I was wondering if anybody could help me with the vibration monitoring of the slow speed bearings in our reduction gearboxes. The input speed is about 3000 RPM but the output is much slower some as low as 50 RPM. The gearboxes are on large chemical reactors I am currently only measuring fans, motors and pumps and have no experience of Gears. Should i measure Velocity or acceleration and what frequency range should I use and what about gearmesh measurement.
Any help you could give me would be great I am relatively new to Predictive maintenance and the equipment I use is about 6 years old.

Thanx.

what type of gearbox is this? Straight spur gears or epicyclic type gearbox? At 3000 RPM you will still get good information at the input side of the gearbox. I would suggest taking both acceleration and velocity measurements at this point. The velocity measurement upto 1-2kHz, the acceleration measurement should capture harmonics of the gearmesh frequencies, possibly 5-10 kHz. If you wish to measure the driven component, then there are a range of alternatives for monitoring slow rotating equipment. If you wish to remain portable, then piezo resistive accelerometers measure to DC (0 RPM). Hope this helps for starters. Steve

Hello Chemic,
I have spent a lot of time monitoring gearboxes with output speeds as low as 2rpm and some machines with even slower speeds than that.
There are many questions, but probably the first would be regarding the stability of the speed and load of these gearboxes - frequently in reactors both of these change during the process.
Secondly, is your interest in these gearboxes due to some failure history?
Best Regards,
Tom Murphy
tom@reliabilityteam.com

If you are using CSI technology, I would suggest using Peakvue. I have had great success with this on low speed equipment. If you dont have peakvue, try high frequency enveloping,or high freq. demodulation, whichever you may have. Set these up with a 400Hz fmax and a 500Hz High pass filter, 1600LOR. Take these readings in both axial and radial directions as most gearboxes have a Timken type bearings on the low speed shaft. That will increase your chances of seeing what you are looking for.

Tom
I am aware that vibration spectras change dramatically under differing loads therefore i try and measure the gearboxes at the same stage in the process (full load) every time. A number of our gearboxes (mainly the Plenty ones) are blighted by excessive noise and worn crown wheels.

Uk Vibeguy
Thanks for the information. I do get good info from the input measuring velocity at 10-1000hz and also measuring the gearmesh frequencies (acceleration at10-10000hz)but the slow speed stuff is unreadable. I use an old PL33 meter and polymanager software.I am, apart from a 2 day SKF course, self taught so any information on point measurement from experienced analysts is greatly appreciated.

Hi Chemic: I just came across your message. Did you get all the info you needed or are you still looking for input? John

Chemic,

Don't forget the value of waveform analysis. Periodic events from bearing faults can often be seen if you select the appropriate Time-max (often set by Frequency-max) and resolution (usually set by lines of frequency resolution). Setting a high-pass filter can clean up the waveform by removing low frequency content.

Walt

John
Thanks for your post. Things were alittle clearer from the other guys replies but the more information I could get the better.

Chemic

Walt
Do you mean analyse the vibration in the time domain rather than frequency?

Chemic

Chemic,

Yes, Waveform Analysis is a Time Domain process.

Walt

Hi Chemic,

I have been monitoring extremely slow speed bearings successfully for many years and I can agree with most of the posts above.

You need to take the following readings on the slow speed bearings:
1) Velocity to capture at least 10x BPIR. You should consider how many averages to take - especially if the speed tends to wander during the measurement. More speed variation means fewer averages unless you use PEAKHOLD averaging.
2) You MUST take a demodulation/enveloping type of reading. Different manufacturers use some different names such as PEAKVUE, SSE, ENVELOPING, DEMODULATION etc. The secret is to use a filter that reflects the natural frequency of the bearing. If this is a fairly small bearing that means between 1 and 4 kHz. If it is a big bearing that could be as low as 500Hz. Keep the spectrum display Fmax frequency as close as possbile to the velocity Fmax.
3) Take a time Waveform but only in G ACCELERATION. That will clearly show the indivdual impacts if the bearing starts to spall. Keep the duration down so that you see about 3 revs of the shaft and the sampling rate should be high enough so that you see the impacts (i.e. above the natural frequency of the bearing).

