See discussions, st ats, and author pr ofiles f or this public ation at : https:www .researchgate.ne tpublic ation3340868 [627754]

See discussions, st ats, and author pr ofiles f or this public ation at : https://www .researchgate.ne t/public ation/3340868
Determination of gas pressu re in vacuu m interru pter based on partial discharge
Article    in  IEEE T ransactions on Dielectrics and Electric al Insulation · July 2007
DOI: 10.1109/TDEI.2007.369518  · Sour ce: IEEE Xplor e
CITATIONS
14READS
382
5 author s, including:
Some o f the author s of this public ation ar e also w orking on these r elat ed pr ojects:
Dischar ge Pr operties and Emitt ed Electr omagne tic W ave Spectrum in L ow Vacuum R egion of V acuum Int errupt er View pr oject
RBD P alm Olein & Envir otemp FR3 Mixt ure View pr oject
Mohamad Kamar ol
Univ ersiti Sains Malaysia
110 PUBLICA TIONS    318 CITATIONS    
SEE PROFILE
All c ontent f ollo wing this p age was uplo aded b y Mohamad Kamar ol on 13 June 2014.
The user has r equest ed enhanc ement of the do wnlo aded file.

IEEE Transactions on Dielectri cs and Electrical Insulation Vol. 14, No. 3; June 2007
1070-9878/07/$25.00 © 2007 IEEE 593
Manuscript received on 24 May 2006, in final form 24 November 2006. Determination of Gas Pressure in Vacuum Interrupter
Based on Partial Discharge

Mohamad Kamarol, Shinya Ohtsuka and Masayuki Hikita
Kyushu Institute of Technology.
Department of Electrical Engineering,
1-1 Sensui-cho, Tobata-ku, Kitakyushu-shi, 804-8550 Japan
Hitoshi Saitou and Masayuki Sakaki
Japan AE Power System Corporation
585, Shimonakamizo Higashi Makado,
Numazu, 410-0865 Japan

ABSTRACT
Partial discharge characteristics in ai r under a low vacuum region were studied to
develop a diagnosis technique to determine the gas pressure in a vacuum interrupter.
The pressures were set at from 1.3 Pa to 2.6 kPa in order to simulate the leakage of gas
into the vacuum interrupter. The stru cture of the vacuum interrupter, the
measurement and the circuit construction fo r the experimental setup were described.
The measurement of partial di scharge occurring inside the vacuum interrupter was
performed with a current transformer, intensified charge coupled device (ICCD) camera and photomultiplier tube . The measurement of partial discharge light intensity
with a photomultiplier tube was more sensit ive compared with that measured with a
current transformer . From this result, an attempt has been made to distinguish the
pressure below and above 260 Pa on the basis of the rise time and peak intensity of
discharge light pulses. A relat ively longer rise time (2 µs) with a smaller magnitude
(less than 0.5 mA) was attributed to a Townsend-like discharge at pressures below 260 Pa, while a sharper rise time ~ (10-100 ns) with a larger magnitude (greater than 1 mA) was characterized as a streamer-like discharge above 260 Pa. In addition, the estimation of gas pressure in a vacuum in terrupter was made based on phase-resolved
discharge characteristics utilizing an artificial neural network.
Index Terms – Partial discharges, di scharge light intensity, low vacuum, vacuum
interrupter, neural network.

