RISA Common Errors...

[WBUW]

I've got a RISA 3d model built up. Used plates, its a pretty simple little support I probably missing something simple.

All the forces are in the X direction, vertical axis is the Z axis.

I have what is basically a column sitting on a plate with 4 pin nodes. I was expecting to see equal (or very close to) equal forces at the nodes in each direction...yet I'm seeing reactions in the X, Y and Z direction.

I did not expect to see any forces in the Y direction...I'm a bit baffled as to why they are there. I did have to connect a pipe to the plates on top of the support so I used 4 rigid sections. Looks like there is torsion in the rigid sections...maybe thats translating through the plates...if thats whats going on I could use a tip on how to connect these elements so it doesn't happen.

I uploaded the file to engineering.com haven't done this before so hopefully it works...was expecting a link or something to appear.

 

[WBUW]

There we go, just missed the final click looks like...

 

[JAE]

I think your vertical column sides (veeerrrry long finite elements) aren't the best modeling technique.

Usually try to limit your aspect ratio of your elemenent to 1.5:1 or prefereably as close to 1:1 as possible.

also - display your plate surface loads - you have a weird side panel surface load that doesn't look right.

 

[JoshPlumSE]

Well, the net reaction in the Y direction should certainly be zero. But, I wouldn't necessarily expect every joint to have a zero Y reaction.

Now, looking at that model, there are problems with the way you've defined your plate elements. The length of some of your plates is 5 in x 163 inches. That's an aspect ratio of more than 30 to 1. That's probably going to do some funny things to the plate element formulation.

To be safe, this aspect ratio should not exceed a value of 9.0. And, personally, I like to keep it close to 3.0 or so.

Also, some of your most critical plates are Tri's. The triangular element forumultion is not as accurate as the quad. It's really just a collapsed quad and can be viewed as a highly distorted quad element. Therefore, I'd suggest using the submesh tri's tool to convert those into quadrilateral elements. Then mesh the rest of the plates 2x2 to match up with the new mesh.

At that point, you should have a better model. You'll still get some Y reactions. But, at least you'll be able to trust the rest of your model behavior.

 

[WBUW]

Thanks Josh and JAE.

Actually, now that I think about it thats kind of funny...finite element analysis by using large elements! It does seem a little incompatable.

I made the changes, an updated model is attached.

X and Z reactions I get since there are forces and moments applied that would elicit that reaction...Y does seem to pretty much cancel out now.

Is the force in the Y direction due to deformation of the base plate? I guess that would make sense and normally when I do hand calcs I don't check that...

 

[JoshPlumSE]

The short answer is that, yes, the Y reaction at each node is due to deformation of the plate. Or, the two way shear resistance of the plate.

The bigger concern to me would be the fact that the SUM of the Y reactions is not equal to zero. Let's see if we can justify that.

I suspect that (for this model), the rigid links are behaving a bit too rigidly and you are getting some "ghost reactions". Ghost reactions occur when the internal stiffness of a member begins to approach or exceed the stiffness that we use for boundary conditions.

The program is usually smart enough to throw up a warning message when this happens. But, in this case, the magnitude of the load is so small (0.1 lbs compared to an applied load of 1000 lbs or so) that it is not detecting the lost load.

To solve this problem, we can change the E, I, A and J values of the rigid links to 1E5 rather than 1E6 or 1E8 which they currently are set to. This seems to eliminate the problem. Does it also reduce the Y reactions at individual joints? If it does, it's not by much.