Help with testing partial? ground fault on floating secondary

[milldrone]

I'm trying to identify / locate what appears to be a partial(?) ground fault on a rather unusual piece of industrial equipment. It's a 2MVA 3 phase transformer with a 4160 primary and an infinitely variable ( between 35 and 250 volt) secondary. The primary is connected in delta. The secondaries are brought out of the can as 6 terminals. The best way to describe the secondary is it's construction could be thought of as a multi turn power rheostat. The secondary is connected in delta (floating)

The purpose of these strange units is to melt glass by energizing electrodes that are submerged in the molten glass pool. Yeah, I thought glass was an insulator, but when it's heated to a "glowing" temperature it conducts very well.

Each phase conductor that supplies the electrodes consists of 12 530mcm locomotive cables. The normal running amperage to each electrode is just under 6,000 amps. The normal voltage phase to phase is around 180 volts.

We have two of these units and one displays voltage to ground measurements like what you would expect from a floating delta system. The problem child has the "vector center" skewed by approximately 15 volts. The measurements recorded yesterday while under load were:
1. A phase 110 volts to ground.
2. B phase 89 volts to ground.
3. C phase 125 volts to ground.

The 89 volts leads me to believe that the phase is not "solidly" faulted to ground.

I have "some experience" troubleshooting "floating ground" and "high resistance ground" systems. I tried a few of my tricks and they failed or confused me.

Extensive resistance and meggar tests have been performed when the equipment was not in production, no "smoking guns" were found.

I tried using a "wiggy" (.1 amp) as a small load to swing the "vector center" from center when the system was under load, this produced no visible change.

I tried using a 500 watt quartz light (4 amp nominal load) to swing the vector center, this produced:
1. When applying the quartz light to ground and A phase, it lowered the voltage 4 volts.
2. When applying the quartz light to ground and B phase, it lowered the voltage 2 volts.
3. When applying the quartz light to ground and C phase, it lowered the voltage 4 volts.

I was able to lift the electrodes out of the melt pool and energize them (at the minimum voltage). My results were:
1. A phase 24 volts to ground.
2. B phase 20 volts to ground.
3. C phase 22 volts to ground.

As you can see the B phase still read the lowest, but what I found interesting is that A & C swapped in intensity, voltage to ground.

My opportunities for testing are severely limited by production. What I would like to get from the community is a list of their tricks or suggestions on what to test.

 

[zbang]

You've megger'd to ground and nothing interesting came up? That was when the rig was cold? If so, could be a part on the B phase that expands enough when hot into a short to ground. Do you have a high-temp IR camera?

(Is this one of the OI plants in the SF area? Just curious.)

 

[milldrone]

You've megger'd to ground and nothing interesting came up?

Correct, the meggar showed nothing unusual.

That was when the rig was cold? If so, could be a part on the B phase that expands enough when hot into a short to ground.

Yes rig was cold. In approximately 50 days I might be able to convince production to let me lift the electrodes and meggar (rig will be hot in temperature). I will add this to my list of things to test.

Do you have a high-temp IR camera?

Yes, I forgot to mention this in the original post. We have an IR team dedicated to IR testing. They are the best I have ever seen. Again, "no smoking guns". The only thing they found unusual was we had a cable assembly too near a gearbox (12 inches away on A phase) this proximity induction heated the gearbox to 205 degrees F. The partial ground remained at the same level after this was corrected.

(Is this one of the OI plants in the SF area? Just curious.)

Not sure what a OI plant is. Not in SF area. Management is very secretive about process, I may have already broadcast too much info already.

 

[GoldDigger]

If the megger shows no resistive connection to ground, there can be significant difference in the capacitive coupling from one winding to another. This will cause an off center voltage to ground and will not be affected by any line to line load that is connected. The amount off center might or might not depend on the output voltage setting.
The clear indication of this kind of coupling is that the available current to a line to ground load will be very limited. Connecting a line to ground test load should make the voltages look much like a corner grounded delta.

 

[Ingenieur]

OI Owens-Illinois

I did work in some glass plants
basically an arc furnace (usually with NG supplemental)

these are hard to work on because it is a continuous process and shutting down solidifies the glass , a real mess

basically all you can do is disconnect and megger the leads
swap leads on the xfmr and see if it follows

iirc the electrodes are consumed/sacrificed and can be adjusted for gap
bad rod or need adjustment?
sometimes you get a bad piece of glass stuck in the path due to a leak/cooling

the bigger question
is it an issue?
is product quality affected?
why not equalize V by adjusting xfmr taps?

i'm sure the data is logged
is it consistent? Or does the low leg move around
did it just crop up? Gradually appear over time?
what are the measurements on the 2nd unit?

