Motor's high vibrations on vertical pump unit 2

[DekiPL]

Maybe you would be able to help me to find the source of vibrations that we've got on na motor that is installed on a vertical pump. High level of vibration is only in one direction (perpendicularly to discharge pipe), highest on NDE. Motor is working with VFD. Motor type: 1LE5603-3BD23-4GR3-Z 315kW 400V 744 rpm. VFD type: SINAMICS G120P 6SL3710-1PE36-6AA0-Z 355kW 400V. Pump maximum speed is 684 rpm.

The casue of high vibrations is peak at 25,75(@604rpm)-24,94Hz(@656rpm). As you can see it's not a multiple of 1x harmonic and with increasing rotation speed the frequency is decreasing. Below 529 rpm peak at ~26Hz is almost invisible. Above 600 rpm vibartion level is inadmissible.

I was thinking that maybe the cause is beating but 1x harmonics are not in the frequency range 24-26Hz. Is it possible that it is related to the VFD? What is your opinion?

https://files.engineering.com/getfi...8c5c-76614b43d076&file=SpectrumsWaveforms.pdf

 

[GregLocock]

Beating is not visible in the frequency domain, you get two spikes separated by the beat frequency (I'd normally say beat frequency is the delta between the two spikes). There seems to be a problem with your linked file.





Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[spciesla]

Just based upon the description (vertical motor/pump, highest amplitude at the top of the motor (NDE), in one direction) suggests resonance. Recommend performing a bump test or impact test (using a modal hammer) to identify natural frequencies.

 

[DekiPL]

Beating is not visible in the frequency domain, you get two spikes separated by the beat frequency (I'd normally say beat frequency is the delta between the two spikes). There seems to be a problem with your linked file.

Try this:

https://files.engineering.com/getfi...8c5c-76614b43d076&file=SpectrumsWaveforms.pdf

Just based upon the description (vertical motor/pump, highest amplitude at the top of the motor (NDE), in one direction) suggests resonance. Recommend performing a bump test or impact test (using a modal hammer) to identify natural frequencies.

Please look to the file linked above.

 

[GregLocock]

Just looks like a 25 hz resonance, but it is at 2.5 order, is there a gear ratio involved?

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[DekiPL]

Yeah, looks like it, but what is the source if it's not related with the 1x harmonic. There is no gear ratio.

 

[JJPellin]

I agree that resonance is the most likely cause. We have successfully mitigated this type of vibration on vertical motors driving vertical pumps by adding a tuned mass damper. Basically, we clamped a piece a steel bar stock to the motor and tuned it by changing the length and adding weights to the end until the natural frequency of the bar matched the dominant frequency of the vibration. This dropped the vibration of the motor drastically. We have done this with single speed motors. I believe it could also work with variable speed motors.

Johnny Pellin

 

[electricpete]

Yeah, looks like it, but what is the source if it's not related with the 1x harmonic. There is no gear ratio.

It could be flow turbulence. Higher speed, more flow, more turbulence. It would give broadband excitation to excite whatever resonance is there.

In that case (broadband excitation), I'd expect the vib frequency to stay roughly constant(at the natural frequency) as you increase speed. But you have slightly decreasing frequency with increasing speed. That's a little strange, but I still think it could be broadband because it's roughly constant frequency.

If I were looking for a potential excitation source whose frequency decreases as speed increases, I could postulate a backward whirl mode which is theoretically predicted for overhung rotors (including most verticals) and is sometimes observed during coastdown like figure 11 and 12 here. Seems unlikely to me and there are there are undoubtledly other possible explanations for the decrease with speed increase (could be random if the experiment was not repeated/repeatable, or could be fluid forces associated with the operating point could load the piping and possibly tune the natural frequency and shift the frequency slightly if the mechanical system acts nonlinear, or there could be other factors like temperature change that shift the resonance with time).

tuned mass damper.... I believe it could also work with variable speed motors.

I think an undamped tuned mass damper would be problematic on a vfd. It can make things worse if you're operating just off the tuned frequency. The undamped dynamic absorber resembles a 2DOF system whose transfer function has a zero at the tuned frequency and sidebands on either side. EDIT - you gave me an excuse to trot out my dynamic absorber presentation. See slides 14, 15, 16. On slide 16 you see large mass is required to move those sidebands farther away from the tuned frequency. Link

=====================================
(2B)+(2B)' ?

 

[GregLocock]

I very rarely use undamped TMDs. Admittedly tuning the damping is a bit of a black art.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[electricpete]

Yeah, I think if you're working with variable speed stuff like automotive applications, damped TMD is the only option. That sounds pretty complicated to me. Luckily for me, at our plant we have mostly fixed speed machinery so I get to stick with the simpler undamped option.

=====================================
(2B)+(2B)' ?

 

[electricpete]

op
1. was the test done only once, or do you have additional data? Was the pattern of decreasing frequency with increasing speed repeatable?
2. If this is the only available data, can you adjust the first slide to remove the viewing angle in the waterfall and view the spectra straight on, all on exact same axes. Since you have labeled the colors we will still be able to tell which is which even without the angle, and the straight-on view will give us better ability to judge the changes in frequency on both the 25hz and the 1x
3. As an alternative to #2 (maybe better if you have the capability), if you have a peak label feature that uses interpolation to improve the estimate of the frequency of the peaks, can you give us the frequency that the software assigns to those peaks (for the peaks near 25hz and the peaks near 1x)

I'm not sure whether or not we'd learn anything useful from excercize 2 or 3, but sometimes you don't know until you try. I can't help thinking that we're handicapped by our limited view of the data and those details might possibly reveal something useful. I would put this request at a lower priority than bump test.

=====================================
(2B)+(2B)' ?