2Xline frequency spikes in peakvue 3

[micjk]

Can anyone explain why I am seeing 2xLf peaks plus harmonics, some with relatively high aplitudes >1.5 G's, in some instances, in my spectral plots when using Peakvue (CSI). I am set up with an fmax of 1000Hz, 1600 LOR, 1 avg., aquired in acceleration, with a 1000 Hz hp filter. This does not show up on all motors. Peakvue detects Stress wave energy, could the 2xLf + harmonics be shorts in windings, rotorbars, etc. simulating metal to metal contact or stress wave energy. I also see it in the waveforms of these particular spectrum.
Regards,
MICJK

 

[electricpete]

That is very common. I don't beleive it is a problem. It may be an indication of soft foot or airgap problem, similar to 2LF in the regular spectrum.

k*2LF in peakvuew is usually associated with a pattern in the regular spectrum of j*RBPF (rotor bar pass frequency) +/- k*2*LF (Line Frequency).

When you apply demodulation (peakvue) to that regular spectral pattern, you will get the k*2LF that you see in peakvue. (that applies to whatever part of the j*RBPF+/-K+/-2LF is above your low-frequency corner paramter in your high-pass filter)

 

[micjk]

electricpete
are k and j representitive of random # of harmonics or something else.

 

[electricpete]

you're right. They are just integers 1,2,3 representing multiple harmonics.

 

[vanstoja]

If your k2LF peaks are coming from demodulation of the
2LF sidebands straddling the Rotor Slot Passing fundamental peak, then 1.5 g's sounds really high to me. What is your motor speed and how much higher in AdBs are the sidebands compared to the RSP peak in a standard F-A spectrum? You can get really high RSP+/-2LF peaks and higher harmonics from stator teeth saturaton if your rotor eccentricity approaches 10% due to either or both shaft loading and bearing centerline offset from stator core bore centerline. This is not too good a finding for motors with rolling element bearings though it may be more normal for motors with fluid film bearings.
What are the possibilities that your Peakview demodulation process is not picking up the RSP 2LF sidebands but rather is picking up harmonics of 2LF=120Hz? This can be caused by several things including rotor eccentricity and phase current unbalance and at this lower frequency a 1.5 g acceleration would be even worse than at the higher RSP frequency. Do you have harmonics of 2LF in a a standard F-A spectrum at the same time they are showing up in Peakvue?

 

[electricpete]

John (vanstoja)

good comments.

In response to your 2nd paragraph - The standard cutoff for the high-pass filter for Spike Energy (Entek's version of Peakvue demodulation) is 500,000cpm. I think Peakvue is similar That would filter out all of the harmonics of 2*LF up to perhaps 70x 2*LF.

(LF = Line Frequency = E).

Even higher harmonics of 2xLF (above 70x 2*LF),if present, would not show up at 1x2XLF, 2x2XLF, 3x2XLF unless the 2XLF frequency is involved in some kind of modulation or impacting. Most typically that is the modulation which gives rise to RBPF +/- 2LF. In theory, if the 2XLF vibration was causing impacting at 2XLF (imagine a very loose bar slamming into the iron), then 2XLF would also show up in the peakvuew spectrum.

So in summary:
Peakvue won't show you any signal associated with any activity less than the hi-pass filter low-frequency cutoff (typically 500,000 cpm). If you look in the regular spectrum for the direct cause of your peakvue vibration you have to look above 500,000 cpm. That would either be 2*LF sidebands around K*RBPF where K*RBPF >500,000cpm (very common), or harmonics of 2*LF above the 70th (very rare in my experience).

 

[vanstoja]

Pete,
Micjk says he's set up for 1000 Hz maximum with a 1000 Hz hipass filter which says to me he can't see anything. I presumed he meant 10000 Hz maximum so he's bandpassing 9000 Hz above 1000 Hz. Your Entek 500,000 cpm hipass is over 8000 Hz which is getting rather high for conventional 2-pole and higher pole number motor rotor slot numbers. If he's really highpassing at 1000 Hz, then 6LF or 6E is nearly there at 960 Hz and I would expect that sloppy filter rolloff might get 6E and possibly also 4E harmonics of 2E.

PS Sent you some stuff on zigzag leakage via snail mail today. Can't remember which forum you brought this up in. I'm getting confused between pumps, piping, pressure vessels, valves, motors, power systems, mechanical acoustics, electrical acoustics, magnetism engineering, etc. Any ideas on how to cope with this brainwork overload?

 

[electricpete]

I didn't see micjk's 1000 hz parameters during my first reply. That 1000hz cutoff plays the same role as the 500,000cpm (~ 8000hz) cutoff which I am familiar with for Entek.

This setup (setting Fmax at same frequency as the low-frequency corner of hi-pass filter) may seem unnatural, but it is very common and useful, for these special purpose Peakvue and Spike Energy demodulation measurements.

It is basically a two step process:
Step 1 – high pass filter to remove everything below F_cutoff
Step 2 – Demodulation (envelop-detect) processing of remaining frequency components which identifies the low-frequency modulating frequencies or triggering frequencies buried in the high frequency signal. Output spectrum is from 0 to F_cutoff

Envelope detection is the same process used to retrieve an audio signal from an am-encoded transmission. The audio signal is much lower frequency then the transmission. For example sin(2pi*15000*t)*sin(100*t) has frequency content entirely above our 8,000cpm cutoff. But when it is demodulated the envelope frequency 100hz is retrieved.

Likewise k*RBPF+/-j*2LF may lie totally above cutoff (for k=6,7, 8?). Signal strength may be very low but algorithm magnifies it.

The primary prupose of these "demodulated" or "enveloping" measurements is for detecting impacting type of events. These events may be almost undetectable at their fundamental frequency and low harmonics: BPFO. 2xBPFO etc. But when processed through Peakvue, the high frequency impact response which is repeatedly triggered at BPFO frequency will produce output of BPFO and low-order harmonics. Since there is no other low-frequency stuff in the spectrum (we have filtered it out), there is a very high signal to noise ratio and rolling element defects can be identified very early.

The fact that this algorithm also happens to produce an output when subjected to k*RBPF+/-j*2LF is a natural result of the processing algorithm. The instrument makers all tell us to ignore it.

I don't know how the Spike Energy/PeakVue magnitudes are related to the traditional spectrum magnitudes. We are cautioned not to place too much importance on these magnitudes but more importance on the pattern and trend.

I probably haven't given a good explanation. I looked for a good link. All I can find is:

http://www.bksv.com/bksv/?LO=000000...00000000000000000000000&type=application/.pdf

 

[vanstoja]

Pete,
I looked up spike energy today and found the connection to "envelope detection" which I am familiar with. Your 8000+ Hz highpass now makes sense since very high frequency resonances especially accelerometer resonances are the target for demodulation to detect hidden impact events. It now seems strange to me that Peakvue would be used with high pass as low as 1000 Hz when accelerometer resonances are of the order of 20,000 Hz give or take a few thousand. Peakvue at 1000 HP may then be reading RSP+/-2LF directly which is unlikely for your Entek set at 8000 HP. I was going to ask about the meaning of demodulated signal levels but you've answered that above. I don't have any clue as to the significance of demodulated signals either. Incidentally, I had a brain lapse talking about 6E at 960 Hz. That needs a 160 Hz power supply which I have not yet invented. Perhaps you've got one in the works. John