VENDOR WARNING

In my company's software, severity is principally described as a time parameter, a minimum safe operating period. Rather than using a fixed measurement interval, we can create measurement routes that only have points with expired minimum safe operating periods. Machines in poor condition get measured more frequently than machines that are in good condition. This can allow "just-in-time" repair. It also offers the option of making a list of all the machines that have minimum safe operating periods less than the next convenient repar date. Safe operating periods can range from up to 200 days or 20% of the MTBF down to one day; the default is 100 days for machines with no faults. If the calculated safe operating period is less than 1 day, the software recommends repairing the machine. Of course, you can choose to make repairs if the safe operation period is greater.

pete - more on your questions.

Crest factor is not an absolute number as it's very dependent on how you make the measurement - bandwidth, etc. plus the repetition rate of the most significant flaw or flaws. Like anything else that not absolute, you can only realistically use it as a guide. In my opinion, the best single tool for assessing severity is the envelope spectrum.

In my opinion, the need to sample at some very high frequency is overstated, especially by one vendor that I won't mention. ;-))

The ringdown rate is determined by the characteristics of the machine components and that generally implies a much lower frequency than you might otherwise expect.

The problem in using peak acceleration is that it is very path loss dependent. Look at the curves in the following link and think about how that might affect your results.

http://vibrotek.com/articles/factors/index.htm

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

To get an idea about one problem with depending on crest factor, look at Figure 3 in:

http://vibrotek.com/articles/sv95/part1/index.htm

Maximum crest factor occurs with a single flaw that is much larger than the background signal. If the number of similar flaws increases, the crest factor decreases.

Pete, In my world trending for bearing faults are the most required prerequisit and it´s rarely given to me, it is difficult to do any better judgement than change bearing if all kinds of parameters indicate that a failure is happening in short term eg. they are sky hi. I support Duncan in the statement that Crest factor and by that other parameters do go down as bearing destruction get worse. If you trend CF you may use it as early warning. I have had good results from PV anyway but not in all cases and I have rarely tried to or had a chance to trend it. Olov

So... The peak acceleration roughly represents the worst individual defect of all defects present (more later**). We divide it by rms to get CF. Add more defects without making the worst one worse and rms goes up, peak stays same, CF goes down.

From that standpoint, peak accelration is better than CF. Also the high frequency attentuation that affects peak but doesn't significantly affect rms will affect CF just as badly as it affects twf peak.

So based on all the above I still think peak accel is better than CF.

** We admit the worst individual defect does not necessarily give highest peak acceleration and peak acceleration can go down... perhaps as some have suggested before the sharp edges get rolled over smooth.

So Peak accel twf certainly has it's limits. So the rule that would be suggested is either/or. Call the bearing if either
1 - pk/pk twf g's exceeds alarm
OR
2 - pattern indicates multiple defects or worsening pattern.

More thoughts or comments?

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

quote:

Originally posted by electricpete:
So... The peak acceleration roughly represents the worst individual defect of all defects present (more later**). We divide it by rms to get CF. Add more defects without making the worst one worse and rms goes up, peak stays same, CF goes down.

A small comment:

The CF makes it possible to highlight the bearings in phase of damage. But when this CF is small, it’s better to make sure that the rms value (or the peak value) of the amplitude is not in progress, which would be synonymous with a deterioration of the bearing in terminal phase.

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