Professional Documents
Culture Documents
Control and
Information Sciences
Edited by A.V. Balakrishnan and M.Thoma
12
lan Postlethwaite
Alistair G. J. MacFarlane
Springer-Verlag
Berlin Heidelberg New York 1979
Series Editors
A.V. Balakrishnan M. Thoma
Advisory Board
A. G. J. MacFarlane H. Kwakernaak Ya. Z. Tsypkin
Authors
Dr. I. Postlethwaite,
Research Fellow, Trinity Hall, Cambridge, and
SRC Postdoctoral Research Fellow,
Engineering Department, University of Cambridge.
This work is subject to copyright. All rights are reserved, whether the whole
or part of the material is concerned, specifically those of translation, re-
printing, re-use of illustrations, broadcasting, reproduction by photocopying
machine or similar means, and storage in data banks.
Under 54 of the German Copyright Law where copies are made for other
than private use, a fee is payable to the publisher, the amount of the fee to
be determined by agreement with the publisher.
by Springer-VerlagBerlin Heidelberg 1979
Printed in Germany
Printing and binding: Beltz Offsetdruck, Hemsbach/Bergstr.
2061/3020-543210
Contents
CHAPTER 1 Introducti0n 1
References 6
CHAPTER 2 Preliminaries 8
2.1 System description 8
2.2 Feedback configuration ii
2.3 Stability 13
2.3-1 Free systems 13
2.3-2 Forced systems 16
2.4 Relationship between open- and closed-loop
characteristic polynomials for the general
feedback configuration 17
References 20
References 169
Bibliography 170
Index 174
1. Introduction
Bode [2] and Evans [3],which were still in use for many
applications.
as a single object in its own right and to ask : how can the key basic
concepts of the classical single-loop frequency-response approach
ily break the flow of the text has been placed in appendices.
References are listed at the end of each chapter in which they
are cited, and also at the end of the text where a bibliography
is provided.
References
r" "~ 1
u{t}=u~(t} I
llYh(t)=
y(t)
I .~
I
I
'
I
S J
of subsystems
10
B2c I A2 tB2DlJ
application.
or i n p u t - o u t p u t description is o b t a i n e d if in e q u a t i o n s
Gi(s ) = Ci(SIn.-Ai)
1
-1Bi + Di (2.1.7)
of o r d e r m.
where
Ac = A-B(k-IIm+D)-Ic
B c = k B - k B (k-iIm+D)-iD
Cc = (Im+kD) -i c
Dc = (k-iim+D) -I D
y(t)
~l 7"'ml -11 1, "7"hi
2.3 Stability
This is because:
to time, and
Willems [3] .
2.3-1 Free systems
definitions
the origin.
A = TJT -I (2.3.3)
c
with
J = Jl
J2
Jk
J. = I. 1
l l
~i ".
hi 1
exp[ Jk (t-to)]
and
0 1 t tr-2/(r-2)'
000 1
where ~ = (t-t O )
and r is the order of the Jordan block Ji" The general
we then have
det[IIn-Ac] = O (2.3.8)
so that
OLCP(1) ~ det[IIn-A ]
= detflInl-Al]det [ IIn2-A 2] . . . .
return-difference operator [ 5 ] .
~(s)
(a)
i 0 0 ....
(b)
Figure 4. Feedback configuration
(a) dosed-loop
(b) open-toop
19
which is equivalent to
20
detF(s) = det
-I
% letlSn l+etSn
: Is L--~C--J, I S k
= det IsIn-A+B (k-IIm+D) -IC :--im+~
0 ]-- det[SIn-A ]
[----ic ...... -,
= det [SIn-A+B(k-iIm+D)-iC]det[Im+k~ (2.4.5)
det [SIn-A ]
Now from equations (2.2.2) we have
A c = A-B(k-IIm+D)-Ic
and it is obvious from equation (2.4.3) that
detF(~) = det[Im+k ~
and therefore under the assumption that det F(~)~O we have
from equation (2.4.5) the following relationship
(2.4.6)
The zeros of the open- and closed-loop characteristic
polynomials, OLCP(s) and CLCP(s), are known as the open-
and closed-loop poles or characteristic frequencies respectively.
Relationship (2.4.6) shows how the matrix-valued
rational transfer functions F(s) and G(s) are intimately
related to the stability of a dynamical feedback system.
The study of such matrices and their eigenvalues opens the way
to suitable extensions of the classical techniques of Nyquist
[7] and Evans [8;9] ; the results of such a study are given
in Chapter 3.
References
det[k-iIm+G(s~
= (3.1.3)
det[ k-iIm+D]
plane), we have
23
~ I n _ ] ......
frequency.
and
?(s,g) ~ det[sI n - S(g)] = 0 (3.2.2)
functions.
