TELKOM
NIKA
, Vol. 11, No. 9, September 20
13, pp.
5293
~52
9
8
ISSN: 2302-4
046
5293
Re
cei
v
ed Ap
ril 5, 2013; Re
vised J
une 1
0
, 2013; Acce
pted Ju
ne 21,
2013
A Signal Pre-processin
g
Algorithm Applied for
Ultrasonic Flow-meter
Rang
-ding Wang
*
1
, Qiang Liu
1
, Chen-to
u Du
1
, Ling Yao
2
1
School of Infor
m
ation Sci
enc
e and En
gi
neer
ing,
Ni
ngb
o Un
iversit
y
, N
i
n
g
b
o
315
21
1, Chin
a
2
Ning
bo W
a
ter Meter CO. L
T
D, Ningb
o, 315
0
00, Chi
n
a
*Corres
p
o
ndi
n
g
author
，
em
ai
l: w
a
ngr
ang
di
n
g
@n
bu.ed
u.cn
A
b
st
r
a
ct
In order to solv
e the prob
le
m
of time differ
e
n
c
e
ultraso
n
ic fl
ow
meter
’
s low
accuracy, aga
inst the
basic ch
aracte
ristics of the sampl
e
data, a data-pr
ocessi
n
g
alg
o
rith
m is
prop
osed. F
i
rst, w
e
use shell sort
do
a d
a
ta
pre-
process
i
ng
to t
he s
a
mpl
e
s, th
en r
e
move
the
error
of the
s
a
mpl
e
sp
ace
b
y
co
mpl
e
x
dig
i
ta
l
filter, and use the error co
mp
ensati
on
al
gori
t
hm to get the final sa
mp
le
re
sults. Amon
g the
m
, the comp
l
e
x
digit
a
l filter
is ma
inly c
o
mp
o
s
ed
by
medi
an
filtering
al
gorit
hm, sl
idi
ng w
i
ndow
, Peters
alg
o
rith
m a
nd
the
w
e
ighted
av
era
ge
alg
o
rith
m. T
h
is ki
nd
of d
a
ta
proc
essi
n
g
a
l
g
o
rith
m ca
n
effectively filter
o
u
t the
error
of th
e
sampl
e
space.
It can also mak
e
a larg
e i
m
pro
v
ement to
the accuracy of ultr
ason
ic flow
me
ter w
h
ile ensur
e
the stabil
i
ty an
d real-ti
m
e.
Ke
y
w
ords
:
ultr
ason
ic flow
me
ter, sort metho
d
, digita
l filter, error co
mp
ens
ation
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
In the indu
strial and
ag
ricu
ltural ap
plications
, flo
w
me
ter is
gen
eral
ly used fo
r
re
al-tim
e
measurement
of fluid, so i
n
the
co
urse
of their work
it was u
s
u
a
lly affected by random fa
cto
r
s,
su
ch a
s
environmental fa
ct
ors
and h
u
m
an facto
r
s.
T
he data
sign
al whi
c
h was measure
d
from
flow meter
contain
s
a large part of n
o
ise.
When t
hose noi
se
s mix with measure
d
sig
n
a
l,
accuracy will
be se
riou
sly
affected, even can
not
work by the
st
rong i
n
terfe
r
e
n
ce
of n
o
ise
s
[
1
].
For the abov
e rea
s
on
s, d
a
ta pro
c
e
ssi
n
g
to the si
gn
al measure
d
by flow meter is beco
m
ing
th
e
essential p
a
rt
in the desig
n
process.
This arti
cle a
nalyze
d
abou
t the features of di
gital signal of the time differen
c
e u
l
traso
n
ic
flow meter
whi
c
h is ba
sed on MSP430F44
9
chi
p
, and prop
osed data so
rting method a
n
d
filtering meth
od [2] comm
only use
d
wh
ich a
r
e corre
s
po
nde
d to it, summa
rized
and integ
r
at
ed
the variou
s t
y
pes of o
p
timal algo
rith
ms, and
im
pl
emented i
n
softwa
r
e, fin
a
lly got the
digital
pro
c
e
ssi
ng al
gorithm
s to m
eet the requi
re
ments of me
asu
r
em
ent sy
stem a
c
cura
cy.
