3.5 MW motor starting problems continue... 7

[ters]

A while ago, I started a thread

http://eng-tips.com/viewthread.cfm?qid=235885&page=1

which many of you contributed to, but that story has not ended yet...

So the issue is that a 3.5 MW motor driving a large fan has problems starting. The starting method is the soft starter. Rotational inertia is high.

We discussed several causes, such as inadequate motor, inadequate starter, week and inadequate power system, etc. But calculations show that although things are marginal this still should work, but it does not.

Today we managed for the first time to put two identical 13/20MVA transformer in parallel. Their impedance is 10.5% at 20MVA but each transformer also has a current limiting reactor 6%. However, 20+20MVA system still would no start a 4MVA motor...

System: two 13/20 MVA transformers in parallel, + about 5 MW of additional load shared between two transformers.

Motor: asynchronous, TECO-Westinghouse 3.5MW, In = 385Amp, 6kV, 50Hz, 1485 rmp, locked rotor current 2300Amp.

Fan: Large rotational inertia. The fan has no load at the present (no air, dampers closed). The Impeller weight is 15,000 lbs, and the Impeller moment of inertia (WR^2) is 66,000 lb-ft^2.)

Starter: constant current, voltage ramp ~10 sec, current limit 470%, starting time 50 sec, after which bypass contactor closes (when then speed is about 80%).

We had some luck starting and running the motor using one 13/20 MVA dedicated transformer but now with 2 transformers in parallel it would not work. We tried three times and none of the starts was successful. Each time the soft starter was manually shut down after about 20 sec since motor was creating too much vibrations and was not accelerating any more.

Problems still could be electrical, but for a change we started to suspect that there may be some mechanical problems, perhaps related to the motor rotor axial movement. If the rotor indeed does try to move, either it may not have enough space to move to the electrical center or the shaft is moving back and forth around its electrical center for some unknown reason and hitting the trust.

The attached charts provide more info. During the voltage rump time, which was about 10 sec, motor runs smooth and with two transformers in parallel the initial voltage drop is rather modest - falls from 6.4kV to 6kV. This is good news, with one transfomer we had much larger initial voltage drops. However, the bad news is that after the voltage is ramped up high enough that the current reached the preset limit (470% In), some vibrations start.

As you can see, the current and voltage outline forms on the charts start at some point having "teeth" like a saw. Once stabilized there is about 3 - 4 “teeth”/second. Just judging by the ear, the frequency of mechanical vibrations is in the same range, meaning current spikes likely correspond with frequency of mechanical vibrations.

Any idea what might be happening?

 

[ters]

,,,

 

[ters]

.,.

 

[Marke]

Hello ters

Perhaps I can join in very late in the discussion with a few questions.
1. I understand that the cables between the starter and the motor are quite long. Are these screened or armoured cables?
2. What speed does the motor/fan get to when the vibrations begin? Is it always the same speed?
3. Is there any natural mechanical resonance in the fan? If so, at what frequency?
4. What is the susceptibility of the motor to harmonics?

Looking at one of your earlier graphs, the oscillation in the current is very sinusoidal suggesting that there is some form of resonance going on.
I have found a situation many years ago, starting a large crusher where the motor would accelerate to part speed and then would vibrate significantly with the shaft appearing to move axially. The motor was dismantled and the rotor was found to be slightly off round. A few thou were skimmed off one side of the rotor, it was reassembled and worked perfectly from then on.

I have also recently been involved with a large fan on a VFD which had a natural resonant frequency within the operating range of the fan and this interacted with the open loop vector control algorithm to cause torsional vibrations which destroyed the fan.

I have also seen similar results when the motor and load are not perfectly aligned, and also when the mounting bed for the motor and/or the load are able to flex slightly.

As this is occuring at a low speed, you should not have a torque issue. What happens if you attempt a start at a much lower start current? does the motor still vibrate at the same speed, or does it pas through this speed?
I would suggest that you try at say 250% start current.
I would also have a series of video cameras and strobes etc to see if you can get a good picture of the mode of vibration and see where it is occuring.

Best regards,


Mark Empson
L M Photonics Ltd

 

[ters]

Mark,

Thank you very much for joining again. I will try to answer rest of your questions tomorrow, it is already late here in Europe. Re cables, there are three cables in the circuit:

1. The cable from the soft starter to the motor is actually very short, only about 15m, 2x120mmsq per phase, screened.

2. The cable from the process switchgear (which is to feed two identical 3.5MW motors some day...) to the soft starter is also very short, only about 20m, 2x120mmsq per phase, screened.

