TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 16, No. 1, Octobe
r 201
5, pp. 65 ~ 7
4
DOI: 10.115
9
1
/telkomni
ka.
v
16i1.849
9
65
Re
cei
v
ed
Jun
e
30, 2015; Revi
sed Aug
u
st
7, 2015; Accepted Augu
st
30, 2015
Compared w
i
th
PI,
Fuzzy-PI and PSO
-PI Controllers of
Robotic Grinding Force Servo System
Adna
n Jabb
a
r Attiy
a
*
1,2
. Yang
Weny
u
1
, Salam Wale
y
Shneen
3
1
School of Mec
han
ical Sci
enc
e and En
gi
neer
ing,
Huaz
hon
g Un
i
v
ersit
y
of Sci
e
n
c
e and T
e
chno
log
y
(HUST
)
, Luo
yu R
o
a
d
103
7, W
uhan, Chi
n
a
2
Alk
w
a
r
izmi En
gin
eeri
ng co
lle
ge/Bag
h
d
ad U
n
iversit
y
, Bag
h
dad, Iraq
3
Huazh
ong U
n
i
v
ersit
y
of Sci
e
n
c
e and T
e
chno
log
y
(HUST
)
/
Univers
i
t
y
of
T
e
chn
o
lo
g
y
,
Bagh
da
d, Iraq
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: rainma
n30
09
@
y
ah
oo.com, me
w
y
a
ng@m
a
il.hust.ed
u
.cn,
Salam
_
w
a
le
y7
3@
ya
hoo.com
A
b
st
r
a
ct
By grin
din
g
pr
ocess, w
hen
an i
ndustri
a
l r
obot is
used t
o
finis
h
a cur
v
ed surfac
e, b
o
th fee
d
mov
e
me
nt an
d contact forc
e must c
ontro
l
l
ed
at the si
milar ti
me
in
or
der that th
e g
r
indi
ng to
ol w
oul
d
mac
h
i
ne the w
o
rk-piec
e
at t
he ri
ght p
o
si
tion i
n
rig
h
t posture w
i
th r
equ
ired forc
e.
A passiv
e
w
r
ist
system is
adv
a
n
ced, i
n
this
p
aper, to co
nfor
m the
sha
pe of
the mach
ini
n
g
prop
ell
e
r
by al
tering its
postu
re
alo
ng w
i
th th
e
surface. T
h
e
prop
ortio
nal-
i
ntegra
l
(PI)
co
ntroll
er, du
e t
o
its si
mplic
ity, robustn
ess,
and
affordab
le
pric
e, is extre
m
ely
often
used
in
practica
l
a
ppl
ic
ations,
but it
is
effective for
li
near syst
e
m
s, as
w
e
ll as, th
e c
h
alle
ng
ing
task
is to fi
nd
its o
p
timal
ga
ins. If the
process
e
s
inv
o
lve
d
h
i
g
h
er or
der
an
d ti
me
delay system
s
,
m
a
ny intellig
ent controllers
were appe
ar
ed. In this paper, to
cope with
nonlinearities,
improve
the c
ontrol
l
er p
a
ra
meters
an
d at
the sa
me
ti
me
mod
e
li
ng
uncerta
inties
o
f
grind
i
n
g
mar
i
n
e
prop
ell
e
r surfa
c
e, a PI torque
cont
roll
er is
propos
ed suc
h
t
hat its opt
i
m
al
gai
ns are
deriv
ed vi
a a
mo
de
r
n
systems bas
ed
on fu
z
z
y
lo
gic
theory and p
a
r
ticle sw
arm o
p
timi
z
a
ti
on al
g
o
rith
m w
h
ich a
r
e used to sol
v
e
vario
u
s eng
in
e
e
rin
g
prob
le
ms. Grinding fo
rce is cont
roll
ed un
der F
u
zz
y
-
PI control
l
e
r
w
h
ich is bei
ng
asse
mb
led
a
n
d
co
mpare
d
w
i
t
h a PSO-PI c
ontrol
l
er to
o
b
tain
w
h
ich c
ont
roller
that
prov
ides
gri
ndi
ng
w
i
th
hig
her q
u
a
lity. T
he co
mp
ared
cont
roll
ers h
a
v
e be
en o
p
ti
mi
z
e
d
tog
e
ther w
i
th the p
a
ra
me
ters of the T
w
o
-
Phase
Hybr
id
Steppi
ng M
o
tor
.
T
he su
ggeste
d fu
zz
y
ru
le
fun
c
tion
and
PSO
alg
o
rith
m
impr
ove th
e res
p
o
n
se
of the c
ontrolled system and
searches
a
high-quality
solution impr
essively. Simulation
and com
p
arison
re
su
l
t
s a
r
e
p
r
e
s
e
n
t
ed
an
d
th
a
t
th
e
p
r
op
ose
d
con
t
ro
l
syste
m
s
a
r
e
co
pi
n
g
we
ll
wi
th
n
o
n
l
i
n
e
a
r
i
t
ie
s
and
uncerta
inties
w
h
ile find PI c
ontrol p
a
ra
met
e
r set e
ffectively, the PSO-PI controller h
a
s a better co
ntro
l
perfor
m
ance w
i
th improved st
ep r
e
sponse for robotic gr
inding forc
e
s
e
rvo
system
. These control
m
e
t
h
ods
w
a
s simul
a
ted
usin
g MAT
L
AB/SIMULINK.