 

[milldrone]

If the megger shows no resistive connection to ground, there can be significant difference in the capacitive coupling from one winding to another. This will cause an off center voltage to ground and will not be affected by any line to line load that is connected. The amount off center might or might not depend on the output voltage setting.
The clear indication of this kind of coupling is that the available current to a line to ground load will be very limited. Connecting a line to ground test load should make the voltages look much like a corner grounded delta.

This was one of my thoughts as well. This is why I loaded phase to ground with a 500 watt quartz light. This put a nominal 4 amps phase to ground. As I noted in my original post this shifted the voltage 4, 2, 4, volts to ground. I'm nervous to try much more. I'm intending to try (50 days from now) when we can lift the electrodes from the melt pool another test with the "wiggy" load to ground then depending on the "wiggy" test results possibly another quartz light test.

A couple of things I left out of my original post are:
1. A phase conductor length is around 25 feet long.
2. B phase conductor length is around 15 feet long.
3. C phase conductor length is around 35 feet long.
4. We try to maintain 18 inches between conductors and ferrous materials. There are some obvious exceptions to this (transformer terminals on the secondary). So I'm thinking the capacitive coupling would be very small.
5. 500 volt meggar test showed 50, 100, 77, megs to ground on A, B, C phase conductors and electrode assemblies. This is with a cold rig, because it takes 2 days to get the transformer and the phase conductors separated.
6. 500 volt meggar test showed just over 1 gig when testing secondary to ground (with control instrumentation disconnected).

 

[milldrone]

the bigger question
is it an issue?

Engineer says it is! My original thoughts were along the lines of GoldDigger's thoughts and it was just a couple of ma to ground. After I performed the quartz light test to ground and saw only a 4, 2, 4 volt to ground difference , I became a little paranoid. A molten glass leak will destroy almost anything in it's path (think china syndrome).

is product quality affected?

Not that we are aware of.

why not equalize V by adjusting xfmr taps?

It's an infinitely adjustable tap and has a single motor adjusting all three phases, and phase to phase voltage is less than 1% difference.

i'm sure the data is logged
is it consistent? Or does the low leg move around
did it just crop up? Gradually appear over time?

From the moment the volts to ground disparity was noticed it remains near the same percentage of difference. It has remained on the same leg. The engineer claims it was not this way when installed.

what are the measurements on the 2nd unit?

The second unit phase to ground voltage is = phase to phase voltage / 1.73.

 

[Ingenieur]

The Z of the glass varies with locarion
higher at the end where the raw materials are input, sand, glass cullet, etc
Lower at the throar where discharged, nice, hot and homogenous

making glass is a craft/art as much as science
the reason rhey haxe individual adjustments is so you can vary the temperature of the mixture as it flows thru the furnace

if the conductors meg good
the xfmr windings are good
the rods are good
the glass is good
not sure where to start

how are the phases arranged in the furnace? From back to throat

what I meant is you have 2 furnaces/xfmrs
do both read the same?

the furnaces I saw all had bottom leaks
basically layers of kiln bricks
they would just hose them from below with water to freeze/plug them

 

[milldrone]

if the conductors meg good
the xfmr windings are good
the rods are good

Check
Check
Check

the glass is good

I'm still suspicious that everything is OK in the melt pool. As you already mentioned "it's a continuous process" and my opportunities for testing are very few and short duration. I'm developing a list of needed tests so that I can capitalize on the time when the power is off. I'm also looking for any tests I can perform while the power is on.

The Z of the glass varies with location (spelling changed for clarity)

I'm very aware of this. I'm not trying to be rude with this last statement. I cannot answer many of your questions with much detail, as it may divulge what my employer want's to keep a secret.

what I meant is you have 2 furnaces/xfmrs
do both read the same?

Yes two nearly identical rigs, each has one transformer. One rig has the expected reciprical sq root of 3, phase to ground voltage. The other rig has the 15 volt skew. Again sorry for the lack of details.