3. The cable from the 110/6kV substation switchgear to the process switchgear is long, ~ 400m. This cable is a bit strange mix of cable types. In the past, there was one three core armored cable 3x240mmsq, which was sufficient then. With new big motors, the power needs changed, so to meet it we added more cables to the same feeder. That feeder now included the original cable of 3x240mmsq three core + new cables 2x240mmsq per phase single core. Could this mix of long three core + single core cables be causing or contributing to some resonance phenomena?

Our only attempt so far to eliminate something which might theoretically contribute to electrical resonance phenomena was to eliminate current limiting reactor (at the substation swithgear - before long cable), but that made no difference.

However, it is now very hard to draw any conclusion what does or does not make a difference since meanwhile our problem progressed from occasional successful starts to no successful start... So, if something failed meanwhile, the problems is probably the same with or without reactor. We have not started motor for about 3 weeks as the substation is not available - due to a major upgrade.

Unfortunately, we do not measure the speed religiously during each start (only manual measurement is available), but it does seem that the motor stalls at about the same speed all the time. That speed may be in the range of 300 RPM. We have not attempted to change the current limit for some time, so we can try lowering the limit and see what is going on next time. We will be in position to do it about one week down the road when the substation should be back online.

Best Regards

 

[edison123]

There are no signs of overheating in the rotor end rings.

While you are at it, could you note down the no. of rotor bars and no. of stator coils. ?

 

[ters]

Here is another chart from my selection... This one is a bit better than those I previously posted since the chart recorder paper speed was 100 mm/sec which provides enough resolution to see at least something - 50 cycles fits in 100mm (4”) of paper.

The top chart is the line voltage; the mid chart represents the voltage on the motor terminals; while the bottom chart is the motor current.

If you take a closer look at the motor voltage, you can see that SCRs, for some reason, for about 8 cycles are chopping only positive half periods, while let negative half period pass through NON-chopped, and then it switches around, so next 8 cycles SCRs are chopping only negative half periods while positive ones are let through NON-chopped, etc.

I think that non-chopped sinusoidal half periods are those which are transparent all the way to their tip, while those which are chopped have black shadowed tips.

Consequently, the whole current chart (or one could say the reference for the current form) oscillates up and down – goes up when SCRs are chopping only negative half periods and goes down when SCRs are chopping positive.

This seems to indicate that something may be wrong with the soft starter, unless the soft starter behaves so due to some severe motor or system asymmetry.

This chart captures the time from 21 to 24 seconds after that start. It was long after the motor stopped accelerating and just stalled at an estimated speed of 150 RPM.

Rest of the chart, which I will post below as well but at a reduced resolution (too long, about 2 meters) suggest the following:

1.The motor accelerated normally and smoothly during first 6 seconds. This approximately matches voltage ramp up time. After 6 sec, the speed was about 90 RPM. During this time SCRs were chopping the voltage (firing) symmetrically - all half periods seems to be equally chopped, either positive or negative ones.

2.Then, from 6th to about 12th sec SCRs would seem to decide NOT to chop some of the sinusoidal half periods. Initially, around 7th second, it started with one positive half period being non-chopped, followed by one negative half period non-chopped, etc, then about 9th second two of each would seem to be non-chopped, then 3 half periods were not chopped, etc, etc. It seemed that during this time the motor was still accelerating but at a much slower pace than during first 6 seconds and it started to develop severe mechanical vibrations.

3. Finally, at about 13th second, the motor completely stalled and an estimated speed of 150 RPM. After that, the frequency at which the thing alternated chopping only positive or only negative half periods remained steady - about three “periods” per second, whereas the “period” is defined by about 16 normal 50 Hz cycles. Mechanical vibrations continued at the constant frequency.

Does this provide any more light to explain what is going on and why?

 

[BJC]

What kind of fan is it trying to start?
Some axial fans have strange problems if started into a closed systems. The air coulumn in front of the fan compresses and unloads. You may need to find a smart fan person to explain it. I have seen fans shed blades because of it.

