Rotor broken bar problem or variable load or vibration issue cause sideband around line frequency on current FFT spectrum?

Edisonindia,

Low frequency vibration can cause swinging current. Because equation of stator current is: I(stator)=I(50 or 60 Hz) + I (50 +/- f_vibration).

with low frequency vibration (from 0 to 5 Hz), you can see the sideband around line frequency likes as Pole pass frequency which can cause fault analysis.

Thank you!

I agree with op, it's always nice to have things to help confirm before you make the call to pull a motor. Would be awfully embarassing to send a motor out only to find it's there's nothing wrong with it.    What kinds of things other than actual rotor problem can cause apparent elevated pole pass sideband around FL in the current spectrum:

  • varying load which happens to vary at a frequency near pole pass frequency.  Check frequency as preciselyy as you can.... the closer the match to exactly pole pass frequency the less likely it is load oscillation (would be a big coincidence to oscillate at exactly pole pass frequency).
  • if number of spider arms are same as number of poles as Aditya reminded us recently (and it explains a pattern on a family of our motors).  Wouldn't change much over time under similar conditions.  
  • some of the literature suggests that dynamic eccentricity (rotor moving in airgap at 1x) might cause pole pass sidebands, although I don't recall seeing any cases of that in our plant or posted here. 

We use current signature analysis by feeding output from clampon direct into our vibration data collector.  It easily shows pole pass sidebands around FL, but doesn't have the dynamic range needed to look for sidebands around 300hz, so we're missing the ability to check for presence or absence of this "swirl effect" as confirmation.

Personally I also look for some of the following indications to somewhat confirm rotor bar problem: growling noise, pole pass sidebands in vibration under load, and change in behavior over the first few hours after startup (again under load), and of course increasing trend of that current sideband.  

Here's a previous topic about swirl frequency:
https://www.maintenance.org/top...idebands-around-5-lf
I think Cando has posted a few examples before this thread also.

Now, a different subject. I notice the second case study has lower sideband much higher than upper sideband around line frequency. As far as I can tell that is often associated with a motor that drives a high inertia load (and of course high inertia loads also tend to make each start harder on the rotor bars).  I'll bet this motor drives something with high inertia like a flywheel, very-large diameter fan, or else a speed-up gearbox that increases the effective inertia of the load.  What does it drive?

Here were some previous threads discussing this idea that higher LSB than USB occurs when driving high inertia load:
https://www.maintenance.org/top...ilures-and-freqyence

https://www.maintenance.org/top...962761908#7481072103

May too late to be relevant, but I don't think either of those two are broken bar cases.

The first spectrum has just too many sidebands to be a rotor problem. My guess would be mechanical modulation.

The second spectrum has only the lower PPF sideband, a true case would give you both the peaks.   

EPete - I do have some case studies of dynamic eccentricity causing PPF sidebands, but this doesn't always happen. Just saw a case last week where a 790 kW motor with sleeve bearings had excessive clearances (0.25 mm, upto 0.16 mm is acceptable). We saw strong PPF sidebands & suspected the bars/clearances. Turned out to be a clearance issue.

Regards,

Aditya

 

 

Aditya posted:

May too late to be relevant, but I don't think either of those two are broken bar cases.

The first spectrum has just too many sidebands to be a rotor problem. My guess would be mechanical modulation.

The second spectrum has only the lower PPF sideband, a true case would give you both the peaks.   

EPete - I do have some case studies of dynamic eccentricity causing PPF sidebands, but this doesn't always happen. Just saw a case last week where a 790 kW motor with sleeve bearings had excessive clearances (0.25 mm, upto 0.16 mm is acceptable). We saw strong PPF sidebands & suspected the bars/clearances. Turned out to be a clearance issue.

Regards,

Aditya

 

 

I would respectfully disagree about significance of high LSB without much USB.  To my mind it indicates high inertia of driven load. We have a series of motors with flywheels (high inertia) and they all have a pattern of LSB much higher than USB (they are also all healthy but I this represents small asymmetry in rotor and large asymmetry would increase both sidebands but keep the LSB higher than USB... ... ok that's just an assumption on my part).  Example attached. The theoretical reasons for expecting LSB > USB on motor driving high inertia are somewhat mystical to me but discusesd in links in papers above.

