Non Planar Plate Elements

[JosefTaylor]

Hi,

I'm getting an error 2034 "Plate may be very poorly shaped", which the help tells me is probably because my quad is not planar.

The plates represent corner supported glass panes which we are only using for load distribution (making use of the plate corner releases), so they're a single element per pane of glass.

The funny thing is that the error only comes up when I run a case with projected area loads. Surface normal loading works fine, and so does self weight. The geometry is complex, brought in from Rhino/grasshopper with some python scripts. The quads there are flat within the 1e-5 meter tolerance in Rhino (versus the 2.5e-4m tolerance in RISA). We've upped the joint coordinate precision in that process to the 14 decimal places allowed in the text file, but it still pops up the error. I'd rather not triangulate them, as that mucks up the load distribution. I've tried super stiff and super soft material properties to see if that helps, but it does not.

Any thoughts?

Thanks,

Josef

 

[JoshPlumSE]

I've not seen the program be wrong on the non-planar plates message before. But, it should happens whenever the model is solved. that's because the check is part of the stiffness formulation which is independent of loading.

That being said, the program uses a similar routine for the member area loads to make sure that they are planar as well. I don't know what version of the program you are using. But, it is a possibility that the version you are using improperly triggers one error message when it should have triggered a slightly different one.

How many area loads do you have? You might investigate the issue by deleting a few of the area loads to see if there is one that is causing a problem. If you know which one causes the problem then you can take a closer look at it to see if it is truly poorly formed or not.

 

[JosefTaylor]

Thanks Josh,

I'm using 8.1.2

About it failing only with projected loads, that's exactly what I thought -- it should be checking planarity only when it is writing stiffness. Could there be a similar check when calculating load projections which has a slightly different tolerance? Perhaps there's a mess with units somewhere; I've noticed RISA isn't the best at keeping quantities up to date with their units.

These plates are near vertical - does RISA have trouble with projecting vertical loads to near vertical faces? My solution, for now, is to only apply the load where panels are not too vertical. Kind of defeats the point of a projected load though!

Looks like 7 out of 644 plates were triggering the error. I pulled my point coordinates back out into CAD and checked the planarity, the worst is 1e-6 m out, which is pretty dang good!

Thanks,

Josef

 

[JoshPlumSE]

So, it is related to some plate elements being out of plane?! I cannot think of any reason why that would ever be a load dependent error message.

Feel free to post your model if you like. I can at least test it to see if the current version 12.0.2 (soon to be 13.0) behaves the same way.

 

[JosefTaylor]

The plate elements are very slightly out of plane, but not anywhere near the tolerance stated in the manual. There are other near vertical panels in this model which don't throw the error.

Yeah, please have a look in the latest version. We should probably get up to date!

 

[JoshPlumSE]

Josef -

Thanks for posting this. It took me awhile digging through and experimenting with the model. But, now that I found it the issue is obvious.

The material that you made your plate elements out of has a young's modulus of zero (or very close to it). This is wreaking havoc with the plate element formulation, but in a way that we did not anticipate. Hence you get an error message, but it is inconsistent and dependent on load. My guess is the plate element formulation works properly (thought the stiffness is so close to zero that you are likely to get instabilities). But, when it comes down to process transverse area loads, the program detects a problem. And, we have code built to catch that problem, but that code was added years ago when we intended to catch poorly formed plates (like bow ties or such) rather than this. Hence the error message you get.

Put in a reasonable E value and you should be okay.