 

[ters]

Here is the entire unsuccessful start (0 - 24 sec). Could not use Engineering.com to upload it (looks like the file is too big) but this link seems to be working:

http://www.sendspace.com/file/c3wfu5

We are starting a centrifugal fan with the moment of inertia of about 70,000 lb-ft^2 (3000 kg-m^2), Double Width, Double Inlet arrangement.

Edison, I will try to count the bars some other day.

Mark, I could not get info today on natural mechanical resonance in the fan and susceptibility of the motor to harmonics, will keep trying...

 

[Marke]

Hello ters

From your latest graph, there certainly seems to be a sinusoidal oscillation that is apparently affecting the firing of the SCRs.
It is difficult to determine what is happening from the graph as it is too compressed. I would like to see the graph showing just half a dozen cycles so that we can see the actual conduction and commutation angles with some detail of the motor current which is definitely not sinusoidal as implied by this graph.

Is it possible that the problem is occurring at 136RPM?
This is one eleventh of the speed and is one of the speeds where cogging can occur due to harmonics. The seventh is usually another difficult spot.

If the motor has a harmonic problem at this speed, then I would expect to see a sinusoidal modulation of the commutation angle of the SCRs which could explain what you are seeing. If you study the approach to this speed, you will probably see the modulating frequency reduce and amplitude increase as it approaches this speed.

I believe that this characteristic is altered by the design of the motor with the skew of the slots and the ratio of stator slots to rotor slots.

If there is a harmonic interaction going on with the motor locking on to a harmonic, then I would expect the problem to get worse as the supply impedance is reduced. At a higher supply impedance and a given current, the conduction angle of the SCRs would be greater and the harmonic content would reduce, reducing the problem.

Best regards,

Mark Empson
L M Photonics Ltd

 

[Marke]

Hello ters

From that latest graph, you can see the oscillations starting at a high frequency and low amplitude and the beat frequency reducing and amplitude increasing as you approach the critical speed.

I am pretty sure that you have a critical speed that is either due to mechanical resonance, or harmonic interaction in the motor. Check the shaft speed and see if it is almost 136.36 RPM (1/11 synchronous speed.)

Best regards,

Mark Empson
L M Photonics Ltd

 

[ters]

Mark, thank you very much for all replies and explanations.

I have found a situation many years ago, starting a large crusher where the motor would accelerate to part speed and then would vibrate significantly with the shaft appearing to move axially. The motor was dismantled and the rotor was found to be slightly off round. A few thou were skimmed off one side of the rotor, it was reassembled and worked perfectly from then on.

Our shaft does tend to move axially towards the fan during the start, but it does not have much room to go - the play in the coupling is small, just enough to allow the shaft to move a bit to the magnetic center depending where it landed during the previous stop.

But, again, why a slightly elliptical rotor would cause the soft starter to, as it seems to be doing when the speed stalls, chop 8 or so positive half periods but conduct the entire time during corresponding negative half periods. and than do another way around...? This phenomena is like superposing normal 50Hz voltage with something like 3Hz voltage of lesser amplitude... so it is probably not due to rotor shape.

Is it possible that the problem is occurring at 136RPM? This is one eleventh of the speed and is one of the speeds where cogging can occur due to harmonics. The seventh is usually another difficult spot.

It is possible. For the last chart I posted, we unfortunately did not measure the speed, but we modeled the motor start and it seems that the model, which tells us what is the speed each second during the start is OK - matches the reality when we have a successful start. So based on that, we can be more or less sure that before the problem starts to develop when the motor reaches at least 90 RPM. It does probably catch a bit more speed in the next second or so, so maybe it is really around 136.

Next time we start we will measure the speed along with voltage and current using a better recorder.

If there is a harmonic interaction going on with the motor locking on to a harmonic, then I would expect the problem to get worse as the supply impedance is reduced. At a higher supply impedance and a given current, the conduction angle of the SCRs would be greater and the harmonic content would reduce, reducing the problem.

If we did not do some damage (or more damage) to the motor and/or the soft starter due the repetitive harsh starts, than your above explanation may be exactly what we are going through…! We were able to start the motor before (more frequently than not), but once we either decided toput two transformers in parallel or reduce the impedance of one transformer circuit by eliminating its current limiting reactor, than there is no more successful start.

So assuming that you really nailed it down, what would be our next step, what measurements we should do to confirm that this is the exact problem, and if yes, how to deal with harmonics in our situation, i.e, what would be a cost effective and practical solution which would not be as demanding as getting a new motor of different design?

