I am the other Alan! I have some bearings that turn in the 15 to 30 rpm range, they are a SKF 22228CC. When they the process is set at a certain speed it is steady without varying. (I aquire the speed before collecting) In order to take meaning full data that I can trend what is the recommended # of averages and L.O.R. that will not take too long to collect.

Regards,
Alan

You didn't mention whose equipment you're using so I'll assume CSI--since that's what I use. Nice, huh? Turning that slow, it's going to take a while to collect meaningful data. Can't escape the math. Definitely a PeakVue setup. 70X Fmax with a 500Hz filter, one average, and either 800 or 1600 LOR. You'll have to experiment to see if the extra lines are worth the extra time. Sample time, t(s)= LOR/Fmax(Hz) so @ 30 rpm (0.5 Hz) and 1600 LOR, t=1600/(70*0.5 Hz) = 46 seconds. That's 23 revs of the shaft--should be plenty.

PeakVue uses discrete fmax settings and in the example Patrick uses it will default to 3000 cpm fmax. I use those setting often when I need to take an extra look at a slow moving roll with excellent results.

I am new to slow speed moving bearings. I let RBM wizard set these points up. I decided to take one Standard point and one Peakvue point per bearing.The standard point is 400 LOR and 5 averages, the peakvue point is 800 LOR and 5 averages, the HP is 1000hz. Obviously it took some time to collect the points, is this time necessary? or should I alter the settings.

Regards,
Alan

You need to consider what you're looking for with this data.....

Are you interested in rotation rate to detect balance and misalignment problems? With a rotation rate of 0.5 Hz, balance probably isn't your main concern....at least not as detected by vibration analysis.

It sounds like bearings are the biggest concern. If this is the case, scrap the conventional measurement and rely on the Peakvue measurement. I'm not sure that 800 lines buys you much over 400 lines while it doubles the data acquisition time.

Regarding the conventional measurement....you need to be certain the input high-pass filter isn't set to 5 or 10 Hz, which would filter out the rotation rate signal. Also, acceleration may be the best measurement, as analog integrators will filter below 5 or 10 Hz.

Jon
Spintelligent Labs

Alan,

I would not trust RBMWizard to set the analysis parameters.

The hp filter should probably be the next highest over the fmax which in the case of slow speed readings will probably be 500 hz.

Dr. Robinson now recommends only one avarage for PeakVue and I doubt that the Wizard knows that. It appears to either have been written by programmers with limited vibration analysis experience or involve such difficult logic that it comes to some questionable setups.

Good Luck,

You would think that the wizard would be able to get it nearly correct - Dr Robinson has been recommending one average for peakvue for at least 5 years.

I would have thought that 800 lines is plenty.

If you have a dual channel data collecter, you can set up to take the readings simultaneously, which will save you a fair bit of collection time, depending on how many points you have to collect.

Has the square root of 2 been discarded? Where does 3, 5, 7 or 9 averages come from?

Jon

I am primarily interested in predicting bearing failure.

Since I am new to both SST, and Peakvue I would prefer to take both conventional and peakvue data until I get a handle on the latter.

Am I missing something in the replies, am I getting good data that takes too long,or is the data useless, and takes too long. It must be one or the other?

How many revolutions should occure before I have reliable data?

Please note my version of RBMware is 4 or 5 years old, so maybe the newer versions have more realistic set ups, which was overlooked on earlier versions

It seems strange that with all the science (one would assume) that is behind the formulation of the program that the recommended settings are not realistic.

Regards,
Alan

Alan,

You need to get 5-10 revolutions of the shaft and resolution good enough to see any anticipated sidebands. On slow stuff it will take a while. I used to do some Yankee rolls that took about 20 minutes each side. With only 1 average now recommended for PeakVue that saves time and also makes the calculations for the duration of the readings easier. Remember that your time is multiplied by the number of averages and that overlapping the averages will save time in averaging.

Setting up for collection with the 628 or 624 adapters using 2 channels is a god idea. Be careful when setting up for dual channel collection as there are many many rules governing what will be collected simultaneously and what will be collected sequentially. The rules are available on CSI's website.

No, the newer versions do no better setting up ap and al sets. They aren't all bad. They will get decent readings on a normal motor, fan or pump (not high enough resolution imo), but when you vary from normal, they don't know what to do. (Or at least they don't know what I would do.)

Good Luck,

quote:

Am I missing something in the replies, am I getting good data that takes too long,or is the data useless, and takes too long. It must be one or the other?

