TELKOM
NIKA
, Vol.13, No
.3, Septembe
r 2015, pp. 8
28~835
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i3.2092
828
Re
cei
v
ed Ma
rch 1
2
, 2015;
Re
vised J
une
2, 2015; Accepted June 2
0
, 2015
Online Correction of the Dynamic Errors in a Stored
Overpressure Measurement System
Wei Wa
ng
1
*,
Zhijie Zhang
2
Schoo
l of Instrument an
d Ele
c
tronics, Nort
h
Univers
i
t
y
of Chin
a, T
a
iyua
n, Chin
a
*Corres
p
onding author,
e-mail: 2002
ww
2002@126.com
A
b
st
r
a
ct
T
he prob
le
m e
n
cou
n
tere
d in
sharp sh
ock te
sting
as a res
u
lt of ina
deq
u
a
te ba
ndw
idth
must b
e
addr
essed
to
obtai
n an a
ccurate overpr
essure pe
ak
valu
e w
hen
me
asuri
n
g
the
steep s
i
g
nal
s of
shockw
aves d
u
rin
g
expl
osio
n
s
. A dynamic c
o
mpe
n
sato
r ca
n effectively a
m
e
nd the dy
na
mic err
o
rs cau
s
ed
by se
nsor
sys
tem c
har
acteri
stics; thus, a
dyna
mi
c co
mp
ensati
o
n
meth
od
bas
ed
on
i
m
pr
ove
d
p
a
rti
c
le
sw
arm o
p
ti
mi
zation
(PSO) al
gorith
m
is
prop
osed
in
this
pa
per. T
h
is
meth
od c
an
effectively ov
erco
me t
h
e
influ
ence of the
initial va
lu
e de
rived w
i
th PSO algor
it
hm
on c
o
mpe
n
sator in
dex. T
he distri
buted a
l
g
o
rith
m is
introd
uced
i
n
to
the
har
dw
are
structure d
e
si
g
n
of t
he
dyn
a
m
ic co
mpens
ator
to faci
litate
the
ap
plic
atio
n of
an
opti
m
i
z
e
d
co
mpens
ator to real-ti
m
e
o
n
li
ne
me
asur
e
m
ent. T
h
is integr
atio
n reali
z
e
s
the
hig
h
-spe
ed p
a
r
a
lle
l
of the dyna
mic
comp
ens
ator
of the sensor
w
i
th field-pro
g
r
a
mmabl
e
gat
e array. Experi
m
ental res
u
lts show
that a h
i
gh-s
p
e
ed p
a
ra
lle
l dy
n
a
mic co
mpens
ator can
a
m
e
n
d
the
dyna
mic
errors i
n
a s
e
n
s
or accur
a
tely
an
d
in a timely
ma
n
ner.
Ke
y
w
ords
:
dyna
mic co
mp
ensati
on
meth
od, i
m
pr
oved
parti
cl
e sw
arm opti
m
i
z
e
d
alg
o
rith
m, distri
b
u
ted
algorithm
,
FPGA
Copy
right
©
2015 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduction
A dynami
c
t
e
sting
sy
ste
m
is in
stalle
d un
de
rgrou
nd to
test
explosi
on
sh
ockwaves. A
sen
s
o
r
interf
ace i
s
locate
d above g
r
ou
nd. The sp
e
c
trum of the p
r
essu
re
sen
s
or cannot
co
ver
the si
gnal
spectrum
und
er th
e influ
e
n
ce
of th
e
cliff front of t
he
sho
c
kwave. Thu
s
,
sig
nal
amplitude i
n
cre
a
ses
sig
n
i
ficantly. The
area
s
se
le
cted by the dynamic te
st
ing sy
stem
are
sho
c
ked
sha
r
ply, and
pe
ak
overp
r
e
ssure
ca
nnot
b
e
dete
r
min
e
d
accu
rately
whe
n
the
se
nso
r
approa
che
s
resona
nt fre
quen
cy. He
nce, the
a
m
plitude fre
quen
cy cha
r
acteri
stics of
a
piezoele
c
tri
c
sen
s
o
r
shoul
d be compe
n
s
ated. Co
mp
ensation met
hod
s incl
ude
inverse filteri
ng,
the collo
catio
n
of
ze
ro
pol
es, a
n
d
sy
ste
m
ide
n
ti
ficati
on [1]. In
add
ition, neu
ral
netwo
rk,
pa
rticle
swarm o
p
timization
(PSO) algo
rith
m, and other algo
rithm
s
are a
dop
ted to improve
comp
en
satio
n
pre
c
isi
on.