You mentioned that  pole pass sidebands around current might indicate dynamic eccentricity in your experience. (I haven't seen it myself).   Do you have any suggestion to rule out dynamic eccentricity when we see elevated pole pass sidebands in current?   I'm guessing the two ammeter trick wouldn't work because the pattern for dynamic eccentricity would resemble the pattern for rotor defect.

 

 

 

Attachments

James I Taylor speaks a lot in his book regarding the relative heights of the upper and lower sidebands but this is a bit mystical to me also! I have always looked for both but am willing to be convinced otherwise. 

I have always understood the modulation to be as a result of a momentary torque reduction as the broken bar(s) pass by a pole as a result of the higher resistance nature of the broken bar circuit. This torque reduction is seen, due to conservation of energy principles, as a momentary reduction in supply current hence the modulation at PP. So I tend to look for the classic upper and lower SB's.

 

Gary 

[quote]I have always understood the modulation to be as a result of a momentary torque reduction as the broken bar(s) pass by a pole as a result of the higher resistance nature of the broken bar circuit. This torque reduction is seen, due to conservation of energy principles, as a momentary reduction in supply current hence the modulation at PP. So I tend to look for the classic upper and lower SB's. [/quote]

Your explanation makes perfect sense to me on an intuitive level.  And I would be happy to accept it as simple amplitude modulation with equal sidebands expected on each side.... EXCEPT for the fact that:

1 - Academic papers state otherwise.

2 - Review of my plant equipment matches the papers. Specifically, the motors we have that drive the highest inertia load (flywheel) are the ones that have this behavior of LSB much larger than USB.

 

One paper here:

https://www.maintenance.org/fil...hompson_T32pg145.pdf

I have quoted from the paper below with my own clarifications in [square brackets] 

[quote]

fsb = f1*(1-2s)     [where f1 is LF]

This [USB] is referred to as a twice slip frequency sideband due to broken rotor bars. There is therefore a cyclic variation of current that causes a torque pulsation at twice slip frequency (2sf1) [pole pass frequency] and a corresponding speed oscillation that is also a function of the drive inertia. This speed oscillation can reduce the magnitude (amps) of the f1(1-2s) sideband, but an upper sideband current component at f1(1 + 2s) is induced in the stator winding due to the rotor oscillation. This upper sideband is also enhanced by the third time harmonic flux. Broken rotor bars therefore result in current components being induced in the stator winding at frequencies given by

Fsb = LF*(1+2s) [/quote]

I'm not claiming this is the entire whole story but there's definitely something to it when the field measurements match what is predicted by the paper (our high inertia motors have low USB<<LSB). If anyone else wants to look for the same pattern USB<<LSB on their own high inertia equipment with high inertia I'll be interested to hear if you guys see the same results I see at our plant. It does raise a question about how we role it into our questions of severity. I tend to think if you take a given motor with a given rotor defect, it would give a higher LSB on a high inertia load than a low inertia load (even though same rotor condition). So the criteria of looking at highest SB fooled us. Maybe we should be adding both sideband magnitudes (amps) together before computing the db and comparing it to a db limit which is  6db lower (factor of 2) than the limit we would use normally use.

I have had a look at the paper and whilst I had some understanding of it I did struggle so will have to take your word for it Pete! 

Not to disagree but I cannot recall any cases I have encountered where there has been any significant difference between the USB and LSB. The majority of mine have occurred on large FD and ID fans (coal fired power station) so perhaps high inertia?

I think in line with Occam's Razor and (more likely!) my inability to as yet absorb this paper I am for now going to stick with my simple if perhaps somewhat incorrect explanation! I have always had an easier time with mental visualization rather than mathematical explanation.

