Going back to an earlier subject of true peak:

quote:

Does this 15G(pk) fault level come from a standard or from experience? The reason I ask is that my primary reference is the GM Vibration Specification, which is one of the few freely available Specs that addresses something other than 1X amplitudes. I've had machines with 30G that are "perfectly normal" and ones with 1G that scare the heck out of me.

First as a clarification I think you're already aware - it is true peak from the TWF or a peak-detecting instrument, not an overall expressed in a peak/0 or pk/pk basis..

I can tell Chris that you are not yet a believer in the relevance of true peak. I am a believer. Here are some references and reasons it can be relevant helping to evaluate rolling element defects.

* 1 - CSI recommends it. The article referenced above - full citation is:
The DESCRIPTION OF PEAKVUE AND ILLUSTRATION OF ITS WIDE ARRAY OF APPLICATIONS IN FAULT DETECTION AND PROBLEM SEVERITY ASSESSMENT
James C. Robinson, CSI, an Emerson Electric Company and James E. Berry, P.E., Technical Associates of Charlotte, P.C. Emerson Process Management Reliability Conference 2001
October 22 - 25, 2001

* Another article "There’s Still Value in Overall Vibration Measurements" By John C. Johnson Balance Plus Wichita, Kansas Enteract 2000 conference: "I place a high value on true-peak acceleration measurement. I started using true-peak acceleration measurements after attending an advanced vibration analysis class in 1993 taught by Nelson Baxter. Talking to Nelson during a break, I learned one of the most valuable methods of detecting defective bearings that I had ever learned before or since. He told me to look at the acceleration time waveform and identify the highest amplitude peak, either negative or positive (this is truepeak acceleration). Amplitudes over 7 g’s for ball bearings and 12 g’s for roller bearings are a strong indicator the bearing is defective. I applied this over the next few years and found it to be very accurate on machines that do not produce a lot of high frequency energy during operation. Taking time waveforms with 1024 samples, the time sample must not be over 200mS (100mS is preferred for most machines). If the time sample is longer than 200mS, the anti-aliasing filters in the instrument will filter out the higher frequencies that defective bearings produce. This can become a problem on low-speed equipment because of the short time span for the time waveform compared to the long time span between the impacts caused by the rolling elements impacting bearing race defects."

I think you will find that using true peak to evaluate impact type events is widespread and not limited to rulling bearings:
* Mark's Standard Handbook for Engineers, 8th edition, (1978), page 5-75 ".Vibrations to be measured can be classified as either 1-Periodic, 2-shock or transient, or 3-random/statistical. Of these periodic motionis best understood and instruments for measuring frequency, amplitude [=displacement presumable], velocity, accelration, or waveslope have been developed. .In the measurement of shocks, these same quantities mentioned above may be of interest; however, in general, PEAK ACCELERATIONS [emphasis added] are of prime importance. In the case of random motions a frequency spectrum of mean-square values is desirable."
* Shipment of large expensive stuff - Before I worked on rotating equipment I worked on large transformers. When we shipped a Generator Stepup Transformer, we included an "impact recorded" which measured the peak acceleration in each of the three axes during shipment. I have Doble paper which recommends this approach for monitoring transformer shipments. Now working on motors, Westinghouse also uses the same approach for shipping reactor coolant pump motors. They are trying to determine the severity of a single impact during shipment and for that the most relevant parameter they use is true peak.
* There is an API document that mentions true peak for evaluating impact type events such as gearboxes and bearings. I don't remember the reference. (anyone got it?)

I have been looking for a math explanation why true peak is most relevant. I don't really have one (I would love to hear one if anyone has an explanation).. But step back and think about it using Buzz' KISS principle. You have periodic impacts that last a very short time and you want to try to quantify how "hard" a given impact is. (not that pattern of the impacts, that's a different question also important to severity). So to find out about one impact, are you going to look at the spectrum? No, the spectrum averages over time (averages the empty time between impacts) and even the portion that remains still gets spread out among multiple frequency peaks. So you turn to the TWF. How do you put a number on a TWF? Peak is the best one I think if you want to find out about the impact itself. The ringdown period is more a reflection of the system being impacted (although certainly the system affects the peak as well.... nothing is perfect.). rms wouldn't make sense... it averages.

Another advantage of true peak is you can share/compare your results to others using different equipment (other than the minor squabbles about how close are you to true peak).

