Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
5
,
Octo
ber
201
9
, pp.
4114~4
129
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v9
i
5
.
pp4114
-
41
29
4114
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Des
i
gn a
nd
p
er
f
orman
ce
c
omp
ar
i
son of
d
i
fferent
ad
aptive
c
ont
rol
s
chem
es
f
or
pitch
a
ngl
e
c
on
tro
l in
a
T
win
-
R
oto
r
-
M
IMO
-
S
ystem
Winst
on
N
et
t
o
,
R
ohan
L
akhani
,
S.
M
ee
n
at
c
hi Sun
da
r
am
Depa
rt
m
ent
o
f
I
nstrum
ent
at
ion
a
nd
Control
Enginee
ring
,
Man
ipal
Instit
u
te of T
e
c
hnolog
y
,
Manipa
l
Aca
d
e
m
y
of
High
er Ed
uca
t
ion,
India
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Feb
16
, 201
9
Re
vised
A
pr
1
0
, 2
01
9
Accepte
d
Apr
25
, 201
9
The
Twin
Roto
r
MIM
O
Sy
st
e
m
is
a
highe
r
orde
r
non
-
li
n
ea
r
pla
nt
and
is
inhe
ren
tly
unsta
ble
due
to
cro
ss
coupl
ing
be
tween
ta
il
and
m
ai
n
r
otor.
In
thi
s
pape
r
onl
y
the
c
ontrol
of
m
ai
n
rotor
is
conside
red
which
is
non
-
li
ne
ar
and
stabl
e
b
y
using
ada
pt
ive
sch
emes.
The
cont
ro
l
p
roble
m
is
to
ac
h
ie
ve
p
erf
e
ct
tra
ck
ing
for
in
put
ref
e
ren
c
e
s
igna
ls
whil
e
m
ai
nt
ai
ning
robu
stness
and
stabi
lit
y
.
Four
ada
pt
ive
sch
emes
were
imple
m
ent
ed
,
two
u
sing
Model
Refe
ren
ce
Adap
ti
ve
Con
trol
und
er
which
MIT
r
ule
and
Modified
MIT
rul
e
are
used
.
Th
e
ot
her
two
using
Adapti
v
e
Inte
r
acti
on
,
namel
y
,
A
dapt
iv
e
PID
and
Approxim
ate
Adapti
v
e
PID
.
It
is
observe
d
t
hat
ad
apt
iv
e
sch
emes
fulfi
ll
al
l
th
e
three
s
y
s
te
m
per
form
ance
req
uir
ements
at
the
sam
e
t
ime.
Modifi
ed
MIT
rule
was
found
to
give
superior
per
form
a
nce
in
compari
s
on
to
othe
r
cont
rollers.
Also
Approxi
m
at
e
Adapti
v
e
PID
was
able
to
stab
il
i
z
e
th
e
m
ai
n
rotor
and
ca
nc
el
the
eff
e
ct
of
cro
s
s c
oupli
ng
bet
we
en
ta
il
ro
tor
and
m
ai
n
rotor
when
oper
a
ti
ng
sim
ult
ane
ousl
y
without
th
e
n
ee
d
for
design
ing
d
e
coupl
ers
for
the
s
y
stem.
Thu
s
the
m
ai
n
rotor
ca
n
be
m
ade
inde
pen
den
t
from
the
stat
e
of
the
ta
i
l
ro
tor
b
y
using Approxi
m
at
e
Adapti
v
e
PI
D
.
Ke
yw
or
d
s
:
Ad
a
ptive
c
ontr
ol
Ad
a
ptive
PID
MIT
r
ule
MR
AC
Robustne
ss
Stabil
it
y
TRM
S
s
yst
em
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
C
or
res
pond
in
g
Aut
h
or
:
W
i
ns
t
on N
et
to
,
Dep
a
rtm
ent o
f In
st
ru
m
entat
ion
a
nd Co
ntr
ol
En
gin
eeri
ng,
Ma
nip
al
Insti
tu
te
o
f
Tech
nolo
gy,
MAHE
, Man
i
pa
l, Udupi,
Ka
r
nataka,
57
6104
, In
dia
.
Em
a
il
:
winn
et
to@
gm
ai
l.co
m
1.
INTROD
U
CTION
The
T
win
Rot
or
M
IMO
Syst
e
m
was
dev
el
oped
by
Feed
ba
ck
In
st
ru
m
ents
Ltd.
an
d
se
r
ve
s
as
a
real
-
tim
e
m
od
el
of
nonlinea
r
m
ult
idi
m
ension
al
s
yst
e
m
.