Neu
r
al n
e
two
r
ks
can
ea
sil
y
be trap
ped
in the lo
cal
minimum
re
gardl
ess of
netwo
rk
sea
r
ch sp
eed
, and Internet
preci
s
io
n is difficult
to improve at latter stage
s of training. Althou
gh
the PSO
alg
o
rithm i
s
a
h
o
listic optimal
algo
rithm, th
e initial
po
sition of th
e
particle affe
cts the
optimizatio
n
results of thi
s
algo
rithm [
2
]. On
the b
a
si
s of the a
f
oreme
n
tione
d prin
cipl
es,
the
pre
s
ent
pap
e
r
p
r
op
ose
s
a
n
imp
r
oved P
S
O algo
rithm
that empl
oys ada
ptive ne
ural
networks to
determi
ne the
optimal initial
value for ea
ch parti
cle
in the partic
le
s
w
arm within the s
h
ortes
t
time.
This alg
o
rith
m eventually yields a holi
s
t
i
c optimal val
ue.
The dynami
c
compe
n
sato
r is no
rmally
of
high ord
e
r, and the
dynamic e
r
ro
rs of the
sen
s
o
r
a
r
e
difficult to am
en
d qui
ckly
and
in real time
a
c
cordi
ng to
th
e programmi
ng ide
a
s of th
e
traditional
di
spla
cem
ent
summ
ation [
3
]. The
current stu
d
y establish
e
s
a p
a
rallel
metho
d
to
develop ha
rdwa
re for d
y
namic com
pen
sation filtering a
c
co
rd
ing to the con
c
e
p
ts of
the
distrib
u
ted al
gorithm. Thi
s
method co
nverts the in
d
e
x of optimal dynamic
comp
e
n
satio
n
filtering
as
obtain
ed
with the
imp
r
oved PS
O
algorith
m
int
o
the
RO
M l
ook-up
table
ope
ration
while
avoiding
the
multiplicatio
n op
eratio
n. Com
pen
sati
on results can b
e
g
ene
rated th
roug
h
the
perfo
rman
ce
of a simpl
e
data ad
dition
operation
af
ter the loo
k
-up table i
s
i
n
trodu
ce
d. T
h
is
pro
c
e
s
s signi
ficantly
in
cre
a
se
s ope
rati
on spe
ed.
Fi
nally, this
m
e
thod
ca
n eff
e
ctively me
a
s
ure
dynamic
com
pen
sators in real time.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Online
Corre
c
tion of the Dynam
ic Errors in a
Stored
Ove
r
p
r
e
s
sure
Measu
r
em
en
t… (Wei
Wan
g
)
829
2. Amendment Me
thod
of Sensor
Dy
namic Errors Bas
e
d on
PSO Algorithm
Figure 1
sho
w
s th
e ge
ne
ral p
r
in
ciple
s
of dynami
c
errors in
se
nso
r
s ba
sed
on the
improve
d
PSO algo
rithm.
y
k
is the
sen
s
o
r
output,
rk
is th
e output of th
e refe
ren
c
e
model,
zk
is the com
p
e
n
sate
d output
, and m is the
compe
n
sator orde
r.
Figure 1. Sch
e
matic of sen
s
or dyn
a
mi
c errors
Acco
rdi
ng to the input vect
or of the algo
rithm,
X
k
can b
e
expre
s
sed a
s
follows:
()
,
(
1
)
(
)
ky
k
y
k
y
k
m
X
(1)
Comp
en
sato
r index is dete
r
mine
d as foll
ows:
01
1
,,
T
mm
WA
A
A
A
(2)
The co
mpe
n
sated output from Form
ula
s
(1) a
nd (2
) is
obtaine
d usi
n
g:
()
(
)
zk
W
X
k
v
k
(3)
W
h
er
e
vk
is the
un
correl
ated
ran
dom
noi
se that
follo
ws the n
o
rm
al d
i
stributio
n. Th
e
mean squa
re
d errors between
zk
and
rk
are determi
ned u
s
ing:
2
0
1
()
(
)
N
k
J
rk
z
k
N
(4)
Adaptive neu
tral networks determi
ne the optim
al i
n
itial value for the PSO algorithm
according to
Formul
a (5
) for co
ntinual
rene
wal.
