Learning Forum

[blackbaddl]

Hi everyone, 

I am having trouble importing a CAD geometry on a HPC System in a Journal-File. The process is started but will be arborted after a while, I already tried it with a lot of different files to eliminate the possibily of a corrupt CAD file.

I am working on a CentOS Cluster 7.6 and I use this command

file import cad , BAExample.scdoc.pmdb , , , mm 

 

and get this result,

 

Importing one geom object per program-controlled and one zone per body ...

Initializing ANSYS component system ...

running Ansys.Meshing.FieldMesher.exe ...

creating FMComponent instance ...

querying IFieldMesher interface ...

AnsysFieldMesher build date: Aug  8 2018 15:00:06

 

argv[0] = /cluster/ansys/v192//aisol/CommonFiles/linx64/ansys.meshing.fieldmesher.exe

argv[1] = -B

argv[2] = /cluster/ansys/v192//commonfiles/CPython/2_7_13/linx64/Release/python/../Ansys/TGrid/CADReaders.py

argv[3] = -a

argv[4] = /tmp/options1578765954.771324.xml

argv[5] = -a

argv[6] = /tmp/options1578765954.771324.xml

    /cluster/ansys/v192//aisol/CommonFiles/linx64/ansys.meshing.fieldmesher.exe started by bt703327 on r05n59 linx64 CentOS Linux release 7.6.1810 (Core) on Sat Jan 11 19:06:02 2020

    using Python 2.7.13 (default, Jun  7 2018, 04:236) 

    [GCC 6.3.0]

    using Ansys.Meshing.FieldMesher build Aug  8 2018 14:555 including TGLib build 2018-03-30 205:20 CEST

    running ANSYS TGrid CADToTGridConverter ...

    setting up parameters ...

    setting up parameters done.

    running conversion ...

    converting 1 file(s) from Workbench to FLTG using output path '/tmp/out1578765954.771324.tgf'

    converting file 'BAExample.scdoc.pmdb' (1 of 1) from Workbench to FLTG using output path '/tmp'

    importing data ...

    importing meshing model from PartMgr from file '/scratch/ansys/runtime/bt703327/39545.mgmt01/BAExample.scdoc.pmdb' ...

    setting enclosure and symmetry processing to False

    setting transfer named selections to True

    setting named selection prefixes to ''

    setting mixed import resolution type to 'SolidAndSurface'

    setting target application to 'FluentMeshing'

    initializing part manager from file ...

#f

 

Warning: An error or interrupt occurred while reading the journal file.

Some commands may not have been completed.

 

 

Meshing> Halting due to end of file on input.

 

The things I noticed are:

   - Is Ansys supposed to open a .exe file on a linux cluster?

   - in the last step, before the process is terminating the RAM of the HPC system is getting flooded extremely.

 

What can I do? The HPC here are a bit helpless and this point is quite critical for my project.

Thank you very much! 

[Rob]

Fluent Meshing should be able to read the pmdb file via the water tight geometry workflow: have you tried using that?  Note, parallel meshing will give you a speed up, but it shouldn't alter how the import & surface mesh behave. 

Edit: I see you're using R19.2 can you try with 2019R3?  Fluent Meshing is being upgraded with each release so the issue may have been resolved.  

[blackbaddl]

Hi, 

I am glad you mention that, I was searching quite a while for documentation on how to use watertight geometry in batch mode and have not been succesfull yet. Where do I find this? I am not able to use a GUI on the HPC

I am only using .pmdb because Ansys told me so. When you try to import a .scdoc on the cluster it says something like "Linux can not read scdoc use .pmdb as workaround". I can use something else when you tell me so. 

I know that there is not a lot advantage from meshing on a HPC but I am working on a automated workflow which includes meshing, solving and basic postprocessing.

I am using 19.2 and I can not change that unfortunately 

Thank you !

[blackbaddl]

All I can try is on a local maschine in batch mode with 19.3 because the HPC System is not offering the 19.3 module yet.

I will try.

[blackbaddl]

Yes it is indeed working on Windows on a local maschine with 19.3 but I do not know how this is comparable to an issue on a Linux cluster. I can not test it with 19.3 on the cluster. For me this is not really a problem with meshing I am not even able to import geometry.

 

EDIT: On local mashine on windows everything is working even in 19.2!

[Rob]

Watertight may not work in batch at R19.2; it's been improved for 2019R3 but the workflow is designed for GUI use.  You can run using scripts, is your journal using any "cx-gui-" type lines? 

The pmdb route is to avoid needing SpaceClaim on LINUX as it's Windows only. 

[blackbaddl]

No the script is not using this kind of commands yet. I have seen these in the automatically recorded journal files. I already considered using them but not tried yet. These are even more confusing than Scheme scripting with formated strings is anyway. It would be an acceptable workaround for me to import the geometry by these commands but not the whole meshing part. That might be possible because you can import the geometry in watertight and change to the outline tap and proceed. I will try it like that 

 

By the way what are the Improve Surface/Volume Mesh task doing? These provide awesome results in reasonable time. For me this is the best part of the whole watertight process. The rest is nice to have and might be good to learn Fluent 

 

Regards

[Rob]

The cx-gui style commands are generated when you save a journal using the GUI; if you're then running without graphics they tend to fail. If the standard scripts are failing in standalone check the paths: I've not seen that reported from commercial users. 

Improve Surface/Volume mesh is doing just what it says: improving the mesh.  

 

[blackbaddl]

No you do not necessarily need a GUI for gui commands. I already excecuted other gui commands in batch mode  just with the fluent -gu flag. Which is with grafic commands and without user interface

 

Yes this is obvious. I wanted to know what it does in detail. I never achieve these improvements by hand. Is it doing General Improve, Collapsing skewed cells, fixing intersections...? Is there a particular order for it?

But this was just something I was courious about. Anyways I did not proceed with my issue at all.