 

[Marke]

Hello ters

If there is a resonance occurring, and the graphs strongly suggest that this is the case, then any feedback system can amplify that resonance if the poles occur at the wrong frequencies and this can make a marginally stable system very unstable. Modifying the gain and frequency response of the feedback system can often shift the poles sufficiently to soften the problem and allow the system to pas through the resonance.
In this case, there is a closed loop, the current control loop and it is possible that this is amplifying the "resonance".
If you revert back to a voltage ramp start with a start voltage at 60% and a ramp time of 120 seconds, you should have an approximation of a current limit at around 400%, slowly increasing.
If the motor accelerates through the "bad" patch, then you have a potential solution. If the motor still vibrates at around the same speed, you have a problem and I would suggest talking to the suppliers about the harmonic susceptibilty of the motor and investigating resonant frequencies of the mechanicals.

Best regards,

Mark Empson
L M Photonics Ltd

 

[electricpete]

If there is a resonance occurring, and the graphs strongly suggest that this is the case,

Sorry to stick my nose in. Maybe I missed something. Which graph are we talking about that strongly suggests resonance? The graph I saw showed current oscillating at a frequency that decreased as the machine accelerated (once again as is expected for rotor defect although inspection doesn't show that... yet... still would really like to see that single phase test).

Resonance means a single frequency where the system gives a large response. Why would the frequency of oscillation change like this for a resonance? For any type of frequency I am familiar with the frequency is relatively constant, not smoothly varying.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

Correction to my last sentence in bold below:
"For any type of resonance I am familiar with the frequency is relatively constant, not smoothly varying."

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[edison123]

Mark

Doesn't resonance require a capacitance ? Or you think the cable and stator winding capacitance is enough for a resonance ? Given the cable length, it is a possibility ?

 

[ters]

Mark, thank you very much for the reply again. We can certainly try a long ramp time and see if it stalls at the same or different speed. We will be in the position to try it maybe a week down the road and I will report back… We hope to be able to measure more things then (and we know now more what to look for) and try starts with different power configurations to see how it affects the start.

Gents, maybe “resonance” was not the most descriptive word for what is going on, nor I think even “oscillations” is the best qualifier for what we see on the chart since the RMS measurement does not show much of the current oscillations. It is only the chart recorder graph showing that most of the time SCRs tend to chop the voltage unequally in one single period, making so the amplitude of the negative vs. positive half period different. Mark did explain in a way which I understand as 50Hz being like a carrier and it is then being modulated by the amplitude due to certain harmonics.

If the motor has a harmonic problem at this speed, then I would expect to see a sinusoidal modulation of the commutation angle of the SCRs which could explain what you are seeing. If you study the approach to this speed, you will probably see the modulating frequency reduce and amplitude increase as it approaches this speed.

But we would need to prove this “beyond reasonable doubt” before we can search for an appropriate solution.

Mark, I do not seem to understand another part of your expanation. If the motor stalls at say 136 RPM, which is one eleventh of the synchronous speed, how do we know which harmonic is causing this? Or to rephase the question in the most general terms, what is the relationship between N fraction of the speed with Nth harmonic, if any?. Or in our case, which harmonic would cause such modulation which makes approximately 8 periods “more positive” and then next 8 periods “more negative”? What I see from the graph is like 50Hz being modulated by something close to 3 Hz...

 

[LionelHutz]

If there is a resonance occurring, and the graphs strongly suggest that this is the case, then any feedback system can amplify that resonance if the poles occur at the wrong frequencies and this can make a marginally stable system very unstable. Modifying the gain and frequency response of the feedback system can often shift the poles sufficiently to soften the problem and allow the system to pas through the resonance.
In this case, there is a closed loop, the current control loop and it is possible that this is amplifying the "resonance".

Whatever you want to call the effect, I would agree with this. It's very likely the problem is an unstable current control loop in the soft-starter. Some unstability in the connected motor system is being amplified by the current control loop. Fix the current control loop and the motor will likely start.

Unfortunately, if you use an open loop voltage ramp and starve the motor for voltage then it can also exibit a similar unstablility.

The number of smart people who understand the feedback theory yet still don't properly impliment a digital feedback loop might surprise you.

 

[ters]

LionelHutz, thank you for your comments. We just received an e-mail from the soft starter supplier, with good comments, but does not point out a possible problem with the soft starter loop control, which you mentioned.

