Possible reasons for VFD line reactor failure?

[jraef]

This is a new one on me. Someone called wanting my help on figuring out why they have had 3 failures of Line Reactors ahead of some 40HP VFDs. 30 years in this industry and I have never once seen a line reactor fail. It's too stupid of a device; wire and iron. What could fail? So far no details, I have a meeting in 2 days in which the customer will give me part numbers and mfrs of the components in the system. Until then, I'm just trying to ferret out possible directions to go with this so that I can sound reasonably intelligent when I talk to them.

So in looking around, I notice that all of the big names in reactor suppliers all say their reactors are "harmonic compensated". I've seen that forever, always just took it at face value. But does that imply then that there is such a thing as a "non-harmonic compensated" reactor? My thoughts are that possibly whomever built this system found some el-cheapo reactor supplier and they are not up to the task, i.e. not "harmonic compensated" somehow. One thing that leads me to that already is that in the sketchy info I have so far, they said they are "2% reactors", and nobody I know of offers 2%, they are either 3% or 5%. So my thought is that if they somehow found a cheap source, it may be that the reactors were not intended for use as line reactors for VFDs, i.e. not harmonic compensated.

Any other ideas? Are there some bizarre interactions with something else, similar to capacitor resonance, that could be causing the reactors to fail? Anyone else see reactor failures?

And yes, my first line of reasoning, until I get numbers and data, is that someone simply picked reactors that are too small. That just seems too obvious though. Considering they have replaced them twice now, one would think that they would have checked that first. Still, I've seen people do things like that...


"You measure the size of the accomplishment by the obstacles you had to overcome to reach your goals" -- Booker T. Washington

 

[FreddyNurk]

I've got an anecdote, nowhere near a line reactor failure, but related enough that it might shed some light on other aspects.

We'd developed a platform that used a particular make of bidirectional drive as an energy dump device under certain operating conditions. Part of the drive equipment included a capacitive filter. Our system was designed to have multiple units in parallel at the site(s).

Subsequent field operations observed overloading of the wiring feeding the capacitor. A number of false starts and distracting conclusions were developed, but we finally worked out that the filter was being overloaded by harmonics, way in excess of design criteria.

The eventual conclusion was that the switching frequency of the inverters was being 'caught' in the filter as it was supposed to, but operating multiple units in parallel gave harmonic currents greatly in excess of what was expected. Whats more, as each drive had its own switching frequency reference, the observed current would rise and fall as the harmonics came into sync on the unit under observation, only to disappear again a second later, which effectively made it impossible to accurately detect on a current clamp.

None of the currents were observed at that magnitude during testing as all the testing had focussed on single units rather than multiple units. In this case, it might be that the reactors are appropriate for single drives but possibly not for multiple units in parallel, although I would expect that the inverter manufacturers would be reasonably aware of that issue, something that wasn't apparent in our case.

 

[davidbeach]

So, in situations like this, question #1 is "Define Failure". With out a failure mode all else is rank speculation.

Funny how I can easily picture you writing that sentence.

Air core? Iron core?

Fused open? Molten puddle on the floor?

Mid coil "failure"? End of coil "failure"?

 

[jraef]

Yes, "define failure" is on my list right below the request for data on part numbers and mfrs. I'm dealing through our salesman who really doesn't yet know how to ask probative questions properly (he's young). That's why I wanted to meet with them myself.


"You measure the size of the accomplishment by the obstacles you had to overcome to reach your goals" -- Booker T. Washington

 

[jraef]

OK, had my meeting, this just got really fun.

Background:
2ea 150HP 480V VFDs added to an existing MCC enclosure by a panel builder, each with their own dedicated line reactor. Consultant specified 5% line reactors, for some unknown reason the panel builder used 2% reactors.

Issue:
What happens is that if they run both motors, no problems. But if they run only one motor, the line reactor in the OPPOSITE (idle) drive cabinet heats up and catches on fire! So it cannot be that the reactor is under sized because that motor is NOT running, ergo it is not drawing current from the VFD. 30 years in this industry and I have NEVER seen a line reactor fail in any way, let alone catch on fire, and to have it catch on fire when the VFD is feeds is idle?
Mind blown...