1 INTRODUCTION
A VACUUM circuit breaker (VCB) is an alternative device
that may be employed to contribute to reducing usage of SF 6 gas
insulated apparatus in the future [1, 2]. There are several classes of VCBs used in power systems such as medium voltage (MV) and high voltage (HV). The improvement with increasing the rated voltage of VCB has been the subject of numerous studies [3, 4]. The trend of research conducted on the improvement of vacuum technology to the higher rated voltage indicates that the use of a VCB in power distribution and transmission will expand. Today, VCB is widely used in electrical power distribution, such as cubicle-type insulated switc hgear (C-GIS) and air insulated
switchgear (AIS), to control and protect power lines and equipment under conditions of ove rload, short-circuit, over-
current and other faults. Since the application of vacuum technology is increasing, the performance of the vacuum interrupter (VI) should be maintained in order to have stable and
reliable system operation. The stability and reliability of the system can be ensured by detecting and improving the performance of the VCB. Detecting the gas pressure is one of the suggested ways to monitor the VCB performance, prevent serious faults, and reduce maintenance costs.
Insulating
envelope Arc shield Metal bellows
Fixed electrode Movable electrode
Main electrode
Figure 1. Structure of vacuum interrupter.

M. Kamarol et al.: Determination of Gas Pressure in Vacuum Interrupter Based on Partial Discharge 594
VCB in C-GIS and AIS are surrounded by SF 6 gas and air,
respectively. VI is the key part of the VCB. Figure 1 depicts the structure of VI [5]. It is structur ed with a cylindrical shape called
an insulating envelope, which is evacuated to a high vacuum as much as 10
-6 Pa. It consists of fixed and movable electrodes, an
arc shield surrounding the electrodes, and a metal bellows to support the movable electrode.
Knowing the gas pressure in VI after long periods of operation
is important for both manufacturers and users. The gas pressure of VI may increase gradually after long periods of operation. During the operation, when the gas pressure of VI exceeds the
order of 10 Pa, partial discharge (PD) may occur and lead to an interruption failure. If no measures are taken, this phenomenon is not detected and it could lead to a serious failure to the VI as well as the operation system. A serious fault to the operation system due to interruption failure brings heavy losses to both manufacturers and consumers.
The estimation of the gas pressure in the VI after a long period
of service is of interest to users. Numerous diagnostic apparatus have existed in the market with a simple indication such as good-bad indicators. The Magnetron method is impracticable to use in the field because detaching of the vessel would be necessary.
Such an operation should be performed only by a well trained operator under guidance of the manufacturer of the device. The diagnostic devices used in the field should be simple to operate and to interpret results. Many studies have been devoted to investigation of gas pressure and the ability of practical application [6, 7]. Although the diagnosis technique for detecting the gas pressure of VI based on PD has been studied recently [8],
there have been no adequate st udies on discharge characteristics
and electromagnetic wave spectrum (EMW) in low vacuum of practical VI. The goal of this study is to develop a reliable diagnostic technique for gas pressure in the VI based on detection of PD. To acquire good knowledge to develop a gas pressure estimation based on PD, an understanding of PD properties in a practical VI is necessary. For this reason, investigation of PD properties and EMW spectrum in practical VI were conducted previously [9-11]. Continuing from previous work this paper deals with the characteristics of PD in the low vacuum region of a practical VI under the rated voltage. The relation between the phase-resolved PD pattern and the gas pressure is also investigated. Eventually the phase-resolved PD
characteristics (
φ-q-n) are used to estimate the gas pressure of
VI with a back propagation neural network (BPNN).