2. Data F
eatures of
the T
i
me Differ
e
n
ce Ultr
asoni
c Flo
w
M
e
te
r
The time
differen
c
e
ultrasonic flow met
e
r
use
s
the ti
me differen
c
e bet
wee
n
ul
traso
n
ic
wave
s in
the
fluid
downst
r
eam
an
d u
p
s
trea
m in
the
sa
me
dista
n
c
e,
while
thi
s
time diffe
ren
c
e
has a
relatio
n
s
hip
with the
rate of fluid flow, so
flo
w
rate can
be drawn if the tim
e
differen
c
e h
a
s
been
mea
s
u
r
ed, we
can
a
l
so
cal
c
ulate
the fluid flo
w
,
the b
a
si
c p
r
i
n
cipl
e i
s
sho
w
n i
n
Fig
u
re
1
.
H1, H2 i
s
a pair of rotati
on tran
smitting and recei
v
ing ultrasoni
c tran
sdu
c
e
r
s, D is the p
i
pe
diamete
r
, V is the fluid flow rate, and
is the angl
e betwe
en
chann
el and t
he axis of th
e
pipeline.
Figure 1. The
Basic Prin
cip
l
e of Ultrasoni
c Flow M
e
ter
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 529
3 – 5298
5294
Assu
ming
the
sp
eed
of ultraso
n
ic in th
e
meas
ured flui
d is C, time
from
H1
to
H2 is
1
t
,
from H2 to H1 is
2
t
.
0
is the circuit del
ay time, which is f
a
r less than p
r
opa
gation ti
me, then:
0
1
sin
cos
/
V
C
D
t
(1)
0
2
sin
cos
/
V
C
D
t
(2)
2
2
2
1
2
sin
2
V
C
DVtg
t
t
t
(3)
In gene
ral i
n
d
u
strial
mea
s
u
r
eme
n
t, pro
p
agation vel
o
ci
ty of ultraso
n
i
c
in th
e liquid
(ab
out
1450m / s in
the water) i
s
larg
er than
the liquid’s,
which i
s
2
2
2
sin
V
C
, s
o
the time
differen
c
e ca
n be simplifie
d as:
tg
C
DV
t
t
t
2
1
2
2
(4)
Therefore, th
e basi
c
eq
uat
ion of the inst
antane
ou
s flow rate
can be
written a
s
:
t
Dtg
C
V
2
2
(5)
Then, the inst
antane
ou
s flow rate i
s
:
t
tg
DC
K
V
D
K
V
S
K
Q
8
4
2
1
2
1
1
(6)
And
1
2
2
1
n
n
K
,
n
ha
s a
re
lationship
with the
Reynol
ds n
u
mb
er,
S
is t
he P
i
p
e
cro
s
s
-
sect
io
na
l
area.
It can
be
see
n
from
(6
), a
s
lo
ng a
s
we
get the
mea
s
ured tim
e
t
, the in
stanta
n
e
o
u
s
flow ca
n be calcul
ate [3].
As ca
n be se
en from Fig
u
re 2, the actua
l
samplin
g time differen
c
e
has fou
r
very
distinct
cha
r
a
c
teri
stics; these a
r
e disp
ersion, di
sorde
r
, rand
o
m
ness an
d
the limited. If
use the di
re
ctly
measured time differen
c
e data to cal
c
ulatio
n the
correspon
ding
velocity and flow, the result is
boun
d to make a great e
rro
r.
