First Turn Failure on submersible pumps

[Marke]

Hi

We are currently hearing stories of a number of first turn failure on submersible pump motors. I know that some of these stories are totally incorrect and the first turn failure is a description given by the pump suppliers for other modes of failure, with no evidence of such, it seems to be a good excuse in some cases.

There are a number of quoted remedies for reducing first turn failure, and most involve the use of dv/dt filters on the output of the VFD to reduce the VFD to a quoted figure when the reality is that it is the peak insulation voltage of the insulation that is the problem.
Reducing the dv/dt will reduce the peak voltage for a given cable length, but from my understanding, quoting a required maximum dv/dt needs to be at a particular cable length and cable velocity factor to have any real meaning. - It is looking like a trend towards proving that the maximum dv/dt is 500V/uS or there will be no warranty on the motor.

Questions are being asked of suppliers about the guaranteed maximum dv/dt of the drive with no information on the installation, cable length etc.

One solution commonly used is a three phase reactor in series with the output of the VFD, sometimes with resistors and or capacitors, often without.
I have not seen any reference to the use of common mode chokes on the output of the VFD being used as a means of reducing first turn failure, but to my warped understanding, we use common mode filters (often referred to as Ferrite Rings) to reduce common mode leakage currents in EMC sensitive environments and also to reduce bearing currents and EDM, both of these are related to the high frequency energy in the output waveform as is the peak voltage impressed on the motor windings.
I expect that the common mode filter as used to reduce EDM would also be effective in reducing the peak voltage at the beginning of the motor windings and therefore reduce the probability first turn failure.

Any comments??

Mark Empson
Advanced Motor Control Ltd

 

[waross]

Is the dv/dt the same at the drive as it is at the motor and are the pump suppliers concerned with the dv/dt at the drive or at the motor?


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

 

[X49]

Replacing submersible motors can be very time consuming and expensive, so I would normally err on the side of caution and use a dv/dt filter at a minimum.

The NEMA MG1 standard for motors has specific recommendations for running motors on VFD's, but these recommendations do not apply to submersible type motors. It is recommended that you follow the manufacturer's recommendations for filtering. A typical dv/dt filter will limit voltage peaks to 1.5 times DC bus voltage at 1000ft, and limit dv/dt to 200V/us, which will meet the requirements of most submersible motors.

There are exceptions, however. I'm currently on a project where the 575V Grundfos well pump motors require a maximum voltage spike of 850V peak phase-phase, which will require a sine-wave filter to achieve, despite a relatively short cable length.

 

[KllrWolf]

That is correct, NEMA MG1 does not address submersible motors besides the shaft and mount face on some of the smaller sizes. It is the environment of the submersible motor that is the problem. The cable has an impedence mismatch at the water/air interface as well as at the motor. As an employee of a submersible motor manufacturer and repair, we can only say that the system designer needs to coordinate with the VFD and filter companies to make sure they get what will match the conditions of the application. As someone who has seen many failures from many applications of a water filled submersible, it is easy to identify when the winding was most likely destroyed due to use of a VFD with mismatched/inadequate filtering versus other electrical problems.

X49: That is a low voltage requirement. I would respond more about it, but I am afraid it would sound like I was talking bad about Grundfos.

 

[ScottyUK]

Hi kllrwolf,

Can you explain a little more about the impedance mismatch at the water / air interface? I had assumed that the cable was screened and / or armoured. Is this not the case, or is there a mismatch even if there is a screen? This latter state is harder to understand.

 

[KllrWolf]

Most of the cable used is not armored or screened. If any is, we are not informed of it, unless the motor is ordered with it. I wish I had more information about any difference if armored/screened cable is used, such as if it made the problem better or worse. The problem is that nobody wants to pay the money for testing when the problem is not that common, and using the proper filtering is considered the cheaper solution that works.

 

[ScottyUK]

I would expect that screened cable would make the problem slightly more predictable. Whether that was better or worse than an unscreened type I'm not sure, but a known problem is usually easier to deal with than an unknown one.

 

[waross]

Given the high capacitive constant of water I would expect the characteristic impedance of the cable to change abruptly at the air/water interface.
Are we talking about the same type of cable? The submersible cables I am used to seeing are composed of three fairly heavily insulated single conductors twisted together.

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

 

[electricpete]

The dv/dt should decrease as the wave propagates along the cable

There is that potential impedance mismatch at motor terminals which increases voltage (and therefore dv/dt). But it is bounded to a factor of 2.

So it seems that specifying a dv/dt at the drive output will limit the dv/dt at the motor to no more than twice as much, regardless of cable configuration (and excluding resonant amplification from synchronized reflections of multiple pulses which Gunnar has said is unlikely).

Also note that max dv/dt and max voltage are both important, but for different reasons. The first tends to stress the turn insulation and the 2nd tends to stress the ground insulation. So it is logical to have a separate limit each.

Sorry if I'm misunderstanding your point/questions.

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

 

[KllrWolf]

The cable used is usually not considered twisted cable. It is usually called ship or marine cable by the suppliers. It is three conductors, with or without ground, encased in an outer jacket. Flat cable is also very common in wells.

The air/water interface is an abrupt change, as is the cable to motor. It is the fact of having two impedence mismatches that makes it so difficult to deal with. We have found that keeping the carrier frequency as low as possible and proper output filtering has been the best way for submersible motors to have a decent life span when on VFDs.

 

[iop995]

There is possible to use integrated motor-VFD? Will eliminate cable issues with VFD output and may use even DC to feed VFD.