Ke
y
w
ords
:
PI controller, F
u
zz
y
-
PI co
ntrol
l
er, particl
e sw
arm
optim
i
z
a
t
i
on (PSO), PSO-PI controller,
force
control, gri
ndi
n
g
robot
Copy
right
©
2015 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
Grindi
ng is a
mech
anical pro
c
e
ss inte
n
ded to rem
o
ve a very thin and even l
a
yer of
material
s on
the outer e
dge of the
workpi
ece.
Field practi
ce
previou
s
ly demon
strated
the
difficulty in g
r
inding
compl
e
x sh
ape
d
workpie
c
e li
ke
a ma
rine
p
r
o
peller o
r
tu
rbi
ne bl
ade
whi
c
h
deman
ds the
end-effecto
r
of the manipu
lator to
kee
p
a stable
cont
act force with
the environm
ent
while the gri
nding tool m
o
ves alon
g the profile
of
the workpie
c
e. To perfo
rm su
ch a task,
con
c
u
r
rently, both force an
d positio
n of the ro
bot sh
o
u
ld be
control
l
ed. The
s
e co
ntrol st
rategie
s
have b
een
classified i
n
to
two
main
systems, im
p
edan
ce
control an
d hyb
r
i
d
po
sition/fo
rce
control. Th
ese meth
od
s d
e
mand
a
n
a
c
curate
dyn
a
m
ic m
odel
of
the m
anip
u
l
a
tor
and
of t
he
conta
c
t force
intera
ction. In
a hybri
d
po
si
tion/fo
rce
con
t
rol sch
e
me o
n
ly path no
rmal directio
n
is
subj
ect to
a
con
s
tant
pressure
or force
co
ntrol.
The
r
e a
r
e
seve
ral
advanta
g
e
s
of this solutio
n
.
Firstly, the
d
i
mensi
onal
variation
du
e
to the
tool wea
r
i
s
com
pen
sated aut
omatically,
a
nd
se
con
d
ly, the nece
s
sity of accura
cy on
the
program
med path is
relaxed
sin
c
e
the assured
conta
c
t from the force co
ntrol loop will re
comp
en
se th
e prog
ram e
r
ror [1].
In the PI controlle
r there
are two pa
ramete
rs: p
r
oportio
nal co
efficient and
integral
coeffici
ent. the PI controll
e
r
ca
n present
indivi
duali
z
e
d
co
ntrol n
e
cessities
by tuning the
s
e two
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 1, Octobe
r 2015 : 65 – 7
4
66
parameters.
Many intelligence algorithms are
suggested to tuni
ng th
e PI param
e
ters. the
plurality pa
rt of control sy
stems u
s
e
reg
u
lar PI c
ontro
l algorith
m
s
with fixed con
s
traint valu
es set
throug
h the
empo
we
ring.
Thi
s
i
s
g
u
id
e to the
ea
ses
of de
sig
n
and
lo
w
co
st but it
rem
a
ins
uncertain
gui
de to m
o
re
compl
e
xity that is m
a
the
m
atical. T
h
e
com
b
inatio
n
of P-I controller
para
m
eters d
epen
d on type of method
s and well
-kno
wn de
sig
n
m
e
thod
s ordin
a
rily dema
n
d
s
a
mathemati
c
al
model th
at can de
scrib
e
t
he control obj
ect dyna
mical perfo
rma
n
ce
pre
c
i
s
ely [2]. In
addition, the
r
e are n
u
me
rous linea
r m
e
thod
s u
s
ed
in the d
e
si
gn
of PI stabl
e
s
d
e
sig
nated
as
con
s
e
r
vative controlle
r any
way, re
sea
r
ch is a di
stin
g
u
ish
ed deal
s with ca
rrie
d
out in the design
of unco
mmo
n cont
rolle
rs
usin
g ne
w computatio
n
a
l
techniq
u
e
s
su
ch a
s
ne
ural netwo
rks
and
fuzzy logi
c [3
]. Dependi
ng
on the ch
ang
e of control
o
b
ject
s pa
ram
e
ters, the p
e
rforman
c
e of t
h
e
system will
chang
e whe
n
this controll
er
is applie
d to a nonline
a
r
control sy
stem
[4]. Fuzzy logic
is wid
e
ly use
d
in pro
c
e
sses where sy
stem dynam
ics is eithe
r
ve
ry compl
e
x or demo
n
st
rat
e
a
highly no
nlin
ear
ch
ara
c
te
r. In orde
r to
get expe
ctan
t control effe
ct, fuzzy co
ntrol rule
s de
si
gn
sho
u
ld ta
ke
full statem
ent
of the
control
spe
c
ialty
whi
c
h
i
s
different
from co
nvent
ional co
ntrol
[
5
].