To
vis
ua
li
ze
the
pa
rts
and
m
otion
s
of
the
TRM
S
t
oget
he
r
with
the
f
or
ce
s
ge
ner
at
e
d
by
the
act
uat
or
s
,
a
m
od
el
of
t
he
T
RM
S
is
seen
in
Fig
ur
e
1.
T
he
TRM
S
co
ns
ist
s
of
a
towe
r
with
a
be
a
m
at
ta
ched
by
two
bea
rin
gs.
T
hese
bear
i
ngs
al
lo
w
t
he
be
a
m
to
m
ov
e
f
r
eel
y
in
the
hori
zon
ta
l
and
ve
rtic
al
plane
wit
hin
s
om
e
lim
it
s.
At
the
tw
o
en
ds
of
t
he
be
am
,
ro
to
rs
are
at
ta
c
hed
w
hich
ro
t
at
e
d
90
degrees
from
e
ach
ot
her
a
re
a
ll
ow
in
g
them
t
o
ge
ne
rate
hor
iz
on
ta
l
an
d
vert
ic
al
thru
sts.
T
he
r
otor
ge
ne
r
at
ing
ver
ti
cal
thr
us
t
is
cal
le
d
the
m
a
in
ro
t
or
.
T
his
enab
le
s
the
m
odel
to
pitch,
wh
i
ch
is
ro
ta
ti
on
i
n
the
ve
rtic
al
plane
arou
nd
the
hor
iz
on
ta
l
axes.
T
he
ro
t
or
ge
ne
ra
ti
ng
the
horizo
ntal
thru
st
is
cal
le
d
the
ta
il
ro
tor.
T
his
enab
l
es
the
m
od
el
to
ya
w,
wh
ic
h
is
ro
ta
ti
on in
t
he horiz
on
ta
l
plane
ar
ound t
he verti
ca
l axis [
1].
The
T
win
M
ot
or
M
IMO
syst
e
m
is
a
hig
hly
non
-
li
near
pla
nt
in
wh
ic
h
th
ere
are
c
ertai
n
s
ta
te
s
that
cannot
be
m
ea
su
re
d,
this
m
akes
desig
ning
of
the
con
tr
oller
a
diff
ic
ult
ta
sk
.
The
popula
r
well
known
sc
hem
es
will
no
t
giv
e
de
sired
outp
ut
char
act
erist
ic
s
if
us
e
d
for
co
nt
ro
ll
in
g
the
pla
nt
and
will
fail
to
sta
bili
ze
i
t
i
n
m
os
t
cases. Ma
ny contr
ol
schem
es h
ave
b
ee
n de
ve
lop
e
d for c
on
t
ro
ll
in
g
the
m
ain
ro
t
or
of the
T
RM
S.
PI
D
’s
a
nd
li
ne
ar
co
ntr
ollers
[2
]
wer
e
no
t
able
to
gu
a
r
antee
global
s
ta
bili
ty
and
ful
fill
desired
respo
ns
e
c
har
a
ct
erist
ic
s.
Mode
l
pr
e
dicti
ve
con
t
ro
ll
er
[3
]
does
gua
ran
te
e
global
sta
bili
ty
bu
t
at
a
pri
ce
of
l
ow
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
El
ec &
C
om
p
En
g
IS
S
N: 20
88
-
8708
Desig
n a
nd p
e
rforma
nce c
ompa
ris
on
of d
if
fe
rent
adapti
ve con
tr
ol sc
hem
e
s
for pit
ch
...
(
Wi
ns
to
n
Ne
tt
o)
4115
degree
of
r
obust
ness
a
nd
po
or
trac
king
pe
rfor
m
ance.
Ro
bust
PID
obta
in
ed
[
4]
usi
ng
K
har
it
on
ov’s
the
or
em
gav
e
good
r
obus
tness
but
poor
trac
king
pe
r
form
ance
and
rem
ai
ns
sta
ble
on
ly
for
sm
al
l
values
of
co
nt
ro
ll
er
gains
.
Using
Sli
din
g
M
ode
Co
ntr
oller
[
5]
sat
isfact
ory
tracki
ng
was
obta
ine
d
bu
t
due
t
o
c
hatte
ri
ng
conve
rg
e
nce
r
at
e
of
syst
e
m
sta
te
s
was
ver
y
low
al
so
the
non
-
r
obus
t
reaching
pha
se
in
SMC
m
akes
syst
e
m
u
ns
ta
bl
e.