G(s) are different from the poles and zeros of the characteristic
poles and zeros of g(s) to the poles and zeros of G(s), and
ti ti-i
Ai(g's) = gi +ail(s)gi + ...... +ait. (s)=O
1 (3.3.3)
t. t.-i
bio(S)g i l+bil(s)g i i + ...... +biti(s) = O (3.3.4)
i=i,2,...,
{gi(s):i=l,2,...,~}.
put into Jordan form [2]. In general this will not be the
Let
O O . . O -aiti(s)
10 . . . O -ai,ti_l(S
(3.3.5)
C(Ai)~ O 1 . . . O -ai,ti_2(s)
O 1 -ail (s)
and is given by
function g(s) :
zero when
bt(s) = O (3.3.10)
30
and w i l l t e n d to i n f i n i t y as
bo(S) + O (3.3.11)
bo(S) = O (3.3.12)
bu(S) bv(S)
may be replaced by
where
bj(s) ~ O . . . . . bm(S) O
g(s) = 0
and (3.3.15)
where b~(s) and bt[ (s) are the monic polynomials obtained
we put
-i
s=z (3.3.16)
so that
it follows that:
function matrix
that
where the mm matrix H(s) and the matrix J(s) are both
unimodular (that is having a constant value for their determinants,
independent of s ). If r is the normal rank of T(s)
(that is T(s) has rank r for almost all values of s ) then
M(s) has the form
]
M(s) = [ M*(s)rr Or, Z-r I (3.3.20)
I
r,r Om-r,m-rJ
with
where:
(i) each ei(s) divides all ei+ j (s) and
T(s) = Hi(s)
Idiag{ ei (s) 7
~i-~-s~}j Jl (s)
r ci(s) t
= (3.3.23)
h i(s) ~ 3i(s)
i=l
where :
(i) {hi(s) : i = 1,2 .... ,r} are the columns of the
matrix H l(s) ;
34
matrix Jl(S)
We know that
r ~ min(,m)
by
(3.3.24)
ks(s)
g(s) =
~(s)
is defined as having zeros at those values of s where (s)
quantities
E(s)
some ~i (s) + O.
definitions [3].
and put
r
ZT(S) = ~ g. (s) (3.3.28)
i=l 1
The following rules [4] can be shown to give the same results
common denominator
known that:
that
Furthermore since
that
38
in which case eG(s) will be unity and the p01e and zero
written as
and
where the pole and zero polynomials for the jth characteristic
Example demonstrating th 9 p o l e V z e r o r e l a t i o n s h i p s
Let
(s-l) (s+2) O
1
G(s) z
i s+l
o] (s+l) (s+2) (s-l) (s+l)
-i 1
s-i s+2
ZG(S) = (s-l)
39
and
1
A 2 (g,s) = g - s+---2 = O
which may be w r i t t e n as
(s+l)g - 1 = O
and
(s+2) g - 1 = 0
-1
Pg2 (s) = b 2 0 (s) = (s+2) Zg 2(s) = b21(s) = 1
-i
e G (s) = (s-l)
which v e r i f i e s the r e l a t i o n s h i p s
2
PG(S) = eG(s) H b.; (s)
i= 1 lO
and
2
ZG(S) = eG(s) i=iH bilJ (s)
occurs only if
This last situation can occur for finite values of s if, and
b (s) = O or D (s) = O,
o g
or both, plus the point s =~.
Solutions of
D (s) = O
g
are called finite branch points of the characteristic gain
S u p p o s e we h a v e a r e p r e s e n t a t i o n of p a r t of one b r a n c h of
a representation is u s u a l l y c a l l e d a f u n c t i o n a l e l e m e n t .
t h a t the i n d i v i d u a l p o i n t s of t h e p a p e r d i s c are m a d e b e a r e r s
initital element is a n a l y t i c a l l y c o n t i n u e d by m e a n s of a
as i l l u s t r a t e d by f i g u r e 6. The p a r t s p a s t e d t o g e t h e r are
t r e a t e d as a s i n g l e r e g i o n c o v e r e d o n c e w i t h values. If a
1
Figure 7 Repeated anatytic continuation
have that:
C-i,
(ii) each analytic continuation of any of these
g(s) = -1 (3.3.38)
frequency loci.
frequency loci
open-loop gain m a t r i x
48
Cut
Cut
-I
-3
G(s) = 1 [s-i s]
1.25(s+i) (s+2) -6 s-2
and g reversed.
different.
functions involved.
=- 0.6
1 O.5
crosses.
52
Charocteristic
.... gain loci
mmmm Cut
Characteristic
gain loci
b C~%
~ Right. half
plane region
Left. half
plane region
. . . . . . . . Choract,erbtic
..... gain loci
Cut. between
branch points.
