What is the Value of a "Bump Test"?

[yogibear1]

My knowledge of vibration analysis is relatively nil. I have replaced four old vertical centrifugal pumps with new pumps and 1200 r.p.m. motors. The installation is on a reinforced concrete slab over the pump wet well. The manufacturer's rep checked everything out before startup and blessed the installation. Upon startup the two inboard units exhibited quite a bit of vibration, whereas the two outboard pumps were good. On the two inboard units we uncoupled the motors from the pump and ran them uncoupled and they shook badly. Next, we swapped motors between outboard and inboard units and ran them uncoupled; again the inboard unit ran rough whereas the outboard unit was okay. We then had a vibration analysis contractor come and check things out. He ran a "bump test" and pronounced the problem being the natural frequency of the vibrating units was very near 1200 (cycles?)the same as the motor r.p.m. and the mating surface of the motor and discharge head moved in the direction it was weakest - sideways to the flow- as the head has a large opening for access to the coupling and shaft seal. He said a possible fix would be some way to strengthen or stabilize the weaker axis of the discharge head. I can't just go and make modifications to this new unit as I also have to deal with the pump vendor and manufacturer to keep things in warranty. Is the "bump test" to determine the natural frequency a good tool in this case? Any suggestions?

 

[electricpete]

Yes, it is a very good tool for determining resonance.

Another sign of resonance you can look for is "directionality".... machine vibrates much more in one direction than the other.

There can be other more subtle indicators observed during a coast-down test.

I would recommend that you monitor vibration while you install some very temporary bracing (2x4's against adjacent structure?). If it's resonance you will likely see a dramatic change. If machine is running at or slightly below resonance frequency, then the vib will likely decrease as you stiffen it and increase the resonance frequenccy. I suspect it will provide you with intuitive proof that a permanent brace will acomplish the same thing.

 

[EnglishMuffin]

It depends what you mean by "bump test". If you mean he just hit it with something and checked the frequency response, in general that might or might not reveal everything you want to know - it probably would be adequate in your case. Ideally, you should hit the equipment in various directions with an instrumented hammer, which should have a tip hardness chosen to excite the frequencies in the range of interest. Using a hammer incorporating a force transducer compensates for the tendency of the excitation force to fall off as the frequency increases. If that's what he did, you can't do much better. If he just hit it with something and knew what he was doing, that would probably work too, especially for such a low frequency.

 

[electricpete]

By the way the units that go with 1200 are cycles per minute (CPM)

 

[EnglishMuffin]

That's what I figured - 20 Hz is a pretty low frequency, right ? Ideally, you would use a pretty soft rubber tipped hammer for that. Once you get really low, say 6 or 7 Hz, that gets to be a problem for some analyzers. But that's another story.

 

[electricpete]

Hi EM - I should clarify my comment was directed at the original poster who questioned the units "1200 (cycles?)"

Some thoughts on technique.
We do quite a bit of bump testing on rotating machines. Mostly we use 2x4's (wood). No calibrated hammer. Sure there may be a theoretical curve which tails off toward higher frequencies and you may beyond flat part. It doesn't matter to us... we are looking just for frequency, not for magnitude (other than shape of the response spectrum).

Most times the resonant frequencies is easily distinguished in the time waveform. We also capture a spectrum. We like to use a high resolution spectrum. Not because we want to distinguish the frequency precisely better than 10 or 20 cpm... but because it provides a means to clearly differentiate resonant peaks from background machinery peaks. Resonant peaks show up like a mountain with gradual slope while background machinery vibration shows up on high-res spectrum as a super-thin vertical line. With lower frequency resolution the background machinery vib peaks become broader and harder to distinguish from resonant peaks. Sure we can also do comparison test without banging but I like the high-res trick.

Of course we test with the machine shut down. I have seen someone try to do a test with the machine running. Their analyser allows capturing two spectra... one while bumping and one while not bumping... then the two spectra were digitially subtracted to attempt to remove the influence from running machine. Results were not pretty and not accurate. I think the signal (bumping) to noise (machine vib) ratio was too low.

I have heard there are more exotic techniques which evaluate some coherence between the force function of an instrumented hammer and the resulting vibration... mathematically separate out the stuff that is not correlated to the impacts.

Sorry it has turned into a ramble. Someone else's turn now...

