dedicated today to find the solution.

I have found it by declaring implicitFeatureSnap true; This way, the rear edge of my hemisphere, and future projectiles, is neatly defined. Right now, the size of the cells on my body's surface are defined by a refinementRegions{ ... } in mode inside;, and I cannot convince SHM to follow the values set in refinementSurfaces{ ... }, and feature{ ... } when explicitFeatureSnap true;.

My workaround for that has been to define a new refinementRegions{ ... } closer to the surface, but this method doesn't allow me to have higher cell size in areas of the body surface where gradients are not strong (to save some mesh size).

Now, my problem is to be able to define a finer mesh around the rear edge of the hemisphere, which I think is related to the cell size around that region.

I'm attaching a snapshot, logs and dictionary.

thanks!
C.

[aerospain]

Dear Kutalmis,

Let me share some initial results with you.

Hemisphere colored with wallShearStress (poor mesh quality near the rear edge)
Streamlines, both 2D from line source and 3D from point source, colored with U
and a z=0 cut plane to show the (very poor) boundary layer mesh.

I am attaching the snappyHexMeshDict too in case you may be able to help me with my next challenge. To be able to increase the level of discretisation near the rear edge. Right now, the fastest solution that comes to my mind is to define a thin cylinder to enclose a region a bit downstream, a bit upstream, and a bit above, the edge, and increase the level of discretisation inside. But, I will be increasing the cell count at the hemisphere rear surface.

Once, I resolve that small challenge, I will be able to perform a mesh convergence study properly.

thanks!
Carlos

[einstein_zee]

Originally Posted by aerospain

Dear Kutalmis,

Let me share some initial results with you.

Hemisphere colored with wallShearStress (poor mesh quality near the rear edge)
Streamlines, both 2D from line source and 3D from point source, colored with U
and a z=0 cut plane to show the (very poor) boundary layer mesh.

I am attaching the snappyHexMeshDict too in case you may be able to help me with my next challenge. To be able to increase the level of discretisation near the rear edge. Right now, the fastest solution that comes to my mind is to define a thin cylinder to enclose a region a bit downstream, a bit upstream, and a bit above, the edge, and increase the level of discretisation inside. But, I will be increasing the cell count at the hemisphere rear surface.

Once, I resolve that small challenge, I will be able to perform a mesh convergence study properly.

thanks!
Carlos

Sorry if it is irrelevant to your question but have you given a shot with cfMesh already?

[HakikiCanakkaleli]

Hi,

I will take a stab at it when I will have some free time.

Thanks for your understanding in advance

[aerospain]

Originally Posted by einstein_zee

Sorry if it is irrelevant to your question but have you given a shot with cfMesh already?

cfMesh not ready yet to work in Ubuntu 17.10. They've told me to give it another shot in a couple of months, they will let me know when new version is ready

Thanks!

[aerospain]

Originally Posted by HakikiCanakkaleli

Hi,

I will take a stab at it when I will have some free time.

Thanks for your understanding in advance

Thanks to you for your time!!!! I'll keep trying other things and sharing them.

C.

[einstein_zee]

Originally Posted by aerospain

cfMesh not ready yet to work in Ubuntu 17.10. They've told me to give it another shot in a couple of months, they will let me know when new version is ready

Thanks!

Are you talking about cfMesh Pro ? cause I don't think using cfMesh itself has anything to do with your linux distribution. You can take a look here "https://cfmesh.com/cfmesh-now-adjusted-to-latest-version-openfoam/" which says that the opensource cfMesh is already adapted to OF4.0. And with a little bit of tweak you might also compile it with OF5.0/1712. This means you can simply download it here "https://sourceforge.net/p/cfmesh/code/ci/development/tree/" and use it

[aerospain]

Originally Posted by einstein_zee

Are you talking about cfMesh Pro ? cause I don't think using cfMesh itself has anything to do with your linux distribution. You can take a look here "https://cfmesh.com/cfmesh-now-adjusted-to-latest-version-openfoam/" which says that the opensource cfMesh is already adapted to OF4.0. And with a little bit of tweak you might also compile it with OF5.0/1712. This means you can simply download it here "https://sourceforge.net/p/cfmesh/code/ci/development/tree/" and use it

Thanks H.,

I've stepped into that page a couple of weeks ago. At this time, I don't have the time to learn how to tweak source codes before compiling them. Too much on my plate these days

But, thank you very much for the links!!!

C.

[aerospain]

Progress update ...

I have decided to use two searchableCylinder to refine regions of my geometry where to expect higher gradients, i.e. front/stagnation area, and rear 90º edge.

Already running mesh convergence, and moderately happy with results.

I'm still concerned with the shape of the BoundaryLayer mesh next to the rear edge. Next step: use relativeSize false. This way, I may prevent shrinking the BoundaryLayer cells. Warning!!! I must be really carefull with first/last cell size, or total BL-layer size. I should NOT forget about surrounding cells level of refinement

Any thoughts???

Attaching images, and example snappyHexMeshDict.