TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 12, No. 9, September
2014, pp. 67
6
4
~ 677
0
DOI: 10.115
9
1
/telkomni
ka.
v
12i9.509
0
6764
Re
cei
v
ed
No
vem
ber 1
0
, 2013; Re
vi
sed
May 10, 20
14
; Accepte
d
Ju
ne 6, 2014
Multimode Speed Control Based on Fuzzy Dec
i
sion-
Making for Automatic Train Operation
Junnian Go
u
Schoo
l of Auto
mation a
nd El
e
c
trical Eng
i
ne
e
r
ing, La
nzh
ou
Jiaoto
ng U
n
ive
r
sit
y
,
Lanz
ho
u Cit
y
,
Chin
a
E-mail: ju
nni
an
@mail.lz
jtu.cn
A
b
st
r
a
ct
Based
on th
e a
nalysis
of trai
n
runn
ing
proc
es
s and tr
ain
motor-drive
n
mo
de
ls, an AT
O co
mp
ou
n
d
control a
l
gor
ithm of
multi
m
o
de fu
zz
y
deci
s
ion
maki
ng i
s
propos
ed, w
h
ich w
ill i
m
pr
o
v
e the effects of
dynamic and static control
of
ATO
system
. Control weighting
coefficients
of PID cont
roller
output and f
u
z
z
y
control
l
er o
u
tp
ut are o
b
tain
ed
by fu
zz
y
r
easo
n
in
g,
fu
zz
y
dec
ision-
maki
ng a
nd dy
na
mic c
a
l
c
ulati
ons of err
o
r
and err
o
r rate
of train spe
ed,
and als
o
the
w
e
ightin
g coef
ficients are us
ed for
the out
put calcu
l
ati
o
n
of
fu
zz
y
decis
io
n-mak
i
n
g
contr
o
l
l
er. Si
mu
latio
n
result
s sh
ow
the sp
ee
d a
nd
accuracy
of A
T
O speed c
ont
rol
are i
m
prove
d
e
ffectively by th
e co
mp
ou
nd
mu
lti
m
od
e d
e
ci
sion c
ontrol
l
er
alg
o
rith
m, w
h
ic
h he
lp to
i
m
pr
ov
e
the co
mfort of trains
a
nd its stop prec
isio
n.
Ke
y
w
ords
:
AT
O, fu
zz
y
d
e
cisi
on-
maki
ng,
mu
ltimod
e control,
speed co
ntrol
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 ultimate goal of ATO (Automatic T
r
ain
Ope
r
ation
)
system is t
o
sele
ct rea
s
onabl
e
operation m
o
de an
d ru
nni
ng tra
c
k auto
m
atically to
complete a
u
to
matic d
r
iving
task
acco
rdin
g to
line co
ndition
s and
extern
al sign
al of ra
ilways in
unm
anne
d ca
se
s.
The main fu
nction of AT
O is
to adjust the
train’s
run
n
in
g spe
ed, and
only on
the basi
s
of accu
rate speed
re
gulation
can t
he
ATO comple
te positio
nin
g
and
pa
rki
ng tasks; T
he spee
d re
gulation i
n
the p
r
emi
s
e
that
passe
nge
r feel comfo
r
t improve the effi
cien
cy of
the train ope
rati
on [1]. Theref
ore, train
spe
ed
control system’s perform
a
nce
w
ill direct
ly affect the developmen
t of the rail transit system.
Re
sea
r
che
r
s have
appli
ed
cla
ssi
cal
co
ntrol
the
o
ry, intellige
n
t co
ntrol
th
eory
and
method
s to train ATO spe
ed co
ntrol
si
nce the
early
60's. G
r
eat
prog
re
ss of the ATO sy
stem
developm
ent
has
bee
n b
r
o
ught by intro
duci
ng of th
e
s
e al
go
rithms and th
eory,
but there is
a
l
so
some l
a
ck of
improvem
en
t. For examp
l
e, PID
appli
c
ation
wa
s in
trodu
ced to
modulate trai
n’s
spe
ed a
c
cu
ra
tely by usin
g
steady-state
error
of
control sy
stem. PID
control, ho
wever there
are
also
disadva
n
tage
s of sl
o
w
respon
se,
accele
ra
tion
not ea
sy to control, and
not meeting
the
requi
rem
ents of comfort.
