high P.I. (with good one-minute I.R.) 3

[electricpete]

Can anyone provide information related to concerns that may be associated with the particular symptom of high Polarization Index P.I.?

For example, epoxy-mica medium voltage machine shows 20 G-ohms at one minute and 150 G-ohm at 10 minutes => PI = 150/20~7.5.
I am familiar with IEEE43 guidance that PI loses significance above 5 G-ohms one minute, since readings above this based on lower current can have error, but still interested in exploring possible meanings of this symptom).

I have heard various places this could indicate dry / brittle / resin-depleted insulation (thumbrule PI > 6 suggests this condition).

My two biggest questions:
1 - Does this "thumbrule" apply only to older Asphaltic systems?
2 - What is the physical reason why insulation systems which are dry / brittle / resin-depleted might have high PI?
3 - Has anyone seen high PI associated with insulation problems (on machine with good 1-minute IR)
4 - Any references?


=====================================
(2B)+(2B)' ?

 

[electricpete]

Here are a few quotes about this phenomenon:

http://ewh.ieee.org/cmte/pes/materials/WorkingGroups/P56/P56 Rev 16 Nashville .pdf

An abnormally high polarization index from an older winding; i.e., a PI in the order of 5 or more for a winding in the 20-year age bracket, could indicate intact, but lifeless, dried-out insulation. An "in-service" failure could result from a sudden fracture of the brittle insulation caused by mechanical shock, such as introduced from a short circuit on the system

http://electromotores.com/PDF/InfoTécnica/EASA/Using Polarization Index.pdf

A ratio above 5 may indicate a very dry, brittle insulation system

http://www.scribd.com/doc/24911026/...ctrical-Insulation-Testing?wwparam=1351686293

**In some cases, with motors, values approximately 20%higher than shown here indicate a dry brittle windingwhich will fail under shock conditions or during starts. For preventive maintenance, the motor winding should be cleaned,treated, and dried to restore...

=====================================
(2B)+(2B)' ?

 

[ScottyUK]

ePete,

Have you done any PD monitoring on these machines? I would expect that severe insulation problems might show up their signature in other ways; the most useful way is through trends but a major problem might be noticeable through a spot reading.

 

[electricpete]

Thanks Scotty.

I do have a specific machine in mind, but it's a long story for another day. I realize you could not make a diagnosis based on limited info. But for purposes of this thread I am just looking to understand possible significance to attach to this one particular symptom (which will be taken with others) and to understand how/why dry aged insulation can result in increased PI (question 2).


=====================================
(2B)+(2B)' ?

 

[Addy71]

ElectricPete,

I have seen this kind of behaviour & have seen the correlation in the capacitance & tan delta tip-ups (see the attached document). The top reading for each motor is of 2012 & the bottom reading is for 2011. Couple of these are planned for rewinds. These motors are 6.6 kV with VPI insulation.

Based on the correlation; I assumed that certain cases of VPI degradation can cause high P.I. I do not know why. However; I've seen many other cases wherein P.I. was normal; tip-ups were high & the degradation could be seen when opened. Thus; it is not a rule.

Regards,

Aditya

 

[electricpete]

Thanks.

I take it from the kw rating these machines are somewhere in the range 4kv to 13.2kv.
Do you know if any of them have stress grading coatings?

Why two rows per machine? The top row is more recent data?





=====================================
(2B)+(2B)' ?

 

[Addy71]

The motors are all of 6.6 kV. I do not know for sure that stress grading coatings are present; but this is quite standard for 6.6 kV motors (certainly for the first two; which are of BHEL make).

The top row is for October 2012, bottom row for October 2011.

Why the query about stress grading? There is a very standard pattern in PD if this damaged, did not show up in any. Please share your experience about what you've seen.

Regards,

Aditya

 

[electricpete]

Thanks Aditya. I will have to study that some more - good data.

