Learning Forum

[Dennis Chen]

Hello, from what I've seen before, Optislang was able to build a sampling plan and create various version of a dyna input file somehow using *case and *parameters.   My questions are

1) Is this true?

2) is it possible to do this without optislang?

3) can anyone share a template of this keyword file such that I may be able to replicate this without optislang?  I can write scripts myself

4) I checked all over the documentation about how *parameter works with *case but I cannot find anything, did I miss anything?

[igandiko]

I would assume optislang or any optimization/DOE tools are generating LS-DYNA k file for each design combination by varying certain parameters. These can be defined using *PARAMETER card. For example, if shell section thickness is being varied, then the only difference in ls-dyna input file of two designs would be the value of parameter defined in *PARAMETER, rest of the input is same. These two jobs are not dependent on each other and can be executed in parallel. 

*CASE allows you to run multiple ls-dyna runs sequentially. If I want to run the above setup sequentially, I may define the input as below. 

*KEYWORD

*CASE_1_BEGIN

*PARAMETER

rthk, 3.0

*CASE_1_END

*CASE_2_BEGIN

*PARAMETER

rthk, 6.0

*CASE_2_END

*INCLUDE

rest _of_model.k

*END

I am not sure why you would need both *CASE and *PARAMETER in the same setup, unless you would like to run sensitivity analysis/optimization using certain paramaters with sequential ls-dyna analysis where for one setup of parameter values, you have two sequential ls-dyna runs with output of case 1 as input of case 2 (for example, dynain files in metal forming type problems). 

[Dennis Chen]

Hi igandiko, to run cases in parallel, what is the proper template?    Also, maybe I just missed it but where in the documentation can I find a more detailed explanations.   I looked at both the keyword for *parameter and *case and also at the corresponding appendix.    I will dig deeper this weekend on this topic.   I've previously looked at sequential runs but I didn't use parameter, only used the dynain files.

I like to build a clearer workflow on how to properly use *parameter and *case to basically do my own DOE setups for example.    Each *Case would be one set of parameter settings and I like to submit them to a cluster to solve concurrently and I am also not sure how to properly control how many cases are submitted at the same time, etc.

In any case, I appreciate any guidance or examples because it will save me a lot of time where I will have to go dig up stuff and figure out haha

Best Regards