Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

Register Log in
  • Congratulations waross on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!

detect hole in fan blade

Status
Not open for further replies.
faquir

faquir

Mechanical
Oct 30, 2014
11
0
0
PT
Hello !

I don't know if this is the right forum but here it goes:

I'm doing a thesis on defect identification through audio signal measured in sound level meter. My professor asked me to apply my program in a ""simple"" case which is a 10 diameter microwave cooling fan with a hole in one of the blades. So I'm suppose
to detect the audio signature of this hole in the spectrum of the fan rotating at 900 rpm. The only hint he gave to me was to
try a dimensionality analysis, considering the hole could be emitting vortex in frequencies of the order U/D. U being the linear speed
of the hole and D the hole diameter. After talking with aerodynamics professor and doing a bit of research I realize the flow in a fan
is "highly tri-dimensional and unsteady" and blade tips can be also emitting vortex, so it's not that simple. I couldn't find anything yet in the spectrum comparing no-hole spectrum with spectra of various hole diameters, just a bunch of random broadband noise and tonal frequencies. If anyone can give me any hint or advice or whatever I would deeply appreciate.

Thank you very much for the attention!
 
Replies continue below

Recommended for you

Sort by date Sort by votes
Can you /hear/ a difference? If not then I suspect that you are going to have a hard time of it.

The big problem if your prof's theory is right, is that vortex streets sound like pink/white noise, as does any other turbulence, which your fan already generates. There may be some time based structure to the additional noise, but it will be heavily confounded by blade passing frequency. Your new noise should have a structure associated with 1/rev, but then so will the fan itself if it is a cheap one!

Unfortunately it comes down to playing around with binning frequencies together and looking in the time domain, possibly wavelet analysis. I hope you have the Matlab Signal Processing toolbox.

One option might be to build a quiet fan of the same performance and rpm, and see if holes in that are audible.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376
 
Upvote 0 Downvote
10 what diameter?

I'd take a shot using a 1X trigger to gather "time synchronous" data. That can do a remarkable job eliminating random info.

A highly directional microphone might make some sense of it too.

Maybe listen with a piece of tubing etc at each blade (front and back) with the non-rotating fan mounted in an airstream?

 
Upvote 0 Downvote
Hello and sorry for taking so much time to reply.

First thank you all for your feedback!

GregLocock: 1st: I can't hear the difference

2nd: I have SP toolbox yes and I already tried to search for that time structure at 1/rev sec, using empirical mode decomposition
and hilbert spectrum but i found nothing, but i have to say I'm no expert here... Do you think wavelets are better? I think one of problems here is there's just to much noise (and yes the fan is cheap :D )

3rd: What do you mean by "binning frequencies together". I'm not native English speaker sorry...

Compositepro: It's 4 blades with a hole in one blade only.

Tmoose: 1st-> 10 cm diameter.
2nd-> I don't understand what is 1X trigger to gather "synchronous data". I'm going to do some research though!

Thank you all!
 
Upvote 0 Downvote
Hello again!

TMoose, sorry for asking again. What I have been seeing about synchronizing data was calculating cross correlation between signals
so it is possible to "align" them in the time axis, using maximum correlation. Beeing this non-stationary and noisy I don't see how it's possible to do that well, or even if it makes sense !? The trigger thing that u are talking about is like for example when a blade passes
a certain point, it triggers the sound meter to start recording? I don't know if this is making any sense but I'm a bit confuse here...
 
Upvote 0 Downvote
Assume for the moment that the noise is not random.

If you had a position sensor that gave you a spike or an edge as exactly the same point passed the sensor for each fan blade, and you had a two-channel recorder for the position sensor and the noise, then you would have recordings of three perfect blades and one holed blade passing the microphone, streamed in time, with the blade recordings in sequence, and with the second channel providing handy delineators.

You could then displace a signal from a perfect blade in time, invert it, and subtract it from the blade + hole signal, leaving you with the hole signal alone.

