hexagon logo
Register Sign in
  • State Not Answered
  • Replies 5 replies
  • Subscribers 4 subscribers
  • Views 119 views
  • Users 0 members are here
Options
Related

How to make the result from the Superelement reduction more accurate?

noravivian.yazdandoostkhosravi
I created a plate with an frequency dependent force from 20 to 1000 Hz. I analysed the plate with SOL 111. For the Frequency Range of Interest for the Real Eigenvalue Extraction and for the Frequency Response solution parameters frequency recovery points I took 20 - 1000 Hz as attitude.
I create a graph out of the reults. The blue graph is the one with the full model. The red one is the result out of a superelement analysis.
How do I get a result with superelement method which is more accurate to the original?
Attached Files (1)
  • 0
    Are you able to supply your input files? If so, I can edit to improve the results...
  • 0
    Of course, here are the input files. I changed the frequenzyrange to 0 - 500 Hz.
    The bdf OhneSuper ist die original model, the one named super is the reduction analysis
    Attached Files (1)
  • 0
    Here are my observations:
     
    1) you do not want to do modal reduction of the residual prior to assembly with the upstream SE, so removed AUTOQSET and explicitly defined SPOINT/QSET for component mode placeholders within each upstream SE section
     
    2) it is always best to include modes to a higher frequency for the upstream SE than the desired accuracy of the assembled modes... I changed your upstreams to solve Component Modes to 2000.0 Hz which is 4x your downstream desired upper frequency of 500 Hz
     
    3) I commented out the TABDMP1 for each upstream SE... if doing modal damping, it is already going to be applied at the assembled modes, so it is over-damping to also appy it to the upstream modes
     
    4) why are you applying a FORCE to GRID 25000071 in SE2 as well as in SE0? I removed this force and all the dynamic loads 'stuff' from SE1 and SE2
     
    5) I converted from the old LOADSET/LSEQ to just pointing to the loading directly
     
    6) converted to expanded case control to better control what is done for each SE
     
     
    A few spot checks seem to show same results between the SE version and NON-SE version...please do your own validations and let me know if I missed anything!
     
    Regards,
    Don
    Attached Files (1)
  • 0
    Dear all,
    I'm really interested in the approach you followed, but not sure if applicable to my case.
    I need to perform a sine analysis where some external superelements have higher damping wrt to residual structure.
    I also need to deliver a reduced model (again, by external SE) form the previous one.
     
    It is very likely that before final delivery, we need to perfom sensitivity analyses changing the damping value
    Therefore I want to apply it "externally" to SEs in the residual run.
     
    In theory I should :
    1. create a SUBCASE for every SE using SUPER instruction assigning damping value: as damping is additive, proper values shoud be chosen wrt residual damping to avoid overdamping for SE;
    2. create residual SUBCASE with nominal inputs.
     
    What I see in your file Super-djg is that in every SUBCASE relevant to SEs, there are three METHOD actually identical each other: do I need to repeat this instruction in order the analysis to work? Do I need to include an SPC as well, even if it doesn't change?
    As far as damping is concerned, I was thinking to use SEDAMP (and PARAM,SEDAMP,YES) for SEs subcases, pointing to each damping table (see annex).
    Please note theat there is also Case 01131684: "sine analysis with multiple SE and different damping".
     
    Thanks in advance and regards,
    Stefano.
     
© 2025 Hexagon AB and/or its subsidiaries. Privacy Policy | Cloud Services Agreement