2 EXPERIMENTAL SETUP
In a practical situation, VI is basically operated under a
closed contact condition and the contacts of VI are opened during current interruption due to over-current flow. Thus, study of the detection of gas pressure based on PD is more important for a closed contact condition rather than an open contact condition. In this experiment the glass vessel of VI was used for a better understanding of the optical PD properties in the low vacuum region and to know the location of PD occurrence as the pressure varies at constant applied
voltage. In addition, since the optical measurement has a high sensitivity and cannot be influen ced by the external circuit structure, it gives an accurate and fundamental knowledge of
the PD phenomena and gas pressure estimation based on the PD. This can be as a reference for the measurement technique based on PD. Moreover before starting the experiment, the outgassing influences were confirmed. The internal gas of VI was evacuated to the specified pressure and kept for a longer time to observe the changes of the internal pressure. The
residual pressure was maintained for 6 h and the measurement was conducted within this period.
Figure 2 depicts the equivalent circuit for the closed contact
condition used in the measuremen t of PD that simulates an
actual configuration condition in C-GIS. A capacitance C
1
was connected to the high voltage side of a transformer in parallel with VI. C
1 has a value of 3000 pF to demonstrate the
capacitance of an actual cable us ed in the field which has 0.2
µF/km. Thus, the practical cable length was 15 m. CS is the
stray capacitance between the shield and the tank wall of an actual C-GIS. The value of the C
S was estimated to be around
10 pF. The distance between the ground plate and shield is fixed at 27 mm in order to maintain the value of C
S. The gap
distance was obtained at the valu e of 10 pF with the electric
field calculation.
The measurement of PD occurring inside VI was performed
with a current transformer (CT), intensified charge coupled device (ICCD) camera and photomultiplier tube (PMT).
The frequency response of the CT covering from 10 kHz to
250 MHz was clamped at the bottom side of C
1. The CT was
measured and found to have a flat frequency response from 3 MHz to 250 MHz. Since PD pulses have a high frequency component over 30 MHz, the distortion of CT can be ignored. The output of CT was connected to an oscilloscope (LeCroy 9362, 1.5 GHz) to measure the PD current occurring inside VI. The light emission of PD was m easured with an ICCD camera
(Hamamatsu model C7772-2). The ICCD camera with a UV filter from 240 to 370 nm of the spectrum range was fixed 1.5 m from VI. The PMT having a spectrum range from 300 to 850 nm was located 11 cm from VI at the movable electrode side near the bellows to measure the light intensity of PD.
The
φ-q-n pattern of PD occurring in the low vacuum region
of VI was measured using a partial discharge measurement device (PARADISE: Shoei Electronics). The output of the CT and PMT were connected to a tuning amplifier. Then the 100V
50kV/100V V 400V/50kV
C1=3000pF
Oscilloscope CT Shield VI
CS=10pF P=1.3 Pa~2.6 kPa
CS: stray capacitance
ICCD
camera Tuning
Amplifier Paradise
PCPMT
PC

Figure 2. Equivalent circuit diagram of experimental setup.

IEEE Transactions on Dielectri cs and Electrical Insulation Vol. 14, No. 3; June 2007 595
output of the tuning amplifier was processed into digital data
using PARADISE and a computer.
The experiment was conducted under an ac applied voltage
(Va). A transformer with a 400 V/50 kV and 75 kVA capacity
was used as the power source and the applied voltage was measured at the third winding of the transformer.

3 E XPERIMENTAL RESULTS AND DISCUSSION
3.1 PARTIAL DISCHARGE AND LIGHT EMISSION AT
VARIOUS APPLIED VOLTAGES
Figure 3 shows the partial discharge inception voltage (PDIV) verses the internal pressu re of VI in air. The PDIV of
VI filled with air was measured using CT and PMT in the pressure range 1.3 Pa – 6.6 kPa. The measurement result indicates that PDIV increases w ith the increase of the internal
pressure. Thus, the pressure dependence of the PDIV is considered. The measurement of PDIV in air with CT was
only detectable in the pressure range of 260 Pa – 6.6 kPa due to the low sensitivity of the measurement device. The PD current measured at 130 Pa had a very small magnitude below the noise level of the measuring device even though the applied voltage was increased to 7 kVrms. The occurrence of
the PD at 130, 660, 1300 and 2300 Pa at 7 kVrms obviously can be seen with an ICCD cam era as shown in Figure 4. The
light emissions of PD seem to scatter on the surface of the glass envelope at both the mova ble (bellows side) and fixed
side of the electrodes. The luminosity of PD captured by the ICCD camera especially at 130 Pa indicates that the PD is still
occurring at this level of pressure though it cannot be detected with CT. The PD at 660 Pa and above seems to occur at the edge of the bellows and the shield of VI. Thus, two different types of PD can be considered for the pressure below and above 660 Pa. These results will be explained in section 3.2.
Figure 5 shows the PDIV verses the number of experiments
in 1 Pa order measured with PMT. The voltages were applied 20 times until PDIV appeared. Then this process was repeated 2 to 3 times by re-evacuating the VI to the same vacuum level and refilling the VI with air to the pressure of 1 Pa order. It is apparent from the results depicted in Figure 5 that the occurrence of PD at 1 Pa order would result in the immediate increase of the internal pressure of VI to pressure giving the minimum PD inception voltage. The immediate increase of the internal pressure may be related to the out-gassing or gas evolution due to PDIV. This information indicates that the measurement of PD is more applicable to estimating the gas pressure and feasibility at the pressure giving the minimum of PD inception voltage.