"l
-0"8~.533 19-2
L
533 1<3,2
sd = ezj8
Then ~ is the smallest (in magnitude) negative real contour
~- -0.05 and 8 = O
so that
sd ~ -0.05
sd = -0.0528
point loci, the root loci are simply the values of s at which
I ~ a l contours
labll~:l thus :
-06
-0.5
-0,4 0-5-
-0.3
4
-i.s ~ -,.o I -o:s..~f-o' ,!o 1.5
:2/ (lll :
0.9 ,'0 I"11~2 ~:3 ~.415
,o
-I.5-
applies.
is based.
that is
and (4.2.4)
OLCP(s)~det[SIn-A ]
where Px(S) has as its zeros those poles of GI(S),G 2 (s), ....
rewritten as
or
have that
detF(s) = ~. ZF(S) (4.2.13)
PF (s)
where 8 is a scalar quantity independent of s; and from
detF(~) = 8 (4.2.14)
ZF(S)
CLCP (s) = Pd(S)PG(S) (4.2.15)
P F (s)
~d~o (s)
i=l
pieced together to f o r m ~ f
. Then when the surface is
3
formed the set of Nyquist D-contours combine to form a set
Trn (s)
Radius r where
S -plane ) ..,......~ r-,-oobut#00
')///4/ / _
f
Re (s)
x Poles
O Zeros
z~ Branch points
therefore be replaced by
Zf. = 0 i=1,2,...,~
l
or (4.2.22)
N(fi,O) = - ~ P (4.2.23)
i=l i=l
axis; and
Z N(fi,O) = Z Zf - Z Pf (4.2.26)
i=l i=l ii=l i
To e s t a b l i s h the n e c e s s i t y of c o n d i t i o n (4.2.24)
suppose that
N(fi'O) ~ -PG
i=l
(a") Z N(fi'O) = - PG ;
i=l
(b") [fi(s) : i=1,2 .... ,} have no zeros on the imaginary
axis;
the Nyquist D-contour set mapped under fi(s) and kgi(s) are
(b") is equivalent to
1
gi(Jw) ~-~ , i=i,2, .... i (4.2.30)
(3.3.35), as
H b/ (s) (4.2.31)
PG(S) = eG(s) i=l Io
branches.
1
(i) Z N(gi' -k ) = - PG;
i=l
critical p o i n t (-kl---+jO);
(3) the n u m b e r of b r a n c h e s of the c h a r a c t e r i s t i c gain loci
Nyquist s t a b i l i t y criterion.
70
that
or
h
N(fi,O) = -Z PG (4.2.36)
i=l i=l
i
where PG~ is the n u m b e r of r i g h t h a l f - p l a n e p o l e s of Gi(s).
that
h
PG = e + Z Pfi + P (4.2.37)
i=l ~ i=l x
Pd = Px (4.2.39)
suppose that
h
7 N(fi,O ) ~ -
i=l il~--1PGi
then from equation (4.2.40) this implies that
Z Zfi ~ O, or e ~ O, or Pd ~ 0
i=l
i~iZfi = e = Pd = O
sufficiency of c o n d i t i o n (4.2.36) is e s t a b l i s h e d .
axis; and
s t a b i l i t y criterion, that
Z N(fi,O) ~ Z N(gi,-!)
k
i=l i=l
4.3 Example
open-loop gain m a t r i x
73
0s l
1.25(s+i) (s+2)
i s i-6s s-2
1
By d e f i n i t i o n the s y s t e m has no u n o b s e r v a b l e or u n c o n t r o l l a b l e
criterion is d i r e c t l y applicable.
loop u n s t a b l e for k = 1 . 2 5
]o.5o
~=2.0.
S ---,,-.- 00
~=0.05. f 60
60
~=-0.05 60 ~=1-0
~ =-2.,
-]o.5o
~ j l O=_o.s
" 0 = -I.0
1
(iv) For -~ = -0.4 the characteristic gain loci pass through
to positive feedback.
References
given by
inverse
g(s) = 1 (5.1.3)
g(s)
and therefore g(s) and g(s) have the same branch points,and
is based.
and
where HG(S) and JG(S) are both mxm unimodular matrices and
MG(S) = diag [ ~l
(s)1(s) ~2
(s)e2(s)-' ~m--~
)em(s) ] (5.1.7)
and
m
ZG(S) = H ei(s) (5.1.9)
i=l
inverted to give
G(s) -1 = JG(S)-IMG(S)-IHG(S) -I (5. i. lO)
m, and given by
O O ... 0 1
O O ... 1 O
E = : . = E
-i (5.1.11)
O 1 ... O O
1 O ... O 0
can be rewritten as
-i -i (5.1.12)
G(s) = JG (s) -IEEM G (s) -IEEHG (s)
or
G(s) -I = H~(s)M6(s)J~(s) (5.1.13)
80
where
H~(s) = JG(S)-iE (5.1.14)
-i
J~Cs) = EHG(S) (5.1.15)
and
M~(s) = EMG(S)-IE
and
m
z~(s) = H ~i(s) = PG(S) (5.1.18)
i=l
where p*(s)G and z~(s) denote the pole and zero polynomials
of G(s) -i respectively.