Fuzzy and e
x
pert co
ntrol
mainly simu
late experi
e
n
c
e fro
m
co
ntrol
experts,
whi
c
h meet
s the
requi
rem
ents for comfort
and fa
st spe
ed, but its
accuracy of
co
ntrol
can
not be
e
s
sentially
cha
nged [2,
3]. Low
sta
b
ility and
slo
w
e
r
l
earni
ng
sp
ee
d are p
r
obl
e
m
s
from neural networks
[4].
Single co
ntro
l method is
applie
d to ATO sp
eed
co
ntrol, and th
e overall effe
ct is not
particula
rly d
e
sirable. M
o
re and
mo
re
re
sea
r
che
r
s are t
r
ying t
o
apply
a variety of con
t
rol
method
s
in ATO,
whi
c
h are com
posit
e
co
ntro
l me
thods. Th
ere
f
ore, the co
mposite
co
ntrol
algorith
m
is the focus of the ATO
sp
e
ed contro
l s
y
s
t
em of the
c
u
rrent and future
res
e
arch.
Based
on
the
tren
d, PID
control
and
fuzzy control
alg
o
rithm
are
int
egrate
d
tog
e
ther to g
e
t a
n
e
w
comp
osite
m
e
thod. Th
e f
u
zzy de
cisi
o
n
-ma
k
in
g ou
tput of different stag
es i
s
a
c
hieve
d
b
y
introdu
cin
g
fuzzy de
ci
sion,
whi
c
h imp
r
ov
es cont
rol a
c
curacy an
d ro
bustn
ess of the ATO syste
m
.
2.
Train Oper
ation Process
and D
y
namic Model of the Train Ope
r
ation
2.1.
ATO Spe
e
d Con
t
rol Proc
ess
In the ca
se
of a fixed rai
l
way line, AT
O, according
to the drivin
g com
m
and
and the
curre
n
t state
of the train,
modulate
s
train’s
sp
e
ed i
n
the differe
nt operating
con
d
ition
s
, while
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Multim
ode Speed Control B
a
se
d on Fu
zzy De
ci
si
on
-M
akin
g for Autom
a
tic… (Ju
nnian G
ou)
6765
different spee
d is co
rrespo
nding to diffe
rent ope
rati
n
g
target of the cont
rol. All of the stage
s of
the train t
r
a
c
tion, coa
s
ting
, spee
d
control and
braki
n
g are un
der the complete
sup
e
rvisi
on a
nd
prote
c
tion of
the ATP(Automatic Train
Protec
tion
)
whi
c
h is
part
of ATO system. The ATO
maintain
s the
actual vehi
cl
e sp
eed to a
certai
n
value
unde
r the limi
t
from the AT
P, and on
ce t
he
train sp
eed
excee
d
s the
ATP’s spe
e
d
limit, emergen
cy bra
k
in
g will be do
ne. Due to the
exce
ssive
grading
brake
(multi-ste
ppe
d
brakin
g m
o
d
e
), the
train
will ta
ke
up
a
line
more lo
nger
time, affecting transpo
rt efficien
cy. Now one
level sp
eed adju
s
tme
n
t has re
pla
c
ed multi-step
ped
bra
k
ing
mod
e
before b
r
a
k
ing an
d pa
rki
ng in
acco
rd
ance with
est
ablished
bra
k
e cu
rve [5]. T
h
is
spe
ed
co
ntro
l mode
do
es not o
n
ly gu
arante
e
th
e
train
po
sitioni
ng p
a
rking,
but al
so
help
to
improve
the t
r
an
spo
r
t effici
ency
of the t
r
ain. Th
e
trai
n
sp
eed
contro
l and
braki
ng
mode i
s
sh
o
w
n
in Figure 1.
Figure 1. Trai
n Speed Cont
rol Ope
r
atin
g Mode
In the case of fixed section
length of the
train operation, railway line con
d
itions,
and the
runni
ng spee
d of each p
o
i
n
t on the line
,
ATO, acco
rding to the traction / braki
ng ch
ara
c
te
ri
stic
curve,
obtai
n
s
a
give
n run
n
ing
sp
eed
curve
by tra
c
ti
on
cal
c
ulatio
n. One
of th
e
main
tasks
o
f
the
ATO is to co
ntrol the train
to track
a limited run
n
ing
spee
d cu
rve
fast and sm
oothly for a h
i
gh-
spe
ed and
stable train op
eration. So the desi
gn of
ATO controll
er is ab
solut
e
ly the resea
r
ch
f
o
cu
s.
