TheRepubl icofIr

aq
Mini s
tryofeducat
ion
Uni veesi
tyofBasrah
CollegeofEngineering
El
ect
ric
alEngi neeri
ngDepar t
ment
ControlSystemsTechnology

PI
DCont
rol
ler
des
ign,t
uni
ngpar
amet
ers,andsi
mul
ati
onf
or4t
hor
der
plant

‫ﺮ‬
‫ﻴ‬‫ﻣ‬‫ﺍﻷ‬
‫ﺪ‬‫ﺒ‬
‫ﺪﻋ‬‫ﻤ‬
‫ﺃﺣ‬ ‫ﺀ‬
‫ﺎ‬‫ﻴ‬
‫ﺿ‬:‫ﻻﻢ‬
‫ﺍﺳ‬
‫ﺋﻲ‬
‫ﺎ‬‫ـﻣﺴ‬
‫ﻢ‬‫ﻜ‬‫ﺗﺤ‬
I
ntr
oduc
tion
Letusi magine,forexampl e,i nacaront hehi ghway.  
Wewantt odr i
veat
130Km/ hwi t
houthavingt opr esst heaccel er
at or.
 Thecr uisespeedcont r
ol
ofyourcarmustbyi t
selfmai ntainthisspeed.  Whenappr oachingas lope,
thesyst em r
eali
zesthatfort hesamepoweratt hemot orlevel ,itnolonger
reachest he130Km/ hsetpoi ntandwi l
laddal it
tl
eboost .
 Yesbuthow
much?  Andhowl ongwi l
littakef ort hesystem t ostabi l
i
zear oundt heset
point
?
Thisist hewholepr oblem ofensl avementandPI D controlisonewayt o
solveit!
Therear emanyot herexampl esofensl avement sthatuseot hervar iables
fortheirsystem l
ikespeed,angl e,positi
on,t ension. ..Butwhat ever,the
underlyingi
deastayst hesame.  
PIDi st hemostwi delyusedr egulatori
nt hei ndustry. 
Thei deaof​ t
​
his
supervisorybodyi st oi ntentionall
ymodi fyt he value oft he errort hat
remainsbet weenthesetpoi ntandt hemeasur ementmade.

Now wecandef inePI D contr

ol(orproport
ional
,integr
alandder i
vat
ive
cont
rol
)asitisamet hodofr egul
ati
onoftenusedf orensl
avement.Iti
sa
syst
em capabl
e ofreaching and maint
aini
ng a setpointthankstot he
measurement
sitper
forms.

Forexampl
e,i
nthec
aseofaposi
ti
oncont
rolt
heer
rorwoul
dbe:
ε =Ω 
setpoi
nt -Ω 
measur
ement

Fi
gur
e1:Ensl
avementwi
thPI
D

P,pr
opor
ti
ona
l
I
nthecaseofpr
opor
ti
onalcont
rol
,theer
rori
svi
rt
ual
l
yampl
i
fi
edbya
certai
ngai nconst antthatshoul dbedet ermi neddependingont hesystem.
Set
poi nt(t) =K  p .
ε(t)
Whi chgi vesLapl ace:
Setpoint(s) =K  p .
ε(s)
Thei deai st oincreaset heeffectoft heer r
oront hesystem sothatitreact
s
fastert or efer
ence changes.  The larger 
the value of  
p ,t he l
argerthe
response consume ener gy. 
However ,t he stabili
ty of t
he syst em is
deterioratedandi nt hecaseofan  excessive 
K p t
hesyst em mayeven
diverge.
Ifwet aket heexampl eofadr if
tingcar ,proport
ionalregul
ati
onconsi st
sof
count eri
ngst eerqui cklytorestorethecar .
at
Rot i t
onPosi r
on(
i ad)