In sub-section 3.3-3 it is shown that the pole and
zero polymomials of G(s) are related to the poles and zeros
of the characteristic gain function set {gi(s) :i=1,2,...,} by
PG(S) = eG(s)i=l
K b ti(s) = eG(s)i~iPgi (s) (5.1.19)
and
(5.1.18) to give
and
and the output as that of the hth subsystem, are all in the
left half-plane.
by:
stable,
G(s) = 1 (5.4.1)
3
(s+l)
point the same number of times, and thereby cancel each other.
the mapping under g(s), from the set of Nyquist curves formed
right.
85
*I~h
20~
-20.
(o)
I.O
/ -~
,o( o,:: f.,~ = 6 o
-0.5 0.5 l.~l
~ I ~ ,03
(b)
relationship
this we obtain
det F(s) = k m det[ s -l]
det[G(s) -I]
=y z~(s) * (5.5.6)
g( )
87
where pf(s) and zf(s) are respectively the monic pole and
and therefore
(5.5.8)
and
z f(s) = z*(s)
r @
p*(s)
)
pf(s) (5.5.9)
we obtain
where:
and
where:
(i) N(gj,O) is the net sum of clockwise encirelements of
corresponding to gj(s);
we have
Z , ,
Z N(ri,O)- Z N(gi,O) = Z Z* - Z Z* (5.5.19)
i=l i=l i=l ri i=l gi
therefore be replaced by
rewritten as
are s a t i s f i e d is p r o v e d as follows
PG = e + Z Z* (5.5.22)
i=l gi
. Z .
Z N(ri,O) - Z N(gi,O) = Z Z* + e-P G (5.5.23)
i=l i=l i=l ri
To e s t a b l i s h the n e c e s s i t y of c o n d i t i o n (5.5.21)
s u p p o s e that
, ,
N(ri,O) - 2 N(gi,O) fl -PG
i=l i=l
91
for c l o s e d - l o o p stability.
of c o n d i t i o n (5.5.21) is e s t a b l i s h e d .
in s t a t e m e n t 1 of the s t a b i l i t y criterion.
in s t a t e m e n t 1 of the s t a b i l i t y criterion.
criterion.
the c o n d i t i o n
, ,
E N(r i, O) = Z N(gi,-k) = - Z G, (5.5.25)
i=l i=l
ZG = e + Z P, (5.5.26)
i=l gi
that
,
F N(ri,O ) ~ -Z G
i=l
for c l o s e d - l o o p stability.
is p r o v e d as follows.
(5.5.32) can be r e w r i t t e n as
h
~iPGi = e + Z Z* + Pd (5.5.33)
i i=l gi
to give
, , h
Z N(ri,O) - Z N(gi,O) = Z Z* + e + Pd - Z=IPGi
i=l i=l i=l ri i
(5.5.34)
w h i c h u s i n g e q u a t i o n (5.5.24) b e c o m e s
, i , h
Z N(gf-k) - 2 N(gi,O) = Z Z* + e +Pd - __ZIPGi
i=l i=l i=l ri i
(5.5.35)
To e s t a b l i s h the n e c e s s i t y of c o n d i t i o n (5.5.31)
s u p p o s e that
, , h
N(gi,-k) - Z N(g,O) ~ -Z PG.
i=l i=l i=l l
is n e c e s s a r y for c l o s e d - l o o p stability.
95
For s u f f i c i e n c y s u p p o s e t h a t
, , h
N(gi,-k) - Z N(gi,O) = - Z PG
i=l i=l i=l i
then f r o m e q u a t i o n (5.5.35) we m u s t h a v e t h a t
Z Z* = e = Pd = 0
i=l r i
(5.5.31) is e s t a b l i s h e d .
(ib) i.e.
i , , h
Z N(ri,O) = ~ N(gi,-k) = - ~ ZG. (5.5.36)
i=l i=l i=l l
Z N(ri,O) = Z N(gi,-k) = Z Z* - Z P*
i=l i=l i=l ri i=l ri
(5.5.37)
rewritten as
Z N(gi,-k) = Z Z* - Z P* (5.5.38)
i=l i=l ri i=l gi
t h e n we m u s t h a v e
h
ZG. = e + ~ Z + Pd (5.5.40)
i=l l i=l 9i
zeros, of gi(s),
h
ZG = e + Z P* + Pd (5.5.41)
i=l i i=l g i
to g i v e
, Z h
Z N(gi,-k) = ~ Z* + e + Pd - ~iZGi
i=l i=l ri i
(5.5.42)
that
Z . h
Z N(gi,-k) ~ - ~ ZG
i=l i=l i
closed-loop unstable. H e n c e we c o n c l u d e t h a t c o n d i t i o n
For s u f f i c i e n c y s u p p o s e that
, h
N(gi,-k) = - Z ZG
i=l i=l l
then f r o m e q u a t i o n (5.5.42) we m u s t h a v e t h a t
Z* = e = Pd = O
i=l r i
of c o n d i t i o n (5.5.36) is e s t a b l i s h e d .
5.6 Example
G(s) are
so that
PG = O and Z G = O.
statements.
-1.875 < k ~ 0
stability criterion.