What I HAVE seen on two occasions was what I assumed was a resonance taking place between two drives, where one drive running caused an adjacent drive that was NOT running to trip on over voltage. That I can understand, based on an interaction between the harmonics of the running drive getting into the idle drive capacitors and over charging them. But in both cases where I have seen that there were NO reactors and adding 3% reactors was the cure. So my speculation is that this is something similar, but because the reactors are only 2%, it isn't enough(?). The thing that doesn't fit is that the idle VFD is not tripping on under voltage,. The ONLY symptom is that the reactor catches on fire!

PS: Nobody was sure where the reactor failed (mid-core, etc.) because of the fire. This has happened repeatedly by the way, the last one caused damage to the VFD. Apparently the first two times they assumed it was faulty reactors.

I love the smell of burning brain cells in the morning...


"You measure the size of the accomplishment by the obstacles you had to overcome to reach your goals" -- Booker T. Washington

 

[Skogsgurra]

A safe and very conclusive measurement would be to check the flux in the reactors. One turn to pick up induced voltage and a simple 100 ms passive integrator will show what happens with core flux.

It sounds like a core heating thing. To be able to judge if that is the case, one needs to look for capacitors (or long cables) to resonate with. Reactor/capacitor combinations often have a tendency to ring somewhere between 5 and 20 kHz and that is an extremely effective frequency when it comes to induction heating. Have seen that a few times and the simple remedy was to put resistors in parallel to the reactors. The resistors reduce Q so the ringing doesn't heat the cores.

I do not think that this is an overload issue, current-wise. The scope is your best friend.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

Here's my theory:
The second drive is in-fact running.

But.

Not correctly. Its output stage(s) are being triggered directly via conducted or radiated emissions from the running drive. The output is nearly random and poorly phased causing major dissipation in the reactor but nothing coordinated that can spin the motor. The motor, being able to dissipate the most energy in the setup, is being abused but has enough thermal mass to not make-a-scene. The non-operating drive can and in this case was damaged by its output stage's uncontrolled operation.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[ScottyUK]

The reactor heating presumably requires current flow. An isolated current probe with decent bandwidth clipped around each of the reactor leads may provide some insight to the origin of the signal.

 

[Skogsgurra]

Careful! I had very little current flow. In my two cases, the heating was purely inductive and less heating when there was load (damping) in the circuit. Also, if you use a current clam, make sure that it has enough band-width. Most current clamps can't show 20 kHz correctly. LEM is almost there, PEM can and i-Prober can. Fischer HF can, yes, but have a high lower BW limit so they will usually not show 20 kHz. The HP and TEK clamps are OK to use.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[ScottyUK]

Yes, I agree that sufficient bandwidth is essential, otherwise it's easy to miss things or for the signal to be something other than how it appears due to transducer limitations.

 

[jraef]

I'm no longer allowed to do any field labor type of work ("we have people for that") so my role is restricted to things I can see, in person or on a screen/paper. But the more I research Gunnar's angle, the more I am compelled to search for (or have my field techs search for) capacitance in the circuit. I'm going to be studying the known one-line diagrams when I get them, if there is nothing on there, I'm going to go physically look around the site. If I can't find any immediate evidence, at least I know what to tell the techs to be looking for.

I'm suspecting something along those lines, but its always resulted in DC bus overvoltage issues in the past, I was unaware of this inductive heating phenomenon. In my meeting with their Engineer the other day it came out that these drives were retrofitted into a situation where they did not have VFDs before, so we are going to look to see if they maybe had bulk or automatic PFC caps on the line side in the original installation and nobody removed them when they changed to VFDs. In one DC bus over voltage issue i dealt with in the past it turned out to be utility line capacitors that they categorically and repeatedly denied existed, until I found them, took pictures, and demanded they explain to me what they were. Then they said "Oh, you mean THOSE capacitors... Well yeah, we have those." The utility added de-tuning air core reactors to their (non-existent) cap bank and the problem went away.

