Has anyone experience of using CSI 2120-2 for model testing?

I am thinking about applying this technique with instrumented hammer to find the transfer function on a piping at different points and frequencies when there is no flow. Later after measuring the vibration on the same piping with fluid flow, can I measure the dynamic forces by using the known transfer function?

If this approach is not correct, what should be done?

Hello Arshad, yes in a number of cases for pipes and valves in power stations to qualify for operation changes and modifications. From very thin high pressure piping to 600 mm steam pipes. You have to apply a lot of understanding for the situation when you make the test and the situation later in operation. Not only the liquid is there, but also pressure and temperature. A first approximation is to consider the weight of the liquid, a second important is to judge the impact of pipe restraints and what eventual insulation will bring of weight and damping.

The tool is not critical as such, but a CSI2120-2 can do the work very well. You should start with small systems that you know reasonably well and build as your experience grows and include several "calibration" activites in close cooperation with the designer.

I have difficulties to believe that you can make a serious estimation of the real system in operation making only dry modal tests. You can detect major problems but not with an accuracy that for instance an authority would accept for endorsing a system as we do here. A mathematical model that is made for a dry system, then adjusted to be as built using your modal analysis results and then in turn modified with the operating parameters is a good way to go. All these efforts should be weighed against the real benefit it would bring.

I know people working with this all day with whom you could eventually discuss this in more detail. Just mail me with a bit of information of your situation and I will transfer it to a competent person. Arne.Lindholm@bluebottle.com

Arshad,
The excitation and response of piping systems can be EXTREMELY complex in general.
If your response is dominated by a frequency traceable to a machine, ie 1X, vane-pass etc, and acting at one point, you might be able to piece together what is going on. Particularly if the excitation is mechanical coupling or transmission rather than flow-induced.
But if the vibration is flow-induced, ie pulsation, valve noise, vortex at a cavity etc, the situation can be very complex. Consider this: The system may be vibrating at one or more natural modes. So the exciting forces are by definiion small relative to stiffness and inertia forces. There may be more than one possible points of excitation. Therefore, a variety of excitations (ie what magnitude at what frequency at what point) can produce responses that appear almost the same. So, extracting "the excitation force" is a tough task indeed.
I would recommend you stay out of this complex undertaking, and diagnose and solve your problem in a more direct way.
Look at the frequency content and the variation with process conditions to identify the likely source. Look at operating mode pattern, to see what the general shape of the vibration is. Do a modal test if you like, to relate the operating vibration to a particular system mode or modes. Decide what is the better approach - reducing the excitation or reducing the response.

This message has been edited. Last edited by: <Ron Hartlen>,

Arshad, I think in general it is a good idea. Although one of the problems I can envision is that transfer function of an empty pipe is going to be different from that of a full pipe.

Secondly, if one is interested in getting force values over a frequency range of interest acting at a certain physical location of a pipe, then when acquiring this transfer function the pipe has to be excited exactly at the same location.

This message has been edited. Last edited by: David_G,

What about converting the acceleration measurement into force? If by F = m a, acceleration gives the dynamic force, dynamic stresses in the vibrating structures or pipework can be calculated, which will help judge the severity of vibration on the basis of endurance limit of material. If this is true, the question is, what mass value to be used, especially in the case of a complex pipework, which has several supports and guides? Any idea?

Actually if you are interested in dynamic stress the best measurement is to use strain guages and measure that fairly directly. The next best alternative is to measure vibration displacement. Companies like Southwest Research Institute and EDI have a long history of relating displacement to stress in pipe works, and have published some of those correlations.

quote:

Originally posted by Arshad:
What about converting the acceleration measurement into force? If by F = m a, acceleration gives the dynamic force, dynamic stresses in the vibrating structures or pipework can be calculated...

Even knowing the mass m would not give the value of dynamic excitation force acting at a certain physical location of a pipework since m*a is just the inertia force. Full equation of force balance is:

F = m*x" + c*x' + k*x
x - displacement
x' - velocity
x" - acceleration
m, c, k - mass, dampening, stiffness
F - excitation force

As it has been mentioned above, there may be other methods of measuring material stress, but I find method suggested by Arshad for measuring the exitation force by using the transfer function and accelerometer to be very practical and original.