Figure 2. The
Actual Time
Differen
c
e u
n
der Static
0
5
10
15
20
25
30
35
-8
-6
-4
-
2
0
2
4
6
8
Sa
m
p
le nu
mber
t
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Signal Pre-pro
c
e
ssi
ng Al
gorithm
Applied for
Ultraso
n
ic Flo
w
-m
eter (Ran
g-di
ng
Wang
)
5295
Time differen
c
e m
e
a
s
ured
is n
o
t a
con
s
tant
value,
namely, ther
e are ma
ny reason
s,
mainly from
the followin
g
aspe
cts. 1
)
Ultra
s
oni
c tra
n
sd
ucers. 2) The wo
rki
n
g con
d
ition
s
. 3)
Secon
d
a
r
y in
strum
ent a
n
d
so
on. Exce
pt the ha
rd
ware
proce
s
sin
g
, we
sh
ould
use rea
s
on
a
b
le
softwa
r
e alg
o
r
ithm to impro
v
e the accura
cy of flowmet
e
r.
3. The Desig
n
of Digital P
r
oces
sing Al
gorithm
There are m
any pro
c
e
s
si
ng metho
d
s
for
mea
s
u
r
e
m
ent data,
based on th
e sig
nal
cha
r
a
c
teri
stics of the time difference u
l
traso
n
ic
flo
w
meter, de
sig
ned a more pra
c
tical di
gital
sign
al p
r
e
-
proce
s
sing
alg
o
rithm
s
. Th
at the te
st
data
ca
n b
e
so
rte
d
, filtered,
error
com
pen
sat
ed.
Thro
ugh the
s
e three p
r
o
c
e
ssi
ng sta
g
e
s, we
ca
n get an ideal
measure
m
e
n
t data for flow
measurement
.
3.1. Sorting Algorithm
There i
s
a
l
o
t of
sortin
g
algo
rithm
s
with th
e time effic
i
enc
y
is very different. More
comm
on so
rt
ing algo
rithm
s
are b
ubble
sort, inse
rtion so
rt, shell
sort, heap
sort, qui
ck
sort,
sele
ction
sort
, merge sort, and so on. Assumin
g
n
is the numbe
r of data sh
ould b
e
sorte
d
, then
:
))
(
(
)
(
n
g
O
n
f
(7)
)
(
n
f
is the times o
f
exchang
e,
))
(
(
n
g
O
is the complexity, while
)
(
n
g
is
a function of
n
.
In the resea
r
ch
pro
c
e
ss of time differen
ce ult
r
a
s
oni
c flowm
e
ter, co
nsid
e
r
ing the
system'
s
p
o
wer a
nd me
asu
r
eme
n
t accu
racy, t
he n
u
m
ber
of sa
mple
s is set to 10,
namely
10
n
, during
cal
c
ulating the
complexity of the so
rti
ng al
gorithm, al
so
took into a
c
cou
n
t the sa
mple
space. Can be
obtai
ned by
cal
c
ulat
ing the
compl
e
xity of the algor
ithm,
Hill
sort i
s
the lowest
averag
e
com
p
lexity, that is
)
(
25
.
1
n
O
. This
so
rtin
g meth
od i
s
particula
rly suitable fo
r
small a
n
d
medium am
o
unts of data, and very ea
sy to implement.
3.2. Digital Filtering Algo
rithm
For
the ch
ara
c
teri
stics of
si
ngle-chi
p
system,
here m
a
in research
several
digital
filtering
method
s
are
as foll
ows. 1
)
Medi
a
n
filtering. 2
)
Weig
hted ave
r
ag
e
filtering
met
hod. 3
)
M
o
ving
averag
e filteri
ng method. 4
)
Peters filtering method
[4-6]. Here
we us
e the Peters formula
1
/
2
1
n
n
x
x
s
n
i
i
(8)
In this
formula,
s
is the sa
mple stan
da
rd deviation,
i
x
is the NO.i sa
mpled data,
x
is
the average
of mea
s
u
r
e
m
ents
n
times, the l
a
st
measurement
, and
n
is th
e num
be
r of
measurement
s, if there is:
s
x
x
new
3
(9)
The re
sult
s of this mea
s
urem
ent is e
ffective, otherwi
se, mea
s
u
r
eme
n
t results is the
rand
om e
rro
r,
shoul
d be
re
moved. In this form
ula,
new
x
the last mea
s
u
r
eme
n
t, if the numbe
r of
measurement
s
10
n
, then (9) sh
ould be rewrit
ten as:
s
x
x
new
2
(10)
If we o
n
ly u
s
e on
e of th
ese filtering
met
hod
s to
deal
with the
a
c
tu
al mea
s
u
r
em
ent data,
it is difficult to
meet the syst
em requi
rem
ents. The
r
efo
r
e, in th
is re
search, we ha
ve used medi
an
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 529
3 – 5298
5296
filtering meth
od, the weig
h
t
ed averag
e filter, mo
ving a
v
erage filteri
n
g method an
d
Peters filtering
to form
a
co
mposite
filter,
it can
en
sure
the
stability
and
accu
ra
cy
whil
e in
crea
sing the
real-ti
m
e
data pro
c
e
s
si
ng, thus in
cre
a
sin
g
the use
f
ulness of the
test system
[7
-
1
0
]
.