Fi
gur
e2:St
epr
esponsewi
thdi
ff
erentKp
Ti
me(
s
)
I
,Int
egr
ate
d
Toproporti
onalcontr
ol,wecanaddt heintegrati
onoftheerr
or. 
Inthi
scase
weobtainPIregulati
on(propor
ti
onalandintegrat
ed).
Theer r
orbetweent hesetpointandt hemeasur ementi
sintegrat
edher e
wit
hr especttot i
me and mul t
ipl
i
ed bya const antwhich mustal so be
det
erminedaccordingtothesystem.
t
Set
poi
nt(
t) =K 
p .
ε(t
) +K 
i τ)
. ε(
  dτ ∫0
Whi
chgi
vesLapl
ace:
 ε(s)
Setpoi s
nt()=K  p 
. s
ε()+K  i 
.
s
Whydoweneedt oaddt hi
sfuncti
onal i
tytoourcontrolbody?
Well,duri
ngasi mplepr oporti
onalcheck,t hereremainsast ati
cerror.
 When
thesystem isapproachi ngitssetpoi nt
,theer r
orisnol ongerlar
geenough
tomovet heenginef orward. 
Thei ntegraltermthuscompensat esforthe
stat
ic err
orand t herefore provides a syst em mor e stablei n st
eady
stat
e. 
Thehi gher  
i  
,t hemor et hest ati
cer roriscorrected.Tot akethe
exampleoft hedrift
ingcar ,thei ntegralterm consist
si naddingasmal l
counter-
steeri
ngblowi nordertorecoverpr operl
y.
at
Rot i t
oci
onVel r
y( s)
ad/

Fi
gur
e3:St
epr
esponsewi
thaPIcont
rol
ler
Ti
me(
s
)
D,De
riva
tive
TogetaPIDcheck,west i
llneedt oaddat er
m. Thisconsi
stsofder
ivet
he
er
rorbet
weent hesetpointandt hemeas urementwi t
hrespecttot
imeand
al
somult
ipl
yitbyaconst ant.
t   t
dε()
Set
point(t) =K p .
ε(t) +K  i
  τ)
. ε(
0
∫
dτ +K  d .
dt
Whichgi
vesLaplace:
ε(s)
Setpoi
nt(s)=K  p .
ε(s)+K i 
. +Kd.s
s
Whydoweneedader ivati
vet erm?
Wel l
,PIcont r
olcanl eadt oover shoot i
ngthesetpoint,whichisnotalways
very desirable( exampl e ofr ever se polari
tyint he case ofel ect
ri
c
motors).
 Thet erm der ivat
ivel i
mi tst his.
 Whent hesystem approachesthe
setpoint
,thisterm br eakst hesyst em byappl yi
nganact ionintheopposit
e
dir
ecti
onandt husallowsf ast erstabi l
izati
on.
Sti
llwit
ht heexampl eoft hecar ,theder i
vat
iveter
mr epresent
sthesmal l
l
ocki ntheopposi tedi r
ecti
onofadj ustmenttostabili
zet hecarwheni t
approachest hepointi twant edrej oin.
at
Rot i t
onPosi r
on(
i ad)

Ti
me(
s
)
Fi
gur
e4:St
epr
esponsewi
thPI
Dcont
rol
ler

I
I.
 Ca
lcul
ati
onofc
oef
fi
cie
nts
Wewi l
lnow seehow t ofindtheval ues​t
​
oassigntothet hreecoeff
ici
ents
(
 K 
i 
,K p , K 
d )ofthePI Dcont r
ol. 
Ther ear
et wowayst odot hi
s,oneby
model
ing,t heotherbyexper imentati
on,knowingthatoftenonecompl etes
t
hefi
rstthroughthes econd.
Model
ingt hesystem consistsi
ndet ermini
ngit
sbehaviorbycal cul
ati
onand
f
rom t
her ededucingpl ausibl
evalues​ f
​
orthecoeff
ici
ents. 
Theexper i
enti
al
appr
oachmeansthatwewi l
lusearealr
esponsef
rom t
hesyst
em t
ofi
rst
adj
ustr
oughl
ythenf
inel
ythecoef
fi
ci
ent
s.

Mode
linga
ppr
oac
h:
Thechoi cet oproceedwiththemodel ingofthesystem ornotisdictat
edby
theinherentconstrai
ntstothesystem.  
Oftenthecompl exi
tyofrealsyst
ems
putsmodel i
ngoutofr each,butinsomecases,i nchemi str
yormechani cs
forexampl e,ther ul
esthatgovernthesy st
em ar esimpleenought hatwe
canat t
emptt omodelt hesyst em. 
Ont heotherhand,i tmaybeonl yin
cases wher et he syst
em i s part
icular
ly cri
ti
calordi f
fi
cul
tt o access
(i
ndus t
ri
alpr ocessheavy,system whichcannotbet akenoutofser vi
ce),i
t
i
sessent ialtomodelthesystem inordert ohaveasuf f
ici
entl
yprecisesetof
coeff
ici
ent stoimmediatel
yobtainaPI Dr egul
atorwhichisfunct
ional.

