Dear Vibration Professionals
Hi!

How do you think about this spectrum that belongs to a centrifuge pump?

Other data:
speed:2986rpm
vane number: 5
direction of measurement: H
Location: Inboard bearing housing of pump
peak:2903.13Hz
If it is resonance, Is it dngerous?

Regards,F

If I have done my conversions correctly, 2 mm/sec at 2900hz is around 6 g's.

The slight bulge in the noise floor in that region might possibly indicate you have a resonance there as you suspect.

If it were mine to analyse, I would put it on a log scale to see the noise floor better. I would also lable the other largest peaks in orders and see if there is a pattern within which this large peak fits. Is it a harmonic of a lower peak? Are there harmonics at twice the freq of this peak? Are there other similar peaks spaced at 1x away from this one (easy to see when they are labeled in orders)? Are there equally spaced sidebands at other frequencies?

Just wondering maybe if the peak in question is slotpass from the motor? Looks like it is awfully close to 58 orders. Sort of looks like it is sidebanded by 1X.

Assuming that 2903 Hz is explained by the speed and bars/slots of the motor (good observation Ralph), do the lower frequency peaks correspond to any fault frequencies of the bearings (assuming rolling element bearings)?
--Hokeith

quote:

do the lower frequency peaks correspond to any fault frequencies of the bearings (assuming rolling element bearings)?

May be, but looks like they might be related to vanepass with 1x sidebands. Be nice if Farahani would cursor them, huh?

Dear all

The associated motor of this pump have 48 stator slots and I don't know the number of rotor bars (maybe=36). So I think it probably doesn't relate to slot passing frequency.
The fmax in this measurement is 5000Hz and so I can't check harmonics at twice the freq of this peak, unfortunately this pump now is in standby and surely when it come back to service I check it.
As you can see from figure 1, there are peaks spaced at 1x away from 2903Hz peak.
Also you can see fundamental harmonics of the spectrum in figure 2. The cursers denote bearing freq. There is an unusual peak at 87.5 Hz. This pump has 5 impellers and vane pass frequency equals to 248.8Hz (there is a peak at this frequency in the spectrum).

Dear all
I can't relate the 2903 Hz peak to any problem other than resonance, how do think about this?

Regards,F

Figure 1:

Figure 2

Hmmm, well so much for slot pass and brg faults. I was starting to think 2X RBPF but that would require 29 bars. Something tells me that 36 is probably not the number of bars but would be more in the ballpark than 29.

In any case, 2.3 mm/sec RMS @ 2903 Hz is nothing to be concerned about (unless you just want to know what it is). I'll go along with your resonance theory. As for 87.5 Hz, not much concern for the magnitude there either, doesn't line up conveniently with some running or electrical freq multiple - I'd guess another resonance.

These measurements are yet taken on the pump housing (as indicated in your first post) or another location?

Sorry, not much of a help today (as if I am ever).
--Hokeith

quote:

I can't check harmonics at twice the freq of this peak, unfortunately this pump now is in standby and

Can you do an impact test on the pump while it is in the stand by position?

Dear Hokeith
Yes, These measurements are from same point.

Dear Ralph
Excuse me. I haven't done any impact test until now, however I am familiar with it. But I have a question:
Does impact test always require special hammer? Can I do it by common light hammers?

Regards,F

Farahani,

"theory" has it that a special hammer is required or at the least, "best", which it probably is best, but I have, and I am sure others have also, had great success with a simple piece of wood. I would not use a metal hammer though, if that is what you are referring to as, "common light hammers".
I use mostly a 2x4 about 3 or 4 feet long or if what I am bumping is fairly "heavy" or massive, I would use a 4x4 piece about 3 feet long. In reality, in my opinion, it doesn't take much force to excite a true resonance. It doesn't have to be hit like you are trying to "kill" it or dive it though the ground.

what is slotpass?, what is it refer to??

Dear Farahani,

Don't think twice! This is a typical case of resonance looking at the graph you show. In my classes, I emphasize that resonance will appear like a "pine tree". In fact, in a class that I held in Tabriz, we did a demonstration and that produced a pine shape exactly in a resonance set up. So I recommend you do a bump test as suggested. But please do remember: You need a blunt object that will set the complete system in action. A sharp metal hammer will set off local resonances which is not what you want. Example: I had to determine the natural frequency of a stack at a cement mill. The expert before me took a sledgehammer to the stack and obtained naturally very high frequencies and decided that there was no resonance problem. Unconvinced the plant called me. I slammed a 1.5 ton big bag filled with cement and slammed it to the stack with a boom crane. (No I was not close to solving the stack problem once and for all) Result: resonance frequency which matched a source of excitement was apparent in the spectrum. It was then easy to solve the problem. Generally if a machine that did not have vibration problems of this sort starts having problems and shows a resonance peak, it means the system is weakened, maybe due to cracks or looseness. In your case, as there are no harmonics, I would not look for looseness. One last point: I always use logarithmic graphs which show details very clearly. Linear graphs are a waste of time if you ask me.
Hoda hafez,
Ibrahim Caglayan, PhD
Ankara TURKEY

Dear all

The relevant pump put into service after some days rest as standby. This is stupendous! The peak at 2900Hz completely disappear in the latter measurement. All the conditions are same same in both measurements. How do you think about this experience?

Dear Dr. Caglayan

Nice to hear from you. I appreciate you for your sound mentions.

Regards,F

What kind of accelerometer and mounting method are you using? Is it a magnet mount and/or hand held?

Best Regards,
Tom Murphy

Dear Tom

My accelerometer is AS-065 from B&K. The mounting is by magnet.

Regards,F

Dear Farahani,

If you see resonance peak one day and it is gone the day after you take it apart and put it back, it is evidence that the resonance was caused by some weakness in the system that was fixed in the process of putting it back.

Here is how it works: We all know that natural frequency of a system is related to SQRT(K/M) where K is stiffness of the system and M is its mass. Obviously the mass of the system does not change unless one forgets to install the rotor ";-)and that leaves us with the stiffness of the system. Due to loosenesses, a crack in welds and God forbid even a crack in a shaft, the stiffness of a system is weakened. Normally designs are made so that natural frequency of a system is at least 15-25% higher or lower than the operational frequency, of course higher being the preferred alternative. Now if that higher stiffness is diminished as explained above, it may be reduced by such an amount that the result of the expression SQRT(K/M) ends up smack on top of the operational frequency.

I think that is what you had before but as a result of successfull rework process, you re-stiffened the system to its design value and you are home free. Good job and congratulations!

Regards
Ibrahim

Dear Ibrahim

Alright! This pump was repaired because of a problem in the free side thrust bearing. Again your opinion is sound. I will check this pump periodically to find whether the peak come back or not.

Thanks.
Farahani

Dear Farakhani,

Did you take any pictures of this machine and the bearing when you took it out? It would make a very nice case history.

Regards
Ibrahim

PS: In fact in your spectra there are clear hints that the resonance relates to bearing fault. You can see that the resonance peak relates to 5th harmonic of a harmonic of the bearing fault at 143 Hz. I strongly recommend that you use "order" rather than Hz or CPM. Doing so allows bearing related faults stare at the analyst whereas with the other units one has to wrestle.

Regards
Ibrahim

Say it is a resonance; on a log plot it has that look. What is the damping? The amplificaiton factor by the half power point (should be done on the graph to get some accuracy) would probably be > 100. The corresponding damping ratio would be .005 or less. This may not be mechanical.

What about the instrumentation? Other logical posibilities are forced response or instability. This high a frequency instability would not likely be purely mechanical either.