TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 12, No. 12, Decembe
r
2014, pp. 80
3
3
~ 804
1
DOI: 10.115
9
1
/telkomni
ka.
v
12i12.67
71
8033
Re
cei
v
ed
Jul
y
2, 2014; Re
vised Septem
ber
29, 20
14;
Accept
ed No
vem
ber 3, 20
14
Prediction the AC Breakdown Voltage in Point/Plane Air
Gaps with Barrier Using Design of Experiments
Abdelgh
a
ni Rouini*
1
, Djillali Mahi
2
, Tahar Seghier
2
Dep
a
rtment of Electrical E
ngi
neer
ing,
F
a
cult
y of Scie
nce a
nd T
e
chnol
og
y,
Univers
i
t
y
Am
a
r
T
e
lidji of L
a
g
hou
at, BP 37G, Lagho
uat 03
0
00, Alger
ia
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: a.rouin
i
@u
ni
v-djelf
a
.dz
1
, d.mahi@m
ail.l
a
g
h
-un
i
v.dz
2
,
t.seghier
@mai
l
.
lagh-
univ.dz
3
A
b
st
r
a
ct
Breakd
o
w
n
vo
l
t
age c
har
acteri
stics of a
i
r g
a
p
s
cons
ider
ed
a
rand
o
m
pr
oces
s w
e
hav
e
inve
stigat
e
d
how
to fin
d
re
l
e
vant statistic
a
l differe
nt l
ead
i
ng to
bre
a
kdo
w
n or w
i
th sta
ndi
ng
duri
ng
a
n
i
m
p
u
ls
e volt
age
test und
er th
e
same c
ond
itio
n
for a
pl
an
e
hig
h
vo
ltage
e
l
ectr
ode
a
nd a grou
nde
d el
ectro
d
e
confi
gurati
on.I
n
the prese
n
t pa
per metho
d
s of the mo
d
e
l
lin
g and pr
edicti
on
of the AC brea
kdow
n voltag
e in poi
nt-pla
ne
ai
r
gaps
are
inv
e
stigated. A
n
a
nalysis
b
a
sed
on
exp
e
ri
me
nts
des
ign
me
thod has be
e
n
d
e
vel
o
p
ed w
i
th
indic
a
tes th
at. Measur
e
m
ents
to conta
i
n s
o
me
rel
e
vant
in
formati
on test
at early
stages
in re
duc
ed ti
me
frame i
n
vesti
g
ation w
e
re do
n
e
on a po
int-pl
ane w
i
th ba
rri
e
r
varying in p
o
s
ition a
nd si
z
e
.
T
he barrier us
ed
w
a
s mou
n
ted
vertically b
e
tw
een the e
l
ectro
des.T
he us
e of
metho
d
o
l
ogy
of exper
i
m
ents
desig
n is one
of
this meth
ods
and pr
esent a
n
origi
n
a
l
ide
a
in high vo
lta
ge pre
d
ictio
n
prob
le
ms seve
ral factors w
e
re
consi
dere
d
,na
m
e
l
y
the
dist
a
n
ce betw
een electro
des a
n
d
differe
nt par
a
m
eters
of th
e
barrier
suc
h
u
s
is
di
me
nsio
n an
d
his hol
es.T
he exper
imenta
l
results are co
mpare
d
w
i
th results from
numerical sim
u
lations.
We
fi
rstl
y p
r
e
s
e
n
t
th
e p
r
in
cipa
l
s
o
f
th
i
s
mod
e
l
th
e
n
we
ap
p
l
y i
t
to
the
stu
d
y
o
f
b
a
rrier e
ffe
ct.A g
ood
agre
e
m
ent h
a
s
been fo
und b
e
t
w
een the computed a
nd ex
p
e
ri
ment
al resu
l
t
s
Ke
y
w
ords
: ex
peri
m
e
n
ts desi
gn metho
d
, bre
a
kdow
n volt
ag
e, barrier, po
int
-
pla
ne ga
p, factors
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
The kn
owl
e
d
ge of the con
d
ition of ionisati
on and p
r
o
pagatio
n of electri
c
al di
sch
a
rge i
s
of great inte
rest to
well
un
derstand
the
méca
ni
sme
s
leadin
g
to breakdo
wn
[1]. The barrie
r
s are
widely used i
n
many high voltage devices.It is well k
nown that the dielectri
c
strength of long air
gap is
signifi
cantly incre
a
sed by
the insertion of an i
n
sul
a
ting ba
rr
ier. The in
sul
a
ting stru
ctu
r
e to
different stress and pa
rticu
l
arly to the discharge ph
en
omena [2, 3].
The effe
ctiveness of the
barrier
dep
e
nds
on the
geomet
ry (th
e
dimen
s
io
n
and the
positio
n of th
e ba
rrie
r
, an
d
the physi
cal
nature
on
th
e
barrier [4, 5]. The inve
stig
ation have
be
en
done
expe
ri
mentally an
d
simul
a
ted i
n
order to
study the b
r
e
a
kd
own ph
e
nomen
a of
solid
d
i
e
l
ec
tr
ic ba
rr
ie
r
.
T
h
e pr
es
e
n
c
e
a h
o
l
e
w
i
th
in
th
e barrier signifi
cantly red
u
ces
th
e diele
c
tric
stren
g
th of th
e sy
stem. Te
sts
are
cond
ucted to
mea
s
ure the
50
Hz AC brea
kd
own volta
ge
of
small
air
gap
[6-8]. The
re
sults
of the t
e
st
seri
es
sh
ow h
o
w the
brea
kd
own v
o
ltage va
rie
s
with
the dista
n
ce
betwe
en the
electrode
s, th
e si
ze
and th
e materi
al of
barrier, th
e re
lative positio
n
of
the electrode
s and the b
a
rrier b
e
twe
en them [9].
