I was curious to know if anyone else is using ultrasound with slow speed bearings and thier techniques. I have one turning 42 RPM that vibration just didn't show much activity. Of course when you start collecting data that resides in the lower domain, it takes alot longer to collect. The trend data didn't show much change but it still looked like something was going on. The temperature of the bearing was up and it was reported to be noisy during the midnight shift. Grease was pumped in and the noise went away. Using ultrasound I went back and recorded this bearing and another exact same bearing doing the same job. I found at 20K, I could hear this the best. (I normaly listen between 32K to 40K for bearing issues)
Q What frequency do you find to be the best for this slow of speed?
Q Is anyone else using some type of program to look at recorded data and what do you find to work the best.
Attached is what I have been using and the recorded data from the good bearing and bad bearing.
Thanks,
Pete

slowspdbrg.doc (254 Kb, 63 downloads)

Pete,

I use 39-kHz with SDT 150 meter for bearings. For bearings at low speeds, I measure dB-max and dB-Min. Max is the maximum dB value (RMS) that occurs during measurement interval. Min is the minimum value or typical low level (friction or unknown backgound) that occurs in the absence of impacts. I find that low shaft speeds make it hard to determine by meter and audio if the impacts (spikes) are periodic and associated with specific bearing fault frequencies. In that case I use a digital strip chart (DasyLab with PC soundcard) to view peak events over one shaft revolution or several shaft revolutions. I don't find that FFT of the ultrasound is that useful for this low frequency data.

What software was used in your attachment? The waveform time scale shows 255.9m. Is this 255.9 millisec full scale or 1/4 of a second? How do you identify periodic events from this setup?

Walt

Hi Walt,
Thanks for your reply. I used the UE spectralizer with my UE 10000 and only used it to compare a known good bearing to a suspect bad bearing. No real science on this particular one. I'm guessing it's 255.9 millisec. I realy wasn't using this to identify any particular frequency but comparing DB and imacting of two identical bearings. You could clearly hear the difference and I wanted to show it on paper. (I used the waveform as it showed the difference much better)
I have however used this spectralizer with FFT and identified fault frequencies with harmonics. I did this one time on a fan bearing I knew had an outer race defect that showed up very clearly on my 2120. I then recorded it with ultrasound and ran it through spectralizer. Of course I already knew what frequency to look for. But it did work.
You'll have to explain the digital strip chart as I have never heard of it. There's alot of things I've never heard and I'll never know unless I ask.
Thanks Walt,
Pete

Good morning,
my problem here is that I do not think you have enough data. For a rotational speed of 0.7Hz, 700mS, the 256mS you have recorded is inadequate. I am an SDT user, so I do not know what you can adjust on your UE unit. You need to have several revolutions so I would suggest that you record approx 2-3 minutes of data and go back up to ultrasound, say 38kHz.
I use this equipment all the time and have produced many presentations for PdM and IMC over the last few years. My slowest success is 120degrees of rotation in 10 minutes!!
Best Regards,
Tom Murphy

Thanks Tom,
As stated, This was just a quick snapshot I could put on paper to back up the vibration data. It convinced "the man" to replace the bearing. They don't always understand spectral data and it sometimes helps to give them an audiable and visual to back up the spectrum.
Now, your last remark caught my interest. I have a thrust bearing on an agitator that turns one rev in 9 minutes. I set up an ultrasound AP set and used the sonicscan by CSI to try to trend this bearing. It has failed on several occasions with no indication. The only way I can tell if it has a problem is by sampling the grease and looking for metal. I have attached the AP setup CSI gave me for you to view. Is there smething I should do different with this or do you think like me, that the sonicscan just sucks.
I would like to try this again using my UE 10000 and trending it using ultratrend. What perameters would you recomend?
Thanks,
Pete

sonicAPset.doc (116 Kb, 30 downloads)

