Dissolved Gas Analysis of

Transformer Insulating Oils by

Headspace/Capillary Gas
Chromatography
Application Note 228-310

Headspace
October 1995

Authors

Introduction

Jocelyn Jalbert and Roland

Gilbert
Institut de recherche dHydroQubec
1800, monte Ste-Julie,
Varennes, Qubec, Canada
J3X 1S1

The decomposition of transformer

insulation materials results in the
formation of gases which then
partially or totally dissolve in the
oil. Among the gases formed are
H2, CO and CO2. These can be
found along with the hydrocarbons
CH4, C2H2, C2H4, C2H6 and C3H8 . By
monitoring the variations in the
concentration of these gases in the
oil, electrical utility industries
diagnose the presence of faults at
an early stage and take preventive
action before failures occur.1,2 This
monitoring is usually done in North
America using Method A of the
D 3612 standard procedure
approved by the American Society
for Testing and Materials (ASTM).3
This method involves degassing an
oil sample into a pre-evacuated
vessel, followed by injecting an
aliquot of these gases into a gas
chromatograph (GC).

Susan M. Brillante
Hewlett-Packard Company
Wilmington, DE 19808-1610

Abstract
The analysis of dissolved gases
in electrical insulating oils is an
efficient diagnostic tool
for routine performance
monitoring of power system
equipment. The analytical
performance of a method using
the HP 5890 Series II GC
system with HP 7694 headspace
sampler with NiCat is evaluated
and compared with that of the
ASTM D 3612, Method A
standard procedure. This
comparison shows that the
headspace technique can be
used to determine dissolved
gases with the same precision
but shorter analysis times and
lower detection limits than the
ASTM method.

Previous work4,5 has shown the

feasibility of using headspace
sampling techniques for
automatically performing a
dissolved-gas analysis. This
technique relies on the
establishment of an equilibrium
between the oil and vapor phases

in a closed vessel. Diffusion of an

analyte from a viscous matrix such
as insulating oil into the headspace
can be accelerated by agitating the
sample during the equilibrium
period.6 Results and conditions for
the analysis of dissolved-gas
standards as well as a series of inservice transformer oil samples are
shown. The analytical performance
of the HP 5890 Series II
GC/HP 7694 headspace sample
system was compared to that used
for Method A of ASTM D 3612.

Experimental
Instrumentation
A schematic of the system used is
shown in Figure 1. Sampling was
done using an HP 7694 headspace,
equipped with a 44-sample outer
carousel feeding a 6-sample inner
carousel in a thermostatted oven
compartment. This sampler is
equipped with a vibrator to
accelerate analyte equilibrium rate
and improve detection limits. An
equilibration temperature of 70C
was used. All samples were placed
in 20-mL glass vials sealed with
perforated aluminum caps fitted
with a rubber PTFE-lined septum.
All vials were purged with Argon

___________
2

prior to the addition of 15 mL of oil

sample. The oil sample was
introduced using a 30-mL
Perfektum syringe equipped with
a three-way valve and a Precision
Glide 18G 1.5 needle. The pressure
buildup was released through a
25G needle. An HP 5890 Series II
gas chromatograph equipped with
two detectors, was used for all
analyses. A thermal conductivity
detector (TCD) was used for the
quantitation of H2, O2 and N2. A
flame ionization detector (FID
[with a nickel catalyst to convert
CO and CO2 into CH4]) was used to
quantitate the amounts of CH4, CO,
CO2, C2H6, C2H4, C2H2 and C3H8.

Table 1a. HP 7694 Headspace Sampler

Conditions
Temperatures:
70C
150C
150C

Sample
Transfer Line
Injection Valve

Pressure:
0.4 bar

Vial

Times:
0.25 min
0.25 min
0.25 min
0.90 min
30 min

Pressurization
Pressure Equilibrium
Expansion in Sample Loop
Injection
Mixing

Mixing Speed:
Power 2 (high)

30 minutes

Sample Preparation

Table 1b. Chromatographic System

HP 5890 Series II GC Parameters
Carrier Gas: Argon @ 12 mL/min

The light hydrocarbons were

separated on a 30 m 0.53 mm
ID Carboxen 1006 PLOT column
(Supelco). A 25 m 0.53 mm
MolSieve 5 PLOT column with
50 m adsorbant was used for the
permanent gases (Chrompack).
Chromatographic and static
headspace conditions are shown in
Tables 1a and b.

Lecture bottles of H2, N2, CH4, CO2,

C2H6, C2H4 and C3H8 (all Research
grade) and Purified grade C2H2
were purchased from Matheson
Gas Products to prepare the
dissolved-gas standards. The
headspace was calibrated using
certified gas mixtures of H2, O2, N2,
CO, CO2, CH4, C2H6, C2H4, C2H2 and
C3H8 purchased from Scott
Specialty Gases in the
concentration range of 100,
1000 and 5000 ppm in Argon.