Hsi
eh
et al.
pre
s
ent
an
o
p
timal p
r
edi
cted fuzz
y-PI gain sche
duli
ng cont
rolle
r
to control
t
he
con
s
tant turni
ng force pro
c
ess with a fixed metal rem
o
val rate und
er variou
s cu
tting conditio
n
s
[6]. On the
other
hand,
for the
same
magnitud
e
of negative
and p
o
sitive
referen
c
e in
put
cha
nge
s, the
execution of
a linear con
t
rol law will
cau
s
e differe
nt resp
on
se
s of a nonline
a
r
system. Many design st
rategies
will
developed to defeat the
disadvantages of linear
P-I
controlle
rs. S
u
ch
meth
od
s create
d
fo
r
obtainin
g
a
g
oal tran
sform
a lin
ea
r P-I
co
ntroll
er i
n
to
unconventio
n
a
l PI cont
roll
ers [7]. To
defeat the
s
e
dif
f
i
cult
ies,
sev
e
r
a
l t
y
pe
s of
mo
dif
i
e
d
PI
controlle
rs
su
ch a
s
ad
apti
v
e PI controll
ers
and a
u
to
tuning were
advanced la
tely [8]. Also, a
cla
ss of non
conve
n
tional
type of PI contro
lle
r appl
ying fuzzy lo
gic ha
s bee
n
design
ed a
n
d
simulate
d for
this pu
rpo
s
e [
9
]. The Particle Swarm Opt
i
mization, i
s
a
n
addition
al p
opula
r
optimal
algorith
m
propo
sed by
Kennedy
an
d Eberh
a
rt [10] in 1995.
It is an algorithm for
swarm
intelligen
ce b
a
se
d on pop
ulation-ba
sed
adaptive
op
timization, swarm theory
and stocha
stic
based o
n
the
simulatio
n
of
animal
so
cia
l
behavio
rs li
ke fish swa
r
ms a
nd bi
rd f
l
ocks.
Comp
a
r
ed
with other
method
s su
ch as ge
netic comp
ut
ation
,
machine l
e
arnin
g
, and
neural netwo
rk
learni
ng, it furnish
e
s bette
r perfo
rman
ce
in co
mp
uting accuracy,
co
mput
ing spee
d, and memo
ry
size in spite o
f
the fact that
the origin
al PSO is
very si
mple with onl
y a few para
m
eters to modify.
Each
parame
t
er in PSO e
x
tremely affects the
pe
rfo
r
man
c
e
of PSO. Although
, it is yet to be
found h
o
w to
determi
ne
suitable valu
e
s
of p
a
ram
e
t
e
rs in PSO t
hat ca
n be
consi
dered a
s
level
optimizatio
n [11,12]. The PSO techniq
ue ha
s a ste
ady meeting
cha
r
a
c
teri
stic as com
pare
d
to
other
stochastic method
s a
nd it
can a
c
hi
eve a high qu
ality solution
durin
g a sh
ort
e
r computatio
n
time [13]. It h
a
s va
rio
u
s im
plementatio
n
s
in
e
ngine
ering field
s
. In
the PI control
l
er
de
sign, th
e
PSO algorith
m
is applie
d to sea
r
ch a
bes
t
PI c
ontrol parameters
[14].
In this p
ape
r, the PI co
ntro
ller h
a
s
bee
n
prop
osed first
,
then tunin
g
i
t
by fuzzy l
o
g
i
c an
d
a particl
e swarm optimi
z
at
ion (PSO) alg
o
rithm to
improve the cont
rolle
r parame
t
ers. The Fo
rce
Control Algori
t
hms are de
scrib
ed in Se
ction 2.
The propo
sed fu
zzy
logic an
d PSO algorith
m
a
r
e
deline
a
ted in
Section
3.
MATLAB sim
u
lation
re
su
lts an
d
some
comp
are results a
r
e
sho
w
n i
n
Section 4. Fin
a
lly, concl
u
si
ons a
r
e ma
de
in Section 5.
2. Force Co
n
t
rol Algorith
m
s
As
sho
w
n
i
n
Fi
gure
1,
the
rob
o
t
e
nd-e
ffe
ctor o
r
tool of
an
ind
u
strial
ma
nip
u
lator in
co
ntact
with a wo
rkpi
ece.
