WHAT ARE FUZZY SETS I MODELS?

_e 0 1'. _'
n 1 d 1,,,1r) cshiP. 111_the sd.
11
• ;
rli·rnging d cgre 1'"){,ree<=: of th P ,;ll'li,Det :.r.r, ,
. 1 l' 11 ,vin<1 a . "
zy set is a combination of the e emrn ' . ··tiC11l among var ious< 1 mnr.iJer'>ntP o: ,.,, • ..•
':'
'I "
uzzy means vagueness, m. t 1er 1'vord tht• 11a11s1 Tl er ',' e tie
1ambiuuous. 1 'i -l , "I"',-,rtainty a..d
'
• I fOl - arc 1 ngul' am '. .
complies that the limits ol l 1e uzzy se1 s ' .csta "'fu nclioll ti ••'·1e'nt1f •XH ,_• -I conta1::1 a,.
1 11
s
elements from the universe in the set· is nH•,i urccl• V - • Lril<e NoW,I( ft1zzy et •,<,'-•\ ,.. Clf tun:.,
ambiguity. A fuzzy set is denoted by a text having Lilqc 1111cer, 5 it.in theun1vcrs1 is J nwm J • h : ·-
• . I S . is a11 r erne1 d 1 s ••. ,Pn'.0 •
thepossible outco me from 1 ntcrva to . uppo . ventinn 1sc · fL i.t'.·.,.y.......
•
O I ·(•" 1
s ' , ,H
seLX ,Uiefunclion "ives0 the 111;:'IJlJling bv XJ (,1) - f(l, l] - 'lhc 11011011 con ,., r1111tP tor ft1'l7-V 9•:.X ;:; g•\t\
uz.z,yverse of discou rsf' U (set ofinput va lues lor tlw 1
u111 X) • <lisrrete a1w
pt' is •
h, ..

(l!

Thefuzzy set theory was initially proposed by a computer scientist totfi A. Zadeh in the ywr of 1qr,;_ Aite-r
that lot of theoretical development has been done in a similar field. Previousiy tl-1e theory n ··-;p st•t
based in dual logic is used in the computing and formal reasoning which involves the '.>o1utiom, ;:-- e ;,<-,
two form such as "yes or no" and "true or false".

Fuzzy logic: Unlike crisp logic, in fuzzy logic, approximate human reasoning capabilities a:·e add1:d ,n ord'•,
:o apply it to the knowledge-based systems. But, what was the need to develop such ·a theory? Trie funy
log,c theory provides a mathematical method to apprehend the uncertainties related to tl:le bll':1$n
cognitive process, for example, thinking and reasoning.and it can also handle the issue of uncc1t,:ii,1ty •. c
lexical imprecision. • •

Example: Let's t.ake an example to understand fuzzy logk. $uppose we need to finchvhetber tbe c...11<E!, ,:i
thc object is blue or not. But the object can have any of the shade of blue depending on the imt.1::.: . of -ii
primary colour. So, the answer would vary accordingly, such as royal blue, navy blue, sky bit:c tu:·qu i-..,
biue, azure blue, and so on. We are assigning the da1:kest shade of biue a value 1 and Oto the v\.h,,t! co lU:
ar tht! lowest end of the spectrum of values. Then·the other shades will range in O to -1 a,:cordin.; :;.;
intensities. Therefore, this kind of situation where any Qftlie values can b_e accepted in a range 01 Oh, 1 1s
termed as fuzzy. • ••

EXPLAIN THE PROPERTIES OF FUZZY SETS

Fuzzy set have a number of properties. Here are definitions of the most import,ml properties. hlt they .ir..:
not necessary for understanding of the course. 1fyou wa11t, youcan go now directly to th llt;).l settion. Le,
X be a set and A, a fuzzy subset of X and LA the membei;ship function characteriiing it. i,A(x) 1s -:alle,11,1c
membership degree of x in A.

DefinWon 1: Let X be a set. A fuzzy subset A.of Xis characterized by a membership fnnction.
fa: X--. [O, 1]. (In theory, it is possible t·ha't the output is greater than 1, bt in pr: nice
it is almost never used.)
N(Jte: This membership function is equivalent to the identity function of a classical set.

Definition 2: The height of A, denoted h(A.), corresponds to the upper boL1nd of th.: co1_10rnain 0f its
membership function: h(A) = sup{µA(x) Ix EX}.

Definition 3: A is said to be normalized iFaqcLonly if h(t\) = 1. In practice, it is exlre•ncly ia.-,:

toworknn nm1-11ormalized fuzzy sets.
 .Definition 4:The support uf A is the set uf elements of X belonging to at least <;n,r,•• /\ li.e the J1,,,n,l;ersh1p
degree ofxis strictly positive). In olher words, 1hc support is the set supp(A) {xf XI pMx.)> O}

Definition 5: Tile kernel of A is th"' set of Pit'11wnts ol X lwlonging enlin:ly tn A 111otlwr word·,th<·
kernt•I noy(A) = {x EX I t/\(x) = 1). By construction, noy(/\) t supp(/\),

Definition 6: A,n1-cut of,\ is theclassicalub l'l olt'lt·111t•nts with a melnh,,r,hq, di'fl'<·1· gJ'P,11!·1 ,h. :, 111
equal toa: a-cut{/\)=-{x EXI rLA(x) "a}.

Definition 7: I.et V be a variable (quality or service, tip 11nount, etc.), X the rnngr: of vnluf', of ,r.,_ ,,n-
uh!1• and TV a finite 01· infinite set of fuzzy sets. A linguistic-v.irlable
corresponds to tll<' triplet (V, X, TV),

Anotlw1· membership function for an average tip through whlch we have included the above p101,1 • ••e, 1•,
presented in figure
. ,
i
Membership \
degree i

a·putvanab.e• Tip

A membership function with properties displayed

\: 3;, ,.., ,,•t:

\' lf
{ _£.oor .G ,ih ,it}
1i-

Linguistic variable •quality of service'

'

I
 LOW ).(ED.rc,r fil(,11

"!'
tior"- --.._--..1.
....
30 70 lO

Graphical representation of a conventional set an<l a fuzzy set

In the classical set, any temperature can only be categorized into one subset, LOV\/, MEDlUM)!' lllGH, and
the h.oundary is crystal clear. But in the fuzzy set such as shown in Figui-e (b), these bound,1iics ber.ome
vague or smooth. One temperature can be categorized into two or maybe ev. n rhr e subsets
:-irnultaneously. for example, the temperature 40°F can be considered to belong to LO\IIJ to « certai..n
degree, say 0.5 degree, but at the same time it can belong to MEDIUM lo about 0.7 d0gr.-ee. Another
interesting thing is the temperature 50°F, which can be considered to belong to LOW and Fl!Gl! tu arcnmd
degree and belong tu MEDIUM to almost 1 degree. The dash•line in Figure b rcpresc1its the r::1ssic.,i SPl
boundary. It is clear that a fuzzy set contains elements which have varying degrees of me1nl1,arship it, tht>
sd, andthis is conti".1sted with the classical or .crisp sets because members of a classical si't cannot he
members unless their membership is full or complete in that set. A fuzzy set aliows a J11,;mber lo h,we a
partial degree of membership and this partial degree membership can be tJrnppcd inlu a .function or a
universe of membership values.

Comparison Chart

I
BASIS FOR FUZZY SET CRISP-SET - l
C O M PARISO N
B a s ic - P -1-·e_s_c-ri_b_e_ct_b_y_va_gtet oa_rrn--:b-i-gu-◊-1-1.-•-+-=-D_e_fi_n-ed-.,.b_y_p_1_·e-ri•;p
and·certain i)roperties. characterl.stics.
Property Elements are allowed to be Element is either the rnembei
partially included in the set. of a set or not.
Applications Used in fuzzy controllers Digital design
L'.!:o'.tg'..'.:ic -+-,:,::.:n:;:fi:.::n:,..:it:.:.:e:...:·V-=a l::,uec_d7'-' ------t.b:::i"':.v'::a:,:lt:--:-1e::-:d:;-"'---- ===-·-_7

•
 l \elA\N
Fuz•n1., LOGIC SYSTE
-"

I he imp\emL•11tat1on ff
l F llZZ·I11c,1t10o1-1 unzmzyvel t,ogciclatsechniq11l ' t0 •11,''1l ,q1phcatlo11,, l'11uin•, 1h,•
tnllow111p,ll1:c
115 1 ,ep Mt•mb,. h' SH,il cl,t •10• ll 1s, d t,t into full\ d,11,,01
1 Fuzzv11111,,1 . ll Phmrtlnns (IVII s,
enn• ro11•,;s -t o• mb.Ill( lllPll•CI ,,,rumt1011sw1th1I, f ,111111
-', Deluzzifit"\l' . Lo derivt• lh<.' 11"'/.)' oulput
' ion- use d1ffer1•nl nwthods t, c,1k11l,1te l',ldt assriciah d n11tp111, , 1,111 tt, 1,
mto" tahll': till' lookup 1,1blt•, l'ick up thl' output Imm the luc,I up 1,,rn

cn:1blemachin
, nu,t
.
ht·
bec t
,h mentioned bet.e, all ma
_c,s to l,\ndll•
1<
based on tlw
climes C,lllcun l'l1t input
pron•ss