Her
ewegi veanexampleofmodel
ingasimplesystem:theelect
ri
cmot
or.
Ast
heequationsofdynami
csandelect
romagnet
ism areknownandusabl
e,
wecandeterminet
heequati
onst
hatgovernthesystem.
Her
earetheequati
ons:

Fi
gur
e5:Mot
orf
unct
ionalschemat
ic

Symbol Par
amet
er Uni
t

J momentofi
ner
ti
aof kg.
m^2
t
herot
or

b motorvi
scous N.
m.s
f
ri
cti
onconst
ant

Ke el
ect
romotivef
orce V/
rad/
sec
const
ant

Kt motort
orque N.
m/Amp
constant
 R el
ect
ri
cresi
stance Ohm

L el
ect
ri
cinduct
ance H

1 b
s2.
Θ(s)= .
KtI
(s)-.
s. s)
Θ(
J J

b 1
s)
Θ( ( J)
s = Kt.
.s2 + .
J
I(
s)

Kt
s)= 2
Θ( .
I(
s)
s +b.
J. s

-
R 1 Ke
s.
Is)= .
( (
Is)+ .
V(s)- .
s.Θ(
s)
L L L

R 1 Ke
(
I ( )
s)
.s + = .
L L
V(s)- .
L
s.Θ(
s)

1 Ke
(
Is)= V(
. s)+ 2 .
Θ(s)
L.
s +R L.
s +R.s

Kt 1 Kt Ke
s)= 2
Θ( . .
V(s)+ 2 . 2 .
Θ(s)
J.
s +b.sL.s +R J.
s +b.sL.
s +R.s

Tr
ansf
erf
unct
ioni
s

s)
Ω( K rad
= [ ]
V(
s) (
J.s +b)
.(
L.s +R)+K2 s.V

Wede
duc
efr
omt
hist
het
rans
ferf
unc
ti
onoft
hes
yst
em:
s)
Ω( A
=
s)
V( 2ξ 1 2
1+ . s
s+ 2.
ω0 ω 0
1
Wi
thA = staticgain
Ke
ξ=R J Dampi ngf
act
or
2 KtKeL

ω0 = KeKt Pur
epul
se
LJ

From thi
s,wecandeducethetr
ansferf
uncti
onoftheslavesystem and
ther
eforetheval
ues​t
oassi
gntothecoeff
i
cient
stoobtainananswert hat
sati
sfi
esus.
Thepur poseoft
hisexampl
eistoshowhowasi mplesystem governedby
equati
onssimpl
eli
neardi
ff
erent
ial
s,gi
vesri
setocomplexvalues​
fort
he
coef
fi
cient
s. 
Wecanthusjust
if
ythepract
i
cali
nter
estoftheexperi
mental
method.

Expe
rime
nta
lappr
oac
h
Herewepr esentt wopossibl
eappr oachesf ordet er
mini ngcoefficient
sby
experience.
Mention:onemet hodwit
ht heuseoft hePI Dregul at
orandt heot herwhi ch
doesnotusei t
.
Weal soi ndicatetheadvantagesanddi sadvantagesoft heset womet hods.
Theexi stenceoft hesecondmet hodi sjust
if
iedbyt hef actthatincer t
ain
contr
ollerstheval ues​
ofthecoeff
ici
entscannotbemodi fied,thisi
st hecase
i
fforexampl et hecontr
olleri
sanel ectroni
ccircuitandt hecoef f
icientsare
deter
mi nedbyt hecomponent s(
resist
ors,capacitors)
.

1.Zi
egl
er-
Nic
hol
sappr
oac
hforc
los
edl
oops
yst
ems(
alr
eadyr
egul
ate
d)

St
abl
eosci
ll
ati
ons
Fi
gur
e6:Obt
aini
ngst
abl
eosci
ll
ati
onsbyexper
iment
ingwi
thdi
ff
erentcoef
fi
cient
s

Coe
ffi
ci
entva
lue
s:
Wenote: 
K u =K p 
osci
ll
ati
ons ,
 P u =Per i
odofosci
l
lat
ions
Ku Pu P 
u
Thenchoose 
K p = ,  
K i
 = ,  K d = .
1.
7 2 8
Thesevalues​
f
​
ort
hecoef
fi
cient
sar
ethosethatZi
egl
erandNichol
sfound
t
oal l
ow dampi
ngt
heampli
tudeoft
heoscil
l
ati
onsby1/4ateachpseudo-
peri
od.

