Hyperstudy Error

User: "Geovane_22233"
Altair Community Member
Updated by Geovane_22233

Hi,

I'm trying to do an optimization with a radioss file. 

But the follow message appears when I try to do it: " ERROR #2237: failed in sampling initial points. The design space could be too narrow, or EPSGRS could be too large."

Before I started this optimization, I checked the file using Hypermesh and I run the model. Everything was ok.

Find more posts tagged with

Sort by:
1 - 8 of 81
    User: "PaolaAG"
    Altair Employee
    Updated by PaolaAG

    Hello Geovane!

    Can you please give me this information?

    What version of Hst are you using?

    How many active variables do you have?  Are they discrete or continuous?

    What are the algorithm settings? (Default ones or did you modified them)

     

    Thank you,

    Regards.

    User: "Geovane_22233"
    Altair Community Member
    OP
    Updated by Geovane_22233

    Hello Geovane!

    Can you please give me this information?

    What version of Hst are you using?

    How many active variables do you have?  Are they discrete or continuous?

    What are the algorithm settings? (Default ones or did you modified them)

     

    Thank you,

    Regards.

    Hi, Paola

    I'm using the version 2022 of Hyperstudy.

    I have just one variable and it is discrete. 

    The algorithm is using the default settings.

    User: "PaolaAG"
    Altair Employee
    Updated by PaolaAG

    Thanks Geovane,

    If the design space is too limited, Hyperstudy may not be able to find enough sample points to start the optimization process. The discrete variable (with not enough values) is causing this behavior. 

    The solution is to increase the design space by allowing more values for the discrete variable.

    If you have one discrete variable with two values you need to add at least one or two more values.

     

    Let me know if it works,

    Kind regards.

     

     

    User: "Geovane_22233"
    Altair Community Member
    OP
    Updated by Geovane_22233

    Thanks Geovane,

    If the design space is too limited, Hyperstudy may not be able to find enough sample points to start the optimization process. The discrete variable (with not enough values) is causing this behavior. 

    The solution is to increase the design space by allowing more values for the discrete variable.

    If you have one discrete variable with two values you need to add at least one or two more values.

     

    Let me know if it works,

    Kind regards.

     

     

    Paola,

    I created two more values for this variable and now it works.

    Thank you.

    User: "Geovane_22233"
    Altair Community Member
    OP
    Updated by Geovane_22233

    Thanks Geovane,

    If the design space is too limited, Hyperstudy may not be able to find enough sample points to start the optimization process. The discrete variable (with not enough values) is causing this behavior. 

    The solution is to increase the design space by allowing more values for the discrete variable.

    If you have one discrete variable with two values you need to add at least one or two more values.

     

    Let me know if it works,

    Kind regards.

     

     

    Paola, I selected the option of GRSM and the number of evaluation of 10.

    But in the step of evaluation tasks, just show 3 evaluations.

     

    User: "PaolaAG"
    Altair Employee
    Updated by PaolaAG

    Paola, I selected the option of GRSM and the number of evaluation of 10.

    But in the step of evaluation tasks, just show 3 evaluations.

     

    Hello Geovane,


    In this specific case the behavior is as expected.

    The way GRSM works is that it needs 3 initial sampling points, and after that for each iteration it needs 2 points. Since you only have 4 values GRSM takes the first 3 for the initial sampling, and does not have enough values for subsequent evaluations.
    What you can do is change from discrete to continuous or select at least 11 values.

    Just out of curiosity, what is the use case? I find it curious that you only use one variable with discrete values.

    Regards.

    User: "Geovane_22233"
    Altair Community Member
    OP
    Updated by Geovane_22233

    Hello Geovane,


    In this specific case the behavior is as expected.

    The way GRSM works is that it needs 3 initial sampling points, and after that for each iteration it needs 2 points. Since you only have 4 values GRSM takes the first 3 for the initial sampling, and does not have enough values for subsequent evaluations.
    What you can do is change from discrete to continuous or select at least 11 values.

    Just out of curiosity, what is the use case? I find it curious that you only use one variable with discrete values.

    Regards.

    I'm using a material id like a variable.

    So I selected the discrete mode to change the material.

    But when I get the results, the output responses have the same values.

    User: "PaolaAG"
    Altair Employee
    Updated by PaolaAG

    I'm using a material id like a variable.

    So I selected the discrete mode to change the material.

    But when I get the results, the output responses have the same values.

    I understand that your end goal is to explore/optimize the material properties. To achieve this I'd suggest you to consider using parameters controlling the material properties rather than the ID. Please look at the example in the tutorial

    https://help.altair.com/hwdesktop/hst/topics/tutorials/hst/tut_hs_1070_t.htm#tut_hs_1070_t

    Regards.