SHM does not follow STL geometry well
3 Attachment(s)
I want to mesh a geometry from STL files with SHM. The resulting mesh looks imo fine at first sight. However it does not follow the STL file properly, which causes problem in the simulation, because I need to use cyclicAMI boundary conditions. Unfortunatly I can not post the geometry or pictures of the whole mesh, because the product is still in development. I am trying to provide as much as possible.
The extracted feature edges are represented in red in the images. A major problem I see with the created mesh, is that the boundaries Upper_Ducts and Lower_Ducts have a significantly different number of faces, although they should have the same geometry. This is indicated in the 2nd and 3rd image below. If you have any further questions about the set-up please ask. Thanks for your help in advance! snappyHexMeshDict: Code:
/*--------------------------------*- C++ -*----------------------------------*\ Code:
/*--------------------------------*- C++ -*----------------------------------*\ |
Just taking a quick look at your settings, I see that all of your refinement surfaces are set to a refinement level of 0 0. For example:
Code:
refinementSurfaces // Surface-wise min and max refinement level. Block is needed to define faces. Code:
Right_Ducts {level (1 2);} |
4 Attachment(s)
Hey Dan, thanks for your reply!
As you recommended I changed the following things in my snappyHexMeshDict: Code:
features Code:
refinementSurfaces // Surface-wise min and max refinement level. Block is needed to define faces. Code:
Lower_Ducts Do you have any further ideas to enhance my mesh? Thanks in advance for any help! |
You're moving in the right direction. Most of the issues with meshing with snappyHexMesh, whether that be resolving features or adding layers, can be fixed with more refinement (edge, surface, and volume refinement). To answer some of your questions:
|
2 Attachment(s)
Hi Dan, thanks for your reply again!
So I took a few steps back and oriented myself at the snappyHexMeshDict of the motorBike tutorial. The first huge difference I noticed was, that the inital block mesh in the motorBike example was much coarser than in my case. I now changed it to Code:
blocks Code:
/*--------------------------------*- C++ -*----------------------------------*\ |
I would get out a pen, a piece of paper, and a calculator and determine the size of all the levels that you are using in your mesh.
I would turn off layer addition for now until you have good snapping and then go back and turn that back on. I woudl also make sure that all your mesh quality settings are their default and work on decreasing the base mesh sizing or refinement until you get snapping. Once you have a mesh resolved, simulate and then adjust your mesh accordingly. |
2 Attachment(s)
Hi Dan,
I was able to get a very nice looking snapped mesh w/o layers. What confuses me is that the snapping quality somehow decreases when I'm increasing the refinement level. The image I attached belongs to the following sHMDict: Code:
/*--------------------------------*- C++ -*----------------------------------*\ Code:
features |
Depending on your application, 1e6 or 2e6 is still a small mesh where the cell count ultimately depends on a balance between desired accuracy and computational resources. Meaning, you will mesh, simulation, check your solution,assess, and remesh. Continue this cycle until your solltuion does not "significantly" change or you run out of computational resources. I regularly run simulations on meshes anywhere from 1e6 to 1e9 cells in order to get the results i need. lastly, the sole reason layers where not added is becasue when added, the mesh quality criteria were not met so it was scaled back. Refine more to get small enough cells or try adding a single layer to start.
|
All times are GMT -4. The time now is 15:36. |