Ol1\t l lt>d to hnguista van,1hl1•s \\1th 1.,

c-r1,p<1111ing ,lit appllt
01 1lass1t,1I d,1t,1 <Ilion
u< h ,..,1·11111·1 (I ur l I,,ur rl, 1
vagut• langll,ll(e input s1H h 1· 'Sonwhow ',,1t1sl11 I' th,• 11 I 11 tit .1 m 111 r,v
,ompom·nts Fm 11,,1"", 1 on• .., n •11
.'l 0111 . 10ner systc•m· lhe i1Ipu t L1•111per,1lurt• ,111d tin 01t,put control v.111.ibl,• 11111,t h , \ 1 , ,,.
c s'1uatec1 1m u1st11 variables s11d1 as llll,11 'Ml-. 111 I 'LOW' and 'l\t, 1 ,' 1 v· •
'h\',•1 1>,.
101 mer is. corrc·s11oncl111g to th c •mput tcmpcr,\turc ,111d till' latter is assoc1,1tt•'d w11h tlw 1c,1,111on
1 w ," d,r 1ieope_ratmg moto1 Besides those convers1011s, ho h the input and the output must, , , lu•
<111wc·rtccl om cnsp data to fuuy datn. /\ll of these job,sire performed by the first step h1z1.1l1c,1tw ..
. In the second step. to begin the fuzzy inference proct>ss one need r11mhi1ll' tht1emlwr»h1pFu11lion
wnh the control rult·s to derive the control output, ,11\ll arrange tho'ie outpuh into " i>I" "d"I th•
lookup table. The control rule is the core ol Lhe f111.7y i11fcrenet' prnn·s, an<l thfl<t' 111l•·s 11 <111, C'I\
rel.11,·d to a hum,10 b< mg's mtuitton and fl:chng. Fo1,.,mple, still 'n thc air
rond111om.,·1,,al,l,Isysttm 1
tlw lt.>mperaturc• is too high, the heater should he turned off, or the ht•at driving motor ,houldh,sl(lwn
,., n which is a human being's mtuit1011 or rommoncn,;e. Different meth,1ds ,;uch .,sC·•nt<•tolGr,1vil:,,
{CO<,) or Mean of Maximum (MOM) are utilized t<, rc1lculate the associated nrntrnloutput, and P.lr!
control output should be arranged into a table called lookup table.
Duringan nctual application, a control output should be selected trom tlw lookup tablL• uc, ,·l1!p•·d II 011
the last step based on the current input Funhermo,e, 1h.1t control output should be rn11,•p1 ltd lrn1•1l.,,
linHIIISticvari,1ble b,1ck to the crisp vanablt• and output to the control 01wr.11ur. Thh pror ";• :,, 1lk11
cleluzzification or -;lt'p 3.
111 most ca-.1·s tlw input variables are mon· th,111 nt .ltmension tor re,11.111:11,1!1lw11:., rnd n:wm·1•tl<H
pl't·lorm fuzzificalio11 or develop a Memlwr hip h111, tion for NCh di1111•D,i<111.1l v,11,,,hi,· "'1'•11·,1\rtl
Pt>rlorm the same operation if the system has mul111,,., 0111put variables.
Summarilya, fuzzyprocess isa process ol cnsp-f11ay-11 isp for a rl'al syst<'111'1IH' o:-igin:11input an,I th,·
tC'rminaloutput must be crisp variables. but the 1111e1111,,diate pro< l'SS 1s I Im'./.} inlcrenc•'prnr, ss ·1h
rea on whyone1weds to change a crisp to ,1 lun:y v n.11,lt is that. lru11, tlw po11\\ ol viL'''d l11u.y conlrol
< r a human bl•mg', intuition. no absolut,l' n '' •xi,tt•cl r r 11w, rid
11o1
/\ny physicalvarnible may contain sonw oth1•1 componenls. h1r 1n 1.i1H1, 1f sonw,Hll'-1 - th,
,emperature here 1s high. This high temperature wnt,1111s some middle :111<1 evn1 \1,.v \l•rnperatitn'
nHtpl onents. FromI his poi nl of view, fuzzy rnntrol use-. universal 01 gloln1l componcl'I , nr,1 1u.,t ,1 linutld

r,rn1:e of components as the classical variables die.I.

With the rapid development of fuuy tt:chnoll,g11·s, different fuuy t:ontrnl st ..1t,. ,1·, l.,:..i: l>,1•11
1
devdoped based on different classical control m.:thoth, such as PIO-luuyro11trol,sl1 j•111:·m,,1h:tu//Y
control, neural fu:t.1.y control, adaptor fuuy c.:0111rnl <111d phase-plan mapping fuu.y wntrol. More ,111<111t0•1
nc\ fuizy control rrategies or combined c:n,;p t1nd l11zzy rnntrol tcchn1qut.:s c1,·1· be1ni\ tl,iv€1o1wd ,11HI
w111
be applied to many .,reas in our society 111 the future.
 ,., ,j·i oi ,wplil,1!it''' 'lb,
I·'uzzy Control f
I
expert ,11 c1nyI l ,•hlt:1 . ,It' ' t'
.l ,c; -,·'11iPt' w·1,1t ,•c ioq
F · , \1 1,·IC'<lg, '
u zzy control rule can be considered as tlw ,n
N 1c ·idJ111 fuc1r11hrn, (IC '..
cl,-, • t ·1 •
I Rules:
- . .I . t ltid ,,. '"h'• 1
10

c • •
nz 1u e 1s r,•prcsrnted by a sequence ol1l H, •,.0 n1 11h-<•rved •- I in. for11 1al! Oll., , nrii:,. • •1: i::c·· ;•, .. l )... 1 •• ,,
" n P ' . P ,, l l •'

.,
wipeh, 1
or o.utput should br 1,1ken in terms of 1!11 1 tin 1·t1lh O
.t ,(eswn
1 or build,, , f < • •. :..c ,,11, ,•t:on.
I cccib acor
,,ulput I< .If. a closl"d-loopTlw I 1 •d Till' 1I I c1·r1· rent
1s11ly
is mni O' m 11
a ,,pp ust·d • to r·s dc•pen< t ,'..,
,IW ·1nu.i,
control system ptuci l •itl
l it
onpc1ch I e • 1 ' I, s t· :,.,y ,·cts ':I
conclusion. IF part I 111
1 •1 11
1, • or ou t!P ll"1
,,,.,... iistic vana!,J,! ·in< 1 11,•
'ur ' .• ••r ' -
on a human bt'ing's knowledge or cxperit>nce. w 1 11 . linguistic I v;in a ) e . 't<
.· t s 1c 1rilwd u s ing • tlI 'e•d,• 1:! ·u· " ' 1
'1 !••.. . Tl "' •
1
,ind Athe fuzzy
THEN IF-Tl!EN
part rule associates
can be utilizt'd atocont I1the
giv,, t 1011
ronr c ". "• 1 .
1 11 . , . theow!Nl d
' ' ge hy using
t 'I" N' lu wh •
it
tlili'ili·, 1•
lw, !f
THEN rule is widely used by tht• fuzzy 111fen•rn (' w 11•111 10tt,mipull I '!
1 1 .
°iht• c1,,wcc !l· i\\(' <• ·t.1.Y
d
male Iics tI1e con,_,,1. I e I"·1gure I wI 01\ I·11u·,11.•.111"•' ·•1 w.iy 10 1'1 •
t·w11of a
111 cc
d.t' ndltH ...• ••, 1 .-,
ll
.11 \, 11 , , ,., still 11 <' 111 • •f ' • •
. II'
n1
,,,put 7 (tt'lllJll'ralu11•) and ,1 fuzzy co11d111n11 1.0 \ , r( w
, \d111ple>. .
p
·T
1.0
Fuzzy Condition
I

0
20 30 40
_,,.. ,.,j
Matching fuzzy input with a fuzzy co'ndition

This calculation can also be represented by tla>funcLion

M(T, LOW)= Support min(PT(x), Pww(x)) (2.16)

Two types of fuzzy control rules are widely utilized for most reai applications. One is fuzzyr pping
rules and the other is called fU'l.,ZY implicatio11 rules.

Fuzzy Mapping Rules:

i·uzzy mapping rules provide a functional mapping between the input and the output using Hngu1sLic
vc1nables. The foundation of a fuzzy mapping rule is a fuzzy graph, which describes the relationship
betvveen the fuz.zy input and the fuzzy output. Sometimes,-in real applk 1tions, it is.very hard tu Je1·ive
t.t"rtain relationship between the input and the output, or the relationship between thos i.,puts,1Ni
outputs are very complicated even when that relationship is developed. Fuzzy mapping rul!:'s _.re a bood
solution for those situations.

Fuzzy mapping rules work in a similar way to hum;:in intuition or insight, and ead1fu· ,:y m ;pp,ng , ul,
only approximat<'s a limited number of elements of the function, so the entire func-ti, . s!i,,uk. lW
approximated by a set of fuzzy mapping rules. Still using our air conditioner system as an exarnpl<;>, a fllz:z •
mapping rule can be derived as

IF.':he temperature is LOW, THEN the heater motor should be rotated F,\<;T I or orher i11put
lemperatures, different rules should bP developed.