Figure 1. Main para
m
eters used in g
r
ind
i
ng poli
c
ies
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Com
pared wit
h
PI
,
Fuzzy_P
I
& PS
O_PI Cont
rollers
of
Robot
i
c…
(
A
d
n
a
n
J
a
b
b
a
r
A
t
t
i
y
a
)
67
The obj
ective
is to make the tool softly into
conta
c
t
with the workpiece, apply
a conta
c
t
force, which is well kept while the tip of
the tool
follows the p
r
ofile
of the part to be grind
ed. The
ordin
a
ry
meth
od to
satisfy the a
bove
obj
ective i
s
to
m
a
ke
the
tool
a
l
ways no
rmal
to profile of
th
e
workpi
ece an
d to control th
e orientatio
n of the end
-eff
ec
tor. At the s
a
me time this
demands
onl
y
the com
p
o
s
ite force in Fn
and Ft dire
ct
ion to be
con
t
rolled. The f
o
rce co
ntrol system stru
ct
ure
can be descri
bed as a position-base
d controller with
an outer fuzzy PI or PSO
PI force cont
rol
loop. The p
r
ofile data fo
r the ma
rine
prop
eller
can be p
r
od
u
c
ed off-li
ne
and sto
r
e
d
in
microprocessor mem
o
ry a
s
a refe
re
nce
path for th
e
robot to foll
ow. The g
r
in
ding tool
can
n
ot
kee
p
the
spe
c
ified
conta
c
t
force by mov
i
ng ba
si
cally along
th
e
p
r
o
peller profile path
on acco
unt
of the geom
etric a
nd di
splacement e
r
rors in
the
p
r
opell
e
r lo
cat
i
on and
othe
r fine erro
rs,
so
need
s to be a
n
external force control lo
o
p
to accompli
sh a practi
cal
path.
3. The
Model
of the T
w
o
-
phase
H
y
brid Stepping
Motor
The tran
sfer
function G
(
s) of the open-loop sy
stem of the two-ph
ase
Hybrid S
t
epping
Motor is
as
follows
[15]:
Gs
(
1
)
Whe
r
e,
A
s
KPis
K
Ii
KH
(
2
)
B
s
s
THs
1
Ls
R
KPis
K
Ii
KH
L
THs
RT
H
L
s
R
s
KPis
K
Ii
KH
(3)
The subdivid
ed drivin
g is
assume
d for
the Hybr
i
d
Steppin
g
Moto
r in ord
e
r to
reach to
the actu
al sy
stem p
e
rfo
r
m
ance pa
ram
e
ter an
d to
de
cre
a
se the i
n
trica
c
y of the
system t
r
an
sfer
function. In si
mulation, the
par
a
m
eters o
f
the two-ph
a
s
e Hyb
r
id
Ste
pping M
o
tor
sele
cted a
r
e
as
follows
:
In
e
r
tia
C
o
n
s
ta
n
t
J
=
25
0
k
g
· m, In
du
cta
n
c
e L=
0
.
33
H
,
Re
s
i
s
t
an
c
e
R
=
8
Ω
, ,
β
= 1,
Coeffici
ent of Viscou
s
F
r
iction B = 0 N · m ·
s/rad, Kpv = 500, KIv = 0, KPi = 5, KIi =
500,
ke = 0.2
5
N · m/A, N = 180, KH = 15 a
nd TH
= 0. The Tra
n
sfe
r
functio
n
will b
e
:
G
s
.
(
4
)
4.1. Conv
entional PI Strateg
y
for Grin
ding Force S
e
rv
o Unit
The mo
st con
v
entional PI controlle
r or lin
ear
PI cont
rol
l
er is de
scri
b
e
d as follo
ws:
Y
t
e
t
Kp
K
i
e
t
dt
(
5
)
Whe
r
e KP is the propo
rtional consta
nt
gain and KI
is the integral con
s
tant g
a
in acco
rdin
g
to
manual
expe
rtise. T
he
sig
nal e
(
t) i
s
th
e er
ro
r
signa
l betwe
en th
e refe
ren
c
e
and the
p
r
ocess
output c(t) it is explain
ed a
s
: e(t) =
r(t)
−
c
(
t).
Table 1. The
effect of Kp and Ki to the controlle
d syst
em
Parameter
Rise
ti
me
O
v
ersh
oo
t
Turni
ng
time
Error
Kp
decrease increase
Small
change
decrease
Ki
decrease
increase increase eliminte
Figure 4. Simulink a
nd Blo
ck di
agram of
PI controller
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 1, Octobe
r 2015 : 65 – 7
4
68
And the PI controlle
r inte
rnal st
ru
cture
is explai
ned i
n
Figu
re 5, th
e input p
a
ra
meters for th
e PI
controlle
r are
Kp, Ki,and th
e output for the co
ntrolle
r i
s
u.
Figure 5. The
stru
cture fo
r PI controlle
r
4.2. Fuzzy
-PI controller fo
r Grinding Force Serv
o
Unit
The
Rule
ba
se i
s
form
ed
based o
n
th
e
followi
ng concepts.
T
h
e
fuzzy rule
s and
the
rang
es of inp
u
t membe
r
shi
p
functio
n
s are ente
r
ed f
r
o
m
the obtai
na
ble data. T
h
e
followin
g
ste
p
s
can rep
r
e
s
ent
the pro
c
edu
re of the sugg
ested fu
zzy lo
gic:
Step 1: Initialize (FIS editor) the input of
fu
zzy logi
c
co
ntrolle
r erro
r (e),
chan
ge o
f
erro
r
(ec) an
d outp
u
t (KP) and (KI), see Figu
re 6
Figure 6. FIS
Editor of Fuzzy-PI cont
roll
er
Step 2: Set the sy
stem wi
th tow fuzzy logi
c
cont
rolle
r (KP,KI) ea
c
h
co
ntrolle
r h
a
s tw
o
inputs, (e,ec) and the o
u
tp
ut are (KP,KI), e input
h
a
s seven fu
zzy
set a
s
soci
ate
d
with it, whi
c
h
sorte
d
as n
e
gative large
(Nl
)
, negativ
e medium
(NM),n
egative
small (NS),
Zero e
rro
r (Z),
positive sm
all
(PS), positive medium (P
M) and
p
o
siti
ve large (Pl
)
, see Fig
u
re 7, 8.
