Gentlemen:

I have attached two data plots (standard velocity & PeakVue)depicting a bearing defect on a 1000 HP Westinghouse Vertical Pump Motor - DE Bearing. Please comment on what you think the problem source may be and severity (based solely on this data). Should the bearing be replaced? All comments are appreciated -
Thanks, Mike

Doc1.doc (350 Kb, 160 downloads) DE Bearing

I would, based primarily on the waveform amplitude at BPFO.

I know some people don't pull the trigger on BPFO, but I guess it depends on how critical this is and how far in advance you need to plan the job. I'm assuming this 1000 hp is no small job and is probably critical to ops. I would pull the trigger now and start planning the replacement.
Obviously the velocity is very low at the fundamental BPFO frq, but those peakvue levels look just like some I have pulled that were in pretty bad shape.

Just my opinion.....

quote:

Please comment on what you think the problem source may be and severity (based solely on this data).

Based solely on the spectrum data, I would say it is in no danger of failing right away.
Strange to see 1X sidebands around an outer race defect in the high frequency range with basically nothing showing in the low frequency range.

The Gs are high in the waveform which might be misleading, which way would be the question, in the waveform or in the spectrum?

The Peakvue waveform really shows some high Gs but the spectrum sort of dies out after about 9 peaks, which is about what the velocity spectrum is showing, between 2K and 4K Hz.

The motor appears to either be heavily loaded or is a variable speed with a speed of 1727 rpm. That would be a slip of 73 rpm. If the speed were 1750 rpm, the peak would be 6X rather than 6.08. Just thinking out loud.

What I don't understand, and probably because I do not have the latest version of RBMware (or whatever they call it now) is the spectrum and waveforms not having any dates showing on them. Nor do I understand why there are labels on the Peakvue called "Expert Spectrum" and "Expert Waveform" and the non Pv spectrum is called route (as it should be) but the waveform is called "alarm waveform". Are these labels coming from the "nspecter" program in the software? Or what is going on with dates?

Only my observation and I may be totally in left field.

Strange indeed to me. I better get me a copy of the latest version before I get lost completely.

The question about the machine being variable speed is a good one. I often see OR defect frequencies at those high orders when the cause is shaft currents. DC motors are notorious for this, but see it a lot in VFD drives also. Generally electrical fluting appears at higher frequencies and as the bearing deteriorates, the grouping of spikes will work its way down in frequency. At some point the bearing will start singing to you. I have watched electrical fluting for as long as 6 months before I threw up the white flag and on a 1000 HP DC motor saw the bearings destroyed in 7 days. Just have to watch the data and look at the rate of change.

Gary B

quote:

What I don't understand, and probably because I do not have the latest version of RBMware (or whatever they call it now) is the spectrum and waveforms not having any dates showing on them. Nor do I understand why there are labels on the Peakvue called "Expert Spectrum" and "Expert Waveform" and the non Pv spectrum is called route (as it should be) but the waveform is called "alarm waveform". Are these labels coming from the "nspecter" program in the software? Or what is going on with dates

I also agree with Ralph. There is something strange about the labels!!! And about the signature, normally peakvue shows the defects very early.. Do you have some HF acceleration readings? Anyway i think you will get enough time for planning for changing the bearing(If it was BPFI it may be severe). Meanwhile monitor the equipment frequently!!

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

I think one version of meter software has options available where instead of selecting the peakvue fmax, filter, lines of resolution, etc settings manually, you can hit one button for peakvue analysis and the fmax, filter, lines of resolution, etc are set based on rpm for you automatically after hitting the one button to acquire peakvue data.

maybe.. probably way out in left field but I think the 2130s have this options also.

I think this beaing is in the early stages of failure, with the way the waveform looks I wouldnt be surprised to see a single small width spallding defect in the outer race

Mike

Mike M.

If severity is defined as length of a spall in the race then having high g's in TWF only does not translate into high severity. Some other aspects have to be taken into account as well. At the inception of a spall, g's is high but so is the frequency. In this case that is exactly what is in the spectrum. As spall worsens (becomes longer) lower harmonics will appear as well.

IMO this bearing is not going to fail at least within 3-4 weeks. Out of curiosity I'd suggest monitoring and trending TWF peak value in PeakVue and frequency content in normal spectrum.