2.2.
D
y
namic Mo
del of the Tr
ain Running
Acco
rdi
ng to the literature
[6], the dyna
mic m
odel of
train run
n
ing,
such as Equ
a
tion (1)
is sh
own:
ds
v
dt
dv
C
W
dt
M
(1)
Whe
r
ein,
s is traveling
di
stance of th
e train; v is
train’
s
spe
ed; t i
s
operation tim
e
of the
trai
n;
C
is
trac
tion/brak
i
ng forc
e, and C =
F is
the trac
tion s
t
ate
,
C = -B is the
braki
ng state
;
W is the dra
g
forc
e; M is
tot
a
l mass
the train.
As the plant,
the elect
r
ic
locom
o
tives
not
only incl
ude the tract
i
on motor
co
ntrolled
dire
ctly, but
also
the train
alt
ogeth
e
r. I
f
take th
e
re
ctified voltag
e a
s
the
inp
u
t appli
ed to
the
traction m
o
to
r, the spe
ed
of the train is the out
put. Take a
se
para
t
ely excited motor to an
al
ysis
of the sp
eed
modulatio
n of
the train [7,
8] as a
exam
ple, the arma
ture
circuit of
traction
moto
r is
s
h
ow
n
in
F
i
gu
r
e
2
.
Figure 2. Arm
a
ture Ci
rcuit of Tractio
n
M
o
tor
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 9, September 20
14: 67
64 – 677
0
6766
Armature circuit voltage ba
lance equ
atio
n of the tracti
on motor i
s
b
e
low:
dd
v
v
di
L
Ri
u
E
u
C
dt
(2)
Whe
r
ein,
L a
nd R,
are
re
spe
c
tively the total in
d
u
ct
ance an
d tot
a
l re
si
stan
ce
of the a
r
mat
u
re
circuit, and i
s
the a
r
matu
re ci
rcuit cu
rrent, is the to
tal voltage a
pplied to the
motor, E is
the
cou
n
ter ele
c
tromotive force
of the motor.
Based
on th
e relatio
n
ship
betwee
n
the
train a
c
celeration and tra
c
tion of the
electri
c
locom
o
tive traction
as
wel
l
as the
relati
onshi
p bet
we
en ea
ch tract
i
on moto
r on
the train a
nd
its
curre
n
t, if acceleratio
n
re
si
stan
ce of trai
n is ig
n
o
red, can th
e total mass of the train be
conve
r
ted
to each tracti
on motor. Derivate Equation (2) a
nd Equ
a
tion (3
) ca
n be got [8].
'
2
1/
1
v
dM
i
M
C
v
uT
T
s
T
s
(3)
Whe
r
ein,
'
''
,,
,
vv
M
i
v
Fv
RM
L
CC
T
T
C
CC
R
coe
fficient of i
n
d
u
ce
d el
ect
r
o
m
otive force,
'
F
C
is
con
s
tant coef
ficient of tract
i
on motor.
Visibly, a se
con
d
-o
rd
er transfe
r fun
c
ti
on
can b
e
u
s
ed to
o de
scribe th
e train drag
system, an
d the sel
e
ctio
n of
the param
eters
can b
e
combi
ned
with the experim
ental data.
3.
The Multi-m
odal Con
t
rol
and Fuzz
y
D
ecision-ma
king Struc
t
ure
of ATO Sy
stem
Thre
e pa
rts
are in
clu
ded
in the ATO
cont
rolle
r o
f
multi-modal
system: the
fuzzy
controlle
r, PID
controller
and fu
zzy de
cisi
on
cont
ro
l
l
er. In Fi
gure
3 i
s
target
speed,
and i
s
the
actual trai
n speed.
Figure 3. ATO System Mu
lti-modal
Con
t
rol Block Dia
g
ram
3.1. Multimode
Control
The natu
r
e
of multimode
cont
rol is t
o
mi
mic stra
tegies of
co
ntrol expe
rts on th
e
transitio
n pro
c
e
ss of the controlle
d syst
em, namely segment control.
