Knee brace issues, unreasonable deflections & drift, **solved**, fixity problem / bug

[thejonster]

I'm designing an independent steel mezzanine level (apx 30x50’ x 15 ft tall) inside of another building. I am having trouble getting reasonable results for loading in the knee brace direction (load acting in direction of strong axis beams and braces pinned m22 minor & m33 major, weak axis columns pinned m22 minor). I am getting unreasonable results for EQ, dead, live. I tried fixing the ends of the beams and keeping the braces but then the columns no longer work (U=1.8 vs U<.50 in my risa run)

The deflection of the beam shown is gigantic, not reasonable. And this isn't how the moment diagram should look, this would only be possible with the joists taking moment out of the beam with torsion but there is no torsion in those joists and any in the view is negligible. There is no moment in the pinned braces, the axial loading are about what I would expect. I have double checked the releases.

I thought that using knee braces would be a good idea to avoid some interesting connection design, but this has me scratching my head. I modeled the same frame, same sizes orientations and releases, in RISA and am getting reasonable results.

I am new to ETABS and would appreciate any insight here. Is there a brace setting I haven't assigned?

I have included a couple of screenshots showing what I'm talking about and have attached my ETABS model.

 

[thejonster]

I realized that the top of column does not need to be released m22 if the beam and brace are pinned. I fixed the ends of the columns and the problem is gone. I'm really not sure why this tripped up the program, seems like csi should add this to their list of bugs or am I looking at this wrong?

 

[Agent666]

I think you're the one who has inadvertently created an instability here, it is not a bug, but more a limitation of the analysis method and numerical issues that might occur when solving the stiffness equations to get forces/deflections when the boundary conditions are invalid. Its the same in any analysis program. Pin everything at a joint and it all goes to the crapper.

The other classic example that I come across often, is with columns with pinned base and only translational restraints (no rotational restraints), and all beams framing into the column with M22 minor axis pins. Then in this configuration the column is free to spin with no restraint. Creates "ill-conditioned" result which basically means the result is unstable and has numerical issues because convergence cannot be obtained.

https://engineervsheep.com