Documente Academic
Documente Profesional
Documente Cultură
Dynamical Systems
2
Stability Theory of Large-Scale Dynamical Systems
1st edition
2014 A.A. Martynyuk & V.G. Miladzhanov & bookboon.com
ISBN 978-87-403-0771-9
3
Stability Theory of Large-Scale
Dynamical Systems Contents
Contents
Preface 8
Acknowledgements 10
Notation 11
1 Generalities 13
1.1 Introduction 13
1.2 Some Types of Large-Scale Dynamical Systems 13
1.3 Structural Perturbations of Dynamical Systems 23
1.4 Stability under Nonclassical Structural Perturbations 27
1.5 Method of Stability Analysis of Motion 31
1.6 Notes and References 34
www.sylvania.com
We do not reinvent
the wheel we reinvent
light.
Fascinating lighting offers an infinite spectrum of
possibilities: Innovative technologies and new
markets provide both opportunities and challenges.
An environment in which your expertise is in high
demand. Enjoy the supportive working atmosphere
within our global group and benefit from international
career paths. Implement sustainable ideas in close
cooperation with other specialists and contribute to
influencing our future. Come and join us in reinventing
light every day.
Light is OSRAM
360
.
4 Impulsive Large-Scale Systems 119
thinking
4.1 Introduction 119
4.2 Nonclassical Structural Perturbations in the Impulsive Systems 119
4.3 Definitions of Stability 122
4.4 Tests for Stability and Instability Analysis 122
4.5 Linear Systems Analysis 138
360
thinking . 360
thinking .
Discover the truth at www.deloitte.ca/careers Dis
Discover the truth at www.deloitte.ca/careers Deloitte & Touche LLP and affiliated entities.
References 226
Index 237
Do you like cars? Would you like to be a part of a successful brand? Send us your CV on
We will appreciate and reward both your enthusiasm and talent. www.employerforlife.com
Send us your CV. You will be surprised where it can take you.
7
Stability Theory of Large-Scale
Dynamical Systems Preface
PREFACE
9
Stability Theory of Large-Scale
Dynamical Systems Acknowledgements
ACKNOWLEDGEMENTS
A. A. Martynyuk
V. G. Miladzhanov
Typeset by AMS-TEX
10
Stability Theory of Large-Scale
Dynamical Systems Notation
NOTATION
AXA Global
Graduate Program
Find out more and apply
1
GENERALITIES
1.1 Introduction
In
2 this Section the notions1. of motion stability
STABILITY THEORYcorresponding to the mo-
tion properties of nonautonomous systems are presented being necessary in
subsequent presentation. Basic notions of the method of matrix-valued Li-
Typeset by AMS-TEX
apunov functions are discussed and general theorems and some corollaries
1
are set out.
Throughout this Section, real systems of ordinary differential equations
will be considered. Notations will be used.
dyi
(1.2.1) = Yi (t, y1 , . . . , yn ), i = 1, 2, . . . , n,
dt
dy
(1.2.2) = Y (t, y),
dt
dy
(1.2.3) = Y (t, y),
Download free eBooks at bookboon.com y(t0 ) = y0 ,
dt
13
for any (t0 , y0 ) T , 0 and is an open connected subset of Rn .
Let y(t) = (t; t0 , y0 ) be the solution of system (1.2.2), definite on the
interval [t0 , ) and noncontinuable behind the point , i.e. y(t) is not
definite for t = , . Then
dy
(1.2.2) = Y (t, y),
dt
Stability Theoryn of Large-Scale
where x R , Y (t, y) = (Y1 (t, y), . . . , Yn (t, y))T , Y : T Rn Rn .
Dynamical Systems Generalities
We will assume that the right-hand part of (1.2.2) satisfies the solution
existence and uniqueness conditions of the Cauchy problem
dy
(1.2.3) = Y (t, y), y(t0 ) = y0 ,
dt
Using solution y(t) and the right-hand part of system (1.2.2) we construct
the vector-function
dx
(1.2.6) = f (t, x).
1.2 TYPES OF LARGE-SCALE SYSTEMS 3
dt
It is easy to verify that the solutions of systems (1.2.2) and (1.2.6) are
correlated as
x(t) = y(t) (t)
on the general interval of existence of solutions y(t) and (t). It is clear that
system (1.2.6) has a trivial solution x(t) 0. This solution corresponds
to the solution y(t) = (t) of system (1.2.2). Obviously, the reduction of
system (1.2.2) to system (1.2.6) is possible only when the solution y(t) =
(t) is known.
Qualitative investigation of solutions of system (1.2.2) relatively solution
(t) is reduced to the investigation of behavior of solution x(t) to system
(1.2.6) which differs little from the trivial one for t = t0 .
In motion stability theory system (1.2.6) is called the system of perturbed
motion equations.
Since equations (1.2.6) can generally not be solved analytically in closed
from, the qualitative properties of the equilibrium state are of great prac-
tical interest. We start with a series of definitions.
(ii) uniformly stable iff both (i) holds and for every > 0 the corre-
sponding maximal M obeying (i) satisfies
(iv) uniformly stable in the whole iff both (ii) and (iii) hold;
(v) unstable iff there are t0 Ti , (0, +) and T0 , > t0 ,
suchfree
Download that for every
eBooks (0, +) there is x0 , x0 < , for which
at bookboon.com
x( ; t0 , x0 ) . 14
inf [M (t, ) : t Ti ] > 0;
(iv) uniformly stable in the whole iff both (ii) and (iii) hold;
(v) unstable iff there are t0 Ti , (0, +) and T0 , > t0 ,
such that for every (0, +) there is x0 , x0 < , for which
4 1. x( ; t0 , x0 )
STABILITY .
THEORY
In the investigation of both system (1.2.2) and (1.2.11) the solution x(t)
is assumed to be definite for all t T (for all t T0 ).
Further, with reference to system (1.2.6) we introduce the notations
xT = (xT T T T n
1 , x2 , . . . , xm ) R , xs Rns ,
f T (t, x) = (f1T (t, x1 ), . . . , fm
T
(t, xm ))T ,
(1.2.7) m
T
g (t, x) = (g1T (t, x), . . . , gm
T
(t, x))T , ns = n.
s=1
System (1.2.6) has the meaning of a large scale system, if for its dimen-
sions being large enough the decomposition to the form
dxs
(1.2.8) = fs (t, xs ) + gs (t, x1 , . . . , xn ), s = 1, 2, . . . , m,
dt
dxs
(1.2.9) = fs (t, xs ), s = 1, 2, . . . , m,
dt
(1.2.10) gs : gs (t, x1 , . . . , xn ), s = 1, 2, . . . , m,
(1.2.11) x( + 1) = f (, x( )),
where N+ {0 , . . . , 0 + k, . . . }, 0 0, k = 1, 2, . . . , x Rn ,
f : N+ Rn Rn , f (, x) is continuous in x. Let solution x( ; 0 , x0 ) of
system (1.2.11) be definite for all N+ and x(0 ; 0 , x0 ) = x0 . Assume
that f (, x) = x for all N+ iff x = 0. Besides, system (1.2.11) admits
zero solution x = 0 and it corresponds to the unique equilibrium state of
system (1.2.11).
The definitions of the dynamical properties of solutions of system (1.2.11)
are obtained by replacing the independent variable t R by N+ in
Definitions 1.2.1 1.2.3 and so are omitted.
Stability (instability) of the equilibrium state x = 0 of system (1.2.11)
is sometimes studied by means of reducing this system to the form
xi ( + 1) = fi (, xi ( )) + gi (, x1 ( ), . . . , xm ( )),
(1.2.13)
i = 1, 2, . . . , m,
(1.2.14) x( + 1) = fOF
1.2 TYPES (,LARGE-SCALE
x( )) + g(, x( )) H(, x( )).
SYSTEMS 7
(1.2.15) xi ( + 1) = fi (, xi ( )), i = 1, 2, . . . , m,
each of the latter can possess the same degree of complexity of the solution
behavior as the system (1.2.11).
Download free eBooks at bookboon.com
1.2.3 Ordinary impulsive large-scale systems The impulsive system
of differential equations of general type 17
dx
= f (t, x), t = k (x),
(1.2.16) dt
if g(, x( )) 0 , this system falls into the independent subsystems
(1.2.15) xi ( + 1) = fi (, xi ( )), i = 1, 2, . . . , m,
Stability Theory of Large-Scale
Dynamical
each of theSystems
latter can possess the same degree of complexity of the solution Generalities
behavior as the system (1.2.11).
dx
= f (t, x), t = k (x),
(1.2.16) dt
x = Ik (x), t = k (x), k = 1, 2, . . . ,
dxj
= fj (t, xj ) + fj (t, x), t = k (x), j = 1, 2, . . . , m,
(1.2.17) dt
xj = Ikj (xj ) + Ikj (x), t = k (x), k = 1, 2, . . . .
T0 S() = [t0 , ) {x : x 0 }, t0 0;
93%
OF MIM STUDENTS ARE
WORKING IN THEIR SECTOR 3 MONTHS
FOLLOWING GRADUATION
MASTER IN MANAGEMENT
STUDY IN THE CENTER OF MADRID AND TAKE ADVANTAGE OF THE UNIQUE OPPORTUNITIES
Length: 1O MONTHS
THAT THE CAPITAL OF SPAIN OFFERS
Av. Experience: 1 YEAR
PROPEL YOUR EDUCATION BY EARNING A DOUBLE DEGREE THAT BEST SUITS YOUR
Language: ENGLISH / SPANISH
PROFESSIONAL GOALS
Format: FULL-TIME
STUDY A SEMESTER ABROAD AND BECOME A GLOBAL CITIZEN WITH THE BEYOND BORDERS
Intakes: SEPT / FEB
EXPERIENCE
T0 S() = [t0 , ) {x : x 0 }, t0 0;
Stability Theory of Large-Scale
(5) functions
Dynamical Systemsk (x), k = 1, 2, . . . , and number satisfy conditions ex- Generalities
cluding beating of solutions of system (1.2.16) against the hyper-
surfaces Si : t = k (x), k = 1, 2, . . . , t 0.
We assume on system (1.2.17) that
(1) xj = (0, . . . , 0, xT T n
j , 0, . . . , 0) R , xj R ,
nj
f = (f1T , . . . , fmT T
) , fj (t, x) = fj (t, x) fj (t, xj );
T T T T
(2) Ikj = (Ik1 , Ik2 , . . . , Ikm ) , Ikj
(x) = Ikj (x) Ikj (xj ), n = n1 +
8 1. STABILITY THEORY
+ nm .
The state of the j-th noninteracting impulsive subsystem is described by
the equations
dxj
= fj (t, xj ), t = k (xj );
(1.2.18) dt
xj = Ikj (xj ), t = k (xj ).
19
Stability Theory of Large-Scale
Dynamical Systems Generalities
1.2 TYPES OF LARGE-SCALE SYSTEMS 9
Let the solution x(t) satisfy one of the conditions ( or ) and be definite
on [a, ), a R. Besides, the solution x(t) is referred to as unboundedly
continuable to the right.
Let the solution x0 (t) = x(t; t0 , y0 ) of system (4.2.1) exist for all t t0
and be unperturbed. We assume that x0 (t) reaches the surface Sk : t =
k (x) at times tk , tk+1 > tk and tk as k .
Definition 1.2.6 Solution x0 (t) of system (1.2.16) is
(i) stable, if for any tolerance > 0, > 0, t0 R+ a =
(t0 , , ) > 0 exists such that condition x0 y0 < implies
x(t) x0 (t) < for all t t0 and |t tk | > , where x(t) is an
arbitrary solution of system (1.2.16) existing on interval [t0 , );
(ii) uniformly stable, if in condition (1) of Definition 1.2.6 does not
depend on t0 ;
(iii) attractive, if for any tolerance > 0, > 0, t0 R+ there
exist 0 = 0 (t0 ) > 0 and T = T (t0 , , ) > 0 such that when-
ever x0 y0 < 0 , then x(t) x0 (t) < for t t0 + T
and |t tk | > ;
(iv) uniformly attractive, if 0 and T in condition (3) of Definition 1.2.6
do not depend on t0 ;
(v) asymptotically stable, if conditions (1) and (3) of Definition 1.2.6
hold;
(vi) uniformly asymptotically stable, if conditions (2) and (3) of Defini-
tion 1.2.6 hold.
dxi
(1.2.19) = fi (t, x, y), i = 1, 2, . . . , q,
dt
dyj
(1.2.20) j = gj (t, x, y, M ), j = 1, 2, . . . , r,
dt
and then
10 1. STABILITY THEORY
Stability Theorythat
it is necessary of Large-Scale
for x(t0 ) x0 (t0 ) < any solution x(t) of the system
Dynamical Systems Generalities
remain in the neighborhood of solution x0 (t) for all t [t0 , ), and point
t0 is not to be the discontinuity point of solutions x(t) and x0 (t).
dxi
(1.2.19) = fi (t, x, y), i = 1, 2, . . . , q,
dt
dyj
(1.2.20) j = gj (t, x, y, M ), j = 1, 2, . . . , r,
dt
and then
Mm = {M : 0 < j < jm j [1, r]},
t t0
(1.2.21) tj = , j = 1, 2, . . . , r.
j
tj
(1.2.22) = j , j = 1, 2, . . . , r,
t1
(1.2.23) j [ j , j ], j = 1, 2, . . . , r,
tj
(1.2.22) = j , j = 1, 2, . . . , r,
t1
(1.2.23) j [ , ],
1.2 TYPES j
j LARGE-SCALE
OF
j = 1, 2, . . . , r,
SYSTEMS 11
dxi
(1.2.25) = fi (t, x, y),
dt
dyi
(1.2.26) i = gi (t, x, y, M ),
dt
dxi
(1.2.27) = fi (t, xi , y i ),
dt
dyi
(1.2.28) i = gi (t, xi , y i , M ),
dt
where
dxi
(1.2.29) = fi (t, xi , y i ),
dt
(1.2.30) 0 = gi (t, xi , y i , 0),
dyi
(1.2.31) j = gi (, bi , y i , 0)
dt
22
Stability Theory of Large-Scale
Dynamical Systems Generalities
12 1. STABILITY THEORY
dxi
(1.2.32) = fi (t, x, y), i = 1, 2, . . . , q,
dt
(1.2.33) 0 = gj (t, x, y, 0), j = 1, 2, . . . , r,
dyj
(1.2.34) j = gj (, b, y, 0), j = 1, 2, . . . , r,
dt1
dxi
(1.2.35) = fi (t, xi , 0), i = 1, 2, . . . , q,
dt
dxi
(1.2.36) = fi (t, x, 0), i = 1, 2, . . . , q,
dt
respectively.
The separation of the time scales in the investigation of the stability of
the system (1.2.19), (1.2.20) is essential since the analysis of the degenerate
system (1.2.29), (1.2.30) and the fast system (1.2.31) is simpler problem
than that of general problem of stability of the system (1.2.19), (1.2.20).
Stability analysis of systems of (1.2.19) and (1.2.20) type under nonclas-
sical structural perturbations is the subject of Chapter 5. In this chapter
the development of the direct Liapunov method in terms of matrix-valued
functions is proposed.
1.6 STRUCTURAL PERTURBATIONS 13
The processes and phenomena of the real world are modeled correctly by
the systems of equations or inequalities only when the model admits small
changes. In other words the phenomenon model is correct provided that
it allows some uncertainties in definitions of both the parameters and the
external effects on the real system or process and at the same time displays
the main properties of the modeled process.
dx
(1.3.1) = g(x), x M,
dt
Download the
determine free vector
eBooksfield
at bookboon.com
on the compact manifold M . The naive specu-
lations above lead to the following notion of structural
23 stability.
Definition 1.3.1 (see Arnold [1]) System (1.3.1) is structurally stable,
if for arbitrary small changes of the vector field the obtained system is
equivalent to the initial one in the sense of fixed dynamical property.
the main properties of the modeled process.
CLICK HERE
to discover why both socially
and academically the University
of Groningen is one of the best
places for a student to be
www.rug.nl/feb/education
dx
(1.3.2) = P x, x(t0 ) = x0 ,
dt
where x Rn and P = C + F (t, x), C is a constant matrix, F (t, x)
is unknown in general matrix function with bounded real elements in the
domain R+ , Rn , is a real parameter. The fact that the system
(1.3.2) is of the form
dx
(1.3.3) = Cx + F (t, x)x
dt
yields that for = 0 the system (1.3.3) does not have structural pertur-
bations and the properties of its equilibrium state x = 0 are completely
determined by signs of real parts of roots of the characteristic equation
(1.3.4) det (C E) = 0.
dx
(1.3.5) = f (t, x, E),
dt
Download
where x free
Rn , eBooks
f : R+atbookboon.com
Rn Rn , is decomposed into m interconnected
subsystems 25
dxi
(1.3.6) = fi (t, xi ) + gi (t, x), i = 1, 2, . . . , m,
dt
where x Rni , g : R Rni Rni , g : R Rn Rni .
part and the terms bearing the information on structural and/or parametric
perturbations. Anyway the parametric perturbations must be small and
such that the solutions of system (1.3.3) must exist on the interval not
Stability Theory
smaller than of on
that Large-Scale
which the dynamics of system (1.3.3) is studied.
Dynamical Systems
This Chetaevs idea is used in the implicit form in modern nonlinear Generalities
dynamics of systems with uncertain parameter values.
dx
(1.3.5) = f (t, x, E),
dt
where x Rn , f : R+ Rn Rn , is decomposed into m interconnected
subsystems
dxi
(1.3.6) = fi (t, xi ) + gi (t, x), i = 1, 2, . . . , m,
dt
where xi Rni , gi : R+ Rni Rni , gi : R+ Rn Rni .
It is assumed that system (1.3.5) and the free subsystems
dxi
(1.3.7) = fi (t, xi ), i = 1, 2, . . . , m,
dt
satisfy the existence conditions for solutions x(t, t0 , x0 ) for all (t0 , x0 )
R+ Rn and f (t, 0) = fi (t, 0) = 0 for all t R+ , i.e. the motions of
system (1.3.5) and subsystems (1.3.7) can be realized on any given time
interval.
In order to take into account the mutual interaction between subsystems
(1.3.7) in system (1.3.5) and the dynamical properties of the initial system
(1.3.5) the binary elements eij of the interaction matrix E are introduced
in the form
1, i-subsystem acts on j-subsystem,
eij =
0, i-subsystem does not act on j-subsystem.
(1.3.8)
16
gi (t, x) = gi (t, 1.
ei1STABILITY . . , eim xm ),
x1 , ei2 x2 , . THEORY i = 1, 2, . . . , m.
dx
(1.4.2) = Q(t, x, P, S), x(t0 ) = x0 ,
American online
dt
where t0 R+ and x0 .
LIGS University
The function = (t; t0 , x0 ) is a solution to initial value problem (1.4.2)
for any (P, S) P S if and only if is a solution of the integral equation
is currently enrolling
t in the
(1.4.3) (t) = xBBA,
Interactive Online 0+ MBA, MSc,
Q(, ( ), P, S) d
DBA and PhD programs:
t 0
pay in 10 installments
(1.4.4) (T )(t) = x0 + Q(, ( ), P, S) d.
/ 2 tyears
Interactive Online education
0
dx
(1.4.2) = Q(t, x, P, S), x(t0 ) = x0 ,
dt
where t0 R+ and x0 .
The function = (t; t0 , x0 ) is a solution to initial value problem (1.4.2)
for any (P, S) P S if and only if is a solution of the integral equation
t
(1.4.3) (t) = x0 + Q(, ( ), P, S) d
t0
t
(1.4.4) (T )(t) = x0 + Q(, ( ), P, S) d.
t0
= T
dx
(1.4.5) = Q(t, x, P , S )
dt
dy
(1.4.6) = Q(t, x, P , S ) + Q(t, x, P, S),
dt
where
18 Q(t, x, P, S) = Q(t,
1. x,
STABILITY x, P , S ).
THEORY
P, S) Q(t,
Q(t, x , P , S ) Q(t, x , P , S ) Kx x
Stability Theory
is satisfied for |tofLarge-Scale
t0 | a, x x0 b, P P c, for given S S.
Dynamical Systems Generalities
H3 . For any (P, S) P S 2 = max(Q(t, x, P, S) for |t t0 |
h) < k < +.
Proposition 1.4.1 Under conditions H1 H3 there exists a unique solution
x(t) = x(t, t0 , x0 , P, S) of system (1.4.1) determined for |t t0 | h, h =
min(a, b/M ) where
were the value 1 characterizes the initial deviations of solutions x(t) and
y(t) of systems (1.4.5) and (1.4.6) for t = t0 , i.e. x0 y0 1 .
Estimate (1.4.7) allows one to show that the solutions of systems (1.4.5)
and (1.4.6) depend continuously on the system structure and/or parameter
only on the finite time interval. Hence it follows closeness of the appropriate
solutions on finite interval.
The problem on closeness of solutions to systems (1.4.5) and (1.4.6) on
infinite interval is a subject of special investigation of theory of stability un-
der nonclassical structural perturbations
1.6 STRUCTURAL which is basic in this monograph.
PERTURBATIONS 19
We add one more assumption to H1 H3 .
H4 . System (1.4.1) possesses a trivial solution x = 0, which is preserved
for any (P, S) P S.
Since further solutions of system (1.4.1) are considered on the infinite
time interval, we recall some conditions ensuring the existence of such so-
lutions.
Proposition 1.4.2 Let vector-function Q(t, x, P, S) be definite and
continuous in the domain of values (t, x) R+ Rn for any (P, S) P S
and in this domain the inequality
For the functions of the class SL we shall cite some definitions which are
applied in the investigation of dynamics of system in the book.
32
Stability Theory of Large-Scale
Dynamical Systems Generalities
NOTES AND REFERENCES 23
n
vij vij
(1.5.5) Dvij (t, x) = (t, x) + (t, x) fs (t, x),
t s=1
xs
where fs (t, x) are components of the vector-function f (t, x) = (f1 (t, x), . . . ,
fn (t, x))T .
In Chapter 2, Sections 2.1 2.5, we will establish the sufficient conditions
for asymptotic stability (in the whole), uniform asymptotic stability (in the
whole), exponential stability (in the whole), and instability of solutions of
nonlinear large scale systems under nonclassical structural perturbations
by applying Liapunovs matrix functions (1.5.1) and its derivative (1.5.3)
or (1.5.5)
Section 1.2 The problem of motion stability arises whenever the engineering or
physical problem is formulated as a mathematical problem of qualitative analysis
of equations. Poincare and Liapunov laid a background for the method of aux-
Join the best at Top masters programmes
iliary functions for continuous systems which allow not to integrate the motion
33rd place Financial Times worldwide ranking: MSc
the Maastricht University
equations for their qualitative analysis. The ideas of Poincare andInternational
further developed and applied in many branches of modern natural
Liapunov were
st sciences.
Business
1 place: MSc International Business
School of Business and
The results of Liapunov [1], Chetaev [1], Persidskii [1], Malkin 1st[1],
place: MSc Financial
Ascoli [1], Economics
2nd place: MSc Management of Learning
Barbasin and Krasovskii [1], Massera [1], and Zubov [1], were a2ndbase place:for
MScDefi-
Economics!
nitions 1.2.1 1.2.3 (ad hoc see Grujic et al. [1], pp. 8 12 and cf.
2nd Rao
Economics
place:Mohana
MSc Econometrics and Operations Research
2
nd place: MSc Global Supply Chain Management and
Change
Sources: Keuzegids Master ranking 2013; Elsevier Beste Studies ranking 2012;
Financial Times Global Masters in Management ranking 2012
Maastricht
University is
the best specialist
university in the
Visit us and find out why we are the best! Netherlands
(Elsevier)
Masters Open Day: 22 February 2014
www.mastersopenday.nl
Section 1.2 The problem of motion stability arises whenever the engineering or
physical problem is formulated as a mathematical problem of qualitative analysis
of equations. Poincare and Liapunov laid a background for the method of aux-
iliary functions for continuous systems which allow not to integrate the motion
equations for their qualitative analysis. The ideas of Poincare and Liapunov were
further developed and applied in many branches of modern natural sciences.
The results of Liapunov [1], Chetaev [1], Persidskii [1], Malkin [1], Ascoli [1],
Barbasin
24 and Krasovskii [1],1. Massera [1], and
STABILITY Zubov [1], were a base for Defi-
THEORY
nitions 1.2.1 1.2.3 (ad hoc see Grujic et al. [1], pp. 8 12 and cf. Rao Mohana
Rao [1], Yoshizawa [1], Rouche et al. [1], Antosiewicz [1], Lakshmikantham an
Leela [1], Hahn [2], etc.). For Definitions 1.2.4 1.2.7, and 1.2.13 see Hahn [2],
and Martynyuk [9]. Definitions 1.2.8 1.2.12 are based on some results by Li-
apunov [1], Hahn [2], Barbashin and Krasovskii [1] (see and cf. Djordjevic [1],
Grujic [3], and Martynyuk [2, 3, 5, 10, 13, 17]).
Discrete systems appear to be efficient mathematical models in the investiga-
tion of many real world processes and phenomena (see Samarskii and Gulia [1]).
Note that yet in the works by Euler and Lagrange the so-called recurrent series
and some problems of probability theory were studied being described by discrete
(finite difference) equations. The active investigation of discrete systems (for the
last three decades) is stipulated by new problems of the technical progress. Dis-
crete equations prove to be the most efficient model in description of the mechan-
ical system with impulse perturbations as well as the systems comprising digital
computing devices. Recently the discrete systems have been applied in the mod-
elling of processes in population dynamics, macro-economy, chaotic dynamics of
economic systems, modelling of recurrent neuron networks, chemical reactions,
dynamics of discrete Markov processes, finite and probably automatic machines
and computing processes.
