I recently took data on two identical 37 kW, 2-pole vertical motors. Motor ‘B’ has a history of repeated bearing failures over the last couple of years. Overall vibration on both motors is quite low. However, Motor ‘B’ has a series of peaks in the high frequency zone (above 1800 Hz). Demodulated data too shows a series of 1X multiples. Two questons:

1.If this is running speed modulation of the bearing tones, shouldn’t the bearing tones too appear in the demodulated spectra? No sign of them. Any other cause for the 1X harmonics in the demodulated data?

2.Next, there are some strange sidebands at 7.6 Hz in the demod spectrum. These are not sidebands of the 1X harmonics though. I cannot relate this to anything in the motor. Any suggestions?

Motor ‘A’ does not show these patterns. I have attached the demodulated & velocity spectra for both motors. Any inputs would be most valued.

Regards,

Aditya

Motor_Data.doc (143 Kb, 58 downloads)

Blue and red are two perpendicular horizontal directions?

If I look at the Demod NDE spectrum in blue, I count maybe sideband spacing of 1/7 of an order... would be around 8.5hz?

The 7hz sideband is where... Demod DE blue?

Acceleration TWF or demod accel TWF might be interesting if you've got one.

I would be interested to see your regular velocity (or acceleration) spectra on log scale to see if there is some recgonizeable pattern. Although on linear scale it looks like they are pretty darned low all the way out to your Fmax around 3600 or 5400hz. I agree with you the magnitudes are small. If not for the history of problems, I'm guessing you wouldn't pay attention to this? I don't think I would give a second look at these low magnitudes if they were at our plant.

Blue is for Motor 'B', Red is Motor 'A', data was taken at the motor NDE in the Radial direction.

I have attached the acceleration time waveforms as well as the velocity spectra on a log scale. Also, a zoomed view of the demod spectra so that the sidebands are visible.

You are right that I would not normally bother at these values. The call in this case was however to solve the bearing failure problem.

Eccentricity, impacting or torque pulsations seem some of the possibilities. Any opinions?

Regards,

Aditya

Additional_data.doc (215 Kb, 31 downloads)

quote:

1.If this is running speed modulation of the bearing tones, shouldn’t the bearing tones too appear in the demodulated spectra? No sign of them. Any other cause for the 1X harmonics in the demodulated data?

To find where these 1x peaks are coming from in the demodulation data, I would suggest finding the carrier that is modulated by 1x in the higer frequency range of the normal acceleration data.

I see the label where the demodulation is "4-20K" 1600 Hz,. What does this mean? Is the demod data taken from 4K to 20 K as the HP Filter or what does it mean? What is your HP Filter or are you using a band pass filter of 4 to 20 K? If you are using a bandwidth filer of 4-20K, change it until you find where the 1x is coming from and then zoom in on regular acceleration data to see what might be there in real life rather than demod.

Like Pete said and you said, the amplitudes are LOW but to aid in finding where the 1x demod peaks are from, find the carrier that has 1x sidebands. I would say they are probably not coming from the bearing defect, but who knows, could be an early bearing defect in the high frequency area.

What happens to the bearings when they "fail"? A RCFA might help find what is going wrong causing these bearing failures.

1x with harmonics frequently show up in the demodulated data. I think they become visible in vibration when there is NO OTHER problems, such as inner/outer race defects, in a bearing which would otherwise bury 1x.

1x with harmonics in demodulated spectrum may indicate insufficient lubrication. So I would try lubing the motor first and see what happens. This pattern by itself does not point to a major problem, IMO.

7.6 HZ sideband is a puzzle by itself and should be addressed but may not be necessarily a red flag.

Could you please post demodulated TWF?

I'd like to theorize that 1x could be false in certain cases. Imagine that with the BP filter set at 4-16 kHZ the accelerometer goes into resonance in this range and the excitation frequency ( random noise) is not a problematic one. This resonance frequency (carrier) will be modulated by the run speed due to the factors such as imbalance.

I think special measures should be taken to ensure proper accel mounting at such a high filter setting.

David

The demod BP filter is 4-20 kHz, with an Fmax display of 1600 Hz. Using a lower filter say 2-10 or 2-20 kHz gives bad data as the twice rotor bar pass frequency & 2FL sidebands come in this range. Hence, twice line frequency & multiples dominate the demod spectrum.

I don't think the signal is due to accelerometer resonance. The same accelerometer was used for both the motors, data taken at identical locations & motor 'A' does not have any peaks in the demod at all.

The Commtest analyzers do not have the facility for demodulated time waveforms.

I do not think lubrication is an issue. I took Acoustic Emission readings on both motors, the values are quite low. Also, the bearings sounded smooth on the headphone output.

Another piece of information - the failures are happening even after changing the motor, they are happening at Location 'B' only. This would indicate Pump/Foundation is an issue. However, vibration on the pump is lower still & demod on the pump bearing with the same setting does not show any 1X harmonics (for either Pump).

This one's giving me gray hairs early!

Regards,

Aditya

quote:

Another piece of information - the failures are happening even after changing the motor, they are happening at Location 'B' only. This would indicate Pump/Foundation is an issue. However, vibration on the pump is lower still & demod on the pump bearing with the same setting does not show any 1X harmonics (for either Pump).

It seems it is not a motor problem. I wonder the 7.6 hz could be a flow induced resonance causing torsional vibrations with a detrimental effect on the motors thrust bearing?
Therefore, will the output voltage of an electronic pressure sensor at the discharge of pump B reveal pressure pulsations at 7.6 hz in the FFT?
Just my two cents.
Regards,
Arie

What is the failure mode for this bearing found upon its disassembly and RCA? Are any photos available?

quote:

The demod BP filter is 4-20 kHz, with an Fmax display of 1600 Hz. Using a lower filter say 2-10 or 2-20 kHz gives bad data as the twice rotor bar pass frequency & 2FL sidebands come in this range. Hence, twice line frequency & multiples dominate the demod spectrum.

Analysis:

It goes without saying, but The Fmax is being "filled" with different data when the BP is 2K to 10K or 2K to 20K than with the 4K to 20K BP.

This means that the 1x is coming above 4K and is not able to be displayed in the 1600 Hz Fmax most likely because the area between 2K and 4K is "filling" the Fmax before the data above 4K is reached. Even with the 2K to 20 K BP which covers the area of high frequency that the 4K to 20K does, it is obvious that the demodulation data is being restricted to below 4K.

I have always had a problem with the reasoning of a small Fmax with a long HP or BP filter. Even if the system is capable of getting good data out to 20K Hz, how does it decide what data to display within a given Fmax. Seems like yours is an example of what I am thinking. Based on this misunderstanding on my part and based on what you are seeing with the different BP Filters, the data you are seeing in the 4K to 20K BP is probably located between 4K and 6K in real time.

Have you tried to take normal non demodulation data at different Fmaxs above 4K to see if you can determine where the carrier is located from which the demodulation data is coming? I see one of your first spectrums is or appears to be out to 6K Hz but is in velocity and autoscaled. Using autoscale on the vertical scale and using velocity instead of acceleration will hinder this analysis.

quote:

What happens to the bearings when they "fail"? A RCFA might help find what is going wrong causing these bearing failures.

Do you have any of the "failed" bearings? This might be more help than trying to understand the spectrums.

Also trying to determine where the 1x in the demodulation is coming from might be useless unless it is coming from an early bearing defect or something like that, but after all, that was one of yhour original concerns, or so it seemed, so finding the answer might not be a total loss.
OMOAICBTW