 

[KllrWolf]

Unfortunately, that would not work with the motors I deal in. You would have to make the VFD be able to be full submerged in a well. On top of that is the size restriction. We have 300 Hp, 2 pole motors that are designed to fit into a 12" diameter well, with the 4 pole version fitting inside a 14" diameter well. I have looked into trying to design something that would work. It appears to work on paper, but the demand for something like this is so small in the VFD world, the cost would be astronomical, and none of the VFD manufacturers we have talked with have any interest in it. It would still be cheaper to get a VFD and true sine wave filter than to build an on motor VFD. Maybe in the future something will come along to make it practical.

 

[waross]

How is the price and suitability of a transformer between the VFD and the motor? I have seen transformers used for large submersible SAGD pumps in the tar patch.
The supply was low voltage and the motors were high voltage (600 V supply, 4160 V motors).
Some pads used high voltage VFDs and filters.
Some pads used low voltage VFDs and transformers.
These pads were all at the same location.

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

 

[KllrWolf]

I have heard of it being done for voltage purposes, however they still used a filter between the VFD and the transformer. I do not know how well a transformer would preform as a filter on its own or the cost of a transformer versus a filter for just filtering purposes.

 

[MagicSmoker]

The dv/dt should decrease as the wave propagates along the cable

Actually, long cables between a VFD and a motor can act as a pulse forming network, significantly increasing the dV/dt of the PWM waveform. This causes significant AC current to flow through the insulation, resulting in dielectric heating, sometimes to the point that the insulation softens and fails.

As mentioned before, an LC filter on the VFD output can help tremendously. I would not expect a common mode filter to help much unless the failures observed are always between the stator windings and ground (that said, it is perfectly reasonable to precede the LC filter with a common mode filter - which is often just a ferrite toroid through which all of the motor cables pass through).

I primarily design motor drives, and the IGBTs in the one I am working on now can slew 650V in 1.0us, and that is *with* ~400nH of inductance in series with each IGBT's output terminal inside the drive (the inductance is for wave-shaping and dynamic current sharing between paralleled modules). This particular drive is for a dedicated purpose and won't be more than 10m or so from its motor, but if that distance were extended out to 30m or more I would totally expect problems with winding insulation and bearings to occur.

 

[KllrWolf]

Do not forget that each time the signal hits an impedance mismatch, there is a voltage reflection. Normally this would only occur at the motor, but with the submersible, you will get one at both the motor as well as water/air interface. This causes the potential for there to be another distortion bouncing between the motor and the cable at the water/air interface. This is another one of the items I would love to be able to get the equipment and money to test, and find practical ways to reduce/eliminate any negative phenomena.

 

[Marke]

Some interesting points made as usual here, but my major question is whether anyone has investigated the use of a common mode filter as a means of reducing the transient voltage on the windings of the motor.

This is a recognized means of reducing EMC emissions on the output of the VFD and EDM bearing currents within the motor.
The EMC leakage is a function of the dv/dt across the capacitance between the stator winding and the stator frame, and one mode of EDM is dependent on the dv/dt across the capacitance between the stator winding and the rotor of the motor.

This suggests to me that the dv/dt at the motor is somewhat reduced and therefore I would expect that the first turn failure would also be reduced.
I have not seen this put forward as a possible solution, rather I have seen dv/dt filters and sinusoidal filters recommended.
Is this because the common mode filter does not work in this instance, or because no one has tried it out?
I guess it is an out of the square type question.

Best regards,
Mark.

Mark Empson
Advanced Motor Control Ltd

 

[KllrWolf]

From what I have seen, the common mode filter does not prevent failures. From discussions I have had with a few VFD manufacturers, the common mode filter as they build into the VFD does not do anything to the signal, and the same filter would be required if the VFD had a common mode filter or not. It is there to prevent anything on the ground line and reduce bearing failures.

 

[MagicSmoker]

Marke - a common mode filter will have _no effect whatsoever_ on the "legitimate" output of the drive. That is to say, the common mode filter is essentially invisible to the current which exits the drive via one IGBT, then goes through a phase winding in the motor and returns to the drive via a second IGBT. In other words, the forward and return currents cancel within the mutual inductance of the common mode filter core.

If you want to shape the dV/dt of the PWM waveform from the drive and you don't have the luxury of customizing the gate driver circuit, then you pretty much have to put in some series inductance and shunt capacitance (ie - an LC filter) on each phase leg. I usually turn to the good folks at Magnetics, Inc. for high performance inductor/transformer applications first (or Micrometals, Inc. for applications which are, shall we say, more cost-sensitive).

 

[Marke]

Hello KllrWolf and MagicSmoker

Thank you for your comments.
My observations here have been that certain drives are more prone to causing 1st turn failure than others.
With two families, I have yet to see genuine first turn failures, although I have analysed pumps with incorrectly reported first turn failures.
One motor I looked at, which was reported to have failed with a first turn failure, had greater than 1MOhm to earth and perfect winding resistances. The rotor would not turn. The ends of the can did not show any sign of swelling or heating.
The problem was found to be plastic extruding through the stator slots into the gap, jamming the rotor.
The problem was due to the iron loss heating the iron with insufficient cooling, causing the plastic insulation to soften and as it expanded, to extrude into the gap.
Cause, extended periods of very low flow. This was due to a change made in the VFD settings a couple of weeks before the failure, and was proven by the Scada.

My experience has been that the drives with good common mode filtering have been less prone to first turn failures. I do not know if this is coincidence or a legitimate byproduct. From the postings here, it would appear to be coincidence.

Thank you and best regards,

Mark Empson
Advanced Motor Control Ltd