Chemist’s Perspective

Sampling Transformer Oils

This three-part article deals with the art of sampling transformer oil.

T
he results obtained from the analysis of an insulating oil sample
can provide unparalleled information concerning the condition of
the insulating materials within electrical apparatus, life assessment,
and the operating condition of the device. However, a sample that is not by Lance R. Lewand
Doble Engineering Company
representative of the bulk oil insulation can provide erroneous infor-
mation which can easily mislead maintenance personnel to incorrectly
assess the condition of the oil or the electrical apparatus. In one case,
samples taken from two transformers showed very high concentrations
of hydrogen and no other gases, prompting maintenance personnel to What is a Good Sample?
give these units priority for diagnostic surveys. It was later found that Simply put, a good sample is
one that is representative of the
the cause of the high hydrogen was a galvanic reaction occurring in the content of the bulk oil insulation.
drain valves in which water was converted to hydrogen because of the Since samples are usually retrieved
interaction of a galvanized fitting with a dissimilar metal. from a drain valve or the attached
sampling cock, preparation of that
area is important to obtain a good
Why Sample? sample. Cleaning the drain valve
For in-service oil-filled electric apparatus, sampling of the oil provides and the sampling cock inside and
a method to determine the condition of the solid and oil insulation as out is the first step in avoiding
well as the operating condition of the apparatus without opening or sample contamination. Cleaning
de-energizing the apparatus. This is especially important in the present the outside of the drain valve is
utility and industrial climates, as equipment outages for out-of-service just as important as cleaning the
testing have become very limited. Sampling provides a means to check inside. The dirt and debris falling
the condition of oil in storage, whether it be new or used, and to de- off the outside of the valve into
termine if it complies with specifications such as TOPS, ASTM D 3487, the sample container during the
IEC 60296, IEEE C57.106, or company specifications. Sampling can also sampling process can contaminate
help to determine: many samples.
Most of the contamination in
1) If accidental mixing of different dielectric oils has taken place; the apparatus consists of water
2) If the method of transportation contaminated the oil; and particles (paper fibers, metal
particles, etc.), which over time
3) If the handling equipment to transfer the oil contaminated the will settle out on the bottom of
product. the apparatus near the drain valve.

Winter 2002-2003 1
This material needs to be flushed out of the system to drain-valve plug or opening the sampling cock will
get to the bulk oil insulation. It is necessary to remove create particulate metals. This is due to the grinding
at least one to two liters of oil from the drain valve, of the surfaces between the valve body and the drain
cap the drain valve, and then flush out the sampling plug or sampling cock. In fact, it is becoming more
cock before proceeding with sampling. On occasion, apparent that that these types of samples should only
two liters will not be sufficient, especially when sam- be retrieved after a minimum of two (and sometimes
pling a nonenergized transformer or certain OCBs and three to four) liters of oil have been passed through
LTCs. Specific sampling techniques and precautions, the drain valve.
especially those dealing with low volume electric The same is true of retrieving a sample for particle
equipment, are detailed in the Doble Reference Book on count where valve debris, whether inside or outside,
Insulating Oils and Gases and ASTM Practices D 923 can severely skew the results. The debris, soot, and
and D 3613. grime that exist on the outside of the drain valve are of
serious consequence, especially in industrial locations.
Good Samples Versus Bad Samples This debris can be easily transferred to the sample
bottle while the sampling process is taking place. This
It is sometimes very clear to the laboratory per-
validates the importance of cleaning the outside of the
forming the analysis on the oil that the sample was
valve prior to taking the actual sample.
taken improperly. For example, the presence of free
Dissolved gas-in-oil analysis is another test im-
water or foreign objects such as insects, pipe sealing
pacted by sampling, drain valve components, and
tape or putty are strong indicators that the drain valve
sampling materials. When galvanic fittings (zinc
was not adequately flushed out prior to sampling.
coated) are used in the drain valve assembly — such
Once analysis has begun and it is determined that
as the drain plug — a galvanic reaction with water can
there is a high or free water content coupled with a
cause very high levels of hydrogen to be produced.
low dielectric strength, with all the other test results
If this residue is not flushed out adequately then it
being acceptable, then it strongly indicates that the
will be transferred to the sample and included in the
proper sampling technique was not adhered to. It may
analysis, causing a level of concern that is not war-
even imply that some chemical reactions were taking
ranted. Galvanic plumbing fittings such as nipples
place in the drain valve that were not representative
can also have the same effect. Brass, bronze, stainless
of the bulk oil insulation.
steel or black iron should be the only materials used.
In addition, drain valve assemblies should not be
Lab Tests Most Easily Affected composed of dissimilar metals as corrosion can result,
As indicated previously, the analytical tests most which may end up in the sample. Debris, water and
easily affected by sampling are dielectric strength and other ionic contaminants also affect the power factor
water content. This is due to the fact that drain valves test. These materials increase dielectric loss, which in-
are usually at very low points in the tanks where de- creases the power factor. Incompatible inorganic and
bris and water accumulation occurs. Water can also organic materials from the drain-valve stem packing
be present as a result of condensation that occurs in or drain-plug sealants can also have the same effect
the drain valve, which is due to the position of the on the power factor.
drain valve on the tank. In most cases the drain valve
protrudes 15 to 30 centimeters (six to 12 inches) away Costs Associated with a Bad Sample
from the main tank. From experiments performed at
In the case of a single sample, the costs for routine
Doble Engineering, it was found that the oil in many of
oil quality analysis and DGA testing are just a very
these valves varies in temperature from eight degrees
small fraction of the total costs associated with taking
Celsius to 15 degrees Celsius cooler than the bulk oil
and analyzing a sample. Some of the items and costs
insulation. When oil or air has an elevated relative
associated with sampling and analysis are:
saturation or humidity and there is a significant cool-
ing, condensation of water will occur. This is exactly ITEM COST ($)
what happens in a drain valve. Other analytical tests Labor to take sample 275
easily affected by sampling are dissolved metals, par-
Materials to take sample 15
ticulate metals, particle counts, dissolved gases-in-oil,
Packaging and shipping cost 8
and power factor.
The concentration of metals, whether dissolved Analysis cost 70
or in a particulate state, are especially impacted by Engineering evaluation of the data (10-15 min) 35
the amount of cleaning performed on the drain valve TOTAL 403
and the amount of flushing that is performed. Debris
that settles to the bottom of the apparatus and subse- Every situation is different, but in many cases the
quently into the drain valve can consist of metal par- analysis cost is only about 17 percent of the entire sam-
ticles. In addition, just the simple fact of removing the pling and data review process. In a situation where the