David G,
By writing out the four terms of the differential equation, you have unintentionally provided support for the "downside" of this, as per my earlier post.
Piping systems usually don't get into vibration trouble unless there's a resonance involved. Even for a single-degree-of freedom system it's very tricky. At or near resonance, the inertia and stiffness terms are very nearly in balance. So the battle to determine limiting amplitude is between the input force and the damping, both of which are much smaller than the other two terms. So by definition you can't get at the force unless you know the damping behaviour definitively. Damping is notoriously tricky.
Now let's go to a real piping system. It's a continuous system, so there may be more than one mode involved. Piping systems can have modes closely-spaced in frequency, but with different mode shapes. Damping is not one term, but rather is spatially distributed, and can involve friction, impacting, given non-ideal supports etc. And as I said above, the forcing may be acting at more than one point.
Have you ever done this? If so, can you provide some detail on how it's done, including how to resolve the points I've raised here?

Arshad,
Until you get some really specific direction on exactly how to do this, don't even consider it.

Ron,

No, I have not done this. But it does not matter now as we are discussing here a possibility of implementation of a new METHOD.
(Actually, it may not be new).

Let me explain how I see it.
First, our goal is to find experimentally force F acting in a particular location of a mechanical system. (Later on it could be used to calculate material strain if necessary). Is it a pipe or any other object, is the object at resonance or not - all this does not matter as far as method is concerned.

Secondly, we'll do it in a way that there is no need to know m, c, and k values. It may even be a more then single degree of freedom object.

But with CSI 2120-2 we can find the transfer function T(f)in frequency domain:

T(f) = F(f)/a(f)
f - frequency
F - excitation force
a - acceleration
(all terms are vectorial)

by exiting the object at the same (or nearly the same) location where acceleration is measured.

Now we can go backwards. By knowing:
1. T(f) as a parameter of the system,
2. measured acceleration a(f)
we can find the force causing this acceleration by doing vectorial multiplication.

F(f)=T(f)*a(f)

Is 2120-2 capable of storing T(f) and then performing multiplication operation? I do not know, but it is certainly feasable.

This message has been edited. Last edited by: David_G,

David - it sounds to me like you are saying you can estimate everything you need from a transfer function without studying the pipe characteristics and deflection shapes.

Some of my questions would be
- #1 How do you make any judgements comparing motion during the bump test and motion during operation if you don't study the shape? (You found a transfer function from point A to point B but how do you know the force causing the movement during operation isn't coming from point C?)
- #2 Ignoring #1 and assuming you have correctly computed a driving force, how you you convert force to stress without knowing the shapes?

I don't know much about modal and operating deflection shape analysis. I think they use bump tests FRF's to help "calibrate" their math models. I trust the experience of Arnie and Ron in this area when they say it is not a simple matter.

Arshad

Here are some references:

Previous thread on piping vibration:
http://maintenanceforums.com/eve/forums/a/tpc/f/3751089011/m/6401026141

Free Articles on piping vibration analysis:
http://www.engdyn.com/papers/papers_piping.htm

Scroll down to modal analysis and operating deflection shape for free papers here:
http://www.vibetech.com/TechPapers.htm

This seems like a non-trivial problem. On the surface, you can do something similar to what David suggested; except this would be a multi-degree of freedom model. This is an inverse force type problem; i.e. you have some displacement (or velocity or acceleration) data and you need a set of forces that would produce these measurements up to some error.

If you proceed with modal testing, you will build a discrete model with simple force inputs most probably. The actual forces in the system will include pressures acting over areas.

David’s approach in the frequency domain looks like K(w) *d(w)=f(w), a multi-degree of freedom, MDOF, system as a function of frequency. It may be more helpful to use a discrete time based model, but the limitations would still be there.

In the second order model (differential equation) listed the '' are usually represented by dots for the time derivative – It is understood that this isn’t possible to type. If you look at the rate of change with respect to physical dimensions of displacement (e.g. strain gauges) together with a model for the pipes you should get stress. This rate of change is given as '' or more precisely as partial derivatives.

