improving meshes

Tipps & Tricks ... / How to ...

Moderatoren: ccad, mz15, auroraIco, Lehrstuhl

Antworten
frans
Import
Beiträge: 23
Registriert: Do 6. Sep 2012, 21:03

tetgen meshing alternative

Beitrag von frans »

hello,

i'm using tetgen to mesh my model, but it creates irregular mesh on a symetrical part and also the size of the mesh is not controllable.
tetgen decides on it's own what size the mesh will be for thin walled parts of the parts,
On the thicker parts of the part it's controllable with the meshsize option during meshing.
I also noticed depending of the volume of the part the meshsize changes on the thinner parts, and so also the final results.
all this makes it very difficult to create a good set of results.

is there another mesher that's better controllable for the thin walls of a part?

looking forward to your suggestions

regards Frans
TAH1712
Mesher
Beiträge: 30
Registriert: Fr 24. Jun 2011, 14:13

Re: tetgen meshing alternative

Beitrag von TAH1712 »

Hello Frans.
edit: Fran, you particularly mention 'thin walled' parts. If you're referring to sheet metal parts i understand
it is normal to extract 'mid planes' and use a surface mesh without 'volume'. I can't say exactly what Aurora can do in this situation, you have to see the theory manual -sorry. However, follow what is said below for better meshing.

Well, you're not the first to scratch their head with the meshes provided using Auroras Tetgen meshing.
Aurora needs a 'good quality' .stl or .stp. and in the manual there are hints as to CAD output parameters to assist.
My software didn't have the output flexibility so i experimented with changing the facet accuracy up / down
globally and changed facets on individual faces etc. all with some limited success. Generally, i got better ( but not ideal) results by viewing the CAD part in 'wires' and changing the facet accuracy on a face by face process and saving/inputting a .stp file into Aurora.

ALL THIS CHANGED when user 'selopez' came up with a procedure that is extremely reliable for a good Aurora input .stl and it also assures a fast FEA mesh using Tetgen of very complex geometries.

Please goto ''forum/V2 freeware/'How to' and choose THE NETGEN/TETGEN CONNECTION by selopez (at page bottom).

open up document ' Glrt19...etc - message will say it cannot be found here ,then asks 'open in new window?'
accept this. ( you may need a google account for this to work - i did!)

Then the document of selopezs' opens up. He has made a tremendous effort to understand what are the
problems with poor quality .stl's / .stp's and why they sometimes crash or freeze Aurora. :evil:

It is a long document and i do not understand it all. HOWEVER, there is a procedure at p11 chapter 2. Optimization of T16 meshes from .stl models (POM).

This outlines free software called NETGEN ( inc. link) that allows you to import .stl / .stp / .igs and visually ( and automatically) analyse them for errors, autocorrect errors and many options for refining meshes globally or individual faces to say improve a facetting on a round / sphere / fillet etc. It can then output an assured high quality .stl which when loaded in Aurora, be converted into 3D Tetgen mesh VERY easily and FAST too.
Following 'Selopezs' guidelines, what you see in Netgen as an external / surface mesh, can be almost exactly
what Aurora's Tetgen 3D volumetric meshing will follow providing you choose a similar maximum edge length
as Selopezs' recommendations.

I see that version 5.0.0 of NETGEN has been published. It has parallel processing support. However, it is not an 'easy windows install' procedure but take a look if you're confident. I use version 4.9.13 - i managed to load it ok so it isn't difficult.
It's fast , reliable, mesh flexable / adaptable. The only down side is there's no usual 'help' :cry:

If you use 'selopezs' guidelines it will work well. It does for me very nicely. :D

hope the helps..

rgds. Tim

ps. something random here. Use version V2a - big benefits. And, if your models are small / medium sized,( see the theory manual for definition of size), i have discovered that the Pardiso runs at x10 ~x14 faster than SICCG / SORCG processes but large amounts of RAM memory is required or it will crash!
selopez
Preprocessor
Beiträge: 69
Registriert: Sa 24. Mär 2012, 03:10

Re: improving meshes

Beitrag von selopez »

Tim

I suppose that you, as I, tried unsuccessfully to install Netgen 5.0 by compiling the binaries files. At http://sourceforge.net/projects/netgen- ... esher/5.0/ there are Windows Installers for 32 and 64 bits. I tried the 32 bits one and it works fine but... at the new Netgen version, "our friend", the "Step / IGES Doctor" has disappeared! I think that this is a kind of bug that will be fixed in future releases. If you install the 64 bits version and find that the "Doctor" is there, please let me know it.

Best regards.
selopez
Preprocessor
Beiträge: 69
Registriert: Sa 24. Mär 2012, 03:10

Re: improving meshes

Beitrag von selopez »

it seems that the link has not been well published. This is the complete one:

http:// sourceforge.net/projects/netgen-mesher/fles/netgen-mesher/5.0/

(eliminate the space after "//")
selopez
Preprocessor
Beiträge: 69
Registriert: Sa 24. Mär 2012, 03:10

Re: improving meshes

Beitrag von selopez »

TAH1712
Mesher
Beiträge: 30
Registriert: Fr 24. Jun 2011, 14:13

Re: improving meshes

Beitrag von TAH1712 »

Hi Selopez. Good to talk to you.

Thanks for the update on Netgen 5.
I will load it up asap although as the mesher, Netgen 4 is already doing a great job!

Concerning the possible shortcomings of not having an 'igs/stp / topology doctor', It is possible to load in Netgen 4, analyse/heal etc. and save out in .stp format to input into Netgen 5 for faster /powerful meshing- that's only theory for now as i've had a few crashes tonight in experiments!

As to my prowess in surface modelling - well that was a very long time ago. I just remember that almost anything was possible to achieve given the prevailing willpower. I'm sure even today, surfaces are more powerful and more adaptable than solid modelling.

All the very best to you.

Tim.
frans
Import
Beiträge: 23
Registriert: Do 6. Sep 2012, 21:03

Re: improving meshes

Beitrag von frans »

Hi Tim and Selopez

Thank you very much for the suggestions.

I'm very happy with the netgen 5 program, all the button's are there to make a realy nice mesh.
I can make them very equal on the thin walls.
Importing is a bit tricky i don't know how "to compile" the cascade step. importer function with the netgen.
Anybody knows a manual to do this for a windows machine?

One thing that works really well in netgen is to export as an abacus file and import this in z88, this assures you get the same mesh in z88 as in the netgen. you can directly start with picking points etc.


regards,
Frans
selopez
Preprocessor
Beiträge: 69
Registriert: Sa 24. Mär 2012, 03:10

Re: improving meshes

Beitrag von selopez »

Hi, Frans

You say that "Importing is a bit tricky i don't know how "to compile" the cascade step. importer function with the netgen."

Just a question. When you installed NETGEN, did you installed also the "Open Cascade kernel"?

best regards
frans
Import
Beiträge: 23
Registriert: Do 6. Sep 2012, 21:03

Re: improving meshes

Beitrag von frans »

Hello Selopez,

How stupid of me, I forgot to mark the checkbox for the occ instalation when i installed netgen :roll: .
Thank you Selopez for pointing me were to look.

regards Frans
Antworten