Friction is one of my pet peeves. ASME B31.3 says the following: "The significance of all parts of the line and of all restraints introduced for the purpose of reducing moments and forces on equipment or small branch lines, and also the restrinat introduced by support friction, shall be recognized." So if you have a support where sliding friction is an issue, the Code requires you to recognize the effects of that friction as it pertains to the forces and moments on equipment connections and small branch lines.
THIS DOES NOT SAY that you MUST include support friction effects on every sliding support! But it's commonly interpreted that way. Thus my beef.
Friction is not a linear phenomenon; therefore analysis of friction effects is stochastic (see, for example, the explanation of friction that you can retrieve from the COADE CAESAR II technical articles). A typical example of a stochastic process is the spontaneous decay of radioactive nuclei. See, for example, "Schrodinger's Cat." This is a famous problem that vividly illustrates the issues of attempting to model statistical based phenomena and apply them to real-world issues where only yes-no answers are typically accepted by others. The process that we use to model friction is similar; it provides results that may happen, but also may not.
If you're in an area of your piping system that is "near" a terminal point or intermediate anchor, the probability that the process used in CAESAR II will calculate a result that is likely to occur is relatively high, and you can use it with confidence. If you're in an area of your piping system that is not "near" to a terminal point or intermediate anchor, I don't have a lot of confidence that the CAESAR II calculation process tells you anything meaningful. But defining "near" for the purpose of this discussion is not really possible. (See Mr. Luf's comment above.)
When you build a model that includes a lot of supports with friction, you often reach a situation where two or more supports alternate between "sliding" and "not sliding", and the model will not converge. This is usually hard evidence that the CAESAR II friction model is not valid in that part of your piping system. The "old" way of accounting for friction on supports is perhaps a better option for that area.
1. Remove friction from one of the supports (the farthest one from a terminal connection or intermediate anchor, usually) and run the analysis.
2. Calculate the value of Mu * support load for that restraint.
3. Consider whether a lateral load of that magnitude acting in any direction can cause you a problem in your piping system. If the answer is no, make a note in your analysis report that you have removed friction from that support and why you consider this to be valid. If the answer is yes, you need to reconsider which support to remove the friction from.
In general, my experience with nonconvergence of analysis models due to friction is that this effect is most severe where the piping is flexible enough to allow relatively large movements. This is why the COADE recommended procedure of reducing the "friction stiffness" is effective; it makes the linearized model of what is happening at an offending node valid over a larger displacement range. But it's not a good option if you're attempting to model what "really" happens. If you don't believe that, go stand beside a hot line supported on shoes as the system starts up and you'll acquire this particular "religion" real fast. Just take a spare pair of undershorts if you've never done this before. It's especially memorable if what you're watching is a steam line in a tunnel with slip-type expansion joints every few hundred feet, and you have to climb a 20' ladder to get out of the tunnel if things go wrong.
Edited by CraigB (05/23/07 10:31 AM)
_________________________
CraigB