The dynamics of discrete-time systems is in the focus of attention of many
experts (see, for example, Aulbach [1], Diamond [1], Elaydi and Peterson [1],
Luca and Talpalaru [1], Maslovskaya [1], etc.).
Many evolution processes are characterized by the fact that at certain mo-
ments of time they experience a change of state abruptly. This is due to short
term perturbations whose duration is negligible in comparison with the dura-
tion of the process. It is natural, therefore, to assume that such perturbations
act instantaneously, that is, in the form of impulses. Thus impulsive differen-
tial equations, namely, differential equations involving impulse effects, appear as
natural description of observed evolution phenomenon of several real-world prob-
lems. Of course, the theory of impulsive differential equations is much richer
than the corresponding theory of differential equations without impulse effects
(see Blaquiere [1], Krylov and Bogoliubov [1], Milman and Myshkis [1], Myshkis
and Samoilenko [1], etc.).
For Definitions 1.4.1 1.4.3 see Lakshmikantham, Bainov, et al. [1], Samoilenko
and Perestyuk [1], Simeonov and Bainov [1], etc.
Original results and the surveys of some directions of investigations are pre-
sented in the monographs by Lakshmikantham, Leela, and Martynyuk [1, 2], Pan-
dit and Deo [1], and in many papers.
34
Stability Theory of Large-Scale
Dynamical Systems Generalities
NOTES AND REFERENCES 25
The physical system can consist of subsystems that react differently to the
external impacts (see Pontryagin [1], Tikhonov [1], Volosov [1], Hopensteadt [1],
Grujic, et al. [1], etc.). Moreover, each of the subsystems has its own scale of
natural time. In the case when the subsystems are not interconnected, the dy-
namical properties of each subsystem are examined in terms of the corresponding
time scale. It turned out that it is reasonable to use such information when the
additional conditions on the subsystems are formulated in the investigation of
large scale systems. The existence of various time scales related to the separated
subsystems is mathematically expressed by arbitrarily small positive parameters
i present at the part of the higher derivatives in differential equation. If the
parameters i vanish, the number of differential equations of the large scale sys-
tem is diminished and, hence the appearance of algebraic equations. This is just
the singular case allowing the consideration of various peculiarities of the system
with different time scales.
Modern analytical and qualitative methods of analysis of singularly perturbed
systems are based on some ideas and results of the classical works by Tikhonov and
Pontryagin. The development of general ideas in the direction is presented in the
papers and monographs by Vasileva and Butuzov [1], Mishchenko and Rozov [1],
Eckhaus [1], Carrier [1], OMalley [1], Kokotovic and Khalil [1], Miranker [1],
Chang and Howes [1], etc.
Section 1.3 Various problems of the stability theory under classical structural
perturbations were studied in many papers (see, e.g. Aeppli and Markus [1],
Arnold [1], Bowen and Ruelle [1], Conley and Zehnder [1], Coppel [1], Cronin [1],
Hale [1], Hirsch [1], Kneser [1], Kaplan [1], Markus [1], Moser [1], Pilugin [1],
Shub [1], Zeeman [1], etc.).
This Section encorporates some results by Arnold [1], Sell [1], Lefshetz [1],
Peixoto [1], Siljak [1], and Chetaev [1], etc.
Section 1.4 We focused main attention on the concept of stability under non-
classical structural perturbations in the sense of Liapunov. We used in the point
the results from monograph by Grujic, Martynyuk and Ribbens-Pavella [1].
Section 1.5 For the details of the method of matrix-valued Liapunov functions
see Martynyuk [13] and Djordjevic [1]. This method has been developed at
the Stability of Processes Department of the Institute of Mechanics of NAS of
Ukraine since 1979 (see Ph.D. thesises by Shegai [1], Miladzhanov [1], Azimov [1],
Begmuratov [1], Martynyuk-Chernienko [1], Slynko [1], Lykyanova [1]).
For the recent papers concerning the topics of Sections 1.2 1.5 see Kramer
26 1. STABILITY THEORY
and Hofman [1].
We note that the two-index system of functions (1.5.1) being suitable for con-
struction of the Liapunov functions allows to involve more wide classes of func-
tions as compared with those usually applied in motion stability theory. For
example, the bilinear forms prove to be natural non-diagonal elements of matrix-
valued functions. Another peculiar feature of the approach being of importance
is the fact that the application of the matrix-valued function in the investigation
of multidimensional systems enables to allow for the interconnections between
the subsystems in their natural form, i.e. not necessarily as the destabilizing fac-
tor. Finally, for the determination of the property of having a fixed sign of the
total derivative of auxiliary function along solutions of the system under con-
sideration it is not necessary to encorporate the estimation functions with the
quasi-monotonicity property. Naturally, the awkwardness of calculations in this
case is the price.
35
Stability Theory of Large-Scale
Dynamical Systems Continuous Large-Scale Systems
2
CONTINUOUS LARGE-SCALE SYSTEMS
2.1 Introduction
dxi
(2.2.3) = gi (t, xi ), xi (t0 ) = x0i .
dt
i , and the functions
Here xi Rni , the state vector of the subsystem C
ni ni
gi : T R R are determined by the correlations
where xi = (0, . . . , 0, xT T
i , 0, . . . , 0) . 37
The subsystems (2.2.3) do not contain structural and/or parametric per-
turbations and bear the main information on the dynamical properties of
subsystems Ci , while the functions
2.2 A MODEL OF NONCLASSICAL PERTURBATIONS 29
where xi = (0, . . . , 0, xT T
i , 0, . . . , 0) .
in the system
dxi
(2.2.4) = gi (t, xi ) + hi (t, x, pi ), i = 1, 2, . . . , m,
dt
describe the effect of the subsystems C1 , . . . , Ci1 , Ci+1 , . . . , Cm of sys-
tem C on the subsystem Ci .
Designate by Hi the set of all possible hi , from
The fact that fij (t, x, pi ) Fi implies that hji (t, x, pi ) Hi for all
i = 1, 2, . . . , m.
The binary function sij : T {0, 1} is applied as a structural parameter
of system (sij : T [0, 1] ). This function represents the (i, j)-th element
of the structural matrix Si : R Rni Nni of the i-th interconnecting
subsystem Si .
If we designate
S1 012 . . . 01m
0 S2 . . . 02m
S = S : S = 21 , 0ij Rni nj ,
. . . . . . . . . . . . . . . . . . . .
30 2. CONTINUOUS LARGE-SCALE SYSTEMS
0m1 0m2 . . . Sm
where Si = (si1 Ii , si2 Ii , . . . , siN Ii ), sij {0, 1}, Ii = diag {1, 1, . . . , 1}
Rni Rni, the dynamics of the i-th interconnecting subsystem Ci can be
deccribed by the equations
dxi
(2.2.5) = gi (t, xi ) + Si (t)hi (t, x, pi ), i = 1, 2, . . . , m.
dt
dx
(2.2.6) = g(t, x) + S(t)h(t, x, P ), P P, S(t) S,
dt
where
T
x Rn , g(t, x) = g1T (t, x1 ), . . . , gm T
(t, xm ) ,
T
h = hT T
1 (t, x, p1 ), . . . , hm (t, x, pm ) .
m
(2.3.3) v(t, x) = vij (t, x).
Let vii = vii (t, xi ) correspond to subsystems (2.2.3) and vij = vji =
dissertation?
vij (t, xi , xj ) take into consideration connections Si (t)hi (t, x, pi ) between
the equations (2.2.3) for all v = j, i, j = 1, 2, . . . , m.
Assumption 2.3.1 There exist
Get in-depth feedback & advice from experts in your
(1) open connected neighbourhoods Nix Rni of the states (xi = 0)
Rtopic
ni area. Find out what you can do to improve
for all i = 1, 2, . . . , m;
(2) functions of :your
the quality N dissertation!
ik R , for all i = 1, 2, . . . , m; k = 1, 2, K
ix + ik
(ik KR);
(3) constants ij , ij , i, j = 1, 2, . . . , m, and a function (t)
R+ ),Help
C(R,Get (t)
Now c > 0, and
(4) a matrix-valued function U (t, x) with elements vii (t, xi ), vii (t, 0) =
0 for all t R+ , and vij (t, xi , xj ), vij (t, 0, 0) = 0 for all i = j
and for all t R+ satisfying the estimates:
(a) ii 2i1 (xi )(t) vii (t, xi ) ii 2i2 (xi )
for all (t, xi ) R+ Nix (for all (t, xi ) R+ Rni ),
= 1, 2, . . . , m;
Go toi www.helpmyassignment.co.uk for more info
(b) ij i1 (xi )j1 (xj )(t) vij (t, xi , xj )
(x ) (x ) for all (t, xi , xj ) R+ Nix Njx
ij i1 i j2 j
Download free eBooks at bookboon.com
(for all (t, xi , xj ) R+ Rni Rnj ) for all i = j,
39 Click on the ad to read more
i, j = 1, 2, . . . , m.
m
(2.3.3) v(t, x) = vij (t, x).
i,j=1
Let vii = vii (t, xi ) correspond to subsystems (2.2.3) and vij = vji =
vij (t, xi , xj ) take into consideration connections Si (t)hi (t, x, pi ) between
the equations (2.2.3) for all v = j, i, j = 1, 2, . . . , m.
Assumption 2.3.1 There exist
(1) open connected neighbourhoods Nix Rni of the states (xi = 0)
Rni for all i = 1, 2, . . . , m;
(2) functions ik : Nix R+ , for all i = 1, 2, . . . , m; k = 1, 2, ik K
(ik KR);
(3) constants ij , ij , i, j = 1, 2, . . . , m, and a function (t)
C(R, R+ ), (t) c > 0, and
(4) a matrix-valued function U (t, x) with elements vii (t, xi ), vii (t, 0) =
0 for all t R+ , and vij (t, xi , xj ), vij (t, 0, 0) = 0 for all i = j
and for all t R+ satisfying the estimates:
(a) ii 2i1 (xi )(t) vii (t, xi ) ii 2i2 (xi )
for all (t, xi ) R+ Nix (for all (t, xi ) R+ Rni ),
i = 1, 2, . . . , m;
(b) ij i1 (xi )j1 (xj )(t) vij (t, xi , xj )
ij i1 (xi )j2 (xj ) for all (t, xi , xj ) R+ Nix Njx
(for all (t, xi , xj ) R+ Rni Rnj ) for all i = j,
i, j = 1, 2, . . . , m.
(t)T T T T
1 H AH1 v(t, x, ) 2 H BH2
(2.3.7) for all (t, xi , xj ) R+ Nix Njx
(for all (t, xi , xj ) R+ Rni Rnj ) ,
where
T
1 = (11 (x1 ), 21 (x2 ), . . . , m1 (xm )),
T
2 = (12 (x1 ), 22 (x2 ), . . . , m2 (xm )),
H = diag [1 , 2 , . . . , m ],
A = [ij ], ij = ji ,
B = [ij ], ij = ji , i, j = 1, 2, . . . , m.
m
m
s
v(t, x, ) = i2 vii (t, xi ) + 2 i j vij (t, xi , xj )
i=1 i=1 j=2
j>i
m
m
s
i2 ii 2i1 (xi )(t) + 2 i j ij i1 (xi )j1 (xj )(t)
i=1 i=1 j=2
j>i
T
= (t)(11 (x1 ), 21 (x2 ), . . . , m1 (xm ))
11 12 . . . 1s
diag [1 , 2 , . . . , m ] . . . . . . . . . . . . . . . . . . . . . . diag [1 , 2 , . . . , m ]
1m 2m . . . mm
for all (t, xi , xj ) R+ Nix Njx (for all (t, xi , xj ) R+ Rni Rnj ).
The estimate from 2.4 above
TESTSin inequality
FOR (2.3.7)
STABILITY is proved similarly.
ANALYSIS 33
D+ v(t, x, )
= lim+ sup{[v(t + , x + (g(t, x) + S(t)h(t, x, P )), ) v(t, x, )] 1 }
0
(b) vij C(R+ Nix0 Njx0 , R) (vij C(R+ Rni Rnj, R))
for all i = j, i, j = 1, 2, . . . , m;
(3) constants i1i , j2i , i3j , i4i (P, S), jmi (P, S), iqj (P, S), kij ,
lij (P, S), rij (P, S), k = 1, 2; l = 3, 4; r = 6, 7; m = 5, 7; q = 6, 8;
i, j = 1, 2, . . . , m, and conditions
(a) Dt+ vii + (Dx+i vii )Tgi (t, xi ) i1i 2i (xi ) for all
(t, xi ) R+ Nix0 (for all (t, xi ) R+ Rni ),
i = 1, 2, . . . , m;
Brain power
for all (t, xi , xj ) R+ Nix0 Njx0 (for all (t, xByi ,2020,
ni nj
R+ R R ) for all i = j, i, j = 1, 2, . . . , m;
xj ) wind could provide one-tenth of our planets
electricity needs. Already today, SKFs innovative know-
how is crucial to running a large proportion of the
(d) (Dx+i vii )T Si (t)hi (t, x, pi ) i4i (P, S)2i (xi ) worlds wind turbines.
Up to 25 % of the generating costs relate to mainte-
m
nance. These can be reduced dramatically thanks to our
+ 3ij (P, S)i (xi )j (xj ), i = 1, 2, . . . , m, systems for on-line condition monitoring and automatic
j=1 lubrication. We help make it more economical to create
j=i cleaner, cheaper energy out of thin air.
for all (t, xi , xj ) R+ Nix0 Njx0 By sharing our experience, expertise, and creativity,
industries can boost performance beyond expectations.
(for all (t, xi , xj ) R+ Rni Rnj ), Therefore we need the best employees who can
meet this challenge!
for all (P, S) P S;
(e) (Dx+i vij )T Si (t)hi (t, x, pi ) j5i (P, S)2i (xi ) The Power of Knowledge Engineering
m
+ 4ij (P, S)i (xi )j (xj )
i=1
i=j
i = 1, 2, . . . , m;
hold true.
for all (t, xi , xj ) R+ Nix0 Njx0 (for all (t, xi , xj ) R+ Rni Rnj )
and for all (P, S) P S, where
m
qii (P, S) = i2 (i1i + i4i (P, S)) +2 i j {j2i + i3j
hold true.
for all (t, xi , xj ) R+ Nix0 Njx0 (for all (t, xi , xj ) R+ Rni Rnj )
and for all (P, S) P S, where
and
m
qii (P, S) = i2 (i1i + i4i (P, S)) + 2 i j {j2i + i3j
j=1
j=i
36 + j5i (P, S) + j
2. CONTINUOUS i LARGE-SCALEi SYSTEMS
7i (P, S) + 6j (P, S) + 8j (P, S)}, i = 1, 2, . . . , m;
1
qij (P, S) = i2 (3ij (P, S) + 3ji (P, S))
2
m m
+ i j 1ij + 2ij + (4lq (P, S) + 6lq (P, S))
l=1 q=1
l=j q=l
m
m
+ (5lq (P, S) + 7lq (P, S)) , i = j, i, j = 1, 2, . . . , m.
q=1 l=1
q=i l=q
Proof Let all conditions of Assumption 2.4.1 be satisfied. Then for the
expression D+ v(t, x, ) we have
m
D+ v(t, x, ) = T D+ U (t, x) = i2 D+ vii (t, xi )
i=1
(2.4.2) m
m
+2 i j D+ vij (t, xi , xj ).
i=1 j=2
j>i
m
m
1
+2 i2 (3ij (P, S) + 3ji (P, S))
i=1 j=1
2
j>i
m
+ i j 1ij + 2ij + (4lj (P, S) + 6lj (P, S))
l=1
l=j
m
+ (5iq (P, S) + 7iq (P, S)) i (xi )j (xj )
q=1
q=i
m
m
m
= ii (P, S)2i (xi ) + 2 ij (P, S)i (xi )j (xj )
i=1 2.4 TESTS FOR STABILITY
i=1 j=2 ANALYSIS 37
j>i
= wT Q(P, S)w
for all (t, xi , xj ) R+ Nix0 Njx0 (for all (t, xi , xj ) R+ Rni Rnj )
and for all (P, S) P S.
Download free eBooks at bookboon.com
This completes the proof of Proposition 2.4.1.
By means of U (t, x) = diag U (t, x) = [v11 (t, x1 ), .44
. . , vmm (t, xm )], a ma-
trix K and a vector Rm we construct the vector function
(2.4.3) x)
L(t, x, ) = K U(t,
2.4 TESTS FOR STABILITY ANALYSIS 37
Stability
We noteTheory
that of
theLarge-Scale
notions of asymptotically contractive sets with respect
Dynamical
to functionsSystems
were discovered by Grujic [1]. Continuous Large-Scale Systems
The following theorem provides our main characterization of stability
under nonclassical structural perturbations.
Theorem 2.4.1 Assume that the perturbed motion equations (2.2.10)
are such that all conditions of Assumptions 2.3.1 and 2.4.1 are satisfied
except for upper estimates of the functions vii (t, xi ) and vij (t, xi , xj ) for all
i, j = 1, 2, . . . , m and moreover
(1) there exist positive numbers i (or i = +) such that the sets
Lii (t) are asymptotically contractive for any i (0, i ) and
any i = 1, 2, . . . , m;
(2) the matrix A is positive definite;
(3) there exists a negative definite matrix Q Rmm such that
1
38 (Q(P,
2.S) + QT (P, S)) LARGE-SCALE
CONTINUOUS (P, S) P S.
Q for all SYSTEMS
2
Then the equilibrium state x = 0 of the system (2.2.10) is asymptotically
stable on P S
If all hypotheses of the theorem hold for Nix = Rni , for radially un-
bounded functions vii (t, xi ) and vij (t, xi , xj ) and for i = + for any
i = 1, 2, . . . , m, then the equilibrium state x = 0 of the system (2.2.10) is
asymptotically stable in the whole on P S.
1
(Q(P, S) + QT (P, S)) Q for all (P, S) P S.
2
Stability
x = 0 ofTheory of Large-Scale
the system (2.2.10) is uniformly asymptotically stable in the whole
Dynamical
on P S. Systems Continuous Large-Scale Systems
The Proof of Theorem 2.4.2 is similar to the proof of Theorem 2.4.1.
Theorem 2.4.3 Assume that the perturbed motion equations (2.2.10)
are such that all conditions of Assumptions 2.3.1 and 2.4.1 are satisfied
with (t) = 1 for all t R+ and
(1) for given functions i (xi ) there exist positive numbers i and i
such that
i xi i (xi ) i xi
with elements
ii = ii i2 , ii = ii i2 ,
ij i j if ij 0 for all i = j;
ij =
ij i j if ij <0 for all i = j;
ij i j if ij 0 for all i = j;
ij =
ij i j if ij <0 for all i = j,
1
(Q (P, S) + QT (P, S)) Q for all (P, S) P S.
2
1 2
1 2 3
ij (P, S) = i j 1ij kij + 2ij kij + i 3ij (P, S)kij
2
s
4 5
+ l j 4lj (P, S)klj + 6lj (P, S)klj
l=1
l=j
s
6 7
+ i m 5im (P, S)kim + 7im (P, S)kim ,
m=1
m=i
s
6 7
+ i m 5im (P, S)kim
Stability Theory of Large-Scale + 7im (P, S)kim ,
m=1
Dynamical Systems m=i Continuous Large-Scale Systems
D+ v(t, x, ) M (Q1 )x2 for all (t, x, P, S) R Nx P S
T
is valid, where M (Q1 ) < 0, and Q1 = 12 Q + Q . Therefore the
equilibrium state x = 0 of (2.2.10) is exponentially stable on P S.
If Nix = Rni for all i = 1, 2, . . . , m, then Nx = Rn and all hypotheses
of Theorem 2.5.6 by Martynyuk [13] are satisfied for all (P, S) P S.
Hence, the equilibrium state x = 0 of (2.2.10) is exponentially stable in
the whole on P S. This completes the proof of Theorem 2.4.3.
Here the vector w and matrix Q(P, S) are defined in the same way as in
Proposition 2.4.1.
1
(Q(P, S) + QT (P, S)) D
2
42 2. CONTINUOUS LARGE-SCALE SYSTEMS
holds at least for one pair (P, S) P S.
Then the equilibrium state x = 0 of (2.2.10) is unstable on P S.
Proof We construct a scalar function (2.3.2) based on a matrix-valued
function U (t, x). Due to Assumption 2.3.1, Proposition 2.3.3 and conditions
(a) of Theorem 2.4.4 the function v(t, x, ) is positive definite and admits
infinitely small upper bound on Nx . By Assumption 2.4.2, Proposition 2.4.2
and conditions (b) of Theorem 2.4.4 the function D+ v(t, x, ) is positive
definite at least for one pair (P, S) P S.
These conditions are sufficient (see Theorem 2.5.7 by Martynyuk [13])
for instability of the equilibrium state x = 0 of (2.2.10).
Example 2.4.1 Let a fourth order system be given, which consists of two
subsystems of second order
dx1 1 sat 0.1x11
= (1 + t)x1 + s11 (t)
dt 1 + t2 0
Download free eBooks at bookboon.com
0.5x22 0
+ s12 (t) + s13 (t) ;
0 490.1x21
(2.4.4)
dx2 1 0.4x11
= (2 + t)x2 + s 21 (t)
dt 1 + t2 0
0 0
+ s22 (t) + s23 (t) ,
(a) of Theorem 2.4.4 the function v(t, x, ) is positive definite and admits
infinitely small upper bound on Nx . By Assumption 2.4.2, Proposition 2.4.2
and conditions
Stability Theory (b) of Theorem 2.4.4 the function D+ v(t, x, ) is positive
of Large-Scale
definite at Systems
Dynamical least for one pair (P, S) P S. Continuous Large-Scale Systems
These conditions are sufficient (see Theorem 2.5.7 by Martynyuk [13])
for instability of the equilibrium state x = 0 of (2.2.10).
Example 2.4.1 Let a fourth order system be given, which consists of two
subsystems of second order
dx1 1 sat 0.1x11
= (1 + t)x1 + s 11 (t)
dt 1 + t2 0
0.5x22 0
+ s12 (t) + s13 (t) ;
0 0.1x21
(2.4.4)
dx2 1 0.4x11
= (2 + t)x2 + s21 (t)
dt 1 + t2 0
0 0
+ s22 (t) + s23 (t) ,
0.4x12 sat 0.2x22
where x1 = (x11 , x12 )T, x2 = (x21 , x22 )T, sat = for || 1 and sat =
sign for || > 1.
In this example P = {0} and the structural matrices Si (t) have the
form
si1 (t) 0 si2 (t) 0 si3 (t) 0
Si (t) = ,
0 si1 (t) 0 si2 (t) 0 si3 (t)
S1 (t) 0
S(t) = , i = 1, 2.
0 S2 (t)
Note that structural changes of the given system within structural set S
are inadmissible in the frames of connected stability (see Siljak [1]) and
equality sij = smn (t) is admissible for all i, m = 1, 2 and j, n = 1, 2, 3.
All possible interactions are described by means of the matrices
where
It is clear that they are asymptotically contractive for any (0, +).
Further we define the functions vij (t, xi , xj ) as follows
for all (t, x) R+ Nx0 (for all (t, x) R+ Rn ) and for all (P, S)
P S, where
for all i = j, i, j = 1, 2, . . . , m.
for all i = j, i, j = 1, 2, . . . , m.
1
((P, S) + T (P, S)) H for all (P, S) P S
2
is satisfied component-wise.
Then the equilibrium state x = 0 of the system (2.2.10) is asymptotically
stable on P S.
If all hypotheses of Theorem 2.4.5 hold for Nix = Rni for radially
unbounded functions vij (t, ), i, j = 1, 2, . . . , m, then the equilibrium state
x = 0 of the system (2.2.10) is asymptotically stable in the whole on P S.
The Proof of Theorem 2.4.5 is similar to that of Theorem 2.4.1.
Theorem 2.4.6 Assume that the perturbed motion equations (2.2.10)
Challenge
are such that all conditions the way we2.3.1
of Assumptions run and 2.3.2 are satisfied for
(t, x) 1 and
(1) the conditions (a) and (c) of Theorem 2.3.1 are satisfied;
(2) the matrices A and B in the inequalities (2.3.7) are positive definite.
Then the equilibrium state x = 0 of the system (2.2.10) is uniformly asymp-
totically stable on P S.
EXPERIENCE THE POWER OF
If moreover Nix = Rni for all i = 1, 2, . . . , m, the functions vij (t, )
are radially unbounded and functions i KR, then the equilibrium state
FULL ENGAGEMENT
x = 0 of the system (2.2.10) is uniformly asymptotically stable in the whole
on P S.
The Proof of Theorem 2.4.6 is similar to that of Theorem 2.4.2.
RUN FASTER.
Theorem 2.4.7 Assume that the perturbed motion equations (2.2.10)
are such that all conditions of Assumptions 2.4.1 and 2.4.3 and hypotheses
RUN LONGER..