For most actual c1pplications, the input van;1lili•s are l01nm1.mly tnnt'e th,111 01w
ci1mension. For example, in our air condilioner sy lellt, the inputs include both l'lltTent winpL"r a,trl' ,11l.l
the change rate of the temperature. The fuzzy control rules should nlso lw L'XLL•nded tu clllvw multiplt:>
mpulS to be considered to derive the output. 'l'nble below is an example of [uny control rnl-=-s ,,pplied
in onr air conditioner system.
 An l'Xampk ul tuny rules

......
T

t
IOW l\lf PIIJM IIH,11
I.OW bbl Mt I 'lllt-1 f,11 lllll
MF,DIUM F,\ST ',I <)\\I '.lllW
HIGII -.l>l\1111 s111w ',l,ClW

. ,•n The.,-r.nc , w. s ;11HI ,·11lumn.s r,'I 11,,(' ,(, n1 t11.,1 11111111-i,, h,· t.•1111w1 ,11111,. i11p111 .,.,,1 th, , ' ,,
'
,'
,,utpu' l ..,. 11Il' ' 111d thos·->•1npuls,llt'tt',1I1"•I ltlll•t•.11tsl11IF· l'lll,N111, 11,,,,111•1'11111,
1np1
l,H1 1'' n111 1d1•rc•tl ,1:; a third dmwn,1,111,11 v.11,.,t,h• 1h,1l i \m,1ttd ,1 1111', 111• I r
l1hl,••1·pHN,1Nlun•) 1' . 11· •1,'.h cd· um11 (cI ,mg!' 1 th'"' I1ll' '11111111,1111,e), ,Htd llt•,11 'l ,,. 11,1111•1•1, ' , , • ·ti
_ l 1"7 P•11t Ill 11 I IIEN ntll'S. 1-111· l''\,nnplt· wlw11 till' rmrcnt 11•111pl'l,1l1111· ,s I.OW, ,ncl lll' curr Tl
' 1i.m l' rate ol 1,1,· 1,•m1wr.1tun• 1s ,1ho LO\-\', the h,·,11,•1 moior's speed ,111,111,I Iv· I,\ 10 111r1
<f,
ll'l\\pt•ratun•.1s s,wn ,1s µossiblt• This can lw n•presl 111l·d by tlw IF-Tl IE'l 111, ,1•:

n
11• tlw IClllPl'l ,tl\11C is l O\\' ancl I he chan1·.•· mil· ol till' l,•nipt>r.ilure i', I.OW, 11·.N IIH'( 011d1><,\f)I\"
,l\ll put(he,\l,•r 111, 11•1 spcNl) should be F.\:,r. All olhc1 1111, s follow ,1si1111l,H -;11,1t1·,:y, whi,h 1' \ ,.,
v II\. Ir toJ hu111.m 'w111g', 111tuilion. ln tlw;.1i1 nmlltiont:1 ,·xampll .s tol,11 0111,n'- rules 1,c,, ·.ti,
1wI 01 thos,e1pph(,1.io1h thal need high co11tcol .i,..:ur:l<Y, tin mput and output ,ho11ld be
d1v1d;•tl1111..
"re small segm,-nt-.. and more fuzzy rules should he applied.

t-11nv Implication Rules

'\ tu,,zy implicationrule describes a gcner:ili;:ell \og1.: imp!icalion 1,•l,,th>11 h,11 bcl',\u•a !.t, ..l .11•
,,uq uts.Tht' lourJ:111011 cf .-1 fuzzy impli< ation rult' s rh..- narrow wns ol fl\, t),1• 1r. I ••;:,. 1m11li< • 11
l'ltl,•,nre rl'l.it,'clt" da sical two-valued lngw Jnd 1,ntlttple valut·d 1,,. 1.:. :,tia U!>illh th, 1 1 ,, h, 1
,, ,t •m as Jtl e,., ll<.'. thC' impltc.1tion h IF tlw h 111p,•r,1ture is LOW, TIii·:-i u,e lwaH', 1n tnr " ti\.l 11 11
h\S' Based 0r1 tl11,;implicationand a tact: th.: tPmpe,,1turc is 1m;H. Th<! 1<• ... • e.11o•, 111,'•
1111th.•,t
,hmtld slow <hl\\ nor rheSLOW can be inferred.

llefm:zificalion a11d the Lookup Table

fh ronclu ion orcontroloutput derived from Uw rn111bi11aLion of 111pulu, 1tpul 1t1..:n1!>1•1 •;wp fun.:ll<•1
.rnd fuzzy rules1sstilla v;:1guc or fuzzy t'lf'ment ;, 1d I it1-; pror.!SS tn called fu, , 111fu c:111.<: ·1 c, 11•,,k.. th.,t
on, usion 01 h111.v output available lo n•;1I 1pphr;J11011s. a d,•t11n1l1< .,lior ,,.·ort!<:s,· lll'l'•il'd Tl:t
,ietuzzificationprnccss1smennt to convPrl tlw httl.},,utput b;:1d 10 tlw, ri,p ur rl; s,ir,11 c.utµ1. toth•·
Ctl 1·rol ohjen .·t• •member, the fuzzy c,,11du-;1011or 1111•p 111s still ,1 hn1:u1,;1it ,.,,,n,1\ilC',.tr tl thi·, 1,1g111--t,
,•;1r1,1bleneed-; toheconverted tothe crisp variable v1,1 the deluz1ilkauon pr ,r :; .T,,rP" d1•t1r1..1ilir i.l1r
1
1 d111iques artc' ommonly used, which arc: Mean of Maximum method, Cente1 >f(,rav1i:, .111,i L111.
,,,.:.h.•t

I Ieight method.
,1ean ofMaxhnmn (MOM) Method
Th1.·MeanofM,1xrn1um(MOM) defuntfk 111011 nwtho.l ro1 putes the ,ivM·.11:1• ot •ho·, !'.1,1y•:Ni·,• ,. "' 1
1
1
ou:puts that han 1hrhighe,;t degrees. For t"<nmpll', tht· fuay conclusion, : tht·hc•,1t('r n1111111 v h ;ot
1tl•c
FAST. Byusingth1.MOM method, this defuuifi..,1t.un t.111 beexpres-;t>d a-.

L·'
MOM(FAST) = !.. !L
r
,,!wrcTisthts· et of output x that has the h1gh(-Sl degrcf inth sel h\ST
 ' ,t I·•.
l 'l"
!,\;l,01I('(1111
'1 · I I. , l,011,11 IJ1 1,1uur1· lwloW ' 1 , . liiri l111111Ic,11
I 111,•111H I
gr..iph1c l'l'prt•st•ntation ol th<' MOM nwlh•H s ,-,
t-'11)'•' 1method ts th,11 it docs 1101 con id1•1 I J11• 1'"''"1 . ol th•· otll P11 !UII'11
n ..·.,, ''JC";
'.·1.1.pt dei•n·<•S• t 1 111'11 .,tl 11d w,, 'I
,. 111
.1•incj • . I 1 11 , 1lw ltw111 ' I. ,re, '111'
'
it only l,1k •s L·are of lhl' 11oint s lh,11 • ', I l . , lit"l1t• t < L . '
• 11 1,,
s.ltlll "
'1lembcn;hip l1tnt·tio11s th;H have dirfl•r(•nlliapv HI •
produce the sauw result.

(al MOM m<'lhod example (b) c:oc; method example

Graphic repres1•111a1ion ol ddu:t.:t.i fit ,1lion l<'l hnlqt1,·

C.-nte,·ofGravity(COG) Method ...I I t·I,·zl'lf in

··i·1leC.enter .o,l.u.r.avtty method (COG) is the most popular defu,:;litica.u•on tCa!CI1n·1cIuc ·•w_d t!i '•\.'It : Y, u1
•. .. . . . • .·
. tL • ,·ofgrav1ty 111!hy s1c:s.
,,t t uI apj 1 pc.it1ons. 1h is method ts s 1m llarto the formula for ca1cu1a t ing Ih e c e1i ' .· . d
1 1 1
The weighted average of the membership function or the center of the gravity of the areaboun0 " ) Y, e _
0

membership function curve is computed to be the most crisp value of the fuzzy quantity. For C"<cl nplc
tor the conclusion: the heater motor xis rotated FJ\ST. Thf' COG output cnn he reprc1s1•111ed as

COG(FASTl"" '
L _LIF.iST (x)

'l"hc Lookup TabJe

The tennirwl product of defuzziflcation i!i the lookup table. Defuzzification tleed,'to he perform,·d f1,),r,ict.
,ubc;et of;i nientlwrship function, both inputs and outputs For instance, in the arr ronditHHh:r s• ,ten, on,· l'<.!
eds to pt'rlor111 dl'fuzzificat1011for eadi c;ub (•t <'I tP111p,•r,1ture input sud1 ,,., LOW, !'VIFOll .,, ·•11d I IIGII
k,sed on the d'SS<xi.it1:d fuzzy rules. The clefuzzifirc1t1011result for e;ich subset: needs lo be stoi-cd
i11 the associated location in the lookup table acrnrding to the currl'nt temperature anti l{<t11pu·,11t11,·
change rate. In the following we use the air ronditioner system as an example io !llt•st.rate tht.>
dt"f11ni11cntio11prnccss and the creation of the lookup t.:ible.

