Wind Turbine Vibrations Project

[steeno]

Hello Mech/Acou/vib people

I am defining a project in Digital Signal Processing.
My theme is vibration in Wind turbines. I am looking
to all sides for

- books
- articles
- web sites

All inputs are of interest.

My level: I have the following mechanical items:
Fundament with tower. Mill hat. Wings.

In the hat: Main shaft, gear, generator.

There are two main purposes with the study:
1. Short Time Measurement of vibration. Alarm levels
for resonant frequencies.

2. Long Time Measurement. Wear on rotating parts.

I have during 2001 developed a DSP programme that can
FFT analyse the measurements of up to 8 accelerometers
placed various strategic places in the wind turbine.

The aim of this years project is to extract vibration data
and to feed data into a decision making program.

Feel free to comment !

 

[GregPerry]

The folllowing specifications may be of interest:
VDI 2056
ISO 2372
BS 46756

 

[steeno]

Ok Greg thanks.
I have asked the univ. lib. to search these standards.

I suspect you are a gear box specialist and that your
references are related to the gear box part of my
question (smart eng-tips feature...)

The cpy I work for have been presented for FAG experts
presentation on gears and ball bearings. I understand
that FAG has a range of FFT analysers for related
purposes.

Is there a web site / book where I can learn the fundamentals of FFT supervision for gears ?

The wind turbine hat is a rough place to be. Mostly
the vib analysis will have to happen on the basis of
(up to 8) fixed accelerometers placed on stratetic places
in the hat. A number of these could be placed on the gear.

 

[brin]

you could start with the vibration institute:

http://www.vibinst.org/

or web site (one of many):

http://web.ukonline.co.uk/d.stevens2/vibcases/chindex.html

 

[steeno]

Ok.
A 30 years old non-profit org with interest groups
all over USA. I must try and find someone there with
insight in wind turbines.

The Stevens homepage has been recommended at other threads.
I quite like its approach. Everything is explained in very
basic terms. Just what I need. I emailed the author approx.
a month ago about wind turbines - he did not have any
contacts related to these machines.

I shall try and hunt Danish sources to the subject too.
Bruel and Kjaer f.eks.

steeno

 

[GregLocock]

You could use synchronous (or gated) sampling, so that all your vibration data would be in at frequencies presented as harmonics of shaft rotation speed, but i think you lose too much information that way. I'd stick with proper ffts, and then use waterfalls and order plots to look at the data.

Cheers

Greg Locock

 

[steeno]

Sorry -

waterfalls and order plots ?

Maybe a book reference would help me here.

Also I could use a few more words about the idea
of relating the sampling freq to the rotation freq
of the main shaft of the wind turbine.

On my desk lies

Steven W. Smith:"The scientist and Engineer's Guide to Digital Signal Processing" 650 pages, 2nd edition 1999 California Technical Publishing, San Diego CA
ISBN 0-9660176-4-1

Very good. Haven't been able to catch this author in
any shortcuts or misleading explanations yet. The
fundamental principles and the mathematical theory
is very well explained, and yet focus is on applied
DSP - how to get safely to programming.

As recommended by good folks on these eng-tips pages:

Leo L. Beranek (Editor), Istvan L. Ver (Editor): “Noise and Vibration Control Engineering : Principles and Applications”804 pages, 2 edition, August 1992 Wiley-Interscience; ISBN: 0471617512

Cyril Harris, Shock and Vibration Handbook, 4th edition, McGraw-Hill, New York, 1995.

And then the university library has promised to
get for me

D. E. Newland, An Introduction to Random Vibrations, Spectral and Wavelet Analysis 477 pages 3rd edition (December 1993) Addison-Wesley Pub Co; ISBN: 0582215846

Now:
What I already have is sampling at a high sampling
rate and then a series of anti-aliasing LP followed
by downsampling.

In this way I move through various freqBands down
to below 1 Hz where I stumble upon the AC-coupling
of the accelerometers.

The chosen AC-coupling cuts off frequencies below
0.3 Hz. Between 0.3 Hz and 0.6 Hz I have to compen-
sate for an unlinear freq-characteristic. But this
is a detail.

How about this one:
"All vibration in a wind turbine will be correlated
with the speed of the wind".

How can correlation analysis be used to remove from
data effects from the changing wind. And what would
be left ? Effects of the mechanical components ?

Thanks.

Steeno

 

[GregLocock]

" waterfalls and order plot"

These are the usual plots we use to analyse rotating machinery. A waterfall plot is simply a stack of spectra at equally spaced increasing speeds, offset slightly in both screen x and screen y so as to produce a 3 dimensional effect.