So if we are convinced true peak is a relevant parameter, the next question is what values should we be concerned about. You have already a bunch of recommendations above: from Buzz, from CSI, from the John Johnson article. They are all not too far apart. Use a factor of 2 (or maybe 1.5) to do rough conversion between true peak and true pk/pk. I think among all of those Buzz' are the higest limits and I certainly agree at 15g's pk/0 (30 g's pk/pk) is severe for a rolling element defect. I think if you got to 6 g's pk/0 you should feel confident there will at least be visible defect on the race (again that's not the only indicator of severity).

2 key points to remember
1 - This parameter (true peak accel) applies to impacts known to be coming from a rolling element defect. If you're lucky you can use your sideband cursor on the TWF to identify the peaks you're looking at are occuring at a nonsyncronous fault frequency. Then you can apply the severity criteria. If you were looking at a reciprocating machine and seeing imapcts at 1x, you would have no business applying these criteria...it's not an impact from a rolling element bearing. If the 30g's you mention is coming from a rolling bearing defect you darned well better be concerned.
2 - True peak only looks at characteristics of one single impact. There are other indicators of the pattern of all the impacts taken together (i.e. spectrum) and trend discussed above also very important.

BTW Chris - I hope you do not view it as a negative that I provided so many comments on this thread. You have helped me on a lot of my questions and I wanted to try to provide a detialed answer from my perpective (fwiw) to return the favor.

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

Pete,
I am not familiar with true peak other than what I have read on this thread. In the article you posted it recommends no more than 200 mS in the time waveform and 100 mS is perferred. What would be the setup for this, 400 LOR/2000 Hz fmax for 200 mS using 1024 samples? How many averages would you take? And am I following you right, the highest peak multiplied by two for true pk to pk? I am interested in trying this and would appreciate some advise on setting up the proper parameters.

Thanks,
Ronnie

Hi Ronnie
I'm not that good at setting up sample collection parameters.

I agree Johnson's stuff about setup was a little weird.

There are 6 interrelated paramters: Fmax, sample rate (=Fmax*2.56), number of samples, duration, bin width, and lines. If you pick any two of these the other 4 are already defined.

As far as Johnson's stuff - 1024 sample points corresponds to 400 lines. Apparently duration was the other thing he was allowed to pick and he didn't want to pick it too long or else his sample rate would be too low. But his formula may not give enough revolutions especially on lower speed machines. More lines helps you to get both: high Fmax (high sample rate) and higher sample duration. His equipment may have been limited on lines. What was he talking about regarding anti-aliasing? I have no idea.

Here's how I would set it up. First remember the two important things to get for a true peak measurement are a high sample rate and long enough duration to capture a representative number of impacts. The CSI article I posted above recommends at least 10 revolutions). On a 3600rpm machine that would around 0.17 seconds. So a good combination would be Fmax=5khz, Lines = 1600. That would give duration =1600/5khz ~ 0.3 sec.

That same setup gives 0.3 seconds which is probably good enough for 1800rpm machine (almost 10 revs). At lower speeds you might want to bump up the number of lines to make sure you get 10 revs.

Number of averages does not matter for a TWF measurement. It collects only one record.

Someone correct me if I'm wrong above.

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

Pete,

Never had a negative vibe from you, I hope likewise it true. This is a learning experience for me, and I hope the same for others reading. The comment from Ronnie indicates that we are actually learning from each other. This is the part about these forums that I enjoy. So...

I did a search on CSI's site and found the document by Robinson and Berry.

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

Curiously, Mr. Berry's name is no longer on the document. However, you can purchase the document from Mr. Berry's company for $99.00 ... ummm ... I smell trouble so I am just going to leave it alone.

I have read some notes from Nelson Baxter (either class or seminar, not sure). I think Nelson is working more along the lines of "absolute" peak rather than "true" peak. The 7g's for ball bearings and 12g's for roller bearings is an interesting idea. Obviously from a marketing perspective the term "absolute peak" would never fly. Sounds more like a reference to Vodka than to vibration.

Mark's SHE from 1978??? You do know they made more editions after #8 right?


All,

There are many "new, improved, wonderful" measurement techniques and arguing who's does what better and whose is true, is really useless in the face of one question: When is a bearing really bad???

My point is that it doesn't matter if a signal processing technique or measurement parameter can detect a defect 10 days or 10 years in advance. What I am looking for is something that will measure remaining bearing life (in days, weeks, months). If I used Nelson's 7g reference, then it would be a matter of projecting when this bearing will cross that amplitude. However, what if it fails before that amplitude or still has viable life after that amplitude? What if it hits 7g's as projected, but there are still 3 months before shutdown?

Machinery vibration analysis should have a direct correlation to the component being monitoring. In other words, how much time is left? What is the status of the bearing?