Temperatures:
FID
TCD
NICAT II

350C
250C
350C

Oven:
40C for 3 minutes
24C/min to 170C for two minutes
24C/min to 250C for five minutes
Isolation Valve timing:
0-3.6 min
3.6-8 min
8-10 min
10-23 min

columns in series
MolSieve bypassed
columns in series
MolSieve bypassed

The set-up used for degassing the

oil is shown in Figure 2a. The oil
flows from the 4-L funnel into a
75-mL degassing bulb, and then
into a 5-L flask where it undergoes
ultrasonication and degasification
by a mechanical vacuum applied at
points A and B. A Pirani II gauge is
used to detect leaks occurring
during the 48-hour degassing
period. The quality of the blank
sample was established by
determining the residual gases;
these results are shown in Table 2.

Figure 1. Schematic diagram of the Hewlett-Packard GC/headspace transformer oil gas analysis system.

___________
3

A primary dissolved-gas standard

was prepared using the configuration shown in Figure 2b. A
precise volume (5, 10 or 20 mL) of
each gas was injected into the
5-L flask containing the blank
sample. The final composition of
the standard mixture is shown in
Table 2. Intermediate standards
were prepared by diluting aliquots
of the primary standard with blank
sample using a 30-mL or 100-mL
Perfektum syringe fitted with a
three-way valve.

Results and Discussions

Sample loop volume
The effect of the sample loop
volume on sensitivity was studied
by expanding the headspace gas
taken from a 15-mL dissolved-gas
standard into loops of 0.5, 1, 2, 2.5,
3, and 5 mL. The measurements
performed with three replicates
each, showed good signal linearity
for all gases up to about 2.5 mL.
After that, the volume of headspace gas was not large enough to
purge and fill the sample loop.
Therefore, a 2.5-mL loop was used
for the rest of the study.
Equilibration time
One of the most important
parameters in headspace analysis
is the time required for the analyte
to reach equilibrium between
the two phases in contact.
This parameter was studied
by measuring the gas phase
concentration of vials containing
15 mL of dissolved-gas standard
as a function of time. After placing
vials in the inner carousel of the
headspace unit, data was collected
at 0, 2, 4, 8, 16, 32 and 64 minutes
equilibration time with high
mechanical agitation. Three
replicates were analyzed at each
specified time. The data revealed
that the analytes reached at

Figure 2a & 2b. Set-up used for the preparation of blank and dissolved-gas samples.
Table 2. Residual Gases in the Blank
Sample and Composition of the
Standard at 294.45K and 758.5 Torr

Gas

Blank
Sample
(ppm)

H2
O2
N2
CH4
CO
CO2
C2H2
C2H4
C2H6
C3H8

<0.6
17.0
24.0
<0.6
1.6
8.8
<0.05
<0.04
<0.04
<0.1

Standard
(ppm)
960

3800
960
960
1900
950
960
960
960

least 90% of their equilibrium

concentration (t90) in 15 minutes
when agitated. Previous data
indicated that without agitation,
this process required 180 minutes4.
However, any time difference less
than 30 minutes is not important
because the chromatography is the
limiting step. As shown in Figure 3,
it took 14 minutes for the gases to
separate, with an additional
5 minutes to purge the columns

and about 6 minutes to return to

initial GC conditions. The equilibrium time was, therefore, set to
30 minutes. Under these conditions, the analysis took 55 minutes
for the first sample and 25 minutes
for each subsequent sample.
Calibration
As expressed by Equation 1, it
is not necessary to use dissolvedgas oil standards to calibrate
the chromatographic response
when the Ostwald coefficient (Ki)
for each analyte is known and a
linear relation exists between COL
and CG in the range of concentrations of interest.
[Equation 1]
COL = CG(Ki + VG/VL)
O
where C L is the initial concentration of the analyte in the oil
sample; CG is the concentration of
the analyte in the gas phase at
equilibrium; and VG, VL refer to the
volumes of the gas and the oil
phases in the headspace vials,
respectively.

___________
4

If Ki and the VG/VL ratio are known,

then COL can be calculated from the
chromatographic determination of
CG. The only response to be
calibrated, therefore, would be CG.
The following Ostwald coefficients
determined at 70C in a previous
study7 were used:
KH2 = 0.07, KO2 = 0.17, KN2 = 0.11,
KCH4 = 0.44, KCO = 0.12, KCO2 = 1.02,
KC2H4 = 1.47, KC2H6 = 2.09, KC2H2 = 0.93
and KC3H8 = 5.4. Calibration of CG
was done by introducing different
volumes of certified gas mixtures
into a headspace vial containing
15 mL of mercury. The vial was
filled with mercury to keep the
VG/VL ratio of 0.67 constant at
ambient temperatures. Calibration
curves were established over the
range of 1 to 50 ppm for all gases
except for O2 and N2, where the
upper limit was fixed at 1400 ppm.
The 1150 ppm calibration range
was inclusive for almost all
headspace concentrations obtained
from equilibration of real
transformer oil samples.

a-FID

b-TCD

Figure 3a & 3b. FID and TCD chromatograms obtained for the primary oil standard (see
composition in Table 2).