To make this iHustration simple, we makt' tw,ol';<;umptions:

l. assumf' H1al lh<.: membership function of the th,rnge rate of the tPmperatun, c,111be dt",111l>Pd cts 111
figure below;
p. only four mies are applied to this air conditioner system, which are

1) IF the temperature is LOW, and the change rate of the temperature is LOW, T11EN the heater
motor speed should be FAST
l IF the temperature is MEDIUM, and the change rate of the tempt-rature is MEDIUM, im.N ihe
heater motor speed should be SLOW
: , IF the tt•mper,Jture is LOW, and the change rate of the temperature is MEDIUM, Tl IEN
the heater motor speed should be FAST
11 IF I he tPmperature is MEDIUM, and the change rate ofthe temperature is LOW, Tl IF the h,•at<'r
motor spet!d should be MEDIUM
 r-; --
I'
1
.:_:'h:,::e ,n e n b1 e:!r s h lp fu n -
ctronol therhange rate oftemp,•r.t1111t•
INPUT-r
emperature change rate
1.0

0
• F/Hour
0 o.s 1 2

th
d •fB;ise o, (' ilSSUffiplion made for the membership function,llld fuz,y uric,;, WC can tllustratc !\iis

t lllZt t1c1a1_t_ron proce-;s using a graph Four funy rule<. can he mterpn•fL'd ,1<; lunclwn.,I cliJg1Jm,,

;i,
1'lOwn hgun• ,'\s an example constd<!r tht current rnput temper,1ture t<i ·, qr ,1nd ti•• 1-- •n e r. t•1 1:
!H' l<!mperalurc i I qF pt•r hour From 1-igur!', It nrn be found thilt th,' point,; ol 111ter-;ect1on hetw,·cn
Inc temperature values of 35 qF and the graph in the firstcolumn (temperature intntl T) hav< 'hl·
lllcmlwrship functions of 0.6, 0.8, 0.5 and 0.8. Likewise, the second column (tc·mperature d1<1nge rate • T)
shows thata temperature change rate of I qF per hour has the membership runrtions of 1.0 0 4, 0.4
and The fuuy output for the four rules rs the intersection of the p,111,.•d values obtairwd from the
graph, en the /\ND result between tlw tt•111per allJr<.'input ,ind tlw tempcr:iture change r-1 c
111p111 /\rrnrding to Equatton this op1•1,1t111n rt ;ult ,hould lw min (0.6 1.0), 111111 (0.8, 0.4},m111
(0.:,, 0.'1,Jind 111in (CJ.I!, I.OJ. wllllh producps to!)(,, O.•l, CH ,rnd 0.f!,n·s1wrtivcly

LOW .t, LOW FAST

/0 JO
!/ \0 WI IIIJ,J IO(JU

MEDIUM
MCDIUM SLOW

o•_u_ .. /\ ..l
10 ) -1, ,,.1_00_ .,..0. 0- -

LOW ·,. MEO/UM '

•
o.s

I
:0 JO
OJ
f
6 I\ l 1
OS
o•
(.00 800

MEDIUM ·ri LOW

OS fv1ED/Uf_..

I
t
II :\
,......,
ol
I
&;\
30 ss ao G 300

:\n 11lustratior. oi fuzzy output calcul,1t1on

 X

Off-line and On·liuc Dcftnzification t v tit"' u l.11',

As• mentioned , the dl'funification process is to <1•t 1.1·1•c th!' dcsirec,l crisp 011tpll
\\,' h ti ,t • fi1li
.
,
II "·.
,nd
'f'H'
on I• 1,t
dcf_uzz,,•1·cat1•on techniqu,• 1hat comhin,•s the nwn1I1ershi1> h.t1 <11011• 1ir, ,:t•,,.1.11!ti '
1
111 •H
r·r
clt"lu·zz1•,•ic 1t1• 011process , ,111 tw further divid('d i1110 IWll r,i Le" oortl"• O
rlPfun,fication I
• ..t , f 1ncttld!!i, ft 1. y i"U( ,..
So-called oll-'111(' dcf11nitir,1tion me 111s I.hat all inp11t ,ind outpu 1
111emht·•· h P ' " • . . tl• Inc
rnd n 1'1
01 I ii ,pplic ,,1· ,.i. _ l
thc ltJokup t.11'!e should be develo;>ed based on the . tie 1"· tint .11linµu: ,,,ul iwlp:i
m Iestimations f h
Pemt>ntallon o t e fua:y logic technique to the aclu:tl app11·ca 1 0n• Tiu,•plt.',lll l .',nd n•,tplli ·;ir tel•'I
1 t '
, '
111 111 1
mcm!wr,hip functions arl' developed ba cd on the :ictmil experienceoi· thc • .,. dcfi11itwll', nf lllP•_11
ll1 t110 ,t
'anges o( a spcc·rh•ed application, and the lookup table:, calcuI c1te<l 1• 1 _ te, r.,m•, ,, f1,1r-'.y
1 • 1 l
lliln ,,,,,, r;,1,i ''
and ou1p11t membership functions. The advantage vt ,his method 1s t.i,-t '. ·>- . ,,re• ,_, , , ,,,_ 111'1<
1

lalndations .,n• perfornwd prior to the real implemc11t,1t1t1n and thl•rt•tore •h.- t.i,.1J • ti, e timat10,1 ol
rnn urning. rl1r disadvantage of this technique is thaL the fuzzy output is onlyb,,se onl ltl
111flllt ·amI output paramcLers, so the conLrol accuracy is 110Las
as th atr J fo••1-Jinc met im •
lu·gh I ·shir, fun,·uons are
lhe on-line
_,d · method has .real-time controllability. l;loth input .and output memhit) r-,h,e , 1,111·nt·· •· .•111.'
d eveIope,., urmg the real•t1me processing of an actual application. Alsc t1ltc 100' lei • 1
'
. . re
. . I d onl)' f11r1v 'If es .:i 1
c.,I c:,Iatccl. ,n r<•al-t rme based on the current actual input and outputs. In th is nwt ,o • , . 1
d cwI oped pn·or lo lh e renl applications. The advanlilgl' of this method 1•s l1tat I ll•g 11t'I•',·0111rol •'H'<"t11o1rc),c, .11
0 111
ht• ob1ai11<'d for a process ,111d the fuzzy outpnt c,1n hl: computed in rral-ti11w. '!'he dis:ldvnnl,lf\\' t '
• h 1 .
me ti10 d 1 st ,II a onger pron.>ssing time is needed and 1 1 1s a somewhat time-consurrn ng • )J'PCt' ' l lnwcvt•1·
with the d1•wlopnient of new computer technologi...s, today much taste! CPU, are av,ulablt•, auli
processing linw i-; no longer a big deal for this method.

l\n;hitecturN, of Fuzzy Logic Control-.

Combining thl di.,cus 101h ,n· made in tht· r1,·v1m1s "'Cllons.. 1 ,11 ull111t 01 .1rd11IL't i ur, of 111uy lo,,1L
rn11trol sy,tc•1111 given here. /\s shown m l;rgurl' hLlow, wl11<h ,s ,1 typkal h11,:,,;y rlosld 11,0:1 cu,,u·ol
system, the' lnpuls are error and error raw, wl11ch ,1ll' wmhmeu hy IJloc-k M to 111put to thl' f•1n.y 111fr·.-\
·111·t• sys,,.m. The lookup table 1s derived based on the nwml L'r hip lunC'llon ol 1npuh, Ilic lJlllpul .111d
the fuz:,,;y rnntrol I ules A 1011trol garn factor G 1s used to tum· t!ll' output ol tht• look:,p tJl>IP to ,
h11i11 dtffcr1.; 11
ouqiut vaht<'S. The interpolation block S is used w s111nc,1h the output cl,·1m·111 ol the l11olrnp ,::ible. ,\
f1•c•dback sig11;1l is obtainrd from the output oftli,• syst,•m

G= ControJ Ga.ill

--- l_ookup
I Table

I Coubol
s

fiiure 2.12. Blc,ck diagnun of a fuzzv cont101 '),·-.tc:ni

 I
0

I 01
,l s\ sl, r I control c1rrnr·,ic\
mu1llp}e look.tp ,.11,lc, 11111." "ntl ,l ,l< 11 1,
111ttied ll 1 l , l l .Ill ell.., er
hiuhI '.v.o \\ lirok
h1, hrs , h m . .
\ 11111 I rgurt•2 13
up t,1bJe • '
app ll:,H1011, thP , s an• developed 111 this rm1 1• st. .
T h, hrnn. S\ rtch between Lhe coarse nlv em,a coarse anu ,t lltw t,1hl,., llu11nr u,c
1
v1a Ut• can be defined by th • anc_ 111t' uwl7ble 1sunde_r the co1111, or tlw 11,pnt, 1rn1 ·mt
C)
.in
.M I
1I controln J es arc• utihn•d anth·. .
e u c'r b c 1 s e_d on.th' re-aJ app11· c a t 1• 0 1 w,, ,et mt 111\ ·,·,hiJ nnc•·Ju,
the SV-.Sh•m 11H' -.-.• IS SY• S Lem Lo s'au,lv,
r.
equ1• re111ent o{ h1/ 1' 1<•· n111·rol ,. ·•u ,, , \ •lwi·
ds.q111ck ,,1.-.
1',ponses or quick ·ac ttons,t h,enarse table is used•.\\ h• en th •• • 1 •
cont10 I acruracy . :•1 "' '' l ,,,
, sm "':'
th •11 mor.e
nwniory·
11
a11 control errors• th e 11·1_1e1 ouk ll" I I1b e• rs sel cted. , 1he :1111tr,v• fu1 tlw, n1t·th1.•I,-.;
ma kl' 15
needed . ct ,Ille tables, and a liltI, l11n:ie1 t1t11, 1:, 1,eedv<' ,1,
dcc1s1on inselectin b to. store both coarse an
• g ta le 111 terms of the rnput en 11r limit value.