3.2 PARTIAL DISCHARGE LIGHT INTENSITY AT
VARIOUS PRESSURES
At a given pressure and applied voltage, PD light pulses
appearing at any phase angles were found to have a similar waveform. The results also reveal ed that the phase distribution 14
12
10
8
6
4
2
0PDIV (kVrms)
100101102103104105
Pressure (Pa)CT
PMT

Figure 3. PDIV as a function of internal pressure. 5
4
3
2
1
0PDIV (kVrms)
20 15 10 5 0
Numbers of experiments13 Pa
39 Pa1.3Pa 1st time
2nd times
3.9Pa 1st time
2nd times
3rd times
6.6Pa 1st time
2nd times

Figure 5. PDIV detected with PMT verses the number of
experiments.
Shield Insulating envelope Electrode Bellow
( a) Pressure at 130 Pa (b) Pressure at 660 Pa

(c) Pressure at 1.3 kPa (d) Pressure at 2.6 kPa

Figure 4. Partial discharge light emission captured by ICCD camera at 7
kVrms.

M. Kamarol et al.: Determination of Gas Pressure in Vacuum Interrupter Based on Partial Discharge 596
of PD light pulses strongly depended on the phase angle of the
applied voltage. These results indicate that one can estimate gas pressure based on the phase angle dependence of PD light
pulse characteristics.
Figure 6 depicts the summarized plot of the PD pulse rise
time as a function of the internal pressure of VI at 2, 5 and 7 kVrms. At the bottom side of the graph are shown typical waveforms of single PD light pulses triggered at the phase where the maximum number of pulses appeared. It is found from the results that two distinct types of PD pulses appear with their own characteristic f eatures. The characterization of
PD light pulses was done on the basis of the rise time and peak intensity [13, 14]. The rise time t
r measured in this
experiment corresponds to 10% to 90% of the peak amplitude
of the light intensity. The results reveal that the PD pulses observed at 13 Pa to 130 Pa of air have a relatively larger rise time (1 μs order) and smaller magnitude (less than 0.5 mA),
while at pressures of 660 Pa and above they exhibit a shorter rise time (10 ns order) and larger magnitude (greater than 1 mA). The pulses with the larger rise time and smaller peak value can be attributed to a Townsend-like discharge while those with the sharper rise tim e and larger peak value are
characterized as streamer-like discharge. As depicted in
Figure 4a, the luminous area of PD in 130 Pa seems to scatter and spread over the glass wall of VI. This may be the consequence of many secondary avalanches in the Townsend-
like discharge. Note that the ga p between the electrode and the
shield is 11 mm under the pres ent experimental conditions.
Von Engel [12] showed that the logarithm of the reduced current in air at constant re duced field (E/p) is a linear
function of the distance up to 6 mm and then rises sharply due to the Townsend secondary effect. From this point of view, since the gap between the elect rode and the shields is 11 mm
under the present experimental conditions, the luminous PD is
considered to arise from the secondary avalanches of the Townsend-like discharge which makes the PD spread out over the glass surface of VI.
On the other hand, as shown in Figures 4b, 4c and 4d, the
luminous area of PD seems to be initiating from the edge of the bellows toward the edge of the shield, indicating a streamer-like discharge. The streamer mechanism of the
inhomogeneous field in air occurs if the number of ions in the avalanche reaches 10
8 which can be expressed as follows [15].
8
010 exp ≥⎟⎠⎞⎜⎝⎛∫ddxα (1)
Note that α is the first ionization coefficient. The distance
between the edge of the bellows and the edge of the shield in VI is 25 mm. Thus, the streamer-like discharge can be explained by the following consideration; from the left side of equation (1), if α is about 8 ion pairs/cm [12] under a constant
reduce field E/p of 1.32 V/cm-Pa, the number of the ions existing in avalanches will reach 4.8 x 10
8 which agrees with
the right side of equation (1).
Moreover, Ramachandra et al [ 14] has interpreted that the
Townsend-like discharge has a smaller magnitude (1 mA order) with a larger rise time (10-100 ns order) while the streamer-like discharge has a la rger peak value (10mA order)
with a shorter rise time (1 ns order). Their experiments were
carried out at an inception voltage and room temperature for various artificial voids of three thin polypropylene films and pressures of 100, 200 and 400 torr order. Although the test sample and the value of the rise time were different, the shapes of the discharge pulses obtained from their experiments are similar to those obtained in Figure 6.
Furthermore, F. H. Kreuger [15] reported that the rise time of
the Townsend-like discharge in a virgin cavity is far slower than several tens of nanoseconds compared with the streamer-like discharge, which has a duration of one to three nanoseconds. They also noted that as a consequence of many secondary avalanches in the Townsend-like discharge, the discharge spread out over a broa d surface area of the cavity.

