Hello guys, here is another trick from behind the dikes .
Recently there have been several interesting discussions about rotor bar condition of an induction motor.
In another recent posting I have proposed a method to determine rotor bar condition that I feel is somewhat unknown so I like to share it with you on this posting devoted exclusively to this method: RBPF AM demodulation technique. Also refer to:
http://maintenanceforums.com/eve/forums/a/tpc/f/7161085912/m/7801080543

The technique is based on on-line vibration measurement and enveloping feature. It is complement to the well-known stator current signature FFT sideband analysis which is also an amplitude demodulation technique. It could be a substitute for this stator current analysis where access to MCC and/or CT is restricted or time consuming.
The basic idea is: If there is a rotor bar problem - either broken Cu bar(s), hi-Ohmic silver solder joint(s) or voids in Alu casting bar(s) - then one of the prime indicators is an amplitude modulation of RBPF with PPF modulation frequency.
This is how it works:
a) If you don’t know RBPF then make a guess: for 2p > 3, 500 .. 1500 hz is a good guess, for 2p < 3, 1500 .. 2500 hz is a good guess for most industrial applications.
b) Select an ACC envelop filter that best includes RBPF. (envelopping in SKF Microlog, demodulation in VB Commtest, PeakVue in CSI, etc, etc. different names, same baby).
c) Select max. no. of lines available 3200 or 6400, 4 .. 8 averaging, frequency span 0 .. 200 hz, show patience as data collecting may take a minute or more.
d) Measuring location does not matter, because modulation index is the same at any location, best results if sensor is somewhere in the middle of the motor frame close to the stator iron core.
Now evaluating the data:
Observe the very left end of the spectrum: is there a frequency component equal to PPF? You can calculate PPF from the 2*Fline and the fundamental speed from high resolution VEL or ACC measurement.
Also observe 1*RPM and multiples in the spectrum: do they come with PPF side bands!?
If answer is 2 times yes then this is a good indicator for a problem. This test can also provide trend information but requires a reproducable measuring point for best results, like a stud glued on frame.
Other electrical / mechanical condition indicators:
aa) You will also observe the 2*Fline, an indicator of air-gap non-uniformity. IMO, don't worry too much about 2*Fline in envellop spectrum. It is always there since motors with perfectly uniform airgap are few. It is usually not an indication for a failure mode.
bb) If belt worn out then you will find belt repetition frequency and multiples in spectrum.
cc) The humming noise should also be audible in PPF.

It is basically a method to get something from an accel where it is not designed for: LF torsional information! If one bar does not contribute to torque development (sorry Howard / Pete ) then this will result in a torque pulsation. A torque pulsation and the associated stator (work, real) current pulsation produces a angular speed pulsation hence a torsional vibration, right?

Personally I am interested in:
aaa) Will it work for DC machines too, winding failure? Not yet been able to verify thoroughly.
bbb) Also the severity question needs to be addressed: from practical experience I believe AM modulation index 1% or 40 dB down is a good indicator, just like the stator current method.

I hope this posting can become a platform to exchange test results.

Regards,
Arie Mol

It would be certainly interesting to develop a new testing method by cooperation on this board. I took lots of readings on the 1400 hp motor with 3 broken rotor bars (see the thread Classical Rotor Bar Problem).
I am attaching a ppt file with several spectra:
First the spectrum with upper frequency of 400Hz. It was shown before, clearly indicating the split peaks.
The second slide is the look at the pole pass frequency. The peaks are separated by 0.848 Hz. Since the motor was running at 1787 rpm, the slip was s=(1800-1787)/1800=0.00722.
I was taught by Motor Doc that:
quote:
___________________________________
Originally posted by MotorDoc:
On another note, the pole pass frequency is either twice the slip frequency or the number of poles times the slip (both provide the same result).
___________________________________________________________
1) Number of poles times slip is: 4* 0.00722=0.02888 (of what????). Probably not.
2) Twice the slip frequency: 30Hz * 2*0.0072=0.432 Hz (????)

Since neither advice gives me what I found I have to assume that
the pole pass frequency is actually: 0.00722* 30Hz *4poles=0.866 Hz. Close enough to 0.848 Hz. Please, let me know what it really is. As I have said, the “pole pass” frequency is all Greek to me.

The third slide is the rotor bar pass frequency (RBPF). The rotor has 38 bars and the 38x peak is really tiny.
The forth slide is a detailed view of the RBPF and the sidebands at 120 Hz. If you were looking for the PPF at 0.866 Hz (or 0.848 Hz), it seems to me that you really have a very little chance to find it in 1131.9 Hz range. Am I right or totally out?
jank

ARIEMOL.ppt (670 Kb, 71 downloads)

Thanks for the post Arie. Definitely an original idea.

I agree with Jan's point that it would be tough to find closely-spaced pole pass frequencies way up around RBPF unless you have bazillion of lines to give you high resolution and high Fmax. Arie suggested to use demod to overcome this limiation.

