Which is more accurate,powergraph on or off?

[Lirock]

Hi,everyone:
The result is different when powergraph being turned on and off.I wonder which is more accurate?

 

[Drej]

It depends. Probably neither. And "accurate" is probably not the appropriate word to use. The integration point solution (obtained from ETABLE) gives the unaveraged and non-extrapolated solution from the element, so this is probably the most "accurate" if you want to use this word. Capturing the (stress) response -- as opposed to being accurate -- depends principally on having the appropriate mesh density for your model, so the most important thing to get right is this. Powergraphics carries out some averaging which the "Full" method does not. For example, it only averages across elements at the surface of a model (hence doesn't matter in this instance for shells or beams), whereas the full method averages across all. Powergraphics also does not average across material boundaries in your model, as well as a few other nuances that you should be familiar with. Take a look at the help file for a complete description (see the /graph command for example). Choose the integration point solutions if you want the fundamental (i.e. unaveraged) element quantites, but remember that the nodal solution (PLESOL and PLNSOL) are averaged and/or extrapolated values from the integration points. Generally speaking, if your mesh is right, these quantites should converge.


------------
See faq569-1083 for details on how to make best use of Eng-Tips.com

 

[cbrn]

Hi,

Drej's explanation is very accurate. I'd only add a strictly personal point of view on how taking best advantage of the way Powergraphics works:
- generally don't rely blindly on the results you "see" whenever powergraphics ON and OFF give very different results: in these cases, the mesh density is not adequate locally; so, either refine locally, or "invent" some hand-made post-processing methods (best-fits, corrections based upon analytical Stress Concentration formulae, etc...)
- powergraphics ON will amplify surface-effect based S.C., while by turning it OFF you will decrease their weight (in our factory, it's a method we intentionally use for that reason because we know it to be more realistic when compared to experimental results. In our case, the best correlation "numeric-vs-experimental" is achieved with "nodal" results with Powergraphics off (for all the "routine" works; of course when we set up "exotic" verification models, we can't rely on this and further investigation is necessary), but in your case it may or may not be the same. You have 5 possibilities (integration points results, elem results w/ Power ON, elem w/ Power OFF, nodal w/ Power ON and nodal w/ Power OFF), so be careful...
- there are cases in which the S.C. is due to non-removable geometric reasons (triple corner, etc...), so that even increasing mesh density will not really help. In these cases, PowerG/ON will worsen the situation...

Regards