3.3φ-n PATTERNS CORRESPONDING TO THE GAS
PRESSURE AND THE IDENTIFICATION BY NEURAL
NETWORK APPROACH
Figure 7a shows the results represented by the φ-i-n patterns
during 300 cycles of the applied voltage taken for the VI filled
with air at 1.3 kPa and 7 kVrms using PMT. The φ-i-n 10-910-810-710-610-510-4Rise time, tr (s)
1012 468
1022 468
1032 468
104
Pressure (Pa)Va= 2 kVrms
Va= 5 kVrms
Va= 7 kVrmsTownsend-like
discharge
Streamer-like
discharge

(a) Rise time of PD light pulses as a function of internal pressure of VI.
-4-2024Light intensity (a.u)
Time (1μs/div)tr= 40ns
2.6kPa
-0.10-0.050.000.050.10Light intensty (a.u)
Time (1μs/div)tr = 2.7μs
133Pa

(b) PD light waveform at 133Pa and 2.6kPa

Figure 6. Rise time of PD light pulse as a function of internal pressure of
VI measured with PMT and typical waveforms.

IEEE Transactions on Dielectri cs and Electrical Insulation Vol. 14, No. 3; June 2007 597
pattern of PD occurrence in the low vacuum region of VI was
measured using a PD measurement device. It is noted that φ
is for the phase angle (horizontal axis), i is for the PD light
pulse intensity (vertical axis) and n is for the number of PD
pulses distributed during 300 cycles of the applied voltage
(their distribution is marked by circles). The light intensity of
the PD was neglected. Only the φ and n patterns were taken into consideration since the intensity of the PD light pulses
varies with the distance of the PMT from the specimen (VI).
Therefore, considering only the φ-n pattern from Figure 7a,
the example of n against the developed grids was plotted in
Figure 7b. The 36 grids were de veloped vertically where each
grid represents 10 degrees of the phase angle. Using the data
of the φ-n patterns, the neural network (NN) approach was
employed to recognize the gas pressure in VI. The neural
network assistant software ( NNAS) with the back propagation
(BP) algorithm technique of NN was adopted in this analysis. Figure 8 shows the structure of the BPNN. The normalized number of pulses N
r in each grid becomes a set of the input of
BPNN. The N r was determined by the ratio n p to n tp which can
be expressed as follows
Nr = n p / n tp (2)
Note that n
p is the number of discharges occurring in each grid
region and n tp is the total number of PD at all the phase angles
(36 grids). The normalized number of pulses verses the 0.50
0.25
0.00
3633302724211815129630
Number of grids, ng0.50
0.25
0.000.50
0.25
0.00Ratio of pulses number, Nr
0.50
0.25
0.000.50
0.25
0.002.6 kPa
1.3 kPa
660 Pa
130 Pa
66 Pa
Figure 9. The ratio of pulse number verses number of grids for
five different pressures with 100 samples each.

