I changed the fixtures of my I-beam (not the one in the above screenshot), and now it seems to work - even if I change the fixtures back to how they were originally. Hopefully this won't be a problem from now on.


Moderators: Feel free to delete this thread if it seems superfluous now.

The problem is obvious in your screenshot. Insufficient constraints allow the model to float into space. You could use soft springs to get around that, but there is really no reason to do that.

I will say the dialog should be more helpful. At least a singularity error would be helpful. You already know it failed to run when the dialog popped up.

TOP
CSWP, BSSE
www.engtran.com  www.niswug.org
www.linkedin.com/in/engineeringtransport

"Node news is good news."

Yeah well, the screenshot shows the second model I made as a test. The first one was an I-beam with more supports. I changed it, and it worked. Then I changed it back to how it was, and it worked. Then I tried the test model (as shown on the screenshot) and that one worked as well.

So, even though it is only restricted from moving in one direction (the x-axis), it is only affected with loads in that same direction. So, I guess it should be fine, right? Or at least now the program runs the analysis without problems and yields the expected results.

Numerical round-off errors can cause this part to fail to solve, despite the load being uniaxial. The single restraint might not be enough.

Certified SolidWorks Professional

You cannot depend on fortuitous results to be consistent. The fact is that a static analysis needs to be fully constrained to prevent rigid body motion. You are working with a math model on a computer which introduces phenomenon not seen in the real world. Round off error is one of those things. Singular and ill-conditioned matrices is another.

The fact is, what you are simulating is a beam with the end sitting on an infinitely slippery sheet of ice. It would be with the application of great skill in the real world that the end of the beam would not go shooting off in some direction when a load is applied.

I use the 3-2-1 rule to start with:

3. Constrain 3 points, a line and a point, or a planar face in one direction
2. Constrain 2 points, or a line in an orthogonal direction
1. Constrain a single point in a direction orthogonal to the other constraints.

TOP
CSWP, BSSE
www.engtran.com  www.niswug.org
www.linkedin.com/in/engineeringtransport

"Node news is good news."

Alright. I didn't realise that the program could succeed in doing a calculation by making a round-off error.

I was trying to make an I-beam to see how the stresses would distribute (compared to manual elastic beam calculations). And so now this is my model:



It seems to me now that this model i statically determined. Am I missing something?  

Run a frequency analysis on this beam. Show the first and second modes.

This model looks like a shell model. Is this correct?

Round off is a natural part of FEA. See Cook.  FAQ1183-1711: What is a good book to learn finite lement analysis from?

TOP
CSWP, BSSE
www.engtran.com  www.niswug.org
www.linkedin.com/in/engineeringtransport

"Node news is good news."

@kellnerp

Why do you go with 3-2-1 constraining? Is it so you can do a hand calc with the same setup? FEA doesn't care if it statically indeterminate, right? Also, shouldn't the BCs be setup how the real life model is?

I'm not correcting you, just trying to learn why, because from the other posts I've seen from you in other threads you seems to be pretty experienced, more than me anyways

@ninjaengineer
3-2-1 constraining is the minimum necessary to avoid having the model drift into space. This is necessary to solve the problem as a static problem.

FEA is very good at statically indeterminate. The problem with the original beam constraints is that the model was set up to be both statically indeterminate and under-constrained at the same time.

Yes the boundary conditions should reflect the actual problem being solved taking into account the assumptions of the FEA method and material mechanics. For example in both problems setups above the beam is being constrained to an infinitely rigid support which is not real world but is commonly done. So one would invoke St. Venant's principle near the fixed ends and take the stresses at the constrained ends with a grain of salt.

 

TOP
CSWP, BSSE
www.engtran.com  www.niswug.org
www.linkedin.com/in/engineeringtransport

"Node news is good news."

Just to quickly elaborate on St. Venant's principle. It's simple really, the idea is that if you take your boundary conditions off of your part of interest - that right there will already be a much more realistic situation.

The further away the boundary conditions (which are always theoretical concepts) are from your model, the more natural you allow your part of interest to flex and bend a little like it does in reality.

http://www.clear.rice.edu/mech403/HelpFiles/WhatIsFixedSupport.pdf

Certified SolidWorks Professional

FAQ1183-1719: What is St. Venant's Principle?

TOP
CSWP, BSSE
www.engtran.com  www.niswug.org
www.linkedin.com/in/engineeringtransport

"Node news is good news."