As a result, the brea
kd
o
w
n voltage varie
s
due to
the size of the barri
ers and the
maximum fla
s
hove
r
voltag
es
are
o
b
se
rv
ed
whe
n
th
e
barriers are
p
o
sition
ed
at the n
earest
p
o
int
to the ele
c
tro
de an
d the
small si
zed
ba
rrie
r
s
be
co
m
e
effective on
ly in very sm
all air g
a
p
s
[10-
16]. An a
nal
ysis
ba
sed
o
n
expe
rime
ntal de
sig
n
m
e
thod h
a
s b
e
e
n
devel
ope
d
whi
c
h i
ndi
cat
e
s
that mea
s
u
r
e
m
ents do
co
ntain
some
relevant info
rmation te
st a
t
early
stag
e
s
in
redu
ce
d
time
frame.
The p
r
inci
ple
of this metho
dology is to
carry o
u
t a schedul
e of exp
e
rime
nts de
si
gned to
obtain the
most a
c
curat
e
inform
ation
for a
sp
eci
f
ic problem
with a mi
ni
mum nu
mbe
r
of
experim
ents [17, 18]. Its a
d
vantage
s
were
proved i
n
differe
nt are
a
s
of
ap
plication, e
s
pe
ciall
y
in
chemi
s
try a
n
d
me
cha
n
ics, whe
r
e a
l
a
rge
num
be
r of paramet
ers have to
be o
p
timized
simultan
eou
sl
y. This meth
od ha
s
been
use
d
in
ele
c
troni
cs, ci
rcuits an
d d
e
vice
s,co
mpon
e
n
ts
and pa
ckagi
n
g
, for the de
sign o
p
timiza
tion in som
e
indu
strial ele
c
troni
cs appli
c
ation
s
an
d for
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 12, Decem
ber 20
14 : 8033 – 80
41
8034
powerful non
linear
control of
po
wer con
v
erters
[19
].
A de
sign
spa
c
e, o
r
regio
n
of interest, m
u
st
be defined, that is, a ran
ge of variabil
i
ty mu
st be set for ea
ch
variable.Th
e
numbe
r of the
variable
s
val
ues can a
s
sume in desi
g
n method is
rest
ricte
d
an
d gene
rally is small [20]. The
desi
gn
metho
d
techniq
ue
a
nd the
nu
mb
er
of level
s
a
r
e to
be
sele
cted
acco
rdi
n
g to th
e n
u
m
ber
of experime
n
ts whi
c
h
can b
e
afforded [2
1, 22].
In this pape
r work, in ord
e
r
to modelling
the AC brea
kdo
w
n voltag
e in point-pla
ne gap
s
arrang
ement
in p
r
e
s
en
ce
of ba
rrie
r
, th
e
experi
m
en
ts de
sig
n
m
e
thod i
s
used.
The
carried
ou
t
experim
ental
re
sults are
take
n to b
u
ild a
model
whi
c
h ta
ke
s into
co
nsi
deratio
n diffe
rent
para
m
eters
such a
s
(the rel
a
tive position
of the barri
er, its hole and
the width of t
he ba
rrie
r) th
at
affect the bre
a
kd
own phe
nomen
a.
2.
Material an
d Metho
d
s
2.1. Material
The expe
rim
ental set
-
up
consi
s
ts of a h
i
gh-
voltag
e te
st tran
sform
e
r 100
kV/5kV
A/ 50Hz,
a capa
citive
voltage divid
e
r. Fig
u
re 1
a
nd Fi
gure
2
(t
he exp
e
ri
ences
h
a
ve
be
en
pe
rform
e
d i
n
the labo
ratory of high voltage Universit
y
of Bi
skra
)sho
ws the arrang
eme
n
t of electrode
s a
nd
insul
a
ting ba
rrier it co
ntain
s
a point–
p
l
an elect
r
od
e
arran
gem
en
t mounted vertical. Th
e HV
electrode
s
co
nsi
s
t a ste
e
l
needl
e point
on copp
er of
coni
cal i
n
sh
ape 3
0
°. Th
e gro
und
ed p
l
an
electrode i
s
a
circula
r
steel
plate of 30 cm long, 2.8 cm diameter.
The
ba
kelitte barriers (
ε
5
.
8
2
are
squ
a
re
s of d
i
fferent width
s
(15
cm, 20
cm an
d 2
5
cm). a
nd diff
erent
hole
s
(5mm, 10mm
and 1
5
mm
)
and its thicknesse
s i
s
1m
m,an alumi
n
um
plan groun
de
d. To cha
nge
the positio
ns
for se
ve
ral ba
rrie
r
s,
carrie
rs ba
kelite are
used.
Figure 1. View of real te
st cell
Figure 2. Sch
e
me of the test cell
The
barrie
r
i
s
mou
n
ted ve
rtically bet
wee
n
the
ele
c
tro
d
es It
s
surf
ace
s
a
r
e
che
c
ke
d
after
each brea
kd
own.
2.2. Method
s
The p
r
in
ciple
of the techni
cal de
sign
of e
x
per
ime
n
ts
consi
s
ts in va
rying levels of
one
or
more factors simult
aneous (whi
ch are variabl
e, discrete or
continuous) in ea
ch test. This
will
help to red
u
ce signifi
cantl
y
the numbe
r of ex
perim
ents to be p
e
rform
ed.
