I modeled two identical beams with the same distributed load. Each beam has a short vertical member connected at each end. The only difference is one beam is a single element. The other one is divided into 10 segments. For static linear analysis, I get the same results for each beam as would be expected. However, when I run a nonlinear analysis, the beam with only one segment gets wildly incorrect values. Here's a screen shot of the Mz bending moments. They should be the same but notice how different they are.
I reported this as a bug but the service ticket was closed. I understand that breaking an element up into more parts can give better nonlinear results but I wouldn't have expected this large of a difference. Also, the loads are such that the beams stay in the linear range. You can look at the deflection of the center node of the subdivided beam and see that it behaves linearly.
I've attached the .std file communities.bentley.com/.../Nonlinear-Test.std so others can try this. Any thoughts on what is happening here?