Be
nef
it
s

 Themet hodiseasytoimplementphysical
lyandf r
om acal cul
ati
on
pointofview
Itcan be appl i
ed to a syst
em al r
eady i
n product
ion and all
ows
automaticadapt
ati
onofther egulat
ortoadaptt
ochangesi nint
ernal
(wear)andexter
nalparameters(envi
ronment
)tothesystem.

Di
sadva
nta
ges

 Thesyst em maybecomeunst ableorgoi ntodangerousstates(f

or
examplef orchemicalsyst
ems)
 Themet hodmayt akeal ongt i
mei ft
hesyst em react
sverys l
owly
(days,weeki nthecaseofcertainchemicalreacti
ons)
For
tunat
ely,manys ystemshave l ow character
isti
ct i
mes( el
ect
roni
cor
mechanical
).
Pr
otoc
ol:
 Appl
yat
imes
lott
othes
yst
ema
ndr
ecor
dit
sre
spons
e.

Fi
gur
e7:St
udyoft
hesyst
em r
esponset
oani
che
Assumi
ng:
 X0 τ
K0 = .
M0τdelay
Wet
henhave:
 K 
p =1.
2 *K 
0 ,
 K 
i =2.τdelay 
and 
K d =0.
5 *τdelay

Be
nef
it
s

 Thi
smet hoddoesnotr
equi
rehavi
ngasyst
em al
readysl
aved,nor
guessi
ngavaluef
orK 
p

Di
sadva
nta
ges

I
tr equi
resmor emat er
ial
:recordi
ngofaresponsecur
ve.
I
tf or
cest oputthesystem "of
fl
ine"
I
ft he system response istoo dif
fer
entfr
om the response shown
above,t hismet hodgivesvalueswhi chmaynotbesui tabl
eatal l
(somesyst emsf orexampledonotar enotatallcontrol
l
ablebyPI D
regulat
ors).

Conc
lus
ion

Theexper i
ment almet hodspr esentedaboveal low goodest imatest obe
made.Gener i
cf orareasonabl ysimplesy st
em ( i
econt r
oll
abl ebyaPI D!).
However ,itisver yof tennecessar yt of i
ne-tunet hecoef f
ici
entsunt ila
system is obtained.r especting the specif
icati
ons. Forthis we use t he
fol
lowi
ngr ules:
I f
 K p 
increasest hent hecl i
mbwi l
lbef asterbutt herewi l
lbemor e
overshoot
I f
 K i
 i
ncreasest hent her i
sewi l
lbef asterandt hest ati
cer rorwi l
lbe
l
owerbutt hesteadyst atewillt
akelongert oreach
I f
 K d 
increasest hent heover shootwi l
ldecreaseandt het i
mef or
establi
shingsteadyst atealso,butthesensi ti
vi
tytonoi sewi l
lincrease.
Othermet hods( notaddr essed):I
nternalModelCont r
ol,Nasl i
nMet hod

II
I.
 Li
mitsa
ndimprove
ments
Aswesawear li
er,aPIDservocanincl
udeacer
tai
nnumberofl
imi
tswhi
ch,
i
fnottakenint
oaccount,cani
mpairthepr
operf
unct
ioni
ngoft
hesyst
em or
evendamagei t
.
4thor
derpl
ante
xampl
e

Lett
het
ransf
erf
unct
i
onbel
owi
sourpl
ant

Bui
l
dingamodelonSi
mul
i
nkwi
tht
hist
ransf
erf
unct
ion

Wi
thoutPI
Dcont
rol
l
ert
her
esponsehasasi
gni
fi
cantval
ueoft
ransi
ntanda
st
eadyerror(
theoutputi
sles
sthan0.
6fr
om i
nput
),butwi
thPcot
rol
l
ee
we'
llr
eli
zethedif
fence.

I
nor
dertofi
ndtheval
ueofPI
Dpar
amet er
sweshal
lusethef
oll
owi
ngt
abl
e
PIDtype Kp Ti Td
P 0.
5Kcr Inf
init
e 0
PI 0.
45Kcr Pcr/1. 2 0
PID 0.
6Kcr 0.5Pcr 0.125Pcr
Byusi
ngMATLAPwesckuchRootl
ocust
ofi
ndKcr=63.
1asshown
bel
l
ow

Tof i
ndPcr ,f
ir
stl
ywetakeKcrasagaint
othesyst
em andpysimulinkt
o
f
indcr i
ti
calperi
odti
me=3.148sec,t
henwefi
ndPcr=( 2*
pi)
/(
3.148)=
1.99719r ad/
s
WhenweuseaPcont
rol
l
erweshal
lfol
l
owt
hef
ir
str
owi
nthet
abl
eKp=
0.
5*Kcr
,andP=Kp

Fi
gurebell
owshowst heresponse,andtheadvant
geofusi
ngPcont r
oll
er,
howitdecr
easesthetimeoftransi
nt,andmakethest
eadyresponsehi
gher
(about0.
65from i
nput).
Togetmoradvant
ageofourcont
rol
l
er,we'
l
laddanot
herpar
amet
er,by
usi
nganint
egrat
edproper
ty.

