General Mechanical

General Mechanical

Parameters demoted to “derived parameters” during the “Design of Experiments” phase of optimisation

    • pilotNodes
      Subscriber

      Hello,

      I am currently trying to perform a response surface optimisation of a parametric design of a beam attached to a larger model. I am using five parameters, each of which correspond to a certain dimension of the beam.

      Prior to running the design points generated by the "Design of Experiments" tab in the optimisation module, I have verified that each of the parameters are properly linked, as the design can be changed via the parameters both from within Spaceclaim and from the project level in Workbench. However, once I start to update all of the design points which the optimisation module generates, it only manages to successfully run about 8 of the roughly 30 design points before two parameters become unlinked from the parameters in the Spaceclaim model. At this point they appear as "derived parameters" as shown in the picture below:

    • pilotNodes
      Subscriber
      Oh and also, I forgot to mention, I am using the academic research license for mechanical, if that is of any help.
    • peteroznewman
      Subscriber
      I have experienced failures in DOE with shape parameters from SpaceClaim because the dimension change caused a face to be consumed which is not recovered when the dimension returns to the original value. That is why I prefer DesignModeler for shape optimization. However, you seem to have checked that this is not the case for your model.
      I don't know what you mean by a parameter being demoted. Can you show an example of what that looks like? Can you upload your .wbpz archive file and provide instructions to reproduce the error?
    • pilotNodes
      Subscriber
      By "demoted", basically what I mean is that the parameter becomes a "derived parameter" (that's what it says when I right click it in the table). I actually don't know what it means, but it appears to shift slightly towards the left in the table when it happens, and what was previously a regular parameter seemingly loses its link to the geometry file. So what this seems to mean is that they can't be used as inputs for the geometry - changing the values of these "demoted" parameters causes no change in the geometry.
      I can't share the project file as it is work related, sorry. But I don't really know how to achieve this error either, it just seems to happen at a random point. I simply have 5 parameters which are each linked to the pulling of five faces in Spaceclaim (and yes, as you say, I am avoiding the issue of the faces being consumed by using limits to the dimensions). I generate the design points using the preview button in the "Design of Experiments" module in the Response Surface Optimisation, and at this point I can see that all of the values for the parameters in each design points abides by the limits I have set up. At this point I can also manually enter the parameter values in each of these design points and still see the geometry correctly change. At this point, all of the parameters are aligned in the table, meaning that those two that have shifted to the left in my picture are initially aligned with the others (and right clicking them causes no mention of derived parameters).
      However, once I actually press the update button to calculate the output for each design point, it only manages to succesfully update the geometry for a few of the design points before some of the parameters inexplicably lose their links to the dimensions. And this occurs for design points which could previously be entered manually successfully. Once this demotion occurs however (which is when I observe the affected parameters shifting to the left in the table), the solver fails to update the geometry for every design point after this. This is noticeable for example through the output mass of the beam, which no longer meets the expected values given the dimensions the program should (but fails to) assign. After the demotion, it is also no longer possible to change the corresponding dimensions by updating these derived parameters, and it is instead only possible to change within Spaceclaim. Within spaceclaim, the parameters still exist, and still have their links, but since the link appears to be broken between the parameter list in workbench and the parameters in Spaceclaim, the solver still fails to change the geometry correctly.
      I guess it is impossible for you to help me since I cannot share my files, but I will let this topic remain here in case somebody else happens to have encountered the same issue, or in case somebody else encounters it in the future. Have a nice day :)
    • pilotNodes
      Subscriber
      I should probably add that in continuing to attempt to solve this problem by redoing the parametrisation once more, my project seemingly became corrupted, showing the same errors as in this thread:
      https://forum.ansys.com/discussion/comment/151241 so maybe that could be an explanation for the issues I am having. I hope this helps anybody who happens to have the same issue (however at the time of writing there seems to be no apparent fix to the issue raised in the other thread, as nobody has answered it yet).
    • peteroznewman
      Subscriber
      I use DesignModeler for shape parameters. It is much more robust than SpaceClaim.
    • pilotNodes
      Subscriber
      Yes I realise that is probably the way to go. Thank you :)
Viewing 6 reply threads
  • You must be logged in to reply to this topic.