3.3. Error Co
mpensa
tion Metho
d
Erro
r com
p
e
n
satio
n
can
be achieved
th
rough a v
a
riety of ways, depe
nding
on the
prin
ciple
ca
n be
divide
d into a
r
ith
m
etic ave
r
a
ge meth
od,
wei
ghted
averag
e met
hod,
interpol
ation
method an
d so on [10].
This article took care of the stability and
real
-time requirement
s of measurement results
whi
c
h the
sy
stem requi
re
d into con
s
id
eration,
use
d
the wei
ghte
d
avera
ge m
e
thod fo
r error
comp
en
satio
n
, at the
sa
me time,
ea
ch
sa
mple
value
of
weig
ht co
efficient
ca
n
be
adj
usted
according to
conditions to
get a different error
com
p
ensation value, improved the flexibility of the
system. After the erro
r compen
satio
n
the m
easu
r
ed data ha
s been fully able to meet
th
e
measurement
requireme
nts of flow meter,
and th
en obtaine
d
final fluid flow thro
ugh
the
measurement
formula.
4. Data Pro
c
essing Algo
r
i
thms
In this
pap
er, microcontro
ller MSP4
30
F449 i
s
th
e
core of
the t
r
an
sit-time
ul
traso
n
ic
flowmete
r system in the hard
w
a
r
e, und
er IAR Embe
dded Workb
e
n
ch envi
r
onm
ent, using C4
30
in the
softwa
r
e to
achieve
the d
a
ta
sorting, f
iltering,
and
erro
r
compen
satio
n
algo
rithm. T
h
e
referen
c
e cod
e
s are as foll
ows.
4.1. Hill Sort Algorithm
void
shell
s
o
r
t( sign
ed lon
g
*shell
D
ata[1
0
], int CONT )
{ int k
=
CONT/2;
//CONT i
s
sa
m
p
le space, and CONT
=1
0;
while(k>0)
{ for(int i=k
;
i<CONT;i++
)
{ signed lo
n
g
t=shell
D
at
a[i];
int j=i-k;
while( j>=0 && t<shellData[j] )
{ shellData[j+k] =
shell
D
at
a[j];
j=j-k;
}
shell
D
ata[j
+
k]
= t;
}
k=
k/
2;
}
}
4.2. Peters F
iltering
cha
r
Peters(signed lo
ng *p
eter[9])
{ signe
d long
sum
=
0;
for(int i=1;i<9;
i++)
sum
+
=ab
s
Fu
ction(pete
r
[i]-peter[0]);
sum
= sum
>
>
1
;
if(sum
>a
bsF
u
ction(pete
r
[8]-pete
r
[0]))
return 1;
else
return 0;
}
4.3. The Error Corre
ction
Algorithm
sign
ed long
JQYes(si
gne
d long *Ye
s
[9])
{ signe
d long
JQY;
Yes
[
J
][8] =
Yes
[
8]<<
2;
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Signal Pre-pro
c
e
ssi
ng Al
gorithm
Applied for
Ultraso
n
ic Flo
w
-m
eter (Ran
g-di
ng
Wang
)
5297
J
Q
Y
=
(Yes
[4]
+
Yes
[
J
][5]+
Y
e
s
[
6]+
Y
es
[7]
+
Yes
[
8])>>
3
;
Yes[8] = Yes[
8]>>2;
return J
Q
Y;
}
sign
ed long
JQOk(signe
d long *O
k[9])
{ signed lo
n
g
JQO;
JQO
=
(O
k[
5]
+
O
k[
6]
+
O
k[
7]
+
O
k[
8]
)
>
>2;
return J
Q
O;
}
sign
ed long
JQNo
(si
gne
d long *No[9])
{ signed lo
n
g
JQ
N;
JQ
N=(No[5]+No[6]+N
o[6]
+No[7])>>2;
return J
Q
N;
}
5. Analy
s
is o
f
Experimen
t
al Results
The test pla
tform is 150
mm diamete
r
time difference ultra
s
o
n
ic flowm
e
te
r while
microcontroll
er MSP430
F
449 is a
s
t
he MCU.
