100 Hz vibration predominant in all 3 axes in a vertical decoupled 50 Hz motor 1

[edison123]

A 1.5 MW, 500 RPM, 50 Hz vertical decoupled motor has predominant 12x = 6000 RPM = 100 Hz vibrations in all 3 axes of radial, tangential and axial measured with a triaxial accelerometer probe. The readings (attached) were taken at both DE (bottom) and NDE (top) in two 90 deg directions (east and south) and all the 12 readings show 100 Hz vibration is predominant.

Are these 100 Hz vibrations due to motor torque pulses, which are known to have two times supply frequency? Why is axial much more than radial and tangential? How to correct it?

Thanks in advance for your valued inputs.

epete - Look forward to your views.

Muthu
www.edison.co.in

 

[electricpete]

Did you check if the three phase currents are balanced? I think unbalanced currents (normally from unbalanced voltage) could produce a 2*LF torque pulsations in a similar fashion that a single phase motor produces 2*LF torque pulsations. Whether that applies unloaded I'm not sure, but it's easy to check if someone recorded the phase currents.

=====================================
(2B)+(2B)' ?

 

[edison123]

Thanks, pete. The decoupled motor currents are indeed unequal and also varying in each phase - 61/69/64 A at 10 am and 63/73/67 A at 6 pm. Motor supply voltages are not that balanced either. Do you think this will cause 12x?

Muthu

http://www.edison.co.in

 

[electricpete]

Unbalanced current can definitely cause 2*LF.

Induction Motor Condition Monitoring: VibrationAnalysis Technique - a Twice Line FrequencyComponent as a Diagnostic Tool by Tsypkin, IEEE 2003

See article above. Note at the above link, you can scroll further down the web page to read the whole thing or download the pdf directly. (actually I can do that, but I'm already registered at that site... it was free to register, just had to give them my email address.... I'm not sure what access is available before you register)

The article demonstrates current unbalance causes 2*LF both through case study #3 and through testing. In fact both case history #3 and the testing established a tangential (torsional) component of the 2*LF that was present with current unbalance. The case study was in the loaded condition, I'm not sure about the testing.

On page 119, he says

In the case of a voltage/current imbalance (asymmetrical voltage/current in a three phase stator winding), the periodic torque component with frequency of 2LF in the induction motor may be substantial [19]. Torque pulsation results in a torsional vibration of the motor stator and rotor.

=====================================
(2B)+(2B)' ?

 

[electricpete]

Here is another paper which very specifically discusses the relationship between unbalanced voltages and twice line frequency vibration

EFFECT OF MOTOR VOLTAGE UNBALANCE ON MOTOR VIBRATION: TEST AND EVALUATION by Toshiba authors made available on the Toshiba website here:

https://www.toshiba.com/tic/datafiles/whitepaper/Effect_of_Voltage_Unbalance_on_Motor_Vibration.pdf

This particular motor happened to be tested with a relatively large voltage unbalance, leading to an elevated 7200 RPM (2 times line frequency) vibration level. Once a set of suspects is identified, additional information can be incorporated into the analysis to either eliminate suspects or confirm the inclusion of the suspected cause. Analysis of 2 times line frequency (2f) will traditionally point to a short list of suspects [6]. Most of these potential causes indicate imbalances in the magnetic field of the motor.

Of the remaining suspects [causes for 2LF vib], eccentric air gap, soft foot conditions, loose stator core, interphase and ground faults, unbalanced voltages and loose top covers each have a set of specific tests that can be performed on the motor to either confirm or dismiss a suspected cause….
Motor voltage unbalance is difficult to diagnose without data of the motor voltage condition during running conditions. Data must be collected with the motor in operation and in its installation. Without having the vibration data collected concurrently with data regarding the voltage on each leg of the motor, analysis of vibration caused by voltage unbalance will be difficult to accomplish. By observing and reviewing the data collected it is easier to evaluate the unbalance of the voltages supplied, and make a determination about whether the power supply is a contributing factor. Removal of the motor from service and sending to a repair shop may, in the instance of voltage unbalance, mask the issue. Moving the motor to a different mounting base on a different power supply may produce a vibration unlike that of the end-user instillation, further complicating the analysis.

