iDocSlide.Com

Free Online Documents. Like!

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Share

Description

Pressure controlled object of distilling tower is complicated, has more effective factors and strong interference, its modeling is difficult. so normal control algorithm can't control it with satisfactory quality. To overcome this difficult, a

Tags

Transcript

Pmceedngs
zyxwvu
f
/E€€
zyxwvutsr
NCON02
zyxwvut
A KIND OF GENETIC FUZZY CONTROL ALGORITHM AND ITS APPLICAION IN DISTILLING TOWER PRESSURE CONTROLLER
zy
ang
XIONG
Dr.’
cniarm~mber1EEE;TIMO
R.
NYBERGDR.’,
Zhi-yu
ZHANG3,
and
Guang-yu
X10NG3
’Nokia Business Infmtlucture, M%M team, FIN-24101, Salo, FINLAND Email: xiong.gang@,nokia.com ’.3.‘Automation and Control Institute, Tampere University of Technology, FIN-33101,Tampere, FINLAND
Abstract:
Pressure controlled object
of
distilling tower
is
complicated, has more effective factors and strong interference, its modeling
is
diffbxlt,
zyxwvutsrq
o
normal control algorithm can’t control
it
with satisfactory quality.
To
overcome this difficult, a kind of genetic fuzzy algorithm
is
put forward.
While
fuzzy logic mimics the human’s imprecise reasoning, the genetic algorithm mimics the evolution
of
the nature. an optimum point always can
zyxwvuts
e
found by
the
genetic fuzzy algorithm. Experimental result proves that, even through there
is
strong interference, the algorithm
still
can ensure actual pressure value
to
fellow the given pressure value quickly and steadily.
i
Key
zyxwvutsrqpo
ords:
distilling towcr, pressurc control, genctic algorithm, fuzzy
logic,
gcnctic fuzzy algorithm
2 1
Fuzzy
Logic
System The basic principle of fuzzy logic system in the form
of
hybrid genetic fuzzy logic
(HGFL)
is shown as Fig.2. The
fuzzy
logic system execute the mapping from
x=(e,i)EXc ’
to
AUEUCX
wher(e,e)
z
re
accurate measurement values, which
re
fuzzied by the fuzzy module,
re
dc-fuzzicd by de-fuuy module. Every input is grouped into
one
of
seven
fuzzy
sets, i.e.
NL
(Negative Large), NMWegative Medium), NS(Negative Small), ZE(Zero), PS(Positive
Small
zyx
,
PMfPositive Medium), PL(Positive Large). Actual value can be gotten by input subordination function. For simple optimization process, every subordination function can be defined by
a,
and
zyxwv
i
:
1
xlqand subset=NL
0
xlqand subsetf NL
1.INTRODUCTION
The
oressure
(PC302 in tower 2) control svstem of
2
(x-q.)yq
<xl-
4
a)
olvent recovering device
[I1
is the object mainly discussed in
this
paper. PC302 is set by energy optimization system of upper controller, the pressure control system mainly keeps
&a,,/ )
=-
8.
a.)
2
the
actual value of PC302 follow
uu
&e
set value
buicklv
”
,
r
,.,
ll\
2
.
and
accurately. The control valve bf PC302 is the
output
valve
on
the top
of
tower
2
(beside Al). Because.input
4x’q’p,)’~(p,
-x gy
V
YI
material fluctuates greatly, it is difficult to keep PCi02 stable. Pressure controlled object of distilling tower is complicated,
has
more effective factors
and
stron
q.,h.)={
interference, its modeling is difficult,
so
normal
control algorithm can’t control it with satisfactoly quality
I]
TO
Overcome
this
dimcult, a kind
of
genetic
’.
1
x>A.and subseePL
0
x+fi.and subsei PL
Where xis eor
e.
Fig3 and Fig.4 separately are the
fuzq
subset where the subordination hnction
c
and
e
arc
21,
and
ug
separately. lgorithm
is
put fowd
where
genetic algorithm
is
used to change the. subordination function. While fuzzy logic mimics the
human’s
imprecise reasoning,
the
genetic algorithm
mimics
the
nature
evolution:
Random
search
1
ur
NM
NS
ZE
ps
PM
p~ method
is
used by
GA
an
optimum point
always
can be found by the genetic fuzzy algorithm.
2
THE
DESIGN
RINCIPLE OF
GFLC
Recently, more and more people use CA to optimize
fuzzy
.
logic
set
‘?I
its basic princi le is shown in Fig. 1. Hybrid
0 0
Genetic
Algorithm
(HGA)
R
is
an
optimization tool, its
emin
e
mar
operation
s
based
on
real nninber,
or
float number, and some special knowledge.
So,
fuzzy logic
set
will change with the change of GA operation. Fig.3 Input subordination function of
e
0-7803-7490-8/02/$17.00O2002
IEEE.
1455