Just an opinion...

Mike M
I am with Martin.. flag it now and get things in order for the change out. It doesn't look good to me. Not saying it's going to fail tomorrow or next week.
Also has this just shown up or has it been getting worse with time?
How does PeakVue inboard Horz. look?
I would be interested.

I agree as well it needs to be identified for replacement and monitored until repaired.

In my limited experience (5 years at a power plant) I have not yet run accross a race defect on lower bearings on vertical machines. There should not be much load unless the machine is misaligned. I am curious whether anyone would interpret this defect differently based on the fact it is a lower guide bearing?

Another thing I am curious about... how close was the reading taken to the bearing? Usually the bearing housing at bottom of a large vertical motor is not accessible, or if it is you have to reach inside near the rotating shaft to get to it...we just take readings on the outside of the frame which is far removed from the bearings.

Combination of low frequencies (less than 6X) and high frequencies (around 90X) are producing beat frequecy shown in the TWF.

Are low frequencies 1X, 2X, 3X and 4X?, Please show enhanced window in this zone.

In conclution, the source of low frequencies (misaligment, looseness, unbalance) maybe yield high frequencies related to ball pass frequencies. Would be good to know bearing number to calculate it.

I believe the solution is correcting low frequencies vibrations, the high are the consequence.

I found out where the "expert" labels are coming from, a CSI built in function for taking a sort of "analyzed" additional set of data but letting the firmware in the meter choose what the Fmax, lines, waveform time, etc. will be. At least that is what I was told by a 2130 user.

But does anyone know where the "alarm waveform" label and the fact there are no collection "dates" on the data is coming from.

This is confusing me. Is this normal in the "uodated" versions of, not only CSI, but other vendors also? How can one keep track of what waveform goes with what spectrum, and when they were actually taken, other than they are coded within the software to be shown in a dual plot? Is this a "bug" in the software?

Where are the CSI engineers board watchers when you need them?

Arne, you are a top notch CSI user, do you know?

Back to basics for me. Based solely on this data I would say this bearing is NOT in immediate danger of failing. Trending would be my recommendation... My concern would be the rate of deterioration. This motor could run like this for years if the cause of the defect is no longer present and the load and lubrication are correct for this type of bearing. Did this defect come out of nowhere? Trends? I know, you said based solley on this data, I just felt like typing.......

Thanks to all for your responses. You ask a lot of good questions and make some great observations. I will answer the questions later as time permits however, I have to respond to Ralph's inquiry regarding the "missing" time stamps. I "removed" the time stamps because I wanted observers to look first at the data as I did when it first surfaced - as I have been monitoring this bearing for several months - more history later. Regarding the "alarm waveforms" it is a mystery to me too and apparently CSI. The only waveforms in my databases that have the "alarm waveform" label have one thing in common and that is I collect peak/phase data utilizing a keyphasor and CSI 410 TTL conditioner. To answer more observation is yes, this pump motor is controlled by a VFD. Thanks - Mike

Not to further derail this thread, but I'm a 2130 user and all my additional data, whether manually setting it up or using the pre-programmed ones, show up as 'Analyze' spectra or waveforms with no labels in the plots. Could be a difference in firmware, though. I think I'm behind a version or two.

quote:

I "removed" the time stamps because I wanted observers to look first at the data as I did when it first surfaced - as I have been monitoring this bearing for several months - more history later.

I may be misunderstanding this quote, but are you saying you have been monitoring this bearing for months prior to the pattern showing you have posted here? If so then we did not get to see what you saw when it first surfaced, we do not have the trend pattern for the months prior. If you did not have the trend either then I guess we are all on the same page.

Patrick,

I don't know what version firmware the guy with whom I talked to yesterday about the "expert" label on his spectrum has, but he said it came from choosing something in the meter that lets the meter do all the setup for acquiring data as oppossed to manually setting it up. Just something CSI wants to advertise that the meter is an "expert", I guess, where the analyst will not get credit for acquiring the data.