Figure 7. Membershi
p
error functio
n
(e
)
Figure 8. membershi
p
ch
a
nge erro
r fun
c
tion
(ec
)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Com
pared wit
h
PI
,
Fuzzy_P
I
& PS
O_PI Cont
rollers
of
Robot
i
c…
(
A
d
n
a
n
J
a
b
b
a
r
A
t
t
i
y
a
)
69
Step 3: Set the input ra
ng
e from -3 to 3,
where the
input e and ec are sh
own in the
followin
g
equ
ations: e
=
{
N
B(-4,-3,-2
), NM(-3,
-
2,-1),
NS(-
2,-1,0
),Z
(-1,0,1
), PS(0,1,2), PM(1,
2
,3],
Pl(2,3,4)}, ec=
{
[N(-2,-1,0),
Z(-1,0,1), PB(0,1,2)}.
Step 4: set th
e output rang
e from 0 to 6.
Fi
gure 9 sho
w
s th
e memb
ership fun
c
tio
n
of the
output variabl
e KP.
Figure 9. Membershi
p
Kp function
Figure 10. Membe
r
ship Ki function
The su
rfa
c
e viewe
r
is sho
w
n in Figure 11
, 12.
Figure 11. Surface vie
w
of Kp of fuzzy
c
ontroller
Figure 12. Surface vie
w
of Ki of fuzzy
c
ontroller
Figure 13. Ru
le bases for F
u
zzy cont
rol
sy
st
em
Figure 14. Ru
le bases vie
w
for Fuzzy co
ntrol
sy
st
em
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 1, Octobe
r 2015 : 65 – 7
4
70
Step 5: The
estimation p
e
rform
a
n
c
e
measur
es a
r
e Rise time
(RT
)
, pea
k oversh
oot
(OV), and
Am
plitude (AM).
Figure
shows
a
n
exampl
e
of a
de
si
red
time respon
se
to imp
r
ove
th
e
respon
se of
the system
by c
hang
e
the controll
er pa
ram
e
ters until we
reach the op
timal
respon
se of the syste
m
.
4.3. PSO Algorithm for G
r
inding Forc
e Serv
o Unit
In a PSO system, particl
es
fly approxima
t
ely in
a multidimen
sion
al search
spa
c
e
with an
adapta
b
le vel
o
city whi
c
h m
odified dyna
mically a
c
cord
ing to its o
w
n and the oth
e
r pa
rticle
s fl
ying
experie
nce. Duri
ng flight, each p
a
rti
c
l
e
modifie
s
its po
sition a
c
cording to its own expe
rie
n
ce
giving the be
st previo
us p
o
sition
(the
minimum fitn
ess value
)
is called p
e
rso
nal be
st (P b
e
st)
whi
c
h
kee
p
s
path of its co
ordin
a
tes a
s
sociate
d
with
the be
st
soluti
on that o
b
tai
n
ed
so fa
r in
the
probl
em sp
ace, P best is reco
rde
d
and
deline
a
ted
as PI=(pi1,pi2,
…,pid), whil
e
the index of
the
best p
a
rti
c
le
among
all th
e neig
hbo
rin
g
pa
rticle
s in
the po
pulati
on is called t
he glo
bal b
e
s
t (G
best), it is deli
neated by the
symbol g an
d it
is the best solution in the wh
ole gro
up [16].
Figure 15. Particle
swarm
optimization
(PSO)
strategy
Figure 16. Th
e stru
cture of Particle
swarm
optimization (PSO) strategy
For a PI- co
ntrolled
syste
m
, there are
often
four indices to ch
ara
c
teri
ze th
e system
perfo
rman
ce:
ISE, IAE, IT
AE and
ITSE. In this paper
we sele
ct ITSE which i
s
de
fined as:
ITSE
te
∞
t
dt
(
6
)
Figure 8. The
flowch
art of the PSO-PI co
ntrol sy
stem
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Com
pared wit
h
PI
,
Fuzzy_P
I
& PS
O_PI Cont
rollers
of
Robot
i
c…
(
A
d
n
a
n
J
a
b
b
a
r
A
t
t
i
y
a
)
71
The modifie
d
PSO framework i
s
de
scrib
ed as follo
ws:
1. Generate the initial velocity and po
sition fo
r ea
ch p
a
rt ran
domly deline
a
ted on
e input-
output
syste
m
controll
er i
n
pa
rticl
e
a
ccordin
g to
its lowe
r
and
up
p
e
r val
u
e
s
of
e
a
ch
PI contro
ller
parameters
in s
y
s
t
em to form parents
[17].
2. For ea
ch p
a
rt in each pa
rticle in the sw
arm, estima
te P best (the
personal fitn
ess.