In accordan
ce
with nation
a
l metrolo
g
ical
verification,
we have g
o
t seven flow
poi
nts’ a
c
tual
ca
librat
i
on re
sul
t
s,
1m³/
h, 1.3
m
³/ h, 1.6m
³/h,
75.48m
³/h, 176.12m
³/h, 250m³/h,
26
5
m
³/h, and show the a
n
a
ly
sis di
agra
m
of 1.3m³/h,
75.48m
³/h, 176.12m
³/h flow point
s.
Figure 3. Und
e
r Static
Figure 4. Und
e
r 1.3m
³/h
Figure 5. Und
e
r 75.48
m³/h
Figure 6. Und
e
r 176.1
2
m³/
h
The
stand
ard
deviation
of t
he time
difference fo
r ea
ch
flow
point i
s
in Tabl
e 1,
in
cludi
ng
both before a
nd after the di
gital pro
c
e
ssi
ng.
1
2
3
4
5
6
7
8
9
10
2.6
2.7
2.8
2.9
3
3.1
3.2
x 1
0
4
t
Sam
p
le num
ber
af
t
e
r
b
e
fore
1
2
3
4
5
6
7
8
9
10
2.6
2.7
2.8
2.9
3
3.1
3.2
x 1
0
4
t
Sam
p
le num
ber
af
t
e
r
b
e
fore
1
2
3
4
5
6
7
8
9
10
600
650
700
750
800
850
900
Sam
p
le num
ber
b
e
fore
af
t
e
r
t
1
2
3
4
5
6
7
8
9
10
-
30
-
25
-
20
-
15
-
10
-5
0
5
10
15
20
Sam
p
le num
ber
t
af
t
e
r
b
e
fore
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 529
3 – 5298
5298
Table 1. Anal
ysis of the Standa
rd Devia
t
ion to the Time Differe
nce
Fl
ow
point
Original Standar
d
deviation
Ne
w
Standa
rd
deviation
static 15.63
9.64
1.3m³/h
63.46
28.72
75m³/h
1466.18
323.09
175m³/h
899.34
381.00
6. Conclusio
n
This a
r
ticle
mainly analy
z
ed a
bout th
e feature
s
of
time differe
nce d
a
ta of
the time
differen
c
e
ultrasoni
c flo
w
meter
whi
c
h
is b
a
sed o
n
MSP430F4
4
9
chi
p
, an
d variou
s type
s
of
cla
ssi
cal
data
pro
c
e
s
sing
a
l
gorithm
s
co
rresp
ondi
ng
wi
th it, desi
gne
d an
d impl
em
ented
pra
c
tical
digital si
gnal
pre
-
processin
g
algo
rithm,
proved
by
a
c
tual test resul
t
s, usin
g thi
s
method
ca
n
be
very effective in removing
error of the
sampl
ed
data,
improving th
e me
a
s
urem
ent accu
ra
cy of
flow meter, e
nhan
cin
g
the stability
and real-time of th
e test system
.