()
(
1
)
(
)
(
1
)
zk
W
k
X
k
b
k
(5)
Whe
r
e
1
Wk
an
d
1
bk
rep
r
e
s
en
t the comp
ensated ind
e
x and thre
shol
d,
respec
tively,
when the c
o
mpen
s
a
ted network
trains
to the
1
k
th s
t
ep.
The comp
en
sated
ind
e
x
a
nd
threshold are ren
e
wed according
to
Formul
as
(6
) and (7)
durin
g the ne
ural net
wo
rk t
r
ainin
g
proce
ss.
(1
)
(
)
(
1
)
(
)
Wk
W
k
X
k
e
k
(6)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 828 – 835
830
(1
)
(
)
(
1
)
bk
b
k
e
k
(7)
Whe
r
e
is the
study facto
r
who
s
e p
a
ram
e
ters
sh
ould
not be excessively larg
e to meet
conve
r
ge
nce requi
rem
ents.
This varia
b
le
typically take
s a value of 0
<
< 1.
We re
ga
rd Fo
rmula (4) a
s
the stan
dard in
the entire p
r
ocess of
ada
ptive neutral
netwo
rk
training.
The
dynamic
com
pen
sator ind
e
x
obtaine
d i
s
the initial val
u
e of th
e PSO
algorith
m
wh
en
training rea
c
hes a sta
ge i
n
which the mean squa
re
d erro
rs and
J are le
ss than a set valu
e or
whe
n
the trai
ning times
re
ach a
certai
n value [4].
Whe
n
we
ap
ply the PSO algorithm to
solve dyna
mic compe
n
sator ind
e
x, we sho
u
l
d
encode the
algorith
m
properly to ge
nerate th
e
p
a
rticle. O
n
the ba
sis
of PSO algorit
hm
cha
r
a
c
teri
stics, a real n
u
m
ber
can b
e
used to
rep
r
esent ea
ch
para
m
eter. I
n
addition, if W
rep
r
e
s
ent
s th
e current lo
cation of e
a
ch
parti
cle, the
n
anoth
e
r pa
rticle th
at correspon
ds to
sho
u
ld b
e
ge
nerate
d
to
re
pre
s
ent
parti
cle sp
eed. Fitn
ess fun
c
tion
F
W
a
s
s
e
s
s
e
s
the p
r
os
an
d
c
o
ns
o
f
th
e c
u
rr
en
t
loc
a
ti
on [5]. Amo
ng the
s
e
variable
s
,
W
is a
n
m-dimen
s
i
onal va
riabl
e
;
therefo
r
e,
should b
e
an
m-dime
nsi
o
n
a
l variabl
e a
s
well so th
a
t
the particl
e
can a
dopt t
h
e
f
o
llowin
g
cod
e
st
ru
ct
ur
e:
)
(
W
F
01
1
,,
m
m
A
AA
A
m
m
m
v
v
v
v
v
v
,
,
,
,
1
2
2
1
0
Figure 2. Cod
e
stru
cture di
agra
m
of
the particl
e in the
PSO algorith
m
The PSO
alg
o
rithm
uses t
he a
daptive f
unctio
n
to
de
termine
the
p
r
os a
nd
co
ns of the
curre
n
t pa
rticl
e
location. O
n
ce th
e alg
o
ri
thm is
com
p
l
e
ted, the o
p
timum solution
obtaine
d is the
smalle
st pa
rticle thro
ugh
o
u
t the operati
on. This
valu
e rep
r
e
s
ent
s the para
m
ete
r
value. Final
ly,
comp
en
sato
r index is de
rived [6].