I believe very much in the value of time data.
time data
The above link is acceleration from a punch press. If you look closely at the negative peaks, you will actually see two peaks. The first (smaller) peak is when the tooling contacts the part. The positive climb was when the main shaft shifts from the top of the journal to the bottom of the journal. The second (larger) negative peak was when the part was actually punched. Using this data technique, I was able to correlate the positive climb with the main bearing clearances. Most of the presses were different sizes and had different bearings, which required a different calibration factor. However, there were four groups of presses that were identical and the calibration factors were nearly the same (within 0.5%).

It would be nice to be able to tell a customer that this being has a flaw in the raceway that is 1 millimeter deep and they can expect no more than 45 days life on it.

What do you think?

This message has been edited. Last edited by: Chris Olsen,

I bought my Marks' used. I have a low-budget library of mechanical textbooks. Along with Mark's ME Handbook I have Den Hartog's Mechnical Vibrations, Fieldings Handlheld Calcuator Programs for Rotating Equipment Design, Harris' Shock and Vib Handbook. All were under $20 and all except Den Hartog were used. I'd recommend any of them although Fielding's book would be at the bottom of the list. (if anyone wants to recommend any other good cheap textbooks let me know)

In routine monitoring, there are two opposite constraints on when to call a bearing:

#1 - Don't call it so early that you won't see visible damage on the races when it's pulled out and cut apart for inspection. Then everyone loses confidence in the vib guy.
#2 - Don't call it so late that it fails or that it takes a nasty turn for the worse that makes everyone nervous enough to require jump-up maintenance.

For #1, I think the true peak is one of the more useful tools.
For #2, the trend and the pattern are more important.

There was a previous thread talking about #1
http://maintenanceforums.com/eve/forums/a/tpc/f/375...011/m/3751074441/p/1

Your question I think is similar to #2 (when will it fail). A lot tougher call to make.

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

Pete,
I agree with the importance of taking enough time to see enough revolutions to capture an event such as a bearing defect. I guess that is where the confusion lies for me. In order to accomplish this, you will most likely defeat the 200 mS mentioned in the article due to anti-aliasing because of too low of sampling rate. I can manipulate the time by adjusting the LOR/fmax but how can I stay in the parameters suggested and get useful data? And which is more important, enough revolutions or stay below 200 mS?
Pete, I am not trying to hold your feet to the fire so if anyone reads this and can enlighten me, please do. Or should I stick with PeakVue which I am just starting to understand, sorta kind of?

Thanks,
Ronnie

Getting enough revolutions is important.

Staying below 200 msec is not important at all.

I believe the reason it was important for John Johnson's specific instrument was that you apparently selected the duration and the lines. With lines fixed at 400 he had to keep duration low to keep Fmax high.

So in summary the two important things are long enough sample and high enough Fmax. On most instruments you control Fmax directly and pick lines so that the sample is long enough.

Thanks Pete. I'll start looking at true peak and see if I can't make it another useful tool.

Thanks,
Ronnie

From what I saw the bearing is not in jeopardy of failing soon. On this one-shot call deal, with a good lube program it may or could make it until Oct. You have defects; how bad are they? I had one glance and without going back to look again on limited data I'd say category 2 bearing flaws. Regardless, I don't see the bearing as failing quickly. Typically over 2 kHz one will worry on the PW Brg at 7 g's. Expect failure at 12 g's 0-P. How often do you do a route monitor? I may have missed running speed but assumption of 1200 rpm was basis. Faster and you'll accelerate there quicker. Monitor closely so you can do a "I told you so" thing.

Recently called a bearing for replacement at 12 g's and they didn't do it and now it's two weeks past due for failure, 18 g's and 35 swing on TWF and will numb you finger at the touch and sounds really bad but the guy said it let it blow. But I have it documented. Last week he replaced 2 motor I said were good (less then 2 g's); so smile and say, "did you hear the one about the farmers daughter and >>>>...??? Smile.

I agree. For the most part it looks like outer race fault and nothing else. I saw a few unidentified frequencies in the acceleration spectrum but those didn't show up in the demod. No 1x sidebands around bfpo harmonics - that's good.

The waveform is 3g's pk/pk. CSI alert would be 6g's pk/pk and alarm 12 g's pk/pk for outer race fault on 1200rpm machine.

If it were a non-critical machine, we probably wouldn't identify it for repair yet at our plant (just continue to monitor). If it were a critical machine, it would be a tougher call. Either monitor a little closer or to be on the safe side identify it for repair at the next routine opportunity. There might be very little visible damage on the outer race. Since the pattern started at installations, Duncan's guess of true brinneling damage during installation seems as good as any.

