Dear Srs.,

I'm analyzing a flare system with the following conditions:

Operation: T=37ºC and P=0,5kg/cm2
Design: T=175ºC and P=3,5kg/cm2

There is a rule here that for flare system we have to perform a modal analysis and ensure that all frequencies are above 2Hz. This is because of a possible two phase flow or other vibration problem.

Since modal analysis (eigenvalue) is a linear calculation, sometimes it is very difficult to get all frequencies above 2Hz. For example, in the ".c 2" attached when the frequencies are extrated without friction effect the first mode is equal to 0,4Hz. In the other hand, if it is considered the friction effect through a "stiffiness factor for friction" equal to 2 (CAESAR II's manual suggest a value beetwen 50 to 100!!) the first frequency increases to 2,5Hz and meets the requeriment (above 2Hz). It's important to note that it's a 50inch pipe diameter and the friction forces are relevant!

Not considering the friction effect will result in criating several other supports (guides and stops) just to meet the frequency requerement.

Appreciate any sugestion or advice in this issue.

Regards,
Roberto Liberato

Rlibrato,

Check these following posts out:

http://65.57.255.42/ubbthreads/ubbthread...=true#Post36458

http://65.57.255.42/ubbthreads/ubbthread...=true#Post38431

Roberto,

Out of main topic, guides and stops are not such a bad idea. Since you can not cover all possible scenarios, is good to have them in place. Take a look on other topic "disaster" or so.

Regards,

Thank you all.

I will place as many support as possible but considering friction in modal analysis. To linearized the system I will account friction with a stiffiness factor of 1 at least.

Regards,
Roberto

Roberto,

I wouldn't use 1 for the stiffness factor for friction. I suggest you try something more like 500 or 1000.

Dear Mr. Dave,

Actually my client requires that I need to remove all friction effect in the input during modal analysis. What's funny is that nobody knows why and from where this rule came, but everybody needs to follow.

I understand that eingenvalue is a linear calculation and for this reason can't handle with nonlinear boundary conditions like friction. But in my point o view, removing friction effect is not quite realistic. Not considering the friction effect will result in criating several other supports (guides and stops) just to meet the frequency requerement.

So I think that adding in each support a stiffness equal to normal load in operation multiplied by friction coeficient will bring a result near to reality (unitary factor). Why do you suggest a factor of 500 to 1000? With a high factor the pipe horizontal movement in the support location (+Y only, for example) will be restricted, is that correct?

Thank you very much for your help.

Regards,
Roberto

Nobody can you give a quarantee , that a vibrition pipe with vertikal acceleration for ever time ( dyn. Event)friction acts.
With friction you system will be stiffer and the eigenvalues higher.
This is maybe not conservativ, if your excitation frequency is low.

But how is that value of 1 applied? CAESAR II takes the normal load from your specified load case, and multiplies it by the coefficient of friction AND your multiplier. The resulting product is the stiffness assigned to two orthogonal springs that are orthogonal to the normal load. This is a stiffness, so to reach this (friction) load, the pipe would have to move one inch (assuming US units). Now, without an applied dynamic load, the calculated natural frequencies (and mode shapes) will not have a set displacement for these friction restraints but I think you would agree that allowing up to one inch of deflection before achieving the friction load lets that remain a rather free boundary condition.
Run the model with a few different values here and see what makes better sense.

Dear Mr. Diehl,

Thank you very much for your time.

What I meant was 1 for the stiffiness factor for friction in the Caesar II dynamic input. I agree with you but I think that considering a high factor will have the same effect as anchoring the nodes where were placed the supports. In other words, the program will report the modal shapes not for the entire systeam but for every element (between two nodes) indenpendetly.

The big issue here is how to take into account friction damping effect in the natural frequency analysis of the entire system (without any applied load). How relevant is that damping?!

Support friction is a kind of Coulomb damping which has a value proportional to the contact force (normal load) and acts against the moving direction. I know that damping add considerable complexity in the solution of the motion but may be approximately lumped into the viscous damping for consideration. So, if it's hard to implement damping mathematically, probably there is any simplified idea (theory, book, or any reference) to approximate friction damping to evaluate natural frequency motion.

Back to my model, remember that I don't know (nobody knows) the caracteristic of the dynamic excitation so my client established the rule that all frequencies must be above 2HZ (they avoid low frequencies - high amplitude). And as I said earlier not considering the friction effect will result in criating several unnecessary other supports (guides and stops) just to meet the frequency requerement. Since it's a big heavy pipe I think friction is not something irrelevant.

Again, thank you very much for all the support.

Any help will be very much appreciated.

Regards,
Roberto

Roberto,

1000 is not equivalent to an anchor. Give it a try and see what you get.
If you cannot verify these results, it sounds to me like guides are required.

Dave,

You are right, not like an achor actually like a guide and stop in every support. With 1000 there is no translational horizontal movement in the nodes only rotational (like a kneecap - i dont know if that's the word).

For example, in the CAESAR II attached with 0 the first mode is 1,17Hz (loop swinging) whereas with 1000 the fist frequency is 15Hz!!

I wish i could use 1000 for this factor since I would attempt the requeriment. The problem is to convince my client that this is the right way. If I had any reference (book, paper, etc) to support that ideia!!

Sorry, no, I cannot "prove" the first mode is 15Hz but then I cannot prove it is 1.17Hz either.

Ok, I see.

Summarizing: the only thing we can say for sure is: without friction (pipeline in a hockey air table) the natural modes frequencies will be XXXX. For the "c 2" attached probably the first mode is between 1.17Hz (min) and 15Hz (max) in different locations of course.

Question without answer:

1) Is friction damping relevant in the modal analysis?
2) What are the frequencies with the friction effect (real world)?

Anyway, I would like to thank you for your help and opinion on this issue.

Regards,
Roberto

Roberto,

You are going down the wrong road. You are incorrectly looking for a PAPER solution to a POORLY supported line for a dynamic event. Treat the line as if the line is in a high Seismic zone.

Rule 1. Add LINE STOP at the CL loop node 520. This is usually a neutral point and is very commonly done. The only problem is getting steel to this point.

Obviously if this criteria was known from the start of the job, an upper tier stress engr (top 10% of stress guys who actually know what they are doing) would have coordinated with C/S early to make this happen.

Rule 2. Dont back off guides near std loops. Keep guides close.

Rule 3. Try to add line stops in each long run of pipe.


Line stops can easily be added at node 210 and 270 without much thermal effect.

If you want higher frequencies for a large system then horizontal supports must be added. Look for inflection points, etc... You are in control of how the line expands. Place supports accordingly.

Bob,

Thank you very much for all your suggestion. It helps me a lot.

I agree that the best solution for this case is to restrict the line avoiding high forces and moments during operation (neutral points). The problem is that not always you have stell structure to support in the point you need.

This system is one of severals that I need to meet the frequency requerement (mainly lines of flare and those connect to reciprocal machines). Imagine how many support I will save if instead of using 0 I uses 1 (for example) for the stiffiness friction for factor. The frequencies are very sensitive to this factor.

Anyway, considering the lack of information on this friction effect modal analysis issue I think I will follow your instructions.

Regards,
Roberto