(a) and (b) of Theorem 2.4.5 are satisfied and moreover, there existREAD MORE & PRE-ORDER TODAY
(1) positive numbersRUN
i andEASIER
i such that WWW.GAITEYE.COM
i xi i (xi ) i xi
where
(P, S) = ij (P, S) ,
ij
= ji for all i, j = 1, 2, . . . , m;
s
ii (P, S) = i2 1i (P, S)ki1 + 2 i j (2i (P, S)ki2 + 3j (P, S)kj3 );
j=1
j=i
1 2 1 2
ij (P, S) = i [1ij (P, S) + 1ji (P, S)]kij + i j 2ij kij
2
s
s
3 4
+ 3lj (P, S)klj + 4im (P, S)kim
l=1 m=1
l=j m=i
for all i = j, i, j = 1, 2, . . . , m,
2
i (or j2 ) if the corresponding multiplier
2 (xi ) (or 2 (xj )) is positive,
i j
kijr =
i2 (or j2 ) if the corresponding multiplier
2i (xi ) (or 2j (xj )) is negative,
for all i, j = 1, 2, . . . , m and r = 1, 2, 3;
Download free eBooks at bookboon.com
i j if the corresponding multiplier
i (xi )j (xj ) is 55
q positive,
kij =
i j if the corresponding multiplier
i (xi )j (xj ) is negative,
2
i (or j2 ) if the corresponding multiplier
2i (xi )
(or 2j (xj )) is positive,
kijr =
Stability Theory of
Large-Scale
i2 (or j2 ) if the corresponding multiplier
Dynamical Systems
Continuous Large-Scale Systems
2i (xi ) (or 2j (xj )) is negative,
for all i, j = 1, 2, . . . , m and r = 1, 2, 3;
i j if the corresponding multiplier
i (xi )j (xj ) is positive,
q
kij =
i j if the corresponding multiplier
i (xi )j (xj ) is negative,
for all i, j = 1, 2, . . . , m and q = 1, 2, 3, 4;
S11 0 0 0 1 0
S11 0
0 S11 0 0 0 1
S= =
0 0 S11 0 1 0
0 0 0 S11 0 S21
0 1
S1 0
= , S = {S : 0 S I4 }.
0 S2
hold.
The matrix
1 0.1
A=
0.1 1
is positive definite and the function (t) = 1 + t2 1 > 0.
Let = (1, 1)T , then given choice of elements vii (t, xi ), i = 1, 2,
v12 (t, x1 , x2 ) of matrix-valued function U (t, x), the elements of matrix
G(P, S) are defined as
1
11 (P, S) = 3.39 + |P21 S21 |;
2
1
22 (P, S) = 1.8 + |P11 S11 |;
3
3
12 (P, S) = 0.88 + |P11 S11 | + |P21 S21 |.
2
= TESTS
(t, x)2.4 1 + t2STABILITY
(t) = FOR 1 > 0 ANALYSIS
for all t R+ . 51
58
Stability Theory of Large-Scale
Dynamical Systems Continuous Large-Scale Systems
52 2. CONTINUOUS LARGE-SCALE SYSTEMS
1 T (P, S)) G
(G(P, S) + G for all (P, S) P S
2
1 T (P, S)) G
(G(P, S) + G for all (P, S) P S
2
holds.
Then the equilibrium state x = 0 of (2.2.10) is uniformly asymptotically
stable on P S.
If, moreover Nix = Rni, functions vij are radially unbounded and func-
tions i are of Hahn class KR, then the equilibrium state x = 0 of (2.2.10)
is uniformly asymptotically stable in the whole on P S.
2.4 TESTS FOR STABILITY ANALYSIS 53
The Proof of Theorem 2.4.9 is similar to that of Theorem 2.4.2.
Theorem 2.4.10 Let perturbed motion equations (2.2.10) be such that
all conditions of Assumptions 2.3.1 and 2.4.4 are satisfied for (t) 1 and
hypotheses (a) and (b) of Theorem 2.4.3 hold and, moreover, there exists
a symmetric negative definite matrix G Rmm such that for matrix
1 T (P, S)) G
(G (P, S) + G for all (P, S) P S
2
is valid, where
(P, S) = kii ii (P, S), ij
G
(P, S) = kij ij (P, S), i = j = 1, 2, . . . , m,
2
i , if ii (P, S) > 0,
kii = i = 1, 2, . . . , m,
i2 , if ii (P, S) < 0,
i j , if ij (P, S) > 0,
kij = i, j = 1, 2, . . . , m, j = j.
i j , if ij (P, S) < 0,
Download free eBooks at bookboon.com
Then the equilibrium state x = 0 of system (2.2.10) is exponentially
stable in the whole on P S. 59
If all conditions of Theorem 2.4.10 hold for Nix = Rni , then the equilib-
rium state x = 0 of (2.2.10) is exponentially stable in the whole on P S.
The Proof of Theorem 2.4.10 is similar to that of Theorem 2.4.3.
G (P, S) = kii ii (P, S), ij
(P, S) = kij ij (P, S), i = j = 1, 2, . . . , m,
2
i , if ii (P, S) > 0,
ii =
StabilitykTheory i = 1, 2, . . . , m,
ofi2 ,Large-Scale
if ii (P, S) < 0,
Dynamical Systems
Continuous Large-Scale Systems
i j , if ij (P, S) > 0,
kij = i, j = 1, 2, . . . , m, j = j.
i j , if ij (P, S) < 0,
Then the equilibrium state x = 0 of system (2.2.10) is exponentially
stable in the whole on P S.
If all conditions of Theorem 2.4.10 hold for Nix = Rni , then the equilib-
rium state x = 0 of (2.2.10) is exponentially stable in the whole on P S.
The Proof of Theorem 2.4.10 is similar to that of Theorem 2.4.3.
Example 2.4.3 Consider the forth order system (S) consisting of two
subsystems (Si ) of the second order
dxi 2
= xi + si1 i (i ), i = (4, 2)(2xi xj ),
(2.4.9) dt 1
i, j = 1, 2, i = j
when the conditions
i (i )i1 [0, +), for all i R, (0, +), = const
hold.
We suppose that
s11 0 0 0
S1 0 0 s22 0 0
S = S: S = = ,
0 S2 0 0 s22 0
0 0 0 s22
si1 {0, 1}, i = 1, 2.
www.sylvania.com
We do not reinvent
the wheel we reinvent
light.
Fascinating lighting offers an infinite spectrum of
possibilities: Innovative technologies and new
markets provide both opportunities and challenges.
An environment in which your expertise is in high
demand. Enjoy the supportive working atmosphere
within our global group and benefit from international
career paths. Implement sustainable ideas in close
cooperation with other specialists and contribute to
influencing our future. Come and join us in reinventing
light every day.
Light is OSRAM
2
Dv(t, x) = Dvii + 2Dv12 = 4x21 4x22 s11 1 (1 )1
i=1
is negative definite.
Thus, all hypotheses of Theorem 2.4.9 are satisfied and the state x = 0
of (2.4.9) is uniformly asymptotically stable in the whole on S.
for all (t, x) R+ Nx0 (for all (t, x) R+ Rn ) and for all (P, S)
P S.
Here the function (t, x), vector w and matrix (P, S) are defined as in
Proposition 2.4.3. 2.4 TESTS FOR STABILITY ANALYSIS 55
m
dxi
(2.5.1) = Ai xi + sij Aij xj , i = 1, 2, . . . , m,
dt j=1
j=i
where Bii are symmetric positive definite matrices and Bij are constant
matrices for all i = j . It can be easily verified that for the functions
S = {S : S = diag [S1 , S2 , . . . , Ss ]},
Si = [si1 , . . . , si,i1 , I, si,i+1 , . . . , sis ],
Stability Theory of Large-Scale 0 sij I,
Dynamical Systems Continuous Large-Scale Systems
where I is a unique matrix of the corresponding dimension, n1 + n2 + +
nm = n, x = (xT T T T
1 , x2 , . . . , xm ) R .
n
vii (xi ) = xT
i Bii xi for all i = 1, 2, . . . , m,
(2.5.3)
vij (xi , xj ) = xT
i Bij xj for all i = j, i, j = 1, 2, . . . , m,
where Bii are symmetric positive definite matrices and Bij are constant
matrices for all i = j . It can be easily verified that for the functions
(2.5.3) the estimates (cf. Krasovskii [1], and Djordjevic [1])
360
2.5 LINEAR SYSTEMS ANALYSIS 57
for all (xi , xj ) Nix0 Njx0 and i = j, i, j = 1, 2, . . . , m
.
hold true, where m (Bii ) and M (Bii ) are minimal and maximal eigen-
1/2 T
thinking
values of the matrices Bii for all i = 1, 2, . . . , m and M (Bij Bij ) are
norms of matrices Bij for i = j, i, j = 1, 2, . . . , m.
We introduce the function
where
T = (x1 , x2 , . . . , xs ),
360 H = diag [1 , 2 , . . . , s ],
.
A = [aij ], B = [bij ], i, j = 1, 2, . . . , m,
(2.5.6)
aij = aji ,
D+ v(x, ) = T D+ U (x),
bij = bji for all i = j.
thinking
m
R+ , > 0,
.
along solutions of the system (2.5.1) is considered.
Proposition 2.5.1 If for the system (2.5.1) there exists the matrix-
valued function (2.5.2) with elements (2.5.3), then the total derivative of Discover the truth at www.deloitte.ca/careers Dis
(2.5.3) by virtue of system (2.5.1) satisfies the estimates
m
dxi
(a) i2 (Dx+i vii (xi ))T 1i xi 2 + 2 1ij (S)xi xj
dt j=1
j=i
Deloitte & Touche LLP and affiliated entities.
for all
Discover (xi , xjat
the truth Rni Rnj , i = 1, 2, . . . , m, S S; d
) www.deloitte.ca/careers Deloitte & Touche LLP and affiliated entities.
+ T dxi + T dxj
(b) 2i j (Dxi vij (xi , xj )) + (Dxj vij (xi , xj ))
dt dt
Download free eBooks at bookboon.com
Deloitte & Touche LLP and affiliated entities. 2i (S)xi 2 + (2ij + 3ij (S))xi xj
Discover the truth
for all (xi , xj ) Rni Rnj , for i = 63
at www.deloitte.ca/careers
j, i, j = 1, 2, . . . , m, Click on the ad to read more
where
T = (x1 , x2 , . . . , xs ),
H = diag [1 , 2 , . . . , s ],
A = [aij ], B = [bij ], i, j = 1, 2, . . . , m,
aii = m (Bii ), bii = M (Bii ), i = 1, 2, . . . , m,
1/2
T
aij = bij = M (Bij Bij ), i = 1, 2, . . . , m 1, j = 1, 2, . . . , m,
aij = aji , bij = bji for all i = j.
2ij = i j (aT
i Bij + Bij Aj ),
i1
3ij (S) =
T
k (i Bki Skj Akj + (Skj Akj )T j Bkj )
k=1
m
T
+ k (i Bik Skj Akj + (Ski Aki )T j Bkj )
k=i+1
Download free eBooks at bookboon.com
m
+ T
k (i Bik Skj Akj 64
+ (Ski Aki ) j Bjk )
T
k=j+1
for all i = j, i, j = 1, 2, . . . , m.
1 2
1ij (S) = T
2 i Bii Sij Aij + (Sij Aij ) ,
T
2ij =
Stability Theory of i j (ai Bij + Bij Aj ),
Large-Scale
Dynamical Systems i1 Continuous Large-Scale Systems
3ij (S) =
T
k (i Bki Skj Akj + (Skj Akj )T j Bkj )
k=1
m
T
+ k (i Bik Skj Akj + (Ski Aki )T j Bkj )
k=i+1
m
+ T
k (i Bik Skj Akj + (Ski Aki ) j Bjk )
T
k=j+1
for all i = j, i, j = 1, 2, . . . , m.
Proof Let for the system (2.5.1) the matrix-valued function (2.5.2) be
constructed with elements (2.5.3). Then we have in case (a)
m T
T dxi
i2 Dx+i vii (xi ) = i2 Ai xi + Sij Aij xj Bii xi
dt j=1
j=i
m
+ i2 xT
i Bii Ai xi + Sij Aij xj = xT 2 T
i i [Bii Ai + Ai Bii ]xi
2.5 LINEAR
j=1 SYSTEMS ANALYSIS 59
j=i
m
1 2 T
+2 xT
i [Bii Sij Aij + (Sij Aij ) Bii ]xj
j=1
2 i
j=i
m
1i xi 2 + 2 1ij (S)xi xj
j=1
j=i
where
1 T
( at(S)
Download free eBooks + (S))
bookboon.com for all S S.
2
65
Then the equilibrium state x = 0 of the system (2.5.1) is asymptotically
stable in the whole on S.
The Proof of Proposition 2.5.2 is similar to that of Proposition 2.3.1.
Stability Theory of Large-Scale
Theorem
Dynamical 2.5.1 Assume that the system (2.5.1) is such that
Systems Continuous Large-Scale Systems
(1) there exists a matrix-valued function (2.5.2) with elements (2.5.3);
(2) the matrix A in (2.5.5) is positive definite;
(3) there exists a negative definite matrix Rmm such that
1 T
( (S) + (S)) for all S S.
2
m
m
m
1/2
D+ v(x, ) M (Cii )xi 2 + 2 T
M (Cij Cij )xi xj
i=1 i=1 j=2
where
i1
Cii = i2 (Bii Ai + AT
i Bii ) + i j AT T
ji Bji + Bji Aji
j=1
m
+ i j Bij Aij + AT T
ij Aij , i = 1, 2, . . . , m,
j=i+1
1 2 T
Cij = i Bii Aij + AT
ij Bii + i j Ai Bij + Bij Aj
2
i1
T
+ k i Bik Akj + AT
ki j Bkj
k=1
j1
T
+ k i Bik Akj + AT
ki j Bkj
k=i
m
T
+ k i Bik Akj + AT
ki j Bkj ,
We will turn
k=j your CV into
an opportunity of1, aj lifetime
i = 1, 2, . . . , m = 2, 3, . . . , m, i < j,
Cij = Cji for all i = j.
= [ ij ], ij = ji , i, j = 1, 2, . . . , m,
T1/2
where ii = M (Cii ) and ij = M (Cij Cij ) are maximal eigenvalues of
1/2
matrices Cii and M () are norms of matrices Cij , i, j = 1, 2, . . . , m.
Do you like cars? Would you like to be a part of a successful brand? Send us your CV on
We will appreciate and reward both your enthusiasm and talent. www.employerforlife.com
Send us your CV. You will be surprised where it can take you.
m
+ i j Bij Aij + AT T
ij Aij , i = 1, 2, . . . , m,
Stability Theory of Large-Scale
j=i+1
Dynamical Systems Continuous Large-Scale Systems
1 T
Cij = i2 Bii Aij + AT ij Bii + i j Ai Bij + Bij Aj
2
i1
T
+ k i Bik Akj + AT
ki j Bkj
k=1
j1
T
+ k i Bik Akj + AT
ki j Bkj
k=i
m
T
+ k i Bik Akj + AT
ki j Bkj ,
k=j
i = 1, 2, . . . , m 1, j = 2, 3, . . . , m, i < j,
Cij = Cji for all i = j.
= [ ij ], ij = ji , i, j = 1, 2, . . . , m,
T1/2
where ii = M (Cii ) and ij = M (Cij Cij ) are maximal eigenvalues of
matrices Cii and M 2.5
() LINEAR SYSTEMS ANALYSIS 61
1/2
are norms of matrices Cij , i, j = 1, 2, . . . , m.
Example 2.5.1 We consider a linear fourth order system consisting of
two second order subsystems
dx1 1 0.5 s21 0 0.5 1
= x1 + x2 ,
dt 0.5 2 0 s21 1 0.5
(2.5.8)
dx2 2 1 s11 0 0.1 1
= x2 + x1 ,
dt 0.5 3 0 s11 1 0.1
vii = xT
i diag {2, 2}xi , i = 1, 2,
v12 (x1 , x2 ) = v21 (x1 , x2 ) = xT
1 diag {0.1, 0.1}x2.
hold.
If T = (1, 1), then the matrix A has the form
2 0.1
A=
0.1 2
68
Stability Theory of Large-Scale
Dynamical Systems Continuous Large-Scale Systems
2.6 GENERALIZATIONS AND APPLICATIONS 63
AXA Global
Graduate Program
Find out more and apply
Stability
where UTheory of Large-Scale
+ (t, x) is positive semi-definite on T matrix-valued function
Dynamical
and a K. Systems Continuous Large-Scale Systems
For the measures (t, x) and 0 (t, x) taking values from the sets M
and M0 respectively the notions below are the generalizations of property
of Movchans metrices (cf. Movchan [1], and Lakshmikantham, Leela, and
Martynyuk [1]).
70
Stability Theory of Large-Scale
Dynamical Systems Continuous Large-Scale Systems
2.6 GENERALIZATIONS AND APPLICATIONS 65
We return now to the system (2.2.7) and assume that the operator (2.2.9)
is contractive for all (P, S) P S. The solution x(t; t0 , x0 ) of system
(2.2.7) is designated by x(t) and its dependence on (P, S) P S is taken
into account.
Definition 2.6.5 System (2.2.7) is
(1) (0 , )-stable on P S, if for every t0 Ti and > 0 there exists
a positive function (t0 , ), continuous in t0 for each so that for
0 (t0 , x0 ) < the inequality (t, x(t)) < holds for all t T0 and
all (P, S) P S;
(2) (0 , )-attractive on P S, if for any t0 Ti and any > 0
there exist (t0 ) > 0 and = (t0 , x0 , ) [0, +) such that
for 0 (t0 , x0 ) < (t0 ) the inequality (t, x(t)) < holds for all
t (t0 + , ) when all (P, S) P S;
(3) asymptotically (0 , )-stable on PS, if it is (0 , )-stable on PS
and (0 , )-attractive on P S.
1
where e = (1, 1, . . . , 1)T Rm and (P, S) = ((P, S)+ T (P, S));
2
(5) there exists a constant m m matrix such that (P, S) for
all (P, S) P S.
Then
(1) system (2.2.7) is (0 , )-stable on P S, if the matrix is negative
semi-definite, the measure is continuous with respect to measure
0 and condition (3a) is satisfied;
(2) system (2.2.7) is uniformly (0 , )-stable on P S, if the matrix
is negative semi-definite, the measure is uniformly continuous
with respect to measure 0 and condition (3b) is satisfied.
The facf that the measure is continuous with respect to the measure 0
yields the existence of a function CK and a constant 2 = 2 (t0 ) > 0
such that Month 16
(2.6.8) (t , x ) (t , (t , x )) for (t , x ) < , I was a construction
0 0 0 0 0 0 0 0 0 2 Mo
where is taken so that
2 supervisor ina const
I was
(2.6.9) (t , ) < . 0 2 1 the North Sea super
Let (0, ) and t R . Since the functions a K and b CK,
0 0 advising and the No
given and t , one can take = (t , ) > 0 so that
0 3 3 0
Real work he
helping foremen advis
International
al opportunities
Internationa
(2.6.10) b(t , ) < a(). 0
ree wo
work 3
or placements ssolve problems
Real work he
helping fo
International
Internationaal opportunities
We take (t ) = min ( , , ). Conditions (2.6.6) (2.6.10) imply
that for (t , x ) <
0 0
0
0
1 2 3
ree wo
work
or placements ssolve pr
0 a((t , x )) v(t , x , w) b(t0 , 0 (t0 , x0 )) < a(),
0 0 0
Download free eBooks at bookboon.com
from which it follows that 72 Click on the ad to read more
(t0 , x0 ) < .
Let x(t; t0 , x0 ) = x(t) be a solution of system (2.2.7) with the initial
conditions for which 0 (t0 , x0 ) < . Let us verify that under conditions of
Stability Theory 2.6
of Large-Scale
GENERALIZATIONS AND APPLICATIONS 67
Dynamical Systems Continuous Large-Scale Systems
of having a fixed sign of function (2.6.2) with respect to measure does not
requre the -signdefiniteness of elements uij (t, x) of matrix (2.6.1).
First we shall prove the assertion (1) of Theorem 2.6.1. Conditions (1),
(2) and (4a) imply that the function v(t, x, w) is weakly 0 -decreasing.
Therefore there exists a constant 0 = 0 (t0 ) > 0 for t0 R such that
for 0 (t0 , x0 ) < 0
The facf that the measure is continuous with respect to the measure 0
yields the existence of a function CK and a constant 2 = 2 (t0 ) > 0
such that
for the solution x(t) with the initial conditions 0 (t0 , x0 ) < . Condition
(3) and the fact that the matrix (P, S) is negative semi-definite in the
domain S(, H) imply that the roots i = i ((P, S)) of the equation
(2.6.12) E ] = 0
det [
satisfy the condition i ((P, S)) 0, i = 1, 2, . . . , m, in the domain
S(, H). Therefore
D+ v(t, x, w)|(2.2.7) eT (P, S)e 0 for all (P, S) P S
1
B(P, S) = (B(P, S) + B T (P, S)) B for all (P, S) P S.
2
m
a((t, x)) v(t, x, w) for all (t, x, w) S(, H0 ) R+
and
m
v(t, x, w) b(0 (t, x)) for 0 (t, x) < 0 , w R+ .
and
Stability Theory of Large-Scale
Dynamical Systems Continuous Large-Scale Systems
m
v(t, x, w) b(0 (t, x)) for 0 (t, x) < 0 , w R+ .
Since the matrix B is symmetric and negative definite, M (B) < 0. The
measure 0 is of class M, so there exists a function K such that
m
(2.6.14) (0 (t, x)) 0s (t, x).
s=1
Therefore
b( )
T () = + 1,
M (B)()
where M (B) is the maximal eigenvalue of the symmetric matrix B(w) and
the function is of class K. We shall prove uniform asymptotic (0 , )
stability on P S of system (2.2.7), if we make sure that a t [t0 , t0 +T ]
exists such that
93%
(2.6.16) OF
MIM
0 (t , x(t )) STUDENTS
< . ARE
WORKING IN THEIR SECTOR 3 MONTHS
If this is not true, then thereFOLLOWING
exists a solution x(t) = x(t; t0 , x0 ) of system
GRADUATION
(2.2.7) with local values 0 (t0 , x0 ) < for which
t
0 +T
(2.6.19) 5 (B)
M Specializations
(0 (s, x(s))) ds m (B)()T
#10 WORLDWIDE
> b( ) 55 Nationalities
MASTER IN MANAGEMENT
Personalize your program FINANCIAL TIMES
in class
t0
Therefore
Stability Theory of Large-Scale
Dynamical
(2.6.15) Systems D+ v(t, x, w) M (B)(0 (t, x)) Continuous Large-Scale Systems
m
for all (t, x, w) S(, H) R+ and for all (P, S) P S.
Further for arbitrary 0 < < H0 we take = () being the same as in
the definition of uniform (0 , )stability. Assume that 0 (t0 , x0 ) < =
min {0 , 1 } and take
b( )
T () = + 1,
M (B)()
where M (B) is the maximal eigenvalue of the symmetric matrix B(w) and
the function is of class K. We shall prove uniform asymptotic (0 , )
stability on P S of system (2.2.7), if we make sure that a t [t0 , t0 +T ]
exists such that
If this is not true, then there exists a solution x(t) = x(t; t0 , x0 ) of system
(2.2.7) with local values 0 (t0 , x0 ) < for which
According to Shaaban [1], Shaaban and Grujic [1], and Grujic, et al. [1],
pp. 341 345, an N -machine power system is decomposed into subsystems,
each consisting of two machines in addition to the comparison machine. The
system is decomposed into (N 1)/2 interconnected subsystems for odd
number N . When considering transfer conductances, mechanical damp-
ing, electromagnetic damping and speed governor action, the mathemati-
cal model of the system is derived, and it is decomposed into (N 1)/2
sixth-order and one second-order interconnected subsystems. If N is even,
then the system is decomposed into (N 2)/2 sixth-order, one third-order
and one second-order sybsystems.
In this section N is odd, without loss of generality. The system
N N
di
= i i ij (i j ) + ki pi Aij ij ,
dt j=1 j=1
(2.6.20)
dpi
= i pi i i , i = 1, 2, . . . , n, N = n + 1,
dt
According to Shaaban [1], Shaaban and Grujic [1], and Grujic, et al. [1],
pp. 341 345, an N -machine power system is decomposed into subsystems,
each consisting of two machines in addition to the comparison machine. The
system is decomposed into (N 1)/2 interconnected subsystems for odd
number N . When considering transfer conductances, mechanical damp-
ing, electromagnetic damping and speed governor action, the mathemati-
cal model of the system is derived, and it is decomposed into (N 1)/2
sixth-order and one second-order interconnected subsystems. If N is even,
then the system is decomposed into (N 2)/2 sixth-order, one third-order
and one second-order sybsystems.
In this section N is odd, without loss of generality. The system
N N
di
= i i ij (i j ) + ki pi Aij ij ,
dt j=1 j=1
(2.6.20)
dpi
= i pi i i , i = 1, 2, . . . , n, N = n + 1,
dt
(2.6.22) x = [1N , 1 , p1 , 2N , 2 , p2 , . . . , nN , n , pn , N , pN ]T
dxN
(2.6.23) = PN xN + hN (x),
dt
where
1
N MN
PN = ,
N N
and N 1
1
{ MN AN j N j (N j )}
hN = j=1 N j jN .
0
Each of the sixth-order subsystems may be written in the general form
dxI
(2.6.24) = PI xI + BI fI (I ) + hI (x) for I = 1, 2, . . . , m,
dt
dxI
(2.6.25) = PI xI + BI fI (I ),
dt
where
2.6 GENERALIZATIONS
and the interconnections hI (x). AND APPLICATIONS 75
0 0 1 0 0 0
0 0 0 1 0 0
0 0 I I Mi1 0
(2.6.27) Pi = I
.