Stress grading generally results in higher tipups for capacitance and power factor/tan delta. We are reluctant to put any absolute limit on tipup when stress grading is present, only whole machine. I understand if the test is done on individual coils, the grading can be guarded, but not for full machines.

=====================================
(2B)+(2B)' ?

 

[electricpete]

Sorry, my sentences got a little mixed up. Correction in bold:
Stress grading generally results in higher tipups for capacitance and power factor/tan delta. We are reluctant to put any absolute limit on tipup when stress grading is present, only whole machineon the trend. I understand if the test is done on individual coils, the grading can be guarded, but not for full machines.

=====================================
(2B)+(2B)' ?

 

[Addy71]

Here another case; this one does worry me. These are identical 30 MW, 11 kV, 2-pole gas turbine generators; commissioned in 1995. The P. I. values for two phases increased drastically for GTG 1. We've opened the generator; cleaned it (solvents + dry ice), dried it & varnished it. The P. I. values remain high even after all of this. Physical inspection did not reveal anything. The tan delta & capacitance haven't changed at all; PD has gone down!

Regards,

Aditya

 

[electricpete]

I see you are using discharge ratio that you discussed before and it also raised a flag for that same machine. Don't ask me what it means. We still have never used that. I think you use a bigger variety of indicators than anyone else I've heard of. That's good for you and I'm sure we can learn a lot from you.

I looked through some articles to see if there was any explanation for this high PI phenomenon. I found one that seems to discuss this: "Influence of Rise Time on Dielectric Parameters Extracted from Time Domain Spectroscopy in the Context of Generator Stator Insulation" by David and Lamarre, IEEE Transactions on Dielectrics and Electrical Insulation Vol. 12, No. 3; June 2005

They present a complicated but (I think) standard formula for current during dc test vs time (equation 9). Then they apply some simplifications. First simplification is the rise time (beta parameter is 0). That results in equation 10:
PI = (alpha +1) / [alpha * 10^(-n)] [equation 10]
where alpha is conduction current normalized by dividing by 60-sec value of absorbtion current.
n is power law exponent for absorption current component = A*t^(-n)
n = 0.9 for unaged epoxy-mic insulation, decreases with age according to experiment.

Here is discussion based on equation 10:

If the direct conduction is negligible alpha~0) in the above equation, the value of the polarization index
is given simply by PI = 10^n. A large number of measurements have shown that n is approximately equal to 0.8
and 0.9 for unaged asphalt-mica and epoxy-mica materials, respectively, which leads to maximum possible values of PI of 6.3 and 7.9, respectively. It is worth mentioning that the exponent n of the power law is often found to decrease with aging [ref 2, 16], leading to a steeper rise of the dielectric losses toward low frequencies for service-aged samples than for the unaged ones. This effect is shown in Figure 5 of [ref 2].
While low values of PI often mean moisture absorption [ref 12], values higher than the maxima mentioned above
could indicate a significant deterioration of the insulation system. Figure 2a shows the dielectric response of an
epoxy-mica insulated bar after it was submitted to a multi-stress aging experiment conducted in a stator bar
test facility [ref 21]; in this case, the value of PI measured from the charge current was 9.1. Such a high value reflects
the fact that the dielectric response was significantly distorted by aging.

.....[2]E. David, R. Taghizad, L. Lamarre and D. N. Nguyen, ‘‘Investigation
on the Low-Frequency Dielectric Response of Groundwall
Insulation of Rotating Machine Windings’’, IEEE Conf.
Electr. Insul. Dielectr. Phenomena CEIDP., pp. 157_160, 2003.

The words at first glance sound like plausible explanation, but there's a disconnect somewhere.
n going down with age means that if we neglect conduction current and rise time, PI~10^n goes down (not up).
i.e. if n goes down, the polarization current decays slower, the ratio IR1/IR10 is smaller, PI is smaller.
It doesn't seem consistent with the words they're using.
I must have misunderstood what he was saying.
Maybe someone can make sense of above?
Else I will just put it aside and come back to it in a few days to see if it makes sense then.