This has been excellent, thank you! If we can't find anything, I'm going to suggest they hire you (Gunnar) to come over here and find the issue. At the very least, we'll get to visit again! I've dropped your name and contact info several times in the last few years, so far no takers, but I keep trying...


"You measure the size of the accomplishment by the obstacles you had to overcome to reach your goals" -- Booker T. Washington

 

[djs]

I have seen some line reactors that had RC circuits associated with them. This is probably the "harmonic compensation". I have had one such reactor catch on fire, but that was caused by poor wire dressing where an insulated wire was touching one of the power resistors. It maybe possible if the harmonics on the buss are very high with just one motor running, the RC circuit will over heat and start a fire. Just something to check for.

 

[jraef]

The ones with what would look like an RC circuit were likely passive harmonic filters, slightly different animal but they do include a reactor. That RC circuit would be tuned to absorb the 5th, plus maybe the 7th harmonic too (depending on what they were willing to spend). They are considering going that route (at the suggestion of a filter vendor), but I'd rather find the real cause before throwing hardware at something. Scatter guns are good for desperation defense, not for hunting.


"You measure the size of the accomplishment by the obstacles you had to overcome to reach your goals" -- Booker T. Washington

 

[Skogsgurra]

Agree. If there is an R involved, it usually means that the resonance either is at a quite low frequency or not present at all. It is parasitic capacitances that should be sought after. Or it could turn out to be something entirely different. Not at all unusual.

Re having a look at the problem on site: Thank you, Jeff. Nice of you to mention my existence. And it would have fitted very well in my plans for this spring, which already included a visit the US. But that trip may be cancelled because of the unpredictable situation with the HSA. I once did a job in Egypt and that caused some problems when I visited the US later on. I have visited the US several times after that, one or two times/year, and I have not had any difficulties at immigration. But the new president has made it a risky business, even for a white Swede with a clean record and family traceable back to around A.D. 1600, to cross the border.

Otherwise, I have all the tools needed (lightweight, in carry-on luggage) for a job like this. Transportation across the North Atlantic is in the order of USD 400 and I have been travelling all over the world all my life. So that wouldn't be a problem either. But the situation in the US doesn't feel good at all and that makes me hesitate, even if the task sounds very interesting and is in line with what I do for a living.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[waross]

A question friends.
When you tune a VFD harmonic filter is it tuned for a harmonic of 50 Hz or 60 Hz or for a harmonic of the switching frequency?


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

 

[Skogsgurra]

Always n*50 or 60 Hz.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[waross]

Thanks Gunnar. That makes sense for the line side of the VFD.
That leads to another question.
When a filter is installed on the output of the VFD, what frequency is it tuned at? How does the filter cope with the varying output frequency of the VFD.
Respectfully.

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

 

[Skogsgurra]

That depends a lot on what kind of filter you use.

The simple motor reactor (a little like a line reactor, but less H) is not tuned at all. Sometimes it needs damping, and that can be achieved with a parallell resistor or (seen that on Siemens drives) by feeding the overshoot via diodes back to the DC link.

Then you have the common-mode filter. Again, no tuning. Just a high-permeability toroid (several, usually). Most guys call it a "ferrite ring", which sometimes is true and sometimes not. A nano-crystalline amorph band ring is much more effective and saturates at much higher levels (up to 1.8 - 2.0 T) than a ferrite with typical values around .3 - .5 T.

And last (and luckily least) there are sine filters. They cause more problems than they solve. Resonance and overcurrent are common problems. I have had a few cases with sine filters where I had to add damping to be able to run without killing the IGBTs and another case where the necessary auto-tuning could not be carried out because the algorithm didn't expect the extra pole that the sine filter created. We just removed the filter and used part of it as motor reactors. Worked very well and the EMI that it was supposed to eliminate was as bad with the sine filter as without it. Also, this was for a huge gas pipe valve in the desert and there was no-one that could be disturbed by EMI within tens of miles.

Sorry, I got carried away. And there's more to be said. But I don't want to tax your patience any further.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.