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.

Yeah, friction is a real pain. I generally run the model and look at which nodes are not converging and then remove friction from those nodes specifically. That generally lets me get a conversion eventually. Once I have updated the model with any needed changes I go back and put the friction back in and try and get it to work. Sometimes I get lucky, and other times I don't.

One thing I have started doing is running both a standard operating case without friction and then a standard operating case with fricition using the controls in the load case editor to change that between run.

Thanks,
Duncan

I've taken to setting the friction multiplier to 0 so that I can get a Q&D™ answer for which supports are lifting off, and start taking friction out for these specific locations until I get convergence, adding friction back in, one at a time, until I get as close to a final answer.

However, generally these are at risers, and because of lift-off, they inevitably have a higher vertical load than the others.

But with no friction, this results in no lateral loads, even though this is fictitious, I've had structural types take the no lateral loads case and run with it. I've also had other structural types average out the lateral loads and run with that. Both are inaccurate, so long as the vertical load at those particular supports are much higher than other "average" supports.