24 MVA 125 rpm Hydro unit - Stator vibration at 104,2 Hz - why?

[Jari T]

Hi.

I measured a bit strange stator vibrations at one hydro power unit. It is a 125 rpm vertical Francis turbine (~24 m head) directly connected to the generator. There were totally four accelerometers installed radially at the stator frame and all measurement locations indicated similar responses. Typically it is the 2X line frequency (100 Hz) that jumps up when the unit is excited. (Typically means another similar units, this was the first measurement at this specific unit so so not know the history) Now the highest component was ~104,2 Hz meaning 100 Hz + 2X rotational frequency (see attached spectra). The grid is very stable so the frequency is not 2xLF. Any ideas what causes this?

Br:
-Jari

 

[Jari T]

to electricpete: Big thanks for answers! I´ll check the linked article in next possible occasion. The mechanical resonance is of course one possibility. The damping of this kind of structure (steel) is low causing a sharp peak (hign amplitude & not much spread in frequency). However impact tests and modal analysis that I have made earlier have shown that even the sharpest peaks have bit of frequency spread so I think that should cause a bit of amplification also in 102,1 Hz. But the previous was only a general comment based on previous tests. Have had no possibility to test this stator by impacting. I had 4 sensors at the stator (but only two radial locations in 90 deg difference) so this actually does not say anything about possible modal shape Regarding the frequency range, around 100 Hz, I think we are still seeing global modal shapes (of course those are totally different about the complexity compared to rigid body modes) that can be measured with a reasonable amount of measurement locations. Local impact tests probably reveals something but I think that revealing global modal shapes needs a hydraulic excitation. I should have measured stator vibrations with much higher frequency resolution during operation. Now the standard measurement point is to 1 kHz with 6400 lines but couple of measurements with 200 Hz & 3200 lines would have shown if there are natural frequency present. (But the time pressure to get machine in operation asap caused that I noticed this next day in the office and did not measure extra spcectra... ) Actually I like very much using high resolution spectra to reveal natural frequencies. The power of this analysis is that you measure during normal operation meaning there is no errors in boundary conditions and excitation compared to modal tests with artificial excitation. Hopefully I get a change to made some more measurements in the future.

to BrianE22: I did not a bump test.