0 0 I I 0 Mi1
I +1
0 0 iI 0 iI 0
0 0 0 iI +1 0 iI +1
Assuming that the free subsystem (2.6.25) contains the six nonlinarities
given by
(2.6.28)
I1 (I1 ) = cos(iI N + iI N iI N ) cos(iI N iI N ),
I2 (I2 ) = cos(iI +1,N + iI +1,N iI +1,N ) cos(iI +1,N iI +1,N ),
I3 (I3 ) = cos(iI ,iI +1 + iI ,iI +1 iI ,iI +1 ) cos(iI ,iI +1 iI ,iI +1 ),
I4 (I4 ) = cos(iI +1,iI + iI +1,iI iI +1,iI ) cos(iI +1,iI iI +1,iI ),
I5 (I5 ) = cos(N iI + N iI iI N ) cos(N iI iI N ),
I6 (I6 ) = cos(N,iI +1 + N,iI +1
iI +1,N ) cos(N,iI +1
iI +1,N ),
80
Stability Theory of Large-Scale
Dynamical Systems Continuous Large-Scale Systems
76 2. CONTINUOUS LARGE-SCALE SYSTEMS
1 0 0 0 0 0
0 1 0 0 0 0
1 1 0 0 0 0
(2.6.31) CIT = ,
1 1 0 0 0 0
1 0 0 0 0 0
0 1 0 0 0 0
(2.6.32)
0
0
N1
I xN 1 + I xN 2 + Ij jN
jJI
AiI j iI j (iI j ) AN j N j (N j )
+
MiI MN
hI (x) = .
N
1
I xN 1 + I xN 2 + Ij jN
jJI
Ai +1,j i +1,j (i +1,j ) AN j N j (N j )
I I I
+
MiI +1 MN
I xN 1 I xN 2
I xN 1 I xN 2
Aij = Ei Ej Yij ,
AI = EiI EN YiI N , AI = EiI +1 EN YiI +1,N , A I = EiI EiI +1 YiI ,iI +1 ,
Excellent Economics and Business programmes at:
iI N = iI N iI N , iI +1,N = iI +1,N iI +1,N ,
iI ,iI +1 = iI N iI +1,N = iI ,iI +1 iI ,iI +1 ,
iI N = iI N , iI +1,N = iI +1 N ,
I = N iI , I = N iI +1 ,
I = iI ,iI +1 N,iI +1 , I = iI +1,iI N iI ,
Ij = iI j N j , Ij = iI +1,j N j ,
of a successful,
international career.
CLICK HERE
to discover why both socially
and academically the University
of Groningen is one of the best
places for a student to be
www.rug.nl/feb/education
Aij = Ei Ej Yij ,
AI = EiI EN YiI N , AI = EiI +1 EN YiI +1,N , I = EiI EiI +1 YiI ,iI +1 ,
A
iI N = iI N iI N , iI +1,N = iI +1,N iI +1,N ,
iI ,iI +1 = iI N iI +1,N = iI ,iI +1 iI ,iI +1 ,
iI N = iI N , iI +1,N = iI +1 N ,
I = N iI , I = N iI +1 ,
I = iI ,iI +1 N,iI +1 , I = iI +1,iI N iI ,
Ij = iI j N j , Ij = iI +1,j N j ,
N N
I = iI + 2.6 + iI j , I = AND
N iI GENERALIZATIONS + N,iI +! +
iI +1APPLICATIONS iI +1,j , 77
j=iI j=iI +1
I = iI N , I = iI +1 N ,
I = iI N , I = iI +1 N ,
I = Mi1
I
1
MN , I = Mi1
I +1
1
MN .
It is obvious that the state vector of the whole system is given now by
(2.6.33) x = [xT T T T T
1 , x2 , . . . , xm , xN ] .
vII (xI ) = xT
I QII xI for all I = 1, 2, . . . , m,
(2.6.35) vIJ (xI , xJ ) = xT
I QIJ xJ for all I, J = 1, 2, . . . , m, I = J,
vIJ (xI , xJ ) = vJI (xI , xJ ) for all I = 1, 2, . . . , m, I = J,
I I I
where QII = [q ], q = q , , = 1, . . . , 5, are symmetric positive
IJ
definite matrices of the order of 5 5; QIJ = [q ], , = 1, . . . , 5 are
constant matrices of the order of 5 5.
It is easy to verify that for the functions (2.6.41) the bilaterial inequalities
(2.6.36)
m (QII )xI 2 vII (xI ) M (QII )xI 2
for all xI NIx , I = 1, 2, . . . , m;
1/2 1/2
M (QT T
IJ QIJ )xI xJ vIJ (xI , xJ ) M (QIJ QIJ )xI xJ
for all (xI , xJ ) NIx NJx , I, J = 1, 2, . . . , m, I = J
are satisfied, where m (QII ) are minimal and M (QII ) are maximal eigen-
1/2
values of the matrices QII ; M (QT IJ QIJ ) are the norms of the matrices
QIJ respectively, NIx is the neighborhood of the state xI = 0.
Download free eBooks at bookboon.com
82
Stability Theory of Large-Scale
Dynamical Systems Continuous Large-Scale Systems
78 2. CONTINUOUS LARGE-SCALE SYSTEMS
the estimates
Q = [q IJ ], Q = [q IJ ], I, J = 1, 2, . . . , m,
q II = m (QII ), q II = M (QII ), I, J = 1, 2, . . . , m,
1/2
q IJ = qIJ = M (QT
IJ , QIJ ), I, J = 1, 2, . . . , m, I = J.
kI rkl
akI
11 akI
12 0 0 0
akI akI 0 0 0
21 22
kI
Stability Theory of Large-Scale
AkI (S) =
a31 akI
32 0 0 0
,
Dynamical Systems kI
a41 akI 0 0
0 Continuous Large-Scale Systems
42
akI
51 akI
52 0 0 0
arkl
11 arkl
12 0 0 0
arkl arkl 0 0 0
21 22
rkl
Akl (S) =
r
a31 arkl
32 0 0 0
,
rkl
a41 arkl
42 0 0 0
arkl
51 arkl
52 0 0 0
r, l = I, J, k = 1, 2, 3.
American online
LIGS University
is currently enrolling in the
Interactive Online BBA, MBA, MSc,
DBA and PhD programs:
Theorem 2.6.3 For the large scale power system (2.6.24) let matrix-
valued function U (x) with elements (2.6.35) be constructed. If the matrix
H T QH is positive definite and the matrix C is negative definite, then the
equilibrium state x = 0 of system (2.6.24) is asymptotically stable under
nonclassical structural perturbations.
The Proof of this theorem follows immediately from the proof of Theo-
rem 2.4.1.
Remark 2.6.2 System (2.6.24) was studied earlier be means of the vector
Liapunov function. The comparison of results of the above monograph with
the approach based on the matrix-valued function shows that the aggrega-
tion matrices in both cases are of the same order. The difference between
the aggregation matrices is that in the determination of the diagonal ele-
ments of the aggregation matrix C the matrices
are required. The matrices (2.6.41) and (2.6.42) appear because of the
presence of the functions
vIJ (xI , xJ ) = xT
I QIJ xJ ,
s
(1) (2) s
dxi
= Sil Ail xl + Sil qil fil il ,
(2.6.43) dt
l=1 l=1
il = CilT x, i = 1, 2, . . . , s,
Download
where il1 ffree eBooks at bookboon.com ni
il (il ) [0, kil ] R+ , Ail are constant matrices, xi R ,
n1 + n2 + + ns = n, kil are constants. Here all matrices
85 and vectors are
(1) (2)
of the corresponding dimensions, and Sil and Sil are diagonal matrices.
By means of the structural matrices
where il1 fil (il ) [0, kil ] R+ , Ail are constant matrices, xi Rni ,
n1 + n2 + + ns = n, kil are constants. Here all matrices and vectors are
(1) (2)
of the corresponding dimensions, and Sil and Sil are diagonal matrices.
By means of the structural matrices
(k) (1)
S = {S : 0 Sil I, Sii = I, i, l = 1, 2, . . . , s, k = 1, 2},
dxi (2)
(2.6.44) = Aii xi + Sii qii fii (sii ),
dt
where
82 ii = CiiT x2.i , CONTINUOUS
xi = (0T , 0T , .LARGE-SCALE
. . , 0T , xT T
. . 0T )T Rn , i [1, s].
i , 0 , .SYSTEMS
(2)
fi (xi ) = Aii xi + Sii qii fii (ii );
s
s
(1) (2) (2)
fi (x, S) = Sil Ail xl + Sil qil fil (l ) + Sii qii [fii (ii ) + fii ()];
l=1 l=1
l=i l=i
dxi
(2.6.45) = fi (xi ) + fi (x, S), i = 1, 2, . . . , s.
dt
(1)
(a) i2 (Dx+i vii )T fi (xi ) i (S)xi 2 for all xi Nix0 , i [1, s];
Download
s free eBooks at bookboon.com
s s
(b) i2 (Dx+i vii )T fi (x, S) + 2 i j {(Dx+i )T (fi (xi ) + fi (x, S))
i=1 i=1 j=2
86
j>i
s
(2)
+ (Dx+j vij )T (fj (xj ) + fj (x, S)) i (S)xi 2
i=1
s
s
along solutions of system (2.6.43) is considered.
(1)
(a) i2 (Dx+i vii )T fi (xi ) i (S)xi 2 for all xi Nix0 , i [1, s];
s
s
s
(b) i2 (Dx+i vii )T fi (x, S) + 2 i j {(Dx+i )T (fi (xi ) + fi (x, S))
i=1 i=1 j=2
j>i
s
(2)
+ (Dx+j vij )T (fj (xj ) + fj (x, S)) i (S)xi 2
i=1
s
s
+2 ij (S)xi xj for all (xi , xj , S) Nix0 Njx0 S,
i=1 j=2
2.6 GENERALIZATIONS AND APPLICATIONS 83
j>i
(k)
where i (S), k = 1, 2, i [1, s] are maximal eigenvalues of the matrices
(2) (2)
i2 [Pii Aii + AT i T i T T
ii Pii + Pii Sii qii kii (cii ) + (Sii qii kii (cii ) ) Pii ];
i1
(1) (2) (1)
i l |{[(Sli Ali )T + (Sli qli kli (cili )T )T ]Pli + PliT [Sli Ali
l=1
s
(2) (1) (2) (1)
+ Sli qli kli (cili )T ]} i l {Pil [Sli Ali + Sli qli kli (cili )T ] + [(Sli Ali )T
l=i+1
(2) (2) (2)
+ (Sli qli kli (cili )T )T ]PilT } i T
+ i2 [Pii Sii qii kii i T T
(cii ) + (Sii qii kii (cii ) ) Pii ]
j1
s
(1) (2) (1)
j l [(Sli AT i T T
li ) + (Sli qli kli (cli ) ) ]Plj + j l [(Sli Ali )T
l=1 l=j+1
i1
(2) (1) (2)
j T
+ (Sli qli kli (cili )T )T ]PjlT + i l PliT [Slj Alj + Slj qlj klj (clj ) ]
l=1
1 2
s
(1) (2) (1)
+
i l Pil [Slj Alj + Slj qlj klj (cjlj )T ] + Pii (Sij Aij )
2 i
l=i+1
(1) (2) (2)
+ (Sij Aij )T Pii + Pii (Sij qij kij
(cjij )T ) + (Sij qij kij
(cjij )T )T Pii
(2) j T (2) j T T
+ Pii (Sii qii kii (cii ) ) + (Sii qqq kii (cii ) ) Pii
1 (1) (1) (2)
+ j2 Pji (Sji Aji ) + (Sji Aji )T Pji + Pjj (Sji qji kji
(ciji )T )
2
(2) (2) (2)
+ (Sji qji kji (ciji )T )TPjj + Pjj (Sjj qjj kjj
(cijj )T ) + (Sjj qjj kjj
(cijj )T )TPjj
respectively. Here
(k)
kij for ij (Sij qij )T Pij xj > 0, i, j = 1, 2, . . . , s; k = 1, 2;
kij =
0 in other cases;
(2)
kii for ii (Sii qii )T Pii xi > 0, i = 1, 2, . . . , s;
kii = (2)
kii for ii (Sii qii )T Pii xi < 0, i = 1, 2, . . . , s;
j1
s
(1) (2) (1)
j l [(Sli AT
li ) + (Sli qli kli (cili )T )T ]Plj + j l [(Sli Ali )T
l=1 l=j+1
i1
(2) (1) (2)
j T
+ (Sli qli kli (cili )T )T ]PjlT + i l PliT [Slj Alj + Slj qlj klj (clj ) ]
l=1
1 2
s
(1) (2) (1)
+
i l Pil [Slj Alj + Slj qlj klj (cjlj )T ] + Pii (Sij Aij )
2 i
l=i+1
(1) (2) (2)
+ (Sij Aij )T Pii + Pii (Sij qij kij
(cjij )T ) + (Sij qij kij
(cjij )T )T Pii
(2) j T (2) j T T
+ Pii (Sii qii kii (cii ) ) + (Sii qqq kii (cii ) ) Pii
1 (1) (1) (2)
+ j2 Pji (Sji Aji ) + (Sji Aji )T Pji + Pjj (Sji qji kji
(ciji )T )
2
(2) (2) (2)
+ (Sji qji kji (ciji )T )TPjj + Pjj (Sjj qjj kjj
(cijj )T ) + (Sjj qjj kjj
(cijj )T )TPjj
respectively. Here
(k)
kij for ij (Sij qij )T Pij xj > 0, i, j = 1, 2, . . . , s; k = 1, 2;
kij =
0 in other cases;
(2)
kii for ii (Sii qii )T Pii xi > 0, i = 1, 2, . . . , s;
kii = (2)
kii for ii (Sii qii )T Pii xi < 0, i = 1, 2, . . . , s;
ckij
84 Rnk is the k-th the vector cij SYSTEMS
component ofLARGE-SCALE
2. CONTINUOUS .
89
Stability Theory of Large-Scale
Dynamical Systems Continuous Large-Scale Systems
86 2. CONTINUOUS LARGE-SCALE SYSTEMS
86 2. CONTINUOUS LARGE-SCALE SYSTEMS
For
For the
the elements
elements of
of matrix-valued
matrix-valued function
function (2.5.2)
(2.5.2) taken
taken in
in the
the form
form
T
vvii (x i ) = xT Ix, ii =
= 1,
1, 2;
ii (xi ) = xii Ix, 2;
T
(x , x ) = v (x , x ) = x1 0.1
vv12 (x1 , x 2 ) = v21 (x1 , x2 ) = x T
0.1 Ix
Ix2 ,,
12 1 2 21 1 2 1 2
the
the estimates
estimates
2
vvii (xi ) xi 2 , ii =
= 1,
1, 2;
ii (xi ) xi , 2;
vv12 (x 1 , x2 )
(x , x ) 0.1 x x2
0.1 x 1 x
12 1 2 1 2
T
are
are satisfied.
satisfied. Let T =
Let = (1,
(1, 1),
1), then
then the
the matrix A corresponding
matrix A corresponding to
to the
the
matrix
matrix AA in
in estimate
estimate (2.5.5)
(2.5.5)
1
1 0.1
0.1
A =
A = 0.1 11
0.1
is
is positive
positive definite.
definite.
For
For this
this choice
choice of of the
the elements
elements vvij (), i, j = 1, 2, we have
ij (), i, j = 1, 2, we have
(1) (1) (1)
(1)
(1) for
for k ki =
= 0:
0: 11 (S)
(1)
(S) = 6,
= 6, 22 (S)
(1)
(S) =
= 0.2,
0.2,
(1) (S) = 1.1 + 0.02 S11 ,
(2) i 12 (S) = 1.1 + 0.02 S11 ,
12
(2) (S) = 0.9 + 0.02 S 22 , 12
21 (S) = 0.9 + 0.02 S22 , 12 (S) = 0.29;
21 (S) = 0.29;
(1) (1)
(2)
(2) for
for k kii =
=k kii =
= 1:
1: 11 (S)
(1)
(S) = 6 +
= 6 + 0.02
0.02 S
S11 , (1) (S) = 0.2 + 0.01 S22 ,
11 , 2
2 (S) = 0.2 + 0.01 S22 ,
(2) (2)
11 (S)
(S) = = 1.1
1.1 ++ 0.02
0.02 S + 0.001 S21, (2) (S) = 0.9 + 0.001 S12 +
(2)
S11
11 + 0.001 S21 , 2 2 (S) = 0.9 + 0.001 S12 +
0.02 , (S) = 0.29+0.011 +0.01
0.02 S22 , 12 (S) = 0.29+0.011 S11 +0.01 S12 +0.005 S
S S S +0.005 S21 +0.007
+0.007 SS22..
22 12 11 12 21 22
The
The matrix
matrix C
C corresponding
corresponding to
to the
the matrix
matrix C in
C in estimate
estimate (2.6.47)
(2.6.47) has
has
the form
the form
4.88 0.29
4.88 0.29 for kii =
= 0;
for k 0;
0.29
0.29 0.68
0.68
C =
=
C
4.859
4.859 0.323
0.323
for
kii = kii =
= 11
for k =k
0.323 0.669
0.323 0.669
and
and is
is negative
negative definite.
definite.
Thus,
Thus, all conditions of
all conditions of Theorem
Theorem 2.6.3
2.6.3 are
are satisfied
satisfied and
and the
the equilibrium
equilibrium
state
state xx==0 0 of
of system
system (2.6.43)
(2.6.43) with
with vectors
vectors and
and matrices
matrices (2.6.48)
(2.6.48) is
is struc-
struc-
turally asymptotically stable in the whole on
turally asymptotically stable in the whole on S. S.
2.7 NOTES AND REFERENCES 87
Section 2.3 The estimates for the class of Liapunov functions applied in this
section were obtained by Martynyuk and Miladzhanov [1, 2]. In this section some
results of Krasovskii [1], and Djordjevic [1] were used.
Section 2.4 This section is based on the results by Martynyuk [6], Martynyuk and
Miladzhanov [1 3], Martynyuk and Stavroulakis [1, 2], and Miladzhanov [1, 2].
Definition 2.4.1 is based on some results of Liapunov [1], Chetaev [2], and Sil-
jak [1 3] (see and cf. Grujic et al. [1], pp. 160, and Martynyuk and Miladzha-
nov [1, 2]).
Section 2.5 The results of analysis of large scale linear system are new. Alongside
the results obtained in Section 2.4 some results by Djordjevic [1] are applied.
Section 2.6 The results of Sections 2.6.1 and 2.6.3 are new as referred to the
application of matrix-valued function and two measures in the investigation of
stability of nonlinear system under nonclassical structural perturbations. Some
results
88 in the direction were presented
2. CONTINUOUS in the paper by
LARGE-SCALE Martynyuk [8]. The anal-
SYSTEMS
ysis of stability with respect to two measures of nonlinear systems without struc-
tural perturbations in terms of matrix-valued function was carried out by Mar-
tynyuk and Chernienko [1]. The investigations of motion stability with respect
to two measures by means of scalar Liapunov function were summarized in the
monograph by Lakshmikantham, Leela, et al. [1], Lakshmikantham and Liu [1].
The mathematical model of energy system studied in Section 2.6.2 was investi-
gated earlier by means of scalar (see Ribbens-Pavella [1, 2]) and vector Liapunov
functions (see Grujic et al. [1], Voronov and Matrosov [1], etc.). The results ob-
tained via the application of matrix-valued function in the investigation of this
model are presented according to Martynyuk and Miladzhanov [7].
The results for large-scale Lure-Postnikov systems under nonclassical struc-
tural perturbations are new.
91
Stability Theory of Large-Scale
Dynamical Systems Discrete-Time Large-Scale Systems
3
DISCRETE-TIME LARGE-SCALE SYSTEMS
3.1 Introduction
Maastricht
University is
the best specialist
university in the
Visit us and find out why we are the best! Netherlands
(Elsevier)
Masters Open Day: 22 February 2014
www.mastersopenday.nl
The object of investigation in this chapter are the models of real systems
and/or processes whose mathematical description is made by the systems of
the first order ordinary difference equations under nonclassical structural
and/or parametric perturbations. For the class of systems (subsystems)
under consideration the notations D and Di are used respectively and the
following hypotheses are adopted.
(3.2.1) P = {P : P1 P ( ) P2 , N+ }.
(3.2.2) xi ( + 1) = fi (, x( ), pi ), i = 1, 2, . . . , m,
ni
where xi R3.2 , AfiMODEL
Fi = OF , fi2 , . . . , fiN }, PERTURBATIONS
{fi1STRUCTURAL x = (xT T T T T
1 , x2 , . . . , xi , . . . , xm )91.
It is assumed that fi (, 0, 0) = 0 for all N+ .
The number N in the definition of the families F and Fi , i = 1, 2, . . . , m,
and the variations of the exponent k = k( ) on the set N = {1, . . . , N },
k( ) N for all N+ , describe structural changes in system (3.2.2).
H5 . The dynamics of the i-th isolated subsystem Di in system D is
described by the finite-dimensional first order ordinary difference equations
(3.2.3) xi ( + 1) = gi (, xi ), i = 1, 2, . . . , m,
Download free eBooks at bookboon.com
where xi Rni, and the functions gi : N+ Rni Rni are determined
by the correlations 93
gi (, xi ) = fi (, xi, 0), i = 1, 2, . . . , m.
gi (, xi ) = fi (, xi, 0), i = 1, 2, . . . , m.
(3.2.4) xi ( + 1) = gi (, xi ) + Si ( )hi (, x, pi ), i = 1, 2, . . . , m,
and taking into account the designations from Section 2.2, one can present
system (3.2.4) in the vector form
(3.2.6) T m (, x( ), P1 , S) x( + 1) T M (, x( ), P2 , S).
T m (, ( ), P1 , S) ( + 1) T M (, ( ), P2 , S)
for all N+ .
Since
T m (, x( ), P1 , S) T (, x( ), P, S) T M (, x( ), P2 , S)
Stability Theory
for all (, of
x, P, S) Large-Scale
N+ Rn P S, every solution of system (3.2.5) is
Dynamical Systems
the solution of inequality (3.2.6). Discrete-Time Large-Scale Systems
For the first differences of the matrix-valued function (3.3.1) and scalar
function (3.3.2) along the solutions of system (3.2.5) we introduce the no-
tation
95
Stability Theory of Large-Scale
Dynamical Systems Discrete-Time Large-Scale Systems
94 3. DISCRETE-TIME LARGE-SCALE SYSTEMS
Proof When the conditions of Theorem 3.3.1 are satisfied, so are the
inequalities
v(, x( ), ) > 0 and v(, x( ), ) 0
and
l
v( + 1, x( + 1), ) v(, x0 , )
2
and, therefore, x0 G. Since v(, x( ), ) is a continuous function, there
exists a > 0 such that B G0 . Therefore, if x0 B , then x0 G
and x( ; t0 , x0 ) G0 B , for any (P, S) P S. The theorem has thus
been proved.
Proof Since all the conditions of Theorem 3.3.1 are satisfied, the state
x( ) = 0 is stable for any (P, S) P S. From the fact that the matrix-
valued function U (, x( )) is a function of class AS() it follows that
there exist functions k K for every k = 1, 2, . . . , m, and matrices
G(P, S) P S (max (G) < 0) such that
or
v(, x( ), ) min (G)T(x( ))(x( )),
Proof Since all the conditions of Theorem 3.3.1 are satisfied, the state
x( ) = 0 is stable for any (P, S) P S. From the fact that the matrix-
valued function U (, x( )) is a function of class AS() it follows that
there exist functions k K for every k = 1, 2, . . . , m, and matrices
G(P, S) P S (max (G) < 0) such that
or
T
v(, x( ), )
3.3 LIAPUNOVS min (G) (x( ))(x(
MATRIX-VALUED FUNCTIONS)), 95
and, therefore,
v(, x( ), ) = TU (, x( ))
such that
97
Stability Theory of Large-Scale
Dynamical Systems Discrete-Time Large-Scale Systems
96 3. DISCRETE-TIME LARGE-SCALE SYSTEMS
v(, x( ), ) 0
and
v(, x( ), ) 0.
Therefore,
The contradiction that has been obtained shows that the statement of
Theorem 3.3.3 holds, as was to be proved.
(3.4.1) uT T T T
1 H AHu1 v(, x( ), ) u2 H BHu2
uT
k = (1k (x1 ( )), 2k (x2 ( )), . . . , mk (xm ( ))), k = 1, 2,
H = diag (1 , 2 , . . . , m ), A = [ij ], ij = ji ,
B = [ij ], ij = ji , i, j = 1, 2, . . . , m,
T
= (1 , 2 , . . . , m ).
(a) i2 {vii (, xii ( + 1)) vii (, xii ( ))} 1i 2i (xi ( )) for all
xi ( ) Ni0 , i = 1, 2, . . . , m;
m
(b) i2 {vii (, xi ( + 1)) vii (, xii ( + 1)) + vii (, xii ( ))
i=1
m
m
vii (, xi ( ))} + 2 i j {vij (, xi ( + 1), xj ( + 1))
i=1 j=1
m
vij (, xi ( ), xj ( ))} 2i (P, S)2i (xi ( ))
i=1
m
m
+2 ij (P, S)i (xi ( ))j (xj ( )),
i=1 j=1
Remark 3.4.1 Here, xii () means that the difference is taken with regard
for subsystems (3.2.5)
(3.4.4) v11 = xT
1 B1 x1 , v22 = xT
2 B2 x2 , v12 = v21 = xT
1 B3 x2 ,
where
C1 = AT
11 B1 A11 B1 , C2 = AT
22 B2 A22 B2 ,
C4 (S) = (A12 s12 )TB1 A12 s12 + 2(A12 s12 )TB3 A22 BBy2 ,2020, wind could provide one-tenth of our planets
Brain power
C5 (S) = AT T T
11 B3 A22 + (A21 s21 ) B3 A12 s12 B3 .
electricity needs. Already today, SKFs innovative know-
how is crucial to running a large proportion of the
worlds wind turbines.