Luru11 fo,
fm.-f

1,Xl'l,J\I

" I 111111111 '-Y d r,ncd 10 ,It , 1111111n, p11y,,,1 1-,h-111 ",

1l1,ll lhh 1>y• w Arr some .: 1oion ut " " , l·1121v 1.111,11111
< 1111110I ,-,,11 ml

.,, 11111 1111,111, 11\ r 11 II t01

, , lu ,
1111•pl. o Jlf,ul .i
\, I 1h
,1,h'I - t c , d knf>Wlccl·: ""' / h,1•1inp
l•n.:11 l'• mpu·-ou1 1rules nt 1•1.tr.i't!•'
,'\tI,•nce o, , .t n - I alU:.t ht> assure l re exlSl.S a solu on•
•
L;,,,,d d?Oll;., , , uon 1, enough - The cor.rrol en,,"llleenng must lc,or. for 't'1',o cnc,1.;:ti' %! ,uor
r.lthi'r lh.rn. l '' mm one
• !bngl" of prt --ion - r-uzn leg c cor.t er
ill't',,lllll
• '"•n, l: • ,ln6 ,. t, and opt e .i of st , t r1.il f Thl'>l (J ' I f>l
ty
,ft ,gmn f.1 : lo_h con roller ra:.lier th:maddro.,_ c ::xpl utij..
 Arcliitecturc of Fuzzy logic Control: .
T h
e
rOII owi.ng di.agram shows the architecture of ' (' wol (FLC).
,oi
l,()gtc

Fuzzy -: ..... .,
..'

()uti,•,t 1,,4r,
y 'lr/,1!Jll!t(

i\1· • • c - ,
' ·,<lJoi omponeuts off LC: f·ollowings are the major components or t h e ,,. , ns"
r•Lr . hown rn tllf':,ibov • fiw 11·"-

1. Fuzzifte1·- The role of fuzzi11er is to convert the crbp input values into fuz,zy v,dP(;;,. . . "' lt
2. Puzzy Knowledge Base - It sto·res theknowledpe,about all the input,output<nz7Y•,· ,!, ,,v,psP,p.•t.L<
also I1as tI1e membership function which define•s> the input vari.ables tu t.11e
• t11z·Jzv ' r·1lf' 1'"'-;;,
•
an<1 "
,
output variables to the plant under control. • .
3- Fuzzy Rule Base - It stores the knowledge about th'e operation of the process of ciormnn. .
1. Inference Engine - It aqs as .a kernel of any FLC. Basically. it sirnulates llurnanut:1,,;,,:ow, by
µerforming approximate reasoning. .. . ..
5. Dcfuzzitler -The role of defuzzifieris to conv:ert the'fuzzy values int0 cri p values getting fromfuzzy
, inference engine. • •

Steps in Designing FLC

Following are the steps involved in design.ing FLC:

• ldentificaliort of variables - Here, the input, output and state variables -µmst lje identified.of the p!an
which is under consideration. • • •
• ·Fuzzy subset configuration - The universe of inf rmation is divided into rnimbei• or f:..!t'.l)' s,)bSi.!\.$
and each subset is assigned a linguistic label. Always make sure that these fuzzy s,ibset:, include ail
the elements of universe. • •
" Obtaining rnembersJ1ip function - Now obtain the membership function for 'cdcl' tllt'JJ "'.'St'I th11
wegetin the above step.
• Fuzzy rule base configuration - Now fo.nnulate the fuzzy rule base by assig.;ring
rcl,rtionstHp between fuzzy input and output. •
• Fttzzification - the fuzzification process is initiated i.n this step.
• Combining fuzzy outputs - By applying fuzzy approximate reasoning, locate .the. fuz:1.y mitput aul
merge them. • •
• !)efuzzification - finally, initiate defuzzificatian prqcess to form a f:risp otitput.

State the Advantages of Fuzzy Logic Control:

■ Cheaper - Developing a FLC is comparatively cheaper than developing model h«"sed or otht,r
controller in terms of performance.
■ Robust - Fl.Cs are more robust than PlD c9_ntro1letsbecause of their capability to cover a huge range
of operating conditions.
• Customizable - FLCs are customizable.
• .Emulate human deductive thinking- Basically FLC is designerl to C'nwla\e human dt'cht<.tlve tlw1ldn ;.
thi;-process people use to infer conclus[on from what Lhey know.
• Reliability - FLC is more rc>liabk th,rn convention.:11control system.
• Efficiency - Fuzzy logic provides more efficiency when applied in control system.
 -. :ipll<',l_tum.s 111 Bus11w:-s: In hus111css, ft1z1.y 111.,.,1,·<. is usc>d in Drcision-111,1king ,;11pp1111 •, tc"• ,,.,
\ • u.111011 in .1 l,1n:,• romp.1t1\'

Apph,.Jtions in lklt'll'<' In' tI •Il •11:·;, . t U•tt} log1r 1s used in Undcrwall'r t,1,-gt•I 1,•111r:11111n•1, i' 11t,11t,1l111 , • •
1

. .. • • • l
Il' l'• 'l ll\lOII nl I lwr111 1111111 11 • • I 1mJgcs ,
, ,,,w.1 Itlt'l1,1on <;11pporl aids, (on11 nl 111 .1 hype•velo•·l'y 1111•·1,cptc,r
u1.1., ,.,•t moddmg ot 'I,\ 1'1) d,•ds1on m,1king.

\pphr.1twns in Elt•ctronirs· Inl'lcctronirs, fuzzy logic i), used in Control of .111tomatll" ci,.l"'J•lfl
1
111" d
:_:11_w1a,., to 111a111t.1in llu11111hty 111 a dt',ln
room, Air conditioning v..,1,•m . Wa;,l,1:.• 11.i ·llll•l 11111r
•:lll'O\\',l\'C O\"t'llS, \'ar11u111rll',llWrs.

A 1phra11ons anI
111111c1 I 1 •lh 1111.rncc field, lua.y logi, 1 U'it•d 111 HanknntL' 11·111•,lt•1• r,•1 ,I, hti <
111,rn,1ge111cn1,Stork 111,1rkt l pn•dicllons.
\:1phc;itwns in l111tu,tri;il Sl'llor: In industrial, fuzzy lo 1c is used in C1·mcnt kiln control tw,11 "xcha1,1;,
rnntrnl, /\ctLv,ltt•d sludgt• w,1stl•watcr treatment proress control, \V,n,•r purific.,thlll pl,11• cun•r
Q .. 1111it.111ve p.u tt•r,n1nalpi-. tor industrial quality ,1ssu1ann•, Control of <Om,t1.tint sat,sfa u,111
1,,obit1:1
.;1 t1rtu1.1l des1g11 Control of\\ ,\ler punticat1on plant\
\ 1pl:rations 111\l.111ut.1etu1ing: In the 111anufart11nng i111h1stry, fuzzy log!c 1s 11st'cl 111 Optimi1.a111111 01 r,,, cs
1
1•1od11ct1011. Opt11nizatwn of mtlk produrtion•
. \pphc,11i,,11s 1111\1,1111w: In th'-' manne field, fuzzy logic bused in /\utopilo1 for ships, oplirn,,1 ruute,•\,!
,•m t.ontr!ll o,l1Ulo1111111011s unden\ ,ltl'r veh1clt•s, Ship stt•ering.
\µp!i ·.lltons 111 1\.hdi1,1l lu•ld· In tlw nwdi1,1I ft,•ld, tuLZy log,, is usetl 111 \!edtt,1i ,hag11t,s1,,· ,11111' 111s
Control of ,1r1,,1·11l pn•s-.1:n• during anesth1:51,1, Multiv,1riahle ro1 ,rol "' a1w:;tl11•,;i-,, imll'l
twurnirntholog1r.1l finding-. 1•1 Alzhe11ner\ p.1t1enh, i{,1d1ology di 1gn1 ,e!>, Fu11y t11fl-1•1H d1,11 ,,
1:11b, ,.:s .11 d p(l.) t,1t<,·.',mcer.

t\ppht'<ltions inTr,111sport,1t1on: In transportation, fuzzy logic is used i,n\utomatw umk,1-\' 0t111d L1 1111
• ''.;r>r;it,on, Tram schedule nmtrol, Railway accelcralton,Braking andstot,ring.