3. Algorithm Verification
Given thi
s
h
a
r
dware d
e
si
g
n
, the
pre
s
su
re
se
nsor i
s
prima
r
ily st
ud
ied th
roug
h
d
y
namic
calib
ration
e
x
perime
n
ts a
nd compute
r
simulatio
n
s.
Then, the
PSO algorith
m
is ap
plied
for
optimizatio
n
according
to
the input
and
output d
a
ta
of the sen
s
o
r
and
of the
referen
c
e
mo
del.
Specifically, the algo
rithm i
s
verified th
ro
ugh
a dyn
a
mi
c calibration
experim
ent that empha
si
ze
s
the pressu
re
sen
s
o
r
. In t
h
is exp
e
rim
e
nt, the
sh
ock tube
sho
w
n
in Figu
re 3
g
enerates a
step
pre
s
sure as
a stand
ard signal that is inco
rpo
r
ated
into the me
asu
r
ed
pre
ssure
sen
s
o
r
b
y
analyzi
ng the
respon
se of
the pre
s
sure sen
s
o
r
output
, calibrating this sen
s
or, a
nd studyin
g the
actual
wo
rki
n
g perfo
rma
n
ce. The curve
depicte
d
in
Figure 4 is th
e actual
mea
s
ured respon
se
curve of the
specifi
c
pre
s
su
re se
nsor.
Figure 3. Shock tub
e
The improve
d
PSO algorit
hm can b
e
used in
the inverse m
odeli
ng of the sen
s
or. Then,
the sen
s
o
r
o
u
tput can
be
equ
ated to
comp
en
s
a
te
d s
y
s
t
em inp
u
t. T
h
u
s
, th
e
c
u
r
v
e
c
a
n be
rega
rd
ed a
s
comp
en
sated
system inp
u
t
and the ref
e
ren
c
e m
ode
l as the ide
a
l
step si
gnal.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Online
Corre
c
tion of the Dynam
ic Errors in a
Stored
Ove
r
p
r
e
s
sure
Measu
r
em
en
t… (Wei
Wan
g
)
831
Furthe
rmo
r
e,
indices a a
n
d
b of the sen
s
or dy
nami
c
compen
sate
d filter are
obtai
ned, wh
ere a
=
[1.1223,
−
0.2
144,
−
0.3
357
, 0.3262] and
b = [1.2817,
−
1.01
31, 0.2
533, 0.538
1].
The
com
pensation result is illust
rated as
curve
2, where the re
sponse ti
me after
comp
en
satio
n
is 1
0
µs, t
he oversh
oot
is 10%,
an
d all re
sult
s
meet the technical indi
cat
o
r
requi
rem
ents.
The simul
a
tion re
sult
s ob
tained in MA
TLAB sho
w
that re
spon
se
is accele
rat
e
d
and that the workin
g freq
u
ency ba
nd broade
ns.
Mo
re
over, noise is effectively removed.
Figure 4. Co
mpen
sated
re
sults fo
r the sensor
4. S
y
stem
Hard
w
a
re
De
sign
Digital si
gnal
pro
c
e
s
sor d
e
vice
s are i
n
com
petent f
o
r the sy
ste
m
whe
n
a d
y
namic
comp
en
sated
filter is used i
n
real
-time ca
se
s with hig
h
requi
rem
ents.
Non
e
thele
s
s, field-p
r
og
ram
m
able g
a
te a
rray (F
PGA)
device
s
serve
as excellent
carrie
rs
for thi
s
filter
given the look-up
table structure and
parallel
proces
sing capability of such devices
[7]. This stud
y design
s
a data storage
system in
which FPGA i
s
the co
re control unit. This
system in
clu
des
comp
one
nts su
ch a
s
power
ma
na
gement, AD
control, data
stora
ge, and
data
transmissio
n
module
s
. T
he ha
rd
ware
circuit struct
ure d
e
si
gne
d
in this stu
d
y
is depi
cted
in
Figure 5.
