AnsweredAssumed Answered

Parametric study needs recalculation for newly added results?

Question asked by Mar Gai on Aug 18, 2016
Latest reply on Aug 24, 2016 by Boyko Tchavdarov

Like • Show 1 Like1
Comment • 7

Hello,

I tried to use parametric studies to determine the pressure drop over a valve for different mass flows. I expected that parametric studies need less computation time than doing the variants manually.

But I discovered two problems:

1. The computation time for each single scenario is similarly long as the computation time for the original project. Is this true?

2. If I have calculated the different scenarios, and afterwards want to look at a different result parameter (e.g. a velocity...), this newly selected result parameter does NOT appear before I recalculate all the scenarios, needing again the full computation time ! Is this normal, or am I doing something wrong?

Thanks

Boyko Tchavdarov Aug 18, 2016 11:49 AM

Correct Answer

Hi Mar,

1) The computational time of the same model depends on the initial conditions and on the convergence criteria. Check that they are the same in your First Point of the Parametric Study. Each next point may take the previous result or not (default initial guess) depending on whether you have check the corresponding box.

2) You have to plan ahead which field variable will be of interest in your Study (Pressure, velocity, turbulent intensity, etc). Although all field variables will be computed in a Parametric Study only the selected filed variables will be output in a result file for post-processing. This saves space & time in writing./reading large result files.

See the reply in context

No one else had this question

Outcomes

Boyko Tchavdarov Aug 18, 2016 11:49 AM

Hi Mar,

Actions

Amit Katz Aug 18, 2016 4:07 PM

Parametric study will re-run the entire calculation. All you save is meshing time and the inconvenience of setting up multiple copies of the same simulation with slight tweaks. After you have your results you can do a trend analysis on the data in a program such as Excel to obtain an interpolation relationship.

Actions

Mar Gai @ Amit Katz on Aug 19, 2016 7:44 PM

Thank you both!

I consider it a bug that you can't add further output variables once the calculation has been done. What did I spend all that computation time for? ;-)

So parametric study is suitable only for small problems where calculation time doesn't matter much, or for very many scenarios where manual takeover of the results becomes a nuisance.

For bigger problems (with not so many different scenarios) I will use batch processing of many (manually adjustet) seperate projects instead, because only that way can I add further output variables later on.

Actions

Amit Katz @ Mar Gai on Aug 22, 2016 9:47 AM

You can set an option to save the complete flow field for every calculation. From that you can later create stand-alone project folders with fully solved flow fields. This takes up a lot of hard drive space.

Actions

Boyko Tchavdarov @ Mar Gai on Aug 24, 2016 5:09 PM

In the issue under discussion I do not see any difference between running several stand-alone CFD configuration and running Parametric Study. More over the latter has advantages (excluding that is not flexible in choosing transferred Initial Data). In both variants you can chose to save all field variables or only some of them. If you are not sure which variables you may need then save all of them. It is a disk-space trade off and result-file reading/writing time trade off.

Actions

Boyko Tchavdarov Aug 19, 2016 10:33 AM

Some more details on the Parametric Study. If ones the Parametric Study is has been run and then repeated it will e-run all the Design Points (runs in the Study) only if "Recalculate" Box is checked under the Scenario Tab.

One disadvantage of the Parametric Study in "What if Scenario" is there is no option to chose arbitrary Initial Data for each Design Point in contrast to multiple CFD-Configuration of the Project (CFD Model) where one can chose any result file (configuration) to transfer the initial data. I have submitted an SPR on the same some years ago to add a choice of selecting any result file for Initial Data of the Design Point in the Scenario Tab but it is not there in SW2017. . For the above reason the Parametric Study is not as usable as it could be. because the time of completing any CFD runs depends strongly of the Initial Data passed to the Solver.

In what's new in SW2017 it is claimed that by thr Optimization variant of the Parametric Study we can minimize maximize a goal (finally !) but I could not find that option. I created an SR on the same and will update this Q by the SW Support team answer.

Actions

Boyko Tchavdarov @ Boyko Tchavdarov on Aug 19, 2016 5:25 PM

It was my misunderstanding about parametric Study being able to Min/Max a Goal within certain range. Unfortunately it is not there yet and it may takes years to have it. What is supposed to be new is "Multiparametric Optimization" (I would not use the world Optimization when the goal is NOT Min/Max value but use the word "Target'). However I don't see that either in Beta.