Fig
ure
1. Twi
n R
otor MIM
O S
yst
e
m
Th
us
it
can
be
con
cl
uded
t
h
at
the
sc
hem
es
i
m
ple
m
ente
d
up
un
ti
l
no
w
ha
d
a
tra
de
off
bet
wee
n
rob
us
tness
,
tra
ckin
g
pe
rfo
rm
ance
an
d
global
sta
bili
ty
.
In
t
his
pa
per
the
ada
ptive
con
t
ro
ll
er
sc
hem
es
i
m
ple
m
ented
fo
r
TRM
S
m
ai
n
r
otor
sat
isfy
al
l
these
thre
e
syst
e
m
char
act
erist
ic
s
toge
ther
i.e.
t
her
e
is
no
com
pr
om
ise
between
r
obus
t
ne
ss,
sta
bili
ty
and
trac
king
perform
ance.
The
se
ada
ptive
sch
e
m
es
are
al
so
e
asy
to
i
m
ple
m
ent an
d re
qu
ire
m
ini
m
um
p
la
nt know
le
dg
e t
o wor
k.
The
m
ai
n
ro
tor’
s
tra
ns
fe
r
f
unct
ion
was
obt
ai
ned
usi
ng
bl
ack
box
ide
ntific
at
ion
,
w
hic
h
is
the
on
ly
thing
we
nee
d t
o
know abo
ut the p
la
nt.
T
his tran
s
fer
fu
nctio
n
was use
d
in the ad
a
ptive sc
hem
es as a r
eferen
ce
m
od
el
and
as
a
par
t
of
th
e
con
tr
oller
it
sel
f.
The
re
sul
ts
ob
ta
in
sho
w
high
r
obus
t
ness,
good
tra
ckin
g
perform
ance an
d g
ua
ran
te
e a
b
so
l
ute stabil
it
y al
l at
the sam
e tim
e w
hich
was p
rev
i
ously
not obtai
ne
d.
2.
METHO
DOL
OGY
2.1.
Ident
ific
at
i
on
Syst
e
m
was
id
entifi
ed
us
in
g
black
box
i
dent
ific
at
ion
.
Cr
oss
co
up
li
ng b
et
ween
the
ta
il
and
m
ai
n
ro
to
r
was
not
co
ns
i
de
red.
T
he
ta
il
r
otor
was
ke
pt
at
zero
a
nd
onl
y
the
m
ai
n
ro
t
or
was
ide
ntifie
d
in
f
orm
of
a
sin
gle
trans
fer
f
un
ct
i
on.
A
PRB
S
(
Pseudo
Ra
nd
om
B
inary
Sig
nal)
was
giv
e
n
to
the
plant
an
d
the
outp
ut
wa
s
analy
zed i
n
M
ATL
AB syst
e
m
identific
at
ion
to
olbo
x
as
shown i
n
Fi
gure
2
.
Fig
ure
2. PRB
S in
pu
t
a
ppli
ed
to
the
TRM
S
and
the
outp
ut
form
the
m
ai
n
ro
t
or
The best est
im
at
e o
f
the t
ran
s
fer
f
un
ct
io
n o
bt
ai
ned
is
ha
ving a
n
acc
ur
acy
of 72.
7%.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
4114
-
4129
4116
(
)
=
0
.
01
1
06
+
0
.
37
68
0
.
258
3
+
0
.
25
28
2
+
1
.
16
+
1
(1)
w
he
re
G(
s
)
re
presents
t
he
ap
pro
xim
at
ed
transf
er
functi
on
of
the
non
-
li
nea
r
m
ai
n
ro
tor
.
T
he
de
gree
to
w
hich
G(
s
)
is
a
fait
hf
ul
re
pr
ese
ntati
on
of
t
he
TRM
S
m
a
in
ro
t
or
c
an
be
see
n
by
F
ig
ure
3
,
w
he
r
e
the
ste
p
res
ponse
of
bo
t
h
the
TRM
S
an
d
t
he
a
pp
roxim
at
ed
m
od
el
wa
s
c
om
par
ed
.
T
he
tra
nsfer
f
un
ct
io
n
ob
ta
ine
d
is
a
cl
os
e
appr
ox
im
at
ion
o
f
the TRMS
m
ai
n
ro
t
or
.