2 NETA WORLD
sample has been determined to be nonrepresentative “IEC 60296: Specification for Unused Mineral Insu-
of the bulk oil insulation, the following additional lating Oils for Transformers and Switchgear” In-
costs may also be incurred: ternational Electrotechnical Commission, 3, rue de
Varembe, Geneva, Switzerland, 1982.
ITEM COST($)
Labor to take original sample 275 “IEEE C57.106-1991: IEEE Guide for Acceptance and
Materials to take original sample 15 Maintenance of Insulating Oil in Equipment”, IEEE,
Packaging and shipping cost for original sample 8 345 East 47th Street, New York, NY, 1992
Analysis of original sample 70
Engineering evaluation of the data (10-15 min) Reference Book on Insulating Oils and Gases, edited by
of original sample 35 the Doble Client Committee on Oil Insulation, 1993,
Additional engineering time to confirm Doble Engineering Company, Watertown, MA.
sample was nonrepresentative 35
Labor to take 2nd sample 275 “ASTM D 923: Standard Practice for Sampling Elec-
Materials to take 2nd sample 15 trical Insulating Oils” in Electrical Insulating Oils
Packaging and shipping cost for 2nd sample 8 and Gases; Electrical Protective Equipment, Annual
Analysis cost of 2nd sample 70 Book of ASTM Standards, Vol. 10.03, ASTM, West
Engineering evaluation of the data
Conshohocken, PA, 2001.
(10-15 min) of 2nd sample 50
“ASTM D 3613: Standard Practice for Sampling Elec-
TOTAL 856 trical Insulating Oils for Gas Analysis and Deter-
mination of Water Content” in Electrical Insulating
Oils and Gases; Electrical Protective Equipment, Annual
The cost of taking a bad or nonrepresentative sam-
Book of ASTM Standards, Vol. 10.03, ASTM, West
ple has more than doubled from the original total.
Conshohocken, PA, 2001.
This is in part due to the fact that review of data from
the second sampling takes longer as there is a more
Griffin, P. J. “Water in Transformers – So What!” Na-
critical and thorough review.
tional Grid Condition Monitoring Conference, May
If the original sample was not recognized as bad, the
1996.
costs associated with that sample can be staggering.
For example, a bad sample could cause a customer Lance Lewand received his BS degree at St. Mary’s College
to try a remedial effort in an attempt to improve the of Maryland in 1980. He has been employed by the Doble Engi-
condition of the insulating oil — such as processing neering Company since 1992 and is currently Project Manager of
the oil through clay or vacuum-processing a trans- Research in the materials laboratory and Product Manager for the
former to remove moisture — then associated costs DOMINOTM product line. Prior to his present position at Doble,
he was the Manager of the Transformer Fluid Test Laboratory and
may skyrocket to between $10,000 and $30,000. This PCB and Oil Services at MET Electrical Testing in Baltimore, MD.
is one reason why Doble always recommends taking a Mr. Lewand is a member of ASTM committee D 27.
second sample to confirm the results of the first before
any remedial activities begin. Other factors, such as
accidental sample switching or misidentification, can
also be the source of an erroneous assessment.
Part two of this series will cover the sampling prac-
tices to follow and the science of sampling.

References
“Items of Interest” in The Doble Exchange, The Doble
Engineering Company, Watertown, MA, USA, Vol-
ume 11, Number 3, September 1993, Page 4.

Transformer Oil Purchase Specification (TOPS), edited

“ASTM D 3487: Standard Specification for Mineral

Winter 2002-2003 3