3. To find the
best PI controllers re
pre
s
e
n
ted
by the b
e
st pa
rts in
th
e swarm, eva
l
uate G
best (th
e
glob
al best fitness) for all varied
parts fro
m
all
particle
s
in the swarm.
4. Determi
n
e
G best of all
particle
s
in t
he pop
ulation
by adding al
l G best for a
ll parts
from ea
ch pa
rticle in swa
r
m.
5. Update e
a
c
h pa
rticle to
form progeny
.
6. Compa
r
e
P best (the p
e
rsonal fitne
s
s) of ea
ch pa
rt of progeny
with their simi
lar part
s
in pare
n
ts an
d cho
o
se the best on
es to fo
rm ne
w offspring to the n
e
xt generatio
n.
7. Determine
G best
(the
global fitne
s
s) of the varie
d
part
s
in the
populatio
n a
s
to the
new offsprin
g
and addi
ng to be the be
st one
s for the n
e
xt generatio
n.
8. Stop if the
stoppi
ng sta
n
dard i
s
sati
sf
ied otherwi
se,
go to step 5. See Figure 8
Each p
a
rticl
e
is treate
d
as a p
o
int i
n
a D-dime
n
s
ion
a
l sp
ace
.
The ith pa
rticle i
s
rep
r
e
s
ente
d
as xi
=
(xi1,xi2,…,xid). Th
e velo
ci
ty
of each
a
gent whi
c
h gra
dua
lly
gets clo
s
e
to
Pbest an
d G
best i
s
re
pre
s
ente
d
a
s
vi=(vi1,vi2,
…,vid), it can
be
adju
s
ted by
the followi
n
g
equatio
n:
v
w
.
v
c
1
.
r
a
n
d
.
p
x
c
2
.
r
a
n
d
.
p
x
(
7
)
Whe
r
e
v
is
current velocity,
v
is mo
dified
velocity, n repre
s
e
n
ts ite
r
ation,
w is
a
n
inertia
weight,
p
=
Pb
es
t,
p
= Gbest, c1 and c2
are two po
sitive consta
nts,
rand ( ) is a random
gene
rated nu
mber with a range of [0,1], and
x
is the
curre
n
t searching point. The sea
r
chin
g
point in the solution spa
c
e
or the cu
rren
t pos
ition can
be adju
s
ted
by the followi
ng equ
ation:
x
x
v
(
8
)
Figure 18. Steps of PSO-P
I controlle
r
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 1, Octobe
r 2015 : 65 – 7
4
72
As recomme
nded
in PS
O, the
co
nst
ants
ar
e
c1
=c2=0.8, Eq
uation
(7
) i
s
used
to
cal
c
ulate
ne
w velocity of the parti
cle a
ccordin
g to
its
p
r
eviou
s
velo
ci
ty and its p
r
e
v
alent po
sition
dist
an
ce
s
of
f
r
om it
s o
w
n
b
e
st
p
o
sit
i
o
n
(
e
x
per
ie
nce) a
nd the
g
r
ou
p’
s b
e
st
expe
ri
ence. Th
en t
h
e
particl
e flies i
n
the dire
ctio
n of a new po
sition a
c
cordi
ng to Equatio
n (8).
Inertia wei
ght
, w in the Equation (7
) is t
o
make bala
n
ce b
e
twe
en
the local
sea
r
ch an
d
global
sea
r
ch ca
pability. It can be a
nonlin
ear fu
n
c
ti
on of time
or po
sitive
con
s
tant o
r
even
positive linea
r [18]. In this p
aper
we set w=0.3.
5. Results a
nd discussio
n
To verify th
e
efficien
cy
o
f
the p
r
op
osed fu
zzy
rul
e
meth
od
a
nd a
pa
rticl
e
swa
r
m
optimizatio
n (PSO) algo
rith
m. With the
MATLAB
SIMULINK, a fam
ous
simul
a
tio
n
software, a
nd
for a co
nventional PI, Fuzzy_PI, and PSO-
PI cont
rolle
r as
sho
w
n in
Figure 1
9
.
Figure 19. Simulink figu
re
of PI, Fuzzy_PI, and PSO-PI controlle
r
Step respon
ses is
sho
w
n i
n
Figure 20, 21, 22, and 2
3
with Kp= 20
and Ki = 40
Figure 20. Step re
spo
n
se of the system
under
PI controlle
r
Figure 21. Step re
spo
n
se of the system
under
PI, Fuz
z
y
-PI c
ontroller
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Com
pared wit
h
PI
,
Fuzzy_P
I
& PS
O_PI Cont
rollers
of
Robot
i
c…
(
A
d
n
a
n
J
a
b
b
a
r
A
t
t
i
y
a
)
73
Figure 22. Step re
spo
n
se of the sy
stem
under PI, PSO-PI cont
rolle
r
First, for the
step respon
se of the syst
em
unde
r PI controlle
r sh
o
w
n in Fig
u
re
20, we
note that the Rise time va
lue = 0.01
6, the ov
ersho
o
t value = 20%
, and the ste
a
dy state error
also
we note
that there is
no
an un
dershoot value while wh
en we
submit Fu
zzy-PI strategy we
rema
rk th
at the rise time value (d
ecre
ase) =
0.0
1
, the oversh
oot va
lue (de
c
rea
s
e
)
= 10% whe
r
e
the system ta
ke
s sh
ort time to rea
c
h th
e
steady state
as sho
w
ing in
Figure 2
1
.
