Ok Guys I have something that goes with the poll that was posted yesterday. I have an exhauster fan, overhung 6 vanes, supplying a Mill. It has a bearing pedestal with two bearings SKF 6319 and SKF 22319C. I have been tracking a bearing fault in this bearing for about a year. The fault was found first in the PeakVue spectrum after I setup the PeakVue point. This means that the fault was there for a time previous to being first noticed.
Now the fault has continued to progress to a point that it is now plainly visible in the wave form and spectrum on a standard data collection point. In the spectrum it is best seen in the upper frequencies still. The fault presents as BPFO only at this point. I feel it is at a CSI Stage 2 Bearing Fault at this time. The Spectrum trending has been relatively flat with a small increase this month and last month. The same goes for the wave form it has remained below 1 g with a crest factor of 5.8. The crest factor prior to this month was below 4
How do y'all place the severity of this fault? I feel that we can schedule this bearing for replacement in 3-4 months, based upon the present state of the fault and the low turning speed. Am I way off base here, am I missing something. They get really pissed when I call something too early and really really pissed when I call something and it fails before they get around to working on it.
For those with CSI 4.8 I have put the past few months date in a small Dbase that you can open and view as you like.

1B_Exhauster_Fan.doc (373 Kb, 102 downloads)

Sorry I am having trouble with the RBM file. If you are interested email me and I can sned it to you email.
bbroad at firstenergycorp dot com

Kelly;
You didn't talk about scheduling. How long between available shut downs to work on this fan?
Will a failure shut down production or cause a safety risk?
Will a failure result in peripheral damage such as shaft, housing, impellor etc.?
If you consider these questions when you look at the clear defect in your data, it may help you to "make the call" as we say.
For example; if I see a bearing defect on a hood exhaust fan here, the replacement is planned as next opportunity with continuous monitoring to prevent catastrophic failure. This is because the last time we did have a failure, production was shut down for 35 hours. Hope this helps.

Ron.

Are you using an accelometer? Do you have acceleration data - spectral profile data @2 kHz and its TWF in acceleration? It should be easy to put within a reasonable time frame for failure; but, the data must be reliable.

Ron,
This is a critical fan as with out it we have to take an immediate derate on the boiler reducing mega watt output costing us mega bucks. Outage is planned for 2007, with some small maintenance outages until then. Usually we can get one of these to work on whenever the mega watt demand is lower ie. Weekends and holidays, but that always depends on other forced outages at this power house and other powers houses in the company.

I guess I want to make sure I repeat the mistake I made last time of calling the bearing too early. They weren't real happy with me.

Sam,
The data is taken with an accelerometer and then integrated to velocity in the spectrum only, the waveform stays in acceleration. The PeakVue spectrum is already in acceleration, however it is acquired though and enveloping process.

Thanks to all
Got a lot to think about, other equipment to analyze.

Kelly:

I'm interested in how you describe "calling a bearing too early"... meaning the defect was not visible?

I'm (all are) faced with similar problems.. how fast will this defect develop between shutdowns.

I think in part you have a PR problem. Ask the critics if they are willing to guarantee that the bearing would make it to the next shutdown.

I looked at your data and here is my sense of it:

1. Its BPFO, so it develops the slowest
2. It is a single defect
3. WF amplitude is low
4. spectral amplitude is low
5. RPM is very slow

I would not be in a panic about it, but then I have only been doing this for a couple years.

If this is critical as you say, I would increase the frequency of data collection.

I'm interested in how others here evaluate your data.

Good comments, Martin.

Note also that for a single fault, crest factor values tend to increase as RPM decreases.

Kelly,

One thing caught my eye: some randomness visible in both normal and PeakVue spectra. This could be coming from underlubrication. Underlubrication may even produce faulse bearing fault frequencies.

Although I personally hate using alarms, both normal and PeakVue TWF amplitudes look low by all standards. The impactive pattern, though, is clearly there and with CF climbing up points out to a possibly developing fault.

I would monitor it for awhile (6 months) if possible, but for now I suggest lubricating if it has not been done already.

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

Dave,
The bearings are in an oil bath and the level is correct I check it Frequently. I guess one factor that may be considered is that two years ago the fans were ungraded to an 80" fan from a 76" fan for performance issues. The bearings were not changed as engineering dertermined them to bearing intolerance with the prejected loads of the new fans. I believe the oil is still the same type.
Maybe the oil is not up to the new loads?
I know that we have had a lot more bearing faults since the fan upgrades.
More Stuff To Check.

Duncan:
Regarding CF in low RPM.
Could you elaborate a bit?

I tend to discount CF in low amplitude WF, but have never factored rpm into that.
MMM

at that point, you have to take into consideration the sticky factors, like how serious your client is about keeping their machinery up and running.

i had a situation like this at a poorly-maintained power plant up here in new england. i had seen the bearing fault for over a year with only a slight increase in amplitude on the quarterly surveys. you have to tread that line of covering your arse, while not sounding too many alarms as to make your client question your analysis.

i think i kept telling them something along the lines of, "there's evidence of a bearing fault in bearing x of machine x. although vibration amplitudes to not indicate a danger of imminent failure, this type of bearing fault does have the potential to rapidly worsen. it is recommended this machine is closely monitored until bearing x can be replaced during the next scheduled outage." - you kind of put the call in their hands at that point.