In the proce
ss of control,
the amount chang
es of th
e obje
c
t controlled det
e
r
mi
ne the state
of the current
object, and t
hen
approp
riate control strate
g
i
es
a
r
e ta
ke
n. The de
sig
n
task of mu
lti-mode
co
ntrolle
r is to
u
s
e
stru
cture as simple as po
ssible an
d co
n
t
rol m
ode an
d param
eters as few as p
o
ssi
ble to achi
eve
the cont
rol re
quire
ment
s.
PID reg
u
latin
g
rul
e
for li
n
ear time
-inva
r
iant
system
s cont
rol i
s
ve
ry effective,
and the
quality depe
n
d
s o
n
the pa
ramete
rs
of the PID c
ontroller. However conventio
n
a
l PID co
ntro
ller
can
not tunin
g
param
eters online, an
d for nonlin
e
a
r time-va
r
ia
nt systems
of the train ATO
operating
system, it can n
o
t control the
system
well.
Simple fuzzy controller
do
es n
o
t have the
integral
unit a
nd it is difficul
t
for it to com
p
lete
ly elimin
ate the ste
a
d
y
st
ate erro
r; there a
r
e
often
small o
s
cillations n
e
a
r
the equilib
rium p
o
int in the ca
se of not en
o
ugh vari
able
gradi
ng. Fo
r the
above rea
s
o
n
s, if the two
control mod
e
s a
r
e co
mbi
ned, ca
n the
multi-mod
a
l controlle
r be g
o
t,
whi
c
h have b
o
th the advan
tage of
the two method
s m
entione
d abo
ve.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Multim
ode Speed Control B
a
se
d on Fu
zzy De
ci
si
on
-M
akin
g for Autom
a
tic… (Ju
nnian G
ou)
6767
A critical pro
b
lem ne
ed to
be solved f
o
r a m
u
lt
i-m
odal
controlle
r of train: id
e
n
tify th
e
controlled
sy
stem state i
n
formatio
n to achi
eve b
e
tter co
ntrol
of the syst
em, and the
s
e
cha
r
a
c
teri
stics reflect th
e
cha
r
a
c
teri
stics of
th
e inp
u
t and
output
resp
on
se. Sp
eed
deviation
e
and e
rro
r ch
ange
rate e
c
are u
s
ed fo
r judgme
n
t of
dynamic ch
aracteri
stic
s, a
nd thus
achi
eve
effective cont
rol.
3.2.
Fuzz
y
Controller Design
of AT
O Sy
st
em
There are va
riety of ways to desi
gn a fu
zzy
controller, and the wid
e
r u
s
e on
e is look-up
table fuzzy co
ntrolle
r de
sig
n
method:
Step 1:
Dete
rmine
the
nu
mber of la
ng
uage
va
ria
b
le
s of i
nput
an
d the
output
of fuzzy
controlle
r, wh
ich i
s
the di
mensi
on of f
u
zzy co
nt
roll
er. Sele
ct e
and e
c
a
s
th
e input lin
gui
stic
variable
s
, an
d u as the out
put of fuzzy controlle
r.
Step 2: Acco
rding to th
e
actual
circu
m
stan
ce
s of a
pplication, de
termine th
e range
of
variation of i
nput and o
u
t
put variable
s
, the quant
i
z
ation level,
factors of q
uantization a
n
d
deci
s
io
n; the rang
e of e, ec and u are all
7.
Step 3:
Definition fu
zzy
sub
s
et
of ea
ch va
riabl
e
(the varia
b
le
rang
e) withi
n
their
quantified
do
main. Firstly, determi
ne t
he nu
mbe
r
o
f
fuzzy
sub
s
et of its ling
u
istic va
ria
b
l
e
s;
se
con
d
ly take approp
riate
membershi
p
function
for l
i
ngui
stic vari
able
s
. The e
and e
c
have
th
e
rang
e of [-3
3], and the range of the
output u is
[-4.5 4.5]; Tria
ngle mem
bership fun
c
tion
is
sele
cted, the
variable
cla
s
sification u
s
ing
{NB NM NS
ZO PS PM PB}.
