[julianschl]

Hey there,

for improving my IC engine simulations I want to connect a block structured cylinder mesh with a cylinderhead mesh created by snappy.

My situation is the following:

So far I'm doing my simulation with a mesh created by snappy and the deforming strategy for piston and valve motion.
The problem is that the mesh quality decreases fast and I have to create a lot of meshes which needs a lot of computing time.

My idea to improve this is to combine a block structured cylinder mesh with the cylinderhead mesh createdy with snappy. This approach would be only suitable for closed valves but it would decrease my computing time significantly beacause I could use the layer addition and removal strategy for the piston motion.

What I've done so far:

I created my meshes. The block structured cylinder with a patch for merging and the snapped cylinderhead mesh with the corresponding patch for merging. (screenshots)

I merged these meshes with mergeMesh and stitched them with stitchMesh.

For using stitchMesh I've tried several options. Default, -partial, -perfect, -toleranceDict and different values in the toleranceDict.
But no matter which options I used the result contained some bad cells according to checkMesh ( non-Ortho, skewness).

So I'm looking for a method to connect these two meshes without changing any of their vertices.
Anyone got some idea for that? I dont know much about sliding interfaces or GGI I just heard about them. Would be one of them a possibility to solve my problem?

Thanks in advance!

With best regards
Julian
CylHead.jpg

liner.jpg

[hjasak]

mergeMesh basically catenates all points, faces, cells from two pieces to make a single mesh. If yo built your mesh in pieces, this is the first step.

stitchMesh will connect all non-matching interfaces using the sliding interface, just like you need it.

Non-orthogonality and mesh quality loss is minimal: in your case, cell size across the interface is very similar. I am happily running 50:1 cell interfaces and it works just fine.

Enjoy,

Hrv

[julianschl]

Thanks for your reply.

Thats what I've done it so far.

I used
mergeMeshes liner_case cylinderhead_cyse

and

stitchMesh -overwrite -partial masterpatch_liner slavepatch_cylhead

It works without any error!

However checkMesh shows an increased number of highly skew faces (20 to be exactly) and in paraview the mesh doesnt look like I expect it to be.

I attached two screenshots. One with activated decompose polyhedra option in paraview and one without decomposing.

It looks like the vertices are stitched which is what I expect that stitchMesh does. But I dont want it to stitch the vertices.

Do I have to set really low tolerance values in a toleranceDict to solve my problem?

/edit:
The default tolerance values:

pointMergeTol : 0.05
edgeMergeTol : 0.01
nFacesPerSlaveEdge : 5
edgeFaceEscapeLimit : 10
integralAdjTol : 0.05
edgeMasterCatchFraction : 0.4
edgeCoPlanarTol : 0.8
edgeEndCutoffTol : 0.0001

I'm getting errors when I try to use low values for point- and edgeMergeTol with the other values as default.


Stitched1.jpgStitched2.jpg

[joshm]

Originally Posted by julianschl

Attachment 31004Attachment 31005

Apologies for being off topic, but they look like the top of some ultra-modern sky scrapers. You'd have to pay architects millions to come up with designs like those!

[julianschl]

I managed to solve my problem.

It seems that the patch of the slave mesh has to be within the patch of the master mesh.
As you can see in the screenshots above the Cylinderhead (slave mesh) had a slightly bigger diameter than my liner mesh (the master mesh). Changing this solved my problem. No more skew faces created by stitchMesh.

[julianschl]

Seems like I should have had a closer look inside my mesh before reporting about a success..

enginetopomesher aborted with thousands of skew and non-ortho faces in the last part of the deforming phase.
It looks like stitchMesh still connects some of the verties inside my mesh like the screenshot shows.

Any idea which tolerance values help me avoiding this?

Stitched3.jpg

[nanavati]

Originally Posted by joshm

Apologies for being off topic, but they look like the top of some ultra-modern sky scrapers. You'd have to pay architects millions to come up with designs like those!

haha, nice one buddy..

[sharonyue]

lamo,

usually I feel monotonous, tedious when I read some threads with no feelings. Then u made me spray rice out of my month into the screen. I really need to simulate how to spray those tiny rices elegantly.