Byusi
ngt hesecondrowofthepar
ameterstablwe'
llgetKp=0.45*
Kcr
,
and
P=Kp=28. 395
Ti=Pcr/1.
2=1. 66374
I=Kp/Ti=17.07174
Wit
hthisparametesther
esponsei
nthefoll
owingfi
gurewil
lappear
s

Wit
hthi
scontr
oll
erwerel
i
zethatt
heti
meoft
ransi
nti
sdecr
easedandt
he
st
eadyst
atbecomesneartot
heinput
.
Togetanext
r aadvant
ageofthecont
rol
l
er,anot
herpar
ameterwi
l
ladded,i
t
i
sthederi
vat
iveparameter
,andthecont
rol
lerwi
l
lbeaPID,asshown
bel
low

Tocalculat
et hevaluesoft
hiscont
rol
l
erweuset
hwthir
drowoft
het
abl
eto
fi
ndthatKp=0. 6*
Kcr
P=Kp=37. 86
Ti=0.5*Pcr=0. 99796
Td=0. 125*Pcr=0. 24949
I=Kp/Ti=37937
D=Kp*Td=9. 44573
Thenther esponsewi l
lbeasshowninfol
l
owi
ngf
igur
e
Concl
usi
on

ThePI Dcont rolsyst em i st odayoneoft hemostusedcont rolsystemsf or

manyr easons.Fi r
stly,i tisver ysimplet osetupandi seffectiveformost
par tofrealsyst ems.I naddi tion,thecal culati
onoft hecoef fi
cientsleaves
thechoi cebet weensever almet hodsofi ncreasingdi ffi
culty.Ont heone
hand,anexper i
ment almet hodandver yeasyt osetupal l
owsqui ckly
obtaincor rectcoef ficientsf orsyst emst hatdonotr equi r
ever ylargedetai l
s
i
nt heser vo.Ont heot herhand,advancedmat hemat i
calmet hodsof fer
techniquesf orobt aini ngt hei dealcoef fi
cientsforapar t
icularsyst em.So
theset ti
ngpl aceofaPI D ser voper hapsbot hf astandef f
icientandal low
optimizationcoef fi
cient sf ort hemostadvancedsyst ems.However ,iti s
i
mpor t
antt onot et hatt hist ypeofser vo- controlisl imi t
edbyanumberof
const r
aints.
First,itmayt urnouti neffec t
ivef orcer tainsyst emswhi chcont ai
nnoi se
(Der i
vati
vecoef f
ici
ent )orwhi char enotl i
near( thePI Dser vobei nglinear,
thenon- li
near ityofasyst em l eadst oinstablesyst em) .Finally,iti
spossi ble
toopt i
mi zet her esponseofasyst em bymul t
iplyingt hecont rols(li
kef or
exampl et hedoubl ePI D ser vo).I ti
sf oral lt
heser easonst hatthismodel
servor emai nst hemostusedi nt heindust rytoday.

Re
fer
enc
es
Moder
nCont
rolSyst
ems:byRi
char
dC.Dor
f&Rober
tH.Bi
shop

Moder
nCont
rolEngi
neer
ing:byKasuhi
koOgat
a

Autonomousrobotcontrolandservo
ht
tp://
clube
lek.
insa
-lyon.f
r/j
oomla/fr/
bas
e_de
_conna
iss
anc
es/
inf
orma
ti
que
/
se
rvo_and_pil
oti
ng_of_robot_a
uto.php

PI
Dus edi
npos i
ti
onand/orspeedre
gulati
onofe l
ect
ri
cmot
ors
,Chr
ist
opheLeLa
nn2007,
ht
tp:/
/www.t
otofweb.
net
/pr
ojet
s/
pid/
rappor
t.pdf

Pr
a c
ti
calPr
ocessContent
ht
tp://
www.contr
olgur
u.c
om/
wp/
p76
.ht
ml