If ratio of +ve to –ve PD > 1.5; PD is close to copper conductors (aging). If PD Ratio < 0.5, PD is on the slot surface region (loose coils, improper stress grading or contaminated end-winding). If PD Ratio is between 0.5 to 1.5, PD is distributed throughout the insulation (thermal degradation or de-lamination).

That's backwards isn't it?
I always remember it using PONI:
Positive predominance -> pd on Outer surface of insulation
Negative predominance -> pd on Inner surface of insulation


=====================================
(2B)+(2B)' ?

 

[electricpete]

One correction
i.e. if n goes down, the polarization current decays slower, the ratio IR1/IR10 I1/I10 is smaller, IR10/IR1 is smaller PI is smaller.


=====================================
(2B)+(2B)' ?

 

[Addy71]

Here’s an explanation of Dielectric Discharge Ratio from Metrel’s “Guide to Modern Insulation Testing”:

The dielectric discharge test is very useful for testing a multi-layer insulation. This test can identify excess discharge currents that occur when one layer of a multi-layer insulation is damaged or contaminated. This condition will not be detected by both the spot test and the polarization index test. Discharge current will be higher for known voltage and capacitance if an internal layer is damaged. The time constant of this individual layer will differ from other layers, causing a higher current than that of a sound insulation.

Regarding PD Analysis; there's always been a confusion between the American & German way of putting it. My statement actually refers to +VE half cycle (-ve PD) & -VE half cycle (+ve PD). We both interpret actual PD in the same way; just different nomenclature.

I agree that the Lamarre paper isn't really shedding any light on the matter. Will keep hunting for an explanation & post anything that I find.

Regards,

Aditya

 

[rhatcher]

Pete,
You raised a good question. While the explanation for a low PI is easily found and understood, there seems to be a lack of information explaining why aged, dry, thermally stressed, or brittle insulation causes a high PI. To further confuse matters, some references I have seen suggest that a dry/brittle insulation results in a low PI.

It is my understanding that for modern insulation systems a high PI is acceptable and indicates a "good" insulation system. This would suggest that the PI of 7.5 described in your original post is acceptable.

That being said, trending is very important as seen in the data provided by Addy71. His data for the test of GTG 1 on Oct 12 shows readings that are within the permissible levels. However, when considering the trend, something has clearly changed. There is an increased PI after several years of consistently lower readings. Since we know that the insulation did not suddenly become better after years of service, the implication is that something has happened to change it in a negative way. When you also consider the decrease in the IR reading and the increase in the DD reading, there is certainly reason for concern.

This would seem to suggest that there is an upper limit to acceptable PI readings on modern insulation systems. However, there may be no single upper limit that can be applied to every case. This supports the importance of trending when interpreting the results of insulation testing.

I have included an attachment that offers good explanations of PI and DD testing and their results.

 

[electricpete]

Here is an IEEE paper posted on the author's website (note kea in the article and the website).

http://www.keaconsultant.bizland.com/papers/technical/Supatra_EIC2011_official.pdf

D. Thermal problem
Fig. 4 presents the dielectric response results at 18C of
ground insulation of a machine which was overheated due to
high load and insufficient cooling. The similar shape of I pol.
and I depol. in fig. 4a refers to absorption current which is
dominant and has one prominent crook. Insulation resistance
and RC versus time are presented to show the characteristics
for this type of aging. The resistance curve in fig. 4b is steep
and P.I. (between 60 and 600 s) in Table II shows the value of
8.35 which is very high. Experienced from many cases, P.I.
(between 60 and 600 s) higher than 7.0 reveals thermal aging
or overheating. This is another case to confirm that the high
P.I. does not always mean good insulation.

When P.I. is higher than 7, thermal aging can be suspected.

I think this somewhat resolves my confusion about the previous article. I was trying to reconcile direction of change of power law exponent with change in PI. This article I just posted figure 4 shows it is not obeying a power law over the duration of the test (which would be straight line on a log log scale), there is a change in behavior (kink in the curve) associated with this condition.