Step 4: Determine the fu
zzy co
ntrol a
n
d
fuzz
y inference rul
e
s. S
u
mmari
ze th
e co
ntrol
experie
nce of
the expe
rien
ced
ope
rato
r
and d
e
rive
th
e set of fu
zzy rule
s, an
d th
e co
rrespon
di
ng
output cont
rol
of each input
can be got. The method
t
o
design fuzzy rule of the
controller of the
ATO system i
s
co
mmon, n
o
longe
r give
n.
Step 5: To st
rike f
u
zzy co
ntrol table, th
at is
the p
r
o
c
ess of p
r
e
c
isi
ng. In this
study, the
wide
r used m
e
thod of ce
nter of gravity is appli
ed.
Thoug
h the a
bove five steps, fuzzy cont
rolle
r of the ATO ha
s been
desi
gne
d.
3.3.
Fuzz
y
Decision Con
t
rolle
r Design o
f
ATO
In order to i
m
prove
the
resp
on
se
sp
e
ed
a
nd
accu
racy
of the
ATO sy
stem,
a fu
zzy
deci
s
io
n-m
a
ki
ng co
ntrolle
r
is built to integrat
e both a
d
v
antage
s of the PI
D co
ntroller an
d fuzzy
controlle
r. PID control
usu
a
lly has
goo
d
cont
rol a
c
curacy, but it ca
n not si
multa
neou
sly take
the
rapid
system
respon
se a
nd sta
b
ility into ac
cou
n
t. Fuzzy co
ntrol has
goo
d
robu
stne
ss
and
fastne
ss,
whil
e it can
not id
entify small e
rro
r an
d
its
control a
c
cu
ra
cy is n
o
t eno
ugh. The
rol
e
of
fuzzy de
ci
si
on-m
a
ki
ng controlle
r is to enhan
ce
the smooth
n
e
ss of the ATO’s velo
ci
ty
modulatio
n, by fuzzy rea
s
oni
ng an
d j
udgme
n
t, an
d
to achi
eve
the cont
rol
of the two control
methods intensity swit
ch.
D
e
c
i
s
i
on-making unit itself
is
als
o
a fuzzy c
ont
r
o
ller
,
w
h
os
e inputs are absolute value of
deviation e a
nd deviation
cha
nge
rate
ec b
e
twe
en the speed fe
e
dba
ck
and a
referen
c
e o
ne,
namely |e| and |ec|. Weigh
t
coefficient
wp
i
d
a
nd
w
f
uzzy
are outp
u
ts for the PID controlle
r and
fuzzy controll
er, whi
c
h me
ans that the fuzzy dec
i
s
io
n
is a dual-ent
ry and dual
-o
utput cont
roll
er.
The
de
sign
st
eps of d
e
ci
si
on-m
a
ki
ng
un
it are
simi
l
a
r
to the fu
zzy
controlle
r’
s, an
d the foll
owi
n
g
only give the different pa
rts betwee
n
the
m
.
The fu
zzy d
o
m
ains for b
o
th |e| and |e
c
|of the de
cisi
on-m
a
ki
ng u
n
i
t are [0
+3],
and the
output
wp
i
d
and
w
f
uzzy
of dec
is
ion-
mak
i
ng
unit is
[
0
+
4
.5]. Fuzzy s
e
ts
{S M
B} ar
e
used to
descri
be in
pu
t and outp
u
t, and the
mem
bership
functi
ons
of the in
p
u
t and o
u
tput
of the de
cisi
o
n
-
makin
g
unit are sho
w
n in
Figure 4. Table 1 sh
ows the fuzzy
deci
s
io
n rule
s. Control weight
coeffici
ents o
f
decisi
on-ma
king u
n
it and
PID contro
ller can be o
u
tp
ut by decisi
o
n rule
s in Ta
ble
1.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 9, September 20
14: 67
64 – 677
0
6768
Figure 4. Dia
g
ram
s
of Me
mbershi
p
Fun
c
tion
s.
(a) i
s
the diag
ram of
membershi
p
function for
inputs |e| and
|ec| while (b
) is the one fo
r out
put
s wpi
d
and wfu
z
zy. S, M and B r
e
sp
ectively
mean
s sm
all, medium an
d big.
Table 1. Rul
e
Table of Fuzzy De
cisi
on-makin
g
Co
ntroller. S, M and B respe
c
tively means
small, medi
u
m
and big.