Figure 5.Ha
rd
ware ci
rcuit structu
r
e
4.1. Pressur
e
Sensor
Cha
r
ge a
m
pli
f
iers h
a
ve be
en incorp
orated into pie
z
o
e
lectri
c p
r
e
s
sure
sen
s
o
r
s
becau
se
of the rece
n
t
developm
e
n
t in integ
r
a
t
ed ci
rc
uit te
chn
o
logy. S
u
ch
se
nsors are
kn
own
as
integrate
d
ci
rcuit
pie
z
o
e
lectri
c (ICP) pre
s
su
r
e
se
nso
r
s.
S
u
c
h
se
ns
or
s
ca
n ov
e
r
c
o
me
t
h
e
disa
dvantag
e
of tra
d
itional
pre
s
sure
sen
s
ors and
exhi
bit a
stro
ng
a
n
ti-interfe
ren
c
e capa
bility d
u
e
to the prese
n
ce
of the in
ternal
cha
r
g
e
amplifie
r [8].
ICP pre
s
su
re se
nsors ha
ve signifi
cant
ly
improved test accura
cy and reli
ability in
compari
s
on
with tr
aditional pressure
sensors. Thus,
ICP
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 828 – 835
832
pre
s
sure se
n
s
ors a
r
e an i
deal choi
ce a
s
sh
ockwav
e
pre
s
sure se
n
s
ors. Thi
s
stu
d
y use
s
the ICP
pre
s
sure sen
s
or from the 113 se
rie
s
m
anufa
c
ture
d by The PCB Comp
any. Th
e rang
e of this
sen
s
o
r
is fro
m
50 p
s
i to 1
000 p
s
i. In de
signi
ng t
he
conditionin
g
circuit
of the fol
l
owin
g si
gnal
s to
ensure
stabili
ty,
the reson
ant freque
ncy of
the sen
s
or i
s
over 5
00 kHz, the resp
on
se time
is
s
h
orter than 1
μ
s, non-li
ne
ar is lo
we
r than 1% FS, and output impe
dan
ce is le
ss than 100
.
4.2. Signal Conditioning
Circuit
Signal
condit
i
oning
circuit
s
conve
r
t the output
sig
nal of the
sensor to
me
et the
requi
rem
ents of the
sub
s
e
quent
acqui
si
tion ci
rcuit.
T
he tran
sform
a
tion
relation
is illust
rated
in
Figure 6. Th
e full output
sign
al ra
nge
of the ICP
p
r
essu
re
se
nsor is
5 V, the offset volta
ge
rang
es from 8 V to 14 V,
and the sam
p
le volt
age o
f
the subseq
uent acqui
sition circuit ran
ges
from 0
V to
2.5 V. T
h
e
r
efore, the fi
rst
step
in
deriving
the
co
nditioni
ng
sig
nal i
n
vol
v
es
excha
nging t
he co
uplin
g sign
al and t
hen ente
r
ing
the scali
ng
circuit. The
desi
gn ab
ove is
premi
s
e
d
on
the co
ndition
of full sen
s
o
r
ra
nge. In
p
r
actice, small
sign
al testing
situation
s
a
r
e
observed. Th
erefo
r
e, we should am
plify the effective output signa
l of
the sensor to fully utilize
the numbe
r o
f
significa
nt A/D conve
r
ter
digits for imp
r
oving testing
accuracy.
Figure 6. Signal co
nditioni
ng circuit an
d
signal tra
n
sf
orm rel
a
tion
4.3. Samplin
g Storage
Circuit
Sampling
sto
r
age
ci
rcuits
quantify the
samp
li
ng pro
c
e
s
s
an
d sto
r
e sho
c
kwave
si
gnal
records. Thi
s
circuit con
s
i
s
ts of a
n
A/D conv
erte
r
and me
mory,
as de
picte
d
in Figure 7.
This
study obtain
e
d
the approximati
on A/D converte
r AD7
482 from the
Analog Devi
ce
s Com
pan
y.
The resolutio
n
of this
con
v
erter i
s
12
b
i
ts
and
ca
n reach 3 M
H
z. Four type
s
of pro
g
ram
m
able
sampli
ng freq
uen
cie
s
a
r
e d
e
tected
in thi
s
scena
ri
o:
2,
1,
500, and 250 kHz.
T
o
accele
rate da
ta
acce
ss, the system utilize
s
static
ra
ndo
m acce
ss me
mory with a storage
cap
a
ci
ty of 2 MW. F
o
u
r
types of prog
rammabl
e system record
ca
pacitie
s are consi
dered a
s
well:2, 1, 512
, and 256 kW.
