hexagon logo
Register Sign in
  • State Not Answered
  • Replies 9 replies
  • Subscribers 4 subscribers
  • Views 77 views
  • Users 0 members are here
Options
Related

How does NODYNRSP do its thing?

anthony.carlson
Does anyone know how the NODYNRSP option of the RESVEC command do what it does?  
 
I haven't been able to find any documentation on this detail.
 
My (wild) guess is a low pass frequency dependent damping is applied to the modes associated with residual vectors to allow them to be suppressed at high frequency but allowed to responded at low frequencies... 
 
Second question on this topic (which is likely related to the first) is why the NODYNRSP option embedded into the RESVEC call. It seems like it could/should be a separate option on top of the RESVEC command.
  • 0
    Unfortunately, The theory and equation behind NODYNRSP are not exposed in the documentation. QRG tries to explain, perhaps not effectively. That is why, the best approach is to always include enough normal modes and manually try to filter high frequency excitation from the loading as the modes are at the high frequency are not properly represented by the mesh density. The reason why NODYNRSP is included in RESVEC command would be because it only applies to residual vectors. Thanks. Sorry, if you were expecting more.
  • 0
    Hi- if my memory is correct, NODYNRSP zeros out the modal mass of the residual vector modes... that way they can still participate in the 'static' portion of the response, but not the 'dynamic'. Maybe someone else can confirm this.
     
    As far as how this gets triggered, it was put into the RESVEC command as an option, allowing it to be turned on or off along with any other RESVEC options. To me, it is nice to have it contained within the RESCEV command to keep all options together.
  • 0
    Don,
     
    Interesting.  
     
    So in the context of external super elements the reduced boundary mass matrix (equation 9-30 in the 2019 superelements users guide) are the terms you are discussing zeroing out the Mqt terms for the resvec dofs or the Mqq terms, or both?
     
    And if I wanted to test this, how do I write the reduced boundary Maa matrix to a human readable text file?
  • 0
    Unfortunately, NODYNRSP only operates at the residual level, so external SE will not be affected.
     
  • 0
    Don
     
    In another thread on the topic of residual vectors you linked to a paper by Ted Rose from 1991 on the subject. There Ted presents the development of residual vectors in the context of fixed boundary CMS. I guess that's why I was asking about the reduced boundary mass matrix in my previous question.
     
    With what you said in your previous reply I'll rephrase/extend my previous question.
     
    Assume residual vectors were calculated during a fixed boundary EXTSEOUT run. Lets say 10 retained fixed boundary modes and 6 inertial resvecs so the q-set has a size of 16. The RESVEC call had the option NODYNRSP.
     
    This superelement is then used alone in a SOL 112 run. In this case I think the assembled M^0_gg matrix should be essentially the same as the M^1_aa matrix as there is no residual structure and a single super element is considered.
     
    At this point similar question as before; are the terms you are discussing zeroing out of the M^0_gg matrix those that correspond to the 6 resvec dofs in M^1_qt partition of M^1_aa or the M^1_qq partition of M^1_aa, or both?
     
    In my example above the residual vectors were created during the superelement generation. For the downstream use of this superelement in the SOL 112 run it seems they don't need to be recalculated to be utilized, right? If this is the case how is the NODYNRSP setting communicated to the SOL 112 run. If the residual vectors must be recalculated in the SOL 112 run then is there any benefit from ever calculating them upstream in the superelement generation step?
  • 0
    Once the external SE is brought into the assembly run, the program no longer knows which modes are residual vector modes- they are all just 'modes'. Then, when these upstream modes are assembled and a new eigensolution obtained, the new modes could be a blend of those upstream modes and residual structure. There is no longer any means of removing just a portion of the 'new' modes dynamic response from the solution... only the residual 'assembled' modes can then have the newly calculated residual vectors appended. And these new residual vectors can be modified via the NODYNRSP tag to remove the modal mass so they don't respond dynamically.
     
    In your senario, you are just bringing down a single component into a residual with no additional structure... but that is not the normal case. Normally, you will be assembling upstream stuff with downstream stuff and doing a new eigensolution.
     
    But even in your senario, you will be resolving the 'system' modes since there are pysical and modal DOF coming down... you will perform a new eigensolution. At this point you can also generate residual vectors that can be affected by NODYNRSP.
     
    Sorry if this is confusing, as I am in a bit of a hurry!
  • 0
    Don,
     
    Ok it is making some more sense here.
     
    A recap of what you said: any residual vectors generated in an upstream superelement, with NODYNRSP enabled, have no bearing on how assembly run residual vectors will dynamically respond or not. This is now clear to me, thanks for that.
     
    To test your idea from your original comment I did the following:
     
    A) I generated a superelement with resvec turned off.
     
    B) In two separate SOL112 assembly runs using only this superelement (no residual structure) I kept everything the same except for:
     
    Run 1: RESVEC,NODYNRSP = YES
    Run 2: RESVEC,DYNRSP = YES
     
    I don't know how to directly dump the modal mass matrices but I did add in MEFFMASS(ALL)=YES hoping to see something different between the two runs
     
    Confirmed in both that residual vectors were generated in both runs (total of 8).
     
    When I looked at the generated .f06 files with a diff tool, they were identical except for the case control echo.
     
    I am not sure I am fully testing your theory. Could you let me know how to dump the modal mass matrices directly?
     
    I can share files if needed.
  • 0
    we took this conversation over to tech support... this specific question was about why the residual vectors were the same with and without the NODYNRSP flag... that is because NODYNRSP is enacted after the residual vectors are calculated, but before they are used in any subsequent analysis... it only impacts the modal mass associated with the residual vectors. I wanted to post this in case anyone was following this discussion.
  • 0
    Don,
     
    Thanks again for the help on this topic and closing the loop back here in the forum.
© 2025 Hexagon AB and/or its subsidiaries. Privacy Policy | Cloud Services Agreement