Naturally, as long as there is anything else that may contribute to this phenomena, equipment manufacturers (either motor or soft starter) will continue pointing to that as the root cause. Their explanation is more in line with Marks explanation that the root cause are harmonics but they are pointing to a bit different stall speed/harmonic and it also goes beyond that explaining that the power system is the main culprit which happens to be just tuned to oscillate. I will take a liberty to copy/paste what they said:

Quote:
In reviewing the waveforms and information you have supplied, the soft starter actually is firing at all times, as evidenced by the continuous output seen on the current waveform at the bottom of the traces. This holds true because the soft starter is a current-controlled device, which means it uses current feedback to monitor the output of power sent to the motor, so current is a good indication of our firing control.

Now, that being said, it can be seen that as the motor is accelerating, it eventually reaches the stall point as you have mentioned. In calculating the period of the oscillation against the source frequency ("about 16" cycles as you stated in Item #3 below), it comes out to a "Triplen Harmonic" of 16.67Hz in relation to the 50Hz source frequency. This means that the resonant point of the motor RPM is quite likely to be near 150 RPM, which is the extended triplen frequency of the RPM versus the source frequency (16.67Hz x 3 squared).

If you look at the 7 to 8-second point of all three waveforms, you can see where the resonance of the motor is beginning to impress itself on the source voltage waveform, the motor voltage waveform and the motor current waveform. As the resonance builds up along the timeline within the motor voltage waveform, it reinforces itself, gaining amplitude and frequency, until it remains at the same values from 13 seconds to 24 seconds. Interestingly though, the resonance is superimposed within the motor voltage, yet can be seen to only affect the p-p values of the source voltage and motor current waveforms. This resonance will create an off-frequency, counter-rotating torque in the motor thereby causing the mechanical vibrations you describe. Also, as you reach the triplen harmonic, you may find that there are high currents in the neutral return conductor due to the high return currents being generated by this issue.

To summarize, the overall power system is acting like a tuned circuit and as you slowly build up RPMs in the motor during starting, the motor will stall as it becomes loaded down within the resonant waveform.

From my standpoint, there may be a couple of ways to deal with this issue. One, is to change the reactance in the circuit, which may be something as simple as jumpering out the 6% line reactors at the feeder. The second possibility is to accelerate the motor as hard and as quickly as possible to try to break past the triplen harmonic point in the acceleration curve of the motor before it loses so much torque it can no longer accelerate the load. Or perhaps, a combination of both will do.

We have seen this issue once before and even after starting the motor across-the-line, the motor still had an oscillation like the one you show on the charts. The fix for that one was that they had to change the feeder system in order to break the harmonic tendency of the overall system to oscillate. I can't say exactly what needs to be done to fix your situation though, as harmonics are never an easy thing to deal with and the conditions vary from site-to-site.
End of quote

 

[ters]

Just to note that although the above explanation does seem to talk about one possible cause in a convincing way, the part I'm not sure I understand at all is:

Also, as you reach the triplen harmonic, you may find that there are high currents in the neutral return conductor due to the high return currents being generated by this issue.

Our 6kV system has no natural conductor. And even if it had one, what it would had to do with a 3 phase motor?

The 110/6kV transformers are WYE-WYE, secondary not grounded.

 

[electricpete]

I just looked at the charts posted 27 April. That is a revealing closer look where it shows the envelope is alternating positive and negative - not the simple amplitude modulation that appeared from the zoomed-out earlier charts that I associated with rotor problem.

Hope you don't mind some rambling questions. That post about triplen harmonic completely lost me. 16.67 is not a triplen harmonic of 50hz, it is 1/3 of 50hz. 16.67*2^2 is on the other hand third harmonic.

If all three phases are identical shifted by 120 degrees of fundamental, then the triplen harmonics of the three phases are shifted by 2*120 = 360 degrees from each other, in other words in phase - zero sequence currents. Those would flow in a neutral connected back to the source but as you point out there is no neutral connection on wye connected motor. On the other hand if connected in a delta winding connection (maybe that's what y9ou have?) you could have zero sequence third harmonic currents flowing in the three phases leading to the motor.

And how did he start with an observation about 16 cycles length of envelope and conclude there was something at 16.667 or a triplen harmonic?

=====================================
Eng-tips forums: The best place on the web for engineering discussions.