hexagon logo
Register Sign in
  • State Not Answered
  • Replies 5 replies
  • Subscribers 4 subscribers
  • Views 63 views
  • Users 0 members are here
Options
Related

Origins of residual vectors

anthony.carlson
Howdy,
 
First time posting here, please be nice :)
 
It is my understanding that the calculation of residual vectors is triggered in a variety of ways (loads, inertial, rvdof, etc). Assuming that is true is there a way determine the origin of each residual vector in a particular run?
 
Thanks in advance for any input.
  • 0
    That is an interesting question... I can't think of anything other than turning the triggers on one at a time and noting what you get... but even that might make it difficult to know what residual vector came from which trigger when all are on at the same time, since these vectors will all be filtered and re-orthogonalized after creation.
     
    How would this information be useful?
  • 0
    Don,
     
    Thanks for the quick reply on this one. Yes that process of elimination approach should be insightful.
     
    Regarding the usefulness of knowing the origin, not sure atm, I am just curious.
     
    Related thought/possibility: I haven't reviewed the underlying math to answer this but is it possible that these residual vectors could be mixed together when they are calc'ed > synthesized > re-orthogonalized. If so then the final residual vector could have multiple origins? Or is it always single origin?
  • 0
    Without knowing the math myself, I'll just be guessing, but I think they do remain single origin (that's not to say two vectors from different sources can't be the 'same' and one get filtered out, leaving the other)... let's see if others on this forum with deeper knowledge will step up and comment!
  • 0
    This question comes up occasionally. Tracking the path of resvecs through the DMAP logic indicates there
    are 7 steps to produce a numerically relevent resvec:
     
    1.Calculate the residual load DP
    2.Filter DP for duplicates – flushes singularities
    3.Filter DP for numerical zero load vectors
    4.Calculate DU physical space residual vectors
    5.Filter DU for duplicates – flushes singularities
    6.Filter DU for numerical zero displacement vectors
    7.Orthogonalize DU to generated modal space residual vectors
     
    Some of these steps have been combined for efficiency. When a resvec is eliminated, the column (each resvec is a column in a matrix) gets removed. I forget if I figured out the logical order of building the initial DP vectors.
     
    I wrote a DMAP to export the resvecs into individual .op2 files for post processing. The DMAP logic was written as generic as possible to work without updates in post V2014 Nastran releases.
     
    I suggest working with one residual vector source using the CASE CONTROL RESVEC options. The bigger the model and the more residual vector sources can really make understanding 'this residual vector from that load' difficult. Key on the displacement output. Maybe run a static SOL 101 with the same loads, each load an individual subcase.
     
    My opinion, if a residual vector is more than small isolated areas of displacement in your model, then you do not have enough dynamic modes. Up your F2 on the EIGRL. Do note that with mode extraction, the residual vectors will be available in the eigenvector output (DISP command) for review.
     
    And although Don did not mention this, but I know he agrees, RESVEC(NODYNRSP)=YES is recommended. Run a model with and without this setting. If the answers are different - the residual vectors are contributing dynamic responses and this is an error.
     
    Attached Files (1)
© 2025 Hexagon AB and/or its subsidiaries. Privacy Policy | Cloud Services Agreement