Fig
ure
3
.
Mo
de
l v
al
idati
on
2.2.
Contr
oller
d
es
ign
Ad
a
ptive
sc
he
m
es
are
e
m
pl
oyed
he
re
to
co
nt
ro
l
the
m
ain
ro
t
or
of
the
TRM
S.
The
a
dv
a
ntage
of
adap
ti
ve
sc
he
m
es
ov
e
r
co
nv
entional
PID
i
s
that
the
valu
es
of
the
c
ontr
ollers
are
not
f
ixed
a
nd
up
dat
e
with
tim
e,
i
t
has
know
le
dg
e
of
t
he
sta
te
s
of
the
pl
ant
wh
ic
h
it
is
con
tr
olli
ng.
Fi
gure
4
s
hows
t
he
co
nt
r
ol
struc
ture
wh
e
re
the
pa
ra
m
et
ers
of
the
con
t
ro
ll
ers
ar
e
not
fixe
d
but
ev
olv
e
ove
r
tim
e
and
h
ow
they
e
vo
l
ve
de
pends
upon
the ad
a
ptive
law fo
rm
ulate
d.
Fig
ure
4
.
Ada
pt
ive
co
ntr
ol str
uctu
re
2.2.1.
The MIT
r
ule
The
MI
T rule
works on MR
AC (
M
od
el
Re
fer
e
nce
Ad
a
p
ti
ve
Co
ntr
ol) w
he
re th
e
outp
ut
of
t
he plant i
s
m
ade
to
f
ollo
w
t
he
ou
t
pu
t
of
a
ref
e
re
nce
m
od
el
as
s
ho
wn
in
Fig
ure
5
.
He
re
t
he
e
r
ror
(
,
)
obta
ined
betwee
n
t
he
outp
ut
of
t
he
re
fer
e
nce
m
od
el
and
the
pla
nt
is
sub
j
ect
ed
to
a
cost
f
un
ct
io
n
w
hich
is
m
ini
m
iz
ed
us
in
g Gr
a
dient
algorit
hm
.
Mi
nim
iz
at
ion
alg
or
it
hm
u
sed
was p
rop
os
e
d by
Wh
it
aker, the
gr
a
dient m
et
hod [6
]
.
Fig
ure
5
.
Th
e
MIT rule
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
El
ec &
C
om
p
En
g
IS
S
N: 20
88
-
8708
Desig
n a
nd p
e
rforma
nce c
ompa
ris
on
of d
if
fe
rent
adapti
ve con
tr
ol sc
hem
e
s
for pit
ch
...
(
Wi
ns
to
n
Ne
tt
o)
4117
̇
=
−
[
1
2
2
(
,
)
]
(2)
=
[
(
)
−
(
)
]
(3)
w
hic
h
is st
at
ed
as [6]
.
̇
=
−
[
(
)
−
(
)
]
(
)
(4)
w
he
re
is
the
a
dap
ti
ve
gain,
(
)
is
the
outp
ut
from
the
re
fer
e
nc
e
m
od
e
l
an
d
(
)
is
the
outp
ut
f
r
om
the
plant. T
he
ada
ptive law
is
give
n
as
:
̇
=
−
(
)
(
5)
-
P
er
fect t
rac
ki
ng
a
nd
fast co
nv
e
r
gen
ce
of
̇
li
m
→
∞
1
∫
[
(
)
{
(
)
}
]
[
(
)
{
(
)
}
]
>
0
0
(6)
w
he
re
(
)
is
the
pl
ant
a
nd
(
)
is
t
he
re
fer
e
nce
m
od
el
,
an
d
are
th
e
gai
ns
of
the
ref
e
ren
ce
m
odel
and
plant
respec
ti
vely
.
Theor
e
m
1:
Under
t
he
co
ndit
ion
t
hat
(
)
an
d
(
)
are
stric
tl
y
sta
ble,
t
hat
r(
t
)
is
bounde
d
a
nd
(
6)
is
sat
isfie
d
,
there
e
xists
a
posit
ive
c
on
sta
nt
∗
su
c
h
that
f
or
al
l
∈
(
0
,
∗
)
gain
ad
jus
te
d
by
the
M
IT
ru
l
e
is
bo
unde
d
a
nd
co
nver
ges
e
xponentia
ll
y
fa
st
to
∗
(
)
as
→
∞
.