It is i n t e r e s t i n g to n o t e that an i n v e r s e N y q u i s t - l i k e
98
~.,, "%.
g-plane
ca.~ 5.o~
\ ..... . /
feedback systems.
References
[6] A.I. Mees and P.E. Rapp, "Stability criteria for multiple-
loop nonlinear feedback systems", Proc. IFAC Fourth Multivariable
Technological Systems Symposium, Fredericton, Canada, 1977.
6. Multivariable root loci
steps towards this end have been made with regard to the
or the equation
and solutions of
F(k,s) = O (6.1.5)
equivalent to
CLCP(s)~det[SIn-Ac]= det[SIn-S(-k-l)] = O
(6.1.6)
I02
Si= 5 (ko+6k)
i\soo
o
Figure 20. To derive a brmuta for the tangent to
the characteristic frequency loci
s (k) ~ a + bk ~ (6.1.10)
r
where a and b are c o m p l e x n u m b e r s , ~ is a r a t i o n a l r e a l
(6.1.10) we h a v e
where
[ = O when e > O
T
= 1 when ~ < 0
In the f o l l o w i n g sections it is s h o w n h o w to o b t a i n
f i n d i n g the a s y m p t o t i c b e h a v i o u r or the a n g l e s of d e p a r t u r e
the p r o b l e m , by a c h a n g e of v a r i a b l e s in the c h a r a c t e r i s t i c
105
To d e t e r m i n e the a s y m p t o t i c b e h a v i o u r of the c h a r a c t e r i s t i c
-i
s=z (6.2.1)
to the e q u a t i o n
or if D is s i n g u l a r
oo = 0 (6.2.4)
the N e w t o n d i a g r a m , corresponds to t h e r e b e i n g no a s y m p t o t i c
n u m b e r of f i n i t e z e r o s as p o l e s a n d t h e r e f o r e we w o u l d not
expect any c l o s e d - l o o p p o l e s to a p p r o a c h i n f i n i t y .
z ~cg ~ (6.2.5)
N e w t o n d i a g r a m is as f o l l o w s [8].
(l-z+2z2-25z3+29z4)g 2 + (z-22z2+199z3-21Oz4)g
-(33z3-594z 4) = O (6.2.6)
the straight line PIP2 makes with the verticle z-axis determines
Zl
4,
3:
C i 2 a
tJ
3P,
0
2 a
~t
is n e c e s s a r y to s u b s t i t u t e z = cg in the terms of
form
x y~ gXlzY 1
x g z + ..... (6.2.8)
oy~ "+ ~xlY 1
x~-x I Xo_l-Xl _x 2 - x 1
Yc-Yl - YO-I-Yl ........ Y2 - Yl ~t
equation of t h ~ a f o r m
Yl
~xgy c + ........ + ~XlY 1 c = O (6.2.9)
Yl
or d i v i d i n g by c (c ~ O) we o b t a i n
Y~-Yl
~xay c + .... +~XlY 1 = 0 (6.2.10)
~t
sit - bit k (6.2.11)
real axis.
~(g,s) = O.
6.2-1 B u t t e r w o r t h Patterns
theory [8]
F (k,s) = O (6.2.12)
(6.2.12) has a p- fold root So, and let the n branches of s(k)
in a n e i g h b o u r h o o d N of k be {s.(k):i=l,2,...,p,...,n}
c l
where {si(kc)=Sc:i=l,2,...,p}.
Finally all the functions Sl(k), ..... Sp(k) will be sorted into
cyclical systems.
the c o r r e s p o n d i n g analytic f u n c t i o n w i l l be s i n g l e - v a l u e d
Suppose the n e i g h b o u r h o o d N is r e p l a c e d by a p i l e of
To d e n o t e an a n a l y t i c function determined by a c y c l i c a l
L cut
kc
Theorem 4
positive integer.
configurations.
~(g,s ) = O (6.3.1)
(6.3.1) becomes
s' ~ ed ~ (6.3.4)
Hd
s - B + bdk (6.3.5)
given as
ed = argument {b d} (6.3.6)
gXs'Y= O , = 0 (6.3.8)
x(g's') = ~xy oo
is o b t a i n e d where p is a c o m p l e x number a n d n is a r a t i o n a l
s - y + b ak~a (6.3.10)
given as
6.4 Example 1
0s l [3s3+s21s 8s22s+l 1
s4+5s3-2s2-44s+40 s3+79s2+44s-868 -4s3-4s2+4Os-32
w h i c h has a c h a r a c t e r i s t i c e q u a t i o n
~(g,s) = ( s 4 + 5 s 3 - 2 s 2 - 4 4 s + 4 0 ) g2 + (s3+l16s-432)g
-12 (s2-2s+2) = O
-12 (z2-2z3+2z 4) = O
we o b t a i n ~ = 1 , and h e n c e the a p p r o x i m a t i o n
z - cg
s - 4k and s - -3k as k + ~
real axis.
o ~
\ S
Pole at s = 1
-i
Putting s' = s - i and g = d in t h e characteristic
equation we obtain
s' ~ e l d
s - 1 + 10.86 k
the p o l e s = 1 is 0 .
