Understanding the Use of Midside Nodes in ANSYS for Structural Analysis

[minger]

OK, I am having a problem. I come from CFD where we do not use "midside nodes". I am currently in a structural position where we use ANSYS for everything. I typically use the 'eyeball' test to see whether or not my mesh needs refining. I assume that it does not need to be quite as fine as a CFD run. With that said, our structures lead insists on using midside nodes, even though any time I try to use them, I get erroneous answers. I've read through the ANSYS help, which compares linear (no mids) and quadratic (mids) elements. It basically just says that it you are using linear elements, make sure the grid is "decent" in areas of high gradients (makes sense). It also even says that for nonlinear structural analses, one typically obtains better accuracy at less expense if you use a fine mesh of linear elements rather than a coarse mesh of quadratic elements.

It goes on even to say that for non-structural analyses, the linear elements are as good as the higher-order elements, and are less expensive to use. Even degenerate elements produce accurate results for these analyses. The current problem I'm doing involves calculating equivalent housing spring rates...easy. I'm applying a small load on a housing and looking for deflections. It seems to me that using linear elements is perfectly fine. More importantly, I'm getting good results.

I guess the question is: what are your thoughts on using midside nodes? When have they hurt/helped? Anytime I've personally tried to use them, they have failed and failed miserably.

Also, what exactly is being solved at the midside nodes? It can't be the entire governing equations, otherwise they would simply be additional nodes/elements. It must be some simple intraelemental equation that's being solved, but I can't find information anywhere. I really want some good background on these things so I can determine if they are worth my time.

thanks a lot,

 

[FredGarvin]

It sounds like you are having issues with coincident nodes being aligned when doing the mesh. That's a guess on my part. I pretty much stink in that aspect of ANSYS.

As far as the mid side node equations are concerned, the mid side nodes increase the DOF of each element. They certainly do increase the complexety of the stiffness matrix and the governing equations. For a 3 sided element with three nodes the governing equations are linear whereas with the midside nodes it becomes a quadratic.

If you look at a shape function for a linear three node element, the shape function looks something like:

\phi (x,y) = \alpha_1 + \alpha_2 x + \alpha_3 y

The same element with 6 nodes looks like:
\phi (x,y) = \alpha_1 + \alpha_2 x + \alpha_3 y + \alpha_4 x^2 + \alpha_5 xy + \alpha_6 y^2

I have to admit that I too don't understand what the need for the mid side nodes is unless you need some kind of localized refinement. Again, I am not a master when it comes to meshing so I can't answer that. There are other stupid things that I see people insist on that are against my teachings, but people insist they are correct. I chalk them up to more than one way to skin a cat. The big thing is, you can't get your mesh to calm down and play nice. And that is not a good thing. I'll see if I can't dig up some more info.

BTW, if you are ever in the need of a good ANSYS book, I would recommend Madenci & Guven's book "The Finite Method and Applications in Engineering Using ANSYS." It was a huge help in my FEA classes and as a reference at work.

 

[minger]

The contact elements shouldn't be a problem. I've talked to the ANSYS guys, and they said that it 'used' to be a problem. The new surface contact I've been told if robust enough to not require coincident nodes.

I agree, its definitely hard to try and fight with someone who is so set in their ways. It seems as if some of the old timers (no offense to anyone) here don't quite keep up on the current state of the art in terms of software and tools. Compound that with the fact that I'm just a newb makes for a lot of wasted time doing things twice.

Anyways, thanks for the reference, I'll definitely check it out.