OVERVIEW: Large motor tripped on ground fault while running. No other protection tripped (not time overcurrent alarm, time overcurrent trip, HDO OC trip, instantaneous overcurrent trip, negative sequence trip). Motor meggered sat (>1000 megaohms, PI > 2) but showed an open circuit on B phase. Disassembly inspection showed that the B phase line side T-lead (T2) was damaged a few feet inside the motor. There was approx 1.5" of copper missing from the T-lead. There was an arc strike on a grounded surface approx 4" from the fault. The location of the fault was not a joint within the T-lead.

MACHINE DESCRIPTION: 13.2kv, 3500hp, 324 rpm, vertical squirrel cage motor in outdoor environment, WP2 construction.

POWER SYSTEM DESCRIPTION: The motor is located more than half a mile away from the switchgear and has surge capacitors in the term box. The 13.2kv system is low resistance grounded by a 20 ohm resistor in the neutral of the transformer which feeds it. This limits ground fault current to 400A or less.

MOTOR HISTORY:
Motor rewound in 2005 with epoxy VPI windings. T-leads are 2AWG silicon-rubber insulated with an insulation temperature rating of 150C by the T-lead manufacturer. Motor winding temperature as measured by the hottest of 6 RTD's in the slot section is less than 200F while near full load following rewind - no change in this pattern at any time since rewind.

Motor was idle (not running) during the period 1/22/09 to 3/10/09. Review of temperature trends indicates winding temperature 10-15C above ambient - suggests space heaters were working. Other checks of space heaters before and after the event were also sat.

Motor was started 3/10/09 and ran until the event.

Rain occurred 3/14-3/15 and ended on 3/15.

TRIP:
Motor tripped on ground fault relay (51G) the afternoon of 3/16/09. It was sunny at the time and no plant evolutions in progress. No other flags were tripped and no disks noted in abnormal position. The motor had been running continuously since 3/10/09.

IMMEDIATE CHECKS: Immediate checks show B phase open-circuited, A and C phases have balanced resistances. Megger is > 1000 megaohms.

PROTECTIVE RELAYING:
(** Protection curves are shown in slide 1 of the attachment.)
51G relay - ground overcurrent - fed from 50/5 donut CT. It is GE type 12IAC77A801A with TD setting 2 and pickup settin 1A.

The following additional relays fed from 200:5 phase CT's:

A: 50/51 IAC66 phase overcurrent relays (one per phase). These are GE IAC66M. These have three functions:
1 - time overcurrent trip (pickup is 5A secondary, TD=2)
2 - HDO = trip at 30A secondary/1200A primary if that level persists for 0.1 seconds or more.
3 - Instantaneous = trip at 50A secondary / 2000A primary (no intentional delay but curve is provided by GE).

B: 50/51BX time overcurrent alarm function sensed on B phase. This is also a GE IAC66M
time overcurrent alarm (pickup is 4.3A secondary, TD=1.5)

46 - current balance relay - GE IJC51E. TD=4, TT=1.15sec, PU = 1.1A, Slope = 5%

Again the only flag of the above that flagged was the 51G ground relay.

SHOP INSPECTION:
Electrical testing at the shop confirmed same results as at the plant (B phase open, megger test sat).

The motor was disassembled and the location of the open circuit was found. It is at the B-phase line-side T-lead (T2) in the endwinding area of the motor (top end of motor which is connection end). The fault location is approx the 1:00 position if 12:00 is the motor term box. The conductor was melted open at this point and the insulation is compromised at this point, and there is evidence of current flow to ground at the closest non-insulated ground location on the side of the stator approx 4" away. There are no other anomalies evident during visual inspection to suggest the presence of any other fault location.

Slide 2 shows overview of the fault location. Slide 3 is zoom-in on fault location. Note that even though there is powder on the surface of the T5 lead (which is believed to be vaporized silicon insulation from the T2 lead), when cleaned up later T5 lead is found to be unaffected.

Slides 4 and 5 show the adjacent location on the stator which shows evidence of arc damage. (approx 4" away).

Slide 6 is closer view of the fault after cleaned up a little. You can see some copper is missing.

Slide 8 shows the pieces involved in the conclusion that 1.5" of copper is missing.

Slide 6 shows an overview of the T-lead. Note that downstream of the fault there is a transition point where the original lead insulation is stripped off of the lead, and further downstream (not even shown here) there is a brazed/crimped connection to the winding. However there was no connection or transition at the location of the fault. There was a mechanical tie within a few inches of the fault, but no evidence of any unusual mechanical stresses in this area.

Slides 9 and 10 show powder from the top bracket directly above the fault. This is believed to be vaporized silicon insulation.

Slides 11 and 12 show excessive bugs within the motor. After stator was later steam cleaned, there was absolutely no PD evident visually anywhere.

Slide 13 shows trails which indicate water droplets at some point in time had flowed radially outward along the blade (cannot be to gravity - must be due to centrifugal force indicating motor was running at the time).

The bugs and water evidence are clearly undesirable conditions and we plan to re-examine our filter changeout practices and other aspects of our enclosure very closely to address them regardless of whether they contributed to this event or not. But the main question is the cause of this event.

REWORK: The stator was steam cleaned and baked. The failed connection was reworked. Motor passed dc step votlage test to 30kvdc (one phase at a time - very linear). Motor passed surge test to 22.6kv. Motor reassembled, dc step voltage test repeated, installed into the plant, running fine at this time.

THE MAIN QUESTION: What do you think caused the failure?