Martin - the crest factor is defined as the peak value divided by the rms average. For a single flaw, the accleration value caused by an impact has a relatively fixed rise and decay time. As speed decreases, the impact caused rise and decay time becomes a smaller part or each revolution, hence the average decreases. The peak value may also decrease as internal bearing element relative velocities decrease, but usually it does so at a much slower rate than the average decreases.

Kelly,

Email the data base to me.

It doesn't look that bad to me, but I would like to check a few things out.

Good Luck,

I forgot that e-amils are not in the profiles anymore. [dannyharvey@machineryanalysts.com]

Duncan:

I think you worded it backwards..

"The peak value may also decrease as internal bearing element relative velocities decrease, but usually it does so at a much slower rate than the average "

if pk decreases slower than RMS, CF goes up and vica versa.

Anyway, Do you have any papers that document lower than average CF for low RPM machines?

That would be interesting to me.

Or is this a math based conclusion?

If the peak value decreases slower than the average value decreases, then the crest factor increases and that's typically what happens. I don't have any papers showiing this.

There's a lot of discussion about CF and TWF amplitudes. I'd like to see a regular, non-correlated TWF. The numbers are completely different.

Kelly, It would appear the data you have posted is an inner race fault, not outer race. If the shaft is turning 893 rpm. Outer race is typically 3-5 times rotational speed. My experience with inner race faults is they show up just above run speed and carry all the way thru the spectrum. Looking at the time waveform, there is definitely a visible fault in the bearing at this time, but it is minor and should be trended for step changes. 6319 and 22319 bearings are decent size bearings, as Martin spoke of in his response the machine turns slow. I would anticipate 40 g's (+20 & -20) is the maximum this bearing should ever get to, the key is how fast will this happen? This bearing could last for years in this condition. I am currently monitoring a 300hp split case pump with inner race fault. Machine only operates 6-8 months per year, trended bearing for 3 years with very little change. Temperature is also a key factor in trending bearing faults. With oil bath lubrication, if heat starts to develop you know metal is coming off the bearing and getting lodged between the balls.
Hope this helps

quote:

Kelly, It would appear the data you have posted is an inner race fault, not outer race. If the shaft is turning 893 rpm. Outer race is typically 3-5 times rotational speed.

Dave Reynolds, If you look in the attached document in my first post you will see that the fault is a BPFO of an SKF 6319. I have also gone to the SKF web site and verified that CSI didn't get the frequency wrong (like that would ever happen ) Also 3x turning speed is alomost directly on the first BPFO fault line. So you are right about it showing up at 3-5 times turning speed.

As far as length of run time, as opposed to the bearing you are watching this exhauster runs at 100% load 24/7 non stop, no rest for the wiery.

I have gotten them to agree to scheduling the bearing replacement in 3-4 months, based upon continued monitoring weekly. In two months we will revisit the schedule based upon the trending of the bearing. If it trends up sharply we go after it as quick as possible. Hopefully my fat stays out the fryer.

quote:

Originally posted by KellyB:
If it trends up sharply we go after it as quick as possible. Hopefully my fat stays out the fryer.

As if to say it's your fault the bearing has a defect................

KellyB,
Your data definitely shows bpfo. It actually shows it a little "too good." The fault peaks and harmonics stand up really clear even in the velocity spectrums. If you press the "1" key and change the velocity spectrum to acceleration, even a blind man could see that fault. What bothers me is that the waveform doesn't have enough Gs to go along with such nice spectral peaks. We see the same thing fairly often around here. I think of it as "crying for grease." After a few shots of grease, the spectral peaks usually disappear. If they reappear in 24-48 hours, you have a problem. If they stay gone, everything's cool. I believe that you said that this bearing is on oil and that you check it often. Could you check it again? Could anything be stopped up where you see oil, but it isn't getting to the bearing? I could be wrong, but this sure has the looks of a lube problem. If the lube is adequate, I think I would watch it for a while. The waterfall doesn't show much change and the waveform Gs would have to get up higher before I would get too worried.

You may want to take a look at your Peakvue AP set. Yours is set for order tracking. We changed ours to frequency base because of what happened to you. Your AP set 17 is set to a 500 Hz highpass filter, but because of order tracking, the 894 cpm pushed you up to a 1,000 Hz highpass filter. You can see it in Plotdata on the right hand side under the time stamp. This is a dangerous thing. It can filter a fault completely away, or it can filter out a lot of the energy (Gs) and lead you to believe that a fault isn't as bad as it really is. You are also set to 800 lines and 4 averages. You may consider going to 1600 lines and 1 average.

It looks to me like you are all on top of this one. I don't think your fat will be in the fire.

David Eason