Whi
l
e increa
sing
the
numbe
r of studied facto
r
s, detecting int
e
ra
ction
s
bet
wee
n
the factors a
nd opti
m
al comp
are
d
to
are spon
se, that is to say aqua
ntity use
d
as
sta
nda
rd
and allo
wing
to easily mod
e
ling re
sult
s.
The tradition
al metho
dolo
g
y for a
n
exp
e
rime
nt
al
de
sign con
s
ist
s
of
four (4
) steps:
first
the p
r
epa
rati
on
study in
cl
uding
the
def
inition of
re
sp
onses characterizin
g
the
o
b
jective
s
, whi
c
h
is the
main
scop
e of thi
s
p
aper an
d the
determi
nat
ion
of facto
r
leve
ls, then th
e
choice of fa
cto
r
s
and exp
e
rim
ental dom
ain
in the
seco
nd ste
p
,
the
pro
posed
model it
self
and finally,
the
mathemati
c
al
model.
2.2.1. Definition of Res
p
o
n
ses Ch
ara
c
teri
zing the
Objectiv
es
We wa
nt
to measure
the
influen
ce of
t
he
follo
w
i
ng
fa
c
t
o
r
s Wh
e
r
e:R
(
c
m
) is th
e r
e
la
tive
positio
n of the barrier, H(mm) is the h
o
l
e in the
barri
er and
W (cm
)
is the width
of the barri
er.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Predi
ction the
AC Brea
kdo
w
n Voltage in
Point/Plane Air Gap
s
with
… (Abdel
gha
ni Roui
ni)
8035
2.2.2.
Choice
of Factors a
nd Expérime
ntal Domain
Determinin
g the field of st
udy is clo
s
el
y
related to the initial kn
o
w
led
ge hel
d on the
physi
cal
phe
n
o
meno
n u
n
d
e
r
study, b
u
t
also
to the
o
b
j
ectives of th
e expe
rime
nt. In a
ddition,care
must be ta
ke
n to minimize
the cost of
the study, exp
r
esse
d as n
u
mber of test
s.
We
recall
now that th
e
results of the
study will
be vali
d
only
on the range of vari
ation of
the factors co
nsid
ere
d
. The
main factors
co
n
s
id
ere
d
in
this plan exp
e
rime
nts are:
Table 1. Leve
l
s of factors studied
Factors
R (cm)
H (mm)
W (cm)
Level – 1
2
5
15
Level 0
3
10
20
Level +1
4
15
25
2.2.3.
A Proposed Mod
e
l
Our
choi
ce fe
ll on co
mpo
s
i
t
e face-ce
n
te
red pl
an
s for
the study of resp
on
se
surf
ace
s
. A
face-ce
n
tered
comp
osite
d
e
sig
n
is
defin
ed by: two
start point
s by
para
m
eter
an
d po
sitioned
on
each of th
e
axes.
The
s
e poi
nts
co
n
t
ribute to
th
e evalu
a
tion.
The
qu
adra
t
ic term
s of
th
e
polynomial
model, i.e., they give information ab
ou
t the curvatu
r
e of the su
rfa
c
e of respon
se:
The total number of tests to
be condu
cted,
N
depen
ds on the number of factors k studi
ed and
the numbe
r o
f
répétitions i
n
the cente
r
o
f
the domain,
n
.
N2
2
.
K
n
(1)
n
3
With
N
2
2
.
3
3
1
7
(2)
Table 2. Co
m
posite pl
an fo
r the
study ce
ntered o
n
three Fact
ors
Experiments
No
Fac
t
or
U
(k
V)
R (cm)
H (mm)
W(cm)
1 -1
-1
-1
65,84575
2 -1
1
-1
62,73575
3 -1
-1
1
62,204331
4 -1
1
1
55,459331
5 1
-1
-1
66,70175
6 1
1
-1
67,38675
7 1
-1
1
67,525331
8 1
1
1
64,575331
9 0
-1
0
66,602838
10 0
1
0
63,572838
11 0
0
1
65,797838
12 0
0
1
65,797838
13 -1
0
0
64,114838
14 1
0
0
69,100838
15 0
0
0
67,526216
16 0
0
0
67,506216
17 0
0
0
67,486216
The la
st thre
e ro
ws
of Ta
ble 2
corre
s
p
ond
s
to a te
st ce
nter
con
s
ide
r
ed
expé
rimental
field, which should b
e
rep
e
a
ted n
0
times to
ensu
r
e certai
n prop
ertie
s
the matrix exp
e
rime
nts.
So that it m
eets th
e
req
u
irem
ent of
unifo
rm
preci
s
ion, en
suri
n
g
a nea
rly consta
nt
varian
ce withi
n
the experim
ental ran
ge.
Whether the vector of model coefficient
s analytical sought:It is defined
by The
coeffici
ents v
e
ctor of the a
nalytical mod
e
l is define
d
as follo
w:
b
X
.X
X
.y
(3)
Whe
r
e
X
are t
he matrix ex
perim
ent,
X
are
the tran
spo
s
e matrix expe
riment,
y
the breakdo
wn
voltage (the resp
on
se).T
h
e
numbe
r of unkn
o
wn para
m
eters (
b
of the polynomial
is determi
ne
d
from the follo
wing formula:
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 12, Decem
ber 20
14 : 8033 – 80
41
8036
b
!
!
!
⇒b
!
!
!
1
0
(4)
Finally, the model is give
n by equation:
yb
∑
b
.X
∑
b
.X
∑
∑
b
.X
.X
(
5
)
yb
b
.
x
b
.
x
b
.x
b
.x
b
.x
b
.x
b
.x
b
.x
b
.x
(
6
)
2.2.4.