Gary

 

I can understand why you'd view it that way and I think I'd view it that way if I hadn't seen for myself what I reported above on our own machines.  I'll offer two more points to consider:

1 - As you mentioned, "high" inertia is a matter of how high.   The machines I referred to are intentionally designed for high inertia to the tune of J=110,000 lbm*ft^2 for this 13.8kv 8,000hp, 1200 rpm motor with flywheel.  To give you another idea of how high the inertia is, for direct-on line start on a very stiff power system (voltage more than 80% at motor terminals during start) the time to accelerate the motor UNCOUPLED is more than 20 seconds !!!.  That is way more than any other motor at our plant including fans  (although we don't have FD/ID fans). The rotor has a special beefy design around the endrings to accomodate the severe start.  Do you know how long it takes your FD and ID fans to get up to speed?

2 - Aditya had mentioned in some old threads that he has only seen this pattern on wound rotor motors.  I don't know the significance of that but I'll propose one explanation and Aditya can tell me whether I'm off base.  Wound rotor motors are used primarily in two applications: 1 - where speed control is required; 2 - where very high inertia loads are started and power system cannot support DOL start.  And while it might not be fair to compare wound rotor motors to squirrel cage, I'm wondering if the motors he saw the pattern on were also driving high inertia loads. 

I'm still interested for Cando to tell us what his motor ("case 2") was driving since that will be another data point in this discussion.  Cando?   

 

 

 

 

 

 

 

Not 20 seconds Pete!

dont doubt your explanation but will have to take take your word for it for now until I can put some more time into understanding it! 

Do you have links to any more papers that explain this?

How does this tie in with the PP sidebands around 1x etc which I presume are due to a torque modulation and hence a modulation of 1x?

regards

Gary

 

 

Appendix B of the linked Thompson paper gives an explanation of how (in absence of speed modulation) the asymmetry in the rotor gives a LSB.   It makes pretty good sense to me.  Here's my brief summary of that:

  • We know the rotor moves at Nr and (for a balanced field) the rotor field moves at (Nsync-Nr) in the forward direction with respect to the rotor. When you add them together you get a field that is in synchronism with stator field.
  • When you add a symmetry to the rotor, what happens to the field? It is no longer a pure forward / "balanced" rotating field. Sitting in the rotor reference frame there is an asymmetry along one axis.   That means there is a forward and backward component. (In a similar way that we know that a single phase motor stator winding has forward and backward components…a well known result from motor textbooks).   Adding the two rotor field components (forward and backward) to rotor speed gives a synchronous field component (from the forward) and a LSB (from the backward) when we account for the number of poles.

What the linked paper by Thompson glosses over is how does speed oscillation transfer some of the LSB to the USB. That phenomenon is described in gory mathematical detail (not particularly easy to understand though) in another paper "AI Techniques in Induction Machines Diagnosis Including the Speed Ripple Effect" by Filippetti/Franceschini/ Tassoni from IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 34, NO. 1, JANUARY/FEBRUARY 1998.   I have a copy of the paper (it comes free with my membership in IEEE IAS) but unfortunately it is copyrighted so I can't post it.  I'll keep an eye open for other references but that's the only one I've got for now that actually attempts to explain the phenomenon. I've seen a few others that mention the same phenomenon without explaining it.

Another thing I just noticed in that Italian speed ripple paper is the phenomenon of MULTIPLE pole pass sidebands around LF in current is attributed to the same speed ripple effect (it is vaguely reminiscent of FM imo).  I'm not accepting that as fact but that's another thing I'll look for. 

One thing that just occurred to me that doesn't make sense. If we accept this speed ripple theory, then any motor with similar magnitude LSB/USB arising from rotor asymmetry should be experiencing speed oscillations  at pole pass frequency. .   But the few times that I've looked with strobe I have never been able to discern oscillations at pole pass frequency.    I guess the magnitude of the speed oscillation is so darned small that (even when integrated over the long pole pass period) the associated change in apparent position under strobe is so small that it is not visually discernible? ... that would be a miniscule speed oscillation all right.   I'm going to start looking closer for pole pass oscillation (even among my healthy motors) with strobe.  And I'm also going to look at that  paper to see if I there are any clues about the order of magnitude of speed oscillation which might be expected for a given set of sidebands.