The fundamental and its harmonics then show up as diagonal lines of peaks, and resonances show up as lines perpendicular to the frequency axis, excited by each harmonic on the way up.

An order plot is a section cut along each harmonic (or order), so you get a plot of say first order (one times shaft speed) vibration against speed.

Campbell's diagrams are also used, these are (roughly) contour plots of the waterfall plot.

"How about this one:
"All vibration in a wind turbine will be correlated
with the speed of the wind".
"

I would guess that that is a bad choice of phrase. Wind speed is a DC measure and cannot be correlated, in a signal analysis sense, with vibration. I think he means that you should always note the windspeed for a given measurement.


Cheers

Greg Locock

 

[steeno]

Thanks Greg

I shall read carefully what you have written about
waterfall plots / 3D plots.

The sentence about the wind speed was my own.
Here is my background for trying the correlation
idea:

I have a readout here of wind speed versus time
recorded at 32 samples/s.

The wind speed lies in the interval 15-25 m/s.
With a heavy DC component of say 20 m/s.

I have performed a 1024-fft on the signal.
It shows a dominant peak at 0.022 Hz and
what looks like a series of harmonics to that
freq ie. .044,.066,.088,.110,.132,.154 etc.

Then I have a plot of the power produced
on that mill. Again a heavy DC, again the same
picture when looking at a 1024-fft: Appearently
the same fluctuation, same harmonic series on
the ripple of the power.

Now:
The vibration of the turbine is different.
The measurement I have is momentum at the
root of the tower. The signal has no DC
Probably because the transducer (accelerometer ?,
strain gauge ? ) was AC coupled - A wind of
20 m/s should produce a DC momentum on the root of
the tower I think.

Anyway:
The measurement shows a very dominant freq of
0.352 Hz. I was told that this is due to the wing
passage of the tower. With 3 wings this means that
the main shaft rotates with a freq of 0.117 Hz, ie.
T = 8.53 sek for the measured wind speed of 20 m/s.

I wish I had some vibration measurements from the
mill hat (Danish terminology here). I am sure the
wing passage of the tower would still be visible
as the dominant vibration.

I shall use the term "correlation" more carefully
in the future. Lets start with "influence". Since
all movement in the wind turbine comes from the
wind I would like to say something qualitatively
about the influence of the wind on the measured
vibration ?

 

[GregLocock]

I know very little about windmills, so I may be completely off base, but:

I would expect the dominant excitation of the structure to be due to blade passing frequency (BPF). The structure will have its own resonaces, which may mean that you get higher amplitudes of some harmonics of this frequency. You will also get some vibrations at harmonics of shaft speed, and at gear mesh frequencies. if you get a strong signal at shaft speed then suspect imbalance or bending or misalignment of the shaft.

Upstream disturbances of the airflow will have a relatively small impact on the structure.

i think your very low fequency signals (and their harmonics) may well be a red herring, caused by your signal analysis procedure, or by a property of the instrument you used to measure windspeed. if you really want to measure these pressure fluctuations use a microphone, although you may have to modify it to get a DC response, or a dynamic pressure transducer.

Since you seem to be Danish why not talk to Bruel and Kjaer? Cheers

Greg Locock

 

[steeno]

Thanks Greg

From now on I shall use the term
blade passing frequency (BPF).

I was thinking about the wind measurement after writing
my last posting yesterday. I was reaching the conclusion
that the wind changes must be abrupt (like a square wave)
in order to produce a series of harmonics. But as you write
it could well be an artefact of the measurement system.

I am sure wind data are abailable for all sorts of places
and conditions. Only I need to find them.

Yes, Bruel and Kjaer. I will find a contact there.
Also there is a public test site for wind turbines
at Risoe, Roskilde. I have mailed a couple of people
there but without any response so far.

Again: There must be an influence from changes in the wind.
I guess I have to divide wind impact in abrupt shortterm
pulses and then longerterm changes. High time to find a
"wind atlas" with these things described.

Ok I shall get back when I have some results.

 

[FIS4AID]

You mentioned FAG In one of your earlier posts.
FAG has now a presence in the States. F'IS is the Industrial Service division of FAG and are located in Miami, FL.
We have been monitoring windturbines in Europe for several years now and have different equipment on the market. For your application the best unit would be our DTECT X1 system. This unit might well do what you need ore even more.

 

[Hatch]

This is a late message but take a look at Stress Wave Analysis monitoring to gauge the health of the wind turbine.

