RISA plate spreadsheet results don't match graphical output

[weab]

I have noticed this several times in doing FEA in RISA. I graphically show the Von Mises stresses using the contours option on the screen and notice the highest stress indicated by colors. Then I look at the plate stresses in the spreadsheet output and sort max to min. The highest stress is much more than that shown on the screen and it's not a rounding issue. For example, my model graphically states the highest stress is around 37 ksi. The plate spreadsheet shows plates with over 41 ksi.

Even more specifically again for Von Mises, the graphical output will show colors indicating the stresses in a plate. The key at the top right indicates the stress values. Based on the color on specific plates, the stress could not be the stress indicated on the spreadsheet results for that plate. Is this typical? Am I misinterpreting the results?

On a different note, why do stresses seem to drop off at the last plate before an intersection. For example, I have modeled a wide flange with wheel loads for an underhung monorail trolley. As expected, the stresses increase from the load application point toward the web. However at the final flange plate element where the flange connects to the web, the stresses decrease somewhat. This would seem clearly wrong. I find similar things in other models like this occurring one plate element from perpendicular intersecting plates.

This is a very simple model so I'd be surprised to find out that I am doing something wrong, but maybe so. Is this a quirk of FEA modeling? Is this a RISA issue? Any help is appreciated.

 

[JoshPlumSE]

Weab -

This is a fairly common question. The plate force results represent the most accurate value for the natural center of that plate. But, that is a single point in the plate. The contours must show values for the ENTIRE plate projected all the way out to the edges and corners. If you compare the spreadsheet values to the contour color you see at the center of the plate, you should get pretty good correlation... most of the time.

The 2nd question likely has to do with the way plate contours work, especially when you have multiple plates meshing together at a given joint (like the web / flange intersection). Accurate values for any plate result (stress, force, moment etc) are only known within the gaussian integration points of the plate. Think of this as the center 1/2 of so of the plat area. But, since contours have to show values all the way out to the edge, we have use some approximate methods for projecting these plate results out to these locations.

RISA does this projection of contour values by setting "nodal" values for each result. So, the program looks at all the plates that connect to a given node and averages the values based on the nearest gaussian integration point for each plate. Now, we can project these values out to all surfaces of a plate mesh. The only drawback is if you have plate elements connecting to the same joint that have vastly different stress levels, then the stress level in the vicinity of that joint is not as accurate as it could be if the plates were viewed independently. The high stress plates show some decreasing stress as they go into that joint and the low stress plates show some increasing stress as they go into that joint. Now, I haven't seen your model at all. But, based on your description, I believe this would explain the behavior that you are seeing at the flange / web intersection.

 

[JAE]

And with all finite element analyses, keep your plates small in areas of focus and stress changes to minimize FEA error.

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

 

[weab]

Thanks Josh and JAE.

Josh, that's quite an explanation.

Right or wrong, I understand all this to say that it is not possible to display totally accurate results or that RISA chooses a simpler methodology/approximation that can sometimes be not all that accurate. I also understand, as JAE says and I already know but don't always recall, that critical areas need a tighter mesh. In this particular case, the tighter meshes benefit areas of intersecting plates by allowing the smaller intersecting plates of the model to be "ignored". That is, if the intersecting plates can be said to be inconsequentially small, then the results of the neighboring plates are close enough to that of the intersecting plates and provide reasonably accurate hot spot stresses.

In my particular case, I placed a node at the web/flange intersection, of course. The first flange mesh width is equal to the root plus half web width. Therefore, I am able to accurately view the stresses at the base of the root (the critical location for local flange bending) before the RISA results go haywire between the root and the web.

I hope this accurately summarizes the what you have stated. Thanks for the assistance.