Thanks Pete,

Your speed ripple  theory reference is what I was alluding to when I asked about the link between the torque/speed modulation we see as PP sidebands around 1x with broken rotor bars in the vibration spectrum and the phenomena we see in the current spectrum. Something that as you know requires hirez/zoom to pull out of vibration data and can be seen quite clearly sometimes with a long time waveform.  Not something I have tried to look at with a strobe but can often be heard. I had put this speed/torque modulation down to a instantaneous effect as the broken bar passes by a pole (reduced current flow in this rotor turn) which seemed to gel nicely with my explanation for PP SB in the current spectrum. However as discussed the PP or 2SF sidebands in the current are apparently not as a result of this so where does this leave use with regard to the vibration phenomena?

And here me bimbling along thiniking I had a pretty good understanding of the rotor bar symptoms. I guess as with a lot of things condition monitoring we dont necessarily need to have a complete handle on the theory to make good judgments!  

regards

Gary

 

 

 

 

 

 

[quote]Not something I have tried to look at with a strobe but can often be heard. I had put this speed/torque modulation down to a instantaneous effect as the broken bar passes by a pole (reduced current flow in this rotor turn) which seemed to gel nicely with my explanation for PP SB in the current spectrum. As discussed the PP or 2SF sidebands in the current are not as a result of this so where does this leave use with regard to the vibration phenomena? [/quote]

The bar is passing in and out of the field poles at pole pass frequency as you originally said. That is good enough for seeing the effect on torque and possibly vibration (more later). But why can't we use that to explain effect on current in a single phase? I would it cannot because it would require us to a assume that a given single phase of the stator winding is equally associated with all field poles at all times, which it is not. An indvidual phase winding (pick one of the three phases) is not linked to each of the forward synchronous field poles. An individual phase winding has both forward and backward mmf's (and is linked to forward and backward flux waves). We can only ignore the backward when everything is balanced.   So we cannot look at interaction of bar with only forward stator field poles to determine effect on an individual phases in the same way we would look at interaction of a bar with poles to determine effect on torque. That's why appendix B explanation is needed. I think.

Going back to vibration, I have three different ideas of how broken bar might be translated to pole pass sidebands around running speed and harmonics.

  • 1 - Bar affects the airgap flux pattern. The anomaly in airgap flux pattern rotates at 1x. It results in pole pass sidebands in the same way as dynamic eccentricity does (a component of unbalanced magnetic pull rotates at 1x and pulses at pole pass frequency resulting in 1x with Fpp sidebands). 
  •  2 - Speed ripple. If 1x speed is modulated at pole pass frequency (due to torque oscillations) then the sidebands would be Fpp around 1x. In my mind related to frequency modulation something like this: http://www.radio-electronics.c...dwidth-sidebands.php
  • 3 - Torque oscillation in combination with asymmetric support stiffness. Picture a motor on a very tall slender base. The torque is enough to deflect the base horizontally.   As the torque pulses the base moves horizontally. That gives you pole pass fundamental in vibration. I guess pole pass sidebands might possible occur if machine alignment is affected by the movement.

 [quote]I'm going to start looking closer for pole pass oscillation (even among my healthy motors) with strobe. And I'm also going to look at that paper to see if I there are any clues about the order of magnitude of speed oscillation which might be expected for a given set of sidebands. [/quote]

I did get a chance to look at three large motors yesterday with LSB ~ USB and no speed oscillation evident by strobing under load. I haven't gotten an opportunity to look at the suspect cast rotor motor from the other thread yet, but soon.

I don't have a good way to calculate but that never stopped me before. I did the best I can just to help get myself in the ballpark.

 I came up with a SWAG equation to estimate peak-to-peak angle variation in degrees:

Theta ~ 18*Tpp^2*X*Fnp/Taccel

where

  • X is ratio of pole-pass sideband peak to RATED current. (if sideband is taken at 50% load, then X will be twice ratio of sideband to measured 60hz peak).
  • Tpp is the time interval associated with pole pass frequency (Tpp=1/Fpp)
  • Taccel is the time it would take for the motor to accelerate from rest to full speed IF the accelerating torque were at 100% rated torque througout. I figure this number would be between 1 - 2 times the actual accelerating time for most motors (since the actual torque will be higher than rated torque through most of the speed range). It's an invented number that saves us from having to deal with unintuitive parameters and units like torque and inertia that we typically can't guess off the top of our head.
  •  Fnp is nameplate speed (rotations per time).

