Mecway crashing

Sture

Hi,

I have ver 18 and my models keep crashing in the solve phase.

Version 11 and 18
Big and small models
Static 3D
Mecway solver

How can I start trouble shooting?

/Sture

kuhl

Hi everyone,
same issue here, also Mecway 18, model with about 380k nodes crashes in modal analysis during "Assembling matrices" and in 3d static during "Applying constraints", both with internal solver using laminate material properties on quad8 and tri6 shells.
And it seems that only one cpu core is used of the 16 that are available.

Is there a log file that would help me find out what is gooing on?
Are there settings necessary to make Mecway use more cpu cores?

Thanks a lot for any advice and thanks to the forum in general; I have already found a lot of useful information here!

Kind regards,
Kuhl

disla

Hi Kuhl,

¿Could you post one the files that "crash"?. Without it, is difficult we can help you.

Regarding the number of CPU there are many references in the forum about increasing the performance and set up the number of cores.


https://mecway.com/forum/discussion/1012/improving-performance-of-ccx-solver

disla

You also have many examples of laminates in the forum with files attached that do not "Crash". Look at this for example:

https://mecway.com/forum/discussion/354/composite-laminates-tutorial

kuhl

Hello disla,
thank you for your reply and the links!
Sadly, I cannot share the model. I have solved a number of test case before (smaller models) on my notebook to do some validation on composites and everything was fine. Then I started modeling on larger scale and I was able to solve the model (again on my notebook) with no issues. Now the model has become quite large so I tried solving it on my more powerful work station. Both a modal and static 3d analysis failed, because Mecway shut down during "assembling matrices" (modal) and "applying constraints" with no reason that I can see. Mecway just closed and I got no error message.
In the meantime I was able to solve the model on my notebook, again with no issues, just taking quite long and also just using one core.
So the question is: why does the model solve on the notebook but not on the workstation?
And why are the work station and notebook only running on one core with the internal solver?

I want to use the internal solver, because cxx has some bugs with mass elemtens (and others) in combination with composites. The threads on solver performance I could fin in this forum all reference ccx, or did I miss something?

Kind regards
Kuhl

disla

-I would check if Mecway has read / write permission on your workstation.

-Play with the temp directory so you are sure Mecway can write there.



-¿have you been able to solve simple models?. I mean, ?what happens if you make your laminate just a simple isotropic shell. ¿Does it fail too?

Regarding the internal solver, you are right, all improvement suggestions are related to ccx.

kuhl

Thanks for the hint regarding the working directory, I'll give it a try tomorrow with a simple model.
Any Idea why the internal solver is only using one core? I thought it should be working on multiple cores?

kuhl

Hi again,
I did try with a simple shell / laminate model just now. 230 nodes.

Mecway closes during solving stage with no error message. It does so with laminate material and isotropic material with the internal solver. Using calculix works, but that is not a real option for me because of other reasons.
If I extrude the shell elements to solids, it solves with no problem with isotropic material.

Now the peculiar thing: If undo the extrude operation to get back the shell elements from the original model, it solves the shell elements with no problem!

If I open any solid model, solve it, then open my shell / laminate model without closing Mecway application, that too solves without a problem!

As soon as I close and reopen Mecway, I am back to square on and the shell only solves if I have previously solved a solid model in the same instance of mecway...

Thats really weird. Any ideas?

disla

"Mecway closes during solving stage with no error message."

¿Do you mean the program close, or the solver process stops and exits without solution?


¿Could you share that simple model that is failing with the internal as it is?
I understand it only fails in your workstation.
Are you using windows on a virtual machine?

kuhl

Hi disla,
yes, mecway closes completely so that I have to reopen the program.
The workstation is running on Windows 10 Pro, no virtual machine.

The simple model is attached.

disla

It works perfectly fine for me.

Seems a hardware/drivers/permissions problem. Probably Victor has some other ideas.

I can only suggest you reinstall MECWAY as administrator.

NOTE: ¿Did you deactivate the license on your laptop before using the new copy on your workstation?.
I'm not sure if you can have one license working on two machines at the same time. Maybe there is some network check that stops Mecway from running.

cwharpe

No problem solving here, either. Merged nodes first, if it makes any difference.....

kuhl

I did have the licence deactivated on the Laptop. I'll try to reinstall as administrator tomorrow.
Thanks for testing!

Victor

Any license problems should be communicated clearly, and not through weird crashes.

@kuhl. Do you mind trying version 15 (https://mecway.com/download/oldversions/mecway150.msi) which was the last version that worked for another person who had these hard crashes while solving.

Victor

Since this seems to be a problem with workstationy computers, I'm wondering if it's trying to use too many cores. Could you set these two environment variables then open Mecway and solve it?

MKL_NUM_THREADS = 1
OMP_NUM_THREADS = 1

Another strange thing - put some solid elements in the model. Another customer had this kind of computer-specific crash with beams but not if there were also solids!

kuhl

Hi all,

over night the workstation was able to solve the large model in modal vibration!


The it seems that Mecway on workstation can only solve shell models if

• it has successfully solved any other model in the same Mecway instance.
  
• if there is a solid element anywhere in the model. I tried that with the simple model
  

Mecway 15 solves without issue.

Changing the system variables does not seem to help.


What are your experiences with the time for solving compared to the time for allocation/assembly of matrices and derivning field variables? Solving seems to be very fast while everything else seems slow in comparison.

Sergio

Just for the record I´m using a CPU Intel(R) Xeon(R) W-2275 CPU @ 3.30GHz, and the Mecway v17 with the internal solver crash if there are only beams elements in the model. If I include a dummy solid element then can be solved without problems.

Victor

@Sture @kuhl @Sergio I've made a patch with an updated version of MKL. The existing one is 10 years old so that might magically fix the problem. If you'd like to use it, download from https://mecway.com/UpdatedMKLForMecway18.zip and extract the two .dll files to where Mecway is installed (typically %programfiles%/Mecway/Mecway18), overwriting what's already there, then start Mecway.

kuhl

Hi Victor,
I'll try and report back.

kuhl

Hi,
thank you for solving the issue, the patch works fine.

Victor

Great news! Thanks for testing it.

Vinicius_Petrucci

the same problem here.
Dual Xeon 2699v3 CPUs

the path worked

Sergio

The same for me, now it works ok. Thanks!