Pole at s = 2
equation we obtain
- 12(s'2+2s'+2)d 2 = O
s' ~ e 2 d
frequency loci
118
s-2-4.Sk
the pole s = 2 is 1 8 0 .
Pole a t s = -4+2j
equation we obtain
s' ~- e 3d
912 - j 3 2 0
88 + j l O 4
frequency loci
(-912+j320) k
s - -4+2j + (88+9104)
-912+j32Q~ o
argument { 88+9104 j = 110.9
Pole at s = - 4 - 2 j
s = -4-2j is - 1 1 0 . 9 .
4i
2,
o
L
\
2 d ( I 2 =d
,\
O ---
0 I 2 "d I 2
8
Figure 28. Newton Figure 29. Newton
diagram for E3(d,s') diagram for X(g,s')
3-0
2.0
I I
6'0 7"0
2/
Figure 30. Complete characteristic frequency loci
120
Zero at s = i ~
obtain
(g,s') = [ S , 4 + ( 9 + j 4 ) s , 3 + ( 1 3 + j 2 9 ) S , 2 + ( _ 5 8 + j 4 6 ) S , + ( _ 1 8 _ j 3 8 ) ] g2
+[s'3+(3+j3)s'2+(l16+j8)s'+(-318+j118)] g
12[s2j2s '] = 0
The N e w t o n d i a g r a m for x(g,s') is shown in f i g u r e 29 f r o m
w h i c h we o b t a i n n = 1 , and h e n c e the a p p r o x i m a t i o n
s' - pg
The c o e f f i c i e n t p is c a l c u l a t e d to h a v e the v a l u e
-318 + jl18
j24
f r e q u e n c y loci
(318-jI18) k -i
s - 1 + j + j24
f318-jl18% _
argument L j24 ~ + 180 E 69.64
Zero at s = l-j
s = l-j is - 6 9 . 6 4 .
c h a r a c t e r i z e d by c o n s t a n t p h a s e and c o n s t a n t m a g n i t u d e
ChQracterJstic
:" ~: frequency loci
(a) I Cutjoining
branch points
(b)
2.1
f 1.95 I ]
4-05 4'0 -3"95 -3"9
iI
J
1.0
0.98
0.9~
0-94
0"9~
0"90
0.90
linear system
where
H(s) ~ R-IGT (-s) QG (s) (6.5.9)
in s.
half-plane solutions of
n (s) = -p (6.5.14)
section 6.2.
6.6 Example 2
s + 2 6],
1
S(s) (s+l) (s+2) (s+3) (s+4)
s + 3 1
Q = I
and
pR = pI
SO that
127
= O
+ (-25s 2 + 256) = O.
-i
The substitution s = z now gives
(l-z 2) 2(l-2z2)2(l-3z~2(l-4z2)2p 2 + z6(l-z2) (I-2z2) (l-3z2)l-4z 2)
(-2+5Oz2)p
+ z14(-25 + 256z 2) = 0
The Newton diagram for this last equation is shown in figure 34,
we obtain
c~
Q.
~o
LQ
3
129
Im (s)
. . . . J~ . . . . . . J .....
a)
I I
-40 -20 20 40 R (sl
(b)
Figure 35. Asymptotic behaviour of the optimal characteristic
frequency loci (plus right half-plane image)displayed
on the Riemann surface domain of q(s)
(a) sheet 1 of the surface
(b)sheet 2 of the surface
130
loci
-i
s~l.12 [ e x p
-
( j -~
k~ )]j p ~ ,k = 2,3,4
and
-I
s~ 1.37 [ e x p j (w+2kz)] p 8 , k = 2,3,4,5.
8
image are shown in figure 35. Note that they have been computed
References
useful.
can be determined.
functions
-~ID(k2.....kq),l
of
where
S (k) ~A (k 2 .... ,kq) -B (k 2 .... kq)~l-llm+D(k2 .... kq) ] -Ic (k2, ,kq)
-kl(S)-i respectively).
The branches of s (kj ), for k 3 real, clearly define the
introduced.
135
parameter surface
for k. imagine that an arc is drawn,
3
o
centre the origin, from a stable operating point kj until
7.3 Example
carriage.