Mathe
m
atical Mod
e
ls
Estimation o
f
model
coe
fficients,we
have u
s
ed
Matlab p
r
og
ram,whi
c
h
gi
ves an
analytical fo
rm of the resp
onse stu
d
ied
surfa
c
e
an
d
they are
cal
c
ulated by u
s
i
ng Equ
a
tion (3)
(Tabl
e 3).
Table 3. Co
efficients of ma
thematical m
odel
Response
Breakdo
w
n
volta
ge
U
(k
V)
Experimental de
sign domain
Constant
67,5262162
R 2,493
H -1,515
W -1,613209
46
R.H
0,94875
R.W 1,11625
H.W -0,90875
R
2
-0,918378
38
H
2
-2,438378
38
W
2
-0,115168
92
So we can
write the mathe
m
atical mo
de
l (experim
ent
al domain
)
a
s
follows:
U
67,5262
162
2
,
49.
R
1,
515.
H
1
,
613209
46
.
W
0
,94875
.
R
.
H
1,1162
5
.
R.
W
0
,
90875
.
H
.
W
0
,91837
838
.
R
2,4383783
8
.
H
0,115168
92
.
W
(7)
The
obtaine
d
re
sult
s
can
be pl
otted to
co
mpa
r
e
th
e me
asure
d
re
sp
on
ses
with the
estimated o
n
e
. For this, it is necessa
ry to
plot the adequ
acy of
the model. Mea
s
u
r
ed
respon
se
s a
r
e placed on
the absci
ssa
and e
s
timate
d re
spo
n
ses
are on the
ordin
a
te (Fi
g
ure
3). The cl
oud
points is alig
ned with the
line y
=
x, which me
ans that accuracy of the
model is p
r
et
ty good.
Figure 3. Gra
ph model a
d
e
qua
cy
The de
scripti
v
e quality of the model is
illustrated
here but a seco
nd an
a
l
ysis of
variance will
possibl
e to verify this concl
u
sion.
54
56
58
60
62
64
66
68
70
54
56
5
8
60
62
64
6
6
6
8
70
v
ex
p
v
m
odè
l
e
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Predi
ction the
AC Brea
kdo
w
n Voltage in
Point/Plane Air Gap
s
with
… (Abdel
gha
ni Roui
ni)
8037
2.2.5. Statisti
cal Analy
s
is
of the Mo
del
The equ
ation
of the empirical m
odel i
s
onl
y a ap
proximatio
n of reality. Estimating
coeffici
ents
o
f
of the seco
nd deg
re
e po
lynomial mod
e
l is ba
se
d o
n
test re
sult
s that are, ea
ch
treatment
pla
n
expe
rien
ce,
the p
a
rticula
r
value
s
of a
ran
dom va
ri
able. Th
e im
plementatio
n
of
statistical tests must allo
wto pass ju
dgment
on the re
sults o
b
tained in n
a
mely a model
descri
b
ing th
e variation
of the re
spo
n
se
in the expe
ri
mental
field. This step
Stat
istical analy
s
is
of
the re
sult
s in
the con
s
tru
c
tion of table
of reg
r
e
ssi
on
analysi
s
and
determi
ning
the de
scriptiv
e
quality of the
model.
Reg
r
e
ssi
on a
nalysi
s
i
s
to ex
pl
ain
the total
cha
nge in
the
re
spon
se from t
he
defined
sum
of squa
red d
e
v
iations bet
ween
the re
sult
s test and the
i
r avera
ge.
Statistical an
alysis of the
model a
s
a
whole is
foll
owe
d
by the con
s
truction
of a statistical
test whi
c
h co
nsi
s
ts in affirming that the model do
es n
o
t allow de
scribe the
α
equal to 5%.
The
reg
r
e
ssi
on an
alysi
s
t
able i
s
u
s
e
d
to ac
hieve im
mediately
cal
c
ulate
the
co
efficient
determi
nation
R
,
R
and
Q
: descri
p
tive quality of the model
will be evaluated, these
take
values follo
wi
ng:
Table 4. Co
efficients of det
ermin
a
tion of
the descri
p
tive quality of the model
R
R
Q
0,989384
0,994678
0,975735
A good descriptive qu
ality is assured be
ca
use the Coefficient
s
R
,
R
and
Q
0
.
9
(nea
r the unity), which
reflect
s
t
he
good q
uality of the model.
F
or a ri
sk of
α
5
%
,
N
P
17
1
0
7
From the Stu
dent table:
t
crit
=
(0,05,7)=
2
.37
(8)
The effect
will be significant at
the risk of
5% if
(
t
2
.
3
7
.
In addition,
t
is
cal
c
ulate
d
by usin
g Equatio
n (8) a
nd (9
)
|
|
(
9
)
Table 5. Te
st of significan
c
e of the coefficient
s
Factors
Effet
t
2
.
3
7
r
e
sults
Constante
67,5262162
560,47
significat
i
f
R 2,493
20,69
significat
i
f
H -1,515
12,57
significat
i
f
W -1,613209
46
13,39
significat
i
f
R.H
0,94875
7,87
significat
i
f
R.W 1,11625
9,26
significat
i
f
H.W -0,90875
7,54
significat
i
f
R
2
-0,918378
38
7,62
significat
i
f
H
2
-2,438378
38
20,24
significat
i
f
W
2
-0,115168
92
0,96
non
significatif
From
T
able
5,
only
the
coeffici
ents p
r
ovidin
g
d
e
scriptive
qu
ality of mod
e
l
will
be
pres
erved.That is
to rejec
t
the c
oeffic
i
ents
(
W
) The re
d
u
ce
d model e
quation b
e
co
mes:
U
6
7
,
52621
62
2
,49
.
