Learning Forum

[Webster]

Hi Community,

I am using the keyword CONTROL_REFERENCE_CONFIGURATION to find the reference configuration of a geometry in LS-Dyna. Yet, the descriptions on the manual is a bit vague. As I follow the descriptions to set up the main script, somehow the computation terminates after iteration 1 with the error messahe 'forrtl: severe (157): Program Exception - access violation' printed at the beginning of iteration 2. This makes me a bit hard to check whether the desired convergence has reached, or not.

Could anyone kindly offer me some insights regarding this keyword? Is there a way to restart my iterations? How to check whether my convergence has reached or not?

Thank you in advance!

 

 

[Ushnish Basu]

The error message indicates an internal error in LS-DYNA. Please try using a recent development version, else try contacting your Ansys representative to see if further specific help can be arranged. 

[Webster]

Thank you Ushnish. Could you please kindly check whether this issue has been solved in the recent dev versions? Which version, specifically? I may need to contact our HPC team to address this issue accordingly.

[Ushnish Basu]

I would need to have an input file to see what the issue is, but unfortunately, Ansys employees are not allowed to download any material from forum posts

[Reno Genest]

Hello Binghuan,

If  you are a commercial customer, please create a support case on the Ansys Customer Support Space (ACSS):

customer.ansys.com

Then, we will be able to get your model and have a look.

 

Reno.

[Webster]

Dear Ushnish and Reno,

Thank you for your suggestions. Very unfortunately, I am an academic user. By any chance, the customer support portal could let research users to post questions...? :-(

[Reno Genest]

Dear Binghuan,

If you are an academic customer with a commercial Ansys or LS-DYNA license, you are eligible for technical support and you can create a support case on ACSS. If you are eligible for technical support, you should see an "Open Case" button on the ACSS website at customer.ansys.com

If not, please check with your Ansys account manager to see if there are other alternatives for you to get support.

 

Reno.

[Reno Genest]

Hello Binghuan,

Which version of LS-DYNA are you currently using? Please make sure you use the latest version R14.1. Are you running SMP or MPP. Could you send us a screenshot of the LS-DYNA banner (top of the d3hsp file)? This will tell us exactly what we need to know.

You can download LS-DYNA here:

https://ftp.lstc.com/user/download_page.html

username: user

password: computer

For R14.1:

https://user:computer@ftp.lstc.com/user/ls-dyna/R14.1.0/

https://user:computer@ftp.lstc.com/user/mpp-dyna/R14.1.0/

 

Thank you.

 

Reno.

[Webster]

Dear Reno,

Thank you for supplying me the latest version. Unfortunately, the problem persists even with R15 (beta version). The error message states 'forrtl: severe (157): Program Exception - access violation' at the beginning of the second iteration.

Do you have any other suspects? I really appreciate your input. 

[Reno Genest]

Hello Binghuan,

It looks like you are running  MPP LS-DYNA double precision with Microsoft MPI. Have you tried with Intel MPI? Have you tried SMP?

We can send you the latest DEV version of LS-DYNA for you to try.

 

Reno.

[Webster]

Dear Reno,

Thank you for your advice. No, unfortunately, the culprit is not the LS-Dyna versions.

I understand that you cannot download any files from the site. Could you please have a look to the screenshot below, or pass the screenshot to an expert?

The red-coded section is where the keyword CONTROL_REFERENCE_CONFIGURATION is deployed. The orange section is how I load my geometry. 

The computational stops after iter 1. 'iter2.inp' exists in my directory but simply not executable with the error message 'forrtl: severe (157): Program Exception - access violation' .

Thank you.

[Webster]

Additionally, may I raise this point to you here, there is a typo in keywords existed in LSPP 4.10 AND 4.11. There is a missing 'r' in 'CONFIGURATION'. This gives wrong warnings when reading in my scripts, and writes out unexecutable keyword to my .k files.

[Reno Genest]

Hello Binghuan,

I let the LSPP developer know; this should be fixed in the next release of LSPP 4.10 and 4.11.

 

Reno.

[Reno Genest]

Dear Binghuan,

Have you tried with the latest DEV solver? If not, send me an email and I will send the solvers to you. Send me a screenshot of the LS-DYNA banner (top of d3hsp file) so that I know which flavor of the solver you need. Here is an example of LS-DYNA banner:

 

Reno.

[Webster]

 

Dear Reno,

Out of any luck, the problem still persists. I have tried whatever I could tune from my side, and different versions of LS-Dyna. Any suggestions from you?

Would it be more stable for me to run LS-Dyna on a Linux machine, rather than a Windows PC? Seems that many tests for LS-Dyna were conducted in a Linux environment.

To add, the key problem is the keyword *CONTROL_REFERENCE_CONFIGURATION_ITER, rather than *CONTROL_REFERENCE_CONFIGURATION. When using the case keyword (case  i = xxx.k) as the input, LS-Dyna automatically generate .inp files (iter1.inp, iter2.inp, etc) for the iteration purposes, with the keyword automatically replaced as *CONTROL_REFERENCE_CONFIGURATION_ITER. This is likely to be the culprit.

Thank you.

 

[Reno Genest]

Dear Binghuan,

Are you able to run without  the "case" option? Copy your model (.k file) and place it in a new folder. Then, make the modification for that case. Repeat for the other variations of the model.

Can you run this way?

 

Reno.

[Webster]

Hi Reno,

Unfortunately, I cannot run the script without the case option. The Sellier’s method or Rausch’s method employed in LS-Dyna are both iterartive optimisation procedures, I cannot run iteration 2 without knowning a priori from iteration 1 - this means we need to fed the result from iteration 1 as the initial guess for iteration 2.

LS-Dyna automatically generate .inp files (iter1.inp, iter2.inp, etc) for the iteration purposes, and from iteration 2, the keyword is automatically set to *CONTROL_REFERENCE_CONFIGURATION_ITER, which is the only method that allows the result from iteration 1 to be fed as the initial guess for iteration 2. 

The problem appears with the keyword *CONTROL_REFERENCE_CONFIGURATION_ITER, which the access violation error is reported, regardless of the version (tried with v13, v14, v14.1, v15), or computation precision (single or double), or parallelism mode (smp/mpp).

I understand you could not download any files from the site. Could you please kindly quickly check the script (attched below as a screenshot) and see if there are any obvious issues with it? Also, what does the error message 'forrtl: severe (157): Program Exception - access violation' indicate? Would you recommend me running LS-Dyna on a Linux machine, rather than on a Windows PC?

Thank you,

Binghuan 

 

 

[Reno Genest]

Hello Binghuan,

You can try running on Linux and see how it goes.

The error message does not mean much; when we get this error message we have to have the model checked by a developer to debug further.

Have you condidered using the Multi-stage approach with dynain.lsda file instead of using cases? Have a look at appendix X in the LS-DYNA user manual.

 

Reno.

[Webster]

Dear Reno,

Thank you for your suggestion.

Unfortunately, I cannot bypass the 'case' option manually - as explained by our manual, the keyword CONTROL_REFERENCE_CONFIGURATION_ITER was built upon case, hence, users do not manually pipeline the iterations, but let the software do its own job.

I will try with the Linux version, although it may take some time to configure the environment on an old workstation.

By any chance, would it be possible for you to kindly pass this issue to the development engineer to take a quick look? I am happy to provide you with the script in whatever way that complies with the regulation.

Thank you,

B

 

 

[Webster]

 

Hi Reno,

Thank you. Here is the banner. I tried with smp. And R15 did not work, either.

And yes, please send me a dev version through (‘lbinghuan@outlook.com’).

Thank you.

Binghuan