Good Day All, I can't place the large peak at 20750 Hz. I am thinking looseness with bearing fit but can't be sure. Bearings appear smooth and machine is running quiet as usual. This is a horizontal separator made by alfa laval...MX486
I would be greatfull for any help. I will not be able to add any files if needed until Monday but I will be checking this board all weekend. I dont want this machine to fail on my watch and I don't want to take it out too soon.
Thanks in advance to all that my shed some light on this for me
Good In = Good Out

10-4_alfa_laval.doc (239 Kb, 91 downloads) alfa laval

Does the unit have a gearbox? An initial thought is a gearmesh frequency that high in frequency.

Is the peak a narrow peak or a wide region? A bearing tone at that frequency is unlikely. Board-band, high frequency noise would be a good indication of bearing problems.

How is the accelerometer mounted? 20 kHz will be in the resonance range of most accelerometers, possibly increasing the apparent level of the peak by 30+ dB.

Jon
Spintelligent Labs

I'm distrustful of the data. The time domain shows almost 80g peak, yet the "bearings appear smooth and machine is running quiet as usual". I believe that kind of amplitude would saturate most of my accels. I would guess that a magnitude of that sort would be obvious to somebody looking at the equipment.

I'd recommend performing a sanity check of the spectra...verifying that expected frequencies are being represented (eg 1x of centrifuge bowl or drive motor speeds).

I am suspicious of the data too. Your regular 4500 data has 126 G-s and the "demand acquisition" data has 156 G-s. That much energy should be ready to launch something to the moon. The "walk-around" data has only 2-1/2 G-s but it was taken 10 months ago and the f-max is lower and something could have been filtered out. What does the trend show for the past 10 months on the 4500? What is the voltage on that sensor? Have you checked the sensor to make sure it is mounted properly? How was the "walk-around" data taken....did they tap into the 4500 sensor or use a seperate sensor? If seperate, was it placed differently than the permanent mounted sensor? I would also be curious what the "walk-around" shows with the same f-max as the 4500. Lastly, as someone else already mentioned, check the data sheet for your4500 sensor to see if the resonance falls in that 40,000 Hz f-max. Your waveform measured 156 G-s, but your sensor probably overloaded at 50 G-s. If that much energy is really there, you should be able to feel it.

David Eason

Hi Nsmkr,
On a similar machine I have seen an eddy current brake bearing failure with an inner ring defect of the NU roller bearing at 280 hz.
So perhaps you should focus on brake bearings as well!
Regards,
Arie Mol, NL

Thanks to all that have responded to my plea for help.
I have learned the following since my last post.
When the machine was initially set up for continuous monitoring we only looked at 120 orders max. I found the peak at the described frequency by using the demand acquisition feature in the online watch part of AMS Suite when this point reached an arbitrary limit. We have not established AP limits yet, do not know why at this point but will make those changes as well.

The machine does not have a brake on it but does have a hydraulic driven planetary gear box for speed differential. The sensors are high temp accelerometers and are screw mounted 90 degrees horizontal and vertical and are tight. Sensor resonant frequency is 18 kHz. Sensor voltage varies from 9.8 to 11. The cables are armored and show no signs of ware or pinch. The values are slightly less in the vertical plane but feel the chances of two sensors failing at the same time are remote at best.. From floor to peak and back to floor the range is 4000 HZ. I agree it should be orbiting Jupiter by now if the amplitudes are valid.

Trending to 120 orders for the past ten months shows insignificant changes in the spectrum and wave forms even though the over alls have escalated. I will change the AP set to look further out. I am inclined to agree with gear mesh frequency on the basis that the wave forms and spectra at 120 orders or less shows little change.

Thanks to all that responded to my question.

The sanity of the data was suspect. Turns out both accels were bad (overheated). On the hand held shaker the max amplitude at the given Hz was .36 g’s as opposed to 1.000. There was still a high amplitude peak at 20750 Hz after replacing the accels but had been reduced by 2/3. The gearbox was the culprit. When checking the oil there was large, blue from heat, pieces of metal stuck to the magnetic fill/drain plugs. This is where the heat came from that fried the accels. The oil in the box was clearly burnt. The gearbox is being repaired/replaced without loss of production schedules.
Good In = Good Out