As far as I'm concerned, the problem with a lot of samples/averages is speed variation--whatever kind of reading you're taking. So you may be getting great data in four or five averages with PeakVue, but it's not necessary to take that much data--that's the beauty of it.

quote:

It seems strange that with all the science (one would assume) that is behind the formulation of the program that the recommended settings are not realistic.

A lot of hard work by a lot of smart people with a lot of experience went into those algorithms for APs and ALs. They're a good place to start if you have nothing else to go by. They still need adjustment, though, based on your conditions and experience with the equipment. Why PeakVue AP sets are set up with anything other than one average is beyond me, too. If you use the PeakVue Assistant in the FreqCalc program (I don't think it's in the early versions of RBMware), it always says 'Only One' average.

I monitored a machine that had many bearings that turned a top speed of 36 rpm and as low as maybe 20 rpm, depending on the fpm speed.

I used an Fmax (fixed) of 50 Hz with 800 lines and 2 averages(to cut down on time) and got what I considered to be very reliable data as far as bearing defects go.

Peakvue is a good tool, but as slow as the bearings turn and a gearbox or some other critical high speed piece of equipment are not involved, the normal data and the speed will give you plenty of time to make a call from a defect, BPFO, BPFI, FTF, and BSF, if checked on a monthly basis, ------now checking every 6 months or so, I would be leary of any setup used.

Using the SST function works great on this slow speed stuff also.

Time is the biggest issue on slow speed data collection. In my opinion, no data is better than data that is questionable, for whatever reason, be it improper setup or whatever, so if one doesn't have the time or patience to collect it, then that is another issue.

I would setup a dual point and collect a normal set and a PV set at the same time, with both having the same fixed 50 Hz Fmax and the PV Bandpass (not highpass) of 50 to 300 Hz.

Attached is a spectrum of a bearing defect using this setup.

Of course this is only my opinion and I could be wrong.

Ralph,

Why do you use the bandpass filter rather than the high pass filter?

I have many times wished for a high pass filter lower than 500 hz. I suppose that is one reason to use the band pass. What is the practical result of filtering below 50 hz?

Thanks,

Danny

quote:

Why do you use the bandpass filter rather than the high pass filter?

I have many times wished for a high pass filter lower than 500 hz. I suppose that is one reason to use the band pass. What is the practical result of filtering below 50 hz?

I am assured, or hope I am, that I cover what is going on from my 50 Hz Fmax and up to the 300 Hz limit. If I use a 500 Hz HP (which I like you wish there were a lower one), a problem might be missed that is occurring well below that high of a frequency. Only my opinion, so don't take it as set in concrete.

Ralph,

Makes sense to me. (That might not be a good thing.) I'll have to try it out.

Thanks,

Danny

If you're using a 500 HP filter aren't you dropping everything by at least 3dB below 500 Hz? Wouldn't only use a 500 HP filter on items of interest above 500 Hz. Am I reading this thing too fast.

Ralph,

Do you remember this conversation?

Any new experience with the BP filter.

I'm going to do some reactor bearings tomorrow that turn 6.7 rpm and might have to try it. I'm set up for a 500 hz right now and if I try the bp filter, I'll have to remove the accels, let them cool and remount them (20 feet up, 350 degrees) plus it is an 80 sec twf so I would like to get it right the first time if possible.

Edit: Read it all over again and I'm going to use the 20-150 hz bp filter.

This message has been edited. Last edited by: Danny Harvey,

Danny,

Have tried it a little since 2005. Have not had a 6.7 rpm shaft though.

Your 20-150 sounds good. What's your Fmax? Can't you set it up as a dual point and get the 500 HP and the BP at the same time?

Good Luck.

I understand that peakvue doesn't lose the energy that the normal fft does through the RMS process (energy under the curve), but I guess I am a little surprised that no one on this thread suggested looking at the raw acceleration time waveform.
I know how 'tricky' it is to get the collection right for peakvue because I see it being discussed here on the board, so how do I know, if that is all I am looking at that I haven't filtered out what I really want?

High pass filter 500 Hz - everything below 500 Hz is filtered - it allows high pass. Conversely, 500 Hz LP or low pass filter allows low pass but filters above 500 Hz.

If clipping isn't a problem and resolution is OK why do you want a filter. Can't you see your area of interest with adequate resolution?

You can buy inline LP filters to attenuate above a set frequency or HP filters to attenuate below a desired frequency point. What am I missing?