To get fuzzy output from the subordination functions
Of
input e and
zyxwvutsrqp
a
set of fuzzy control rule related to fuzzy execution and combination
rule
inference.
Normal
fuzzy rulcs
zyxwvutsr
se
IF.. .THEN.. structnrc For cxamplc: IF
e
is
NS
and
zyxwvutsrq
is
PS:
then
AU
is ZE. Where.
NS,
PS
are
the fuzzy subset of e and The actual output
value
of
zyx
uzzy
controller is computed from special input signal by specified fuzzy
rule.
To
make
System
zyxwv
e
fuzzy
output
set
de-fuq process is necessary, correspond output subordination function (Fig.5) is defined inequation
1.
separately;
zyxwv
correspond seven
fuzzy
subsets,
i.e.
NL,
NM,
NS,
zyxwvu
S,
PM,
PL.
A
fuzg
model based on
fuzzy
rule
is described in Table.
1.
zyxwvutsrqpon
R
N m
NS ZE PS PM PL Table.1 IF-THEN
rule
1
The
rules
in Table. lcan be expressed
as:
RI
:If
e
is
NL
and
e
is
NL
then
Au
is
NL
R,
:Ue is
NL
and
e
is NM then
Au
is NL R,:UeisNLand
e
isNLthen,Atr isNL
R,,
:U
e
PL
and
e
is PL then
Au
is PL
...
...
or,
R =RI
UR2
U...UR4,
(2)
zyxwvu
t/,
NLNM
NS
ZE PS PM PL
,
A%
Aur
0
Fig.) Subordination function
of
output
So;
the precondition value of every rule is computed out, ind applied to itsxonclusion parl.
The
result of every f.uzy subset
is
assigned to
the
output of every rule. Triangle subordination function of output is gotten
from:
uiA,,,(x)
=
min{z{,(x),u,(x)}
(3)
Where:
A
and B
are
the fuzz)- subset of
e
and
e
The fuzzy subset
of
all
output parameters consist of a independent
fq
ubset.
The
subordination function
of
combination output is:
y,,,,, x)
=
max{u, x),u, x)}
(4)
Where:
C,
is
the
fuzzy
subset ofAu
2.2
HGA
(Hybrid
Genetic Algorithm)
GA
is necessary to apply optimal fuzzy set for control
[61
The evolution process happens on the chromosomes. which represent the values of
4
nd
8
nput subordination fimction of the
fuzzy
subset. Its actual value codes
are:
c
=
{a,.,
Lt
:...
a,,,.
p
a<.,
4
...a;,,,
8:
a,,,
p. .:...
a .,
Where:
u',nand/3en
re
nput subordination function
of
the
u'4m?dD2n
re
input subordination ruuction of
the
a undpun
re
input subordination function
of
the
fuzzy subset of
error
nrh
fuzzy subset of the changes of error
nm
fuzzy subset of
error
h
f
Azr
;
Evolution
and
genetic Operation:
Father
1
ather2
Son
Fig
6
Operahon
1
one
point exchange
1456
Father1
zyxwvutsrq
Father2
[
Fig.7
Operation
2
zyxwvutsrqp
wo
point
exchange
,
,
Natural evolution
(or
the evolution happens
in
other place) isn't
a
fixed process.
In
fact,
natural
evolution pmcess
zyxwvu
s
a process to compete
for
resources, where those units
that
have better
life
ability are easier to
zyxwvutsrq
mvive
and spread their gene material. Steady-state reproduction method is applied, where only
one
chromosome is reproduced in every generation
zyxwvutsrqp
ne point exchange1']:
A
new generation
is
born when father
1
is replaced by father 2
in
a randomly selected point (Fig.6).
Shift
lift
rids
Shift
right
side
Fig.8
Operation
3-
Subset shift
Shifl
liR
comer
Shift
rightcoma
..
.
..
.
.,
,
Fig.9
O&ation44orner
Shift
,,
TNO
point exchange
I:
It is similar
lo
one point exchange; exceptltl~al wo division points are randomly selected.
A
new generation
is
born after the segment behveen the
two
division points of father
1
is replaced by father
2
(Fig.7).
Subset
Shift:
randomly
to
left side (or right side)..It
shif
in parallel
to
keep the subset shape,
zyxwvutsr
s
FigX. Comer
Shift:
The operation
shift
randomly the sclccted
fuzzy
subset L steps. L is also selected.randomly to left corner
(or
right comer),
zyxwvutsr
s
Fig.9.
Subset Expansion:
The
operation
shift
mndonlly
L
steps.
L
is
also selected
.
Similar to operation
3,
but nmst
sld
o both
&&tion
simultaneously (Fig. 10).
8.
,,.
..
Fig.10
Operation
5
Expansion
of
sub
set
2.3
System Performance Adaptability evaluation:
The quality of fuzzy subset used in control
is
chging. The performance
of
control systcm
varies
with the adaptability standard
of
genetic algoritlun
For
most
dynamic control system, over shoot, stable error,'and rcaction time
a?
often considcrcd, To reflect the benefit
of
fuzzy
.
control subset,
three
parameten. are. mainly considered in
the