Also
Energy
in
r
(t)
sh
oul
d
be
local
iz
ed
w
her
e
(
)
an
d
(
)
have
si
m
il
ar
fr
eq
ue
ncy
res
pons
es
.
If
t
hese
c
ondi
ti
on
s
are
sat
isf
ie
d
then
perfect
trac
king a
nd f
ast
c
onve
rg
e
nce
of
̇
is o
btained
[
6]
.
-
Stabil
it
y wit
h
la
rg
e
ad
a
ptati
on g
ai
n
-
̇
=
−
[
(
)
(
)
−
(
)
(
)
]
(
)
(
)
(7)
wh
e
re
(
0
)
=
1
.
(
7) ca
n be
re
wr
it
te
n
as
(
)
=
2
2
+
2
(
)
(8)
Stabil
it
y
fo
r
la
rg
e
a
dap
ta
ti
on
gai
n
is
prov
e
d
with
r
oot
lo
cus
te
c
hniq
ue
w
he
re
the
bo
unde
dness
of
(
)
is
ob
ta
ine
d.
T
he
or
em
2:
Th
e
MIT
r
ule
with
r(
t)=R
has
infin
it
e
gain
m
arg
in
(i.e.
f
or
al
l
posit
ive
value
s
of g an
d
R,
the
ad
a
ptive la
w
i
s stable in
de
pe
nd
e
nt
of
if and
only
if [6]
.
−
2
<
a
rg
(
)
<
3
2
∀
∈
ℝ
(9)
Stabil
it
y
will
beco
m
e
ind
e
pe
nd
e
nt
of
the
a
dap
ti
ve
gain,
s
yst
e
m
re
m
ai
ns
sta
ble
for
al
l
adap
ti
ve
gains
if
the
above c
onditi
on is sati
sfie
d.
Bl
ock
diag
ram
for
t
he
MI
T ru
le
as sho
wn in Fi
gure
6.
Fig
ure
6
.
Bl
oc
k diag
ram
fo
r
t
he
MI
T
r
ule
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
4114
-
4129
4118
2.2.2.
Modifie
d
MIT
r
ule
Ther
e
a
re
cert
ai
n
lim
it
a
ti
on
s
wh
e
n
us
in
g
MIT
r
ule.
T
he
m
agn
it
ud
e
of
the
gradie
nt
changes
as
we
desce
nt
to
the
m
ini
m
a.
But
if
la
rg
e
num
ber
of
sa
ddle
po
i
nt
s
are
present
th
en
the
de
rivati
ve
bec
om
es
ze
ro
a
nd
it
s
m
agn
it
ud
e
al
so
beco
m
es
zero.
T
hus
it
m
a
y
ta
ke
s
om
e
tim
e
to
esca
pe
t
hese
point
s
m
akin
g
t
he
c
onve
rg
e
nce
slow
e
r
[7]
.
T
he
se
lim
it
ation
s
are
ov
e
rco
m
e
by
Mod
ifie
d
MIT
ru
le
in
w
hich
norm
al
iz
e
d
gr
a
die
nt
m
eth
od
is
us
e
d
therefo
re
the
directi
on
of
the
gr
a
dient
is
pr
ese
rv
e
d
b
ut
the
m
agn
it
ude
is
ign
or
e
d.
T
he
N
or
m
al
iz
ed
MIT
Rule usi
ng No
r
m
al
iz
ed
al
go
rit
hm
is g
iven
as
=
−
+
′
(10)
her
e
=
a
nd
(
>
0
)
whic
h
is a c
onsta
nt
[
8]
2.2.3.
Adapt
i
ve
P
ID
The
A
da
ptive
PI
D
w
orks
by
us
in
g
a
dap
ti
ve
interact
ion.
A
dap
ti
ve
i
nterac
ti
on
w
orks
on
the
pri
ncipa
l
that
a
syst
e
m
can
be
bro
ke
n
dow
n
int
o
nu
m
ber
of
s
ubsyst
em
s
(1
,
2,
3,..,n
)
and
the
i
ntera
ct
ion
of
t
hese
s
yst
e
m
causes a
da
ptati
on. Acco
r
ding
to ad
a
ptive
int
eracti
on
the
or
y
four s
ubsyst
em
s ar
e con
si
de
red he
re.