Thus, if inequality (3.4.5) is satisfied and there exists a negative
Up to 25semi-
% of the generating costs relate to mainte-
1 nance. These can be reduced dramatically thanks to our
definite matrix G such that (C(S) + C T(S)) G for all S S, then
systems all condition monitoring and automatic
for on-line
2 lubrication. We help make it more economical to create
conditions of Theorem 3.4.1 (3.4.2) are satisfied and the zero solution of the
cleaner, cheaper energy out of thin air.
large scale discrete-time system (3.4.3) is stable (asymptotically Bystable).
sharing our experience, expertise, and creativity,
industries can boost performance beyond expectations.
3.4.2 The Problem DB In this section we propose the solution Therefore we need the best employees who can
of the
meet this challenge!
problem of nonlinear dynamics of discrete-time systems.
Problem DB Assume that the discrete-time system D is obtained in re-of Knowledge Engineering
The Power
sult of composition of the interconnected subsystems (3.2.2). It is required
to determine stability conditions of different types for the equilibrium state
x = 0 of system (3.2.5) in terms of the dynamical properties of the intercon-
nected subsystems (3.2.4) without additional information on the dynamical
properties of the isolated subsystems (3.2.3).
where
C1 = AT
11 B1 A11 B1 , C2 = AT
22 B2 A22 B2 ,
C4 (S) = (A12 s12 )TB1 A12 s12 + 2(A12 s12 )TB3 A22 B2 ,
C5 (S) = AT T T
11 B3 A22 + (A21 s21 ) B3 A12 s12 B3 .
m
i2 {ii (, xi ( + 1)) ii (, xi ( ))}
i=1
m
m
+2 i j {ij (, xi ( + 1), xj ( + 1)) ij (, xi ( ), xj ( + 1))}
i=1 j=2
j>i
m
m
m
ii (P, S)2i (xi ( )) + 2 ij (P, S)i (xi ( ))j (xj ( ))
i=1 i=1 j=2
j>i
where
where
C1 (S) = AT T T
11 B1 A11 + (A21 s21 ) B2 A21 s21 + 2A11 B3 A21 s21 2B1 ,
C2 (S) = AT T T
22 B2 A22 + (A12 s12 ) B1 A12 s12 + 2(A12 s12 ) B3 A22 2B2 ,
C3 (S) = AT T T
11 B3 A22 + (A21 s21 ) B3 A12 s12 B3 .
Thus, if, for system (3.4.3), the matrix function with elements (3.4.4)
is constructed, inequality (3.4.5) is satisfied, and, in addition, there exists
a negative semi-definite (negative definite) matrix G such that 1 (C(S) +
2
C T(S)) G for all S S, then the conditions of Theorem 3.4.3 (3.4.4)
are satisfied and the zero solution of the multidimensional discrete system
(3.4.3) is stable (asymptotically stable) on P S.
Example 3.4.3 Consider a discrete system of the fourth order which
consists of two subsystems of the second order described by the following
systems of equations
104
Stability Theory of Large-Scale
Dynamical Systems Discrete-Time Large-Scale Systems
104 3. DISCRETE-TIME LARGE-SCALE SYSTEMS
For system (3.4.7), we construct a matrix function U (x) with the elements
vii (xi ) xi 2 , i = 1, 2.
The matrices
1 0.5 1 0.5
A= and B=
0.5 1 0.5 1
The matrices
Stability Theory of
Large-Scale
1 0.5 1 0.5
A=
Dynamical Systems and B= Discrete-Time Large-Scale Systems
0.5 1 0.5 1
is negative definite.
Since the conditions of Theorem 3.4.4 are satisfied, the zero solution of
system (3.4.7) is asymptotically stable on S.
In conclusion, note that the construction of a scalar Liapunov function
for the multidimensional discrete system (3.2.5) is an important and dif-
ficult problem in the theory of discrete systems. The application of the
matrix-valued function U (, x( )) to the construction of the scalar func-
tion v(, x( ), ) simplifies the problem to a certain extent due to weakened
requirements on the components vij , i, j = 1, 2, . . . , m. This, in turn, en-
ables one to more adequately take into account the correlation between the
independent subsystems (3.2.3).
We emphasize that the suggested method for the analysis of the stability
of the multidimensional discrete system (3.2.5) is distinguished by its sim-
plicity and generality, and all established sufficient conditions of stability
and asymptotic stability are represented in terms of the property of special
matrices to have a fixed sign.
(3.5.1) x( + 1) = A x( ) + f (x( ), ),
(3.5.2) x( + 1) = (A + E) x( ),
106
Stability Theory of Large-Scale
Dynamical Systems Discrete-Time Large-Scale Systems
106 3. DISCRETE-TIME LARGE-SCALE SYSTEMS
3.5.1 Scalar approach We assume that for the matrix A the condition
|i (A)| < 1 is realized for all i = 1, 2, . . . , n. In this case the Liapunov
equation
(3.5.3) ATP A P = G
(3.5.5) x( + 1) = Ax( )
be asymptotically stable. If
where
m (G)
(G) = 1/2 1/2
,
M (P G)M (P ) + M (P )
then the uncertain system (3.5.2) is asymptotically stable.
Here E = sup Ex, x = (xTx)1/2 is the Euclidean norm of
x1
3.5 GENERALIZATIONS AND APPLICATIONS 107
vector x.
It is known (see Sezer and Siljak [1]), that (G) takes the largest value,
if G = I . The expression (3.5.6) is a robust bound for the system (3.5.2),
obtained in the framework of scalar approach.
(3.5.10) xi ( + 1) = Ai xi ( )
(3.5.11) AT
i Pi Ai Pi = Gi , i = 1, 2,
(3.5.12) B1 B2 < 2 1 2
1/2 1/2
where i = (M (Pi Ii )M (Pi ) + M (Pi ))1 , Pi are solutions of
the Liapunov matrix equations (3.5.11) for the matrices Gi = Ini ,
Ini are ni ni identity matrices, i = 1, 2.
1/2 1/2
+ M (Pi )Ei xi + M (Pi )(Bi + Ui )xj
1/2 1/2
(3.5 GENERALIZATIONS AND APPLICATIONS
i M (Pi )Ei )xi + M (Pi ) (Bi + Ui ) xj ,
109
where i, j = 1, 2, i = j.
Here we use the known inequality (see Sezer and Siljak [1])
1/2
(pTP p)1/2 (q TP q)1/2 M (P )p q
1/2 1/2
w11 w22 w12 w21 = [1 M (P1 )E1 ][2 M (P2 )E2 ]
1/2 1/2
M (P1 )M (P2 )(B1 + U1 )(B2 + U2 )
where d = (d1 , d2 )T, z = (x1 , x2 )T, W = (wij ) is 2 2 matrix with
the elements
Stability Theory of Large-Scale
Dynamical Systems 1/2
i M (Pi ) Ei if i = j, Discrete-Time Large-Scale Systems
wij = 1/2
M (Pi )(Bi + Ui ) if i = j.
1/2 1/2
w11 w22 w12 w21 = [1 M (P1 )E1 ][2 M (P2 )E2 ]
1/2 1/2
M (P1 )M (P2 )(B1 + U1 )(B2 + U2 )
1/2 1/2
> 1 M (P1 )(1 )1 2 M (P2 )(1 )2
1/2 1/2
M (P1 )M (P2 )(B1 + )(B2 + ) = 2 1 2
1/2 1/2 1/2 1/2
M (P1 )M (P2 )(B1 + )(B2 + ) = M (P1 )M (P2 )2
1/2 1/2
M (P1 )M (P2 )(B1 + B2 ) + 2 1 2
1/2 1/2
M (P1 )M (P2 )B1 B2 = a2 b + c.
1/2 1/2
c = 2 1 2 M (P1 )M (P2 )B1 B2
1/2 1/2
= M (P1 )M (P2 ) 2 1 2 B1 B2 > 0
ij :
C xij ( + 1) = (Aij + Eij ) xij ( ) + (Bij + Uij ) xik ( ),
(3.5.17)
i, j, k = 1, 2, j = k,
where xij Rnij, Rni = Rni1 Rni2, Aij , Eij Rnij nij, Bi1 , Ui1
Rni1 ni2, Bi2 , Ui2 Rni2 ni1,
Ai1 Bi1 Ei1 Ui1
Ai = , Ei = .
Bi2 Ai2 Ui2 Ei2
(i) (i)
where Mjk , Fjk Rnij nlk, i, j, k, l = 1, 2, i = l.
We extract from (3.5.17) the independent components
are asymptotically stable, i.e. there exist unique symmetric and positive
definite matrices Pij , which satisfy the Liapunov matrix equations
(3.5.21) AT
ij Pij Aij Pij = Gij , i, j = 1, 2,
by formula
vi (xi ) = di1 vi1 (xi1 ) + di2 vi2 (xi2 ), i = 1, 2,
where dij are some positive constants. We introduce 2 2 matrices Wi =
(i)
(wjk ) with the elements
(i) i ij if j = k,
wjk = 1/2
M (Pij )(Bij + i ) if j = k.
by formula
vi (xi ) = di1 vi1 (xi1 ) + di2 vi2 (xi2 ), i = 1, 2,
where dij are some positive constants. We introduce 2 2 matrices Wi =
(i)
(wjk ) with the elements
(i) i ij if j = k,
wjk = 1/2
M (Pij )(Bij + i ) if j = k.
where
(1) (1)
w21 d11 w
D= d = (d11 , d12 , d21 , d22 )T R4 : (1)
< < 22
(1)
,
w11 d12 w 12
(2) (2)
w21 d21 w
(2)
< < 22
(2)
,
w11 d22 w 12
(1) (1) (2) (2)
w22 w21 w22 w21
d1 = (1) (1)
, d2 = (2) (2)
.
w11 w12 w11 w12
(i) (i)
Proof As the matrices W1 and W2 are the M-matrices, then w11 , w22 >
(i) (i)
0, w12 , w21 < 0 and consequently,
(i) (i)
w22 w21
(i)
> (i)
> 0.
w12 w11
Let us denote
i
(3.5.27) mi (di ) = 1 1 , i = 1, 2,
di M (Pi1 ) + M
2 2
(Pi2 )
113
Stability Theory of Large-Scale
Dynamical Systems Discrete-Time Large-Scale Systems
114 3. DISCRETE-TIME LARGE-SCALE SYSTEMS
RUN FASTER.
RUN LONGER.. READ MORE & PRE-ORDER TODAY
RUN EASIER WWW.GAITEYE.COM
115
Stability Theory of Large-Scale
Dynamical Systems Discrete-Time Large-Scale Systems
116 3. DISCRETE-TIME LARGE-SCALE SYSTEMS
3.5.4.1 Scalar approach Let us compute the matrices and constants oc-
curring in the framework of the scalar approach (see Theorem 3.5.1):
1.336149 0.008512 0.032104 0.000737
0.008512 1.017019 0.000708 0.007761
P = ;
0.032104 0.000708 1.068495 0.002057
0.000737 0.007761 0.002057 1.016891
Let
1 = 0.5, 2 = 0.125.
Download values
Numerical free eBooks
of theatcorresponding
bookboon.com constants are:
116
M (P11 ) 1.333333, M (P12 ) 1.015873, 11 = 0.5, 12 = 0.875,
M (P21 ) 1.066666, M (P22 ) 1.015873, 21 = 0.75, 22 = 0.875,
1 0.320718, 2 0.096261.
trix (3.5.32) and as a result we get:
We shall set 1 = 0.05, and 2 = 0.006. In this case for the matrices W1
and W2 we get the expressions
0.288675 0.069282 0.096824 0.011360
W1 , W2 .
0.060474 0.440958 0.011086 0.110239
Section 3.2 The large scale discrete-time system with structural perturbations
is composed according to the approach presented in Chapter 2. As in the contin-
uous case this allows the application of the generalized Liapunov direct method
based on matrix-valued function in the construction of the corresponding stability
conditions.
Section 3.3 In this section the matrix-valued function is applied within the
framework of the scalar approach. The section is based on the results by Mar-
tynyuk [12].
Section 3.4 The main results of this section are due to Martynyuk and Mi-
ladzhanov [1], and Martynyuk, Miladzhanov, and Muminov [1].
Section 3.5 This section is based on the results by Lukyanova and Martynyuk [1].
The comparative analysis of the obtained results is carried out using some results
by Sezer and Siljak [1].
118
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
4
IMPULSIVE LARGE-SCALE SYSTEMS
4.1 Introduction
dxi
= fi (t, x, pi ) for t = k (xi ),
(4.2.1) dt
xi (t) = Iik (xi ) + Iik (x) for t = k (xi ), k = 1, 2, . . . ,
www.sylvania.com
We do not reinvent
the wheel we reinvent
light.
Fascinating lighting offers an infinite spectrum of
possibilities: Innovative technologies and new
markets provide both opportunities and challenges.
An environment in which your expertise is in high
demand. Enjoy the supportive working atmosphere
within our global group and benefit from international
career paths. Implement sustainable ideas in close
cooperation with other specialists and contribute to
influencing our future. Come and join us in reinventing
light every day.
Light is OSRAM
dxi
= wi (t, x) for t = k (x),
(4.2.2) dt
xi (t) = Iik (xi ) for t = k (xi ), k = 1, 2, . . . ,
where xi = (0, . . . , 0, xT T n
i , 0, . . . , 0) R .
dxi
= wi (t, x) + Si (t)ri (t, x, pi ) for t = k (x),
dt
(4.2.3)
xi (t) = Iik (x) + Iik (x) for t = k (x),
k = 1, 2, . . . , i = 1, 2, . . . , m,
dx
= w(t, x) + S(t)r(t, x, P ), t = k (x),
(4.2.4) dt
x = Ik (x), t = k (x), k = 1, 2, . . . .
The matrix
The functions i (x) and a number satisfy the condition which excludes
the beating of the solutions of system (4.2.4) on the surfaces
(Sk ) : t = (k (x)), k = 1, 2, . . . .
The possibility
Download of beating
free eBooks of the solutions of the equations on the surfaces
at bookboon.com
(Sk ) : t = (k (x)), k = 1, 2, . . . , causes an essential difficulty when large
121 in the general case
scale impulsive system (4.2.4) is investigated. Moreover,
there is no continuous dependence of the solutions of large scale impulsive
system (4.2.4) on the initial conditions which would be uniform on finite
interval. This fact requires the notion of stability in the sense of Liapunov of
Throughout this chapter, we will assume that for each (t0 , x0 ) T0
Stabilitythere
D(), Theoryexists at least one solution of (4.2.4) for all (P, S) P S
of Large-Scale
which satisfies
Dynamical Systems the initial condition x(t0 ) = x0 . Impulsive Large-Scale Systems
nj nj
Stability Theory of NLarge-Scale
jx = {xj R : xj < hj0 } R
Dynamical Systems Impulsive Large-Scale Systems
of the states xj = 0, j = 1, 2, . . . , m, hj0 = const > 0;
(2) functions j1 , j1 : Njx R+ , j1 , j1 K;
(3) constants aji , bji , j, i = 1, 2, . . . , m, and a matrix-valued function
U (t, x) = [vji (t, )] with the elements
where
DU (t, x) = [Dvji (t, )], thinking
j, i = 1, 2, . . . , m,
due to the system (4.2.4). For the details see Section 1.4.
Proposition 4.4.1 If all conditions of Assumption 4.4.1 are satisfied,
then for the function (4.4.1)
(4.4.3) uT T T T
1 H AHu1 v(t, x, ) u2 H BHu2 , for all (t, x) T0 Nx ,
where
360 uT
1 = (11 (x1 ), 21 (x2 ), . . . , m1 (xm )),
.
uT
2 = (11 (x1 ), 21 (x2 ), . . . , m1 (xm )),
thinking
H = diag [1 , 2 , . . . , m ], A = [aji ], B = [bji ],
aji = aij , bji = bij ,
360
j, i = 1, 2, . . . , m,
Nx = x R : x < 0 , n
thinking
0 = min hj0 .
.
j
(4.4.3) uT T T T
1 H AHu1 v(t, x, ) u2 H BHu2 , for all (t, x) T0 Nx ,
where
uT
1 = (11 (x1 ), 21 (x2 ), . . . , m1 (xm )),
uT
2 = (11 (x1 ), 21 (x2 ), . . . , m1 (xm )),
H = diag [1 , 2 , . . . , m ], A = [aji ], B = [bji ],
aji = aij , bji = bij , j, i = 1, 2, . . . , m,
j (0) =AND
4.4 STABILITY m (0) = M (0) =ANALYSIS
INSTABILITY 0 125
and
m
0 < m (v(t, x, )) 2j (vjj (t, xj )) M (v(t, x, ))
j=1
are satisfied;
(1) (2)
(3) constants j , j (P, S), ji (P, S), j = i = 1, 2, . . . , m, and
(1)
(a) j2 Dt vjj + (Dxj vjj )Tfj (t, xj ) j 2j (vjj (t, xj )) for all (t, xj )
T0 Njx0 , j = 1, 2, . . . , m;
m
(b) j2 (Dxj vjj )TSj (t)rj (t, x, Pj )
j=1
m
m
+2 j i {Dt vji + (Dxj vji )T[fj (t, xj ) + Sj (t)rj (t, x, Pj )]
j=1 i=2
i>j
and
j=1
m m
+2 ji (P, S)j (vjj (t, xj ))i (vii (t, xi ))
Stability Theoryj=1 of i=2
Large-Scale
Dynamical Systemsi>j Impulsive Large-Scale Systems
for all (t,xj , xi , P, S) T0 Njx0 Nix0 P S, j = i, j = 1, 2, . . . , m.
and
t = k (x), k = 1, 2, . . . .
(4.4.6)
Dv(t, x, ) M (Q)m (v(t, x, )), for all (t, x) T0 Nx0 ;
(4.4.7)
Dv(t, x, ) M (Q)M (v(t, x, )), for all (t, x) T0 Nx0 ;
125
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
4.4 STABILITY AND INSTABILITY ANALYSIS 127
where
an opportunity of a lifetime
C = [cji ], j, i = 1, 2, . . . , m, cji = cij ,
(1) (2)
cjj = j + j , cji = ji , j = i, j, i = 1, 2, . . . , m.
Do you like cars? Would you like to be a part of a successful brand? Send us your CV on
We will appreciate and reward both your enthusiasm and talent. www.employerforlife.com
Send us your CV. You will be surprised where it can take you.
where
v(k (x), x + Jk (x)) v(k (x), x) = T [v(k (x), x + Jk (x)) v(k (x), x)]
m
= j2 {vjj (k (x), xj + Jkj (x)) vjj (k (x), xj )}
j=1
m
m
+2 j l {vjl (k (x), xj + Jkj (x)), xl + Jkl (x)) vjl (k (x), xj , xl )}
j=1 l=2
l>j
m
= j2 {vjj (k (xj ), xj + Jkj (xj )) vjj (k (xj ), xj )}
j=1
m
+ j2 {vjj (k (x), xj + Jkj (x)) vjj (k (xj ), xj )}
j=1
m
+ j2 {vjj (k (x), xj + Jkj (x))
j=1
m
m
+2 j l {vjl (k (x), xj + Jkj (x)), xl + Jkl (x)) vjl (k (x), xj , xl )}
j=1 l=2
l>j
m
m
(1) (2) (2) (2)
j j (vjj (k (xj ), xj )) + j j (vjj (k (xj ), xj )
j=1 j=1
m
m
+2 jl j (vjj (k (xj ), xj ))l (vll (k (xl ), xl ))
j=1 l=2
l>j
m
= clj j (vjj (k (xj ), xj ))l (vll (k (xl ), xl )) = uT
k Cuk , k = 1, 2, . . . .
j,l=1
(4.4.10)
v(k (x), xj + Jkj (x)) v(k (x), x) M (C) m (v(k (x), x));
(4.4.11)
v(k (x),
Download xj + Jkjat(x))
free eBooks v(k (x), x) M (C) M (v(k (x), x))
bookboon.com
(4.4.10)
v(k (x), xj + Jkj (x)) v(k (x), x) M (C) m (v(k (x), x));
(4.4.11)
v(k (x), xj + Jkj (x)) v(k (x), x) M (C) M (v(k (x), x))
(1)
(a) j2 vjj (k (xj ), xj + Jkj (xj )) j j2 (vjj (k (xj ), xj ))
for all xj Nji , j = 1, 2, . . . , m;
m
(b) j2 {vjj (k (x), xj + Jkj (x)) vjj (k (xj ), xj + Jkj (xj ))}
j=1
m
m
+2 j l vjl (k (x), xj + Jkj (x), xl + Jkl (x))
j=1 k=2
m
(2)
j j2 (vjj (k (xj ), xj ))
j=1
m
m
+2 jl j (vjj (k (xj ), xj ))l (vll (k (xl ), xl ))
j=1 l=2
l>j
T
(4.4.12)
130 v(k (x), 4.
x +IMPULSIVE
Jk (x)) uLARGE-SCALE
k C uk , k = 1, 2, . . . for all
SYSTEMS x Nx ,
where
respectively. 128
Here M (C ) is a maximal eigenvalue of the matrix C .
Corollary 4.4.3 is proved in the same way as Corollary 4.4.1.
Assumption 4.4.5 Assume that the conditions (1) and (2) of Assump-
Corollary 4.4.3 If all conditions of Assumption 4.4.4 are satisfied and
(1) M (C ) < 0;
Stability Theory of Large-Scale
(2) (C ) > 0,
DynamicalMSystems Impulsive Large-Scale Systems
then for all k = 1, 2, . . . and for all x Nx0
respectively.
Here M (C ) is a maximal eigenvalue of the matrix C .
Corollary 4.4.3 is proved in the same way as Corollary 4.4.1.
Assumption 4.4.5 Assume that the conditions (1) and (2) of Assump-
tion 4.4.2 are satisfied and in the inequalities of condition (3) of Assumption
4.4.2 the inequality sign is reversed .
Proposition 4.4.5 If conditions of Assumption 4.4.5 are satisfied, then
for expression (4.4.2):
AXA Global
Graduate Program
Find out more and apply
m
(2)
= j {Dt vjj + (Dxj vjj )Tfj (t, xj ) + (Dxj vjj )TSj (t)rj (t, x, Pj )}
j=1
m
m
+2 j l {Dt vjl + (Dxj vjl )T(fj (t, xj ) + Sj (t)rj (t, x, pj ))
j=1 l=2
l>j
which has an open intersection with the plane t = const for each t T0 .
Moreover, the positiveness of the functions ujj (t, xj ), j = 1, 2, . . . , m, is
a necessary condition for the positiveness of function v(t, x, ), therefore
m j for every j = 1, 2, . . . , m by the condition of Assumption
j=1
4.4.8 has non-zero open intersection with the plane t = const adjoining to
the origin. This proves the assertion (a) of Proposition 4.4.8.
The boundedness of functions uji , j, i = 1, 2, . . . , m, implies that the
matrix U (t, x) is bounded, but then the function v(t, x, ) constructed by
formula (4.4.1) will be bounded as well.
1
(G(P, S) + GT(P, S)) Q
134 2
4. IMPULSIVE LARGE-SCALE SYSTEMS
for all (P, S) P S, t = k (x), k = 1, 2, . . . ,
is satisfied element-wise;
(4) the matrix C is negative semi-definite or equal to zero (i.e.
M (C) 0).
Then the zero solution of large-scale impulsive system (4.2.4) is stable
on P S.
If condition (4) is modified as follows
(5) the matrix C is negative definite (i.e. M (C) < 0),
then the zero solution of large scale impulsive system (4.2.4) is asymptoti-
cally stable on P S.
In this connection for each pair (P, S) P S the conditions being suf-
ficient for the stability of the zero solution of large scale impulsive system
(4.2.4) are satisfied on P S.
If instead of condition (4) of Theorem 4.4.1 the condition (4.2.4) of the
Theorem is satisfied, then Proposition 3 and Corollary 4.4.2 yield
v(k (x), x + Jk (x)) v(k (x), x) M (C)M (v(k (x), x)), k = 1, 2, . . . Month 16
I was a construction
and therefore the conditions being sufficient for asymptotic stability are
Mo
satisfied for every (P, S) P S. Hence, the zero solution of large scale supervisor ina const
I was
impulsive system (4.2.4) is asymptotically stable on P S.
the North Sea super
Theorem 4.4.2 Let large scale impulsive system (4.2.4) be such that
(1) on the domain T D() Hypotheses 1, 2 and 4 are satisfied;
advising and the No
he
helping foremen advis
0
T
(2) the matrix A is positive definite (i.e. (H AH)
m Real> 0);
work
International
al opportunities
Internationa
ree wo
work
or placements ssolve problems
Real work he
helping fo
International
Internationaal opportunities
ree wo
work
or placements ssolve pr
Download free eBooks at bookboon.com
and the conditions of Assumption 4.4.3, Proposition 4.4.3 and condition (4)
of Theorem 4.4.1 yield
In this connection for each pair (P, S) P S the conditions being suf-
ficient for the stability of the zero solution of large scale impulsive system
(4.2.4) are satisfied on P S.
If instead of condition (4) of Theorem 4.4.1 the condition (4.2.4) of the
Theorem is satisfied, then Proposition 3 and Corollary 4.4.2 yield
and therefore the conditions being sufficient for asymptotic stability are
satisfied for every (P, S) P S. Hence, the zero solution of large scale
impulsive system (4.2.4) is asymptotically stable on P S.