 The form of the equation suggests what may have already been obvious: the oscillation is more likely to be noticeable when  sideband magnitude is high, motor slip is low (high Tpp), nameplate speed is high (shows up better on 2-pole than 4-pole for the same slip and other parameters), and Taccel is low (low inertia motors). The variation of all of these is 1st order (linear or inverse) except Tpp which appears as a square term.   I view Tpp as a function of slip only (independent of rated speed or number of poles of motor) since Fpp = 2*s*LF regardless of number of poles.  So the four variables on RHS of the equation can be considered as independent variables.

 I used computer algebra equation (attached) to make it easier on me (but probably tougher for you to read...sorry). Plugged in some values and could easily come up with values where the oscillation is not noticeable or is noticeable.

 Again it is very much a SWAG as there is one questionable assumption (*) that ratio of oscillating torque to rated torque is equal to ratio of sideband current to rated current.  And I haven't even checked for obvious errors yet. But the value for me is that it makes it easier for me to believe that there can be speed oscillation going on which is enough to balance out those sidebands but still not enough to see with a strobe.

EDITED TO ADD - * I reread that article and it roughly matches the above assumption except I should be using the sum of the USB and LSB. in the ratio.  i.e. ratio of oscillating torque to rated torque would be ratio of (LSB+USB) to rated current.  The variable X should be redfined based on that sum of LSB+USB.   Can't wait to strobe check that cast rotor motor from the other thread.. should be able to do that next week. 

Attachments

but why can't we use that to explain effect on current in a single phase? I would it cannot because it would require us to a assume that a given single phase of the stator winding is equally associated with all field poles at all times, which it is not.

 

Ahh now I see of course,  this makes sense to me!

With regard to vibration and the three possible causes. Does it necessarily need to be limited to one? I think explanation 2 and 3 could quite easily occur in conjunction. 

Explanation 1 - Would the effects of this not rotate giving a rotating axis of vibration so would we not see some secondary modulation?

regards

Gary

Cando

Based upon the signatures, you are using a PdMA tester.

Concerning the 'swirl effect,' at this point they are the only ones using this as part of their diagnosis.  In my experience, I have not really seen it with the technologies I use, with broken rotor bars.

In your top example - Even if one of the sidebands fell around PPF, the sidebands around the line frequency appear to be more related to harmonics of the first peak sidebands around the line frequency.  What is the context of this operation?  Is it a belted or geared machine? 

You also have a significantly high noise floor in both the top and bottom pictures.

The bottom one - OK.  But I tend to find that the right or left PPF peak will be more dependant based upon the type of filter (ie: Hanning or Flat Top) and log spectrum (absolute or relative).  I don't remember if you have a selection to change that setting in the PdMA system.  I determine if I have an issue based upon whichever peak is higher and not a focus on the upper or lower sideband.

Wide bases in the frequency peaks mean that you have either a varying load or a lot of peaks.

Sincerely,

Howard W Penrose, Ph.D., CMRP

MotorDoc.com

Aditya posted:

May too late to be relevant, but I don't think either of those two are broken bar cases.

The first spectrum has just too many sidebands to be a rotor problem. My guess would be mechanical modulation.

The second spectrum has only the lower PPF sideband, a true case would give you both the peaks.   

EPete - I do have some case studies of dynamic eccentricity causing PPF sidebands, but this doesn't always happen. Just saw a case last week where a 790 kW motor with sleeve bearings had excessive clearances (0.25 mm, upto 0.16 mm is acceptable). We saw strong PPF sidebands & suspected the bars/clearances. Turned out to be a clearance issue.

Regards,

Aditya

 

 

Hello ADITYA,

Your comments are always very enlightening, i am a beginner in reliability and motor diagnostics. i will appreciate if i could have you as a my guild should the need arise.

orugbosamuel@gmail.com

many thanks.

 

Add Reply

Likes (1)
electricpete
×
×
×
×