Orde
r an
d d
enote the o
u
tput of the PID co
nt
roll
er
and a fu
zzy
controlle
r re
spectively,
then the
outp
u
t of de
ci
sion
-ma
k
ing
unit
can
be
re
pre
s
ente
d
by th
e
wei
ghted
av
erag
e al
gorit
hm
of formula Eq
uation (4
).
**
wpid
Upid
wfuzzy
U
f
uzzy
U
w
p
id
wfuzzy
(4)
4.
Simulation and Analy
s
is
Acco
rdi
ng to the literature
[9], select ap
p
r
op
riate pa
ra
meters for th
e kineti
c
mod
e
l of the
train an
d the
obje
c
t model
whi
c
h i
s
conv
erted to a
sin
g
le tra
c
tion m
o
tor sy
stem.
The si
mulatio
n
obje
c
ts prese
n
ted as fo
rmu
l
a Equation (5) belo
w
.
2
0.07128
()
0.4356
0
.0324
Gs
ss
(5)
Figure 5 sho
w
s the
simul
a
tion re
sult
s of a r
unnin
g
train at a
certai
n train
se
ction.
Subgraph
(a
) to Subgraph
(d), i
s
re
sp
e
c
tively
corre
s
pondi
ng to t
he ATO fu
zzy control, AT
O
threshold
de
cisi
on-ma
king
co
ntrol, th
e
ATO
PID
co
ntrol
and
AT
O multimo
d
e
fuzzy d
e
ci
si
on-
makin
g
cont
rol [10]. F
r
om
sta
r
t-up,
the
train
ex
pe
rie
n
ce
s
accel
e
rated tractio
n
, sm
ooth t
r
a
c
tion,
coa
s
ting a
n
d
bra
k
ing an
d
parking
se
ssi
ons [11, 12].
The whole ru
nning p
r
o
c
e
s
s is compl
e
ted
unde
r the su
pervisi
on of ATP system. Subgraph (e)
to Subgrap
h
(h) sho
w
the
partial mag
n
i
f
ied
area
s (co
rre
spondi
ng to
a
r
ea
(A) to
a
r
e
a
(D))
of the
accele
ration
t
r
actio
n
and
smooth tractio
n
control
effect
of the
conv
ersi
on. F
r
om
(e
)
and
(f) it
ca
n
be
se
e
n
that th
ere
are
al
ways sl
ight
oscillation of
the fuzzy
con
t
rol and the t
h
re
shol
d co
ntrol un
der the
train tra
c
tion
bra
k
ing
curv
e;
(g)
pre
s
e
n
ts
the co
ntrol ef
fect of
PID control, an
d t
here i
s
static
error, too. (h) p
r
e
s
ent
s
the
effect of a
multi-mod
a
l fuzzy co
ntrol
of deci
s
io
n
-
makin
g
control. It can be
see
n
that fuzzy
deci
s
io
n-m
a
ki
ng co
ntrol
ca
n maximally track the
target spe
ed
cu
rve pro
d
u
c
ed
by ATP. As a
0
0.
5
1
1.
5
2
2.
5
3
0
0.
2
0.
4
0.
6
0.
8
1
D
e
gr
ee
of
m
e
m
ber
s
h
i
p
S
M
B
(
a
)
D
i
g
r
a
m
o
f
m
e
m
b
e
r
sh
i
p
f
u
n
c
t
i
o
n
f
o
r
f
u
zzy d
e
c
i
s
i
o
n
i
nput
s
|
e
|
and |
e
c
|
0
0.
5
1
1.
5
2
2.
5
3
3.
5
4
4.
5
0
0.
2
0.
4
0.
6
0.
8
1
D
e
gr
ee
of
m
e
m
ber
s
h
i
p
S
M
B
(
b
)
D
i
g
r
a
m
o
f
m
e
m
b
e
r
sh
i
p
f
u
n
c
t
i
o
n
f
o
r
f
u
zzy d
e
c
i
s
i
o
n
o
u
t
p
u
t
s w
p
i
d
and w
f
uz
z
y
|
e
| S
M
B
|ec
|
wpid
/
w
f
uzzy
S B/S
M/S
S/B
M S/M
M/M
M/S
B S/B
M/B
B/B
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Multim
ode Speed Control B
a
se
d on Fu
zzy De
ci
si
on
-M
akin
g for Autom
a
tic… (Ju
nnian G
ou)
6769
result, the control a
c
cu
ra
cy of the train is
im
pro
v
ed, maintai
n
ing the t
r
ai
n’s
comfo
r
t and
improvin
g the
accura
cy of the train’
s po
sitioning and
p
a
rki
ng.