Section III goes on to explain that unbalanced voltages cause a negative sequence (reverse rotating) component of the field which leads to twice line frequency vibration. “As the two sequences are opposing one another, this interaction leads to the two times line frequency vibration phenomenon associated with unbalanced voltage operation.”

Section IV provides results of testing motors with various controlled levels of voltage unbalance. Tables 2 through 7 provide results for individual motors. The column headings for each table are: % voltage unbalance, average radial vibration (ips), overall vibration (pu), 2x vibration (pu) where in this case pu is the vibration normalized so that the vibration with no unbalance is 1.0.

You and I know that by today’s typical interpretation of notation “2x” would mean 2x running speed. However I contend that is not the case for these tables, for these tables the heading 2X refers to 2x line frequency. The reason I say this is:
1. Nowhere do they talk about twice running speed in the entire paper.
2. All over the paper they talk about twice line frequency vibration
3. They don’t even break out 1x running speed as a column header, so why on earth would they break out 2x running speed.

With all that said, the results of the testing vary widely among motors. The Table 2 motor shows a strong trend of increasing 2LF with voltage unbalance but some of the others show decreasing trend and yet others show non-monotonic trends. The test data seems very inconclusive to me due to the wide variability, but clearly the authors who prepared the paper believed that unbalanced voltage leads to increased 2LF vibration. There may (must?) be other unknown factors associated with the specific motors or test conditions that explain why the individual motors acted differently.

New subject - axial vibration. I did some thinking about whether unbalanced currents (from unbalanced voltage) could result in axial 2*LF, and I say yes it can. But we have to assume first that the rotor iron is offset axially from the stator iron slightly. That is a condition that we wouldn't notice under balanced conditions because it results in no vibration (see below) but we would notice under unbalanced condition because it results in 2*LF axial force (see further below).

Why does the axially-offset rotor cause no vibration when currents are balanced? The 12 pole stator field It is like having 12 equally spaced magnets (N S N S N S N S N S N S) circulating around the circumference. Each one exerts a vector force on the rotor. The force includes radial and axial components. The radial forces cancel out by symmetry (I'm assuming perfectly centered airgap throughout my discussion). The axial forces (due to rotor off-center axially) do not cancel out. As the poles rotate, the radial components still cancel out but nothing changes in terms of axial force. There is a constant dc axial pull on the rotor, but it doesn’t change over time as the field rotates, so no vibration.

Now what happens when you add unbalanced currents. As mentioned in the Toshiba paper directly above, you can represent the field from the unbalanced currents as a positive sequence (forward rotating) and negative sequence (reverse rotating) set of fields. (it is similar to the principle of sequence analysis of three phase, except we disallow zero sequence because the homopolar flux path is assumed to be high reluctance). So now you have to imagine two sets of twelve magnets each, one set rotating in the forward direction and another smaller set rotating in the reverse direction. When the N magnets of the forward rotating field line up with the N magnets of the reverse rotating field, the fields add and the axial pull is the strongest. When the N magnets of the forward rotating field line with with the S magnets of the reverse rotating field, the fields subtract and the axial pull is the weakest. So there is a varying axial pull which can result in vibration.

What is the frequency? First we need to figure out how fast the poles move. Let's say at t=0 the snapshot of the forward rotating field is located as follows: N pole at 12:00, S pole at 1:00, N pole at 2:00… (12 pole motor). We know from the fundamental relationship among poles, frequency and speed that in one electrical cycle (0.02 seconds) the forward pole that was at 12:00 at time t-0 will move to 2:00 at time =0.02 sec. And (halving that) it means it would only take 0.01 sec to travel from 12:00 to 1:00 position and (halving that again) 0.005 sec to travel from the 12:00 to the 12:30 position.

So let's assume that at t=0 the poles of the forward and reverse sets of fields are lined up for maximum force.
Now let’s start the clock and watch where the poles go and what effect it has on force (12 pole 50hz motor):

[tt]=== Time 0 sec =====
Forward 12:00N / 1:00S / 2:00N / 3:00S/ 4:00N
Reverse 12:00N / 1:00S / 2:00N / 3:00S/ 4:00N
Forward and reverse poles are lined up and therefore add together for MAXIMUM AXIAL FORCE.

