Corona Detection in a Generator

[TheBlacksmith]

We are working with a new design medium voltage generator and are concerned about corona. To ensure long life, we know corona should be minimized and there are insulation systems, etc. that are supposed to minimize/manage corona. We need to know what initial testing is being done to detect corona. Not interested in trending over long periods (yet), need a subjective up front test. Ultrasonic, looking for noise discharge of corona and power factor tip up are the only two I have found.

Blacksmith

 

[electricpete]

Try searching the forum and the web for partial discharge testing. That's a fairly easy test gaining wide acceptance. Manufacturer's include Iris Power and Cutler-Hammer. There is a trial-use IEEE guide on partial discharge.

With stator disassembled (end-bells and rotor removed, stator energized), there are some good tests to locate problems:

Ultrasonic (UE Systems Ultrprobe)
emi (TVA) probes
black-out test (view energized stator in a dark room).

There are also special camera's (laser-vision?) that let you view the corona in broad daylight (accomplishes similar thing to blackout test).

You asked for subjective but I think you meant objective. The results of all of the above tests are open to interpretation, but for PD, the vendor will aid you in establishing a reasonable acceptance criteria for you machine.

DC Step voltage test has some opportunity to detect the types of problems you are looking for. Slowly increase voltage in steps and plot current vs voltage... look for a nonlinearity. See IEEE95

 

[TheBlacksmith]

ElectricPete,

Thank you. Yes, I meant objective; sometimes I get things backwards. I agree about the tests being open to interpretation. However, the vendor has suggested testing a sample winding and extrapolating the results to the entire machine. That's just a little too much interpretation for my tastes, hence the question.

Blacksmith

 

[electricpete]

Don't overlook coil testing as an effective tool in your bag of tricks. (Although I understand you may not be comfortable with using it to the exclusion of whole-machine tests....I'm assuming you're getting some kind of whole-machine hi-pot also... that is a virtual necesssity).

There are two general types of resin application (B-stage and VPI) and the degree of testing coverage varies.

If they use a fully-cured B-stage (resin-rich) coil process, then every coil has it's full dielectric strength before you put it into the machine and you can subject 100% of the coils to power factor tipup and even hi-pot and surge testing (to the extent that you don't degrade coil reliability).

If they use a global VPI (Vacuum Pressure Impregnated) process, then the coils don't have full strength until after they are already installed in the slots and the connections are made (the whole thing gets impregnated as a unit).

Some advantages of this approach are:
1 - It is virtually the ONLY way to test the turn insulation of the interior coils since externally-applied pulses are rapidly attenuated while testing from terminals (of course interior coils are not exposed to severe stresses in service either).
2 - For 13.8kv machines, it is the only way to get a representative power factor tipup reading. The stress control coatings can be guarded out while testing and individual coil but not while testing the whole machine.... their nonlinear behavior interfere's with results.
3 - It can be helpful to test at least one of the sample coils to destruction (by dc hi-pot or surge test). Then cut open the coil at the point of failure and look for weaknesses (insulation voids). Possibly section other coils at the same location.

 

[TheBlacksmith]

ElectricPete,

I think I follow your logic.

1) The machine will be assembled with green coils, the global VPI applied - that is why I think a coil test will not represent the finished product. We won't be certain the whole machine is up to snuff and the test coil may fall short of the finished VPI in its test, so the manufacturer will appear to have failed to provide a good design.

2) We plan on hi-potting an individual coil. All agree it is foolish to risk a finished machine should there be a catastrophic failure.

Problem is the manufacture is asking us what tests we want, rather than volunteering info. Another supplier is not an option - I'm just one the engineers that has to "make it work". That is the basis for my question(s).

Blacksmith

 

[electricpete]

You're right that the green coils prior to VPI deprive you of an opportunity to perform 100% testing of hi-pot and surge testing on every coil.

With global VPI, a very good option is to request that a few coils be set aside and placed in a pair of "mock" slot which have similar dimensions to machine slots. Then the mock coils/slots are subject to the VPI process along with the stator. Then the mock coils are subject to surge and/or dc hi-pot test until failure. Hopefully failure occurs above established minimums (2*KV+1). Either way you can try to identify the point of failure (possibly apply a current and look for smoke at the point of the short). Then section the coil and look for voids.

If you can't produce a failure, you should secton the center of the slot (axially) on the mock coil since impregnation is poorest there, as well as the bends in the end-turn area where the tape insulation tends to bunch and create voids.

I think a full ac or dc hi-pot at the factory is almost a necessity. If you don't do that, then you're limited in your ability to do subsequent field hi-pot testing even at reduced voltage. And the reliability of the machine will forever be questionable.

If you're concerned about the schedule impacts of a hi-pot failure, then stick with the dc hi-pot for two reasons:
#1 - less collateral damage in the event of test failure since the current trip setting is very low (in contrast ac hi-pot current trip must be set high to supply capacitive current.
#2 - You can use a "step-voltage" method on the dc test where you increase voltage in perhaps 15 uniform steps. Hold one minute at each step and record leakage current. Plot current vs voltage. A trend in current vs voltage which increases more than linearly is an early indication of problems (usually end-turn contamination). At that poinot you should stop the test and decide what action to take (no damage has been done if you don't go the point of test set trip). Even with this method, some slot failures do not give much warning. End turn leakage failures give more warning.