Arshad,

Why do you want to know the input force for an existing pipe system? Generally pipe dynamics are evaluated for structural failure and transmitted vibrations and sound to sensitive locations. Pipe vibrations and strain/stress measurements are used for structural failure evaluation. The nuclear power industry is a good example of where considerable time and effort is spent for predicting (calculating) piping response to machinery and seismic input forces.

I have a 2120-2 and impact hammer, and I have conducted structural tests on pipes. I have measured dynamic strain, operating deflection shapes, and transient events. I have added weights (mass) and supports/flexible joints (stiffness) to change dynamic properties and reduce vibration levels. Sometimes its easy and occasionally it is very frustrating for the reasons that Bill and others have stated.

The idea behind model testing on the gas pipeline is to find an accurate, fast and more convenient way of determining vibrating stresses under the influence of acoustic pulsation and machinery dynamic forces.

Model testing provides me natural frequencies, mode shapes and forcing function at different frequencies. My hope is that if I know the ODS of a vibrating pipeline, I will know exactly where to take measurements while making a model test. Secondly I know that piping is a passive system, which vibrates only when dynamic shaking forces are exciting its natural frequencies. Thirdly in a gas pipelines where the bulk modulus of the medium is very less, the effect of pipe contents can be assumed negligible. So if I apply some impact forces on a non operating pipeline close to elbows or locations of differential areas, where the shacking forces actually act, and measure the vibrations in x, y and z directions at the points of maximum deflections known from the ODS, I will be able to find the transfer functions at those ODS. Converting transfer function at actual vibration and ODS will give me the vibratory force. I can use the axial component of that force to find the stresses in axial direction by using the pipe x-section area, but not sure what area to be used to convert the radial component of the forces into vibratory hoop stresses, I need help on this.

Comparing the measured stresses with endurance limit of piping material will tell if the vibration is acceptable or not.

Beside, I need your option to built some trust over this method, I am going to make several experiments and comparing their results with conventional techniques of computerized dynamic simulations as well as strain gage testing.

Thanks everyone for your valuable comments.

quote:

I can use the axial component of that force to find the stresses in axial direction by using the pipe x-section area, but not sure what area to be used to convert the radial component of the forces into vibratory hoop stresses, I need help on this.

Depending on the mode shape, I would think radial vib and force would not likely be related to hoop stress but would instead be more likely related to axial stresses associated with bending. These bending stresses are maximum on the o.d. of the pipe and are proportional to 2nd derivative of deflection with respect to axial distance (beam approximation)

This message has been edited. Last edited by: electricpete,

Sure seems like a lot of bother when you can do the same thing easily with strain guages...

Measuring strain seems to make sense given that max stress/strain for bending occurs on the outside. More direct measurement means much fewer assumptions/approximations and much less chance for error. Arshad did mention that he planned to use this modal analysis as a kind of independent check of other analyses including strain gage.

I've never done either one - modal analysis or strain gage measurement.

Just out of curiosity - can strain gages hook up to portable vib data collectors or you need a different data aquisition system? Do they have a reasonable bandwidth i.e. can capture frequencies up to perhaps 100 hz?

I note that in paper 20 on the Engineering Dynamics link above they talk about direct strain measurements.

In most of the other papers at that Engineering Dynamics link, they talk more about traditional vibration measurements (displacement and velocity). In particular paper #27 they say what I think Ron has been saying for awhile - max stress is roughly proportional to max velocity and the proportionality constant varies over a fairly small range for a wide range of configurations. That certainly might be useful as an easy ballpark sanity check on the other results.

Pete,
Yes straingauges can measure very high freq. many KHz (if you get them stick on good, gluing is very important) temperature is a limitation, have used it for pile driving steel piles and yes you can use a normal analyser if there is a bridge amplifier first, you actually measure the resistance change in the gauges when pulling them so the "wire" get thinner so some amplification is reqd. you need to turn off the xducer power. Olov

Arshad,
With a better idea of what you are trying to accomplish, I think you would benefit from reading my article at http://www.sstusa.com under "Technical Articles".
It addresses what you are interested in, gives insights into how to assess a system and plan measurements etc. There is no more direct way to get to an understanding of susceptibility to high dynamic stresses, relating vibration to stress, and identifying important features of layout and support.
If you wish, contact me by E-mail.
ron.hartlen@sympatico.ca