1
2
i
I1
2
i
I1
j
J1
j
J1
3
5Input layer Hidden layer Output layer
Consideration of
input neuronPressures
Nrin each grid
(ng1…ng36)
*100 samples for
learning algorithm
*100 samples for
testing algorithm
Variable item:
Number of iterations : 200
Hidden layer : 3, 5 &10 2
466 Pa
130 Pa
660 Pa
1.3 kPa
2.6 kPa
Figure 8. Structure of back propagation neural network. 0.07
0.06
0.05
0.04
0.03
0.02
0.01
0.00Mean squared error
200 180 160 140 120 10080 60 40 20
Number of iterations 3 Neurons in HL
5 Neurons in HL
10 Neurons in HL

Figure 10. Recognition error of the BP NN in discrimination of the gas
pressure for 200 iterations. PD light intensity, i (a.u)
360 300 240 180 120 60 0
Phase (degree)
(a)Typical ø-i-n pattern
250
200
150
100
50
0Number of PD pulses, n
3633302724211815129630
Grids region
(b) Development of 36 grids

Figure 7. (a) Typical ø-i-n pattern in air at 1.3kPa and 7kVrms
measured with PMT. (b) Development of 36 grids.

M. Kamarol et al.: Determination of Gas Pressure in Vacuum Interrupter Based on Partial Discharge 598
number of grids for five different pressures with 100 samples
each is plotted in Figure 9. The N r with 100 samples was
assigned for a learning and testing algorithm for different pressure levels. The outputs of the BP network were 66 Pa, 130 Pa, 660 Pa, 1.3 kPa and 2.6 kPa. The number of neurons
in the hidden layer was varied to obtain the optimum
successful rate of the NN.
Figure 10 demonstrates the recognition error of the BP NN in
discrimination of the gas pressure for 200 iterations of 3, 5 and 10 neurons in the hidden layer. The result shows that after 120 iterations, the recognition erro r of the NN is perceived to
be negligibly small for 10 neurons of the hidden layer compared to 170 iterations for 3 neurons of hidden layer.
Table 1 summarizes the recognition success rate of the BP
NN in the discrimination of gas pressure at 100 and 200 iterations for 3 and 10 neurons of the hidden layers. From the experiment results, 10 neurons of the hidden layer have a higher successful rate compared to that for 3 neurons of the hidden layer. Note that regarding the number of neurons in the hidden layer, more than 10 does not give much difference from that of 10 neurons. The average of the successful
recognition rate of the vacuum degree from 93 to 98 % is obtained, indicating a cognitive ability of vacuum degree estimation based on PD pattern recognition using the BP NN approach.
4 CONCLUSIONS
The determination of gas pre ssure based on partial discharge
was investigated. From these experimental results, the following conclusions were drawn: 1) The PD pulses in VI filled with air for a pressure below
260 Pa having a relatively larger rise time (2 µs) with a smaller magnitude (less than 0.5 mA) were attributed to a Townsend-like discharge, while those with a sharper rise time (10-100 ns order) and larger magnitude (greater than 1 mA) were characterized as a streamer-like discharge at pressures above 260 Pa.
2) The estimation of gas pressure based on partial discharge
phase-angle characteristics using the NN approach gave a recognition success rate from 93 to 98 %. Thus the neural network algorithm is applicable to the discrimination of
gas pressure in VI using