Actions

7 Replies

Boyko Tchavdarov Aug 18, 2016 11:49 AM
Unmark Correct
Correct Answer
Hi Mar,
1) The computational time of the same model depends on the initial conditions and on the convergence criteria. Check that they are the same in your First Point of the Parametric Study. Each next point may take the previous result or not (default initial guess) depending on whether you have check the corresponding box.

2) You have to plan ahead which field variable will be of interest in your Study (Pressure, velocity, turbulent intensity, etc). Although all field variables will be computed in a Parametric Study only the selected filed variables will be output in a result file for post-processing. This saves space & time in writing./reading large result files.
Like • Show 2 Likes2
Actions
Amit Katz Aug 18, 2016 4:07 PM
Mark Correct
Correct Answer
Parametric study will re-run the entire calculation. All you save is meshing time and the inconvenience of setting up multiple copies of the same simulation with slight tweaks. After you have your results you can do a trend analysis on the data in a program such as Excel to obtain an interpolation relationship.
Like • Show 2 Likes2
Actions
- Mar Gai @ Amit Katz on Aug 19, 2016 7:44 PM
  Mark Correct
  Correct Answer
  Thank you both!
  
  I consider it a bug that you can't add further output variables once the calculation has been done. What did I spend all that computation time for? ;-)
  
  So parametric study is suitable only for small problems where calculation time doesn't matter much, or for very many scenarios where manual takeover of the results becomes a nuisance.
  For bigger problems (with not so many different scenarios) I will use batch processing of many (manually adjustet) seperate projects instead, because only that way can I add further output variables later on.
  Like • Show 1 Like1
  Actions
  - Amit Katz @ Mar Gai on Aug 22, 2016 9:47 AM
    Mark Correct
    Correct Answer
    You can set an option to save the complete flow field for every calculation. From that you can later create stand-alone project folders with fully solved flow fields. This takes up a lot of hard drive space.
    Like • Show 1 Like1
    Actions
  - Boyko Tchavdarov @ Mar Gai on Aug 24, 2016 5:09 PM
    Mark Correct
    Correct Answer
    In the issue under discussion I do not see any difference between running several stand-alone CFD configuration and running Parametric Study. More over the latter has advantages (excluding that is not flexible in choosing transferred Initial Data). In both variants you can chose to save all field variables or only some of them. If you are not sure which variables you may need then save all of them. It is a disk-space trade off and result-file reading/writing time trade off.
    Like • Show 0 Likes0
    Actions
Boyko Tchavdarov Aug 19, 2016 10:33 AM
Mark Correct
Correct Answer
Some more details on the Parametric Study. If ones the Parametric Study is has been run and then repeated it will e-run all the Design Points (runs in the Study) only if "Recalculate" Box is checked under the Scenario Tab.
One disadvantage of the Parametric Study in "What if Scenario" is there is no option to chose arbitrary Initial Data for each Design Point in contrast to multiple CFD-Configuration of the Project (CFD Model) where one can chose any result file (configuration) to transfer the initial data. I have submitted an SPR on the same some years ago to add a choice of selecting any result file for Initial Data of the Design Point in the Scenario Tab but it is not there in SW2017. . For the above reason the Parametric Study is not as usable as it could be. because the time of completing any CFD runs depends strongly of the Initial Data passed to the Solver.

In what's new in SW2017 it is claimed that by thr Optimization variant of the Parametric Study we can minimize maximize a goal (finally !) but I could not find that option. I created an SR on the same and will update this Q by the SW Support team answer.
Like • Show 1 Like1
Actions
- Boyko Tchavdarov @ Boyko Tchavdarov on Aug 19, 2016 5:25 PM
  Mark Correct
  Correct Answer
  It was my misunderstanding about parametric Study being able to Min/Max a Goal within certain range. Unfortunately it is not there yet and it may takes years to have it. What is supposed to be new is "Multiparametric Optimization" (I would not use the world Optimization when the goal is NOT Min/Max value but use the word "Target'). However I don't see that either in Beta.
  Like • Show 1 Like1
  Actions

Parametric study needs recalculation for newly added results?

Outcomes

Related Content

Recommended Content