136
ndutum
PlV~ll iage
kJ k2
/I/I////
Figure 37. Inverted pendulum
positioning system
by
L' = J + mL 2 (7.3.2)
mL
137
is the m o m e n t of i n e r t i a of the p e n d u l u m w i t h r e s p e c t to
the c e n t e of g r a v i t y .
form
and u s n g the n u m e r % c a l v a l u e s
F -i
- 1 s
M
1 - 1 kg -I
M
(7.3.4)
-2
= 11.65 s
L'
L' = O.842m
it can be f o u n d [2] t h a t
parameter (mass) g a i n m a r g i n = 1 kg
parameter (mass) p h a s e m a r g i n = 60
~. :~(M) ,.,~
4,
-ifD
--70
,z ReC~
t I0
"-~C
5"d~
-4
J..~cr43
M~
~cM)
~.C
IOaB
] 18o"
-4. o ~ o
1 1"70-
...~c
-,4o~
this work are not only applicable to gain and frequency, but
a deeper understanding of t h e s e d e v e l o p m e n t s by c o n t r o l e n g i n e e r s
possibly in the s t u d y of s t a b i l i t y u n d e r m u l t i - p a r a m e t e r
variations.
a s s o c i a t i o n of a m u l t i v a r i a b l e s y s t e m w i t h one or p o s s i b l y
more algebraic f u n c t i o n s e a c h of w h i c h is d e f i n e d on an
to be an i m p o r t a n t c h a r a c t e r i s t i c of a s y s t e m and it is felt
of a m u l t i v a r i a b l e s y s t e m to a s y s t e m w h i c h is e f f e c t i v e l y a set
s y s t e m w o u l d be e q u i v a l e n t to r e d u c i n g an m - s h e e t e d Riemann
surface, w i t h a g e n u s n u m b e r g r e a t e r t h a n or e q u a l to zero,
c o u l d be a f r u i t f u l line of r e s e a r c h . In a p p e n d i x 7 an
p o i n t on a g a i n s u r f a c e c o r r e s p o n d s to a b r a n c h p o i n t or
v i c e versa.
142
References
References
of the form
A(q,v) = O (AI.2)
function v(q).
P ~
by A y r e s [2J for finding the r a t i o n a l canonical form of a
Definitions: If the v e c t o r s
X,GX,G2X,...,Gt-Ix (A2.1)
are l i n e a r l y independent but
2
X,GX,G X,...,Gt-Ix,GtX (A2.2)
are not then (A2.1) is c a l l e d a c h a i n of length t having X
as its leader.
of C m and Cm_ 1 ;
and so on. Then, for
t.-I tj+l-iXj+l ;
E = [ Xj,GXj .... G 3 ~ j + l , G X j + 1 .... ,G
t -i
... ; X m , G X m ..... G m Xm ]
145
length.
(s-3) -2 0
2 2
(s+l) (s+l)
-i - (s+2) 1
2 - - 2 s+l
(s+l) 2 (s+l)
146
1
then GX = (s+l) X
and
= s+l O O
0 0 s
(s+l) 3
1
O 1 (s+l)
are
1
(g,s) = g s+l
2 1
~(g,s) = g (s+l) g (s+l)
Note that these are also obvious from the dependence relations
A~pendix 3. The d i s c r i m i n a n t
the d i s c r i m i n a n t of an e q u a t i o n of the f o r m
Method 1 (Barnett [ 3 ] )
a(g) and c ( g ) , g i v e n by
n-l+
a(g) = a o g n + alg ... + a n
m m-i
c(g) = Cog + clg + ... + c m
aO aI a2 ... an 0
0 a aI ... an_ 1 an
m
rows
a -i a
n n
R[a(g),c(g)] = .oo
co c I .. Cm_ 1 cm
. , o C
o cI c 2 0 . cm O
ooo n
rows
co C1 c2 ..-
L e t the d e r i v a t i v e w i t h r e s p e c t to g of the p o l y n o m i a l
by
_ 1 R [ a(g) a' (g) ]
Dg(ao'''''an) a
0
the d i s c r i m i n a n t D g ( a o , . . . , a n) is zero.
of ~(g,s)
Consider the p o l y n o m i a l
n algn-i n > O
a(g) = dog + + ... + a n ,
the e x p r e s s i o n
Dg(ao'''''an ) = ao2n-2p
where P is a d e t e r m i n a n t given by
P = So ~I "" On-1
... (~
G1 o2 n
= n
o
aI + ao~ 1 = 0
2a 2 + al(l 1 + ao~ 2 = 0
+ + = 0
anO m a n_ i m + l + .. aoOn+ m
Dg(S) = . 2t-2
DO (S) P
to an o p e n - l o o p gain m a t r i x G(s)
axis except for possible cuts along the real axis. Also,
equations.
p r o c e d u r e d e s c r i b e d in s u b - s e c t i o n 3.3-4 where i n d i v i d u a l
151
I m (s)
5- plane
indicates continuity
between eigenvalues
Re (s)
X x indicates continuity
BrQnch points
/- i
~T--~-
i
< I--" Branch point
t-- --~--
(a) (b)
Figure 43. Cuts parallel to the imaginary axis
(a) finite (b) infinite
R = Real invalue
to, and a "large" distance away from, the real axis. The
45. With this new approach the only possible cuts are
Ira(s}
,Y, X X X X
~K X
S -plane
) :K
I
z z :_
Re is)
Cut
3_o
Cut
e
;~;w---.~'~': R o o t loci
Cut
A5.1 Introduction
Suppose that g(s) has four branch points, and that each
[s)~ ~-=Bronchpoint
~ branch
_
-
Cut between
points
Argument.
