Help with testing partial? ground fault on floating secondary

[Ingenieur]

imo can't be any ground current
it is a delta
no X0 (zero seq reactance is infinite or limited by coil insul 1000 M)
at 180 V 0.18 uA
not even factoring in cond insulation!
that would not produce the unbalanced mag you observe

 

[GoldDigger]

I think the reason for the ungrounded delta secondary is the classic one.
Losing the heat source unexpectedly with the furnace full of glass could be catastrophic for the equipment.
With an ungrounded circuit a single fault to ground anywhere will not disable the heater, allowing time to do a controlled shutdown (even finishing the batch normally) and then make the necessary repair.

 

[Ingenieur]

I think the reason for the ungrounded delta secondary is the classic one.
Losing the heat source unexpectedly with the furnace full of glass could be catastrophic for the equipment.
With an ungrounded circuit a single fault to ground anywhere will not disable the heater, allowing time to do a controlled shutdown (even finishing the batch normally) and then make the necessary repair.

Yepper
In fact you can operate like that indefinitely since ph-ph is unchanged
no shock hazard
until a second ph faults then sc fault or if in the path shock
this used to be common in mining
now wye with a NGR is required

 

[gar]

160131-1635 EST

milldrone:

You did not previously mention the ground strap. I believe this will be the source of the current path producing unbalance of the voltages.

Put an AC current meter around the ground strap. By your use of the singular I assume there is only one ground strap. Initially you may or may not see much cutrrent at all, hopefully near zero. Apply your tungsten lamp load from one phase to earth. The ground strap current should change by about the amount of current thru the lamp load. If without the lamp load the current is zero, then with the lamp load the lamp current should equal the ground strap current.

I believe you have a low impedance path from the glass melt to the metal kettle. You do not have a uniform electric field in the glass melt to the kettle and therefore the unbalanced leakage voltages relative to ground.

Does the other glass melter have one or more ground straps?

How does the glass initially get hot enough to conduct to use resistive heating via current thru the glass?

.

.

 

[Ingenieur]

Re:ground strap
if current bled from rod to glass to vessel/strap
where would it flow to?

assume you put a phase on the vessel/strap what would happen? (other 2 are connected as usual)

 

[gar]

160131-2114 EST

Ingenieur:

If there is no path to earth other than the ground strap, then there is no current flow. Within the variable transformer there may be some resistive leakage, but possibly more capacitive leakage. Under normal conditions these would be very small compared to 6000 A.

When milldrone connected his 500 W bulb, then there was a flow of about 4 A. This is small compared to 6000 A, but measurable, and it provided a method to estimate the leakage path source impedance (electrodes to glass, to brick or thru brick, to metal kettle, to grounding strap, and thru earth.).

Possibly this is an indication of a failure in the kettle lining.

.

 

[milldrone]

Does the other glass melter have one or more ground straps?

How does the glass initially get hot enough to conduct to use resistive heating via current thru the glass?

.

.

They all have a single ground strap.

When they "fire off" a melter they load the area between the electrodes with a special low melting point glass, then heat it with oxy mapp torches.

 

[milldrone]

I think the reason for the ungrounded delta secondary is the classic one.
Losing the heat source unexpectedly with the furnace full of glass could be catastrophic for the equipment.
With an ungrounded circuit a single fault to ground anywhere will not disable the heater, allowing time to do a controlled shutdown (even finishing the batch normally) and then make the necessary repair.

That might be the intent, but all of the SCR controlled units fold when there is a ground fault.

 

[milldrone]

160131-2114 EST

Ingenieur:

If there is no path to earth other than the ground strap, then there is no current flow. Within the variable transformer there may be some resistive leakage, but possibly more capacitive leakage. Under normal conditions these would be very small compared to 6000 A.

When milldrone connected his 500 W bulb, then there was a flow of about 4 A. This is small compared to 6000 A, but measurable, and it provided a method to estimate the leakage path source impedance (electrodes to glass, to brick or thru brick, to metal kettle, to grounding strap, and thru earth.).

Possibly this is an indication of a failure in the kettle lining.

.