Maybe. Have fun with it.

(There are ways to develop the delineators from the audio, sometimes, but they make my head hurt.)



Mike Halloran
Pembroke Pines, FL, USA
 
Upvote 0 Downvote
Where exactly are you measuring the sound? Can you get closer to the fan? I would want to be able to synchronize two microphones to one good blade and one bad blade and do some comparative analysis. I would start with getting so close to the blades that the mikes can pick up the whoosh of the blade going by. Start there; if you can't pick up anything under a completely focused condition, then there may not be anything to pick up. But, you should ALWAYS start with a case where there is something that can be clearly detected and work from there.

TTFN
faq731-376
7ofakss

Need help writing a question or understanding a reply? forum1529


Of course I can. I can do anything. I can do absolutely anything. I'm an expert!
There is a homework forum hosted by engineering.com:
 
Upvote 0 Downvote
Hi faquir,

A very brief description of "time synchronous" averaging as often available used by FFT spectrum analyzers for machinery vibration diagnostics page 6 here.
A single channel instrument with a trigger input is sufficient. Mine uses a photocell TTL pulse.

This HP/Agilent application note shows the general effect of time synchronous averaging to "average out" random signals.
Page 44 in section "linear averaging" here -

Lets say the hole makes some kind of 1 per rev acoustic pulse.
I'm WAGuessing that TSA spectrum will show a once per rev peak, and a bunch of integer harmonics as is typical of a pulse train.
Perhaps that would only be useful in comparison to an unholy fan.
Nicely spaced pieces of shiny tape ( one per blade ) ///might/// help discern some of the individual blade characteristics, but I'm thinking all the blades would still be considered 1X, so the aberrant blade's "boing" or "ppfffft" would still just modify the average
 
Upvote 0 Downvote
I'd think it would be easier to see the loss of blade balance, via fan bearing or mounting point accelerometers.

Past that...have you tried cepstrum analysis (taking an FFT of the FFT)? Used for bearing wear and similar.

Last idea, I would try putting the mike on the upstream side of the fan, if for no other reason than you might get less flow noise hiding your hole signature there.

 
Upvote 0 Downvote
Once again thank you for all your feedback I'm very grateful.

I'm measuring the sound in a lab with some background noise. The mic is in downstream approx. 20 cm from the fan, but as long as I get closer I can use a windscreen, so no problems in noise there. I've used the suggestion to measure both with the mic perpendicular to fan plane and parallel to same plane, and I'll see what I get. If I don't see anything I'll start the synchronizing suggestions.
By the way this time I drilled several holes and the last size is very big and I'll start comparing this one with the no-defect blade. I think there is a difference one I hear both sounds.

Again thank you all!
 
Upvote 0 Downvote
Forgot to say, there is vibration noise of course of the plates that support mic and fan structure and other noises that i suppose come from the fan beeing old and the lubrication is not the best. The former is is pretty audible but the latter is very low.
 
Upvote 0 Downvote
Status
Not open for further replies.

Similar threads

stormloader
  • Locked
  • Question
Blade Helix definition? 2
Replies
16
Views
49
stormloader
stormloader
faerosp
  • Locked
  • Question
Industrial practices in CAD modeling of machined parts 1
Replies
16
Views
33
faerosp
faerosp
EdwardNigma
  • Locked
  • Question
How can I self-learn finite element analysis in a group that does minimal FEM work? 6
2
Replies
26
Views
121
Sparweb
Sparweb
SpaceEngineer4
  • Locked
  • Question
Conservatism in/Limitations of Shock Response Spectrum Analysis 1
Replies
2
Views
19
Sparweb
Sparweb
SteveAero
  • Locked
  • Question
Shear transfer via fasteners in torsion in built-up, thin-walled, open sections
Replies
11
Views
35
Stress_Eng
Stress_Eng
Back
Top