==== Time 0.005 sec =====
Forward ******** 12:30N / 1:30S / 2:30N / 3:30S/ 4:30N
Reverse 11:30N / 12:30S / 1:30N / 2:30S / 3:30N
Forward and reverse pole are opposing and therefore subtract for MINIMUM AXIAL FORCE.

=== Time 0.01 sec =====
Forward ***************** 1:00N / 2:00S / 3:00N/ 4:00S / 5:00N
Reverse 11:00N / 12:00S / 1:00N / 2:00S / 3:00N
Forward and reverse poles are lined up and therefore add together for MAXIMUM AXIAL FORCE.
==== ===== =====
[/tt]
The above shows that the period of the axial force (time between maximums) is 0.01 sec which corresponds to a frequency of 100hz or 2*LF (under assumption that rotor iron is not axially centered on stator iron).

I feel pretty good about my above theoretical discussion of axial force...meaning I don't think there's anything controversial/questionable in it. But whether it really explains what's going on in your particular motor, who knows. The Toshiba paper illustrates how widely different motors can act.


=====================================
(2B)+(2B)' ?

 

[electricpete]

As I often do in these threads, I have more wandering thoughts to add.

We can use the forward / backward waves concept discussed above to analyse a lot about behavior in presence of unbalanced voltages.

[ul]
[li]We have discussed before two different causes of 2*lf vibration... one from off-center airgap (for certain motor winding designs) and one from rotating deformation of the stator core. If we added on top of those conditions a voltage unbalance, I would think that the off-center airgap vibration would be magnified by the voltage unbalance, but the rotating stator deformation might not. [/li]
[li]The forward / reverse waves also explain the 2*LF torque pulsations IF we assume the forward wave is the rotor field and the reverse wave is the stator negative sequence field. The interaction of these two particular fields would give a torque component pulsating at 2*LF. However at no-load conditions, there is no rotor field (not rotor current). So from theory I wouldn't really expect to see torque pulsation at no-load. (maybe it's already obvious there can't be any torque in absence of rotor field). But then again you measured tangential 2*LF uncoupled at no-load, I would think it requires some explanation other than torque pulsations from unbalance (because those don't exist at no-load).... [/li] [/ul]
Following up on the last bullet above, what could be the possible explanations for tangential 2*LF at no-load, possibly in presence of voltage unbalance:[ul]
[li]In a horizontal motor I think the mounting configuration and asymmetric H/V support stiffness can couple radial vibration into tangential (for example rocking mode), but you're talking about a vertical motor so I'll have to toss that idea aside.[/li]
[li]On the "far-out-there" side, I recall that some authors have explained the reason why magnetostriction is a big factor in transformer vibration but not a big factor in rotating machinery vibration is that in rotating machinery there are local expansion and contraction of the backiron as the wave rotates around the backiron, but the total length around the backiron circumference stays constant (contrast that to transformers where the wave doesn't rotate, it's just a stationary pulsating wave so the entire core shrinks and expands along the axes of each core leg). If you believe that, and you realize that forward/backward wave in presence of unbalance in rotating machinery gives an element of pulsating wave in the backiron similar to transformers, then maybe you would begin to expect that the unbalanced-current motor to act more like a transformer and vibrate along the circumferential axis of the backiron due to magnetostriction. Personally I can understand how magnetostriction from rotating wave keeps the total circumferential distance constant, but I never understood how that in itself would translate into lack of vibration. I think that what would be more relevant than the total circumference would be the distance between frame ribs. Maybe (?) the arc distance between frame ribs are selected such that magnetostriction from a rotating wave at fundamental pole pitch always gives a constant length between ribs, thus minimizing transmission of core magnetostrive movement to the frame. For example if you have one rib per pole, there is always one pole between ribs (to be nit-picky when the pole is exactly aligned with the rib you have two half poles between ribs). Or if you have 1 rib per 2 poles there are always 2 poles between ribs. Either way the length of the iron between ribs stays constant as the poles pass by. BUT if you have unbalanced excitation, then you now have forward and backward waves rotating around the back iron, alternately coming in and out of spatial phase at a time frequency of 120hz, the total length around the circumference would vary at 120hz as would the length between ribs (regardless of rib spacing). That would put alternating tangential stresses on the frame ribs giving tangential frame vibration. But again this may be a little on the far-out side, I may be way off base.... for one thing I haven't looked at any motors to compare the number of frame ribs with number of poles[/li]
[li]My bottom line, I really don't have any good ideas about what might be causing the tangential 2*LF that you are seeing at no load. [/li]
[/ul]


=====================================
(2B)+(2B)' ?