Figure 7. Block di
agram of
the samplin
g
storag
e ci
rcu
i
t
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Online
Corre
c
tion of the Dynam
ic Errors in a
Stored
Ove
r
p
r
e
s
sure
Measu
r
em
en
t… (Wei
Wan
g
)
833
4.4. D
y
namic
Compens
a
ted Filter Ba
s
e
d on FPGA
To p
r
od
uce t
he dyn
a
mic
comp
en
sato
r in FPGA,
we mu
st divid
e
this array
into six
subm
odul
ars:
AD control, data co
ntrol,
shift regi
ster,
look-up tabl
e, cumul
a
tive sum, and d
a
ta
stora
ge mo
du
les.
a)
The AD
cont
rol mod
u
le
mainly en
sures the p
r
e
c
ise sam
p
ling o
f
AD7482 u
n
der a
pre
s
e
t
sampli
ng rate
, as well a
s
the timely output of
the AD conversion
re
sult from AD74
82 data.
b)
The data
co
ntrol mo
dule
prima
r
ily facil
i
tates
the
co
operation of
AD with oth
e
r mod
u
le
s in
latchin
g
the AD74
82 conve
r
sio
n
re
sult wi
thin
the FPGA to perform
other func
tions
[9].
c)
The shift re
g
i
ster mai
n
ly shifts the da
ta
of each registe
r
acco
rding to the orde
r of the
comp
en
sated
filter. This register the
n
store
s
t
he AD result
s in the lowe
st part of each regi
ste
r
[10].
d)
The loo
k
-up
table mod
u
l
e
mainly sto
r
es
all p
r
ob
able typing
cal
c
ulatio
n result
s for th
e
comp
utation
of the dyn
a
m
i
c
comp
en
sat
ed filter
i
ndex
and
to loo
k
up the
preci
s
e calculation
result input at that moment according
to the typing and
result outp
u
ts [11].
e)
The cum
u
lati
ve sum modu
le prima
r
ily shifts
the sum
m
ation of the out
put data in the look-up
table to perfo
rm the multipl
i
cation fun
c
tio
n
throug
h ad
dition.
f)
The data sto
r
age mod
u
le mainly store
s
the co
mpen
sated re
sults i
n
FLASH accordin
g to the
time sequ
en
ce pre
s
ente
d
in the FLASH
chip h
and
boo
k.
4.5. USB Dri
v
e
The
stora
ge
record mod
u
l
e
displays
a
USB interfa
c
e. Th
e ci
rcuit of this m
odule i
s
establi
s
h
ed
with the ai
d
of
a USB
drive
and i
s
co
ntrol
l
ed
by micro
c
ontrolle
rs.
Thi
s
study uses
the
FT245
R US
B drive pro
d
u
ce
d by Fut
u
re Te
ch
nolo
g
y Device
s I
n
ternatio
nal.
The tra
n
smi
s
sion
rate of thi
s
dri
v
e ca
n rea
c
h
8 Mbp
s
. M
o
reover, the
dri
v
e ha
s a
256
B re
ceive b
u
ffer an
d a
12
8
B
sen
d
buffer.
This
USB dri
v
e can
sup
p
l
y
powe
r
thro
ugh a
USB bus o
r
thro
ugh
the system.
This
hard
w
a
r
e d
e
s
ign a
pplie
s
bus
sup
p
ly powe
r
metho
d
to redu
ce system power con
s
umptio
n
.
In
addition, the
deco
uplin
g netwo
rk
co
nsists of ma
g
n
e
tic bea
ds
a
nd ha
s the capa
city to store
electri
c
ity. Th
e USB i
s
pl
aced b
e
twee
n t
he b
u
s
and
F
T
245
R, a
s
sh
own
in Fi
gure
8. The
FT2
4
5
R
curre
n
t is 1
5
mA in a no
rmal wo
rking
environ
ment;
this current i
s
lo
wer th
an
that in su
sp
e
n
d
mode.