Hi Pete,
I have absolutely no idea what all of those set-up numbers mean in the AP set - it looks like double dutch to me but I am sure it is meandingful to someone.
First and foremost, forget FFT and forget averaging. These are your two biggest enemies.
Similarly I have no knowledge of the UE10000 or the ultratrend package of which you speak.
Going back to basics though, if I have a shaft rotating at 1/9rpm, I would like to have at least 3 rotations' worth of ultrasound data which is 27 minutes. I would also like to have a marker in the recording to give me a once-per-rev (though this is not likely to be a measured signal and more likely to be me knocking the bearing at a particular point). Once I have this I can review the data to look for certain problems. Depending upon the lubrication method used on this bearing I might be able to look for a rub, for air pockets in the grease, etc.
I hope this is helpful?
Will you be going to IMC? I will be helping out for a while in the SDT learning lab and also chairing some fo the sessions. Perhaps we could spend more time discussing it then.
Best Regards,
Tom Murphy

This message has been edited. Last edited by: Tom Murphy,

I'm no expert on ultrasound, but you've got something turning at .0018hz, or .11 rev/min or 1 rev/9 min. Collecting a sonic spectrum to only 40 orders according to your AP set means the Fmax is only 40X .0018 hz or .07 Hz.

Ultrasonic refers to higher frequencies and usually above a human's hearing range. This is likely below human hearing.

I don't think your Fmax is nearly high enough.

Other opinions? Maybe this AP isn't what I think it is.

Joe, When using an ultrasound device for this application, the user will take the signal from the ultrasound device via the headphone output, then using an adaptor the user will adapt the ultrasound receiver to the CSI, IRD, COMMTEST, DLI, etc...
The ultrasound device or receiver, receives the high-frequency, amplifies, then converts (or heterodyne) the high-frequency to a low-frequency so that the user through the headphones (speaker) output can hear the sound or signal.
So when integrating the two instruments you should interface the two devices. Also, if you record from the ultrasound receivers headphone jack an audio tape recording or wavefile, you now have a converted high-to-low signal to run through Spectrum Analysis software or Wave Form Analysis.

Tom,
Thanks for your honesty. This is a setup given to me by CSI. Like I said, I get no usefull data from this whatsoever. As far as getting 27 minutes of data or 3 revs, it just wouldn't work on this application. The only place I have to get close enough to the bearing is the rotating housing. All stationary iron I feel is too far away. (High frequency doesn't travel very far) I would attach it to the shaft but limited on room as it's inside a support housing.
Jim, I couldn't have explained that better if I tried! It would be nice if another CSI user would chime in with thier experience on this.
So far, I seem to get more from my UE then my sonicscan. I even set up an AP set for trending coal partical size in our pulverizers. It's suppose to increase in DB as the partical get larger. I have seen no difference. As you can probably tell, I don't use the sonicscan anymore and focus more effort with the 10000.
Does the SDT come with some kind of analysis program or route setup like UE?
Pete

Hi Pete,
there is a little bit of urban myth I feel about how much high frequency energy can or cannot be transferred. If you had a simple metal rod, you could transmit ultrasound down that rod quite happily. Granted, low frequency waves will alwasy travel huge distances without attenuation, but that is not the point. The key parameter to remember is attenuation at boundaries. If you can find a path from the rotating to static parts which has few boundaries, metal/air/grease/metal/oil/metal/paint/dirt being a typical path, you should have reasonable transmission.
With the SDT system I can have a threaded sensor much like an accel which I can mount on the machine and cable out to a safe location. Can you do that?
It may be hit and miss, but you should normally find one good place. I will have someone tap the shaft or something with a steel rod to generate some ultrasound and then see if I can find a good measurement place.
Amplitudes are of a different scale here. Remember that the difference between a good bearing and a bad bearing may be over 20dB but the key difference will be the nature of the signal rather than purely the amplitude.
I use a minidisc recorder for these applications with manual gain control to prevent clipping and automatic gain crush. An empty disc has 74 minutes of recording time, once I have the data I need I can edit it easily enough and my vibration program (Adash DDS) will import a wav file so that I can analyse, process and store the signals inside my master condition monitoring database.
I hope that some of this rambling helps!
Tom Murphy