φ-n patterns. In practical situations, it is expected that on ce the leakage of
gas into the VI occurs, the gas pressure of VI will increase immediately so as to pass the pressure giving the minimum of PD inception voltage; i.e. reachi ng the 1 Pa-10 Pa order. In
this sense, investigation of PD properties over 1 Pa is also meaningful and important to diagnose the pressure level of VI.
REFERENCES
[1] H. Okubo and S. Yanabu, “Feasibility study on application of high
voltage and high power vacuum circuit breaker”, 20th Intern. Symposium
on Discharge and Electrical Insulation in Vacuum, Tours, France, pp.275-278, 2002.
[2] T. Kawamura, M. Meguro, H. Hama and T. Yamagiwa, “Industrial
Outlook: How to reduce SF
6 use and emission – Various aggressive
approaches to realize less SF 6 environment”, 10th Int. Symp. on Gaseous
Dielectrics, chapter 74, 2004.
[3] S. Giere, M. Kurrat and U. Schumann, “HV dielectric strength of
shielding electrodes in vacuum circuit breakers”, 20th Int. Symp. On
Discharge and Electrical Insulation in Vacuum, Tours, France, 2002.
[4] H. Saitoh, H. Ichikawa, A. Nishijima, Y. Matsui, M. Sakaki, M. Honma
and H. Okubo, “Research and deve lopment on 145 kV/40 kA one break
vacuum circuit breaker”, IEEE/PES T&D Asia Pasific, PN-13, 2002.
[5] N. H. Malik, A. A. Al-Arainy and M. I. Qureshi, Electrical Insulation in
Power Systems , Marcel Dekker, Inc. pp. 204, 1998.
[6] W. W. Watrous jr., “Method and apparatus for measuring in vacuum
circuit interrupter”, US-Patent No. 3575656, 1971.
[7] Z. Ziyu, J. Xiuchen, J. Zhijian, Z. Ji yan, and S. Jing, “Study on internal
pressure measurement of vacuum interrupter”, 19th Intern. Symposium
on Discharge and Electrical Insulati on in Vacuum Xi’an, pp. 775-778,
2000.
[8] G. Yonggang, X. Guozheng, H. Yu long and L. Weidong, “On-Line
monitoring the vacuum degree of vacuum interrupter by partial discharge”, 12
th Asian Conference on Electrical Discharge, Shenzhen,
China, pp.195-197, 2004.
[9] M. Kamarol, S. Ohtsuka, H. Saitou, M. Sakaki and M. Hikita,
“Discharge properties and emitted el ectromagnetic wave spectrum in
low vacuum region of vacuum interrupter”, IEEJ Trans. Power Energy, Vol. 125, pp.797-802, 2005.
[10] M. Kamarol, S. Ohtsuka, H. Saitou, M. Sakaki and M. Hikita,
“Discharge phenomena in low vacuum region of glass tube vacuum
interrupter under AC applied voltage”, Intern.l Symposium on Electrical Insulating Materials (ISEIM), pp.36-39, 2005.
[11] M. Kamarol, S. Ohtsuka, H. Saitou, M. Sakaki and M. Hikita,
“Investigation of discharge character istic for diagnosis of degree of
vacuum in glass type vacuum interrupter”, 14
th Intern. Symposium on
High Voltage Engineering (ISH), Beijing, pp.171, 2005.
[12] A. Von Engel, Ionized gases , Oxford University Press, Great Britain,
1965.
[13] D. G. Kasten, X. Liu, S. A. Sebo, D. F. Grosjean and D. L. Schweickart,
“Partial discharge measurement in air and argon at low pressures with and without a dielectric barrier”, IEEE Trans. Dielectr. Electr. Insul., Vol. 12, pp. 362-373, 2005.
[14] B. Ramachandra and R. S. Nema, “D ischarge mechanism in voids at
sub-atmospheric pressures”, IEEE Intern. Sympos. Electr. Insul., USA, pp.165-168, 1998.
[15] F. H. Kreuger, Industrial High Voltage , Delft University Press,
Stevinweg, Netherlands, Vol. 1, pp. 117-133, 1995.