,\i:plt,,ilioll'> 1
P.1tternRe.-ognition and Cl,1sstf1cation In !',1tte1n Rt ··og 1,tw11 .md Cl,1<:,ilic :11.,} I 1,,
•l'P
1 11
, ,s •d in Fu z y log1, ba l•d spel'l h
rl•cogmt1un, h1zn logic hc1sPd, H,1mh•. ntllli,.:1·tco1;1· "· 1,Iu l \ ,.,,.
i 1 ecf f,1(1.il d,.u.icteristit an,1lys1s, Command analysis, h1zzy im,1ges a,,h, ,:11111'1.1I .11, J''"t 1 n , .i
p c'.·t•ntion ba:. d 011 fuzzy logicre,1soning.
 -, I......................... ,,

, LECTU . , ·JC('.l5AO:thll,)
1,0 ·'
RE NOTES FOR NEURA.L NETWORl{S AND
FU " zzy (
UNIT- 5 . . , 5y<;_II:Mi
•-
£l!ZZ)'1nr-,r" Fuzzy I.OGie APPLICATIONS TQfilillRL
' !1£.LEMENTATrON FOR INDUCTlON [1,10J"Oll CQNTJJ..Q.1
\
cctor rontrol I . ,Ii1 ••' 11·1••· 't'll lt
I'.I '[I • '•• ,
,, '1 I , <no\,n clS f1eld-orient('d rontrol (HJ() h ;i van. bll- lrequr1H> • j l ·, rwc '" th'• 101,
f '
·' ere l ie stator cur..'ent<: 0 a three-pl,a e t\C ele cu 1r 1110Lol' are •kro,\'lt <•W' , • 11
c •n 1 • ,.: 11(,t
'' iponents that can b<' .•• . 1• . f ti1c•' ni.i<>nt Lie 1111 • 0
.ell >l, •
, u 'tsua Jzed with a vector. One co,unonenl de ' <> 'U• f<- fl' '
incs1 111 I e llu)\ ,ll1t 1 1 ' 1·
1
., PP ementary 1s the to 1- 1-h • 1 , !'·'
SJJccifi d b h. que. e control sy!:item of tbe ctrive calculate ' ' , ,,; fy111c .11
ro
1
. ,e_ Yt. e drive's speed control the corresponding current component • e,f-•' '.'1 .... 1 ,, th..,,'
)·c'IO on,onal-1ntegral (l'I) controllers are employed to keep the measured current rol; ;, 11 ' •,- , 1 to
f'l encc values The I •ct i I- ·e d n 1 , " '
•
S wi .,,..
. • pu se-w1 th modulalton of Lhe ,·ariable-frequer,cy L iv • • '' ,,.. .., r-.,r. ,...,
tc j ,ng according to th st
1 f • r,1 ·un· r' , ,:'l r<, 1'' • •
u<;ed to manage th AC
P. ator voltage references Lh;:;t a1e the output o tuc ' ' ' ., . ,. n<' •
1
e , 1 , , , ·d Jo' .. .-.h 11'1.01, ,1 •
1 10
motor apphcatio t' sync ronous . and inauction motor . It was origmally. < CVL ,1,c l .. , !' . f 111
.,r1tc.
ns nat are requ ired lo operate smoothly over the bu1st ing SP< e( 1'1,11,11·' "m' •ne1 •l
lorqufc!at zero speed, and have high dynamicperforn·ance includino fast acceler,1tion a1,d Jecelea, uon.
H o • • I · • · " " , • •o r •s 1 •
_ . _ w e v e i , I t is Jec om mg mcreasingly attractive for lowe rperformance appli-cations aJ:;,, d.i, ,u '' ' : ., _ -
0U,1
1
:size, coSl aJ1<l power consumption reduction dominance.It 1sexpectedth t with 111crea!':mg comp,i c,non '.
power of the microprocessors it will ultimately nearlyuniversally relocate single-variable scalar volt<;-pe•i
llertz (V/t) control.
f'tJZZYLOGIC CONTROU,ER

1 uay f()gir provides a ,-turdy framework for achiev111o robust allC\ simple solutions a1nid difleretil
,ippma< hes of intelligent .-:ompulatJon. Fuuy model is a ,ollecuon of IF. TH[N rnlcs wilh i11dbli11guhh:1llic
fllndrc,m•s that ll5t· a fuuv reasoning such ,1,; Sug<'no and Mamdani mndels. '>ug-011O type syst('lllS c,111 [l;
11',cd lo 111odcl any inferPJ1Cl' sysu:m in \\'h1• I the o nput mcmbeiship fu11rt1ons ar!' Pirlwr line,11ot e,;nst.inf
wlw,r•,1 M.1md,1111 typl' JI oduct>s eilhc:1 111 ·.1 ,, uni nca11ulplll. The !tizzy logic cont roller consist:: or tou,
sr.1gc:s, f11zzifl<:ntio11 <JI 111puts .1nd dcnvauon uf I tiles, mfen•ncc mech,1111 111,111d d· -f11,..(1ficatkrn. 1-'l1zzy
lo;;i, S\ ':lt·111 ,11<· tOl!l'cltvt- 11111ttcm approx1m,1lJ011:; In gt•ru.• 11 the g,,al of ful.'I.) 10,.:1.: -,_ysL'-1111s l·.1 yield ,1
st::l u1
Ot1lp111s for g1v(•r1 inputs 111 a 11011-lirw.11 systl'm, lack.mg using ,rny mothl'm,1t1c,1I 111,>d"I ln11 liy it1:>i11g
I 111g11,,s1c rulC's. A gener.tl !,rock diagram of Fuz.7-y logic 1s!';hown Fig 1.

!i_YSTEl\l MODEL

""' bl l k mo ci,1 ot· the induction motor system with the conrroller be developed using the: p:iwe
I lit' C . .
sy<.tt'IJ\
I t•r ,f'dl. omc. s.c,,,111,•..- -.1 •s.vste,m signal proces.sm1o1 too!buxes & trorn the tundamenul
1<1\\
r b,t
runtl1011'> I\ ,111..
.. - klt -aiT. 't1 \!atlab / Simulink. In th.s p.;iper, plots _of vultage t< rq11e, speed loc1d & .. 1 ''. ·ca,e
, 1I ll . 011
11
•r th-->S
' . J!llll
1
I _. of I
ti •,w• with the controller and the wavetorms are obs<-r, LO on the eq,11valei.t scope
• ·,ms The entire. svsrem modelled .m S'1muh. nk t.s a c' ,os e
· d . d I r
n lollt"(l ,1:, tUrll I
, . rn · ., 1. ' lf,Op t t>e ' -cntn.1
,
• ha.c
Ir"" ,·ll11n1n° ... . : 1,11111 _.,,1
Q -

• •
J c, • sti,1g of thenlants controllers, samplers, comparators, feedback systems, tllP mux. d0-mu,-,
5 'tt.:'' cons1 r • • . , ,
:, . . dders gain b!,'c s multipliers, docks, subsys•ems, integrators, stat...-spacf m(,CWIS, -;: ,::1t,teim,
:,UJlllllt'IS,a ·• '
 ts u
tht!
I\\
·Sug utplll 1
\Ot \ ... u
in ob"t
the ue t<'
settle .,. -alut c
uveness of the d . It 1 a
ac:s cun: s take Jess ch he
e ra n or r contro J

•
 r 11. I '"'
\HJ

1-· ,., ,,,-... "

J1f]l /!\ ·-ti
ii ,.

FUZZYLOGICCONTROL APPROACH:

Thfueuy logic apprc)ach offers a simpler, quicker and more reliable solution than conventionai
The FLC has three main 11/ocks. The first one is Fuzzification block, which rnciclLfies crisp inputs (input v
tediruqu..:s
lue, tram re:.1! world) into linguistic variables to enable the.inp\.1t physical sign;;il to.tLSe the rule-base
through membership funcnons. The second blocik is Rulc base, where fuzzy inputs are compared, arn1 the ;.·,i:
tMi!et makes the decision based on the membershipfunction$ oaf ch,input.The last one is befltzzificatrnn
block. whici1converts b,'Ck the fuzzy outputs of the rule-base.to crisp ones and seJec:ts membership
ftmctions fur the different control outputs from the rule-base..

.;_,.1.'.,.t..,..t.
i
 Convcnttonal l)('slgn \1eth,H.lology

------....
Understand physk.al system and )

D••vf"lop a Lmc,u Model of Plant,

Sensor and A u t

d tontrollcr

0 o, for tt
r

S l""lU ate, Deoug, anrl lmplemer.t

esi -- _ -
 r' ·:,c Speed Control Algorithm:
,S.ie,p 1:. The •Switclie( \ 1·e \uctance motor• !;peed signal l,s1111plecI•
,
,itcp 2• · (• alcul<a"t s p• eed error and change in speed t>JTOI'.
tep 3: Determine,,1",:,'r•.zy sets f•or speed error.
?"'P 4: Detrrmi\1e membership ror speed r1T01
t<'p 5:. Determine iui,:y set fm change in Sj1!'t'il error.
i 'r (,-:etcnnint ll\P111hl'r ltip for change inIH'l'd t'ITOI
Pp 7: t:mdrng control action <1s per fuzzy rul<' and r:,1lnrl,1Ll' error.
s.ql H·Sending rnnlrol COll\ll\and lo the syslel,lIl ll('!"c.1Jr-11latio11of t;:t'l"OI',

iPJ)Ul and output vdriablcs ol'fuzzy memb<'rsllip fnncUnn al C' selectedM follow<;,
PD-Positive Big
PM-Posltivc l'v\edium
l':,,-P0 i!ive Small
!'l8-N,,g.11ivc Big
Nl'd•Neg.1tive ivkd\um
NS-N<."galive Sm:111
? 7.ero
The tri;i,1gular shaped function is chosen as nicrnbcrshii1(unctio11s 13f:cause of tl1e best tontropt erfonp;mce
;;nd simplicity. • •