The system respon
se
of PI
controller
tuni
ng
usi
n
g Parti
c
le S
w
arm O
p
timization
is
shown in Figure 22 show
that the
system will
reach t
he
stability quickly than the sy
stem
under
PI & Fuzzy-PI controll
er a
n
d
the pea
k overshoot (d
e
c
rease) = 7%
while the Ri
se time = 0.013
so
that the syste
m
got good resp
on
se. See
Table 2 and
Figure 23.
Table 2. Step
resp
on
se pe
rforman
c
e
for
PI,Fuzzy
-PI & PSO-PI controlle
rs
Control Metho
d
Overshoot
(%)
Rise Time
(s
)
Stead
y
State E
r
ror
PI
20
0.016
0.083
Fuzz
y
_
PI
10
0.01
0.035
PSO-PI (I
TAE)
7
0.013
0.027
Figure 23. Step re
spo
n
se of the system
under PI, Fu
zzy_PI, and
PSO-PI controller
6. Conclusio
n
In this pa
per
resea
r
ch works
are ta
ke
n
for gri
nding f
o
rce controlling. Wh
en a
grindi
ng
whe
e
l grin
ds
prop
elle
rs a
s
a free-fo
rm surfa
c
e by
a
robot, the gri
n
ding force,
at the mentio
ned
machi
n
ing po
int, must be controlle
d in the normal di
re
ction in ord
e
r that both grinding force a
n
d
feed movem
ent coul
d be
controll
ed. T
he modeli
ng,
control a
nd
simulatio
n
of the Two-Ph
ase
Hybrid Step
pi
ng Motor
hav
e been
done
using th
e so
ftware p
a
cka
g
e MATLAB/SIMULINK. T
he
desi
gn ha
s b
een remarke
d
that the system re
sp
on
se is improved
by setting the paramete
r
s of
the tran
sfer f
unctio
n
until
we obtai
n an
optimal
re
sp
onse of the
system afte
r perfo
rming th
e
value of PI con
s
tants
achi
eved fro
m
the fuzzy
rule metho
d
and PSO. The sub
m
itted
methodol
ogy
gives better performan
ce
in the pea
k
overshot, rise time, and the stea
dy-state
err
o
r. The r
e
spo
n
s
e
rem
a
rke
d
from the
PSO-PI
controlle
r ha
s a small over sho
o
t, small stea
dy
state
e
r
ror,
a
nd small rise time
than
PI and ha
s
smal
l
over sh
oot, small stea
dy state
e
r
ror
th
an
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 1, Octobe
r 2015 : 65 – 7
4
74
Fuz
z
y
-
PI
con
t
roller.
Ho
we
v
e
r, it is
r
e
m
a
rk
ed
fr
om
the
simulatio
n
that the
con
t
roller pe
rforms
improve
d
in
Fuzzy-PI (th
e
over shoot
, rise
time, settling and
steady state
better
than
PI
controlle
r.
Ackn
o
w
l
e
dg
ements
This
work wa
s pa
rtially su
pporte
d by the Nation
al Basi
c Re
se
arch Prog
ram of
China
(201
4CB
046
704).
Referen
ces
[1]
Jianj
un W
a
n
g
, Hui Z
han
g, T
homas A F
u
hlbri
g
g
e
.
F
o
rce Contro
l T
e
chno
log
i
es for New
Robot
i
c
Appl
icatio
ns.
IEEE Internatio
nal C
onfere
n
ce
on T
e
chnolo
g
i
e
s for Practical
Robot App
lic
ations. 20
08.
[2]
W
R
H
w
a
ng, W
E
T
hompson.
Desig
n
of i
n
tell
ige
n
t fu
zz
y
log
i
c co
ntroll
ers
u
s
ing
ge
netic
al
gorith
m
s
. I
n
Proceedings of
the 3rd IEEE
Confer
ence on Fuzz
y
S
y
stem
s, IEEE
World Congress on Computational
Intelli
genc
e. 19
94: 138
3-1
388.
[3]
Z
u
lfatman, MF
Ra
hmat. Ap
pli
c
atio
n
of s
e
lf-tuni
ng f
u
zz
y PI
D co
ntroll
er
on
in
dustria
l h
y
dr
aulic
actu
ato
r
usin
g s
y
stem i
dentif
ic
atio
n ap
proac
h.
Interna
t
iona
l Jour
na
l
on S
m
art Se
ns
ing
and I
n
tell
ig
ent Syste
m
s.
200
9; 2(2).
[4]
Seema C
hopr
a, R Mitra, Vija
y
Kum
a
r. Auto tuni
ng of fu
zz
y
PI type c
ontrol
l
er usi
n
g
fuzz
y
log
i
c.
