I am trying to model a square building (basically a large enclosed gazebo) with hip roof (vaulted ceiling) and a cupola.  

The top of the roof structure (at the cupola) has a compression ring (multi-ply LVL) and the bottom of the roof (at the eave) I have another large multi-ply LVL continous around the perimeter which is my tension ring (ends are pinned.)

The roof rafters are sheathed on the top and bottom.

There are perforated shear walls to resist lateral forces.

Currently I have the rafters and LVLs modeled, not the roof diaphragm.

I plan to use a steel plate (L shape w/vert leg) at the corners to tie the LVLs with a diagonal angle to control thrust.

My boundary conditions are at what would be stud locations in the shear walls and have rollers parallel to the rafters on that side.  Reactions in the vertical and transverse directions.

However, the P-delta analysis of my model will not converge.(I have another model with the shear walls also.)

If I delete the diagonal angle, then the LVLs fail in shear at the corners.  I have not incorporated how the plate will affect that though.

Sketch is attached.  Any ideas?


    

If your boundary supports are vertical only, then there is an instability for the entire model to slide around, and that could manifest itself as a P-Delta convergence issue. Adding some horizontal resistance periodically can solve that. Or you can model some portion of the supporting structure below (i.e. the studs or walls) and fix the base of those members.

If any of the members are released in Torsion, that also tends to be problematic in framed hip-type structures.