R
1,515
.
H
1
,
6132094
6
.
W
0,9487
5
.
R.
H
1,11625
.
R
.
W
0
,9087
5
.
H.
W
0
,
918378
38
.
R
2,43837
838
.
H
(10
)
In this ca
se the co
nfiden
ce interval of an effect is given by:
b
1
,96
.
0,
189
lower limit,
b
1
,96
.
0,189
Higher limit.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 12, Decem
ber 20
14 : 8033 – 80
41
8038
3.
Resul
t
s and
Anal
y
s
is
The valid
atio
n of the
re
sul
t
s given
by t
he mo
del i
s
t
o
che
c
k whet
her th
e a
s
su
mptions
made in de
pa
rture of expe
riments a
r
e well verified.
Validation
can b
e
carrie
d pe
rformi
ng
test comple
mentary o
u
tside the te
sti
ng pla
n
experim
ents t
o
validate the
model beh
avio
r obtain
ed b
y
the experim
ental de
sign.
In our
ca
se
study, we to
ok the
mad
e
test to
stu
d
y the influen
ce of pa
ramet
e
rs Apa
r
t
from testin
g t
he expe
rime
n
t
al desi
gn. th
e re
sult
s
of th
ese te
sts are
comp
ared
with re
sult
s of the
mathemati
c
al
model .
3.1. Influenc
e of the
Rela
tiv
e
Position Barrier on the Bre
akdo
w
n
Voltage
Figure 4
sh
o
w
s the exp
e
ri
mental a
nd
p
r
edi
cted
bre
a
k
do
wn
voltag
e a
s
a fu
nctio
n
of the
relative p
o
siti
on of th
e b
a
rrier valu
es of
the
b
a
rrier
fo
r
different wid
t
h
(1
5cm, 20 cm and
25
cm
)
of the barri
er. The hole in
the barri
er i
s
(5 mm
).Different di
stan
ces bet
wee
n
the point an
d the
barrier
(0 to 5
cm) were stu
d
ied.
The bre
a
kdo
w
n voltage o
f
the gap-poi
nt barrie
r
pla
n
significantl
y
improved whe
n
the
point tou
c
h
e
s the b
a
rrier, t
hen it
decre
a
s
e
s
. As th
e
we
clo
s
e
ba
rrier plan,th
e
optimum
on t
h
e
brea
kd
own voltage corre
s
pond
s to the ca
se wher
e the barrie
r
is contacte
d with
the point.
The predi
ctio
n values are
situated bet
wee
n
the two boun
dari
e
s of the area
of risk of
5%.This mea
n
s that the founde
d model
by t
he experi
m
ental de
sig
n
method is p
e
rfect.
Figure 4. Influence of the re
lative positio
n of
barri
er o
n
the brea
kd
o
w
n voltage: di
fferent width
(a)15
c
m, (b)2
0 cm and
(c)
25cm, the hol
e (5mm
)
3.2. Influenc
e of the Hole in the Bar
r
ier on the
Br
eakdo
w
n
Voltage
In this test,
hole
s
in the
middle of th
e ba
rrie
r
vari
es
(5mm,1
0
mm and
15
mm). The
relative po
sition of the barrier value
s
for
4 cm, differen
t
width (15
c
m
,
20 cm and 2
5
cm
).
We
see f
r
om
Ffigure
5 a
s
perfo
ration
o
f
the
barrie
r
affects con
s
i
dera
b
ly
well
i
n
sul
a
ted
the indee
d, when the b
a
rrier is in
cont
a
c
t with
the po
int, with cent
ered
hole di
a
m
eter (5mm).
In
this ca
se the discha
rge d
o
e
s not pa
ss t
h
rou
gh t
he h
o
le but take
s the path edg
e of the barri
er
–
plan.
Thus di
scharge do
es
not t
end to
go through th
e
sma
ll diamete
r
ho
le, pro
bably b
e
ca
use
of the con
c
e
n
tration loa
d
the middle of
the
barrie
r
, and then
see
deflected of
zone h
o
le.T
his
explain
s
the barrier
cont
ri
bution of the gap to
improve the rigidity diele
c
tric
syst
em.
This ca
n b
e
interp
rete
d
by the fa
ct
that
increa
si
ng the
diam
eter of
the
hole.The
predi
ction val
ues
are situ
ated between t
he tw
o bo
und
arie
s of the area of risk of 5
%
.
-1
-0
.
8
-0
.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
0.
8
1
64
.
5
65
65
.
5
66
66
.
5
67
67
.
5
68
R
e
l
a
t
i
v
e
po
s
i
t
i
on
of
t
h
e b
a
r
r
i
er
(a
)
Br
eak
dow
n v
o
l
t
age (
k
V
)
H
i
gh
er
l
i
m
i
t
Lower
l
i
m
i
t
M
e
as
ur
e
d
Pr
e
d
i
t
-1
-0
.
8
-0
.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
0.
8
1
63.
5
64
64.
5
65
65.
5
66
66.
5
67
67.
5
68
R
e
l
a
t
i
v
e
pos
i
t
i
o
n of
bar
r
i
er
(b
)
B
r
eak
do
w
n
v
o
l
t
ag
e
(
k
V
)
H
i
gher
l
i
m
i
t
L
o
w
e
r
lim
it
M
eas
ur
e
d
P
r
edi
t
-1
-0.