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

Not directly related to the topic of this post, but a question I am wondering about on your displayed spectrums.

Is there an advantage or disadvantage of having the vertical scale raised off the frequency scale floor? I always have had mine set at zero. Just wondering. I think commtest comes defaulted to this setting.

Chris
If you are running an in house program then you have but one customer
and he sees you daily. Right? Things you know or should know to make this call
are 1) bearing loads 2)bearing speed 3) rated speed for the type
of lube 4) the impact it could have on employees, the environment and/or production.
This is all you need to make the call.

I have watched defect like this one for 2 plus years in equipment but I also have
Equipment I would call this a take action in 60 days and others 30 days.

It is not always about more toys to do the job but what does your experience
tell you.

"When is a bearing really bad" may can be compared to the question "When is a woman really beautiful" ?

I guess it is in the eyes of the beholder, in this case it is in the eye of the customer.

Many want to see definate bearing problems, some to the extinct of rollers about to fall out while others realize the bearing may look fine until broken down where inner race and roller spallding can be observed.

We try to call bearings that need to be repaired on the upcoming shutdown, because they may not run thru the upcoming shutdown until the following schduled shutdown 6-8 weeks later.

We try to put our report out 2 weeks prior to shutdown date to allow for planning, this is challenging us to call bearings that may not run another 10 weeks.

The biggest plus to me is getting the bearings taken out during the shutdown, breaking each one down, taking pictures and sending out a followup report on items worked on and bearing condition as found.
Helping to keep unscheduled downtime of the machine to or below a minimumn of 1% yearly, we currently get no harrassment from management for our calls

Mike

Pete - Thanks for your comments, I am going to experiment more (always playing with things)

Sam - So the bearing is bad, but not bad enough to change immediately in your opinion. (I agree)
I undertand about the customer issues. I've had some like that in the past. Very Frustrating.

Ralph - I like the raised floor (CommTest default) because I have several machines that require me to be looking "down in the mud" for bearing frequencies. I have used log scale with other software to see these in the past, but I prefer linear scale.

MikeH - In your case early detection has a high importance, this is where specialized measurement techniques have their greatest value. As I understand your post, your definition of a defective bearing is one that flaws visually noticeable upon bearing disassembly. (correct?) But if you detect a bearing problem, such as this early stage defect, is it "bad"?

Barry - Sixteen-Eleven provides customized services which are controlled by customers. The customers are allowed to pick and choose various technology options. In this instance, the customer has specified a semi-annual measurement with a velocity FFT at 400 lines and 24kCPM Fmax. The customer does not feel that a 1200 RPM FD Fan is of high enough importance to have more frequent monitoring (his call). Of course, I am in the facility on a monthly basis and tend to be a bit more curious than I should be. I pass by this particular machine while performing a monthly vibration route (and take GOOD monthly data on it - hehe)

Your comment about how experience is a factor in making the call, is very true. I only have 15 years in the consultant seat, but learn new things all the time. I guess that is why I love this work!

To All,

Here is the fun part ... the customer's report contains the velocity FFT as first presented (with parameters specified by customer). He is now looking at a confirmed bearing frequency on a fan that will shutdown his plant if the bearing fails (FD Fans are sort of important to power plants). The comment in his report states "confirmed bearing frequency". The plant can not come offline until October, as they are under contract and will pay some big penalties. The customer is VERY nervous and the fact that his boss (who brought me in 10 years ago) also gets a copy of the report does not help matters any. I KNOW he is just squirming about this issue (from conversations with his mechanics).

Our conversation this morning went like this ...
"Hey Chris, I was thinking maybe you should start monitoring the FD Fan on a monthly basis."
To which I replied that it would be no problem and that I won't even charge him extra for it. I think he was expecting an "I told you so", but it never came. (I recommended monthly monitoring originally)

Taking the extra effort to include customers in the PdM process and by being "overly curious" about vibration (syptoms and causes) makes our approach slightly different than the "cookie cutter" programs that other companies sell. I know we're not unique on this, as most of the other smaller, regional operations are also very customer focused. (Most of them also post here)

Just for the record - I have previous failure data on this machine, this bearing in fact. When the current frequencies and amplitudes were present last time, the bearing was run another 18 months without any problems. False brinelling was confirmed on several bearings from the same stockpile and this issue has been addressed as a "reliability improvement". They have reduced parts inventory levels and improved storage conditions for replacement parts.