 

[Ingenieur]

A few more ?'s if you can answer them

xfmr tap changing or induction regulator
if regulator or tap could the phases be out of synch?
I guess tap since you compared it to a rheostat

why 6 leads? I assume sec only, is that prim and sec total?
I would think
6 if induction regulator, 3 coils to be configured as delta or wye
3 if tap changing
9 if tap changing, 6 for the coils and 3 for the taps

how are the rods connected
delta
wye, floating
if wye are a bus bar or the vessel used for centerpoint

 

[gar]

160130-2230 EST

milldrone:

Using your voltages 110, 89, and 125 and assuming that the line to line voltages are equal, and the phase angles of the source voltage are equally spaced (120 deg), then graphically I get your line to line voltage to be close to 185.3 V. The voltage to theoretical neutral is 113.27 V. The voltage from theoretical neutral to your measured neutral is 21 V at an angle of 92.8 deg relative to the theoretical neutral to line A.

My neutral offset of 21 V is in the ballpark of your 15 V.

We need to know the current to each electrode, and phase angle when you get the readings mentioned above.

From your description the special transformer is possibly a motor driven ganged Variac. If it is, then there would be brush contacts, and almost certainly these Variacs would be on the primary side to reduce brush current. Another possibility is a set of transformers with variable coupling, but this is also a variable internal impedance.

Whatever the design there is the possivility of a tracking problem. Your equal line to line voltages does not iimply a tracking problem.

Can you find out from the transformer manufacturer what is inside?

.

 

[milldrone]

A few more ?'s if you can answer them

xfmr tap changing or induction regulator
if regulator or tap could the phases be out of synch?
I guess tap since you compared it to a rheostat

Three cores inside can, secondary is a spiral of "chevron" shaped cross section material. Tap slides on the inside of the "V", and spirals up and down.

A few more ?'s

why 6 leads?

One end of coil and sliding tap are brought out X three windings.

A few more ?'s if you can answer them
how are the rods connected
delta
wye, floating
if wye are a bus bar or the vessel used for centerpoint

Floating delta connection.

 

[zbang]

Floating delta connection.

Hmm, so no "neutral point". Any particular reason the secondary isn't grounded? You could install a zigzag transformer + grounding resistor; would stabilize the voltages and make measuring phase-ground more meaningful. Would also give a better idea of the leakage.

Don't discount some difference in the transformer, either. Could be a minor difference in one coil.

Have you looked at the harmonics? (I doubt that's the issue, just mentioning it.)

I don't think anyone's asked- is this a new issue or has it always been like this and was only recently noticed?

 

[Ingenieur]

Where were the V-gnd readings taken
xfmr?
rods?
with full i flowing?
length of cables ( sorry if I missed it ) edit:saw it

since both ends are delta no gnd current should flow?
grounding should only have an effect similar to a corner gnd delta
with such low potential ~110/185 gnd/ph and your megger readings the leakage ( probably ph-ph) is on the order of uA
In fact you megger'ed at 5X the operating potential

interesting issue
but no idea of the solution lol

still not clear on the xfmr connections
6 leads
3 for the taps
3 for the coil low end
are the 3 coils brought out and external delta connected?
are the any internal coil connections?

if no internal you need 9 leads
if internal you don't need the 3 non-tap ones
???

 

[Ingenieur]

After sketching it out I think I see what they are doing
the hi end of the coil is left open
and the drlta made externally with 2 jumpers

 

[gar]

160131-0904 EST

milldrone:

You have short cables and I wildly guess that capacitance per cable to ground is not greater than 0.001 mfd (2.6 megohms reactance at 60 Hz), but suppose the capacitance was 0.1 mfd, an impedance of 26,000 ohms, then this is still insignificant in comparison to your test loads. 500 W at 120 V is 29 ohms. You do not have an unbalanced cable capacitance problem, nor a normal high resistance shunt resistance leakage problem unless it only shows up under hot conditions.

Did you measure the same unbalanced voltage condition at both the transformer and electrode ends?

I tried using a 500 watt quartz light (4 amp nominal load) to swing the vector center, this produced:
1. When applying the quartz light to ground and A phase, it lowered the voltage 4 volts.
2. When applying the quartz light to ground and B phase, it lowered the voltage 2 volts.
3. When applying the quartz light to ground and C phase, it lowered the voltage 4 volts.