8
-0.
6
-0.
4
-0.
2
0
0.
2
0.
4
0.
6
0.
8
1
61
62
63
64
65
66
67
68
69
Rel
a
t
i
v
e
pos
i
t
i
on of
t
he barr
i
er
(c
)
Br
ea
k
d
o
w
n v
o
l
t
ag
e
(
k
V
)
H
i
gher l
i
m
i
t
Low
er
l
i
m
i
t
M
eas
ured
Pr
e
d
i
t
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Predi
ction the
AC Brea
kdo
w
n Voltage in
Point/Plane Air Gap
s
with
… (Abdel
gha
ni Roui
ni)
8039
Figure 5. Influence of the hole in the mid
d
le of
the barrier o
n
the breakdo
wn voltage: different
width (a
)1
5cm, (b)2
0 cm
and (c) 2
5
cm
)
3.3.
Influen
ce of th
e Wi
dth of the Ba
rrier on the
Brea
kdo
w
n
Voltage
In this test shouts, the dif
f
erent di
stan
ce
s between
electrode
s a
nd ba
rrie
r
(2
cm, 3cm
and
4cm
)
, th
e hol
e in th
e
middle
of the
barrie
r
for (5
mm), differen
t
width
of the
barrie
r
(15
c
m,
20cm a
nd 25
cm).
we see cl
earl
y
, the breakd
o
wn voltag
e is very
low
with the increa
si
ng of large
wi
dths of
the barrie
r
an
d very big
with t
he de
crea
sing
of them. This
can
be
explained
by the fact that the
scree
n
plays
a geomet
ric o
p
sta
c
le.
From Fig
u
re
6, the breakd
o
wn voltage de
crea
ses with in
creas
i
ng of the barrier
width
s
.The predictio
n
valu
es are
situate
d
between th
e two
boun
da
ries
of the a
r
ea of ri
sk of 5
%
.
This me
an
s that the model
founded by t
he experi
m
en
tal desig
n me
thod is pe
rfect.
Figure 6. Influence of the width
of the barrier o
n
the breakdo
wn
voltage: different
relative
positio
n of the barrier
-1
-0
.
8
-0.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
0.
8
1
66
66
.
5
67
67
.
5
68
68
.
5
69
69
.
5
70
H
o
l
e
of
t
h
e
m
i
dd
l
e
t
h
e
bar
r
i
e
r
(a
)
B
r
ea
k
d
ow
n
v
o
l
t
ag
e
(
k
V
)
H
i
g
h
e
r
limit
L
o
w
e
r
limit
Me
a
s
u
r
e
d
P
r
edi
t
-1
-0
.
8
-0
.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
0.
8
1
64
65
66
67
68
69
70
H
o
l
e
i
n
t
he
m
i
dd
l
e
of
t
he b
a
r
r
i
er
(b
)
B
r
eak
dow
n v
o
l
t
ag
e (
k
V
)
Hi
gher
l
i
m
i
t
L
o
w
e
r
limit
M
eas
u
r
ed
P
r
edi
t
-1
-0
.
8
-0
.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
0.
8
1
62
63
64
65
66
67
68
69
70
H
o
l
e
i
n
t
he m
i
dd
l
e
of
t
he ba
rr
i
e
r
(c
)
B
r
e
a
k
d
o
w
n v
o
l
t
ag
e(
k
V
)
H
i
gher
l
i
m
i
t
Low
er l
i
m
i
t
M
eas
ur
ed
P
r
edi
t
-1
-0
.
8
-0
.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
0.
8
1
61
62
63
64
65
66
67
W
i
dt
h o
f
t
h
e ba
r
r
i
e
r
(a
)
B
r
eak
do
w
n
v
o
l
t
age
(
k
V
)
H
i
gh
er
l
i
m
i
t
Lo
w
e
r
l
i
m
i
t
M
e
as
ure
d
Pr
e
d
i
t
-1
-0
.
8
-0
.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
0.
8
1
60
61
62
63
64
65
66
67
68
W
i
dt
h
of
ba
rr
i
e
r
(b
)
B
r
eak
do
w
n
v
o
l
t
a
ge
(
k
V
)
H
i
g
h
e
r
lim
it
Lo
w
e
r
l
i
m
i
t
2
M
e
as
u
r
ed
Pr
e
d
i
t
-1
-0.
8
-0
.
6
-0
.
4
-0.
2
0
0.
2
0.
4
0.
6
0.
8
1
54
55
56
57
58
59
60
61
62
63
64
W
i
d
t
h of
b
a
r
r
i
er
(c
)
B
r
e
a
k
d
ow
n v
o
l
t
age
(
k
V
)
H
i
g
h
e
r
limit
Low
er
l
i
m
i
t
M
e
as
ure
d
Pr
e
d
i
t
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 12, Decem
ber 20
14 : 8033 – 80
41
8040
Con
s
id
erin
g experim
ental con
d
ition
s
in whi
c
h we h
a
ve worked an
d approximations we
have made in
the numeri
c
a
l
model, we can say that
re
sults a
r
e satisfactory and
we note that the
experim
ental
results a
r
e well within
the range of 5% confiden
ce e
r
ror.
In experi
m
ent
s d
e
si
gn m
e
thod, verifi
cati
on
experim
e
n
ts
sho
u
ld
b
e
d
one
to
d
e
termin
e
optimum
con
d
itions
and
compa
r
ed
with
experim
ental
values. In thi
s
stu
d
y, all value
s
are wit
h
in
confid
en
ce le
vels, as a result of the verification exp
e
ri
ments.
