|
[Sponsors] |
March 31, 2017, 15:02 |
Meshing on hpc using ansys mesh
|
#1 |
Senior Member
Join Date: Mar 2013
Location: Germany
Posts: 357
Rep Power: 14 |
Hi All
I am using Ansys mesh for meshing my geometry. And I have a huge geometry which is not axisymetric or so, hence I need to do the entire geometry using ansys mesh- Unstructured mesh. The problem here is the time. Im my local machine it takes hours to get a mesh done of 18 million elements. And hence I find it waste of time. My desktop is 16 GB RAM and 8 core and eventhough I give parallel option Ansys doesnt utilize it completly, dont know why. But now my question is, can i do the meshing in HPC cluster with some batch scripting and automating the meshing process on cluster ? Kindly let me know if there is any way to do that. My second concern is that , I just have fluid volume to mesh, but during the meshing process the Ansys workbench update model status shows Status: Meshing Solids... (30%) and so on...but where did this solid come from or is it just considering the volume as solid ? Thanks |
|
April 1, 2017, 06:57 |
|
#2 |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,871
Rep Power: 144 |
Have a look for parallel meshing options. There are not many available, but ANSYS has a few options. And if parallel meshing was easy then it would already be in there.......
Check you are not running out of RAM. Or split it up into sections and mesh them separately. Otherwise you will just have to be patient and wait. |
|
April 1, 2017, 07:05 |
|
#3 |
Senior Member
Join Date: Mar 2013
Location: Germany
Posts: 357
Rep Power: 14 |
What do you mean by parallel meshing ? HPC or just doing the meshing in my computer parallel. I selected all the CPUs and still it sometimes just take 50% of my RAM and 6 CPU instead of 8 or sometimes 8 but its still slow.
Secondly what do you mean by '' split up into sections and mesh them separately ? '' You mean to split the geometry into different parts and how can I merge them in CFX to apply the solver ? And what about the interface etc ? Could you give me an example. I just have a long pipe with a swirl diffuser at the end. And I am taking the internal fluid volume and meshing it in Ansys Mesh using Workbench platform. So how I can do it into two parts ? Thanks |
|
April 1, 2017, 07:42 |
|
#4 |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,871
Rep Power: 144 |
By split them up I mean just chop the geometry in half and join it up later, probably with a GGI. This is not desirable in most cases and personally I would just put up with a long meshing time.
|
|
April 1, 2017, 07:50 |
|
#5 |
Senior Member
Join Date: Mar 2013
Location: Germany
Posts: 357
Rep Power: 14 |
GGI ? And how to join it later and where ?
The problem is I am ok to wait but then after waiting for hours when I want to convert the ansys mesh to cfx mesh or fluent mesh .msh format it says the conversion failed contact the support...and obviously my mesh was of 31e6 nodes. So 'I dont know what to do. And also I have a doubt..When I just mesh the fluid volume of my mesh in the meshing it shows meshing solids 30 % ..etc but why ? Why solids ? |
|
April 1, 2017, 20:29 |
|
#6 |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,871
Rep Power: 144 |
The mesher does not know if a solid body is assigned a fluid or a solid. It just calls them all solids.
Before you go too much further, please watch your memory usage during a meshing operation. If you are running out of memory you have to fix that before considering anything else. |
|
April 1, 2017, 20:35 |
|
#7 |
Senior Member
Join Date: Mar 2013
Location: Germany
Posts: 357
Rep Power: 14 |
Thats the problem , I have only 16GB RAM and it gets full while running Ansys mesh and I have to wait for long time or it gets crashed and I am wasting time on that hence I wanted a solution to mesh it in HPC cluster using some batch script!! For example StarCCM from CD Adapco has that facility where we can create a macro and do the meshing on HPC with script.
|
|
April 2, 2017, 07:57 |
|
#8 |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,871
Rep Power: 144 |
The easiest solution is to get more memory. Alternately, other meshing strategies use less memory, for instance the ICEM meshing approach is much more memory friendly. Note you can use the ICEM mesher in ANSYS Mesh by selecting patch insensitive meshing.
Of course, you need to think about whether your mesh really needs to be that big as well. |
|
April 3, 2017, 10:08 |
|
#9 |
Senior Member
Join Date: Jul 2011
Location: Berlin, Germany
Posts: 173
Rep Power: 15 |
Just wondering a bit about "hours of meshing for 18 Mio Elements"....
I have a core i7 Quad Workstation with 16 gb RAM and meshes with 18 Mio. Elements certainly do take some time but not HOURS not even one Hour... What does your geometry look like? Have you set your refinement conditions properly, so that only relevant parts have a fine mesh and regions without significance are meshed coarser? It sounds as if you were using the Workbench and the Workbench mesher? |
|
April 3, 2017, 10:28 |
|
#10 |
Senior Member
Join Date: Mar 2013
Location: Germany
Posts: 357
Rep Power: 14 |
Well My geometry is a 200 mm pipe with 7mm Diameter and at the end of the pipe is a Difuser which is 20mm length and that is the most important region of interest. this pipe i used to make the flow fully developed, thats it. And the only options in meshing (Workbench Meshing) I gave was Inflation layer, body sizing with the region close to the difuser and a body sizing with minimum size as one value. So once I give all these with a fine mesh it creates a mesh of around 18e6 elements in 4 or 5 hours and my memory shows full around 15GB out of 16GB RAM.
|
|
April 3, 2017, 10:29 |
|
#11 |
Senior Member
Join Date: Mar 2013
Location: Germany
Posts: 357
Rep Power: 14 |
Or even sometimes that my program crashes
|
|
April 3, 2017, 10:37 |
|
#12 |
Senior Member
Join Date: Jul 2011
Location: Berlin, Germany
Posts: 173
Rep Power: 15 |
Did you try defining elemnt sizes on the different surfaces with an element growth rate. Beginning with quite coarse elements and then refining only the regions of interest or regions that look too bad?
Or try using icem maybe this helps...Personally I prefer ICEM because it gives me the feeling of better beeing in control of what happens during the mesh process. |
|
April 3, 2017, 10:42 |
|
#13 |
Senior Member
Join Date: Mar 2013
Location: Germany
Posts: 357
Rep Power: 14 |
Yeah ICEM is good enough provided I have enough time to spend on it and learn it. And I am doing unstructured mesh and hence I thought its better to use the easy and decent one
Btw I havent tried this element growth rate starting from inlet to outlet |
|
April 3, 2017, 20:05 |
|
#14 | |
Super Moderator
Glenn Horrocks
Join Date: Mar 2009
Location: Sydney, Australia
Posts: 17,871
Rep Power: 144 |
Quote:
The alternative is to use different meshing approaches which use less memory - ICEM is likely to be more memory friendly. |
||
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
[Commercial meshers] Problems with ANSYS mesh conversion | tdog | OpenFOAM Meshing & Mesh Conversion | 1 | March 31, 2016 18:36 |
Mesh Seeds in ANSYS WorkBench Meshing | smcp | ANSYS | 1 | March 12, 2012 14:16 |
[ANSYS Meshing] Export Gambit mesh into something Ansys workbench meshing can read? | amad11 | ANSYS Meshing & Geometry | 1 | February 21, 2012 13:26 |
How To save a created mesh file in Ansys Meshing | ashtonJ | CFX | 4 | January 7, 2012 23:04 |
[Other] Ansys meshing - periodic mesh for fluent | jsm | ANSYS Meshing & Geometry | 4 | September 24, 2010 15:25 |