I was out in the plant with my UE instrument and decided to record wave files from a few of our large vertical motors, the motors themselves not the bearing housings. They files were interesting to listen to but I'm not sure where I'm going with the endevor. My question is: has anyone developed a program where they regularly record motor wave files and if so do they think it's time well spent or just a waste of harddrive space? Any input would be appreciated.

Hello rgf12,

Would you care to share your wave files with the group? Or a sample of 2-3?

In collecting this data what was your goal? You mentioned you were checking the motor itself, not the end bearings. Were you hoping to hear electrical issues, or overall mechanical noise? Or were you just curious to hear something?

I was just looking to see what I could here from the motor itself because Operations reported the pervebial "funny noise" when they started the pump. Vibration didn't show any abnormalities so I decided to record a wave file. I'm trying to get some handle on if it is worth my while to record motor signitures on a regular basis and if there is any history out there that could give me some insight into this effort.

I don't use ultrasonics much, so take my comments with a grain of salt.

I think pointing an ultrasonic gun at a motor has the same value as any other piece of mechanical equipment. We may be able to localize the source of the noise based on directionarility. Bearings, rub, or looseness might possibly create noise that could be narrowed down by ultrasonics similar to other mechanical equipment.

There can be electromagnetic noise produced - primarily k*RBPF +/- m*2*LF where k and m are integers. imo that particular frequency pattern doesn't tell you much of diagnostic value. Some motors make a whole lot of electromagnetic noise by design and often times that noise increases with load. If you see 2*LF, 4*LF, 6*LF, 8*LF etc (on log scale), then that would be evidence of looseness driven by a magnetic force (could be loose core, loose windings, or something else). I assume the frequencies which show up in audible are the same ones that show up in vibration although of course there are factors affect which ones get amplified when monitoring through a certain path.

According to the literature, there is a chance of locating partial discharge in switchgear with ultrasonics. In theory I guess it might be applied to motors but I wouldn't hang my hopes on that until I had seen some successful demonstrations.

You may also find various forms of aerodynamically created frequencies similar to vane pass frequencies. For example rotor end fan blades going past a certain point.

So the ultrasonics gives you directionality. I don't think routine use of the tool gives you spectrum analysis. That limits what you can do. If the noise is coming from the air inlet our outlet what does it mean? Probably that it is the most direct path to the inside of the motor... not much more.

I share your pain of trying to track down "noises" reported by operations. If we can pick out the unique noise ourselves from among the machinery drone, or if we can pick up a bearing or looseness on vibraiton, than we have something to work with. But often we don't hear what they hear. There are a number of problems in that people don't hear the same thing, and sometimes noise such as bearing may come and go. If I ever get my act together, my plan would be to record the noise over the audio range using simple microphone (perhaps directional) as wave file. Then one can do FFT, TWF, envelope analysis autocorrelation on the audio to help figure out exactly what it is that people are hearing. Also serves as a reference for comparison when we collect a catalogue of noise. It is not just a matter of diagnosis - there is also the important function of communicating with the operators for example to verify you have heard the same noise and to assure them we have taken their concern seriously.

Thanks for the input. What I've been doing is taking a contact probe and putting it on the motor case (away from the air inlets and outlets) and recording the wave file and using the software to look at the TWF and spectrum. I guess I will select a few motors and record some data. I will post a couple of files in a week so readers can listen to what I've recorded.

Pete - Did you develop a spreadsheet to perform autocorrelation of TWF data?

Hi Steve,
Indeed I have made a lot of progress on my "bearing post-processing" spreadsheet (attached) and it certainly does auto-correlation.

The "instructions" tab tells you how to import from E-monitor into the spreadsheet (brings in the data, the point label, order reference, and a few other things).

If you don't have E-monitor, you can cut/paste your data starting at B11 (time) and C11 (acceleration or velocity).

More instructions in that tab but basically the grey buttons in the TWFdata tab do what their name says. Push the auto-correlate button and read the prompts to perform autocorrelation.

Then go to plot_TWF's tab to view the data.

There is a lot more that can be done with the spreadsheet as you can see from the other buttons. One upgrade that is sorely needed when I get time is to code in another FFT routine - excel's FFT is limited to 4096 points which usually kills the resolution.

In my next post I will discuss a little more about the machine whose data currently is loaded into the spreadsheet. Auto-correlation very clearly shows inner race bearing defect, when it is tougher to pick out on some of our other indicators.

This message has been edited. Last edited by: electricpete,

BrgPostProcessR3a.xls (4,109 KB, 22 downloads)

Attached is a powerpoint with overview of some predictive data for our CD motor 11 lower bearing which appears to have an early stage bearing defect. Auto-correlation (slide 19) shows is very nicely even though you can't pick it out in TWF (slides 8 and 9), somewhat difficult to pick out in spectrum (multiple slides). It didn't show up in my home-made demod spectrum I suspect due to resolution limitations. We don't have any demod spectrum collected yet.

Actually, I suppose I should move this part of discussion into the vib forum. Don't want to steal a thread or go too far off-topic.

CD11A.ppt (1,444 KB, 30 downloads)

There seems to be a problem with labeling the TWF's in the TWF plot tab. I will fix that and re-post.

Here is a corrected spreadsheet (labelling of curves in plotTWFtab is fixed - will display whatever curves are selected to "TRUE" way down below the plot in cells E36+)

BrgPostProcessRc.xls (4,280 KB, 29 downloads)

rgf12,

What instrumentation have you used to record the sound output from your US instrument ( BTW what kind of US meter?).
Is it a MP3 recorder, magnetic tape recorder, or a digital box, such as vibration analyzer?

I have tryed 1st and 2nd methods ( consumer type of recording instruments). It appears they don't have the required frequency range. Recording geterodyned sound TWF with vibration analyzer does provide valuable information, but the question is :'How to play back the recorded TWF ?'

Thanks,
David

Sorry about the bad typing-I use an UE 10000 that records wave files and Spectralizer Software to play the files. The software gives you a TWF and spectrum for analysis.
Electricpete-a great tool for TWF analysis! Do you know of a way of dumping data from CSI RBMware into this spresdsheet?

This message has been edited. Last edited by: rgf12,

quote:

Do you know of a way of dumping data from CSI RBMware into this spresdsheet?

If you know how to get the data out of RBMware into excel, I could probably tell you how to set up the spreadsheet. Do you know how to get your data out of RBMware into excel? (I don't use that RBMware)

The UE 10000 data is stored in WAV format. You can actually play it back with any software that accepts the format. I have no idea how you would convert this to something acceptable to Excel. The UE software is somewhat limited. The best way to perform a comprehensive analysis is to take the headphone output and intput it to your spectrum analyzer. Another way to look at the data is with 3rd party software. Spectraplus is a useful program than can read wav files.

El'Pete,

quote:

If you know how to get the data out of RBMware into excel, I could probably tell you how to set up the spreadsheet. Do you know how to get your data out of RBMware into excel? (I don't use that RBMware)

In CSI MasterTrend (assumee still available in RBMware) "Plot Data" has a "List" function for Waveform plot that provides 8-columns of Time-amplitude (4-sets) that could easily be imported/pasted into Excel and then reformated.

Walt