 

[edison123]

Thanks, pete.

"Of the remaining suspects [causes for 2LF vib], eccentric air gap, soft foot conditions, loose stator core, interphase and ground faults, unbalanced voltages and loose top covers each have a set of specific tests that can be performed on the motor to either confirm or dismiss a suspected cause…."

We went to site day before yesterday. Reconfirmed the thrust collar at the top was fixed properly and there is no rotor vertical skew. Soft foot was checked for all 12 bolts during decoupled run and ruled out as cause.

The decoupled motor tangential vibrations at the top and bottom and 90 degree apart were high between 4 to 6 mm/sec rms. Radial and axial were less than 1 mm/sec rms.

Voltage unbalance was 1% while the current unbalance was as high as 13%.

Spectrum analysis showed 12x (= 6000 RPM = 100 Hz) contributing 75 to 80% of vibrations in all axes. The tangential vibrations dropped to less than 0.6 mm/sec within 2 seconds of power cut off. I am suspecting eccentric air gap but unfortunately, the OEM has not given any provision to measure the air-gap.

We are still brainstorming this. Your inputs, as always, are very useful. Thanks.



Muthu

http://www.edison.co.in

 

[electricpete]

ok, so now I'm understanding (I think) that your original data was under load, but the latest post includes results of uncoupled data. But the uncoupled data still includes high tangential component with barely any radial. That result is still confusing to me for reasons I mentioned in my last post (I don't think torque pulsations are possible at no load).

It seems like you have checked all the boxes for troubleshooting. Air gap seems like the only practical thing left to check if you can get to it.

Personally I can see how uneven air gap might give 2*LF vibration (especially if the winding includes multiple circuits per phase) and how it that vibration could be magnified by a voltage unbalance. But I don't see how radial offset airgap would give torsional vibration or axial vibration...

On the other hand I'm still leaning toward the idea unbalance would cause axial vibration in presence of axial offset, and it's hard to come up with other explanations for vertical 2LF on a vertical motor. Let's think about possible cause of axial 2*LF on vertical. We talked about axial resonance at 2*LF which we have discussed for horizontal motors... but I've never heard of it on a vertical and seems unlikely to have resonance on both endbells due to difference of construction of upper and lower endbell. You had axial 2*LF vibration of a horizontal motor (if I remember right) where I suggested endhsield resonance check and that was not the cause... did you ever figure out that cause? Going back to my pet theory: I know you said "Reconfirmed ....there is no rotor vertical skew."... does that mean you verified that the rotor iron was centered axially with respect to stator iron? Even if you did, I'm reluctant to give up the idea that the unbalance combined with axial offset is causing it, maybe there are features of the steel frame which are different on top than on bottom which could possibly affect the pattern of the end fringing fields which are contributing to that pull.

Let's go back to the tangential vibration under no-load. That's a headscratcher for me... I threw out one wild guess above (magnetostriction). But another thought... we talked about two different varieties of simple radial 2*LF... one is airgap asymmmetry and the other is rotating p-pole deformation of the stator frame (which ordinarily is less likely at higher poles due to higher stiffness bending to that shape, but lets allow it anyway). I'm wondering if that rotating deformation of the stator core might possibly translate into tangential vibration of the stator frame? I'm not sure exactly how that would work but seems remotely plausible.