Mike merten and all other friends, read the next link :
First , select Vibrary
Second select Rolling element bearing and
Third , follow green arrows

http://www.vibronurse.com/vibronurse.html

Hi guys,
With this Vibration Waveform, can anybody say how much damage the bearing has? I read somewhere that we can find the length of the race defects from the time waveform. Maybe by calculating the time the defect is in the load zone.. Inputs please...

I am a little skeptical of determining the length of the defect from the TWF. But I would love to hear more if someone wants to explain it.

Just to mention some other aspects of judging bearing severity that have been talked about.

CSI has some limits related to true pk/pk time waveform value:

http://www.compsys.com/drknow/aplpapr.nsf/06b6f5a4de2ea...050d1e7?OpenDocument

Their limits for an outer race defect would be 6g's pk/pk alert 12 g's pk/pk alarm. You have somewhere in the neighborhood of 20 g's pk/pk. Also recognize your measured pk/pk values may be slightly less than what peakvue would measure if your sample rate is not high enough.

The true pk/pk value of acceleration has limitations. I believe the primary usefulness is to help determine when you can first expect to see visible damage (and imo when you should first identify a bearing for routine replacement on a critical machine).

As far as how long can you let it go without undue risk of failure (this seems like a risky judgement to try to make but apparently is sometimes required), true pk/pk acceleration is not the best parameter.

Certainly some other important things to look at as you try to judge the severity:
Type of defect - cage is worst, outer race is mildest, inner ring and ball somewhere in the middle.
Number of defects evident in twf - evidence of more than one impact at a fault frequency would be slightly more concern (yours only shows one impact at fault frequnecy)
Number of types of defects - more types of defects present would indicate a worse problem. As damage progresses you might see not just harmonics of bpfo, but also harmonics of bpfi, and ftf sidebands around harmonics of 2*bsf. This may also show up as a general increase in complexity of the spectrum.
Trend - worsening trend of magnitudes, and patterns of defects.

Some people mention an increase in sidebanding as an indication of severity. I don't know exactly why it would be (a little skeptical that it is showing increased clearances which cause increase forces) but I tend to agree with increased sidebands mean increased severity.

Some people have mentioned that in extreme cases when bearing clearances are changed by the wear, then the fault frequencies will change for a fixed speed. I haven't seen a bearig get this bad.

Gentlemen:

First off, I would like to thank all of the contributors to this discussion. Your input, recommendations and observations are appreciated. I will now attempt to close the loop by offering more historical information. The original data plots I posted were collected 13-DEC-05 - the bearing had only been in service for 3 months at this time. “Gary B” referenced electrical fluting as BPFO was limited to the high frequency range. I agree with this assessment however trying to establish root cause and identify the source of electrical current passage through the bearings is still yet undetermined. The other 2 identical motor/pump arrangements with VFD’s do not experience this phenomena. These motors are equipped with insulated bearings – (upper kingsbury thrust and lower guide bearing – SKF 6036). Since DEC-05, shaft grounding brushes have been installed in an attempt to prevent further occurrences. The bearing was replaced 08-MAR-06 and new baseline data was collected. The lower guide bearing shows classic signs of initial fluting to both inner and outer races (will post pics later). As most of you pointed out, the bearing was not in imminent danger of catastrophic failure and had a lot of service life remaining. Our PdM recommendation was to monitor the bearing for further degradation and no recommendation was made to replace however, management’s decision to replace the bearing was the only way “we” would know if the shaft grounding brushes were the “fix”. I have attached a document with historical data plots for 13-DEC-05 to present - Will post bearing pics later as I get them. - Thanks, Mike

15MAR06.doc (2,020 Kb, 22 downloads)

Mike:
You might note whatGary B said about the Bearing in a 1000 Hp motor with Electrical Fluting :
"At some point the bearing will start singing to you. I have watched electrical fluting for as long as 6 months before I threw up the white flag and on a 1000 HP DC motor saw the bearings destroyed in 7 days. Just have to watch the data and look at the rate of change."

7 days notice is better than none, but for critical applications, that is usually not enough. Hopefully day 8 was not the end of the last shutdown.
Gary: Was there a planned shutdown between when the fluting was detected and bearing failure? If so, then wouldn't it have been better to change out the motor then?

I found the comments regarding "calculating time in load zone" interesting, and I think it would be good for another thread.....Which I'll be glad to start.