 

[EnglishMuffin]

Sometimes you have to be careful about the "tailing off", because there might be a significant peak which gets reduced. I like to know what frequencies I am putting into the structure, and I like to display the true dynamic compliance. A soft rubber hammer, for example, might put hardly anything in above 200 Hz. That's what I meant about the guy knowing what he was doing. Just a personal opinion - but I admit not likely to be an issue in this case.

 

[Tmoose]

CSI gear used to offer something called negative averaging for "bump tests" with running equipment. A look at the time wave form of the bump test often reveals a messy period during the actual semi-soft face dead blow hammer hit, followed by a nice decaying sinewave from which the frequency can be pulled directly 1/(seconds/cycle). If the structure can be felt "ringing" after the hit and the 5 lb deadblow is hanging by my side I am pretty confident I will get some usable data. With vertical pumps and a host of other cantilevered machinery it is a pretty safe bet a basic bump test will be very helpful, if depressing for the poor vendor.

The open (____) shape of the pump volute can be responsible for most of the flexibility in vertical pumps. I tried to push for epoxied on thickeners for a few troublesome pump volute housings, but never got the chance to try it. I had made some simple FEA models that made me think it would work slick.

 

[electricpete]

Interesting aspect of testing some vertical deep draft pump where the discharge is above the water level and motor far above the water level. If the machine is shut down then the water level drops to the suction water level which is far lower than during operation. One would think that would have a dramatic effect on the bump test results.

A question for the group - Would you trust a bump test with machine shut down in this situation?

 

[EnglishMuffin]

Good point - could affect the vibrating mass some - it all depends on the relative sizes of things and how quickly the water drains back after you stop the pump - if in fact it does.

 

[GregLocock]

No I wouldn't. There are all sorts of ways of getting around this, but probably the simplest is to switch to swept sine and a shaker. With enough oomph you can burn through the operating vibration.

Another way would be to use many many hits at non synchronous intervals and do time domain averaging on the resulting waveforms. I haven't done this but it should work.

Cheers

Greg Locock

 

[ANTOINE]

Maybe you ought to take care of the dynamical effect due to the rotating inertia as weel.
I have seen cases where the modes shifted enough to tell you that everything is safe in static condition, and get a vibration issue when the machine is running.

To get enough force, you can as well use a cable attached to the structure and cut it when it is sufficiently tensed. Intrument it with a force transducer (to monitor your static input) and measure response synchronously (watch out for the cable flying away, this may be hazardous), then read time signals as well as transfers and coherences to know where your mode is.
Maybe simpler : if you can increase pump speed gradually, then record acceleration at various location synchronously with pump speed and build the Campbells.

Thomas.

 

[EnglishMuffin]

I realize I'm probably displaying my ignorance, but what on earth is a Campbell ?

 

[electricpete]

I have heard a lot of discussion (including articles) of importance of considering gyroscopic effects upon rotor stiffness for horizontal overhung machines.

I personally have never heard any discussion of gyroscopic effect being important for determining resonance of vertical machines. That doesn't mean it's not important, just makes me wonder.

My general impression from testing of our eight 3500hp 324rpm power plant circ water pumps/motors... we have done many simple bump tests with pump off (no water), with 2x4 (not instrumented hammer) and not considering any gyroscopic effects. Our results I believe have been good based on the fact we the machines we find with highest vibration have resonance closest to running speed.... and we can predict decrease in vibration by adding stiffening. We try temporary bracing prior to adding permanent bracing.

This weekend by adding bracing we reduced a 324rpm motor from 32 mils peak-to-peak to 2 mils peak-to-peak. Simple bump test peak went from 330rpm to 420rpm.

 

[GregLocock]

I do love these interlaced threads - EM, a Campbell's plot is a plot of amplitude vs frequency vs rpm, with amplitude shown as square or circles. These line up diagonally on the forced orders, and across the horizontal showing resonances, and vertically when something goes wrong (grin).




Cheers

Greg Locock

 

[EnglishMuffin]

GregLocock : Thanks - can't really visualize it from your description but will look it up. Learn something every day!

 

[blackeye]

You'll find a Campbell's Diagram on page 2 of the following link:

http://www.mts.com/nvd/Software/pdf/300225-01TransientPost.pdf

Not a very interesting example because there are no obvious resonances in what looks like a strong 2nd order excitation, probably from a 4 cylinder engine knowing the source of the data.