Did ground insulation fail first leading to copper damage? If so, did the bugs and or water contribute? Other possible contributors? (this scenario seems unlikely since the ground relay is sensitive and should trip before significant damage occurs, especially given that no other relays tripped).

Did the conductor fail first leading to insulation damage? What would cause the conductor to fail? (this scenario seems unusual since this was not location of connection... wouldn't expect conductor to fail in the middle... also would expect a resistive conductor failure to show up during starting rather than steady state running).

ForPostSmall2.ppt (1,161 KB, 66 downloads)

Did the relay record the event? Is the event downloadable from the relay?

It might be worthwhile testing your relays to verify that it would trip under their other settings to keep from chasing down a false path. Have you checked for product recalls or product notices? We have had several recalls and service advisories on GE protective equipment, some of which were for possible fail to trip failures.

This one resembles to some degree the chicken and the egg situation...

Just pure speculation tells me that it all started with local insulation breakdown rather then copper wire local degradation.

One reason: it is more likely that insulation will break down LOCALLY then a braided copper wire would break LOCALLY in absence of wire connection in this location.

Another consideration: IF it was due to wire degradation, then some temperature protection would have detected it first. Even with the heaters operating during the downtime there was still a possibility that moisture could have penetrated insulation in some areas.

Another consideration: If it was the wire overheating so intensly that it eventually degraded insulation, you would have seen it in the area adjacent to arcing, which is not the case, I guess.

Just an opinion. Hard to prove based on the available evidence.

My guess would be a material fatigue. The billions of tiny moves resulted in the break of the T2 lead. (4 years at 120 Hz equals about 15 billion moves). The ground fault was just secondary. It helped to shut the motor down and prevented more damage.
I would not blame anybody for this fault. The rewind looks like a very professional job. And the repair was probably fairly cheap as well. On the top of it, the long weekend is coming.
jank

EPete

Is there any chance that any of the 'white powder' might have signified any level of PD in that area?

Judging from the level of contaminants in the winding, you might consider that some level of humidity was still present in the machine. If there was any level of treeing or PD activity at the point of fault, and with the voltage value at the T-lead itself, it may have hit a point where the air ionized enough to create an arc path.

I saw something like this in a fault in Australia. The investigation found that there was PD activity between two conductors and a high humidity condition triggered the arc failure, which, in this case, was between two adjacent conductors.

Was there any level of PD monitoring on this machine?

Thanks for the comments.

Wally – we have no fancy micro-processor relays unfortunately. Just 1970's vintage electromechanical relays. We will think about testing them.

David – I agree if it is the conductor it is not a melting due to high current along the entire cable up to the fault. In the conductor-failed-first scenario, it would require a localized high resistance such as from multiple strands broken for some reason.

Motordoc – If there was some partial discharge prior to the fault the evidence of it was erased by the fault. The large amount of material white/grey powder remaining is a result of the fault itself in my opinion, not something leading up to the fault (partial discharge only gives a small amount of powder). Also after We cleaned up the stator we saw no signs of partial discharge on any vulnerable areas in the remainder of the winding.

We do have Iris partial discharge monitoring (via coupling capacitors) on this machine. Since rewind in 2005, the B phase partial discharge level has been about 300millivolts and relatively stable while the other two phases 150 milivolts. The pattern is "classical pd" centered at the 45 degree and 225 degree position, and also negative temperature coefficient (goes down as temperature increased), which to me indicates the partial discharge likely comes from the slot section. It does not exceed our trending threshhold of 400 millivolts and has been stable and so was not/is not a concern. The partial discharge now that the machine is back in service following repair has the same pattern and levels which again reinforces my opinion that the higher level on B phase was not related to the fault.


Jan – Your thinking is similar to mine. It is not likely that arcing from a ground fault (in the ground-fault-first scenario) could cause this much copper damage. But my thinking is based on a thumbule which I don't fully understand.

Here's the THUMBRULE: In general we don't expect to have any significant copper damage during a ground fault unless it is accompanied by turn-to-turn fault or phase to phase fault (which we clearly didn't have in this case.).

That thumbrule has been told to me several times by many motor shop personnel and I have observed it to be correct in all the failures I have seen up until now. But I don't particularly understand whether there is any explanation/proof of this thumbrule.

I have heard reasoning roughly along the lines of: "the ground fault relays are sensitive and the ground is cleared before copper has a chance to melt.". And while I agree the fault is cleared long before the copper will melt by simple copper I^2*R losses in a healthy (full-cross-section) wire, there is also the matter of the arc temperatures which are much much higher.... they could melt copper also couldn't they? I would be interested in your thoughts on the basis of the thumbrule.

Now a related thought is that we might make some small inferences about the amount of copper melting which is possible from this arc (arcing must have melted the copper in the insulation-failed-first scenaio) by observing what kind of melting went on at the other end of the arc where it hit the stator frame steel. As shown in the attachment to this post, the bulk of the crater is red (same color as paint and rust in other areas of the stator). There is only one small point of shiny steel color. So I think that large crater was pre-existing and the arc just caused that one steel-colored blemish in the lower left hand corner. Now IF copper and steel had similar melting points, I would say this is pretty good evidence that the arc didn't melt the copper. Unfortunately steel melts around 3000F while copper melts around 2000F so it is not quite so clear what conclusions can be reached looking at this steel damage on the other end of the arc.

Any other thoughts or comments?

This message has been edited. Last edited by: electricpete,

CraterArc.ppt (109 KB, 32 downloads)