4
. Conclusion
Investigation
a
l re
sult
s
sh
own
that the
expe
riment
s de
sign
meth
od i
s
a
u
s
ef
ul tool to
sea
r
ch pa
ram
e
ters i
n
fluen
cing the obj
ect
i
ve fi
xed and
optimizatio
n. Met
hodol
ogy of
experime
n
ts
wa
s giving g
u
idan
ce o
n
th
e effects
of different facto
r
s. It con
s
ist
s
in a first
step
to make
se
a
r
ch
para
m
eters i
n
fluen
cing t
h
e obj
ective fi
xed. For
this,
the u
s
e
of t
he meth
od
resp
on
se
su
rface
turns outto b
e
highly effe
ctive, becau
se
it allowsto
cl
assify the effect of pa
ram
e
ters in the o
r
de
r
of import
ance. The
study
on the
effect
s of fa
ctors
h
a
s
on the
on
e han
d a
nd
restri
cting
ch
o
o
se
the para
m
ete
r
s of the de
sired mo
del o
n
the
other
hand to defi
ne the limits of the universe
spe
e
c
h
.
A model
with
a goo
d de
scri
ption of the
studi
ed
system
thus h
a
ve be
en defin
ed, a
nd led
to improvement of system
perfo
rmance.
The modelling of the diel
ectri
c
breakdown in point-pl
an
arrang
ement
with insulatin
g
barrie
r
in a
i
r inte
rval h
a
s
bee
n inve
stigated. We
h
a
ve pro
p
o
s
e
d
a
mathemati
c
al
modelli
ng
by usi
ng
experim
ents d
e
sig
n
meth
o
d
. It lead
s
to analy
z
e t
he
intera
ction
s
b
e
twee
n different pa
ram
e
te
rs:
distan
ce
b
e
twee
n el
ectrode
s, diam
eter of th
e b
a
rri
e
r
and the rel
a
tive position. T
h
is metho
dol
ogy has p
r
ovi
ded go
od re
sults and h
e
lp
s to minimize
the
c
o
s
t
of s
t
udy, express
e
d in number of tes
t
s
.It has
been s
h
own, that the us
e of this
methodolog
y
is very u
s
eful
for tra
c
king t
he influe
nt o
peratin
g pa
ra
meters on
in
sulatio
n
relia
bility and for
the
lifetime mode
lling. In this study we ca
n say t
hat the methodol
ogy of experiem
e
nt desig
n sh
o
w
a
good pe
rfo
r
mance to investigation
s
i
n
the anal
ysis of differe
nt discharge
steps of the
air
interval.
Referen
ces
[1]
Holtzhausen JP, Volsov WL.
High
V
olta
ge E
ngi
neer
in
g Pra
c
tice and T
h
e
o
r
y.
Stellenb
osc
h
. 2014.
[2]
Haddad A, Warn D.
Adv
a
n
c
ed i
n
Hi
ghV
oltag
e
En
gin
e
e
rin
g
. T
he ins
t
itution of
eng
ine
e
rin
g
a
n
d
techno
lo
g
y
, L
o
ndo
n, 200
7.
[3]
Bourek Y, Mokhnac
he L, Nait
Said
N, Kattan
R. Determinati
on of Ioniza
ti
o
n
Con
d
itio
ns C
haracter
i
zin
g
the Br
eak
do
w
n
T
h
reshol
d
of a P
o
int
Pla
n
e
Air Interv
al
U
s
ing
F
u
zz
y L
o
g
ic.
Electric P
o
wer Systems
Research
. 20
1
1
; 81: 203
8-20
47.
[4]
Adami
a
k K, A
ttan P. Sim
u
la
tion
of D
i
scha
r
ge
Coro
na
in
Poi
n
te P
l
an
e
Co
nfigur
atio
n.
Jour
nal
o
n
Electrostaic
. 2
004; 61: 8
5
-98.
[5]
Ducass
e O, Pa
pag
eorg
h
i
ou
L, Eich
w
a
l
d
O, S
p
y
r
on
N,
Yo
usfo M. Critic
al A
nal
ysis
o
n
t
w
o
Dimenti
o
n
a
l
Point-to Pla
ne
Streamer Simu
latio
n
Usin
g F
i
nit Eleme
n
t Me
thod a
nd F
i
nit
Volum
e
Metho
d
.
IEEE Trans
Plas
m
. 200
7; 3
5
(5):12
87-1
3
0
0
.
[6]
Gallimi
berti
I, Bachi
egg
a G,
Bou
d
io
u A.
Lal
anch
e
P. F
ond
ament
ale
Processus
in
Lon
g a
i
r Ga
p
Dischar
ge,
Co
mpte r
end
u ph
ysiqu
e
. 200
2; 3(1): 1135-
13
59
.
[7]
Chen GU, Liang
ZW, Bin FJ, et al.
I
n
flue
nce
of R
od E
l
ectrod
e Struct
ure
on S
u
itch
i
ng I
m
p
u
l
s
e
Dischar
ge C
h
a
r
acteristics of Rod Pl
ane
air Gap
. Procee
di
ng oh th
e CSS
E
. 2011: 12
0-1
27.
[8]
Kojma H, Hoti
a K,
T
w
ata T
,
Ha
yakar
a
N, Yanj
ita N, et al.
L
’
infl
uenc
e o
f
Gap Lenth o
n
Disch
ar
g
e
Cha
nne
l Prop
agati
on a
nd Br
eakd
o
w
n
Méca
nis
m
e i
n
Air
. Procee
din
g
17
th
ISH, 2011, D3
7
[9]
Arvello
L, et al
. A Ne
w
Static
Calcu
l
atio
n of
the Streamer
Regi
on for L
o
ng Sp
ark ga
ps
.