This what concerns me. We have IR camera scanned everything and cannot find anything. Perhaps a power off resistance check to ground while the glass is hot will tell us something. I'm not sure I can convince production, but I'll try.

 

[milldrone]

160131-1635 EST

milldrone:


Put an AC current meter around the ground strap. By your use of the singular I assume there is only one ground strap. Initially you may or may not see much cutrrent at all, hopefully near zero. Apply your tungsten lamp load from one phase to earth. The ground strap current should change by about the amount of current thru the lamp load. If without the lamp load the current is zero, then with the lamp load the lamp current should equal the ground strap current.



.

I will also try this on Monday. I'm suspecting that even though there is one ground strap there will be multiple ground paths.

 

[gar]

160131-2210 EST

milldrone:

I think I would get several, about 5, cheap hardware store space heaters, and a multi-outlet power strip. The heaters will have their own on-off switches. As I mentioned these will draw about 10 to 13 A.

Connect the outlet strip from phase B to ground (earth). Measure all three line to ground voltages, and the voltage across the outlet strip. Measure current to the outlet strip, and current in the ground strap.

The outlet strip or whatever concoction you use in its place needs to be good for 100 to 200 A, depending upon how many heaters you actually use. This scheme will allow incremental increases in load current.

Use a Hall sensor to scan over the surface of the kettle to see if you can identify the current path in the kettle. This might allow finding where the current originates from inside the kettle if it is in a localized area.

A Hall sensor I have with 5 V DC excitation has an unfiltered noise level of about 3.3 mV with a Fluke 27. When I isolate a #12 conductor with 0.3 A flow the Flukre reading increases to about 7 mV with the Hall sensor orientated for maximum output, and very close to the wire.

A narrow band 60 Hz filter could improve detection of of low current densities.

My total home load was about 2 kW at the time of measurement and about balanced. Above, lever side, of some of the breakers I was reading about 25 mV from the Hall device. Usually I see a small amount of current on my incoming 1.25" copper water line. Tonight I did not see anything except the Hall noise level.

The current density in the kettle surface would determine how successful this approach would be.

.

 

[gar]

160131-2414 EST

milldrone:

One more method to search for current paths is to use a test coil. This is physically larger than the Hall device but I can get somewhat greater sensitivity.

A coil I have has 7000 turns of #41 wire on a 1/2 x 1/2 inch paper core. There is no iron. The length is about 3/4" and outside is about 1 1/4 x 1 1/4 inches. With a 1 ufd shunt capacitor the noise level is below 100 microvolts. With this I read about 1 mV by my water line.

.

 

[gar]

160201-1019 EST

milldrone:

More thoughts:

Magnetic and electric fields have direction. Current flow has direction.

Using a magnetic sensor the orientation of the sensor provides information on the direction of current flow. With either my Hall device or my air core coil I can get useful direction information.

You do have to worry about unwanted magnetic fields affecting the desired measurement. Unwanted fields need to be minimized relative to the desired field. This can be done by location of the sensor relative to the desired field (close), and maximizing the excitation to the wanted field.


Another method is voltage drop measurements. If you map out equipotential lines, then current flow is perpendicular to the equipotential line.


I tried two quick experiments this morning. I used a scrap piece of aluminum about 18" x 30" x 1/16". Injected and received about 5 A 60 Hz at two concentrated points asymmetrically located. I was happier with the coil magnetic sensing than the voltage sensing. Using voltage sensing you need a coaxial cable to a sharp probe with the shield connected to the other probe point, and from that other probe point to the meter is needed a coaxial cable. The coax cable is to minimize induced voltage from external magnetic fields. But still the voltage drop measurement can be influenced by external magnetic fields. The other probe point would be your ground strap.

Much of this approach is based on field mapping taught in A. D. Moore's senior design class where we spent 1/2 semester drawing field maps. A. D. Moore had a significant role in Tau Beta Pi in the 1920s. See http://www.tbp.org/about/F85FirstCentury.pdf

.

 

[gar]

160201-1129 EST

Measured at 1 kHz the inductance of my 7000 turn coil is 1.25 H and the Q is 2.5. DC resistance is about 2.54 k. #40 wire is 1.049 k per 1000 ft. Thus, more than 2000 ft of wire on this coil.