I assume this means from 110 to 106, 89 to 87, and 125 to 121. For each of the three measurements it would be interesting to see what the other two voltages were for each.

If I assume that each of these paths to ground (earth) have comparable series source impedances, whatever they may be, then I would expect your quartz load to produce proportional changes. It sort of does. But tungsten lamp resistance changes with voltage. A nominal 1500 W spacer heater might be a better test load, about 10 ohms and changes only about 10% from 0 V to full voltage..

Using the information we have I ball park the source series impedance at about 28*4/(125-4) = about 0.1 ohm for the apparent leakage path. A change of voltage of 107 to 89 V = 18 V. Thus, what appears as leakage current is about 18/0.1 = about 180 A.

If my approach has any validity, then the problem is in the glass tank and electrode area. Is the glass tank really insulated from ground?

Others have ask some good questions.

.

 

[milldrone]

After sketching it out I think I see what they are doing
the hi end of the coil is left open
and the drlta made externally with 2 jumpers

Correct!

 

[milldrone]

Hmm, so no "neutral point". Any particular reason the secondary isn't grounded? You could install a zigzag transformer + grounding resistor; would stabilize the voltages and make measuring phase-ground more meaningful. Would also give a better idea of the leakage.

I don't know the reason it's not grounded. We have 6 melters total, 2 use this unique transformer arrangement the other 4 use SCRs on the secondary. All of the melters are floating. The other four have a star configuration but are not grounded.

I asked the engineers about a zig zag and they said no.

 

[milldrone]

160131-0904 EST

milldrone:

You have short cables and I wildly guess that capacitance per cable to ground is not greater than 0.001 mfd (2.6 megohms reactance at 60 Hz), but suppose the capacitance was 0.1 mfd, an impedance of 26,000 ohms, then this is still insignificant in comparison to your test loads. 500 W at 120 V is 29 ohms. You do not have an unbalanced cable capacitance problem, nor a normal high resistance shunt resistance leakage problem unless it only shows up under hot conditions.
.

Your wild guess is what I'm suspicious about. One possibility is that the melt pool is a giant voltage divider and we have a "small" current "leak" to ground nearest to the B electrode.

Production started the melter up under emergency conditions on a weekend, and I missed the opportunity to witness the voltage to ground readings as the pool heated up.

160131-0904 EST

milldrone:

Did you measure the same unbalanced voltage condition at both the transformer and electrode ends?

.

Yes, I cannot measure the electrode end, but I can measure the cable to electrode arm connection. C phase (the longest) appears to have about a 1 volt drop between the transformer and the electrode connection. The other phases are less

160131-0904 EST

milldrone:

I assume this means from 110 to 106, 89 to 87, and 125 to 121. For each of the three measurements it would be interesting to see what the other two voltages were for each.

.

Yes! That would be an excellent test while the power is on. I should be able to give you the results of this on Monday.

160131-0904 EST

milldrone: A nominal 1500 W spacer heater might be a better test load, about 10 ohms and changes only about 10% from 0 V to full voltage..

.

I had considered this, but my access is very limited. I'm not sure I can make a connection capable of a nominal 13 amps.

160131-0904 EST

milldrone:
Is the glass tank really insulated from ground?
.

You have touched on what seems to be some of the insanity (in my mind) of this install. The tank sits on insulated legs, but they have installed a 500mcm ground strap! But wait, the glass is separated from the metal tank by firebrick, and the tank is water cooled. So in theory, if any glass seeped out to the metal tank it would be chilled, thus no current.

 

[Ingenieur]

I calculated phase voltage assuming sequence/ang remained the same
after adjusting for a phase-neutral shift of 30 deg (actually came to 28)

a-b 164 ang 0
b-c 186 ang 112
c-a 195 ang -118 (+242)
phase and mag as expected (within tolerance of the calcs/measurement)

but you measured a phase-phase V all within 1% (which btw is very tight)
more confused lol happens often, ask wifey

imo looks more like a V drop issue (not cond see below) than corner grounding
especially from megger
measured <10 uA at 500 V avg 7
derived based on measured
1-2 uA phase-phase

cable spec per 1000'
R 0.033
XL 0.026
in air 728 A
not adjusted for multiple cond
using your l and 6000 A... V drop range 0.5 to 1.16 nadda 1/2% or less

lugs, rod or xfmr
xfmr tap synch
???
good luck
let us know what you find