-
Pr
op
or
ti
onal
w
it
h
outp
ut
1
-
In
te
gr
al
with
outp
ut
2
-
Der
i
vative
with
ou
t
pu
t
3
-
(
)
, estim
a
te
d
tran
sfer f
un
ct
io
n o
f
the
p
la
nt
W
it
h
t
he
Non
-
li
near
pla
nt
interact
io
n
t
he
interact
io
n
of
these
subsyst
e
m
s
will
giv
e
rise
t
o
th
e
adap
ta
ti
on.
is
the
a
da
ptive
ga
in,
e
is
the
e
rro
r
si
gnal
,
u
is
th
e
re
fer
e
nce
in
put,
g(t
)
is
t
he
im
pu
lse
res
ponse
of
the
syst
em
an
d
1
is
the
wei
gh
i
ng
facto
r.
Using
the
the
or
y
of
a
da
ptiv
e
interact
io
n
t
he
PID
co
ntr
ol
le
r
al
gorithm
b
eco
m
es
̇
=
−
1
̇
[
]
1
(11
)
̇
=
−
1
̇
[
]
2
(12)
̇
=
−
1
̇
[
]
3
(13)
wh
e
re
o
denot
es
f
un
ct
io
nal
com
po
sit
ion
,
1
is
the
ada
ptati
on
gai
n.
1
,
2
,
3
re
presents
the
out
pu
t
of
the
pro
portion
al
,
t
he
i
nteg
ral
a
nd
de
rivati
ve
tra
ns
fe
r
f
un
ct
io
n
blo
c
ks
,
re
sp
ect
ively
.
0
is
the
pla
nt’s
outp
ut
is
the
com
m
and
input.
T
is
a
ca
us
al
f
unct
ion
al
relat
ion
s
hip
be
tween
plant
’s
input
a
nd
out
pu
t.
̇
[
]
=
is
the
Fr
ec
het d
e
rivat
ive
[
9]
.
The
or
em
3:
C
onditi
on
for
adap
ti
ve
inte
rac
ti
on
is
that
the
inp
ut
as
well
as
the
ou
tp
ut
s
hould
be
an
integra
ble
signa
l.
The
ap
plica
ti
on
of
ada
ptiv
e
interact
ion
re
qu
i
res
a
crit
ic
al
con
diti
on
tha
t
sh
ou
l
d
be
sat
isfie
d
wh
ic
h
is t
hat th
e Fr
ec
het
der
i
va
ti
ve
of the
im
pu
lse
r
es
pons
e
of the syst
em
m
us
t exist
.
li
m
‖
∆
‖
→
0
‖
[
+
∆
]
−
[
]
−
̇
[
]
∆
‖
‖
∆
‖
=
0
(14)
Fo
r
li
near t
im
e
inv
a
riant
plant
w
it
h
t
ran
s
fer f
un
ct
io
n G
(s) th
e Fr
ec
het
der
i
va
ti
ve
is gi
ven a
s [9]
̇
[
]
=
∫
(
−
)
ℎ
(
)
=
(
)
∗
ℎ
(
)
0
(
15)
(
)
(
)
(
16)
w
he
re
(
)
is
the
est
i
m
at
ed
plan
t
transfer
funct
ion
a
nd
sat
isfi
es
these
c
ondit
ion
s
he
nce
it
c
an
be
us
ed
f
or
con
t
ro
ll
in
g
the
m
ai
n
ro
t
or
of T
RM
S.
T
he
a
da
ptiv
e law
s ar
e
giv
e
n
as
foll
ows
[9
]
̇
=
(
×
(
(
)
∗
1
)
−
1
̇
1
)
̇
=
(
×
(
(
)
∗
2
)
−
1
2
)
(
17)
̇
=
(
×
(
(
)
∗
3
)
−
1
3
)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
El
ec &
C
om
p
En
g
IS
S
N: 20
88
-
8708
Desig
n a
nd p
e
rforma
nce c
ompa
ris
on
of d
if
fe
rent
adapti
ve con
tr
ol sc
hem
e
s
for pit
ch
...
(
Wi
ns
to
n
Ne
tt
o)
4119
The
sc
h
em
e for
a
dap
ti
ve
PID
is sh
own
i
n
Fi
gure
7.
Fig
ure
7
.
Bl
oc
k diag
ram
f
or
a
dap
ti
ve
P
ID
2.2.4.
Ap
pr
oxim
ate
a
d
ap
ti
ve
P
ID
The
co
ntr
oller
can
be
desig
ned
without
usi
ng
G
(
s
)
i.e.
know
le
dg
e
of
the
plant.
