> This is really a can of worms, but I've always seen more failures in
> connections than with spans. Assuming the joint to be fully rigid makes
> the connection loads conservative even though the span response may be a
> bit less so. Over here in machinery-ville, that's as it should be.
Why more failures in connection? Because your connection design assumes a
perfect rigidity. And, consider how much stiffness the real connector can
provide. The real stiffness is always less than the assumed infinity. When
the difference is significant, the connector may have an unallowable
deformation, and then is torn away. That may cause a connection failure.
Certainly, unallowable deformation may re-distribute the internal forces.
Than may create another problems.
> Besides, who really knows how rigid an as-built connection actually is?
Your connection design should answer this question. Connector has a
dimension, that has a stiffness. If your connector stiffness is infinity,
that means a perfect rigid connection. In the real situation, consider how
much stiffness your connection can provide.
Partially rigid analysis may output "connector load" and "REQUIRED connector
stiffness" for connection design. The design procedure starts at the
connector load (similar to the traditional method), and finds a connector
dimension that provides a sufficient strength for connector loads.
By the selected connector dimension, we can claculate the connection
stiffness, and then compare the "calaulated connector stiffness" and
"REQUIRED connector stiffness". If the REQUIRED connector stiffness cannot
be satisfied, the design procedure needs to re-select a bigger or other
connector dimension, or restarts analysis to reduce the REQUIRED connector
Traditional design only considers the connector load (but assumes the
connector stiffess is infinity). Partially rigid design considers the
connector loads and stiffness, and may provide a safer structure.