A Failure Investigation: Caesar's Predictions vs. Reality

I am in the midst of a failure investigation and Caesar is making predictions that do not make sense to me. I'm thinking that this is a mathematical problem, i.e. a function of the way C2 calculates stresses. I wanted to see what you guys thought if you have a minute.

During a recent rainstorm in this area, runoff water pooled beneath the footings of several overhead pipeway column supports. The water pooled under the supports because the drain culvert that serviced the area was clogged with debris. The soils in this area are collapsible when wet, so when the soil got saturated from the runoff, it collapsed downward about a foot, taking the footings with it. This caused large movements (displacements) resulting in severe bending of the piping attached to the pipeway.

My client asked me to examine the situation and make any recommendations. My analysis to this point has been to model the failure by modeling the piping attached to the supports and then forcing the large displacements at the support nodes using a cnode. The support (restraint) uses the cnode as the displacement node.

Well, I built the model, and the stresses that C2 predicts in the displacement load case are far in excess of the ultimate strength of the pipe. C2 is predicting the max stress as 200+ ksi. This is A106B pipe which has min. specified ultimate strength of 60 ksi.

To me, this says that the pipe should have fractured, but the pipe has not in fact physically fractured, it is just 'all bent up'. I am sure that it has gone plastic althought I have not yet done the calculations to verify this. The pipe also has not locally buckled or 'kinked' from the bending. I could back out the principal stresses and calculate the von Mises' stress to check that criterion too but it still will be far in excess of 60 ksi.

I know that C2 will only do what it is told. I am confident that my model is as close a mathematical representation of the actual physical situation as I can make it. I don't think there's anything else I can do to make it more accurate. My point here is that I'm trying to get confidence in C2's results. I guess one conclusion to be drawn here is that C2 is best left to the design situation in which the material of the piping system is still in the elastic range, assuming 'failure' is defined as the proportional limit of the material, the allowable displacement stress range of B31.3 notwithstanding. Is this strictly a mathematical manifestation? Or am I forgetting something from Mechanics of Materials 403?

Thanks very much in advance for any help/ideas/suggestions/random thoughts.


------------------
Thanks,
Pete
-----------------------
Thought for the day:
Good judgment comes from experience;
Experience comes from bad judgment.

CAESAR II (as well as all other pipe stress programs) assumes that the pipe is linearly elastic, which means that will not calculate an accurate measure of the real stress state in a case where gross plasticity exists. If you recall the stress-strain curve with the straight elastic portion, then plasticity (with continuously changing stiffness) and failure, in CAESAR's methodology, the linear portion is extrapolated indefinitely, no matter whether the material is actually yielding or not. Therefore, CAESAR's linear approximation of stress can be much higher than the true stress state (as it is in your case).
In addition to this inapplicability of the program, with gross distortion of the cross-section, there are changes in flexibility of the piping itself, which will typically add to the over-estimate of the actual stress level present. CAESAR does not adjust stiffness to accomodate any ovalization of piping. This effect is more pronounced for larger diameter, thinner walled pipe.
To acheive an accurate numerical prediction would probably require non-linear FEA.
You may be wise to consult with an expert in materials and failure analysis (accurate assessment may best be done by other means).
Or maybe you can just replace the pipe & fix the support problem?

I agree whole heartedly with the previous post, also I suggest that you take time out to do a literature search on this subject matter.

I commend you for realizing that the beam element model is giving you "funny" results. I wish more people looked at output reports and though about what they were saying.

However, by declaration of your state of puzzlement you also indicate being outside the depth of your knowledge, I am certain however that you can fill in you voids.

Check API RP 579, its pricey at $600, but it looks like it may adress some of these issues (No I have not yet been willing to cough up $600, but I know I will sooner or later, I hope to attend a seminar on this this fall)

Other points to ponder ASME B31.3 and B31.1 are "design codes" for new design and their coverage does not directly adress this type of evaluation. Although B31.3 has some discussion of settlement in it.

And finally, the B31.3 and B31.1 codes will sometimes allow the calculated secondary displacement stresses to go beyond the yield of the material, however they and all "pipe stress analysis, beam element programs" provide only one-dimensional beam elements that act elastically even beyond the yield point. This "simplified" method when combined with the codes has demonstrated time and again to provide a safe design for most (but not all) piping systems.

Hope this helps, and one more thing despite the twisted and magled condition of the pipe, if it does not have to undergo a significant amount of cycles or very rigorous duty it might be OK as is, hence the phrase "Fitness for Service" evaluation.

Best Regards,

------------------
Best Regards,

John C. Luf

This post reminds me of one of my very first piping flexibility analysis jobs where the calculated stresses were on the order of 1E6 psi on a system I was troubleshooting. I knew this stress level was impossible so I dug into the background of the program (a mainframe program preceeding CAESAR) to understand how the calculations were made.

This action was a revelation to me. I learned 2 things: 1) piping programs like CAESAR II assume infinite elasticity (as was mentioned in a previous post), and 2) always look at the results to understand whether the model is behaving right (i.e. have any physical laws been violated, do calculated displacements match actual movements in the field, are restraint loads inactive [liftoff] or exceedingly high and do they match reality, can I duplicate program results with some simple manual calculation methods within acceptable accuracy, etc.).

As I work with young engineers I have found a tendency for them to unquestioningly accept computer calculated results (...it must be right, that's what the computer calculated...) without trying to understand whether their model matches reality. One really needs to watch out for the GIGO trap. (GIGO = Garbage In, Garbage Out)

By the way, members of API get a 50% discount on any API publications. This brings API-579 to about $300, close to the price of the volumes of the ASME piping codes.

Thanks for your responses and comments guys. This is a great forum for this type of discussion, no?

As a postscript to this, the underground part of the piping was excavated. When the soil was removed, the particular line in question did not completely 'spring back' to its installed position, which confirmed my suspicion that the pipe had gone plastic and suffered permanent deformation. So my Caesar analysis was not toally useless in that it did permit me to predict that the line had gone plastic. We decided to abandon the line on place and re-pipe that part of the process.

I learned long ago that Caesar is nothing more than a very fast calculator and is only as good as the data given it. I plan to use thgis case as a training point when training others in the use of the software. As KS Taylor said, I find my biggest obstacle in training others to do pipe stress work is to look past the program results and to remember the fundamentals of solid mechanics. The results of any pipe stress program must be interpreted through the lens of one's education in the fundamentals, through experience, and through actual field results.

Thanks again for your comments.


------------------
Thanks,
Pete
-----------------------
Thought for the day:
Good judgment comes from experience;
Experience comes from bad judgment.

This thread includes many pearls of wisdom:

• Piping program use simple 3D beam elements. You can't evaluate system behavior outside the capabilities of this element.
• Piping programs assume linear elastic material behavior. If your system has gone plastic, the CAESAR II results will be off.
• Piping programs typically don't account for second order effects (P-delta).
• Last but not least, the computer should be used as a tool, not a Guru.

Very, very true!

An excellent discussion. Thanks for contributing to the fourm.

PS: there is a little more to it than that, but you're essentially correct.

------------------
Regards,
Richard Ay (COADE, Inc.)



[This message has been edited by rich_ay (edited June 22, 2000).]