Journal on
electrostatic
. 2
012; 70:
15-1
9
.
[10]
Ottia K, Kgma H, Hay
a
kaw
o
N,
et al.
Imp
u
ls
e Bre
a
kdow
n M
e
ch
anis
m
B
a
se
d
on
Disc
harg
e
Propa
gati
on Pr
ocess Un
der n
on Un
ifor
m Ele
c
tric F
i
eld in Ai
r
. CEIDP. 2011
; 5. B-9
[11]
Roui
ni A, Mah
i
D.
Investigati
on o
n
the Insu
latin
g
Barri
ers Influenc
e in th
e Air Gap Po
i
n
te-Pla
ne b
y
F
i
eld Esti
mata
tion i
n
Pres
e
n
ce Sp
ace
C
harg
e
usi
ng
F
i
nite Vo
lu
me
, X
II An
nu
el
Se
mi
na
ri
o
o
n
Automation Indust
riel an
d Instrumentati
on, T
ang
ier. Maroc. 201
4.
[12]
Dascal
e
scu
L, Miha
lcio
u A.
Norme i
n
tern
ation
a
le
.Ins
u
lati
n
g
li
qui
ds D
e
ter
m
inati
on of th
e bre
a
kd
o
w
n
voltag
e of a ind
u
strial freq
uen
c
y
test met
hod,
Second E
d
itio
n, CEI 156, 19
95 (in F
r
enc
h).
[13]
Boub
ake
u
r A, Mokhn
a
che
L, F
e
liac
h
i A. T
h
e
o
retica
l Investi
gatio
n on the B
a
rrier Effect on
Pointe-Pl
a
n
e
air g
ap Br
eakd
o
w
n
V
o
lta
ge B
a
sed
on Stre
a
m
er Criteri
on.
I
EE Proc SCI
Meas, Technol
. 2004;
151(
3):
167-
174.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Predi
ction the
AC Brea
kdo
w
n Voltage in
Point/Plane Air Gap
s
with
… (Abdel
gha
ni Roui
ni)
8041
[14] T
opalis F
V
, Da
nikas M G. Bre
a
kdo
w
n
i
n
a
i
r
gaps
w
i
th S
o
li
d Insul
a
tin
g
Ba
rrier Un
der Im
puls
e
Volt
ag
e
Stress.
F
a
cta
Unversit
as, SA, Electrical Eng
i
ne
erin
g.
200
5; 18: 87-10
4.
[15] MA Abd Al
la
h, Sa
yed A
W
a
rd
, Amr A Youss
e
f. Effe
ct of F
unction
all
y
Gra
d
ed Mater
i
al
of
Disc Sp
acer
w
i
t
h
Prese
n
ce
of Multi-Co
ntamin
at
ing P
a
rticles o
n
Electr
ic F
i
eld
insi
de
Gas I-nsulate
d
Bus Duc
t
.
Internatio
na
l Journ
a
l of Electr
ical
a
nd Co
mp
uter Engi
ne
erin
g (IJECE).
201
3; 3(6): 831-8
4
8
.
[16]
MA Abd Allah,
Say
e
d A
Ward, Amr A Y
o
ussef. E
ffect of Co
ating
of E
a
rthed
Encl
os
ure
and
Mu
lt
i
Contam
inati
ng
Particles
on Br
eakd
o
w
n
Volta
ge i
n
sid
e
Gas
Insulate
d Bus
Duct.
Internati
ona
l Jo
urn
a
l
of Electrical a
n
d
Co
mp
uter En
gin
eeri
ng (IJECE)
. 2014; 4(4
)
: 471-48
5.
[17]
Atkinson AC, D
onev AN. Opti
mum Exp
e
rim
e
ntal
Des
i
gns,
Oxford U
n
ivers
i
t
y
Press, Ne
w
York, 1992.
[18]
Montgom
er
y
D
C
. Desig
n
and
Anal
ys
is of
Experime
n
ts, Joh
n
W
ile
y& Sons,
Inc. 2005.
[19]
J F
auch
e
r, P
Maussi
on: On-l
ine E
l
ectric
al Q
ualit
y Impr
ove
m
ent of a
Si
ngl
e-P
has
e Bo
ost Rectifier
w
i
t
h
F
u
zz
y
Co
ntrol
l
e
r and E
x
p
e
rim
ental D
e
sig
n
s,
Electro
m
oti
o
n
. 200
6; 13(3): 22
1-23
2
.
[20]
Andu I, Nafiu
LA, Ilimiese PK. Matrix Approac
h
to th
e E
x
p
e
rime
ntal
D
e
sig
n
, of R
e
sp
onse
Surfac
e
Method
olo
g
y
,
Journ
a
l of scie
n
c
e Educati
on a
nd techn
o
l
ogy
. 200
9; 2(1).
[21]
Bisgo
a
rd S.
Must Process
be
in st
astic
a
l co
ntrol
bef
or con
ducti
ng
desi
g
n
E
x
per
iments,
Qua
lit
y
Engi
neer
in
g.
2008; ASQ20(
2)
: 143-17
6.
[22]
T
i
mont gomer
y Doug
las. Desi
gn an
d Ana
l
ysi
s
of Exper
imen
ts. 8
th
ed, Hodoken, Nj, Joh
n
W
i
l
y
&So
ns
.
201
3.
Evaluation Warning : The document was created with Spire.PDF for Python.