I still haven't decided when to let this guy off the hook and tell him things will be OK. I've already had a conversation with his boss and he told me to "just leave him dangling".

Take Care,

Plotting on a dB or log vertical axis scale is usually a more useful way to look down in the mud as it allows you to a better way to assess the height of a spectral line relative to the noise floor of the measurement. The noise floor of a normal spectrum is usually set by the resultant of all the non-linear forces that happen to the machine.

In our software which predicts a minimum safe operating period, the time for the next measurement modifies/reduces the scheduled time for the next measurement. The next scheduled date this gets closer than you are comfortable with or less than the next scheduled down date, assuming you don't want to have shutdowns outside of the normal schedules, you do the repair then - condition based maintenance.

VENDOR WARNING

quote:

Originally posted by Chris Olsen:
MikeH - In your case early detection has a high importance, this is where specialized measurement techniques have their greatest value. As I understand your post, your definition of a defective bearing is one that flaws visually noticeable upon bearing disassembly. (correct?) But if you detect a bearing problem, such as this early stage defect, is it "bad"? Take Care,

Chris, your understanding is not correct.
If you read my post correctly, you would recall where I said, our challenge is to report bearings that need to be replaced on the upcoming shutdown, because they may not run thru the shutdown up to the following scheduled shutdown.

If I was to call a bearing for repair and inspection only showed signs of watermarks, I would clearly have reported the bearing for repair too soon.

We list priorities as ( 1 ) - repair on upcoming shutdown, ( 2 ) repair as time permits, ( 3 ) those with early defects present, currently being monitored more frequent than routine shutdown surveys.

There is actually no answer to your question, when is a bearing really bad. but the question I would ask is.. when will a bearing risk damage to equipment, life or risk unexcepted downtime.

I have seen bearings totally locked up, but the equipment still running, with the shaft rotating on the packing or inner race of a locked up bearing.

Is this bearing bad? well yes.. but is the equipment still running? yes again.

My question would be when a bearing condition causes risk to elements one wishes to obstain from.. whether it is further damage to components, damage to life, health, etc or risk downtime of production.

my 2 cents.

Mike

mike - that's a very practical and reasonable 2 cents.

Lockup often results in conditions where the bearing/housing is thermally unstable. I've seen locked-up oil lubricated bearings that ran ok as long as the bearing still had plenty of oil but once the oil got low, it got very hot and seized. I saw one of these on my first consulting job some thiry-five years ago. I was hired to investigate another problem, a press roll with a bad cover, but the millwrights couldn't resist pointing it out. They also showed me a felt roll that ran with no rolling elements left inside the bearing - just right dynamics! I've also seen locked up bearings that ran with slipping inner races - these usually result in shaft replacement.

As I mentioned earlier, our software forecasts a minimum safe operating period that is also used to schedule the next diagnostic measurement. When the forecast minimum safe operating period drops to less than one day, a REPLACE THE BEARING recommendation is issued automatically - that's our definition of failure. If you choose to replace it earlier for other reasons such as the ones you listed, that's your choice. We give you the info to help make that choice.

DEFINITE, EMPHATIC, VENDOR WARNING ;-))

This message has been edited. Last edited by: Duncan Carter,

Mike,

Your two cents is worth much more! I think you have indeed answered my question: When is a bearing bad?

quote:

when a bearing condition causes risk to elements one wishes to obstain from.. whether it is further damage to components, damage to life, health, etc or risk downtime of production

Because our instruments and software are getting better all the time, we are going to be able to detect bearing "problems" that are not yet at the point where they fit the definition. It is a common mistake by inexperienced vibration people to report a "bad bearing" simply because they are able to detect a bearing frequency. It is also something that I have seen people do who are, shall we say, "motivated" to call bad bearings so that their repair shop stays busy. (This is a sore spot for me, as I see too often, but this is another issue)

For what it's worth, I worked with an engineer that used the phrase "self corrective conversion" when refering to bearings running on the inner race.

Take Care,

Chris
Sorry to wast time on this board but I assumed from your post you were in working in an Inhouse program. Ance again I apologize if
I had know I would not have wasted your time or
mine.

quote:

Chris
Sorry to wast time on this board but I assumed from your post you were in working in an Inhouse program. Ance again I apologize if
I had know I would not have wasted your time or
mine.

Barry Crawford

Barry,

This forum is a place where people share knowledge and learn things. My time here is never wasted, I am sorry if you felt that I wasted your time on this discussion.

I am not sure how being a consultant versus an in-house prgram applies. Aren't we all people that enjoy working with vibration, balancing & alignment?

Sorry again for wasting your time.

Take Care,