The web site of 1 of the companies involved is:

http://www.swantech.com

It looks promising from the continuous monitoring perspective where you can compare trends and amplitudes. I'm just not sure about diagnostic use of the technique. C. Hugh (

http://www.Hatch.ca)

 

[steeno]

Thanks
from the swantec.com homepage:
...marketer of sophisticated ultrasonic diagnostic systems that accurately evaluate the operating health of industrial machinery. SWANTECH's proprietary, patented and field-tested Stress Wave ANalysis (SWAN™) technology is revolutionizing the global condition monitoring market by providing a cost-saving machine defect detection solution that is replacing traditional vibration-sensing and lubrication solutions that have existed for over 25 years....

I saw the other day an article about B&K Medicals ultrasonic
equipment - I hadent thought it might pop up in my wind
turbine project.

Just now I am trying to understand the theory behind wavelet
analysis. Do anyone have experience good/bad/opinions
on this mathematical tool (a bit like fourier - and
there is also a fast algorithm to do the analysis realtime...)

Thanks for the reply. I shall keep my eyes and ears open.

steeno

 

[GregLocock]

I can't see much point in ultrasonics unless they just mean listening to the (ultra) high frequency creaks and groans and bearing defects - we used to call that Acoustic Emissions, and I haven't heard of it for 15 years!

Wavelets are fine, but are still ruled by the B=1/T rule, which I've always liked because it links signal analysis back to Heisenberg's Uncertainity Principle, ie Nuclear Physics.

and here's a very good web page on wavelets:

http://engineering.rowan.edu/~polikar/WAVELETS/WTtutorial.html

Wavelets won't be much help to you, I think. Cheers

Greg Locock

 

[SwanMan]

steeno,
Are you still interested in information on your wind turbine vibrations project? If so, I have data and information to assist you with the "wear on rotating parts" area. I have been monitoring nearly 100 wind turbines in the field for about 14 months now and have learned a lot.

 

[SoCal]

steeno,

SWANTECH's main use in wind turbines is detection, trending and isolation of mechanical defects in the drivetrain. SWAN's high frequency acoustic sensing technology filters out background levels of vibration and audible noise, and provides a simple, trendable graphic representation of machine health.

SWAN's position in the wind industry is favorable and should be looked to for information regarding the "long time measurement" you mention as one of the purposes of your project.

Here is a link to a wind turbine case study from SWANTECH's website,

http://www.swantech.com/applications_monitoringwind.html

Regards,

Dave Batista
SWANTECH

 

[steeno]

Hi Dave, Greg and all

I am still working on my wind turbine project.

Its been divided a bit in focus:
1) Wavelet analysis, therory and background

This part of the project focuses on getting
*fast* vibration measurements with a short
observation time. But its true that wavelets
are not the obvious choise as long as the
vibrations are sine-wave shaped.

2) The FFT long term monitoring nearly finished
as projected. Here I still need to work in
practice with the waterfall plots that Greg
mentioned in an earlier mail. I still believe
that the analysis should reflect the variable
shaft rotation speed situation. Thanks for
explaining the background for these plots.

As for Swantec I have noted the web link. I
will check in and have a lok around. Yes - I
am short of monitoring data - real world data.

Steen

 

[WIM32]

Steeno,

Maybe check out Mecal.

http://www.mecal.nl/blue/index.asp?nav=windturbines.

Their windturbine division does all certifications for the big "wind" companies, I don't know if the are into measuring though.

 

[8ptbuck]

steeno
I have been doing my own study on wind turbines for a year now. I was originally called in to collect data on several generator bearings on a i.8 meg units. This expanded to gearboxs as soon as I took my first set of readings. I am presently monitoring several sites throughout North America on different models and sizes. My work on these units involves vibration analysis, phase analysis ODS, natural frequency testing,run-up coast-down gear box inspections bearing replaements failure analysis etc. All testing performed is up in the hat as you call it with the units on-line. I refer to the hat as the cell.
to truly understand what is going on with wind turbines the analyst must perform all of the above operations. As you probably know most of these units will run free while not on-line depending on wind conditions. The mentality here is if the unit is not under load then damege cannot occur. I believe critical speeds and natural frequencies are seriously overlooked here when the units are not on line and up to speed. Also the alignments are set-up to fall into place once the unit is at production speed (1800). If you go with an online system or you plan on monitoring a unit continusly pay close attention to the unit when not at operating speed. I gave a small presentation to some grad students at a local college on vibration analysis and what to look for on these units. maybe you were one of the attendees. Good luck with the study and let us know how it goes. By the way the equipment I use performs all of the above is handheld and about seven pounds. This makes it pretty easy when I climb the 80m towers.