One of the biggest predictable problems I see with convergence is when I have specified friction at a support that lifts off. It happens any time we have supports on either side of a riser that turns to the horizontal on the ends, or have an in-line support on the riser.

In the hot condition, the vertical section of pipe grows, lifting off the support at the top. In the cold condition, the vertical section of the pipe shrinks, lifting off the support at the bottom.

The only reliable way I've found to get through the convergence is to turn friction off at the support locations altogether. As such, this means that the support loads are under-reported at these locations, and the best I can really say is to multiply friction loads from adjacent supports by the ratio of vertical loads.

Current ways to solve:
• Play with the friction stiffness for the entire model for all load cases.
• Play with the friction value for the supports for all load cases. (Which is what I do when I turn it off)
• Play with all friction value for all supports for a given load case.

Two ways to correct this could include:
• An auto-removal of the friction value on a per load case basis when the vertical load is 0.
• User to specify friction as an array that coincides with T.

Thanks for reading.