Risa 3D vs RAM Elements Vs spMats

[MKarr]

I am trying to model a section of a retaining wall that is very close to an elevator pit and some steel columns in Risa 3D. I believe that my model is correct but the engineer I am doing this for is not sure about the Mxy result and more specifically adding it to Mxx and Myy for reinforcement checks (he is more of a RAM Elements guy).

To compare results we created dummy models in Risa 3D, RAM Elements, and spMats to compare values.
The dummy model is a simple 10x10x1 (3ksi N.W.) footing with a 50k point load at the center, and pinned supports at each exterior corner.

The overall global reactions are the same, and shell/plate moment contours appear to be similar in shape. The values of each one however are different.

Mxx contours are as follows:
Risa: 19 kft/ft
spMats: 24 kft/ft
Ram: 30 kft/ft

All models are set up the same, and from what we can tell have the same settings as far as cracked/uncracked concrete.

So if the overall reactions are the same, why are the internal moments so different?

Thank you for your time.

 

[JAE]

Are some factored and some not factored?

Check out Eng-Tips Forum's Policies here:
faq731-376

 

[JAE]

I get about 20 ft-kip/ft with RISA 3D.
This is with self-weight included and a load factor of 1.2 on everything.

Check out Eng-Tips Forum's Policies here:
faq731-376

 

[JoshPlumSE]

Well, hard to tell. Different plate element formulations will give different results. Some thoughts on this:
1) I believe SP Mats used to use a thin plate element formulation. So, as the mat becomes thicker and shear deformation becomes more important, the results would not be as accurate. That may still be the case. I'll follow this up with a post referencing a technical paper on our website that talks about this.
2) I have no idea what element formulation RAM elements is using.
3) The amount of meshing required to produce the most accurate results can vary with the element formulation. So, even if the element is a good element, it may (or may not) require additional meshing when compared with another. It wouldn't surprise me if there is more variation in the Mxy moments with mesh than there is with Mx or My. Especially as you get into the corners where the rate of change of Mxy is high.

 

[JoshPlumSE]

Here it is:

Finite Element Considerations for Modeling of Mat Foundationsink

Written by Dr. Umesh Puri (a former employee of RISA) around the time we released RISAFoundation version 1.0. So, that was a number of years ago. Therefore, I cannot remember the conference where it was presented.

Take a look at the output for "software B". Does it look like PCA Mats? PCA Mats was was the predecessor to SpMats, I'm pretty sure. Now, that doesn't mean that the current version of SPMats using the same plate formulation, of course.

Note:
The result differences noted in the paper, while they confirm the basic thesis of the paper, were not as dramatic as we had thought they would be. So, while the difference the OP found may be partially due to FEM formulation, I have my doubts about whether this would account for the entirety of the difference.

 

[MKarr]

Thank you both for the responses.

JAE: Risa and spMats have 1.0 factor with the one point load and self weight turned on. I'm not too familiar with Ram Elements, but best I can tell that is all that is being considered as well. All models have a 1.0 factor from what I can tell (couldn't figure it out in Ram, but the joint reactions are the same so I feel confident it's only 1.0). I got similar results to what you got when I ran my model with a 1.2 factor. I did notice that if I rendered the Mxy contours in both Ram and Risa that the magnitudes of the contours were the same. I also found that if I divided the Mxx results in Ram by 1.4 the result agrees more with what Risa and spMats are producing.

JoshPlum: I'll also do some looking into on points 1 & 2, as to 3 I did some information compiling of Mxy results from my Risa model. I noticed from the first few mesh iterations that smaller plates that came from larger plates tended to have decreasing values the finer the mesh got until I got to my finest mesh, then they jumped back up and the values became more erratic. Also as I said above the Mxy plots in Risa and Ram were very similar. Thank you for the link, I'll read through it. That one figure does look like spMats.

On a side note, while I'm not crazy about Ram Elements, I do think its contour plots are a lot cleaner than Risa's.

 

[JoshPlumSE]

I re-read a portion of this and it's making more sense to me now that my head cold is cleared up.

Where are the maximum values for those contours that differ so much? If they are right at the support lines then I know what the issue is...

RISA bases its contours on the plate element forces only. This is great for the values at the center of plates. It's even good for projecting the values between adjacent plates. But, it's not so great at the type of discontinuities that occur at the very edges of a model. Therefore, some programs will throw in a contour "correction" for known "boundary values" at support locations or free edges. RISA doesn't currently do this, which probably explains the difference. Though I did write up an enhancement request for this a couple of year ago. I will add you to the list of users who have requested this.

 

[MKarr]

Somehow all my plates got renumbered after I compiled my spreadsheet, but from what I recall from yesterday it was the plates closest to the supports and center. The plate forces for the plates in between those areas varied with some of them having the highest magnitude. The ones with the high values were also near the edges.

I also originally created footings with soil springs as opposed to the pinned corners. The results between those two models varies greatly.

 

[MKarr]

So I have determined why the results between Elements and Risa are so different. Risa uses thick plates as the link JoshPlum shared above states, and Elements uses thin plate assumptions. The difference in my moment magnitudes between the programs is similar to those shown in one of the figures in the above link, and my deflection in Risa is higher than in Elements again as stated in the link. So now I am just trying to find a correlation in my results to determine which moment to use.