A lot of discussion has taken place on this forum about filter settings for Peakvue so I have an example I just wanted to share. I took several readings on both fan bearings and the one I am posting is next to the fan, overhung belt driven. The speed is marked but do not know bearing numbers. They are pillow block ball bearings. Both bearings were so bad you could hear them without a screw driver Anyway the part I found interesting is the amplitude in the two Peakvue readings. The first is 1000 Hz Fmax 1000 HP filter and the second is 1000 Hz Fmax 2000 HP filter. The amplitudes in the first waveform are almost tripple that of the second. All three of these readings were taken on the same bearing without moving the accelerometer. I Know I can't give a technical explanation but I think it has something to do with the 2000 HP filter filtered out the energy between 1000 and 2000 Hz. And obviously from the spectrum, a lot was going on in that range.
I guess if there is a question or point in all of this is the importance of the filter setting based on the fmax?
Just thought some would find this as interesting as I did.

Thanks,
Ronnie

Sample_with_different_HP_filters.doc (236 Kb, 58 downloads)

As the defect reaches the final stages you have a lot lower frequency energy generated. just like in regular velocity data the fundamental bearing defect frequency doesn't usually show up until towards the end of the bearing life. Yes you have filtered out a lot of the energy by using a 2000HP filter in the later stages of bearing failure. The size of the bearing can have an effect on which filter can make you have different amplitudes in the waveform. A bearing with large rollers will emit lower frequency energy than one with smaller rollers, Dr. Robinson and Jim Berry's paper on PeakVue explains why this occurs. This is the reason unless I have something in particular that I am trying to filter out (high gearmesh for example) I always use as low filter setting as I can. Still don't understand Skip's 6X fmax comment.

How were you mounting the sensor? A two pole magnet on a rough or painted surface will not pass the energy to the sensor. Also, what size (about) was the rollers?

I was using a two pole magnet. The bearing housings were still in good condition, no added paint, corrosion or buildup. The magnets were setting firmly on the side of the bearing and not moved between any of the three readings posted. I guess from the recent posts about filter settings, equal to or some mutilple above fmax, had me curious so I tried some experimenting. I think the difference between the two PeakVue readings is in the filter settings but then I am still a novice at this PeakVue stuff and could be misinterpreting it. I certainly appreciate the feed back.

Thank You,
Ronnie

PS
The shaft size was about two inches in diameter and the bearings were single row. That should give you some idea of the bearing size. Sorry I can't be more specific than that but if it is important to anyone, I will make a phone call and get the bearing info. This was such an easy call there didn't seem to be any need for further investigation.

Thanks,
Ronnie

Jim Crowe--

Would this (increased low frequency amplitudes and decreased high frequency amplitudes in advanced stages of bearing degeneration) be the reason that PeakVue/HFE amplitudes drop at this point in remaining bearing life (around the same time you can audibly hear it)?

This would seem to make sense to me, at least intuitively...

Thanks--

Tony

Tony,
If I have been watching a bearing defect trend up in PeakVue and then the waveform amplitude drops, that is when I start getting worried. Typically this is when you can see it in the regular velocity data (I have seen the fundamental bearing defect with harmonics and sometimes looseness at this time in velocity). When using a 2 pole magnet you want to be careful when choosing the HP filter. If you choose to high of filter you could be trying to get information that the sensor is incapable of getting or at the very least have a lot lower waveform amplitudes than is really there.

Thanks, Jim.

I've seen the waveform amplitudes drop off, also, at the "end of it's life," but I never really associated it with the "moving down" of the frequency peaks. It only makes sense.

Tony

Jim
Do you that is possibly why my amplitudes were different in the two Peakvue waveforms? It makes sense that a two pole magnet would not work as well at higher frequencies. But is also seems to make sense that some of the energy was filtered out by the different PeakVue settings based on the energy that was present in the spectrum between 1000 and 2000 Hz.
That is a good point about the waveform amplitude dropping after an increased trend in the PeakVue. I will definitely be watching for that one.

Thanks,
Ronnie

The data below was taken with a 2 pole magnet. On suspected defect.

Ralph,
It looks like the defect was generating most of the energy near 10Khz. The next step up it really dropped off. What amplitudes did the waveforms show? I'm not saying you can't pick up defects with a 2 pole mag for a lot of faults. Did you take the same data using a flat magnet on a good surface to see how the waveform amplitudes compare? We had a problem a couple years ago with some grease we had changed to and I took a lot of data. Stud mount, flat magnet mount, 2 pole mag mount, 100 Mv/g, 10 Mv/g. and found a great difference when the high frequency generated by poor lubrication was measured using the 2 pole mag. If you are looking for regular defects like bearings a 2 pole is ok, but if you are watching for early lubrication problems you won't get it with one. Probably the best advice I can give using PeakVue data is consistency in location and mounting. Even if the amplitudes are lower you still rely a lot on the trend of the peak waveform amplitude. Just make sure you know what defect you have in the waveform. I have had an outer race defect in the spectrum with very high g's in the waveform, autocorrelated the waveform and found that only about 10% of the energy was from the bearing defect, the rest was lubrication. After the bearing was greased you could still see the bearing defect in the spectrum but the waveform amplitude dropped to a reasonable number.