I am having trouble getting a CAESAR II model with a fairly large number of pipe supports to converge. The technical reference manual suggests altering the frictional stiffness if friction is thought to be a problem, however I am uncomfortable doing this as I have discovered in the past that doing this can cause a system that was previously code compliant to no longer be code compliant, and vice versa.

Any other suggestions?

Thanks

There are several "friction tolerances" that can be adjusted. I suggest that any adjustments ultimately be made to the configuration file so that they can be a documented part of the analysis, rather than an on the fly adjustment made during the solution.

Please see this post in this Discussion Forum for details on solving a friction convergence issue.

Convergence problem is due to system non linearity and need not be due to friction alone.In your question, you have highlighted about the convergence problem , but is it solely due to friction ? When the iterative processs goes on to solve {P} = {K}{X}, you can see by using F2 why there is non convergence. You may get message like "old state open, new state closed ", this obviously indicates that one directional support like +Y or support with gap can be the cause. If only you get messages like "old state sliding, new state non sliding ", the problem is with friction.However for different node points you may get different message.

Now coming to your question on Friction stiffness, we have to admit that using numerical techniques, we have to resort to approximation.When a pipe starts sliding ( a change from static friction to kinetic friction ) the laws of static equilibrium do not hold good , so how are we solving a staic analysis problem in that situation? The answer is engineering approximation. I fully agree that a change in friction tolerance can make a system " pass or fail".But we have realise that Friction itself is indeterministic, say at the same location you need not have the same friction coefficient all along the thermal history, it will vary depending on surface condition ( but not on surface area) and the kinetic friction coefficient is function of velocity also.So using 0.3 does not mean it is 0.3 all along or along all the points.

So the entire process is not which will simulate the exact friction history that the system undergoes. CAESAR or any pipe stress software attempts to address this stocastic problem using the method originally suggested in Sobieczanski's paper.Besides CAESAR II allows the user to try with not just one but several options of friction stiffness.The objective of friction analysis should be to get a feel of the frictional effects on the sytem, but it need not be taken as a 100 % representation of the actual friction history.

What I will recommend you to do is run a file with friction and one without friction and compare the results. You may find that your system was having advantage due to friction !


Anindya Bhattacharya

Stress Analyst

Bechtel Corporation

Anindya has answered it technically very well. Some practical solution to solve iteration problems are as below,
-Make the +Y supports as Y which are creating iteration problem (You can see it by F2)
-Change Rod hangers to +Y support
-Neglect friction at some supports, specially which are far from sensitive equipments.

Do it stepwise until it runs succesfully, afterward check following things against our adjustments,
-Check that there are no upward forces at supports which we changed +Y to Y.
-Displacements at supports which we have changed rod hangers to +Y should not exceeds 3°angles, as normally specified for rod hangers.

Most of the times this solves iteration problems.
Hope it helps.

dnyanesh,
Linde