Figure 5. Tra
c
tion an
d Bra
k
ing Simul
a
tion Cu
rves of
a Train’
s Run
n
ing Pro
c
e
s
s
Partial Magnif
i
ed Tra
c
tion a
nd Bra
k
ing
Curves
5. Conclu
sion
ATO system
under the
monitori
ng of ATP is
a typical no
n-lin
e
a
r, multi-vari
able, and
compl
e
x syst
em. Single control st
rateg
y
can har
dly rea
c
h go
od control effect. The multi-m
o
de
controlle
r of
ATO tra
c
tion
and
speed
modulatio
n
i
s
de
signe
d b
a
s
ed
on
ba
sic fuzzy contro
ller,
PID controll
er and fuzzy deci
s
i
on
controller, and sim
u
lation
is done
as
well. Control
requi
rem
ents in different operatin
g co
nd
itions of
tracti
on, coa
s
ting,
spe
ed-adju
s
ti
ng co
ntrol an
d
bra
k
ing a
r
e
met by the weight output of the fuzzy
deci
s
io
n. The
result
s sho
w
that multi-mode
train spee
d control of fu
zzy deci
s
ion
-
m
a
kin
g
can a
c
hieve goo
d e
ffect unde
r di
fferent wo
rki
ng
con
d
ition
s
an
d control stag
e.
Ackn
o
w
l
e
dg
ements
This re
se
arch is finan
cial
ly sup
porte
d
by the
tech
nology
re
sea
r
ch
a
nd
dev
elopme
n
t
prog
ram
of Ministry of
Rail
ways, China (G
ra
nt NO.20
11X00
8-D) an
d th
e Youth Sci
ence
Found
ation of
Lanzhou
Jia
o
tong Univer
sity, China (G
rant NO.2012
035).
Referen
ces
[1]
T
ang
T
,
Huang
LJ.
A
survey
o
f
contro
l algorit
hm for automatic train operati
o
n
. Railw
ay Jo
urna
l
. 2003
;
25(2): 98-
10
2 (in Chi
nes
e).
[2]
Hiro
y
as
u Oshima, Seiji
Y
a
suno
bu, Shin-ic
h
i Sekin
o
.
Automatic train operati
on s
y
ste
m
based
o
n
pred
ictive fuzz
y co
ntrol.
Artificial Intel
lig
enc
e fo
r Industria
l
Applicati
ons
Pr
oc of the Int W
o
rkshop.
Cochi
n
.
IEEE Press.
1988: 4
85-8
9
.
[3]
Xi
e LJ, Nin
g B.
Applicati
o
n
of fuzz
y
con
t
rol in the a
u
t
omatic train
oper
ation.
Au
to
ma
ti
on
a
n
d
Instrume
ntatio
n
. 1999; (4) (in
Chin
ese).
[4]
Satoshi S
e
ki
n
e
, Naok
i Imas
aki,
T
s
unekazu End
o
.
Appl
i
c
ation
of fuzzy n
eur
al n
e
t
w
ork contro
l to
automatic train
operatio
n an
d tuning of control rule.
T
he 4th IEEE Int Conf on F
u
z
z
y
Syste
m
s
.
Piscataway: IEEE Publis
h
. 1
995: 17
41-
174
6.
[5]
Liu HW
, Z
hao
HD, Jia LM.
A
stud
y
on th
e contro
l a
l
g
o
r
i
thm for auto
m
atic train op
eratio
n Chi
n
a
Acade
my of R
a
ilw
ays Scie
nc
es.
2005; 1
7
(3)
:
557-58
0 (in C
h
in
ese).
[6] Y
A
NG
G.
High
-
spee
d
mag
l
ev
train driv
in
g c
ontrol
l
er d
e
sig
n
. Procee
din
g
s
of the 27t
h Ch
ines
e Co
ntro
l
Confer
ence. K
unmi
n
g
，
Y
u
nn
an
，
C
h
in
a. 20
08; 454-
45
7. (in Chi
nese).