Putting it all together, the first article tells why we expect PI<7 if epoxy insulation is healthy (therefore follows power law with slope that assures PI<7). The 2nd article shows an example of the deviation that causes PI>7

=====================================
(2B)+(2B)' ?

 

[electricpete]

It is my understanding that for modern insulation systems a high PI is acceptable and indicates a "good" insulation system. This would suggest that the PI of 7.5 described in your original post is acceptable.

You may be right, but I'm inclined not to agree so I'd like to challenge you on this. Have you seen any machines with PI this high?


=====================================
(2B)+(2B)' ?

 

[rhatcher]

Pete,
A challenge? You should know by now that if you have a question about information that I have posted, you can simply ask.

When you "challenge" me, it sounds like we have to face off at twenty paces with slide-rules at the ready.

Yes, I have seen a few PI readings greater than 7.5 and many that were at least 6. However, in every case these were on motors with older windings having many years of service and having IR readings that were much lower than you are reporting. In almost every case, visual examination of the winding indicated an aged insulation system. In some cases varnish treatment improved the PI to the range of 2-4 but, in some cases it did not.

Otherwise, I cannot say for sure if I have seen anything like you describe with very high IR readings and a PI of 7.5 or greater. This is primarily because, per IEEE 43 and EASA recommendations, we disregard the results of any PI with a one minute reading greater than 5000 Meg-ohms such as given in your example.

Of course, regarding the portion of my post that you quoted, I said it was my "understanding", not my "experience", that this was a good reading. The basis for my experience is given above. The basis for my "understanding" follows.

The standard for modern insulation systems is that PI ratios are evaluated using the scale: <1 = "poor", 1-2 = "questionable", 2-4 = "okay", and >4 = "good". This scale does not place an upper limit on what constitutes a "good" reading. When you combine this with the fact that IEEE has concluded that IR readings should be disregarded when the one minute value exceeds 5000 M-ohms, all indications are that the reading that you described is a good reading according to modern standards.

However, this does not include consideration of the results of trending. If you read the rest of my post beyond the initial section that you quoted, you will see that I did identify Addy71's Oct 12 PI readings as indicating a problem based on the trending. I also went on to conclude my post by suggesting that there probably is an upper limit to what constitutes a "good" PI reading but that trending is necessary to identify this. Again, this is because of the absence of a standard that establishes a fixed upper limit.

Trending is the key to identifying the meaning of PI results like you reported. What you quoted was the setup to introducing this idea, what followed was the point that I intended to make. I apologize if my post did not make this clear.

By the way, the paper that you cited on 7 Nov was pretty (lps) good. It would be great if work like this led to adoption of a new set of standards that include a fixed upper limit to what is "good" for PI readings.

Until then and in the absence of extensive testing as described in the paper, trending is a simple and definitive way to determine if there is a problem. If the PI increases over previous readings, a negative change in the insulation has occurred. Testing should continue at an increased frequency (if possible) until the slope of the trend (PI or IR) indicates that a rewind is justified.

What does the trend for your motor show?

 

[Addy71]

I would agree with EPete that extremely high P. I. values do indicate insulation degradation. I have attached data of two identical 9.2 MW, 13.8 kV motors that I tested some months ago. These have been examined and visual indications of degradation & PD observed. A spare motor is being planned. The client is monitoring their condition using an ozone measuring instrument.

I sometimes wonder if the IEEE 43 remark about 5 GOhms was because older instruments could not read well in such a range & needs to be expunged now. The new draft standard still retains this clause but does add many new sections on qualitatively assessing the insulation (copy attached in the next post).

Regards,

Aditya

 

[Addy71]

IEEE 43 Draft Standard

 

[electricpete]

Thanks Ray and Aditya
I could not read the file from previous post.

=====================================
(2B)+(2B)' ?