Evaluations not influenced by design variables

[Carlos Almeida 0]

Hello,

I'm following the tutorial "Shape Optimization using CAESES and STAR-CCM+" and everything is ok until I try to run the  Sobol algorithm. Then, it gives the following error message:

"There are evaluations that are not influenced by any design variable."

If I click "continue", then my models are generated, but all my evaluations are "nan". What could be causing that?

 

Best regards,

Carlos

 

[Carlos Almeida 0]

Please also refer to this error.

[Carl Benz 5]

Hi Carlos,

Please make sure, that you have created an evaluation parameter which takes a value from a result file (@ 5:00 of the video tutorial) and make sure that you have set this parameter as an evaluation in the sobol settings (@6:54). If the parameter has a fix value and no reference like:

 Runner.getResults().getTable("cd.csv").getElementAt("cd")

you may get the Message "There are evaluations that are not influenced by any design variable.".

 

CAESES checks every dependency before running software-connections in an optimization algorithm. It is to save resources. If there is no reference to any value from a resulting file of a software-connection, this application will not be executed.

Best,

Carl

[Carlos Almeida 0]

Hello Carl,

I did create the evaluation parameter and it looks like this: starCCMconnectorSYM1.getResults().getTable("RTsym.csv").getElementAt("RTsym"), where starCCMconnectorSYM1 is my Software Connector and RTsym is the parameter extracted from the result .csv file.

I've set the parameter as an evaluation in the Sobol settings, but I still get the same error.

 

Best regards,

Carlos

 

 

[Carl Benz 5]

Hi Carlos,

could you share your projectfile, please? I'll have a deeper look.

 

[Carlos Almeida 0]

Hello,

Yes. Can you tell me your e-mail, please?

Another thing, when I open my CAESES project, I get the options in the screenshot below. Maybe that has something to do with it? If I select "baseline", I lose a lot of work inside of CAESES.

I've connected the 2 softwares effectively and can see the output from Star-CCM inside of CAESES, but even after my simulation is completed, all of my evals are NaNs.

 

Best regards,

Carlos

[Carl Benz 5]

Hi Carlos,

please use a personal message (click on my profile -> message).

Out of your last question about starting CAESES in Baseline or in designOfExperiments I assume, that you made a lot of work not with the baseline design. For CAESES each design is its own instance, and they are not connected with each other. So if you doubleclick on a design from a design engine the name of the design on top of the objecteditor gets a green background with a (by default) closed lock.

For sure you can make some changes by unlocking the design, but as I said already it will not infect the baseline design. As any design engine starts from the baseline design, this might be the cause for the original issue you reported.

If you want to continue working on a specific design from a design engine, you need to make it the baseline-design first: Go inside the design you want to make baseline (doubleclick -> green bar) and "Save Current Design As" a new clean project (without results from the previous design engines).

I hope this helps.

Best,

Carl

[Carlos Almeida 0]

Hello Carl,

That helped a lot, thank you. I'm back to the baseline without any losses.

But I'm still having the same problem, even after making a new Software Connector. I don't see the option to send files via messages, so I'm attaching it here.

 

Best regards,

Carlos

 

 

[Carlos Almeida 0]

These are the errors I'm getting after the run is finished:

 

 

[Carl Benz 5]

Hi Carlos,

you get the message about not influenced evaluations, because you called a sobol to minimize the parameter eval_rt. As a sobol is only a randomized design engine and no optimization algorithm, the message is true and informing. If you click "Yes", it will simply run the sobol. If you don't want to get this message you have to set the objective to "observer" (default).

I can't reproduce your simulation, because of the .sim file, which is only referenced and not attached in your project. But the reference itself is not set correctly. If you want to call the .sim file for each design inside the results directory (as you set in the argument section of you computation) you simply need to empty the string of the "Save directory". Then CAESES copies the file from "Filename"-location to "Save directory" which is always relative to the design folder. 

 

PS: To test, if a software connection has all inputs (referenced or edited files) needed to be ready for any design engine, it is helpful to delete the "manual_results/baseline/<connection>" folder (or make a backup if you think you need it) and let CAESES do the job. If any file is missing (or wrong referenced), you will notice.

Best,

Carl  

   

[Carlos Almeida 0]

Hello,

I'm having this problem when I try to run the Software Connector only, without a specified design engine.

I've deleted the manual_results/baseline/Connector1 folder and let CAESES create it. I can see the outputs from Star-CCM inside CAESES and the folder with the results is created (manual_results/baseline/Connector1), but it seems like it can't open the .png and the .csv files, but they're there. The path referenced by the error is correct:

*** INFO FGenIntReader : Could not open file /IP/Home/carlos_almeida/CAESES_connection3/.../manual_results/designOfExperiments/starCCMconnectorSYM1//rf.csv for reading [parsing process aborted]

*** INFO FResultsGenInt : File read error [Failed to open file /IP/Home/carlos_almeida/CAESES_connection3/.../manual_results/designOfExperiments/starCCMconnectorSYM1/pressure.png]

 

But it can't access it. I've tried a lot of things, including using an empty string on "Save directory", as well as setting the "Save directory" as the path to the /Connector1 folder. Maybe I have some problem with FGenIntReader and FResultsGenInt?

 

Thank you for your response.

 

 

Best regards,

Carlos

 

[Carlos Almeida 0]

I've figured it out. The project folder and files should be on my local machine.

 

Best regards,

Carlos