Theorem 4.4.2 Let large scale impulsive system (4.2.4) be such that
(1) on the domain T0 D()
4.4 STABILITY ANDHypotheses
INSTABILITY 1, 2ANALYSIS
and 4 are satisfied; 135
(2) the matrix A is positive definite (i.e. m (H TAH) > 0);
(3) there exists a negative definite matrix Q Rmm such that the
estimate
1
G(P, S) + GT(P, S) Q for all (P, S) P S
2
is satisfied;
(4) M (C ) > 0
(5) the functions k (x), k = 1, 2, . . . , satisfy the inequality
sup min k+1 (x) max k (x) = > 0,
k xD() xD()
where < 0 .
If there exists a constant 0 > 0 such that for every (0, 0 ] the
functions m (y) and M (y) satisfy the inequality
M (C )M ()
1 dy
(4.4.23) ,
M (Q) m (y)
then the zero solution of large scale impulsive system (4.2.4) is stable on
P S.
If instead of inequality (4.4.23) for some > 0 the inequality
M (C )M ()
1 dy
(4.4.24)
M (Q) m (y)
is satisfied, then the zero solution of large scale impulsive system (4.2.4) is
asymptotically stable on P S.
The proof of Theorem 4.4.2 is similar to that of Theorem 4.4.1.
Theorem 4.4.3 Let large scale impulsive system (4.2.4) be such that
(1) on the domain T0 D() Assumptions 4.4.1, 4.4.2 and 4.4.4 hold;
Download free eBooks at bookboon.com
(2) the matrix A is positive definite (i.e. m (H TAH) > 0);
(3) there exists a matrix Q Rmm for which
133
1 for all (P, S) P S;
(a) (G(P, S) + GT(P, S)) Q
2
(b) M (Q) > 0;
M (C )M ()
1 dy
(4.4.24)
M (Q) m (y)
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
is satisfied, then the zero solution of large scale impulsive system (4.2.4) is
asymptotically stable on P S.
The proof of Theorem 4.4.2 is similar to that of Theorem 4.4.1.
Theorem 4.4.3 Let large scale impulsive system (4.2.4) be such that
(1) on the domain T0 D() Assumptions 4.4.1, 4.4.2 and 4.4.4 hold;
(2) the matrix A is positive definite (i.e. m (H TAH) > 0);
(3) there exists a matrix Q Rmm for which
1 for all (P, S) P S;
(a) (G(P, S) + GT(P, S)) Q
2
136 4. IMPULSIVE LARGE-SCALE SYSTEMS
(b) M (Q) > 0;
(4) M (C ) > 0;
(5) the functions k (x), k = 1, 2, . . . , satisfy for some 1 > 0 the
inequality
(6) there exists a constant 0 such that for every (0, 0 ] the func-
tions m (y) and M (y) satisfy the inequality
1 dy
(4.4.25) 1 + .
M (Q) m (y)
M (C )M ()
1
(G(P, S) + GT(P, S)) L for all (P, S) P S;
2
134
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
4.4 STABILITY AND INSTABILITY ANALYSIS 137
Dv(t, x, ) 0, t = k (x), k = 1, 2, . . . ,
is satisfied.
Under the conditions of Assumption 4.4.6, Proposition 4.4.6, Corollary
4.4.5 and condition (4) of Theorem 4.4.4 one has
93%
(6) there exists a constant 0 such that for every (0, 0 ] the func-
tions M (y) and m (y) satisfy the inequality
OF MIM STUDENTS ARE
m (C )m ()
WORKING1 IN THEIR SECTOR dy 3 MONTHS
(4.4.26) FOLLOWING 1 + .
m (L) MGRADUATION
(y)
MASTER IN MANAGEMENT
STUDY IN THE CENTER OF MADRID AND TAKE ADVANTAGE OF THE UNIQUE OPPORTUNITIES
Length: 1O MONTHS
THAT THE CAPITAL OF SPAIN OFFERS
Av. Experience: 1 YEAR
PROPEL YOUR EDUCATION BY EARNING A DOUBLE DEGREE THAT BEST SUITS YOUR
Language: ENGLISH / SPANISH
PROFESSIONAL GOALS
Format: FULL-TIME
STUDY A SEMESTER ABROAD AND BECOME A GLOBAL CITIZEN WITH THE BEYOND BORDERS
Intakes: SEPT / FEB
EXPERIENCE
(6) there exists a constant 0 such that for every (0, 0 ] the func-
tions M (y) and m (y) satisfy the inequality
m (C )m ()
1 dy
(4.4.26) 1 + .
138
m (L)
4. IMPULSIVE M (y) SYSTEMS
LARGE-SCALE
Then the zero solution of large scale impulsive system (4.2.4), (4.2.5) is
unstable on P S.
Theorem 4.4.5 is proved in the same way as Theorem 4.4.4. Besides, for
each pair (P, S) P S all conditions of Theorem 1.4.6 with the functions
are satisfied.
Theorem 4.4.6 Let large scale impulsive system (4.2.4), (4.2.5) be
such that
(1) Assumption 4.4.1, 4.4.5, 4.4.7 and 4.4.8 hold in the domain ;
(2) the matrix A is positive definite (i.e. m (H TAH) > 0);
(3) there exists a positive definite matrix L Rmm for which
1
(G(P, S) + GT(P, S)) L for all (P, S) P S;
2
(4) m (C ) > 0;
(5) for some constant > 0 the functions k (x) satisfy the inequality
sup min k1 (x) max k (x) = > 0, < 0 ;
k xD() xD()
(6) for a constant > 0 the functions m (y) and m (y) satisfy the
inequality
1 dy
.
m (L ) m (y)
m (C )m ()
Then the zero solution of large scale impulsive system (4.2.4), (4.2.5) is
unstable on P S.
This Theorem is proved in the same way as Theorem 4.4.4. We note
that all hypotheses on instability of Theorem 1.4.6 are satisfied with the
functions
dxj
= x3j + 0, 5x3i + 0, 25Sj1 x3j + 0, 3Sj2 (t)x3i ,
dt
Download free eBooks attbookboon.com
= k (x1 , x2 ),
(4.4.27)
xj = xj + xi , t = k (x1 , x136
2 ),
j, i = 1, 2; j = i,
For the system (4.4.27) we construct the matrix function U (x) with the
elements
vjj (xj ) xj 2 , j = 1, 2,
v12 (x1 , x2 ) = v21 (x1 , x2 ) 0, 5x1 x2 .
The matrix
1 0, 5
140 A = LARGE-SCALE SYSTEMS
4. IMPULSIVE
0, 5 1
is positive definite.
Let T = (1, 1), then for the above-mentioned matrix-valued function
U (x), the elements of matrix G(S) (see (4.4.4)) and matrix C (see (4.4.9))
have the form
and
Here Bii are constant positive definite matrices, and Bij are constant ma-
trices.
We introduce the following assumption.
Assumption 4.5.1 Assume that there exist
(1) the matrix-valued function (4.5.2) with the elements (4.5.3) and
(4.5.4);
(2) the constants aji , bji , i, j = 1, 2, . . . , m, satisfying the estimates
(a) aii xi 2 Uii (xi ) bii xi 2 for all xi Nix , i = 1, 2, . . . , m,
(b) aji xj xi Uij (xi , xj ) bji xj xi for all (xi , xj ) Nix
Njx , i = j, i, j = 1, 2, . . . , m.
and
Here Bii are constant positive definite matrices, and Bij are constant ma-
trices.
We introduce the following assumption.
Assumption 4.5.1 Assume that there exist
(1) the matrix-valued function (4.5.2) with the elements (4.5.3) and
(4.5.4);
(2) the constants aji , bji , i, j = 1, 2, . . . , m, satisfying the estimates
(a) aii xi 2 Uii (xi ) bii xi 2 for all xi Nix , i = 1, 2, . . . , m,
(b) aji xj xi Uij (xi , xj ) bji xj xi for all (xi , xj ) Nix
Njx , i = j, i, j = 1, 2, . . . , m.
uT = (x1 , x2 , . . . , xm ), A = [aij ],
B = [bij ], H = diag [1 , 2 , . . . , m ].
where
uT = (x1 , x2 , . . . , xm ),
G(S) = [ ji (S)], i, j = 1, 2, . . . , m,
ji (S) = ij (S),
(1) (2)
jj (S) = j + j (S),
ji (S) = ji (S), j = i, i, j = 1, 2, . . . , m.
2
(4.5.11) 1 1
M (B)v(x, ) u m (A)v(x, ).
140
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
144 4. IMPULSIVE LARGE-SCALE SYSTEMS
144 4. IMPULSIVE LARGE-SCALE SYSTEMS
Proposition 4.5.3 If for the system (4.5.1) the condition (1) of As-
Proposition
sumption 4.5.1 is4.5.3 If for
satisfied, thefor
then system (4.5.1) the
the function condition
(4.5.5) when (1)
t = of As-
k (x),
sumption 4.5.1 is satisfied,
k = 1, 2, . . . , the inequalitiesthen for the function (4.5.5) when t = k (x),
k = 1, 2, . . . , the inequalities
(4.5.12) v(x + Jk (x), ) v(x, ) uT k Cuk ;
(4.5.12) v(x + Jk (x), ) v(x, ) uT k Cuk ;
and
and
(4.5.13) v(x + Jk (x), ) uT C u ,
T k
k
(4.5.13) v(x + Jk (x), ) uk C uk ,
are satisfied, where
are satisfied, where
uTk = (x1 (k (x)), x2 (k (x)), . . . , xm (k (x))),
uTk = (x1 (k (x)), x2 (k (x)),
m . . . , xm (k (x))),
Jk (x) = Jki xi + m J
kij xj ,
Jk (x) = Jki xi + j=1 Jkij xj ,
j=i
j=1
j=i
C = [cij ], cij = cji , i, j = 1, 2, . . . , m,
C = [cij ], cij = cji , i, j = 1, 2, . . . , m,
C = [cij ], cij = cji , i, j = 1, 2, . . . , m,
C = [cij ], 1/2 cij = cji , i, j = 1, 2, . . . , m,
T
cii = M (Cii ), cij = M 1/2
(Cij Cij T
), i = j, i, j = 1, 2, . . . , m,
cii = M (Cii ), 1/2
cij = M (Cij Cij ), i = j, i, j = 1, 2, . . . , m,
M
1/2
() is a norm of matrix (),
M () 1/2 is a norm of matrix (),
c ii = M (Cii ), cij = M
1/2
T
(Cij C T ), i = j, i, j = 1, 2, . . . , m;
ij
c ii = M (Cii ), cij = M (Cij
Cij ), i = j, i, j = 1, 2, . . . , m;
and
and
m
T T m JT B J
Cii = Jki B
T ii
+ B J
ii ki + J B J
T ii ki
ki + kij
T jj kji
Cii = Jki Bii + Bii Jki + Jki Bii Jki + j=1 Jkij Bjj Jkji
j=i
j=1
m m j=i
m T m T T
Excellent
+ Economics
(Bij Jkji +and Jkji Business
Bij) + programmes
(Jki B J +at: Jkji Bij Jki )
T T ij kji T
+ j=1 (Bij Jkji + Jkji Bij) + j=1 (Jki Bij Jkji + Jkji Bij Jki )
j=i
j=1 j=i
j=1
j=m i m j=i
+ m m (J T B J + J T
Blj Jkli ), i = 1, 2, . . . , m;
kli
T lj kji kji
T
+ l=1 j=1 (Jkli Blj Jkji + Jkji Blj Jkli ), i = 1, 2, . . . , m;
l=i j=1
l=1 j=i
l=i j=i
T T T
Cij = Bii Jkij + Jkij T
Bii + Jki B J + Jkij
T ii kij T
Bii + Jki
m C ij = B J
ii kij + J B
kij ii + J B J
ki ii kij + J kij Bii + Jki
+ m (J T B J T T T
ll klj + Jkli Bll Jkli ) + Bij Jkj + Jki B + Jki
T ij
B J
T ij kj
The perfect start
kli
T T
+ l=1 (Jkli Bll Jklj + Jkli Bll Jkli ) + Bij Jkj + Jki Bij + Jki Bij Jkj
l= i,j
l=1
l=i,j
of a successful,
international career.
CLICK HERE
to discover why both socially
and academically the University
of Groningen is one of the best
places for a student to be
www.rug.nl/feb/education
and
Stability Theory of Large-Scale m
Dynamical Systems Impulsive Large-Scale Systems
T T T
Cii = Jki Bii + Bii Jki + Jki Bii Jki + Jkij Bjj Jkji
j=1
j=i
m
m
T T T
+ (Bij Jkji + Jkji Bij) + (Jki Bij Jkji + Jkji Bij Jki )
j=1 j=1
j=i j=i
m
m
T T
+ (Jkli Blj Jkji + Jkji Blj Jkli ), i = 1, 2, . . . , m;
l=1 j=1
l=i j=i
T T T
Cij = Bii Jkij + Jkij Bii + Jki Bii Jkij + Jkij Bii + Jki
m
T T T T
+ (Jkli Bll Jklj + Jkli Bll Jkli ) + Bij Jkj + Jki Bij + Jki Bij Jkj
l=1 4.5 LINEAR SYSTEMS 145
l=i,j
m
T T T
+ (Bil Jklj + Jkli Blj + Jki Bil Jklj + Jkli Blj Jkj )
l=1
l=i,j
m
m
T
+ Jkli Blr Jkrj , i = j, i, j = 1, 2, . . . , m;
l=1 l=1
l=r
m
m
v(x + Jk (x), ) v(x, ) = Uii xi + Jki xi + Jkij xj
i=1 j=1
j=i
m
m m
m
+2 Uij xi + Jki xi + Jkil xl , xj + Jkj xj + Jkjl xl
i=1 j=1 l=1 l=1
j>i l=i l=i
m
m
m
Uii (xi ) 2 Uij (xi , xj )
i=1 i=1 j=1
j>i
m
m
T m
= xi + Jki xi + Jkij xj Bii xi + Jki xi + Jkij xj
i=1 j=1 j=1
j=i j=i
m
m m
T m
+2 xi + Jki xi + Jkil xl Bij xj + Jkj xj + Jkjl xl
i=1 j=1 l=1 l=1
j>i l=i l=j
m
m
m
xT
i Bii xi 2 xT
i Bij xj
i=1 i=1 j=1
j>i
m
m
= xT
i
T T
Bii Jki + Jki Bii + Jki Bii Jki + T
Jkji Bii Jkji xi
i=1 j=1
j=i
m
m
+2 xT T T T
i Bii Jkij + Jkij Bii + Jki Bii Jkij + Jkij Bii Jki
i=1 j=1
142
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
146 4. IMPULSIVE LARGE-SCALE SYSTEMS
m m
m
T T T T
+ Jkli Bll Jklj + Jklj Bll Jkli xj + xi Bij + Jkij + Jkji Bij
l=1 i=1 j=1
l=i,j j=i
m m
m
T T T T
+ Jki Bij Jkji + Jkji Bij Jki + Jkli Blj Jkji + Jkji Blj Jkli xi
j=1 l=1 j=1
j=i l=i j=i
m
m
+2 xT
i
T
Bij Jkj + Jki T
Bij + Jki Bij Jkj
i=1 j=1
j>i
m m
m
T T T
+ Bil Jklj + Jkli Blj + Jki Bil Jklj + Jkli Blj Jkj + JliTBlr Jrj xj
l=1 l=1 r=1
l=i,j l=r
m
m
m
= xT
i Cii xi + 2 xT
i Cij xj
i=1 i=1 j=1
j>i
m
m
m
1/2
M (Cii )xi 2 + 2 T
M (Cij Cij )xi xj = uT
k Cuk ,
i=1 i=1 j=1
j>i
k = 1, 2, . . . .
Inequality (4.5.13) is proved in the same way.
Corollary 4.5.1 If all conditions of Proposition 4.5.3 are satisfied, then
for the function (4.5.5) for t = k (x), k = 1, 2, . . . , the following estimates
hold true
(4.5.14) v(x + Jk (x), ) v(x, ) v(x, )
where
M (C)1
M (B) for M (C) < 0;
=
M (C)1
m (A) for M (C) > 0;
and
(4.5.15) v(x + Jk (x), ) v(x, ),
where
M (C )1
M (B) for M (C ) < 0,
=
M (C )1
m (A) for M (C ) > 0.
The assertions (4.5.14) and (4.5.15) follow from Proposition 4.5.3 and
the inequality (4.5.11). 4.5 LINEAR SYSTEMS 147
is true, where
G(S) = [ ij (S)], ij = ji , i, j = 1, 2, . . . , m,
ii = 1i + 2i (S), i = 1, 2, . . . , m, SS
1
ij = ( (S) + 1ji (S) + 2ij (S) + 2ji (S) + 3ij (S) + 3ji (S)),
2 1ij
i, j = 1, 2, . . . , m, i = j.
Qi = i2 (Bii Ai + AT
i Bii ), i = 1, 2, . . . , m; 143
i1
T
Pi = i j (Bji Sij Aji + (Sji Aji )TBji )
j=1
G(S) = [ ij (S)], ij = ji , i, j = 1, 2, . . . , m,
ii = 1i + 2i (S), i = 1, 2, . . . , m, S S
Stability Theory
1 of Large-Scale
Dynamical (1ij (S) + 1ji (S) + 2ij (S) + 2ji (S) + 3ij (S) + 3ji (S)),
ij = Systems Impulsive Large-Scale Systems
2
i, j = 1, 2, . . . , m, i = j.
Qi = i2 (Bii Ai + AT
i Bii ), i = 1, 2, . . . , m;
i1
T
Pi = i j (Bji Sij Aji + (Sji Aji )TBji )
j=1
m
+ i j (Bij Sji + (Sji Aji )TBij ), i = j = 1, 2, . . . , m,
j=i+1
American online
LIGS University
is currently enrolling in the
Interactive Online BBA, MBA, MSc,
DBA and PhD programs:
and
where
and uT, Jk (x), cij , cij , cii , cii are defined as in Proposition 4.5.3.
and
Theorem 4.5.1 Let the system (4.5.1) be such that the matrix-valued
function (4.5.2) is constructed with the elements (4.5.3) and (4.5.4) and
(1) the matrix A in (4.5.6) is positive definite, i.e. m (A) > 0;
(2) there exists a matrix Q such that for the matrix G(S) the estimate
1 T
(G(S) + G (S)) Q for all S S
2
is satisfied component-wise;
(3) the matrix Q is negative semi-definite or equal to zero, i.e. the in-
equality m (Q) 0 holds.
Then the zero solution of the system (4.5.1) is stable in the whole on S.
If instead of the condition (3) the following condition is satisfied
(4) the matrix C in (4.5.12) is negative definite, i.e. M (C) < 0,
then the zero solution of the system (4.5.1) is asymptotically stable in the
whole on S.
The proof of Theorem 4.5.1 is similar to that of Theorem 4.4.1.
Download free eBooks at bookboon.com
Theorem 4.5.2 Let the system (4.5.1) be such that the matrix-valued
145 and (4.5.4) and
function (4.5.2) is constructed with the elements (4.5.3)
(1) the matrix A in (4.5.6) is positive definite, i.e. m (A) > 0;
(2) there exists a negative definite matrix Q Rmm such that
1 T
(G(S) + G (S)) Q for all S S;
Then the zero solution of the system (4.5.1) is stable in the whole on S.
If instead of the condition (3) the following condition is satisfied
(4) the matrix C in (4.5.12) is negative definite, i.e. M (C) < 0,
Stability
then the Theory of Large-Scale
zero solution of the system (4.5.1) is asymptotically stable in the
Dynamical
whole on S. Systems Impulsive Large-Scale Systems
Theorem 4.5.3 Let the system (4.5.1) be such that the matrix-valued
function (4.5.2) is constructed with the elements (4.5.3) and (4.5.4) and
(1) the matrix A in (4.5.6) is positive definite, i.e. m (A) > 0;
(2) there exists a matrix Q+ Rmm for which
(a) G(S) Q+ for all S S
(b) M (Q+ ) > 0;
(3) M (C ) > 0;
(4) the functions k (x) satisfy for some 1 > 0 and for all k = 1, 2, . . .
the inequality
k (x) k1 (x) 1 .
Theorem 4.5.5 Let the system (4.5.1) be such that the matrix-valued
function (4.5.2) is constructed with the elements (4.5.3) and (4.5.4) and
(1) the matrix A in (4.5.6) is positive definite, i.e. m (A) > 0;
(2) there exists a matrix Q Rmm such that
1
(a) (G(S) + GT(S)) Q for all S S;
2
(b) m (Q ) < 0;
(3) the matrix C is positive definite, i.e. m (C ) > 0;
(4) for some constant 1 > 0 the values k (x), k = 1, 2, . . . , satisfy
the inequality
k (x) k1 (x) 1 , for all k = 1, 2, . . . .
Example 4.5.1 Let the system (4.5.1) be a fourth order system decom-
posed into two subsystems of the second order which are defined by the
matrices:
1 1 2 1
A1 = , A2 = , A12 = A21 = I2 ;
1 1 1 2
(4.5.21) Jki = diag {1, 1}, i = 1, 2; Jk12 = Jk21 = 101 I2 ,
S = {S : S = diag {S1 , S2 }, Si = [Si1 , Si2 ],
Sii = I2 , Sij = sij I2 , 0 sij 1, i = j, i, j = 1, 2},
vii (xi ) = xT
i I2 xi , i = 1, 2;
v12 (x1 , x2 ) = v21 (x1 , x2 ) = xT
1 10
1
I2 x2 .
For T = (1, 1) R+
2
the matrices
1 0.1 1 0.1
A= , B=
0.1 1 0.1 1
For this choice of the elements of the matrix-valued function U (x) we have
For T = (1, 1) R+
2
the matrices
1 0.1 1 0.1
A= , B=
0.1 1 0.1 1
For this choice of the elements of the matrix-valued function U (x) we have
The matrices
1.8 2.1
G(S) Q =
2.1 3.8
and
0.968 0.099
C=
0.099 0.968
are negative definite which is confirmed by the estimate
M (Q) = 0.474 < 0; M (C) = 0.867 < 0.
Thus, all conditions of Theorem 4.5.1 are satisfied and the zero solution
of the system (4.5.1) with matrices (4.5.21) is structurally asymptotically
stable in the whole on S.
This section deals with two problems. In the first problem we establish
conditions under which the stability of solutions with respect to two mea-
sures in the continuous system under nonclassical structural perturbations
implies the same type of stability of solutions to the impulsive system under
nonclassical structural perturbations. In the second problem we establish
sufficient stability conditions for the system of Lurie-Postnikov type in the
presence of impulsive and nonclassical structural perturbations.
Further the functions of the class SL0 are applied (see Definition 1.4.3)
which were constructed in terms of matrix-valued functions U (t, x).
For the function (4.6.3) for (s, x) G0 , G0 = G0k , Gk0 = {(t, x)
k=1
R+ Rn : k1 (x) < t < k (x)}, t0 s t, we define the expression
Here
Maastricht
University is
the best specialist
university in the
Visit us and find out why we are the best! Netherlands
(Elsevier)
Masters Open Day: 22 February 2014
www.mastersopenday.nl
Here
y(t, s, x) = x,
4.6 GENERALIZATIONS AND APPLICATIONS 155
y(t, s + , x + Q(s, x, P, S)) x + Q(s, x, P, S),
and the expression (4.6.4) becomes
(4.6.5) D+ v(s, x, ) = TD+ U (s, x),
where
D+ U (s, x)
= lim sup {[U (s + , x + Q(s, x, P, S)) U (s, x)] 1 , 0+ }.
In the expressions (4.6.4) and (4.6.5) the total derivative of the function
(4.6.3) is computed element-wise.
Further we designate
151
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
156 4. IMPULSIVE LARGE-SCALE SYSTEMS
for any t > t0 , provided that s [t0 , t) and (s, x) S(, ) G0 . Here
M () is the maximal eigenvalue of the matrix B , M (B ) 0, and
1/2
uT = (u1 (y), . . . , u1/2
m (y)).
of the assertion made is not associated with the presence of the impulsive
perturbations of the system (4.6.1) and is carried out in a standard manner
(see, e.g. Lakshmikantham, Leela, et al. [1]). Therefore, this case is omitted
here. Assume that t (k , k+1 ] for some k, and moreover, for at least
1/2
uT = (u1 (y), . . . , u1/2
m (y)).
Stability
Then theTheory of Large-Scale
conditions (4.6.7) (4.6.13) yield
Dynamical Systems Impulsive Large-Scale Systems
(4.6.14) a((t0 , x0 )) v(t0 , x0 , ) b(t0 , (t0 , x0 )) < a().
if t [0 , k ].
According to the choice of (0, ) and by estimates (4.6.15) we have
(4.6.12) yield
On the other hand, by conditions (4.6.7) and (4.6.17) we have the in-
equalities
v( t, x( t), ) a(( t, x( t ))) = a(),
0
for all (s, x(s)) 4.6
GGENERALIZATIONS
and for all (P, S)AND
P APPLICATIONS
S. 159
v(t0 , y(t; t0 , x0 ), ) b(t0 , (t0 , y(t; t0 , x0 ))) b(t0 , (t, y(t; t0 , x0 ))).
154
This proves the asymptotic (0 , )stability of solutions to system (4.6.1)
for all (P, S) P S.
Actually, when the function (4.6.3) constructed in terms of the matrix-
valued function U (t, y) is applied, the condition (6) of Theorem 4.6.1 be-
For the sufficiently large t t0
v(t0 , y(t; t0 , x0 ), ) b(t0 , (t0 , y(t; t0 , x0 ))) b(t0 , (t, y(t; t0 , x0 ))).