If you do all of that you can use the following rule of thumb:

1) If there are no spikes in velocity and no spikes in envelope and no increase in carpet levels then the bearing is OK.
2) If you see a rise in carpet level in envelope but without spikes you have lubrication problems.
3) If you see spikes in envelope AT THE BEARING DEFECT FREQUENCY but no spikes at the same frequency in velocity then you have marking of the bearing.
3) If you have spikes in velocity AND in envelope at the bearing defect frequencies then you have spalling of the bearing.

Good luck and let us know how you get on.

Ron Frend

Dear Chemic,

As vibration monitoring is base on acceleration, velocity and displacement, normally only one/two data is needed from this 3 data.
For high speed (3000rpm) normally velocity and acceleration used for analysis.
For low speed (50rpm) normally only displacement used in judging the bearing condition.
If you have a FFT analysis vibration measurement device, its good enough for monitor the machine condition.

Ron
Thanks for the excellent advice. Sometimes all the measurement types for different software becomes a little confusing but you translated them into my language. I will set up some routines this weekend and let you know how I get on. Thanks again.

Chemic

Ron is right.
An easy way to find the correct filter settings for envelope, peakview etc, is to take a 0-20kHz acceleration and set the filters to include the hump, because this is where the resonances are, and the resonances carry the modulation signals.
Byron

In general Byron is correct. You do not want to filter out bearing component resonance modulated by bearing frequency fault. At the same time you do not want GMF or bearing fault harmonic to pass through HP filter ( I mean here demodulation methods). This sometimes becomes tricky. Also, if a bearing has no fault you won't see the hump in 0-20 khz range at all. In other words you want to play with different Fmax and HP filter if demodulation method is used.

In general, for a GB you want to set up three types of measurement. One will be designed to see 3x GMF in the spectrum, the other - high res in the TWF for impacts, and third -for slow shafts/bearings (as pointed out in the above posts, demodulation will be the best).

David

For vibration monitoring of bearings you can measure using low frequency accelerometer with velocity as unit of measurement.
Fmax is the maximum frequency to be scanned by the instrument in CPM and for antifriction bearings it should be 60 X rpm & for gears it should be 3.25 X Gear Mesh Frequency.
For condition of antifirction bearings, you should measure spike energy (gSE) and is independent of speed.Permissible value is below 1.00.If higher than 1.00, go for replacement of bearing irrespective of vibration has increased or not.

I am taking measurements in a gearbox of a milling machine which output speed is around 60 RPM. In order to capture the Gear Meshing frequency I calculated the RPM (output shaft) and the #teeth, GMF=RPM * #teeth, to select the maximum frequency I considered this Fmax = GMF * 4/60 (Hz) (4 = number of harmonics that i want to see), to the resolution I considered this resol=(RPM/4)/60 (Hz), with these parameters I am working with a accelerometer which have a sensitivity 4mV/g and 2-5000Hz to linear range frequency. I hope It can help you.

Some of the info here is of great help.
However i need some additional advice. The equipment i am looking at runs at 4rpm for 13 revolutions and then stops. I have used peakvue and shock pulse to some varying sucess, however i was wondering about accelerometer selection for peakvue and this speed. Many accels roll off at 2.5 hz, so missing most of the low frequency stuff, has anybody used DC accels that start from 0 or lower roll off accels,. Any advice greatly appreciated.

Jason

Jason,
for my money, I would not even bother starting with vibration and just go straight to ultrasound. I also feel that for 13 revolutions, you are really close to the limit of finding anything in just one cycle. You might need to consider combining events to give you more data.
If I was using a really good LF accel and a suitable LF instrument I would be looking at approx. 20 minutes of continuous data to give me a good answer at 4rpm. Hence my suggestion to go straight to ultrasound.
If you need any further info, let me know.
Best Regards,
Tom Murphy
tom@reliabilityteam.com