We have vertical multistage pump driven by electric motor on the combined cycle power plant.                                                                                                                    

Motor details are: 4 poles, 1482 rpm, 300 KW, 50 Hz AC motor.              

During our routine condition monitoring, we are detected sound coming from the motor. We are taken vibration measuring, the amplitude at the motor 1.58 mm/s rms 1RH.

I 'm truly need someone who help me to explain well this issue.

you find attached some spectrum FFT's velocity and acceleration, knowing that I don’t know the number of rotor bars. I suspect that we have a problem of out magnetic center. 

thx  for all


Photos (4)
Original Post


1. What is history with this pump? Has this noise suddenly appeared or was it simply noted by data taker?  What have operators noted?  Audible noise from a motor is not that unusual and doesn't necessarily mean its a problem.  (The resolution around the 640 hz and 2280 hz peaks seems low).

2. When the pump is shutoff, does the noise SUDDENLY disappear?  Watch a "live display" of the spectrum during a shutoff if you can; suggest you watch 0-2KHz as a minimum.  

3. Read thru this post    https://www.maintenance.org/top...tor-rotor-bar-number

Regards    Jim P

IVIBR8. Thanks for the reply, in fact the noise was appeared since major inspection of the motor (this motor has been operating since 2011 but the noise increase day after day).
during the shutoff, I can take the measuring vibration because I'm not always available. but if you see the spectrum, there are peaks in the 1X, 2X & 3X. it seems misalignment no?
As well, there are slide bands around 2280 Hz +- 60Hz. 
what do you think about this spectrum, is it normal?

If this seems to be an issue that is getting worse (at least audibly), then can you quantify with past data?   or is what you provided the only data you have?

Harmonics could be the result of several things; misalignment is just one of the possibilities. I'm having a hard time reasoning how that could result in the noise getting worse with time (unless its causing increased clearances). IMO, the level of 2X and 3X (in relation to 1X) just doesn't bother me at this time (i.e. with no previous data to go by - I have no way to trend).  Others may have a different opinion. 

I still believe your best COA at this time is to take data during securing power to the pump.  That will give us all here in the forum a better idea of what to look at next.

At this point in time, I wouldnt be too concerned about the sidebands. Depending on the transient test, this may or may not be something we can look at.  Lets start simple.


I wanna thank you a lot JP. in the past, we were inspected by external company which follow our machines, each 3 months, they gave us only the result without spectrum (mean ok or not ok), therefore, we decide to make our technical office. it has to be said, we are beginner but we make many training courses about in.
If I understand well, that we cannot consider this spectrum as a problem even if there is a noise come from the motor, in fact, I should follow it during shutdown of the pump or take another test. I'll try in the father.


Can you get more data with more resolution?

After a quick look, I see what appears to be 1/2x with at least 5 harmonics following the fundamental of 1/2x. Can you confirm if this is or is not 1/2x

The 235 Hz peak appears to be modulated by 1/2x also. Confirm this?

There are several peaks scattered in the plot that need to be identified also. 

Can you give us a little more info?

Thanks and Have a Great Day,


thanks guys for your reply. about bearing defect, I don't think that there is a problem come from bearing:
BPFI = 13.15*25 = 328.75 Hz
BPFO=10. 85*25 = 271.25 Hz
BSF= 5.16*25= 129 Hz --> no one coincide with our peaks.
Concerning the 1/2 x is 12.5 Hz, there are harmonics around the 1/2x wherein attached be low resolution about in.
About 235 Hz,  you have a reason, there are sidebands  1/2x, you find attached the resolution around it.
if you pick up something useful, don't hesiate to clarify it.



Photos (2)
ai.engineer posted:

i can't take more than this resolution around 1/2x. what about 235Hz ?? 

What parameter settings are you using?  As Ralph mentioned, a higher res recording will show a better picture.

Based off of your first posted pictures, It looks like your using a Fmax of 3000 Hz with Lines of resolution of 400 or maybe 800.  Are you saying that with your data collector, you are unable to increase your lines of resolution to 1600, 3200, 6400, 128,000?

An Fmax of 3000 hz with 400 lines of res, would give you a 22.5 hz separating freq.

At 800 lines of res would be 11.25 Hz and 1600 lines would be 5.6 Hz.


I must say I was kind of leaning the same way as Ralph. With a 50hz line frequency, looking at the multiple half orders and busy noise floor and low amplitude indications of a bell curve I was considering this is indicative of a rub. Although the frequency identifiers are not identified in the spectral data you indicated a bearing issue is not suspected however this may be attenuating from the pump. I would perform a visual inspection with a strobe, check the bearing temperatures and review the process parameters from the reading sheets. It may also help if you have access to an ultrasonic system since the first indication was audible. If walking by you could hear it then it is low frequency (below 20Khz). Perhaps listening at a higher frequency will help to confirm. Otherwise collecting phase data would be helpful, it will work even with low resolution. Not sure if this helps, good luck.


you find attached an acceleration Waveform for the spectrum. 

The bearing temperatures beteween 40 - 50 °C, or the trip alarm in 85°C, therefore our parameters are OK. Concerning The resolution , i'm trying to increase it into software when i did the diagnostic. Knowning that i 'm begineer in vibration analysis, does he anyone explain to me, how can i use the right resolution ?? you can see in the picture below.




Photos (3)

Resolution = (F.S. X WNF X2)/# Lines

F.S. = Frequency Span;   WNF = Window Noise Factor (Uniform=1.0; Hanning=1.5; Flat Top=3.8); 

Because of the "muddy" spectral data you've provided, the forum members are asking to improve the resolution (make it smaller) by reducing F.S. or increasing # Lines.  Your analyzer seems old or limited if it can only provide 800 lines max. Can it do a zoom analysis?


After looking at your pictures of your "setup" (if that is what it is), it appears you are taking a16,000 Hz Fmax with 12,800 Lines.

When you go from 16,000 Hz down to 1,000 Hz from the same collected data, the lines are reduced to near 800.

If you could change the setup (not your measurement point setup, but collecting extra data) and collect data at 1,000 Hz Fmax or even 500 Hz Fmax with 3,200 or 6,400 lines, the data on the low frequency range might would be more useful for "fine tuning" your analysis.

See the attached doc file.


The "new" data posted looks sort of like the old data, a wide bandwidth.

I am sure you know already, that changing a setup will not have any effect on data which is already stored. It only has an effect on the data collected after the setup is changed. 

If this is new data, did you reload the route before taking the data?.

Thanks and Have a Great Day,



Photos (1)


"i wonder, what is the usefulness of analyzer has Fmax = 1000 Hz and 800 lines ??"

It is usable, but sometimes one needs to have more resolution to do a better analysis job.

Can you setup a point with a 200 Hz Fmax and 800 lines? This should help confirm the 1/2x if it is there.


Thanks and Have a Great Day,


This basically eliminates rotor problems.But I do hear a small high frequency change at approximately once per second, but it is "buried" by the louder noise. One has to basically "tune out" the steady noise to hear it. More like a squeal.

Have you some new data taken in the last day or two?

Does your software not allow you to take a 200 Hz spectrum even though you are limited to 800 lines?

Thanks and Have a Great Day,


Add Reply