0
200
40
0
600
0
50
100
150
200
250
300
ti
m
e
(
s
)
s
peed,
v
k
m
/
h
(
a
)
A
T
O
f
u
zzy c
o
n
t
r
o
l
0
200
40
0
600
0
50
100
150
200
250
300
ti
m
e
(
s
)
s
peed,
v
k
m
/
h
(b) A
T
O
dec
i
s
i
o
n-m
a
k
e
i
n
g
c
ont
rol
of
t
h
res
hol
d
0
200
400
600
0
50
10
0
15
0
20
0
25
0
30
0
ti
m
e
(
s
)
s
peed,
v
k
m
/
h
(c
) A
T
O
P
I
D
c
ont
rol
0
200
40
0
60
0
0
50
10
0
15
0
20
0
25
0
30
0
ti
m
e
(
s
)
s
peed,
v
k
m
/
h
(
d
)
A
T
O
m
u
l
t
i
m
ode f
u
z
z
y
dec
i
s
on
-
m
ak
i
n
g c
o
n
t
ro
l
190
195
200
20
5
210
280
285
290
295
300
305
ti
m
e
(
s
)
sp
e
e
d
,
v km
/
h
(
e
)
A
T
O
f
u
zzy c
o
n
t
r
o
l
190
195
20
0
20
5
210
280
285
290
295
300
305
ti
m
e
(
s
)
sp
e
e
d
,
v km
/
h
(f
) A
T
O de
c
i
s
i
on
-
m
ak
ei
ng c
o
nt
rol
of
t
h
res
h
o
l
d
19
0
195
20
0
205
210
28
0
28
5
29
0
29
5
30
0
30
5
ti
m
e
(
s
)
sp
e
e
d
,
v km
/
h
(g) A
T
O
P
I
D
c
o
n
t
r
o
l
19
0
195
200
20
5
21
0
28
0
28
5
29
0
29
5
30
0
30
5
ti
m
e
(
s
)
sp
e
e
d
,
v km
/
h
(h) A
T
O
m
u
l
t
i
m
ode
f
u
z
z
y
dec
i
s
on
-m
ak
i
ng c
o
n
t
r
o
l
(A
)
(B
)
(C
)
(D
)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 9, September 20
14: 67
64 – 677
0
6770
[7]
Du QX.
Desi
g
n
an
d calc
ul
ation
of
locom
o
t
i
ve a
u
tomatic
adj
usting s
y
st
em.
Journ
a
l o
f
Southw
est
Jiaoto
ng U
n
ive
r
sity
. 1977; 2(1
)
: 1
1
1
-1
18 (i
n C
h
in
ese).
[8]
Don
g
HR, Gao B, Ning B.
Ada
p
tive fuzz
y
con
t
rol
for speed control s
y
st
em of
automatic train operati
on.
Journ
a
l of dyn
a
mics an
d cont
rol
. 201
0; 8(1): 87-9
1
(in C
h
in
e
s
e).
[9]
W
ang HP, Yi
n
LM, She
LH.
Pickin
g
run
n
i
n
g data
b
y
us
in
g PLC to
id
ent
if
y
th
e kin
e
tic
mode
l of the
train.
Co
mp
ute
r
Measure
m
ent
and Co
ntrol
. 2
003; 11(
3): 230
-235(i
n
Ch
ines
e).
[10]
Gao B, Dong
HR, Z
hang
YX.
Speed adj
ustme
n
t brakin
g of train
operatio
n system base
d
on fu
z
z
y
-
PI
D
sw
itching contr
o
l
. Proc of the 6th Int Conf on Fuzz
y
S
y
st
em and Kno
w
ledge Discov
e
r
y
.
T
i
anjin, 2009:
577-
580.
[11]
Don
g
HR, Gao
B, Ning B, Z
h
ang
Y
X
. Spe
e
d
ad
ju
stment
b
r
akin
g bas
ed
o
n
fuzz
y
PID so
ft S
w
itchi
n
g
control of the
A
u
tomatic
T
r
ain
Operatio
n. Con
t
rol and D
e
cisi
on.
201
0; 25(5)
: 794-80
0.(in C
h
in
ese).
[12]
Meng L
T
, Jia SF
.
A
multim
ode
intelli
gent cont
rol method.
Ele
c
trical Drive
Au
tomati
on
. 20
07
; 29(2): 1-4,
1
1
. (in Chi
nes
e
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.