Internatio
na
l jo
urna
l of compu
t
ationa
l cog
n
iti
on.
200
8; 6(1).
[5]
Ma Qiuj
ie, Sh
i
Jingz
hu
o. F
u
z
z
y
PID
Spe
e
d
Contro
l of T
w
o Ph
ase
Ultra
sonic M
o
tor.
TEL
K
OMNIKA
Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng.
2014: 6
560-
65
65.
[6]
Hsieh CH, Chou JH
, Wu YJ.
Optimal
predicted fuzzy
PI gain
sch
ed
uli
ng c
ontrol
l
er
of co
n
s
tant turni
n
g
force sy
stems
w
i
t
h
fix
ed metal removal r
a
te.
Int J Adv Manuf T
e
chnol.
20
0
2
: 714-7
21.
[7]
Jiho
ng Le
e. On
metho
d
s for
improv
in
g
perfo
rm
ance
of PI-t
ype f
u
zz
y l
ogic
co
ntroll
ers.
IEEE
Transactions On Fu
z
z
y
System
s.
19
93; 1(4).
[8]
Raja
ni K Mud
i
a, Chanc
ha
l D
e
yb, T
s
u-T
i
an Lee. An
impr
o
v
ed auto-t
uni
n
g
scheme for
PI controllers.
ISA T
r
ansactio
n
s.
2008: 4
5
-5
2.
[9]
W
M
T
ang, G Chen, R
D
L
u
. A modifie
d
fuzz
y
PI
con
t
roller for a fl
exibl
e
-j
oint ro
bot arm
w
i
t
h
uncerta
inties.
Int. J.
Fu
z
z
y
Sets Syst
. 2001: 109
–1
19.
[10]
J Kenn
ed
y, R
C
Eberh
a
rt.
Particle sw
arm
opti
m
i
z
at
io
n
. Proc
eedings IEEE I
n
ter
national Conferenc
e on
Neur
al Net
w
o
r
ks IV. 1995: 19
42-1
948.
[11]
Juin
g-Shi
an
C
h
io
u, Shu
n
-H
u
ng T
s
ai, Min
g
-T
ang Li
u. A
P
S
O-based
ad
a
p
tive fuzz
y PI
D-control
l
ers.
Simulati
on Mo
dell
i
n
g
Practice
and T
heory.
2
012; 26: 4
9
-59.
[12] Z
hang
Xi
n, Da
ng Jia
n
w
u
, Li
u
Min. Parameter Op
timizatio
n
of PID Controller Base
d on PSO for Multi-
leaf Co
llim
ator.
T
E
LKOMNIKA Indones
ia
n Jo
urna
l of Electri
c
al Eng
i
ne
eri
n
g.
2013: 6
127-
613
4.
[13]
Gaing ZL. A Particle S
w
arm
Op
timization Approach fo
r O
p
timum Design of PID Cont
roller
in AV
R
Sy
s
t
e
m
.
IEEE Transactions on Energy Conv
ersion.
200
4; 19: 384-3
94.
[14]
Liu Y, Z
hang J
,
W
ang S.
Opti
mi
z
a
t
i
o
n
Desi
g
n
Based o
n
PSO Algorith
m
for PID Controll
er
. 5th W
o
rl
d
Con
g
ress on In
tellig
ent Co
ntro
l and Autom
a
ti
on. 200
4; 3: 24
19-2
422.
[15]
XU
W
enq
ia
ng,
YAN Ji
anh
on
g. Deriv
a
tio
n
of
T
r
ansmissio
n
F
uncti
on M
ode
l of T
w
o-
p
hase
H
y
bri
d
Steppi
ng Motor
.
Space Electro
n
ic T
e
chn
o
lo
gy
.
2011, 3: 50-5
3
.
[16]
Lalit
ha MP,
R
edd
y
VCV,
Us
ha V. O
p
timal
DG p
l
acem
e
n
t for mi
nimu
m rea
l
p
o
w
e
r
loss
in r
a
d
i
a
l
distrib
u
tion s
y
s
t
ems using PS
O.
J
T
heor App
l
Inf T
e
chnol
. 2
010; 13(
2): 107
-116.
[17]
Sa
w
s
a
n
Gharg
hor
y
,
Ha
nan K
a
mal. Mod
i
fie
d
PSO
for Optimal T
uning
of F
u
zz
y
PID C
o
ntroll
er.
IJCSI
Internatio
na
l Journ
a
l of
Co
mputer Scie
nce I
ssues.
201
3; 1
0
(2): 462-
47
1.
[18]
Mahmu
d
I
w
a
n
Soli
hi
n, Le
e
F
ook T
a
ck, Moe
y
Lea
p K
e
a
n
.
T
uni
ng
of P
I
D Contr
o
ll
er
Using
Partic
le
Swarm
Optimiz
a
tion (PSO)
. Proceed
ing
of the Interna
t
iona
l Confer
e
n
ce on Adv
a
n
c
ed Scie
nce,
Engi
neer
in
g an
d Information T
e
chn
o
lo
g
y
. 201
1: 458-4
61.
Evaluation Warning : The document was created with Spire.PDF for Python.