T
heorem
4:
The Frec
het de
rivati
ve
c
a
n be
appr
ox
im
at
e as [10
]
̇
[
]
=
ℎ
(18)
wh
e
re
ℎ
is
the
i
m
pu
lse
res
ponse
s
ub
sti
tuti
ng
in
(18
)
the
ada
ptive
la
w
s
bec
om
e
ind
epende
nt
of
(
)
.
T
he
a
ppr
ox
im
at
e algori
thm
g
iven
as i
n
[
10]
is
̇
=
−
1
̇
=
−
2
(
19)
̇
=
−
3
is t
he
a
da
ptive
g
ai
n, e is
the e
rror sig
nal and
1
,
2
,
3
are
ou
t
pu
ts
from
f
irst t
hr
ee
s
ub
syst
em
.
2.3.
Algori
th
m
Using
the
ab
ove
ada
ptive
la
ws
sim
ulatio
ns
wer
e
pe
rfor
m
ed
in
Ma
tl
ab
Si
m
ulink
env
i
our
nm
ent
with
the
non
-
li
nea
r
TRM
S
plant
pro
vid
e
d
by
Fe
edb
ac
k
In
st
rum
ents
Ltd.
thi
s
m
od
el
is
a
rep
li
ca
of
t
he
r
eal
plant
and
was
de
sig
ne
d
us
i
ng
gr
ey
box
m
od
el
li
ng
.
Figure 8
s
how
s the i
m
ple
m
entat
ion
o
f
T
he M
IT
ru
le
in
S
i
m
ul
ink
wh
ic
h
is
done
directl
y
by
us
ing
the
ada
ptiv
e
la
w
(5)
.
Mo
dif
ie
d
MIT
r
ul
e
is
i
m
p
lem
ent
ed
in
F
igure
9.
It
is
i
m
ple
m
ented
us
in
g
th
e
ada
ptive
la
w
(
10)
.
Fig
ur
e
10
s
hows
the
A
dapt
i
ve
PID,
t
his
can
be
im
plem
ented
us
in
g
(
17)
in
L
aplace
dom
ai
n.
Figure
11
sho
ws
A
ppr
ox
im
a
te
Ad
aptiv
e
PID
i
m
ple
m
ented
us
i
ng
(
19)
w
it
ho
ut
the r
e
fer
e
nce
m
od
el
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
4114
-
4129
4120
Fig
ure
8
.
Th
e
MIT rule
Fig
ure
9
.
Mo
dified
M
IT
r
ule
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
El
ec &
C
om
p
En
g
IS
S
N: 20
88
-
8708
Desig
n a
nd p
e
rforma
nce c
ompa
ris
on
of d
if
fe
rent
adapti
ve con
tr
ol sc
hem
e
s
for pit
ch
...
(
Wi
ns
to
n
Ne
tt
o)
4121
Fig
ure
10.
A
da
ptive P
I
D
Fig
ure
11
. A
pproxim
at
e
a
dap
ti
ve
P
ID
3.
RESU
LT
S
A
ND AN
ALYSIS
All
si
m
ulati
on
s
wer
e
perf
orm
ed
on
t
he
m
a
in
ro
t
or.
Cr
os
s
couplin
g
has
no
t
been
c
onsidere
d
an
d
ta
il
ro
t
or
will
b
e
ke
pt stat
ion
a
ry
unti
l st
at
ed
ot
he
rw
ise
.
3.1.
PID
c
ontr
oller
The
PID
wa
s
tun
e
d
usi
ng
Ro
ot
locus
te
ch
ni
qu
e
for
the
m
a
in
ro
t
or
.
Fi
gur
e
12
s
hows
th
e
PI
D
ste
p
respo
ns
e
a
nd
t
rack
i
ng
res
ponse
to
re
fer
e
nce
input,
a
la
r
ge
ov
e
rs
hoot
with
os
ci
ll
at
or
y
be
hav
i
our
to
ste
p
input
and an
ina
blit
y t
o
trac
k refe
re
nce inp
uts.
(
)
=
3
.
9
2
+
0
.
2
+
2
(20)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
4114
-
4129
4122
(a)
(b)
Fig
ure
1
2
. (a)
Step
respo
ns
e
(
b) Refere
nce
t
r
ackin
g wit
h
c
onve
rtio
nal PID
3.2.
