STAAD Triangles

[Althalus]

I tried asking Bentley this question and they never answered.

I've noticed that if I use a triangular plate in a STAAD model, the resulting plate stresses are unusual. For some reason, a few of the stresses are completely uniform. All three corner stresses and the center stress are all the same value.

This doesn't happen with rectangular plates or trapezoids. Only the triangles.

Any explanations?

 

[SWComposites]

This a 3 noded triangle? If so, then that is exactly how they are formulated. See a good FE theory text.

 

[lexpatrie]

The heck is a four noded triangle?

What you're describing sounds like a constant strain triangle, not sure why any commercial software would have those besides backward compatibility to say, a 1960 version of the program.

https://en.wikipedia.org/wiki/Constant_strain_triangle_element

 

[SWComposites]

Sorry, was thinking solid tet elements.

 

[JoshPlumSE]

The heck is a four noded triangle?

Well, some programs use a true 3-node plate formulation. Other programs (like RISA, I believe) use a 4-node / quadrilateral formulation. I remember the referring to their triangular plates as a "collapsed quad". Also, when I worked there, we suggested that triangular plates should be avoided except where you made mesh transitions. Hopefully in places that were lower stress.

Now, I don't know what STAAD is doing with their Tri's. You should check their documentation.

 

[271828]

No familiarity with STAAD.

I'm not sure if you're referring to a "plate" as in out-of-plane degrees-of-freedom only, or a "shell" with in- and out-of-plane DOFs.

In SAP2000, triangular shells work well for out-of-plane loads, but poorly for in-plane. For example, if you model a beam with vertical triangular elements and add a vertical load, the computed deflection won't be even close to what you would get from manual calcs. Doesn't matter how fine the mesh is. Repeat this with rectangular shells and the results are close. This is discussed at length in their documentation. If you're using a shell, you might be seeing the same thing.

 

[WARose]

Is this for a in-plane load or out-of-plane? What you are describing sounds like classic CST behavior. But I have tested STAAD's triangle elements (especially for out-of-plane loads that induce bending) and they do not behave like CST elements.

It could be that your mesh is so fine this is occurring.

 

[Althalus]

these are plate/shell elements. Staad doesn't make a distinction. They are either line elements (beams/columns) or they are plates. The plates can be designated with various thicknesses and give the properties of steel or concrete. In this case, concrete.

This is for out-of-plane loads.

I have trouble believing that it is supposed to behave this way. Why don't the quadrilaterals display this behavior?

 

[lexpatrie]

Because 4 node elements have more degrees of freedom (8).

Most of the older finite elements have published information on them in the various journals, including some of the (perhaps older) proprietary software programs promoting their elements. I want to say the origin is largely aerospace and it migrated into structural engineering, but I'm not a historian.

 

[SWComposites]

Some FEM history:

https://www.odonnellconsulting.com/....,Deflection Analysis of Complex Structures.”

 

[TLHS]

Does this happen at all levels of mesh resolution? I could see this happening at really small mesh sizes with constant loads and without stress concentrations. Basically, the differential across the element is so small as to not significantly change. I suspect this isn't your situation though.

I have generally not used triangular plate elements in STAAD because I don't trust the post-processing part of the software that will transform things into the global axes for directional moments and things, after having gotten burned by it before. This is a dumb thing to worry about in modern software, but I'll manually draw rectangular plates if necessary so that I can align the axes in ways that let me process the results better.

Maybe post the std file, or (if it's small) just past it into a post.