Figure 8. USB drive circuit
5. Testing Through Expe
rimentatio
n
The
assem
b
l
ed te
sting
sy
stem i
s
ill
ustrated in
Fi
g
u
re 9. An
ICP
sensor is in
stal
led in
the
cente
r
of the
mech
ani
cal
shell, an
d th
e
sen
s
itive
su
rface
is lo
cate
d at th
e
sam
e
level
a
s
the
up
-
surfa
c
e of th
e shell. A pro
t
ective cover
is situat
e
d
at the perip
hery
of the up-su
rface; this cover
is con
nect
ed to
the scre
w thread, and
the contro
l p
a
nel i
s
un
de
r t
he
cover.
Thi
s
control p
a
n
e
l
mainly con
s
i
s
ts of a power switch, cha
r
ging inte
rfa
c
e
,
USB interface, and state
indicato
rs. T
he
swit
ch i
s
turn
ed on p
r
io
r to testing. Th
en, the
testin
g system
rel
oad
s the working p
a
ramet
e
rs
automatically. In the proce
ss, th
is
syste
m
enters the trigge
ring
state. If
the working paramete
r
s
must be am
e
nded, then th
e system
can
conn
ect a
co
mputer th
rou
gh the USB i
n
terface an
d can
prog
ram p
a
rameters u
s
in
g spe
c
ific
so
ftware.
On
ce
the prog
ra
m is com
p
le
ted, the working
para
m
eters o
f
the system can b
e
refres
hed an
d store
d
internally in
E2PROM.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 828 – 835
834
Figure 9. Image of the testing syste
m
Fi
gure 10. Image of a ce
rta
i
n bomb du
rin
g
actual te
sting
Figure 10
sh
ows the im
ag
e of
a static explosive experim
ent
invo
lving a certai
n bom
b.
The mea
s
u
r
e
d
height of the wo
ode
n suppo
rt unde
r which a ste
e
l plate lies
is 1.5 m. The
proje
c
tion
of t
he
cent
ral
axis of
the
bomb
is se
le
cted
a
s
the
explo
s
io
n cente
r
. Th
e
testing
syste
m
applie
s thi
s
center
as th
e
center
of
a
circle that ra
diate
s
out
wa
rd in t
h
ree
directio
n
s
. Fou
r
te
stin
g
dots a
r
e lai
d
out in ea
ch
dire
ction a
s
t
he set radii
(distan
c
e
bet
wee
n
the te
sting dot an
d
th
e
explosi
on
ce
nter) at 5, 7,
10, and
15
m
.
Figure 11
in
dicate
s the
a
c
tual te
st dat
a of a
ce
rtain
dot
in direction
1.
The
re
sult su
gge
sts that th
e pea
k
sho
ckwave ove
r
p
r
e
s
sure
ge
neral
ly confo
r
ms to
the p
r
in
ciple
s
of mo
noton
e
re
du
ction
be
cau
s
e
the
testing dot
s
are
f
a
r from th
e
e
x
plosio
n
cent
er.
This
finding als
o
reflec
ts
the transmiss
ion
cha
r
a
c
t
e
ri
st
ics of
t
he
expl
osive sho
c
kwave.
Figure 11. Actual testing curve
6. Conclusion
This
study p
r
opo
se
s a
d
y
namic
com
pen
sator
de
sign metho
d
based on th
e PSO
algorith
m
. Output and in
put are o
b
tai
ned thro
ugh
sen
s
o
r
cali
bration and a
r
e maximize
d
to
prod
uce an
optimize
d
dynamic
com
p
ensator. The
effect of modeling e
rro
rs on the dynamic
comp
en
satio
n
of the
sen
s
or i
s
avoid
e
d
in
th
e a
b
sence of
dyna
mic m
odeli
n
g
thro
ugho
ut t
he
pro
c
e
ss.