Stability Theory of Large-Scale
Dynamical Systems
Hence it follows that Impulsive Large-Scale Systems
lim v(t, x(t), ) = 0
t
m
where k (Ck ) 0, k (Ck ) = , C(R+ , R+ ), (0) = 0, (s) > 0
k=1
for s > 0, k (Ck ) are the maximal eigenvalues of some matrices Ck , k =
1, 2, . . . . Then Theorem 4.6.1 is developed as follows.
Theorem 4.6.2 Assume that conditions (1) (5) and (7) of Theorem
4.6.1 and condition (6*) are satisfied. Then (0 , )stability of system
(4.6.2) implies asymptotic (0 , )stability of system (4.6.1).
The proof of this theorem is similar to that of Theorem 4.6.1.
Note, that for the impulsive system under nonclassical structural pertur-
bations it is reasonable to consider the set of measures discussed in Section
160
2.6.1. 4. IMPULSIVE LARGE-SCALE SYSTEMS
(1) m (2) m
dxi
= Si Ai x + Si qi fi (i ),
dt
=1 =1
i = cT
i x, i = 1, 2, . . . , m, t = k (x), k = 1, 2, . . .
(4.6.20) m m
xi = Jki x + bi gi (i ),
=1 =1
i = cT
i x(k (x)), i = 1, 2, . . . , m, t = k (x), k = 1, 2, . . .
where
1
i fi (i ) [0, Ki ] R+ ,
1
(i
) gi (i i ] R+ ,
) [0, K
dxi
ii ), t = k (xi )
= Aii xi + qii fii (
(4.6.21) dt
xi = Jkii xi + bii gii (
ii ), t = k (xi ),
Download free eBooks at bookboon.com
where 155
ii = cT
ii xi ,
ii ciiT xi (k (xi )),
= i = 1, 2, . . . , m.
In order to simplify system (4.6.20) we introduce the designations
to system (4.2.4) are obtained by replacing x in (4.2.4) with xi , where
xi = (0, . . . , 0, xT T ni
i , 0, . . . , 0) R :
where
ii = cT
ii xi ,
ii ciiT xi (k (xi )),
= i = 1, 2, . . . , m.
In order to simplify system (4.6.20) we introduce the designations
= cTAND
4.6 GENERALIZATIONS x , APPLICATIONS
i i i
161
(b)
dissertation?
1/2
for all xi Rni, i = 1, 2, . . . , m;
1/2
M (Pij PijT)xi xj uij (xi , xj ) M (Pij PijT)xi xj
for all xi Rni, xj Rnj, for all (i = j) = 1, 2, . . . m,
Get in-depth feedback & advice from experts in your
where m (Pii ) are the minimal and M (Pii ) are the maximal eigenvalues
topic area. Find out1/2what you can do to improve
of the matrices Pii , and M (Pij PijT) is the norm of the matrices Pij .
Usingthe
thequality of your dissertation!
matrix-valued function (4.6.23) and the constant vector =
m
(1, 1, . . . , 1) R+ we construct the function
(4.6.26) v(x, ) = T U (x)
and considerGet Help
its total Now
derivative
(4.6.27) Dv(x, ) = TDU (x),
where
DU (x) = [Duij (xi , xj )], i, j = 1, 2, . . . , m,
along the solutions of system (4.2.4).
Proposition 4.6.1 If the estimates (4.6.25) are satisfied, then for the
function (4.6.26) the two-sided inequality
uT = (x1 , x2 , . . . , xm ),
A = [ij ], B = ij , i, j = 1, 2, . . . , m,
ii = m (Pii ), ii = M (Pii )
1/2
ij = ji = ij = ji = M (Pij PijT).
Proof The proof of Proposition 4.6.1 follows from Djordjevic [1] (see
also Martynyuk and Stavroulakis [1]).
Corollary 4.6.5 If inequality (4.6.28) is satisfied, then
and for
m (A) > 0, M (B) > 0,
(4.6.30) 2
1
M (B)v(x, ) u 1
m (A)v(x, ).
Pii Aii + AT i T i T T
ii Pii + Pii qii kii (cii ) + (qii kii (cii ) ) Pii ;
i1
T T
(1) (2) i T
Si Ai + Si qi ki (ci ) Pi
=1
(1) (2) i T
+ PiT Si Ai + Si qi ki (ci )
m
(1) (2) i T
+ Pi Si Ai + Si qi ki (ci )
=i+1
T T
(1) (2) i T
+ Si Ai + Si qi ki (ci ) PiT
m
T
(2) i T (2) i T
+ Pij Sji qji kji (cji ) + Sji qji kji (cji ) Pij
j=2
j>i
j1
T T
(1) (2) i T
Si Ai + Si qi ki (ci ) Pj
=1
m
T T
(1) (2) i T
+ Si Ai + Si qi ki (ci ) Pj
=j+1
j1
(1) (2)
+ PiT Sj Aj + Sj qj kj
(cjj )T
=1
m
(1) (2)
+ Pi Sj Aj +
Sj qj kj (cjj )T
=i+1
T
1 (1) (1) (2)
+ Pii Sij Aij + Sij Aij Pii + Pii Sij qij kij
(cjij )T
2
T
(2) j T j T j T
+ Sij qij kij (cij ) Pii + Pii qii kii (cii ) + qii kij (cii ) Pii
158
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
164 4. IMPULSIVE LARGE-SCALE SYSTEMS
T
1 (1) (1) (2)
+ Pji Sji Aji + Sji Aji Pji + Pjj Sji qji kji
(ciji )T
2
T
(2)
+ Sji qji kji (ciji )T Pjj + Pjj qjj kjj
(cijj )T
i T T
+ qjj kjj (cjj ) Pjj , i = 1, 2, . . . , m, j = 1, 2, . . . , m,
respectively.
Here
T
(2)
kij for ij Sij qij Pij xj > 0, i, j = 1, 2, . . . , m,
kij =
0 in other cases;
T
kij for ii qii Pii xi > 0, i = 1, 2, . . . , m,
kij = T
kij for ii qii Pii xi < 0, i = 1, 2, . . . , m.
The proof is carried out in the same way as that of Proposition 4.6.1.
Proposition 4.6.3 If all conditions of Proposition 4.6.2 are satisfied,
then for (4.6.27)
(4.6.31) Dv(x, ) uTu, for all (x, S) Rn S
where
(1) (2)
= [ij ], i, j = 1, 2, . . . , m, ii = i + i (S ),
ij = ji = ij (S ), S S, i = j, i, j = 1, 2, . . . , m,
is the constant matrix such that
(2) (2)
i (S) i (S )
and
ij (S) < ij (S ).
Brain power
The proof of this Proposition is similar to that of Lemma 3 inelectricity
and Stavroulakis [ 1 ] (see also Miladzhanov [3]).
Martynyuk
By 2020, wind could provide one-tenth of our planets
needs. Already today, SKFs innovative know-
how is crucial to running a large proportion of the
worlds wind turbines.
Corollary 4.6.6 If inequalities (4.6.29) and (4.6.30) are satisfied,
Up to 25then
% of the generating costs relate to mainte-
for (4.6.31) nance. These can be reduced dramatically thanks to our
systems for on-line condition monitoring and automatic
m ()1 M (B)v(x, ) for M () < 0,
(4.6.32) Dv(x, ) lubrication. We help make it more economical to create
m ()m (A)v(x, ) for M ()cleaner,
1
> 0. cheaper energy out of thin air.
The proof follows from Proposition 4.6.5 and Corollary 4.6.5. By sharing our experience, expertise, and creativity,
industries can boost performance beyond expectations.
Therefore we need the best employees who can
meet this challenge!
uT
k = (x1 (k (x)), x2 (k (x)), . . . , xm (k (x)));
m
m
1
Jk (x) = Jki x + bi gi (i );
=1 =1
= [ij ], ij = ji , i, j = 1, 2, . . . , m;
= [ij ], ij = ji , i, j = 1, 2, . . . , m;
1/2
ii = M (ii ), ij = M (ij T
ij ), i = j, i, j = 1, 2, . . . , m;
1/2
ii = M (ii ), ij = M (ij T
ij ), i = j, i, j = 1, 2, . . . , m;
m
m
T
ii = Pii Jkii + Jkii Pii + T
Jkji Pjj Jkji + Pii bi i T
ki (
ci )
j=1 =1
m T m
m
i T j T
+ bi ki (
ci ) Pii + Jkji Pjj bj kj (
cj )
=1 j=1 =1
m T
j T
+ bj kj (
cj ) Pjj Jkji
=1
m
m T
m
+ bj
kj cjj )T Pjj
( bj
kj cjj )T
(
j=1 =1 =1
m
m
+ Pij Jkji + T
Jkji Pij + Pij
bj kj (i T
cj )
j=1 =1
m T m
+
bj kj (i T
cj ) Pij + T T
Jki Pj Jkji + Jkji Pj Jki
=1 =1
m
m T
T
+ Jkji Pij bj
kj cij )T +
( bj
kj cij )T Pij Jkji
(
=1 =1
160
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
166 4. IMPULSIVE LARGE-SCALE SYSTEMS
m T m
+ bi i T
ki ci )
( Pij bj
kj cij )T
(
=1 =1
m T m
+
bj kj (i T
cj ) Pij i T
bi ki (
ci ) , i = 1, 2, . . . , m;
=1 =1
m
T
ij = Pii Jkij + Jkji Pjj + Pii bi i T
ki (
ci )
=1
m T m
+ bj
kj cjj )T Pjj +
( T
(Jki T
P Jkj + Jkj P Jki )
=1 =1
m
m
m T
+ T
Jkri Prr r T
cr ) +
br kr (
br kr (r T
cr ) Prr Jkri
r=1 =1 =1
m
m
+ (Pi Jkj T
+ Jkj Bj ) + Pij bj
kj cjj )T
(
=1 =1
m T
m
i T j T
+ bi ki (
ci ) Pii + bi ki (
ci )
=1 =1
m m
m
+ bi i T
ki (
ci ) Pij + T
Jkj T
Pr Jkrj
=1 =1 r=1
r=
m
m T
m T
+ T
Jkrj Prj bj
kj cjj )T +
( bi i T
ki (
ci ) Pir Jkrj
r=1 =1 =1
r=i,j
m T m
i T j T
+ bi ki (
ci ) Pij bj kj (
cj ) , i = j, i, j = 1, 2, . . . , m;
=1 =1
ii = Pii + ii , ij = Pij + ij , i = j = 1, 2, . . . , m.
Here
kij if the corresponding multiplier is positive;
kij =
0 in other cases.
where
M ()1
M (B) for M () < 0,
=
M ()1
m (A) for M () > 0,
and
where
M ( )1
M (B) for M ( ) < 0,
= at bookboon.com
Download free eBooks
M ( )1
m (A) for M ( ) > 0,
161
hold true.
The proof follows from Proposition 4.6.4 and Corollary 4.6.6.
For system (4.6.1), the following stability problem is formulated. It is
necessary to formulate conditions related to the coefficients which appear
M
=
M ()1
m (A) for M () > 0,
and
Stability Theory of Large-Scale
Dynamical Systems Impulsive Large-Scale Systems
(4.6.41) v(x + Jk (x), ) v(x, ),
where
M ( )1
M (B) for M ( ) < 0,
=
M ( )1
m (A) for M ( ) > 0,
hold true.
The proof follows from Proposition 4.6.4 and Corollary 4.6.6.
For system (4.6.1), the following stability problem is formulated. It is
necessary to formulate conditions related to the coefficients which appear
in the system and also to introduce structural perturbation, such that the
trivial solution of system (4.6.1) is asymptotically stable in the whole on S
for an arbitrary function f of the class under consideration.
We shall introduce the following notions.
Theorem 4.6.4 Let system (4.6.1) be such that the matrix-valued func-
tion (4.6.23) is constructed with the elements (4.6.24) and
(1) the matrix A in (4.6.28) is positive definite, i.e. m (A) > 0;
(2) the matrix in (4.6.32) is negative semi-definite or equals to zero,
i.e. M () 0;
(3) the matrix in (4.6.33) is negative definite.
By Theorem 1.4.4 from Chapter 1 for (a) (c) the zero solution of system
(4.6.1) is asymptotically stable in the whole on S. Since here Nix = Rni,
i = 1, 2, . . . , m, and Nx = N1x . . . Nmx = Rn .
Theorem 4.6.5 Let system (4.6.1) be such that the matrix-valued func-
tion (4.6.23) is constructed with the elements (4.6.24) and
(1) the matrix A in (4.6.28) is positive definite, i.e. m (A) > 0;
(2) the matrix in (4.6.31) is negative definite, i.e. M () < 0;
(3) the matrix in (4.6.34) is positive definite, i.e. M ( ) > 0;
(4) the function k (x), k = 1, 2, . . . , satisfies the inequality
sup minn k+1 (x) maxn k (x) = > 0.
k xR xR
m (A) m (A)
ln 1 + ,
M () m ( )
For this example, the elements of the matrix-valued function (4.6.23) are
taken in the form
uii (xi ) = xT
i I2 xi , i = 1, 2;
u12 (x1 , x2 ) = u21 (x1 , x2 ) = xT
1 0, 1 I2 x2 ,
Section 4.2 Dynamics of the impulsive systems for the given model of nonclassi-
cal structural perturbations was first studied by Miladzhanov [3, 6]. These inves-
tigations were further developed and refined by Martynyuk and Miladzhanov [6]
(see also Martynyuk and Stavroulakis [1 3]).
Sections 4.4 4.5 These sections are based on the results by Martynyuk and
Miladzhanov [6], Martynyuk and Stavroulakis [1 3] and Miladzhanov [3, 6]. In
the investigation of linear impulsive systems some results are used which were
obtained in Chapter 2 for linear continuous systems under nonclassical structural
4.7 NOTES AND REFERENCES 171
perturbations.
Section 4.6 The evolution analysis of the impulsive system with respect to two
or more different measures (see Leela [1], Lakshmikantham and Liu [1], Mar-
tynyuk [11], etc.) is a possible direction of generalization of the method of
matrix-valued Liapunov functions for the system with impulsive and nonclassical
structural perturbations. The results of Section 4.6.1 are new (see Martynyuk
and Chernetskaya [1], and cf. Kou, et al. [1]). In Section 4.6.2 we use general
results of this Chapter and the results by Martynyuk and Miladzhanov [6], and
Martynyuk and Stavroulakis [1].
5
SINGULARLY PERTURBED LARGE-SCALE SYSTEMS
5.1 Introduction
describe the dynamics of the i -th isolated subsystem Fi0 of system F , and
the equations
dyi
(5.2.4) = gi (, bi , y i , 0, Pq+i , Sq+i )
dti
i = 1, 2, . . . , r.
H4 . Dynamics of the whole system F is described by the equations
dxi
= fi (t, x, y, Pi , Si ), i = 1, 2, . . . , q,
dt
(5.2.5)
dyi
i = gi (t, x, y, M, Pq+i , Sq+i ), i = 1, 2, . . . , r,
dt
q
r
where xi Rni, ni = n, yi Rmi, mi = m, q + r = s, the
i=1 i=1
parametric perturbations Pi , i = 1, 2, . . . , q, and the structural matrices
Si , i = 1, 2, . . . , s, are determined in the same way as in Section 2.2. Here
i (0, 1] and the set of all admissible values of M is designated as
Moreover
Mm = {M : 0 < i < im , i [1, r]},
where im is an admissible upper value of i .
If in the system of equations (5.2.5) all i (formally) form a zero set,
then the equations
dxi
= fi (t, x, y, Pi , Si ), i = 1, 2, . . . , q,
(5.2.6) dt
0 = gi (t, x, y, 0, Pq+i , Sq+i ), i = 1, 2, . . . , r,
dyi
(5.2.7) = i gi (, b, y, 0, Pq+i , Sq+i ), i = 1, 2, . . . , r,
dti
Download free eBooks at bookboon.com
167
If in the system of equations (5.2.5) all i (formally) form a zero set,
then the equations
Stability Theorydxofi Large-Scale
Dynamical Systems = fi (t, x, y, Pi , Si ), i = 1, 2, . . . , q, Singularly Perturbed Large-Scale Systems
(5.2.6) dt
0 = gi (t, x, y, 0, Pq+i , Sq+i ), i = 1, 2, . . . , r,
dyi
(5.2.7) = i gi (, b, y, 0, Pq+i , Sq+i ), i = 1, 2, . . . , r,
176 dti 5. SINGULARLY PERTURBED SYSTEMS
t t0
(5.2.8) ti = , i = 1, 2, . . . , r.
i
ti
(5.2.9) = i , i = 1, 2, . . . , r,
t1
(5.2.10) i [ i , u ], i = 1, 2, . . . , r,
dxi
(5.2.12) = fi (t, xi , 0, Pi , Si ),
dt
dxi
(5.2.13) = fi (t, x, 0, Pi , Si ), i = 1, 2, . . . , q,
dt
respectively.
dxi
(5.2.12) = fi (t, xi , 0, Pi , Si ),
dt
dxi
(5.2.13) = fi (t, x, 0, Pi , Si ), i = 1, 2, . . . , q,
dt
respectively. 5.3 STABILITY ANALYSIS 177
In the qualitative analysis for the given class of large scale systems the
question whether different time scales ti are interconnected is of importance.
General purpose of our investigation is to obtain conditions under which
stability of zero solution of the initial system is implied by stability of
some independent degenerated subsystems and stability of the independent
subsystems describing boundary layer with allowance for the qualitative
properties of interconnections between the subsystems.
where
such that
(a) ij i1 (xi )j1 (xj ) vij (t, ) ij i2 (xi )j2 (xj ),
(t, xi , xj ) R Nix Njx , i, j = 1, 2, . . . , q, j i;
(b) q+i,q+j i1 (yi )j1 (yj ) vq+i,q+j (t, ) q+i,q+j i2 (yi )j2 (yj ),
(t, yi , yj ) R Niy Njy , i, j = 1, 2, . . . , q, j i;
(c) i,q+j i1 (xi )j1 (yj ) vi,q+j (t, ) i,q+j i2 (xi )j2 (yj ),
(t, xi , yj ) R Nix Njy , i, j = 1, 2, . . . , q.
s
By means of the matrixfunction (5.3.2) and the constant vector R+
we introduce the function
D+ v(t, x, y, M ) = T D+ U (t, x, y, M ),
(5.3.4) def
D+ U (t, x, y, M ) = [D+ vrk (t, . . . )], r, k = 1, 2, . . . , s.
uT T
1 A(M )u1 v(t, x, y, M ) u2 B(M )u2 ,
(t, x, y, M ) R Nx Ny M,
is satisfied, where
170
Stability Theory of Large-Scale
Dynamical Systems Singularly Perturbed Large-Scale Systems
5.3 STABILITY ANALYSIS 179
uT
k = (1k (x1 ), . . . , qk (xq ), 1k (y1 ), . . . qk (yq )), k = 1, 2,
A(M ) = H TA1 (M )H, B(M ) = H TA2 (M )H,
H = diag {1 , 2 , . . . , s }, s = 2q,
A11 A12 (M ) A11 A12 (M )
A1 (M ) = T , A 2 (M ) = T ,
A12 (M ) A22 (M ) A12 (M ) A22 (M )
A11 = [ij ], ij = ji , A11 = [ij ], ij = ji ,
A12 (M ) = [j i,q+j ], A12 (M ) = [j i,q+j ],
A22 (M ) = [ij q+i,q+j ] q+i,q+j = q+j,q+i ,
A22 (M ) = [ij ]q+i,q+j , q+i,q+j = q+j,q+i ,
i for i = j,
ij = i, j = 1, 2, . . . , q.
i j for i = j,
q
q q
+2 i q+j j vi,q+j (t, xi , yj ) i2 ii 2i1
i=1 j=1 i=1
q
q q
2 2
+2 i j ij i1 (xi )j1 (xj ) + q+i i q+i,q+j i1 (yi )
i=1 j=2 i=1
j>i
q
q
+2 q+i q+j i j i1 (yi )j1 (yj )
Challenge the way we run
q+i,q+j
i=1 j=2
j>i
q
q
+2 i q+j j i,q+j i1 (xi )j1 (yj )
i=1 j=1
RUN FASTER.
RUN LONGER.. READ MORE & PRE-ORDER TODAY
RUN EASIER WWW.GAITEYE.COM
q
q q
+2 i q+j j vi,q+j (t, xi , yj ) i2 ii 2i1
i=1 j=1 i=1
q
q q
2 2
+2 i j ij i1 (xi )j1 (xj ) + q+i i q+i,q+j i1 (yi )
i=1 j=2 i=1
j>i
q
q
+2 q+i q+j i j q+i,q+j i1 (yi )j1 (yj )
i=1 j=2
j>i
q
q
+2 i q+j j i,q+j i1 (xi )j1 (yj )
180 5. SINGULARLY
i=1 j=1
PERTURBED SYSTEMS
172
Stability Theory of Large-Scale
Dynamical Systems Singularly Perturbed Large-Scale Systems
5.3 STABILITY ANALYSIS 181
q
+ 2i q+i i (Dx+i vi,q+i )T fi + (Dy+i vi,q+i )T gi
i=1
q
q
+ i j Dt+ vij + (Dx+i vij )T fi (t, x, y, Pi , Si )
i=1 j=2
j>i
+ (Dx+j vij )T fj (t, x, y, Pj , Sj )
q
q
+2 q+i q+j i j Dt+ vq+i,q+j
i=1 j=2
j>i
is true, where
i = 1, 2, . . . , q; 173
Dynamical Systems
+ 10i (P, S)), i = 1, 2, . . . , q; Singularly Perturbed Large-Scale Systems
q+i,q+i (M, P, S) = 2i (P, S) + 5i (P, S) + 11i (P, S)
q
+ i 6i (P, S) + 12i (P, S) + j 13i (P, S) ,
j=2
j>i
i = 1, 2, . . . , q;
i,q+i (M, P, S) = 7i (P, S) + i 8i (P, S), i = 1, 2, . . . , q;
ij (M, P, S) = 1ij (P, S) + i 2ij , i = 1, 2, . . . , q, j = 2, 3, . . . , q, j > i;
q+i,q+j (M, P, S) = 3ij (P, S) + i 4ij (P, S) + i j 5ij (P, S),
i = 1, 2, . . . , q, j = 2, 3, . . . , q, j > i;
i,q+j (M, P, S) = 6ij (P, S) + i 7ij (P, S) + i j 8ij (P, S),
i, j = 1, 2, . . . , q, i = j.
Proof Let all conditions of Assumption 5.3.2 be satisfied. Then for the
expression (5.3.4) we have
q
D+ v(t,x, y, M ) = i2 Dt+ vii + (Dx+i vii )T fi (t, xi, 0, Pi , Si )
i=1
+ 2
q+i i Dt+ vq+i,q+i 2
+ q+i (Dy+i vq+i,q+i )T gi (, bi, y i, 0, Pq+i , Sq+i )
+ i2 (Dx+i vii )T fi + q+i
2
(Dy+i vq+i,q+i )T gi + 2i q+i i Dt+ vi,q+i
q
q
+2 i q+j j Dt+ vi,q+j + j (Dx+i vi,q+j )T fi (t, x, y, Pi , Si )
i=1 j=1
j=i
+ (Dy+j vi,q+j )T gj (t, x, y, M, Pq+j , Sq+j )
q
1i (P, S) + 3i (P, S) + 9i (P, S) + i (4i (P, S)
i=1
q
2
+ 10i (P, S) i (xi ) + 2i (P, S) + 5i (P, S) + 11,i (P, S)
i=1
q
+ i 6i (P, S) + 12 (P, S) + j 13,i (P, S) i2 (yi )
j=2
j>i
q
Download
free eBooks at bookboon.com
+ {7i (P, S) + i 8i (P, S)}i (xi )i (yi )
i=1 174 Click on the ad to read more
q
q
+2 {1ij (P, S) + i 2ij (P, S)}i (xi )j (xj )
i=1 j=2
j>i
Stability Theory of Large-Scale
Dynamical Systems 5.3 STABILITY ANALYSIS
Singularly
183
Perturbed Large-Scale Systems
+ (Dy+i vi,q+i )T gi (t, xi, y i, M i, Pq+i , Sq+i )
2
+ i2 (Dx+i vii )T fi + q+i (Dy+i vq+i,q+i )T gi
+ 2i q+i i (Dx+i vi,q+i )T fi + (Dy+i vi,q+i )T gi
q
q
+2 i j Dt+ vij + (Dx+i vij )T fi (t, x, y, Pi , Si )
i=1 j=2
j>i
+ (Dx+j vij )T fj (t, x, y, Pj , Sj ) + q+i q+j i j Dt+ vq+i,q+j
q
q
+2 i q+j j Dt+ vi,q+j + j (Dx+i vi,q+j )T fi (t, x, y, Pi , Si )
i=1 j=1
j=i
+ (Dy+j vi,q+j )T gj (t, x, y, M, Pq+j , Sq+j )
q
1i (P, S) + 3i (P, S) + 9i (P, S) + i (4i (P, S)
i=1
q
+ 10i (P, S) 2i (xi ) + 2i (P, S) + 5i (P, S) + 11,i (P, S)
i=1
q
+ i 6i (P, S) + 12 (P, S) + j 13,i (P, S) i2 (yi )
j=2
j>i
q
+ {7i (P, S) + i 8i (P, S)}i (xi )i (yi )
i=1
q
q
+2 {1ij (P, S) + i 2ij (P, S)}i (xi )j (xj )
i=1 j=2
j>i
q
q
+2 {3ij (P, S) + i 4ij (P, S) + i j 5ij (P, S)}i (yi )j (yj )
184 i=1 j=25. SINGULARLY PERTURBED SYSTEMS
j>i
q q
+2 {6ij (P, S) + i 7ij (P, S) + i j 8ij (P, S)}i (xi )j (yj )
i=1 j=1
j=i
= uT G(M, P, S)u.