Ackn
o
w
l
e
dg
ement
This
wo
rk i
s
sup
porte
d by
Zhejian
g
sci
ence & techn
o
logy preferred p
r
oje
c
ts
o
f
China
(201
0C110
25
), Zhejian
g
province e
d
u
c
a
t
ion depa
rtme
nt key proje
c
t
of China (Z
D2009
012
).
Referen
ces
[1]
E Mansard, E Kouam
e, R. Battault. T
r
ansit time ul
traso
n
ic
flo
w
meter: vel
o
cit
y
Profil
e esti
mation.
IEEE
Internatio
na
l Ul
trasonics Sy
mposi
u
m
. 200
5: 763-
766.
[2]
Contro
lotron
’s
w
i
de
be
am. Ca
vit
y
-free
u
l
tras
onic
flo
w
m
e
ter
s
ach
i
eve
g
a
s
custod
y tr
ans
fer accur
a
c
y
performa
nce.
Pipe
lin
e an
d Gas Journ
a
l
. 20
03; 230(
4): 93-
95.
[3]
Willatzen
M.
Fl
o
w
acoustics
mode
lin
g and
i
m
plic
ati
ons
for
ultras
onic
flo
w
meas
ureme
n
t
base
d
on t
h
e
transit-time me
thod.
Ultrasonics
. 2004; 41: 8
05-8
10.
[4]
Nori
yuk
i
F
u
ru
ic
hi, Hir
oshi S
a
t
o
, Yoshi
y
a
T
e
rao. A n
e
w
ca
li
bratio
n facil
i
t
y
for
Water flo
w
rate at high
Re
yn
olds n
u
m
ber.
F
l
ow
Measure
m
e
n
t and I
n
strumentati
o
n
. 2009; 20: 38-
47.
[5]
Yang
Hu, W
a
n
g
Ch
en
g. Sim
u
lati
on
Rese
ar
ch o
n
Ultras
o
n
i
c Guid
ed
Waves Det
e
ction
o
f
Metal R
o
d
Buffer S
y
stem
Bon
d
in
g Qu
al
it
y
.
T
E
LKOMN
I
KA Indo
nesi
a
n Jo
urna
l
of E
l
ectrical
En
gin
eeri
n
g
. 20
13
;
11(3): 12
51-
12
57.
[6]
MW Li, GS Lv
, YG Hu.
Research o
n
i
m
pr
ovin
g the accu
racy of
the ultrason
ic flow
-me
t
er w
i
th time
differenc
e method
. Intern
atio
nal co
nfere
n
ce
on electric
al
a
nd contro
l en
gi
neer
ing. 2
010:
256-
259.
[7]
S Kother Moh
i
de
en, S Arumuga Per
u
mal
,
M
Mohamed
Sathik. Image De-n
oisi
ng
usin
g Discrete
Wavelet transform.
Internati
o
nal J
our
nal
of
Co
mp
uter Sci
e
nce
and
Netw
o
r
k Security
. 2
0
08; 8(
1): 21
3-
216.
[8] J
Carlson.
Ultrasonic particle velocime
try in m
u
ltiphase flows
. Proceedin
g
s of the IEEE internati
o
n
a
l
ultraso
n
ics s
y
mposi
u
m. 200
2: 740-7
43.
[9]
Paul
us Insa
p
Santosa. C
o
st
and B
e
n
e
fit of In
formation
Search
usin
g
T
w
o
differe
nt Strategies
.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2010; 8(3): 1
9
6
-19
9
.
[10]
Roy
BVB S
i
m
o
rangkir, Achmad Munir. Numerica
l Design of
Ultra
-Wideband Print
e
d Antenna for
Surface P
enetr
a
ting
Ra
dar A
pplic
atio
n.
T
E
L
K
OMNIKA Indones
ian
Jour
n
a
l of
E
l
ectrica
l
Engi
ne
erin
g
.
201
1; 9(2): 341
-350.
Evaluation Warning : The document was created with Spire.PDF for Python.