The
dist
ribute
d
a
l
gorithm
effe
ctively
ame
n
d
s
th
e dyna
mic errors
in
se
nsors
at
high
spe
ed a
nd
online. T
he
experim
ental
re
sult in
dicates
th
at
a high-sp
eed
parall
e
l
dyn
a
mic
comp
en
sato
r can am
end d
y
namic e
rro
rs of sensors a
c
curately an
d
in real time.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Online
Corre
c
tion of the Dynam
ic Errors in a
Stored
Ove
r
p
r
e
s
sure
Measu
r
em
en
t… (Wei
Wan
g
)
835
Referen
ces
[1]
Liu Q, Ca
o GH. T
he Modeli
ng R
e
ferenc
e
and th
e
D
y
n
a
m
ic Comp
ens
ated Al
gorithm
of the Error
s
W
h
iteni
ng Se
n
s
or.
Control T
h
eory an
d App
l
i
c
ation
. 20
09; 2
6
(3): 256-
26
0.
[2]
T
i
an W
J
, Liu JC.
A New
Optimi
z
a
ti
on Alg
o
rith
m for Dyn
a
mic Co
mp
en
sation of Se
ns
ors
. Second
Internatio
na
l C
onfere
n
ce o
n
Comp
uter Mo
d
e
lin
g an
d Simu
latio
n
. 201
0; 1: 1707-
17
10.
[3]
W
u
J, Z
hang Z
J
.
Real-ti
m
e C
o
rrectio
n for Sensor
’
s
Dyna
mic Error Based
on DSP
.
IEEE Internationa
l
Instrumentatio
n and Me
asur
e
m
ent T
e
chnol
o
g
y
C
onfer
enc
e. 2011; 1: 63
3-6
38.
[4]
W
ang F
,
Lv F
Z
.
Applicati
on
of Non
lin
ear A
daptiv
e
F
ilter i
n
the Si
gn
al P
r
oc
essi
ng of M
agn
etosticti
v
e
Guide
d
W
a
ve T
r
ansducer.
Chin
ese Jo
urna
l of Sensors a
n
d
actuators
. 201
0; 23(11): 1
594
-159
8.
[5]
Che
n
S. Dig
ita
l
IIR filter des
i
gn us
ing
partic
l
e s
w
arm
opti
m
izatio
n.
Int J Mode
lli
ng Id
e
n
tification and
control
. 20
10;
9(7): 327-
33
5.
[6]
W
u
XJ, Z
han
g D. Ne
w
meas
u
r
ement metho
d
of
round
ness
error bas
ed o
n
particle s
w
a
r
m
optimizati
o
n
algorithms.
Chi
nese Jo
ura
n
l o
f
Sensors an
d actuators
. 20
0
7
; 20(4): 83
2-8
34.
[7]
Z
hong YQ, Z
hao L. Des
i
gn
a
nd imp
l
eme
n
ta
tion of rece
iver
front-end filter
in
w
i
r
e
less se
n
s
or net
w
o
rks.
Chin
ese Jo
urn
a
l of Sensors
a
nd actuat
ors
. 2009; 22(
7): 103
4-10
39.
[8]
Sascha E, Alfred L. D
y
namic
Uncertaint
y fo
r Compens
ate
d
Secon
d
-Ord
er S
y
stems.
Sensors.
20
10;
10(8): 76
21-
76
31.
[9]
F
an BS, T
an GZ
, F
an SS. Comp
ariso
n
of
T
h
r
ee Differe
nt 2-D S
pac
e
Vector PW
M
Algorit
hms a
n
d
T
heir F
P
GA Implem
entatio
ns.
Journ
a
l of Pow
e
r T
e
chno
lo
gie
s
.
2014; 94(
3): 176-
189.
[10]
Sutikno
T
.
FPGA for rob
o
tic appl
icati
o
ns:
F
r
om androi
d/huma
n
o
i
d robots to artificial me
n.
T
E
LKOMNIKA (T
eleco
m
mu
ni
cation, Co
m
put
ing, Electro
n
ics
and Co
ntrol).
201
1; 9(3): 401
-402.
[11]
Harikris
hn
a D.
D
y
n
a
mic sta
b
ilit
y
en
ha
nce
m
ent of
po
w
e
r s
y
st
ems us
i
ng
ne
ural-
net
w
o
rk c
ontrol
l
e
d
static-compe
ns
ator.
T
E
LKOMNIKA (T
eleco
m
mu
nicati
on,
Co
mp
uting,
E
l
ectronics an
d Contro
l)
. 20
13;
10(1): 9-1
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.