Adap
tive
c
on
tro
ll
ers
3.2.1.
Adap
tiv
e PID a
nd
a
p
p
rox
im
at
e
PI
D
a)
Step
r
esp
onse
1)
Ad
a
ptive
PID
The
va
riat
ion
of
ada
ptive
gai
n
with
1
a
nd
it
s
eff
ect
on
t
he
s
yst
e
m
was
stu
died
as
s
how
n
in
Ta
ble
1,
the
ra
nge
of
va
lues
for
w
hich
syst
em
rem
a
ins
st
a
ble
or
sti
ll
has
trac
king
ca
pa
bili
ty
.
Step
re
spo
ns
e
f
or
=
0
.
004
with
it
s
P
I
D
a
nd
er
r
or
cha
ra
ct
ersti
cs
is
s
hown
in
Fi
gure
13,
PID
val
ues
conve
rg
e
to
s
om
e
final
value
s
with e
rror re
duci
ng
t
o
ze
ro.
T
able
1.
Var
ia
t
ion
i
n
ℎ
1
and it
s e
ff
ect
o
n t
he sy
stem
1
Sy
ste
m
0
.00
4
0
Stab
le
0
.00
9
0
Un
stab
le
0
.00
4
>
0
Tr
acki
n
g
los
t bu
t Stable
(a)
(b)
Fig
ure
13
. (a)
Show
s
the
stab
le
step r
e
spo
nse
, (b) T
he
P
ID
values
w
it
h er
r
or for
stable
re
sp
onse
2)
Appro
xim
at
e
Ad
a
ptive
PID
Fo
r
a
ste
p
in
p
ut
syst
em
beco
m
es
un
sta
ble
f
or
>
0
.
0013
as
see
n
f
ro
m
Table
2
.
The
sta
ble
a
nd
un
sta
ble
ste
p
r
esp
on
se
for
di
ff
e
ren
t
val
ues
of
is
s
how
n
in
Fig
ur
e
14,
f
or
high
value
of
a
dap
ti
ve
ga
in
the
syst
e
m
b
reak
s
dow
n
int
o osci
ll
at
ion
s.
T
able
2.
Var
ia
t
ion
i
n
an
d
it
s e
ff
ect
on the
syst
e
m
Sy
ste
m
0
.00
1
3
Stab
le
0
.00
5
Un
stab
le
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
El
ec &
C
om
p
En
g
IS
S
N: 20
88
-
8708
Desig
n a
nd p
e
rforma
nce c
ompa
ris
on
of d
if
fe
rent
adapti
ve con
tr
ol sc
hem
e
s
for pit
ch
...
(
Wi
ns
to
n
Ne
tt
o)
4123
(a)
(b)
Figure
14. (a)
Show
s
stable a
nd (b) u
ns
ta
ble step
res
pons
es
b)
Re
fer
e
nce
t
rac
king
T
rack
i
ng for a
dap
ti
ve
P
ID in
F
igure
15
a
nd
tracki
ng for ap
pro
x
a
da
ptive P
I
D
in
F
ig
ure
16
.
Fig
ure
15. Re
f
eren
ce
trac
king
w
it
h
Ad
a
ptive
PID
f
or γ=
0.
Fig
ure
16. Re
f
eren
ce
trac
king
w
it
h
Appro
xim
at
e
Ad
a
ptive
PID
f
or γ=
0.03
3.2.2.
A
dapti
ve
c
ontrolle
r
using mo
del r
eference
adap
tive
c
ontrol
a)
Step
respo
ns
e
1)
MIT Rule
The
sta
bili
ty
of
the
syst
em
f
or
diff
e
re
nt
va
lues
of
is
sho
wn
i
n
Table
3,
it
can
be
sai
d
that
the
syst
e
m
re
m
ai
n
s
sta
ble
f
or
−
0
.
35
≤
≤
0
.
35
.
Stable
an
d
unsta
ble
ste
p
re
spo
nse
for
MIT
R
ul
e
are
show
n
i
n
Figure
17 the
val
ues of
are in
acco
rd
a
nce
wi
th Ta
ble 3.
Table
3.
Var
ia
t
ion
i
n
γ
and its
eff
ect
on the
syst
e
m
Sy
ste
m
0
.35
Stab
le
0
.6
Un
stab
le
-
0
.35
Stab
le with in
v
erse
r
esp
o
n
se
Evaluation Warning : The document was created with Spire.PDF for Python.