As was mentioned, current signature analysis is viewed as a more reliable indicator of rotor bar problems than vibration. I believe the reason is that vib is susceptible to false alarms. I have seen many motors that show pole pass sidebands around 1x in vib but not high pole pass sidebands around LF in the CSA. I'm not sure what the cause is... dynamic eccentricity? (we don't have the dynamic range to check the current spectrum way up around RBPF as some instruments do to look for eccentricity in the current). One wonders if these other pole pass sidebands will be susceptible to those same false alarms or not.

A question for Arie - have you ever found the pole pass frequency on the demod but NOT around 1x? Or vice versa?

This message has been edited. Last edited by: electricpete,

Jank

Here are both methods of calculating PPF. For this example, we will use a 1755 RPM (4-pole) motor.

First method: twice slip frequency

2*((Ns-Nr)/Ns)*LF

2*((1800-1755)/1800)*60 = 3Hz


Second method: #of poles

(p*(Ns-Nr))/60

(4(1800-1755))/60 = 3Hz

The tendency from the standards/certification development groups such as ISO to use the twice slip frequency method. However, both end with the same values.

Howard

Thank you, for the lesson, I needed it! I will stick with the first formula because it has been giving me the correct results in the past. It is also a living proof that the “pole pass” frequency has nothing to do with the number of poles.
jank

Good news, I may have a chance to try this method out towards the end of the month. We have a 800 hp motor, 2 pole that appears to show pole-pass sidebands around 1X and 2X (slide #1). We performed an online PdMA test in June, but we didn't see the pole-pass in the current spectrum. However, the current was approximately 49 amps (99 FLA). The pump is tested quarterly on mini-flow. We are going to run it at a higher flow rate during our shut-down and repeat our vibration and MCSA measurements.

I'm trying to figure out the number of rotor bars. The high-frequency spectrum (slide #2) shows a peak at 122,175 cpm (2036 Hz) with 120 Hz sidebands and a harmonic at 244,350 cpm. This would correspond to 34 rotor bars, which seems low to me. What do you think?

We use Entek for our vibration data collection, so I have a choice of spike-energy (demod) filters. I can use 500, 1000, 2000, or 5000 Hz filters. I'm leaning to setup a measurement using the 1000 and 2000 filters. 6400 lines from 0-200 Hz with four averages should take around 2 minutes.

2MP141.ppt (102 Kb, 51 downloads)

I think you have correcly diagnosed the number of rotor bars as 38 based on the spectrum.

I was also surprised that a machine this big would have only 38 bars. Then I took a look at attached database of rotor bars and stator slots for various motors uploaded by JanK. I have set it up to display three machines that have very close hp and speed rating to yours and that have 40 bars. If you remove the filter condition on column I (by selecting "ALL" in cell I1) and browse the data you'll see most machines in this size range have more bars. So imo, 38 bars is unusual, but not unreasonable.

8561086992_Jans_rotor_bar_list1.xls (135 Kb, 88 downloads)

Pete,
It is a great list! I would be very proud of it if I were responsible for it. But I am not. So as a compensation I am attaching a file with some general rules for choosing the number of rotor bars. It is from my favored book by B.Heller: Harmonic Field Effects in Induction Machines.
jank

good_rotor_bar_numbers.pdf (69 Kb, 79 downloads)

We did a test last night on our 800 HP, 2 pole motor with suspected rotor bar issues. We were able to get measurements with the current at approximately 78 amps (99 FLA). See the attachment for our vibration demod and MCSA data.

The vibration data clearly shows the PPF sidebands around 1X and 2X as well as the PPF peak and harmonics. The current spectrum shows the PPF sidebands, but the values are only 50 dB down. Any comments?

2MP141_Hi_Flow.ppt (144 Kb, 57 downloads)

Steve...

Stunning. Thank you very much. For me, this is an excellent comparison between vibration and current signature indications of potential rotor degradation. It is helpful for me, because we don't routinely collect current signatures...and are left with primarily vibe data and offline testing to determine rotor condition.

Although I admit to only infrequent use of MCSA, my education tells me that 50dB down is not insignificant. Thumbrules I recall are 54 and 45 dB down (transposed 4 and 5) as alert and action levels. Do experienced MCSA users agree?

In Arie's original posting, he indicated that PPF sidebands of running speed harmonics indicate a problem...but that 2XLF in the demod spectra is not (correct me if I misinterpret, Arie). That big 7200 peak in Steve's demod spectra could be either...as it is 2XLF and the 1st order PPF sideband of 2X shaft speed. I would like to hear Arie's interpretation of that peak.

I am curious, Steve, of the units in the demod spectra. I am an SKF user, and am used to seeing acceleration units in our enveloped spectra. Your units appear different. I would also be interested to know if your data included an acceleration spectra with high enough Fmax to show RBPP? If RBPP was evident...what was its amplitude?