Mohamad Kamarol was born in 1971. He received
the B.Eng. (Hons) degree from the University Technology Mara, Malaysia in 2000. He joined Malaysia Science University (USM) with a University ASTS Fellowship in 2002. He received the M.S degree from Kyushu Institute of Technology in 2005. He is now working on his Ph.D. degree in electrical engineering at Kyushu Institute of Technology, Japan.

Table 1. Summarization of average successful rate for 3 and 10
neurons of the hidden layer (HL) for 100 and 200 iterations.

3 neurons of HL 10 neurons of HL Pressures
(Pa) 100
iterations 200
iterations 100
iterations 200
iterations
66 90.4 % 93.1 % 90.6 % 92.6 %

130 92.7 % 94.9 % 92.8 % 94.7 %

660 90.7 % 94.2 % 96.5 % 97.7 %

1300 95.0 % 96.6 % 96.7 % 97.7 %

2600 94.5 % 96.6 % 97.0 % 97.9 %

IEEE Transactions on Dielectri cs and Electrical Insulation Vol. 14, No. 3; June 2007 599
Shinya Ohtsuka was born in 1971. He received the
B.S. and M.S. degrees from Kyushu University, Japan,
in 1994 and 1996, respectively. He finished his Doctor Course of the Dept. of Electrical and Electronic Systems Engineering at Kyushu University in 1998. He was a research fellow (DC1) of the Japan Society for the Promotion of Science (JSPS) from 1996 to 1998, then became a JSPS Research Fellow (PD) in 1999. Since 1999, he has been an Assistant at Kyushu Institute of Technology, Japan and promoted
to Associate Professor at the same uni versity in 2006. His research interests
include the insulation properties of environmentally-benign gas as an SF
6
substitute, PD detection technique for in sulation diagnosis of power apparatus
and superconductivity engineering. He ha s a Doctors Degree of Engineering.
Dr. Ohtsuka is a member of the IEEJ, the IEEDJ and the Cryogenic Society of
Japan.
Hitoshi Saitoh was born in 1970. He received the
B.S. degree in nuclear engineering from Hokkaido
University in 1995 and the Ph.D. degree in electrical engineering from Nagoya University in 2003. He worked for 6 years at Meid ensha Corporate Ltd. In
2002, he joined the Japan AE Power Systems Corporation, and is currently engaged in R&D on switchgears. He is a member of IEE of Japan.
Masayuki Sakaki was born in 1956. He received the
B.S. degree in electrical engineering from Tokyo Science University 1979. He worked for 22 years in Switchgear Factory at Meidensha Corporate Ltd. In 2002, he joined Japan AE Power Systems Corporation, and is currently a Senior Manager of the Switchgear Development Department. He is engaged in R&D on switchgears. He is a member of IEE of Japan.

Masayuki Hikita (M’96-SM’00) was born in 1953.
He received the B.S., M.S. and Dr. degrees in electrical engineering from Nagoya University, Japan, in 1977, 1979, and 1982, respectively. He was an Assistant, a Lecturer, and an Associate Professor at Nagoya University in 1982, 1989, and 1992, respectively. Since 1996, he has been a professor of the Department of Electrical Engineering, Kyushu Institute of Technology. He was a Visiting Scientist at the High V oltage Laboratory at MIT in the USA from
1985 to 1987. Dr. Hikita has recently been interested in research on
development of diagnostic techniques for electric power equipment. He is a
member of the Japan Society of Applied Physics and the IEEJ.

View publication statsView publication stats