Table 3. Detection Limits in ppm

ASTM D 3612
Gas

Headspace

H2
O2
N2
CH4
CO
CO2
C2H2
C2H4
C2H6
C3H8

0.6
11.0*
11.2
0.06
0.09
0.1
0.05
0.04
0.04
0.2

5
50
50
1
25
25
1
1
1
1

20
500
500
1
2
2
1
1
1
1

a-FID

Figure 4a & 4b. FID and TCD chromatograms obtained for the intermediate standard
of lowest concentration (1 ppm H2 and CXHY, 11 ppm CO2, 3 ppm CO, 50 ppm O2 and 250
ppm N2).

Analytical Performance

dissolved-gas standard shown in

Table 2. Intermediate samples
were prepared by successive
dilution of the primary standard
with blank oil until a final
concentration of 1 ppm for all
gases was reached; except for O2,
N2, CO and CO2 where the final
concentration was their residual
level in the blank oil (see Table 2).

The linearity of the response was

confirmed over a range of 1 ppm to
the concentration of the primary

Typical chromatograms for the

sample with the lowest concentration studied are shown in

*Estimated from the H2 response.

b-TCD

Figure 4. These results were used

to establish the detection limits
(S/N = 3) which were lower than
the best values reported for ASTM
D 3612 (Method A or B) for
all analytes (see Table 3).
The precision of the method was
assessed by obtaining three
consecutive measurements at low
(3.2 ppm H2 and CXHY, 4.8 ppm CO
and 15 ppm CO2) and medium
(239 ppm H2 and CXHY, 241 ppm CO
and 487 ppm CO2) levels of

___________
5

The accuracy of the measurements

was assessed by analyzing 35 field
samples with the headspace
sampling technique and by
Method A of the ASTM procedure.
The data from the two methods
were then compared. As
demonstrated in Figures 5 and 6,
very good correlation was
observed for the gases generally
used for detection of the
decomposition of transformer
insulation materials. As shown in
Figure 5, headspace sampling was
sufficiently sensitive to produce
quality results for H2 over the range
of 0 to 20 ppm. Figure 6 shows the
correlation between the two
methods for oxygen. All other
analytes showed similar
correlations.

360
Headspace (ppm)

340
320
60
50
40
30
20
10
0

n = 31

0 10 20 30 40 50 60 320 340 360

ASTM D 3612 (ppm)

Figure 5. Correlation for hydrogen.

References
4
Headspace (x104 ppm)

dissolved gas. An overall precision

better than 5% RSD (relative
standard deviation) was observed
for the medium levels. This is
equivalent to RSDs obtained for
ASTM D 3612 Method A in the
laboratory4. Precision at low levels
was better than 10% RSD for the
HS method. Comparison to ASTM
D 3612 (Method A or B) results
were not possible since ASTM data
was not available for the low levels
investigated in this study.

1. Galand, J.; Thibault, M.; Viale,

F.; Samat, J.; Vuarchex, P. Rev.
Gen. Elect. 1972, 81, 727-739.

2. Duval, M.; Langdeau, F.;

Blanger, G.; Gervais, P. Minutes
of the 55th International
Conference of Doble Clients, 1988,
10, 7.1-7.10.

2
1
n = 35
0
0

ASTM D 3612 (x10 4 ppm)

Figure 6. Correlation for oxygen.

Conclusion
A comparison of ASTM Method
D 3612 A standard procedure and
the equivalent GC headspace
technique, shows that the GC
headspace sampling technique
compares favorably with the
vacuum gas extraction method.
Thirty-five samples of transformer
oil were analyzed using both
methods and the GC headspace
results showed excellent
correlation with the ASTM method.
The high sensitivity of detection
achieved and the fact that the GC
headspace system is automated are
major advantages. The GC
headspace technique can be used
to assess dissolved gases from any
equipment where the total gas
content is very low. The
mechanical vibration of the
samples using the HP 7694
headspace sampler permitted the
analyses to be performed in a
much shorter time with less
handling than Method A of the
ASTM D 3612 standard.

3. American Society for Testing

and Materials Method D 3612,
Annual Book of ASTM Standards,
1993.
4. Leblanc, Y.; Gilbert, R.; Duval,
M.; Hubert, J. J. Chromatogr.,
1993, 633, 185-193.
5. Leblanc, Y.; Gilbert, R.; Jalbert,
J.; Duval, M.; Hubert, J. J.
Chromatogr., 1993, 657, 111-118.
6. Westendorf, R. G.; Lehan, H. J.
Technical paper presented at the
41st Pittsburg Conference, New
York, NY, 1990.
7. Jalbert, J.; Gilbert, R.
Unpublished results, 1994.
8. Jalbert, J.; Gilbert, R,
Determination of Dissolved Gases
at the ppb Level in Transformer
Oils by Static Headspace
Analysis, submitted to IEEE
Transactions on Dielectrics and
Electrical Insulation, July 1995.

Copyright 1995
Hewlett-Packard Company
Printed in USA 10/95
(23) 5963-7118E