Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment  
spciesla

Sr. Member
Registered:
Posts: 107
Reply with quote  #1 
The attached document shows a series of enveloping spectrums from the main bearing of a wind turbine that operates around 13 - 15 rpm.  The data comes from a permanently mounted accelerometer (100 mV/g).   The second and third plots show a series of 1X spaced harmonics with low amplitudes. 

What would cause a series of 1X harmonics in the enveloping spectrum?

Could this be an artifact of the enveloping algorithm being applied to a signal with nothing but a noise floor?  

 
Attached Files
pdf Enveloping Spectrums.pdf (98.46 KB, 51 views)

Danny Harvey

Sr. Member / Moderator / Supporter
Registered:
Posts: 1,402
Reply with quote  #2 
Steve,

I don't have any experience with wind turbines but I have seen similar spectra using PeakVue.

In my case, there was a long series of harmonics of running speed like your data but it was at motor speed which was about 1500-2000 or so.

It was a bad bnc connector and my strobe was what was causing the harmonics. If I turned the strobe off while monitoring, the peaks immediately disappeared. When I replaced the OEM cable with a CTC cable, the strobe no longer had that effect.

Good Luck
electricpete

Sr. Member
Registered:
Posts: 647
Reply with quote  #3 
So, the spectrum of the envelope of your rectified timewaveform has many harmonics.

Could in theory be impacting at 1x with ringdown frequency above Fmax of your regular spectrum.  Although it looks a little odd to me.

Could in theory be that your TWF envelope varies at 1x even though the contents of that envelope is broadband. Let's say you had a rub for approx 1 sec every time a piece (may one particular blade) passed a certain particular location.   You have a one second burst of random TWF every 4 seconds (for 15 rpm).  The envelope would be full of 15rpm harmonics.

That's a lot of guessing. Look at the actual TWF or enveloped TWF to get a better idea. 


Beatnik

Sr. Member / Supporter
Registered:
Posts: 183
Reply with quote  #4 
I'd also love to see your peakvue and normal TWF.

Often on very slow machines it's hard to see 1x impact in the traditional waveform, but if you see 1x impact in peakvue and it's spectrum, there might be a problem that we associate with 1x impact. Looseness in the bearing, cracked shaft, loose tapper, bad coupling etc.
Vibe-Rater

Avatar / Picture

Sr. Member / Moderator / Supporter
Registered:
Posts: 1,397
Reply with quote  #5 
Hi Spciesla,

About 10 - 12 years ago I did a lot of work on windturbines.  Complicated beasts.  Anything from straight input stage to double epicyclic.  You basically have to understand the design and have access to the drawings. These were mainly Vestas machines. 3 different gearboxes. Hansen, Flender and something or other Lohmann & Stolterfoht, obviously German.  Picked quite a few input stage bearing faults. Confirmed by inspection on changeout. Just used normal acceleration / velocity data.

Also picked a main turbine shaft bearing.  It was a decent size.  240/630 if I remember correctly.  Climbed up it with my CSI gear and a 404B.  First time I had seen this machine. Very tough climb. Told client it had a bearing cage fault but because no history. My trusty SKF Engineer pulled the generator side bearing cap back. And voilla a broken cage.

Very interestingly the concern was raised by Germany (halfway across the world from Australia). The bearing had temperature sensors. The last few years shows very definite seasonal changes in temperature. But the average had been increasing. I.e on an upward slope over time.    We were heros.

I do place a lot of decision making weight on the normal acceleration / velocity spectra. Basically was using those successfully before Peakvue even existed.   Don't get me wrong peakvue is good for certain situations but don't rely on it soley - as CSI/Emerson recommends. 

Re your plots, looks like you are using Omnitrend???  Demodulation / Peakvue / Enveloping etc are all demodulation based.  There are differences off course.  It brings rotational events to the low end of the spectra.  A lot of people jump to looseness but reality is that anything rotational at the upper end of the spectrum is "brought down to Earth" and looks like looseness.

Anyway, not sure if any of that was helpfull.  BTW it is amazing how variable the speed of a windturbine is - DURING EACH REVOLUTION. Order tracking is the go for these machines.  Speed can be corrected using the prominent 3x (3 blades) and as each blade passes the tower it entraps / compresses air causing a pulse. So don't be working in Hz/CPM - yes that old chestnut. rgds
Vibe-Rater

Avatar / Picture

Sr. Member / Moderator / Supporter
Registered:
Posts: 1,397
Reply with quote  #6 
Spceisla,

Congrats on your Mobius Cat4.  Just noticed on the Mobius website. rgds
spciesla

Sr. Member
Registered:
Posts: 107
Reply with quote  #7 

Sorry about the delay in my response.  I have attached another set of plots for your consideration.  These plots come from Bently Nevada’s Adapt.Wind software.  The measurement parameters are not configurable by the end user.  I have two sets of plots that include the Envelope Spectrum, Asynchronous Time Waveform (320 ms), Spectrum of the 320 ms time waveform, and a Synchronous Time Waveform (3 revolutions).  The vibration data is sorted by “Mode” which corresponds to output of the generator.  In this document, the first plots (Figures 1 – 4) are from Mode 3 operation (55 – 70% maximum power).  The second set of plots (Figures 5 – 8) are from Mode 4 operation (70 – 85% maximum power).  You can see the Mode designation in the upper right-hand corner of the plots.

Nothing really new here, although I am seeing the same behavior on the Upwind Main Bearing as well.  I’m not sure if this indicates that the issue is “real” or if it has to do with the Enveloping spectrum.  One item that was brought up to me was that wind turbines use a DFIG (Double Fed Induction Generator).

https://en.wikipedia.org/wiki/Doubly-fed_electric_machine

The engineer that told me about this suggested that the “chopping frequency” associated with the DFIG convertor may be affecting the envelope measurements.  However, I’m not sure that the converter on the generator would affect the Main Bearings (at the other end of the machine) and not the generator or gearbox measurements.

 

Eletricpete – “…Could in theory be that your TWF envelope varies at 1x even though the contents of that envelope is broadband. Let's say you had a rub for approx 1 sec every time a piece (may one particular blade) passed a certain particular location.   You have a one second burst of random TWF every 4 seconds (for 15 rpm).  The envelope would be full of 15rpm harmonics.”

I thought you were onto something here as the timewaveform in Figure 2 shows the behavior you describe (a burst of energy for part of the time record).  However, the timewaveform in Figure 6 does not have this same behavior, but the Envelope Spectrums (Figure 1 and Figure 5) are similar.

 

Beatnik – “…I'd also love to see your peakvue and normal TWF.

Often on very slow machines it's hard to see 1x impact in the traditional waveform, but if you see 1x impact in peakvue and it's spectrum, there might be a problem that we associate with 1x impact. Looseness in the bearing, cracked shaft, loose tapper, bad coupling etc.”

 

The “normal” waveforms are shown in Figures 2, 4, 6 and 8.  No Peakvue though.

 

Vibe-Rater – You are right, wind turbines are complicated beasts.  Most of the ones I look at have a single planetary stage and two parallel stages.  However, I have seen some double planetary stage machines.  We are working in Orders, but it can be difficult sometimes to figure out what shaft the software is assigning “X” to in any given plot.  Thank you for the recognition of my Cat IV!

 
Attached Files
pdf Main Bearing Envelop Spectrum_Async TWF_Async Spectrum_Sync TWF.pdf (320.58 KB, 21 views)

Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.