With 1 ufd the resonant frequency is about 140 Hz. For 60 Hz I need about 5.5 ufd. I verified the resonant frequency by measurement. Note: this is a low Q circuit.

With 1 ufd the residual 60 Hz in my basement is around 10 microvolts. A foot from my open main panel about 2 mV.

Adding capacitance to create resonance did not improve sensitivity, but reduced it. This is because of the high internal resistance of the coil.

.

 

[Ingenieur]

160131-2114 EST

Ingenieur:

If there is no path to earth other than the ground strap, then there is no current flow. Within the variable transformer there may be some resistive leakage, but possibly more capacitive leakage. Under normal conditions these would be very small compared to 6000 A.

When milldrone connected his 500 W bulb, then there was a flow of about 4 A. This is small compared to 6000 A, but measurable, and it provided a method to estimate the leakage path source impedance (electrodes to glass, to brick or thru brick, to metal kettle, to grounding strap, and thru earth.).

Possibly this is an indication of a failure in the kettle lining.

.

my bet is the path was ph-ph via vessel to a different phase rod (or both)
how was the light connected? Ph to frame ground? Same as strap
frame-strap-glass-rod(s)
wonder what the lamps i and v were?

no way does the xfmr windings have 4 A of leakage at 100 V
he meggered the xfmr windings at 1 G Ohm, <uA's at 100 V

messy Issue

 

[Ingenieur]

My WAG
The rods are coupled to the earth via glass/vessel/strap to frame/earth
ramdomly to varying degrees based on location/glass properties/etc
so you have a partially corner grounded delta in which the degree of ground and the corner(s) are not fixed

that and some V drop issues
pull the strap and see what happens

 

[milldrone]

Folks, it's been a busy day, everyone wanted a piece of my backside! Production would not allow me to shutdown the power for testing! I do however have some values while we were running.

Problem child transformer
phase to phase volts A-B 171.3, B-C 171.8, C-A 172.8
volts line to ground A 104.3, B 76.4, C 113.3
Amperes A 5100, B 5426, C 5282

Second transformer
phase to phase volts A-B 174.7, B-C 174.9, C-A 175.9
volts line to ground A 103.7, B 100.1, C 101.7
Amperes A 5072, B 5382, C 5382

 

[gar]

160202-0028 EST

milldrone:

Using your data for the problem transformer I plotted vectors.

The line to line voltages defined a triangle. The three corner points defined a circle. The radius of the circle is 99.29 V. The center of the circle is the theoretical neutral.

From the corner points circles of the measured vectors were drawn. These do not interesect at a point, but the intersections create a 3 point circle. This circle has a radius of 5.25 V. A vector from the theoretical neutral to the center of this apparent neutral is 23.3 V at an ange of about 103 deg from the vector from the theoretical neutral to line A.

Correlates well with your previous measurements.

I did not draw out the good transformer data but it looks good.

.

 

[Ingenieur]

...
phase to phase volts A-B 171.3, B-C 171.8, C-A 172.8
volts line to ground A 104.3, B 76.4, C 113.3
Amperes A 5100, B 5426, C 5282

the glass Z (load) is not balanced
Ph V/(i/sqrt 3)
0.058
0.024
0.057
not a balanced load
shifts center point

The load on the 'good' transformer is balanced
0.035
0.032
0.033
almost perfect

I previously calculated the phasors based on the phase shift being consisrent on all 3 phases
when I back calculated the ph-ph V the phase angles were maintained within a few degrees when normalized for the 30 deg ph to N shift

 

[GoldDigger]

the glass Z (load) is not balanced
Ph V/(i/sqrt 3)
0.058
0.024
0.057
not a balanced load
shifts center point

The load on the 'good' transformer is balanced
0.035
0.032
0.033
almost perfect

I previously calculated the phasors based on the phase shift being consisrent on all 3 phases
when I back calculated the ph-ph V the phase angles were maintained within a few degrees when normalized for the 30 deg ph to N shift

Can you calculate where between the two lower voltage electrodes the hypothetical fault in the vessel insulation might be from those vectors?