!_: hie 1.: Rule, based fable

EfE i'<l5 '• rts ,zo PS P&. PH
1

\},'f
' :'.\'B t-113 NB r-.'B NB m.1 N$ ze.
NB ?-413 NB 1\'M NS zo PS'
I
! '\'\i
:t--,13. NB Ni? NS zo pg: PM
r-•·-::-
I- :-,
r·---·.--
1_fo ffBI"'.\ NS zo PS PM PB
P-5 M
1 NS zo vs PM P.$
PM zo PS PM PB PB.-
NS P
P B
Fuuifi ation:zThis is considerc-d the first step to be programmed, the PLC uses linguistic Ve:': iat:ks i;1,;tr._,d
0f nurnE•ricaJ variables. So, rhc error input.signals can be assigned asegativr.) Very l3ig (NVB), Neg 1tiv Si
( :-rn), Negative Medium (N_M), Negative Small{NS), Zero (ZE), Positive SrnaH(P::;), i'os\i::ve Medinm (Pl\\),
Poc:rtive Big (PB), Positive Very Big (PVfl). The triangularinembership function is \ sed for fu,.-,jfkation ac:
:J,owninfigure. The process of fuzzification cotivert numei•ical variable (real ni.nnber) to ;i !im,(1istic
•,·:uial1li:' (fuzzy set). The change of error, \ti/hid, used .as th seconi:l input for fuzzy system,.,lso co1iverted
fromnumerical value to a lirig11istic variable atcordingtothe triangular membership. •
D1.,fu:zzification: The fuzzy logic rules genet·ate the demanded output in a li11guistic variahle:
the:;""vRrinblt>, must betransformed to crisp output (real number), This step is thP. defuzzillcation;
the nwmb •r h•p fonctions used in this study for defuzzification arc shows in Fig. 7. Tiv- out·ptt\ signal c:m
b0 a ;;ign •,d •._. [>,!rcn11 iy Low (EL). Very Low(VL), Low {L), UnderMed.ium (UM), Me<Umn (MJ, i\bove
Mt>u,um lAMl. lH h
(H). Very J-ligh tVH), Extremely High (EH), which represent the modulation tndcx (m1,,.-.;here (0 s m 1 ).
Tlicrc ;r, thrNi different methods can be tJS-ed for membership defuzzific.1tlon, Center of t\rt>a(C(>A}
Hisect0r, or Middle of Maximum (MOM). The center area (COA) isonsidcrNl the most popul.lr nwthod, so,,
is used for defuzzification in this study, whic)l is presented ,in Equation,
.(• ' .. L1\,1p(uj) <tJJ
u n.r - X"'fl ,- ')
6,-!J /!tHJ
11-a, d(Jt,ting tb· J'),C, tlw SH.M cun-enl Lratk tlw rdcr1•11<'li' inal with ni1ii,t>,:m va:U·,i1 ;a1Te11t r l'l'i '
u mJi 1r ng wiLlt uwJitional curr nt contrnl tccb11iqves1 aml henu•. lhe tc·rqUt 1 ipph's duri1t;:. tlw r.rnciuclirni
t>t•.-lodc,Fc,ac.:h phase of the motor were 1•ecl\1ced. The controllel' was tesit:d <1l difkrent lo,1d conditions 111d
:<V,th dJ! rellt turn on aJ1glcs.
 n--
Understand phy!>ic --- ....
and control requirements ;
>
L
sign the contr lle-r ----.

- Il
ng fuzzy rules -_.,)

---
Simulate, Debug and implemen
he design ../

l•.;zzv Logic Controller design is a three-stage process. lt comprises of fuzzification, inferC'nce mechanism
1 d dC'ft1Z7ification st,iges. To design the controller, fir tly, membership Junctions for the ··rpt,l vai 1?.ble
i c•1-r,x, .ind integral of error must be specified). Secondly, the fuzzy inference system must be defined-which
, "'l!>isb of a sclies of "lf .....thcn....." linguistic rules. Then finally, the membership functions tor tl1e output
:11usl be selected. A structure of a fuzzy logic controller 1::.shown in figu1e below.

Fuzzy controller architecture

[ Rul Base J
Ref.ysign;I -.rL
Inference ► Dcfuzzifier
_ - Fuzzifier Engine

Je

Tilt fuz; y logic tnntroller has three main components:

1. Till· inft'1ence mechanism incorporates Lne rule base which holds the knowfcttgc in tlH forn1of ,1
sel
·,,rrule!- of how
best lo ..:onlf'OI the syst m. c1nd l'Valuate which of th<'coat rol rules tire r< le\, rnt.11
rile wrrcnt Uine and then decides whaL rhe rnp11t t,l the plant should he.
2. 1'he fuuifiC,llion interfi.lce that simply modifies the inputs so tbat they can he interpreted,rnd
, ornp;irrd to the rules in therul_e base.
3. ;·t,r defuaitkation interface thaLconverts the conclusion reached by the interterencP mrchanism
·nto lhe inputs to the plant

F()r theiuuy logic AVRcontrol, the error brtwe<>n the reference voltage Vret .:ind the termin:11voltage Vt i.r
v, ai'd!'If integral oCrhe t>rrorVI which i, thr difference bt>l\••cen tht i111rnrcti:1tt- ,incl prcviou, '.'oltagf'
nror
, . J !a!
,.va:ues e r,1!·<:idered
th as
, <n. l t ' ! ! . c t l O theIr:inputs
I Vt. to the fll7,,y
thb controller, conrro!lt'I'
dE'ven whrll· tlw Pulp11t is thept·1·ihl·d outnui
luzq suhs1°Ls
1 1 'r,
c11e cihl!:>l'll.
\ .I. ,
 Mcmhcrship fu.111.:lioi,s for the error

---
'I -- -
'J
V NV
NV NL
NV NB NM
NV
NS
NV
zNV PS
NL
-- NV
NL NV
- NV NV NL NB
f-.

·- f- -- NV -- -
l' \)8 "JV NV NV jty_y NL ,_.J\::B-rNM
Nf 1 '---
NV I NL
·--··
NS
I NV N .'
vf-
·jNV
I
NB
I NV ·~ z·-
izNi NV NV NL NB NJVf NS .., ,.
-- NM
NL NB NS ··- l.
fv,
n-
.....
'S ....,... -,
,MN

NL ---· ·NM -
-- ,.....Z.
NS
_._ PS
l§."-- PM -
z
-
,_,

!p l\ I- NS
t -·- ..NI3 , - ---- ,
. M. -NMTM P S
!NS Z. . . . . ...PS PM I PB -_
p PM
pB
-··•·--
. ·PL
'is !7 PS I • -J-·-t .---
PM
p
·-
L
- ...
\'......L.{:_ I PS
>----+-- - -
PM .
- r ,-mrt 1- ;
-. ,.. _..t
PB

Defuzzification operaLes on the implied fuzzy sets prod_lH..ed by the inference rnt:ch rnism and 1oo!\1h!>Ws
the 1 effeci·s to provide t}le most cenain controller output which is Lhe output of the pla11t Willely. c;;nterl 1
,tn•.: method (centroid) is used for defuzzifict1tion according to the membership ltmction of the 0uqn1t
l'h affangement of the structure results in a co1r1puliltionally Jess lntenstv control .ilgorithrn. An"riv•r
signil1cant adv, ·1tagt oC luzzy logic b that it c;111t•a ily ;1c'("()nH11odarl' additin11,1l iuput signab F,t'n\ I ,,g
Conrrnf provid,·, ;1 rnnv;,ni<'llt means to dewlup t·hc conlrnller which ca11 ,in t11\\•11odat(" th( nnu •11\ •,11
natl!l ed the exc1tc•r•ge111°r:1tor system. Fuzzy logic on the other liand is not witlll>llt any d b<1,k, toe :.t'l
t),;1ck is 0bserved on the stt ady-state error which can be cleared by tuning the gains.
0
 I \l'I llli
I llfil' \!L,_UJE...DESlGN
-"LUI.ILl.-l.!Of. £!Uzy l&iolC CO . tITll.QJ,J,fJl.EQil ILHJ...J lrJFlNITE Bl.!S IU
YSTEM;
ll i, pos,1hlt' th ll ls .
t qut•1Hly sub)l'tlt•d to ,ud ' . ie oper,llion,,1 n•quln•ments th1• i • • •
li, ,lI
' rnnsd,tltined
d111by ti and l"l.',lctivt' powrr lo ,d
1lu,1011 ot '-tlUI
< •.re:>s I l',ll I nr.tma, h111emfim•ct,u, br. sy t.--,1,
rlwui•h \he lncl1t • ,1,11 loach ronnei:t1•d lo tht• g1111 1l • mg am also as•' ies111l ol ,;ucld,•n i•t• lusion o1
, ll
' • 1011.1 P1 ,110111011.11 1111•· • ,I (1'1l conI mav l x h1h1t h>I\ r 1 t:, • •
,II It\ ,md t.h 1 I I oll1.1 1111 I • •·qttt•tu.:y 'ow,·r 11l<1t1on,
· t ,h , ol prl'rt '• ,111c11.. th1 prolI• ti
111nl1l1l h•dllll\J\lt'' , • '>t' 111,tt1w111.11tt,11 mnd,-1, nl tth s, . •tm '' ,.s1.1c1.•tt•d rn,11
,nft•1 •n • •· .t t nt,,•,-.11, "' th.11 tlw ,,.,,. II•· ,1111, ti \\,111<1il!>th,t , ,n. . , . nt"llt,"•'
t l' s sll'm 1:- ,Illell1•1li\l' ah i m u t •, sp,111111",1,l,1!.' "I , llun I I rI re I- L
lht> ,.,,llt m,llhe111,1t1t,1I 11\lltll'I is ;,tl11•11:,1 .111l ,l111ll' 1111 • ,111h indle p1ohl1·111,s,s,,L, it• ,Jw1'l ,' lu 1sI, 1 w!ac
no ,l\',tt •• > 1• 01 ,,..1,•11 1S ,,1\,1·11,11I1C I'• ll'Jll 11;,1111,11• 11 l U Oil• e