Thank you very much for posting, Steve. It is useful to see a real empirical comparison between MCSA and vibration.

George,

Thanks for the comments. I was hoping that some of the electrical experts would have commented by now. In regards to your questions, let me give them a try.

I'm guessing that the 7200 cpm peak is 2xLF based upon the 2xLF sidebands around the RBPF and 2xRBPF peaks in the normal velocity spectrums. Our highest frequency reading went to 3000 Hz (360,000 cpm) and there appears to be some sidebands around 3xRBPF, but we can't see that peak. The TWF data is showing a lot of modulation.

The units of the demod spectrums are g's. I'm not sure what unit you are seeing. If you zoom in or print out the plots, you can see the units in the upper left-hand corner. Reading across the top of the spectrum plot, you have the scale (dB), units (g's), filter (1kHz gSE - g's Spike Energy), and detection (Peak-Peak).

We have velocity spectrums that show the rotor bar peak (~122,000 cpm). The maximum amplitude is 0.001 ips for 1xRBPF and 0.002 ips for 2xRBPF. The peak of the 2xLF sideband around the 3xRBPF is 0.003 ips.

As far at the MCSA data showing PPF sidebands of 50 dB's down. Based upon guidance from PdMA, the rotor condition is considered moderate. The advisory is to continue to trend. Using the formula by Thompson (articles referenced in the other rotor bar thread), the number of broken rotor bars is estimated as:

n = 2R / (10^(dB/20)+ p) = 2*34 / (10^(50/20)+ 1) = 0.21

n = # of broken bars
R = # of rotor bars
db = Average amplitude of PPF sidebands around line frequency current
p = number of pole pairs

Thanks,
Steve

Hey Steve,

Couple of questions:

What filter are you using in collecting your vib info?

In your Emax MCSA data what does it look like out around 300Hz?

Marty

Marty,

For the demodulated vibration data, I collected data using a 1kHZ and a 2kHZ gSE filter. The data appears the same. For the normal vibration measurements, we use a 318 cpm (5.3 Hz) HP filter.

I updated the attachment to include the MCSA data around 300 Hz.

Thanks,
Steve

2MP141_Hi_Flow.ppt (166 Kb, 35 downloads)

To All,
I know I posted this a while back. It's a case history I put together that correlates CSA with vibration at both the PP area and the RBP area at 30%, 60% and 80% load.
Steve,
I didn't see a large increase in the PP sideband modulation amplitudes with the increase in load. Unfortunately, you were just under 80%. The motor in my case history did not really show up as a bad rotor until it hit 80%.

rotorbarcase1.doc (1,282 Kb, 48 downloads)

Ron,
Thanks for posting. I feel a lot better prepared to deal with vibration evidence of suspected rotor problems.
The picture of the damaged rotor made me consider how much of the damage was due to resistive heat from the damaged rotor bars, vs frictional heat from contact with the stator? There looks to be some striations within the damaged area that could occur from rubbing contact? Any insight?

George,
Absolutely no rub involved. This is just a cast aluminum rotor. The marks I believe you are talking about are the laminations. The vibration on this rotor was low (see the plots). This rotor probably had high porosity in the end ring, right at the junction to the bars. I have enclosed another picture of a rotor from a similar machine to show how a bad pour can fail like this one.

Thanks guys for the responses. I think I have to respond to a couple of questions raised.
I apologise for being late, surely I have not lost interest however December is a busy month isn't!
31/Oct, Pete: No and no:
I have always found PPF and harmonics along with a series of PPF harmonics around 1*RPM and 2*RPM and 3*RPM, etc.
Steve's contribution posted on 01/Dec is an excellent example of how the spectra should look like and this also shows the close relationship to the MCSA method.
02/Dec, George:
IMO the 2*Fline and her harmonics and the PPF and her harmonics have nothing to do with each other, although these girls have the same mother. The common denominator is RBPF amplitude modulation. When rotor bar(s) are in bad condition the modulation will have PPF frequency components in demod. When air gap is non-uniform the modulation will have 2*Fline frequency components in demod.

I would like to add some additional info here.
Trick 1:
I think one should be carefull when 2-pole motors are involved: the mechanical 2*RPM and the electrical 2*Fline may create a beat at PPF that has nothing to do with defect rotor bars. Then how to descriminate between a rotor bar defect or a beat when 2-pole motor is involved?
I think here MCSA will be the judge.
Trick 2:
I use both the 2*Fline (always in demod as very few motors have perfect uniform airgap) and the PPF to calculate actual motor slip in order to estimate shaft power (shaft power is proportional to slip). I have seen several times that with slip (load) variations the PPF amplitude may change significantly indicating rotor bar condition can be temperature (load) dependant.

Thanks to all for the contributions to this posting!
Arie Mol

quote:

I use both the 2*Fline (always in demod as very few motors have perfect uniform airgap) and the PPF

Oops!
PPF = 1*RPM (hi-resolution 125 hz / 3200 lines)