A(f,s) = 0
and f is defined.
G e n e r a l i z e d Residue T h e o r e m
in ~, we have
respect to ~.
161
to give
A5.3 Calculation of R e s i d u e s
represented by
~(s) t (A5.3.2)
i.e. f(s) = (s-Pi) Pi
-t
f'(s) Pi
+ ~'(s) (A5.3.3)
f(s) ~pi ) ~(s)
!
<s)
and since is a n a l y t i c in the n e i g h b o u r h o o d of D.
"l
~(s)
s = Pi"
For a zero s. of o r d e r t the a l g e b r a i c function can
i zi
be r e p r e s e n t e d by the series
co
t
f'(s) = zi S' (s)
+ (A5.3.6)
f(s) (s - z i) e(s)
f'(s)
and we see that f(s) has a s i m p l e ~ l e of r e s i d u e
t at s = z..
z. 1
1
To d e t e r m i n e the r e s i d u e s at a b r a n c h p o i n t it is
procedure is as follows.
In the n e i g h b o u r h o o d of a finite b r a n c h p o i n t b.
l
form [8],
163
oo
expansion
have that
o0
f(s) =
dn(r s _ ~ i ) n , do ~ O (AS. 3. iO)
n=O
the substitution
s = b. + x r (A5.3.11)
l
become
f(x) = Z dnxn , d_tb ~ 0 (A5.3.12)
n=-t~i Pi
and
164
co
A 1 Sf'(s) ds
(residue)bl = 249 - cf(~ (A5.3.16)
A 1 / f ' (s)
(residue)b. = 2~j I f ( s ) ds
i C
_ 1 / f' (x) dx ds
27 f(x) as
c
X
f'( (s)
of s----~
f at a branch point b.1 is therefore equal to the
f'(x)
f(x) is e a s i l y o b t a i n e d f r o m its s e r i e s e x p a n s i o n . F o l l o w i n g
p o i n t is a zero of o r d e r t b l;
zi
otherwise f'(x) is a n a l y t i c .
f(x)
equation (A5.2.2) we h a v e t h a t
2~j f(s) i i Z Pi i Pi
~n
= WZ -~P (A5.4.1)
as t h e i r o r d e r s indicate.
N(F,O), we f i n a l l y o b t a i n the r e q u i r e d e x t e n d e d P r i n c i p l e of
the A r g u m e n t in the f o r m
N(F,O) = # Z - ~ p (A5.4.2)
166
equation ~(g,s)=O
C o n s i d e r the c h a r a c t e r i s t i c equation
~(g,s) = O (A6. i)
s - bk ~ (A6.2)
s = p + bk ~ (A6.3)
or 1
k : (A64)
The d e p e n d e n c e of the c l o s e d - l o o p p o l e s on k is g i v e n by
F(k,s) = O (A6.5)
k and p. S i n c e the o r d e r of s w i l l in g e n e r a l be m u c h
g r e a t e r t h a n the o r d e r of k in e q u a t i o n (A6.5), we w i l l c o n s i d e r
Z(p,s) = 0 (A6.6)
-i
we r e q u i r e p for s =~, and so w e put s=z in e q u a t i o n (A6.6)
to g i v e
~(p,O) = O (A6.8)
167
Example
Consider the open-loop gain matrix
s4-s3+2s2-25s+29
[ 41s2-s-91 33s2-170s+l18
which has a characteristic equation
A(g,s)=(s4-s3+2s2-25s+29)g2-(-s3+22s2-199s+210)g +(-33s+594)=O
-33(-l+22z-199z2+210z3) (-2p+p2z)-33(22-199z+210z 2)
+(-33+594z) (-40+6zp2-4z2p3+z3p4)+594=O
188
Therefore
(p, O) = 3 3 2 - 6 6 p - 3 3 x 2 2 + 3 3 4 p + 5 9 4 = O
f r o m w h i c h we find
p=-14.5
Let us s u p p o s e t h a t f r o m an o p e n - l o o p g a i n m a t r i x G(s),
we h a v e o b t a i n e d an a l g e b r a i c e q u a t i o n in t e r m s of the c o m p l e x
F ~ f(g,s) = O (A7.1)
derivatives of F:
and
t ~F ~
<T J -- o CA7.3
g
If we c o n s i d e r g as a f u n c t i o n of s, the t o t a l d e r i v a t i v e
of F w i t h r e s p e c t to s is
dF ~F ~F dg (A7.4)
--ds + ds
g s
Alternatively we can c o n s i d e r s as a f u n c t i o n of g, a n d o b t a i n
dF ~F SF ds
s g
8F DE dg (A7.6)
= ds
g s
and
-- ds
d-~ (A7.7)
s g
that
(~_FF~ = O or dg = 0
%~gJ s ds
(~~F
)s g = O or ds
dg o
surface b r a n c h point.
References
Bibliography