adaptahility evaluation:
1)
over shoot'
ewe,
2)
peak time value
zyxwv
;
(3)
Stable average
square
ermr
e_.
So,
adaptabiliq.is defined as:
F
=.
qe
+
a2tp,
a3e:.~
2
(6)
.
Where
zyxwv
i
s the parameter weight. Colony size is about
2 c9?
..
..
~
3.
EXPERIMENT
ESULT
3.1
GA
performunce
To
prove the.optimization ability
of
Genetic Algorithm.
50
generations
are
built. The average'adaptability
of
every
new
generation
is closer to'the minim.um value
tk
heir.father generation.
,This
trend proves that, hc latest generation
bas
the'best performance.
The
best cbromosome:is: C=[12 17
20
21
5
46
33
86
65 71.60
78
77
87
18
21~17 2
7
73
47 91
6
79 19 77 76 92 17
28
11-
5X
11
78 46 64
5
86
11
98 98 991
3.2
Closed loop performance
Two groups
of
experiment
are
ananged.
One
group sets PC302as 0.237MP, the experiment
result
proved that the actual value
of
PC302 can recover to
0.237MP
even when there exist a great change in input material
of
tower 2, the fluctuation
is
lees than
5%.
Another group applied square wavefor the set value
of
PC302 wave amplitude is
0.03Mp,
maximum
value is
0.207MP,
minimum
value is
0.237MPa,
experiment result proved
that
the actual value of
PC302
can follow up
its
set value quickly and accurately.
..
1457
4.
zyxwvutsrqp
ONCLUSION
zyxwvuts
xperiment& result proves that,
even
through there
is
strong
interference, he algorithm still can ensure the actual pressure value
to
fellow the given pressure value quickly and
steadily.
GA
is proved to be successful
to
optimize
zyxwvu
uzzy
logic control.
In
fact, Gcnetic
Fuzzy
Algorithm
not
zyxwvu
nly
zyxwvutsr
an
control pressure of distilling tower, but
also
solve other optimization problem existing
in
other applications
I.
Reference:
[I]
Gang
zyxwvutsrq
iong
The
Architecture,
Intelligent
Tools
of
CIPS and its Application
Doetoral
Dissenation
of
Shanghai
Iiao
Tong
Univeniy.
1996.10
Karr
C.L. and
Stxiley
D.A.
Fuzzy
logic
and genetic algorithms in time-varying
control
problems. ProcNAFIPS-91:28S-290 Meredith D.L.,
Kma
K.K.
and Karr
C.L.
The
use
of
genetic
algorithms
in
the
design
of
zyxwvutsrqpon
uuy
logic
controllers
.PROC.
WNNAlND
91.695-702
Park
U,,.
Kandel
A.
and
Langholz
U.
Oendic-Based
New
Furzy
Reasoning Mod& with Application
to
fuuy
control.
IEEE
Tmr.
zyxwvu
yst.
Man and Cybernetics. 1994.VolZ4,No.l:39-27
[SI
Davis
L. Handbook
of
Genetic Algorithms.
Van
Nostrand Reinhold.1991 [6] Goldbsrg
D.E.
Gendic
Algorithm
iii
Search
Optimization, aid Machine Learning. Addison-Wesley. 1989
[7l
Holland J.H. Adaptatim in
Natural
and Artificial
Systemr.
Univ-ily
of
Michigan
Press
1975
[SI
Fshehm LJ.,Caruana R.A. and Schaffer
J.D..
Biases
in the
emover
landscape.
Proc.
3rd
Int
Conf.
Genetic
Algorithm
.1989:10 19
Schaffer J.D..
Caruana
R.A.
and Eshelman L.J.
et
al..
A study
of
control
parametar affecting
online
performance
of
genetic
algorithms
for
funetion
optimization.
Roc
3rd
Int.
Cod.
Genetic
Algorithms.
198931-60
121
[3]
141
[9]
Acknowledgment:
'This research
is
partly supported by several resources. 'The project supported by Academy
of
Finland: Integrated Modeling and Overall Optimization
of
Strategic Automation System
used
in Modem Pulp Paper Industry . The project
of
Process Integration Program supported
Finnish
TEKES:
Novel
Information Integration Technologies
WIT)
for
Strategic Automation System (SAS) used in Modem Pulp Paper Industry .
NEPPI
Project
of
TEKES:
Networked Enterprise for Pulp Paper Industry. VPPE-project supported
by
China-Finland government cooperation research program
2000-02:
Modeling
for
Virtual Enterprise and
E-logistics
Supporting Technologies, Chinese partner: Academic
Prof.
Cheng Wu, CIMS ERC. Tsinghua
University,
China. Fuzzy Logic System
.............................................................................
Fig.] The structure
of
Genetic
Fuzzy
Logic
Controller (GFLC)
Fuzzy
Fuzzy
Inference Maclune
=
zyxwvuts
e,
zyxwvut
Inpul
t
t
t
Input subordination Fuzzy
Rule
Base
Output
subordination Function Function
Fig.2
A
Fuzzy
Logic System
1458

We Need Your Support

Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...Sign Now!

We are very appreciated for your Prompt Action!

x