If you have multi-channel analyser capable of measuring phase between acceleromters, it might be interesting to check the phase of the 2*LF at point 180 opposite on the stator frame. When radial 2*LF is present (I guess that's only under load), I would think that the off-center airgap will give in-phase motion of mechanically-180-opposite sensors (when corrected for sensor orientation), while in contrast rotating stator deformation would give 180-out phase of mechanically-180-opposite sensors (again corrected for orientation). I haven't given too much thought to what we could learn from the phase of the tangential vibration. Maybe for kicks if you're really bored, map the radial and tangential 2*LF phase around the circumference (on accelerometer stationary, the other one moved in small increments around the circumference to check 2*LF phase from the first) to try to build an ODS of the 2*LF movement. What does it tell you again I'd have to give some more though... I would think a 2-pole pattern of phase variation around the circumference would be consistent with uneven airgap while a 12-pole pattern of phase variation around the circumference points towards all the other things I discussed (rotating stator deformation, magnetostriction) and also that 12-pole pattern would be consistent with torque pulsations in presence of unbalance and some rotor current. Most of this assumes the forces on the core map directly onto vibration of the frame, which may be a wonky assumption... the configuration of frame ribs may have a role to play... likewise it could complicate things if the frame itself if square (vs round).

I guess we could apply the same logic to the axial vibration. If the problem were an axial offset combined with voltage unbalance, then I would tend to expect top and bottom movement to be in phase at 2*LF when corrected for sensor orientation. If they are out of phase, it seems inconsistent with that scenario... I can't come up with any other scenario at the moment. You could also check magnitude and phase at a few points on the each endbell to see if there is any pattern emerging there.

Those previous two paragraphs are a little bit pie in the sky. I was just thinking about what else could possibly be checked after you checked just about everything else. It's up to you whether you want to invest the effort... I doubt it would change any of the actions that could be taken for this motor but it might possibly help give a dubious theory to explain it or if you're lucky maybe even a plausible explanation.
=====================================
(2B)+(2B)' ?

 

[waross]

A thought. I was under the impression that the rotor in a vertical motor was held in the magnetic center by the magnetic field.
The thrust bearing should be low enough to allow for some heat expansion or it may be pushing the rotor off of the magnetic center.
Any variation in the magnetic field will allow gravity to move the rotor slightly and result in vertical vibrations.
The vertical vibration may be an indication that the cause is magnetic, not mechanical.
Has a poorly cast squirrel cage been considered?

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[Strong]

Test for a natural frequency of motor frame that is a twisting mode. Test with impact force-response at top of the frame in tangential (corner) location. Conduct operating deflection shape (ODS) test while running to map relative motion of motor frame/case. I found and solved this issue on a vertical motor-pump at nuclear power plant many years ago.

Walt

 

[edison123]

Thanks Pete, Bill and Walt (long time, so see)

Sorry for the delayed response. Was busy at site re-erecting the motor and taking a few decoupled trial runs. After motor re-erection, the tangential vibration started from 1.9 mm/sec rms from cold condition and went up to 5.6 mm/sec rms after 5 hours at the bottom. 12x (100 Hz) was contributing 80% of all tangential vibrations. Radial and axial remained under 1.5 mm/sec rms throughout.

Yesterday, since the radial and axial vibs were very low, I asked the client to stop wasting time about the 'high' tangential vibrations (which in any case is not defined in any standards) and to couple the pump and take full load trial.

And boom, in coupled full load condition, all four tangential vibrations (at top and bottom at 90 degrees apart) dropped to less than 1.5 mm/sec rms at the cold start and actually reduced to 1.1 mm/sec rms after 2 hours and stayed there after 7 hours of continuous full load run. Even radial and axial vibrations were reduced from 1.5 mm/sec decoupled to under 1.2 mm/sec rms under full load.

I even asked the client to start another adjacent identical pump and run it on full load to see if it had any effect on my motor's vibrations and found it had no effect at all.

The client is very happy since this particular pump has a history of high vibrations and they have never seen such low vibration values. I guess this is one of those rare cases, where coupling the motor to load attenuates whatever the bugaboos the motor was having.

Pete - I am sure you will have a go at it on how the hell coupling the motor to pump reduced the tangential vibrations so drastically. I look forward to it.

Muthu

http://www.edison.co.in

 

[electricpete]

Thanks for that feedback. That's a long crazy case study, who knows what the answers are.

Pete - I am sure you will have a go at it on how the hell coupling the motor to pump reduced the tangential vibrations so drastically. I look forward to it.

Yeah, that's a head scratcher considering I don't know how we had tangential 2*LF at no-load to begin with.

