optimzation ends prematurely because of failed runs

Melih Ayyildiz_22506
Melih Ayyildiz_22506 Altair Community Member
edited January 31 in Community Q&A

Hello,

I wan to run an optimzation in HyperStudy. I am using a FluxMotor model on which I am changing some geometric parameters to find an optimal motor for my case. These geometric parameters are listed as input variables with a lower and upper bound within HyperStudy. I have around 20 geometric inputs. Because of this, there is a high chance that for a given set of inputs, the actual motor would not be physically possible because of an intersection of lines or faces. I see this especially during a DOE run.

FluxMotor would give me a notice if a specific motor would not be possible if a certain input is too big. I found out that within FluxMotor I can find the geometric constraints for the magnet and slot shapes. I have added these constraints within HyperStudy. I was able to reduce the number of failed runs in a DOE but there still sets of input variables that cause a failed run. The runs failed because of intersections of lines and faces. 

I started a optimization (MOGA solver) with this model but unfortunately the run ended prematurely with this notification:

ERROR #1000: failed analysis is more than 90% of the population.
 
              ***** OPTIMIZATION STOPPED DUE TO FATAL ERROR *****

Is there a way to disable this function? I want to run the optimization regardless of the percentage or number of failed runs.

 

Tagged:

Answers

  • Adriano_Koga
    Adriano_Koga
    Altair Employee
    edited January 31

    hi,

     

    HyperStudy allows you to ignore failed run at some extent.

    Usually at the optimzation setup, you will find an option 'ignore failed analysis'.

    But you can imagine that if there are a lot of failures, it is not possible to proceed, as your case, with 90% of failure.

     

     

  • Melih Ayyildiz_22506
    Melih Ayyildiz_22506 Altair Community Member
    edited January 31
    Adriano A. Koga_21884 said:

    hi,

     

    HyperStudy allows you to ignore failed run at some extent.

    Usually at the optimzation setup, you will find an option 'ignore failed analysis'.

    But you can imagine that if there are a lot of failures, it is not possible to proceed, as your case, with 90% of failure.

     

     

    Usually at the optimzation setup, you will find an option 'ignore failed analysis'.

    I already have this option turned on. Unfortunately, it doesn't make a difference for me.

Categories