UPFC ronti·ot archiH•cl un• for .in 18/1nfinilc hu-,"Y"'"'"

x. .. ,v
\· X V,
w;i- _rrrN, I

,.
- vsc-r vsc s \· . I

l '(

\ I

J_ i

[
V1,

♦
m. cS, m, 68

TlwI i.Clwing ,1 control ted1111que lo deal \\Ith approxim,ll1· data to arrive al ,111agreeable result it ran
lw put u11th:rservicewherethere 1s .111 expert,1t1on with ,1 n•rt,iin range or dt•grec of arceptability. In till
111,111.igenwnt ol UPFC, the l·LC can offer ,1 st.1hl<' oper,111on trarking the wt points imd kPeping the,1ctu,1\
v,11lws within cloSt' vidnity of the desired set pomts The fuzzy logic systt•111 in 1t-; ba ir fonn h,. t 1, '
tollow111g steps.

• r:11z1.yl1rat1011,
• lnll'rence and Ot•cision making using the rule bas1•
• Dl'luz1il1c.1lilln.
lul ol these three basic steps of the FLC the Fuzzificauon prorcss and the Inter •nee ,md <ll•n-:11 nm 11'11•'
pl'OCl'',S tll'pendlargely on the perception and expenenn• of the dt•signer.Th,1, theperfonna11c,• 1111.h,·1 cI ,,
1 lunruon of tlw c,perit•ncc of the operator ,rnd hence thl' enure rontrnlsv,1cm 1,s\l t, d1s1 ,... , 1
th•signcr. Furtht•r 1wrfor111,rnce improvenwnt is pos-;1h1t! h\ integ1,1llnl!. t'w FLC will,11,1at1hc1,1l 1·e11· 1!
iH•tw 1rk systl'lll (1\NN) to forma adaptive nc'Ul'l) luay mterence systt•m{,\NH:{). 111,\NI IS, o11wu1.1 111'l\,orl<
·s fra1med to do till job of learning and gaining expenencc Once the lc·,1rn111,} ll>Ct''>" i-.. 11\ 1t,ll ', ,
Nl•lwl>rk r,111 he Sl'l to decide on the ranges for various linguistic van,1blei.and tlH II u\ ,•rbp. In \NFIS
:.}:,tem till' ANN isused tocoin the rules based on the t!xperimental data supµlit·cl tu .i,t \. \'!1t1, ..t
c,ipability of an ANN is incorporated with the FL the drawbacks of the convr11t1on,1I rt.Co•• 1,'ll\
Jn•h:nl,u
thl' designer i,slleviated. A. Training data collccnon In the present dt•sign lour rl c., wel'c l1r,t dl',1r,1wl! ,II\-\
lhl' sy tcrn wa1s put 11nde1 Sl'l'Vicc. There were a set of lour t•1 ror e1-ro1 r.itc 011tp1:sl .l1w 1\l,l\'h111.1. ll.1t, ,
('l, nl thl' e tour dataets wen•individually ust'Cl to train an ANI-\S• 111L .111d thw, louru•••\,; ol \ \·H'-
\\'l'lt' th-veloped. Tlw rel.1ll'd scrt:t·n shots as th,•y appear step ill I in l\li\ J'l.,\B S\M\li IN r- l'll'1111111\lt'lll .u t·
:,,1t1\\'tl

in ligurcs below.
 ANFISstnil'iure

r:un«t{IM,i (!'iOai:.!)>1·3 "'.414U f{( 0

-,,, "': '· -
j • • ·---------------------------------------- i• -••. -•-·•--".:.. •••-·
•- •
,I

J·:
1.
:•r .,
t

·.1,- ·
,{

. :. ' t,
'
-..-...... ..-...a:lk- h-.dll..«&-"..,._i/' ..._. blod".,. ;'ol....,....,_, ,-.,_, , '•, '

.2000 itoou toto s;;bu -'·:i•t.VJ:

FHH!11;,rrey (1-1.z) 1)(t

P :1J9.!1naucc c:0111parison of two controUers fox

J->ara1netcr Fuzzy .A.NFJS
- -- -
pdelivt?red co Loc-1.d
.--------,----,---t--
1.4 1.- .r J ·;
---:--1---------------------1
(fDolivcrcd to LoBd l .U:2 1.1):)i.1
f l=--at St>UlCC
1
u.:J t
'i1-0 ) n,t: PCC l . 1-v0(11(J·

Algorithm for tlNFIS:

lt'p t • D<"fine the linguistic var'i.ibles and terms (inilializarion}
l1e'P J: lonstru..:t the nw111bership fu11ct1ons (initializ.ition)
'.->11.'P 3· Construrt the rule lnse (inil!ali :auonJ
\tep !: ( ,mvert •Tl:,!' inµut delta to fuzzy values using the membership funLLH•ns (tuz,.ificalion}
St t'P :,: 1Taluatt the rule:; J/1 the rule base tinference)
\tt,p h· Combin, the n::sulls ofeach rule (inference]
tep 7:Convert the output data to non-fuzzyvd]ues (defuz2ification)
..
11
, ingui lic variJble • '
• ti.rnl i l il,1t1on v,1t,.1bh ·s (i\ll M·
• USJ b· 1t 1t,y·ari,1bles
·
(S
' ,
,li ly, Nnt1l')
1
.
•
• Likelihood variabl o1net1 11cs, l•_requently, Alw,1,•,;)
es (Possible, Like!), lcrtain)
1

I he two most important concepts witl . I' .'

rule The
• •
·ct eI y sect fuzzy if-then
wt
• ruleis:
un ·uuyLog11·(FL)areLin •
gu,.,t,e vari,1hle an,I ltrity ,f-then
• 2 IF (lo,1d is more than too-high)·1·'1l:N .
L. l.
IF (I(i°acl . :loo hi_. ll) :II EN command
•.ff ' ' (i-;
Olllll1,111l
Reduct' theIS Load.
1 t,11111.
1 • O,ld is '.Jo, m,il) 1 II EN eonunancl is 110-chan,;1.

Table 1 Samole test results

Sub A 8 C Fuzzy Rem<1rn .
Station I

Too Over Normal

' J

Substallon B
-
Low Load Connected -t-thri
substation A
Low Low Over Substation C
Load Connected ,.,.;1;1
substation A
Over low Too Substation A
Load Connected vr.t.,
Amps
. LO\v
substation (.
Over Nor I high St;bstahci11A
Load· Connected W,ll,
substation 8 I
Normal High Over Substation C
' load Connected \o'.r.fi
' substation A
High Overload Norm..il Substation 8 ·'-
Connected v/iih
I
I I substat:o n C J
- -

S111ce by usmg FL merhocl we can detect the locaUon of l,rnlt exactly 1t saves the ume of ti operatur x•d th,;
fault c,111 be quickly rectified and the power can be supplied to the affected area Hllme-ci1,1td) f1.i I lhc ;;r, 1
• .vherc there is extess of power available by means of resource sharing. As fuzzy logic is si·np'. ,intl qui(k i'1
tlctecting the power fault in the power system it is widely t<tsed. No,·111,11iy ead1 and t..!VCt \ jl'J\\,•. ,;11 I
:-,ubsl,1tion is connected wlth Load Control Centre. Load Cont1:01 Centre "'{iii be m ,., t ,: ., : th· ,t• ,,f
<;ubsl,1tiuns every second. If fault occurs in one Substcltion that will b<: obta1,H::d by Ll, i d -1 ,. 1t1,,l • t..!lllll·
'.hr◊wih SCAOA (Supervisory control and Data Ac:quisitit11 land 111formed 10 lwr. cny subst, ,1n, "h1rb
ili c cvnnected with the faulty Substation, then only powc,· will be shared to foulty subst<1' 11t In k·· D,1t.
I

fralfic is a major drawback because at the same time many substations lllay fail when ...:ornmlttli ·iti0n deb:
"ill dtcur .md so the resource recovery and load balancing will be delayed, p1:..iplP wil'.! ,: · " ,ho:
:
,,ower, cost loss. The ANFIS will overcome the present status of power sharinl\ .rnd t n,: d... . ::. •,>\\,
ysten1 method which eliminates the need for Load Control Centre.