One thing I'll say is that as load increases, the fundamental airgap flux density goes down due to increased current through the stator leakage reactance.
In other words as we increase load current, there is more stator voltage drop due to things like flux encircling the stator end-turns (not linked to the rotor) which means less voltage remaining to create magnetizing current and magnetizing flux. (in terms of the equivalent circuit, the series stator leakage reactance is upstream of the parallel magnetizing reactance branch). That decreasing flux density would affect many of the flux paths in the machine, presumably including whichever fluxes are causing this vibration. It would include the backiron fluxes that I postulated could be related to magnetostriction tangential vibration.... but in that case why was it changing over time as things warmed up? ... I have no clue.

I'm curious if you have any ideas of what we can learn from this particular machine?

=====================================
(2B)+(2B)' ?

 

[edison123]

Pete

All AC machines (both induction and synch) and DC machines (except possibly series field) are constant flux machines as long as V/Hz in AC machines and field AT in DC machines is constant. So change in flux from no-load to load can be ruled out.

While I can say the 100 Hz in this case was most likely caused by uneven radial air gaps, I have no idea (or theory) how it vanished when the motor was coupled.

Muthu

http://www.edison.co.in

 

[waross]

"Constant flux machines."
I understood that the flux path in the air gap stretched and lengthened under load.
Hence commutating poles in DC machines and a change in reactive current under load in AC machines.
Does the flux remain constant despite the distortion in the air gap?

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[edison123]

Bill

Yes, the flux gets distorted by armature reaction under load and the compoles aka interpoles (and compensating winding in low voltage, high current machines) to correct this flux distortion by armature current (hence their connection in series with armature). The end result is a constant flux for a constant field AT.

The same concept applies to ac machines also like increased field current in synch machine under load to counteract armature reaction.

Muthu

http://www.edison.co.in

 

[electricpete]

All AC machines (both induction and synch) and DC machines (except possibly series field) are constant flux machines as long as V/Hz in AC machines and field AT in DC machines is constant. So change in flux from no-load to load can be ruled out.

Constant flux density with load changes is an approximation. It is a great approximation for a transformer, but not quite as good for an induction motor since the p.u. leakage reactances are higher in motors than in a trnasformers (there is leakage reactance associated with endturns, leakage reactance associated with zig zag flux and other leakage reactances associated with cross slot flux). The load current flowing through the leakage reactance HAS to create some voltage drop which HAS to reduce the voltage available for creating flux to some extent. Again, you can see this again from the equivalent circuit, where the magnetizing branch is downstream of the stator leakage reactance. As you increase the load current, the voltage drop accross the stator leakage reactance increases, the voltage available at the magnetizing branch decreases, and the magnetizing current decreases. This means lower flux densities in the airgap and other locations.

I think the equivalent circuit says it all, anything else I say will muddy the waters. At the risk of muddying the waters, maybe there is another way to visualize it. Lets just look at the endturn leakage reactance since that is the easiest to visualize. Imagine if you could take all the endturns (or at least their voltage drops) and move them external to the motor in the power supply path to the motor (that shouldn't make any difference because they don't affect airgap flux and they don't participate in motor operation. Now you have three air coil inductors external to the motor (one per phase). As you increase the current draw (by increasing the load), there is higher current through those external inductors creating higher voltage drop which reduces the voltage at the motor terminals to some extent (even though the voltage drop accross inductor is in quadrature with the motor main resistive/load voltage which reduces thie effect). Lower voltage at motor terminals means lower volts/hz and less flux.

It is admittedly a small effect. The vector nature of the problem minimizes the change (the leakage reactance voltage drop is in quadrature with the rotor load-related voltage drop which is resistive). Maybe if I get time I will estimate how much the airgap flux density drops on a typical motor from no load to full load assuming constant terminal voltage (*).

* And although I didn't mention/consider it before, it's also possible that the increase in motor load current decreases the actual motor terminal voltage due to increased voltage drop in the power supply path (upstream cables and transformer). The effect of series inductances in the power supply path is qualitatively the same as the effect of series inductance associated with stator leakage flux (they both reduce the voltage available to create airgap flux).

=====================================
(2B)+(2B)' ?