Theorem 5.3.1 Let the perturbed motion equations (5.2.5) be such that
all conditions of Assumptions 5.3.1 and 5.3.2 are satisfied and
(a) matrix A(M ) is positive definite for any i (0, i1 ) and for
i 0, i = 1, 2, . . . , q;
(b) there exists a matrix G(M ) which is negative definite for any i
i2 ) and for i 0, i = 1, 2, . . . , q, such that for the matrix
(0,
G(M, P, S) determined in Proposition 5.3.2 the estimate
is satisfied.
Download free eBooks at bookboon.com
Then the equilibrium state (xT, y T )T = 0 of system F is uniformly
asymptotically stable for any i (0, i ) and for175i 0 on P S,
where i = min {1,
i1 ,
i2 }.
If, moreover, Ni1 Niy = Rni +mi, the functions ik , ik , i , i are of
class KR, then the equilibrium state (xT, y T )T = 0 of system F is uniformly
asymptotically stable in the whole for any (0, ) and for 0 on
i2 ) and for i 0, i = 1, 2, . . . , q, such that for the matrix
(0,
G(M, P, S) determined in Proposition 5.3.2 the estimate
Stability Theory of Large-Scale
G(M, P, S) G(M ),
Dynamical Systems (M, P, S) M P S. Singularly Perturbed Large-Scale Systems
is satisfied.
Then the equilibrium state (xT, y T )T = 0 of system F is uniformly
asymptotically stable for any i (0, i ) and for i 0 on P S,
where i = min {1,
i1 ,
i2 }.
If, moreover, Ni1 Niy = Rni +mi, the functions ik , ik , i , i are of
class KR, then the equilibrium state (xT, y T )T = 0 of system F is uniformly
asymptotically stable in the whole for any i (0, i ) and for i 0 on
P S.
Proof Under conditions of Assumption 5.3.1, Proposition 5.3.1 and con-
dition (a) of Theorem 5.3.1 the function v(t, x, y, M ) is positive definite for
any i (0, i1 ) and for i 0 it is decreasing on Nix Niy . Con-
ditions of Assumption 5.3.2, Proposition 5.3.2 and condition (b) of Theo-
rem 5.3.1 imply that the expression D+ v(t, x, y, M ) is negative definite for
any i (0, i2 ) and for i 0 for each (P, S) P S.
These conditions are sufficient for uniform asymptotic stability of the
equilibrium state of system (5.2.5) for any i (0, i ) and for i 0
on M P S since all conditions of Theorem 7 from Chapter 1 of the
monograph by Grujic, et al. [1] are satisfied.
In the case when Nix Niy = Rni +mi , the function v(t, x, y, M ) is
positive definite, decreasing and radially unbounded. This fact together
5.3 STABILITY ANALYSIS 185
with the other conditions of the theorem prove the second statement.
Remark 5.3.1 From the condition of matrix A(M ) positive definiteness
and matrix G(M ) negative definiteness the values i1 and
i2 are deter-
i = min {1,
mined respectively, since i2 } is the lower estimate of
i1 ,
the upper bound of the admissible i so that M = {M : 0 < i < i ,
i = 1, 2, . . . , q}.
Example 5.3.1 Consider nonlinear and nonstationary 8-th order system
consisting of two interconnected 4-th order subsystems described by the
equations
dxi
= (1 + sin2 t)(x3i + 0.1 yi3 ) + 0.2si1 (t)yj3 cos2 t,
dt
(5.3.5) dyi
i = (1 + sin2 t)(yi3 + 0.1 i x3i ) + 0.2s2+i,1 (t)x3j cos2 t,
dt
i, j = 1, 2, i = j,
where
Download free eBooks at bookboon.com
176
A11 = diag (1, 1), A22 (M ) = diag (1 , 2 ),
A12 (M ) = diag (0.1 1 , 0.1 2),
where
186positive definite5.for
is SINGULARLY
any i (0,PERTURBED
1] and for SYSTEMS
i 0, i = 1, 2.
ii (M ) = 2 + 0.26 i , i = 1, 2;
2+i,2+i (M ) = 1.8 + 0.06 i , i = 1, 2;
i,2+i (M ) = 0, i = 1, 2; ij (M ) = 2+i,2+j (M ) = 0.01i ,
i,2+j (M ) = 0.2(1 + i ), i, j = 1, 2, i = j.
Moreover, the matrix G(M ) is negative definite for any i (0, 1] and
for i 0, i = 1, 2. Therefore, by Theorem 5.3.1 the equilibrium state
(xT, y T )T = 0 R8 of system (5.3.5) is uniformly asymptotically stable in
the whole on M S, , where x = (xT T T 4 T T T 4
1 , x2 ) R , y = (y1 , y2 ) R ,
S = S : S = diag (S1 , S2 , S3 , S4 ),
1 0 0 0 1 0 1 0
Si , i = 1, 2, 3, 4 .
0 1 0 0 0 1 0 1
5.3.2 Uniform time scaling In the case of uniform time scaling the
www.sylvania.com
system (5.2.5) is represented as
dxi
= fi (t, x, 0, Pi , Si ) + fi , i = 1, 2, . . . , q,
dt
(5.3.6)
1
dyi We do not reinvent
= i gi (, b, y, 0, Pq+i , Sq+i ) + i gi , i = 1, 2, . . . , r,
the wheel we reinvent
dt
where
light.
fi = fi (t, x, y, Pi , Si ) fi (t, x, 0, Pi , Si ),
Fascinating lighting offers an infinite spectrum of
possibilities: Innovative technologies and new
gi = gi (t, x, y, M, Pq+i , Sq+i ) gi (, b, y, 0, Pq+i , Smarkets
q+i ), provide both opportunities and challenges.
An environment in which your expertise is in high
and demand. Enjoy the supportive working atmosphere
within our global group and benefit from international
i [ i , i ], 0 < i < i < +, 1 = 1 = 1 = 1.
career paths. Implement sustainable ideas in close
cooperation with other specialists and contribute to
To study system (5.3.6) we make some assumptions.
influencing our future. Come and join us in reinventing
light every day.
Light is OSRAM
5.3.2 Uniform time scaling In the case of uniform time scaling the
system (5.2.5) is represented as
dxi
= fi (t, x, 0, Pi , Si ) + fi , i = 1, 2, . . . , q,
dt
(5.3.6)
dyi
1 = i gi (, b, y, 0, Pq+i , Sq+i ) + i gi , i = 1, 2, . . . , r,
dt
where
fi = fi (t, x, y, Pi , Si ) fi (t, x, 0, Pi , Si ),
gi = gi (t, x, y, M, Pq+i , Sq+i ) gi (, b, y, 0, Pq+i , Sq+i ),
and
i [ i , i ], 0 < i < i < +, 1 = 1 = 1 = 1.
uT T
1 A(1 )u1 v(t, x, y, 1 ) u2 B(1 )u2 ,
(t, x, y, 1 ) R Nx Ny M,
where
uT
1 = (11 (x1 ), . . . , q1 (xq ), 11 (y1 ), . . . , r1 (yr )),
uT
2 = (12 (x1 ), . . . , q2 (xq ), 12 (y1 ), . . . , r2 (yr )),
m (A11 )m (A22 )
1 = min 1, ,
M (A12 AT12 )
179
Stability Theory of Large-Scale
Dynamical Systems Singularly Perturbed Large-Scale Systems
5.3 STABILITY ANALYSIS 189
thinking
q
+ 1ip (P, S)i (xi )p (xp ),
p=1
p=i
(t, xi , P, S) R Nix0 P S, i = 1, 2, . . . , q;
2
(b) q+j 1 Dt+ vq+j,q+j
2
+ q+j j (Dy+j vq+j,q+j )T gj (, ,y, 0, Pq+j , Sq+j )
r
2
1,q+j (P, S)j (yj ) + 1,q+j,q+l (P, S)j (yj )l (yl ),
l=1
l=j
(t, yj , j , P, S) R Njy0 M P S, j = 1, 2, . . . , r;
360
q
r
(c) i2 (Dx+i vii )T fi + 2
q+j j (Dy+j vq+j,q+j )T gj
.
i=1 j=1
thinking q
q
+2 i p Dt+ vip + (Dx+i vip )T fi (t, x, y, Pi , Si )
i=1 p=2
360
.
p>i
+ (Dx+p vip )T fp (t, x, y, Pp , Sp )
thinking
Discover the truth at www.deloitte.ca/careers Deloitte & Touche LLP and affiliated entities.
r
r
+2 q+j q+l 1 Dt+ vq+j,q+l
j=1 l=2
l>j
+ j (Dy+j vq+j,q+l )T gj (t, x, y, M, Pq+j , Sq+j )
+ l (Dy+l vq+j,q+l )T gl (t, x, y, M, Pq+l , Sq+l )
q r
+2 i q+j i Dt+ vv,q+j +1 (dxvi,q+j )T fi (t, x, y, Pi , Si )
i=1 j=1
+ j (Dy+j vi,q+j )T gj (t, x, y, M, Pq+j , Sq+j )
q
(2i (P, S) + 1 3i (P, S))2i (xi )
i=1
r
+ (2,q+i (P, S) + 1 3,q+j (P, S))j2 (yj )
j=1
q
q
+2 (2ip (P, S) + 1 3ip (P, S))i (xi )p (xp )
i=1 p=2
p>i
r
r
+2 (2,q+j,q+l (P, S) + 1 3,q+j,q+l (P, S))j (yj )l (yl )
j=1 l=2
l>j
q r
+ (1,i,q+j (P, S) + 1 2,i,q+j (P, S))i (xi )j (yj ),
i=1 j=1
(t, xi , yj , M, P, S) R Nix0 Njy0 M P S.
D+ v(t, x, y, 1 ) uT Cu + 1 uT Gu,
(t, x, y, 1 , P, S) R Nx0 Ny0 M P S, j [ j , j ],
holds, where
181
Stability Theory of Large-Scale
Dynamical Systems Singularly Perturbed Large-Scale Systems
5.3 STABILITY ANALYSIS 191
M (C)
1 = min 1, .
M (G)
The proof of Proposition 5.3.6 follows from the analysis of the inequality
i, j = 1, 2; i = j,
where t, xi , yi R, M = {M : 0 < i < 1, i = 1, 2}, M = diag {1 , 2 },
2 = 12 , 2 = 1, so that 2 [ 12 , 1], Sij = Sij (t) [0, 1], i, j = 1, 2.
The elements of the matrixfunction (5.3.10) are taken as follows
Do you like cars? Would you like to be a part of a successful brand? Send us your CV on
We will appreciate and reward both your enthusiasm and talent. www.employerforlife.com
Send us your CV. You will be surprised where it can take you.
Let T = (1, 1, 1, 1), . Then the matrices A11 = A22 = diag {1, 1} are
positive definite and the matrix
A11 1 A12 0.2 0.2
A(1 ) = , where A12 = ,
1 AT
12 1 A22 0.2 0.2
1i = 1, i = 1, 2; 13 = 4; 14 = 1; 2j = 0, j = 1, 2, 3, 4;
31 (S) = 0.01(S31 + S42 ); 32 (S) = 0.01(S32 + S42 );
33 (S) = 0.002S11 + 0.003S22 ; 34 (S) = 0.003S12 + 0.002S21 ;
212 (S) = 0; 312 (S) = 0.01(S31 + S32 + S41 + S42 );
234 (S) = 0; 334 (S) = 0.002(S11 + S21 ) + 0.003(S12 + S22 );
113 (S) = 0.2 + 0.04S11 ; 213 (S) = 0.05 + 0.2S31 ;
114 (S) = 0.1 + 0.06S12 ; 214 (S) = 0.05 + 0.1S42 ;
123 (S) = 0.2 + 0.06S22 ; 223 (S) = 0.05 + 0.1S32 ;
124 (S) = 0.1 + 0.04S21 ; 224 (S) = 0.05 + 0.1S41 .
1 = min {1, 2, 1, . . . } = 1.
holds true, where uT and G(M, P, S) are defined in the same way as in
Proposition 5.3.5.
The proof is similar to that of Proposition 5.3.5.
Theorem 5.4.1 Let the perturbed motion equations (5.2.5) be such that
all conditions of Assumptions 5.3.1 and 5.4.1 are satisfied and
(a) matrices A(M ) and B(M ) are positive definite for any i (0, i )
and for i 0, i = 1, 2, . . . , q, where i = min {i1 , i2 };
(b) there exists a matrix G(M ) which is positive definite for any i
(0, i3 ) and for i 0, i = 1, 2, . . . , q, such that for the matrix
G(M, P, S) defined by Proposition 5.3.7 the extimate
is satisfied.
Then the equilibrium state (xT, y T )T = 0 of system (5.2.5) is unstable
for any i (0, i ) and for i 0 on P S, where
i = min {1, i , i3 }.
D+ v(t, x, y, 1 ) uT Cu + 1 uT Gu,
(t, x, y, 1 , P, S) R Nx0 Ny0 M P S, i [ i , i ],
1
1 = min {1, m (C)m (G)}.
1
1 = min {1 , T
1 , M (A11 ) M (A22 ) M (A12 A12 )},
q
dxi
= Ai xi + (Sil1 Ail xl + Sil2 Ail yl ), i = 1, 2, . . . , q,
dt
l=1
(5.5.1) q
dyi
1 2
i = Bi y i + (i Sq+i,l Bil xl + Sq+i,l Bil yl ), i = 1, 2, . . . , q,
dt
l=1
where Ai , Si , Ail , Ail , Bil and Bil are constant matrices, all matrices and
vectors are of the corresponding order, and Sil1 , Sil2 , Sq+i,l
1 2
and Sq+i,l are
AXA Global
diagonal matrices, i (0, 1], i = 1, 2, . . . , q. Let
1
Si1 1
Si2 ... 1
Graduate
Si,i1 0 S
Program
... S
1 1
i,i+1 iq
S2
i1
2
Si2 ... 2
Si,i1 Find
J out more. .and
2
Si,i+1 . apply
2
Siq
Si = 1 1 1 1 1 ,
Sq+i,1 Sq+i,2 ... Sq+i,i1 J Sq+i,i+1 ... Sq+i,q
2 2 2 2 2
Sq+i,1 Sq+i,2 ... Sq+i,i1 J Sq+i,i+1 ... Sq+i,q
i = 1, 2, . . . , q, S = diag {S1 , S2 , . . . , Sq }.
q
dxi
= Ai xi + (Sil1 Ail xl + Sil2 Ail yl ), i = 1, 2, . . . , q,
dt
l=1
(5.5.1) q
dyi
1 2
i = Bi y i + (i Sq+i,l Bil xl + Sq+i,l Bil yl ), i = 1, 2, . . . , q,
dt
l=1
where Ai , Si , Ail , Ail , Bil and Bil are constant matrices, all matrices and
vectors are of the corresponding order, and Sil1 , Sil2 , Sq+i,l
1 2
and Sq+i,l are
diagonal matrices, i (0, 1], i = 1, 2, . . . , q. Let
1 1 1 1 1
Si1 Si2 ... Si,i1 0 Si,i+1 ... Siq
S2 2
Si2 ... 2
Si,i1 J 2
Si,i+1 ... 2
Siq
i1
Si = 1 1 1 1 1 ,
Sq+i,1 Sq+i,2 ... Sq+i,i1 J Sq+i,i+1 ... Sq+i,q
2 2 2 2 2
Sq+i,1 Sq+i,2 ... Sq+i,i1 J Sq+i,i+1 ... Sq+i,q
i = 1, 2, . . . ,LINEAR
q, S =SYSTEMS
diag {S1 , S2 , . . . , Sq }. 197
k
S = {S : 0 Sjl 2
J, Sii1 = Sq+i,i = 0, Sii2 = Sq+i,i
1
= J,
i, l = 1, 2, . . . , q, j = 1, 2, . . . , 2q, k = 1, 2},
(a) m (Pii )xi 2 vii (xi ) M (Pii )xi 2 , xi Nix0 , i [1, q];
2 2
(b) m (Pq+i,q+i )yi vq+i.q+i (yi ) M (Pq+i,q+i )yi ,
yi Niy0 , i = 1, 2, . . . , q;
1/2 1/2
(c) l,M (Pij PijT )xi xj vij (xi , xj ) M (Pij PijT )xi xj ,
(5.4.4)
(xi , xj ) Nix0 Njx0 , i, j = 1, 2, . . . , q, i = j;
1/2 T
(d) M
Download (PeBooks
free q+i,q+j )P
at q+i,q+j yi yj
bookboon.com vq+i,q+j (yi , yj )
1/2 T
M (Pq+i,q+j )Pq+i,q+j yi yj , (yi , yj )187
Niy0 Njy0 ,
i, j = 1, 2, . . . , q, i = j;
1/2 T
(e) M (Pi,q+j Pi,q+j )xi yj vi,q+j (xi , yj )
1/2
(P PT )x y , (x , y ) N N ,
(a) m (Pii )xi 2 vii (xi ) M (Pii )xi 2 , xi Nix0 , i [1, q];
2 2
(b) m (Pq+i,q+i )yi vq+i.q+i (yi ) M (Pq+i,q+i )yi ,
Stability Theory
yi Niyof
0,
Large-Scale
i = 1, 2, . . . , q;
Dynamical 1/2 Systems T 1/2
Singularly Perturbed Large-Scale Systems
(c) l,M (Pij Pij )xi xj vij (xi , xj ) M (Pij PijT )xi xj ,
(5.4.4)
(xi , xj ) Nix0 Njx0 , i, j = 1, 2, . . . , q, i = j;
1/2 T
(d) M (Pq+i,q+j )Pq+i,q+j yi yj vq+i,q+j (yi , yj )
1/2 T
M (Pq+i,q+j )Pq+i,q+j yi yj , (yi , yj ) Niy0 Njy0 ,
i, j = 1, 2, . . . , q, i = j;
1/2 T
(e) M (Pi,q+j Pi,q+j )xi yj vi,q+j (xi , yj )
1/2 T
M (Pi,q+j Pi,q+j )xi yj , (xi , yj ) Nix0 Njy0 ,
198 5. SINGULARLY PERTURBED SYSTEMS
i, j = 1, 2, . . . , q,
where m (Pii ) and m (Pq+i,q+i ) are minimal eigenvalues, M (Pii ) and
M (Pq+i,q+i ) are maximal eigenvalues of matrices Pii and Pq+i,q+i respec-
1/2 1/2 1/2
T
tively; M (Pij PijT ), M (Pq+i,q+j Pq+i,q+j T
) and M (Pi,q+j Pi,q+j ) are
norms of matrices Pij , Pq+i,q+j and Pi,q+j respectively.
When estimates (5.5.4) are satisfied for function (5.3.3) with elements
(5.5.3) the bilateral inequality
takes place.
Here matrices A(M ) and B(M ) are defined as in Proposition 5.3.1,
uT = (x1 , x2 , . . . , xq , y1, y2 , . . . , yq ),
1/2
ii = m (Pii ), q+i,q+i = m (Pq+i,q+i ), ij = M (Pij PijT ),
1/2 T 1/2 T
q+i,q+j = M (Pq+i,q+j Pq+i,q+j ), i,q+j = M (Pi,q+j Pi,q+j ),
ii = M (Pii ), q+i,q+i = M (Pq+i,q+i ), ij = ij ,
q+i,q+j = q+i,q+j , i,q+j = i,q+j , i, j = 1, 2, . . . , q.
s
Let T = (1, 1, . . . , 1) R+ , , then the expression of total derivative of
function (5.3.3) with elements (5.5.3) is
where
z = (xT T T T T T T
1 , x2 , . . . , xq , y1 , y2 , . . . , yq ) ;
C(S) = [cij (S)], i, j = 1, 2, . . . , s;
G(M, S) = [ij (M, S)], i, j = 1, 2, . . . , s; s = 2q.
i1
cii (S) = Pi Ai + AT
i Pii + PliT (Sli1 Ali )) + (Sli1 Ali )T Pli
l=1
q
+ Pil (Sli1 Ali ) + (Sli1 Ali )T PilT , i = 1, 2, . . . , q;
LINEAR SYSTEMS 199
l=1
i1
cq+i,q+i (S) = Pq+i,q+i Bi + BiT Pq+i,q+i + T
Pq+l,q+i 2
(Sq+l,i Bli )
l=1
q
2
+ (Sq+l,i Bli )T Pq+l,q+i + 2
Pq+i,q+l (Sq+l,i Bli )
l=i
Download free eBooks
2 at
T
bookboon.com
T
+ (Sq+l,i Bli ) Pq+i,q+l , i = 1, 2, . . . , q;
i1
188
cij (S) = cji (S) = Pij Aj + AT
i Pij +
1
PliT (Slj Alj )
l=1
j1
1
+ (Slj Alj )T Plj + 1
Pil (Slj Alj ) + (Sli1 Ali )T Plj
LINEAR SYSTEMS 199
Stability Theory of Large-Scale
Dynamical Systems i1
Singularly Perturbed Large-Scale Systems
T
T 2
cq+i,q+i (S) = Pq+i,q+i Bi + Bi Pq+i,q+i + Pq+l,q+i (Sq+l,i Bli )
l=1
q
2
+ (Sq+l,i Bli )T Pq+l,q+i + 2
Pq+i,q+l (Sq+l,i Bli )
l=i
2
+ (Sq+l,i Bli )T Pq+i,q+l
T
, i = 1, 2, . . . , q;
i1
cij (S) = cji (S) = Pij Aj + AT
i Pij +
1
PliT (Slj Alj )
l=1
j1
1
+ (Slj Alj )T Plj + 1
Pil (Slj Alj ) + (Sli1 Ali )T Plj
l=i
q
1
+ Pil (Slj Alj ) + (Sli1 Ali )T PjlT , i, j = 1, 2, . . . , q, j > i;
l=j
ij (M, S)real
I wanted responsibili
= ji (M, S) =Ijoined
j ij (S), MITAS
i, j = 1, 2,because
. . . , q, j > i; www.discovermitas.com
Maersk.com/Mitas e G
for Engine
(S) =
q
P
I wanted real responsibili
(S 1 B ) + (S 1 B )T P
, j > i = 1, 2, . . . , q;
Ma
ij i,q+l q+l,j lj q+l,j lj i,q+l
l=1
q+i,q+j (M, S) = q+j,q+i (M, S) = i q+i,q+j (S) + j q+i,q+j (S),
i, j = 1, 2, . . . , q, j > i;
Month 16
I was a construction Mo
supervisor ina const
I was
the North Sea super
advising and the No
Real work he
helping foremen advis
International
al opportunities
Internationa
ree wo
work
or placements ssolve problems
Real work he
helping fo
International
Internationaal opportunities
ree wo
work
or placements ssolve pr
Download free eBooks at bookboon.com
where
uT = (x1 , x2 , . . . , xq , y1 , y2 , . . . , yq ),
G(M ) = [cij + ij (M )], i, j = 1, 2, . . . , s, s = 2q.
1/2
ci,q+j = M (ci,q+j (S )cT
i,q+j (S )), i, j = 1, 2, . . . , q;
ii (M ) = i M (ii (S )), i = 1, 2, . . . , q;
q+i,q+i (M ) = i M (q+i,q+i (S )), i = 1, 2, . . . , q;
T 1/2
ij (M )free
Download = eBooks
ji (M ) =
at j M (ij (S )ij (S )),
bookboon.com j > i = 1, 2, . . . , q;
1/2
q+i,q+j (M ) = q+j,q+i (M ) = i M (q+i,q+j (S 190
T
)q+i,q+j (S ))
1/2
+ j M (q+i,q+j (S )q+i,q+j
T
(S )), i, j = 1, 2, . . . , q, j > i;
1/2
i,q+j (M ) = j M (i,q+j (S )i,q+j
T
(S ))
LINEAR SYSTEMS 201
1/2
ci,q+j of
Stability Theory M (ci,q+j (S )cT
= Large-Scale
i,q+j (S )), i, j = 1, 2, . . . , q;
Dynamical Systems ii (M ) = i M ( (S )), i = 1, 2, . . . , q; Singularly Perturbed Large-Scale Systems
ii
q+i,q+i (M ) = i M (q+i,q+i (S )), i = 1, 2, . . . , q;
1/2
ij (M ) = ji (M ) = j M (ij (S )ij
T
(S )), j > i = 1, 2, . . . , q;
1/2
q+i,q+j (M ) = q+j,q+i (M ) = i M (q+i,q+j (S )q+i,q+j
T
(S ))
1/2
+ j M (q+i,q+j (S )q+i,q+j
T
(S )), i, j = 1, 2, . . . , q, j > i;
1/2
i,q+j (M ) = j M (i,q+j (S )i,q+j
T
(S ))
1/2
+ i j M (i,q+j (S )i,q
T
j
(S )), i, j = 1, 2, . . . , q.
v3+i,3+ifree
Download = yiT 2Jy
(yi ) eBooks , v3+i,3+j (yi , yj ) = 0,
at ibookboon.com i, j = 1, 2, 3, i = j;
vi,3+j (xi , yj ) = xT 1
i, j = 1, 2, 3; 191
i 10 Jyj , j = diag {1, 1, 1}.
It is easy to see that for these elements
vii (xi ) xi 2 , i = 1, 2, 3;
v (x , x ) 0.1 x x , i, j = 1, 2, 3, i = j;
i ij ij
0 2
k
Sjl = diag {sjl k, skjl }, k = 1, 2, l = 1, 2, 3, j = 1, 2, 3, 4, 5, 6;
Stabilityk Theory of1 Large-Scale
0 sjl 1, sii = s3+i,i = 0, s2ii = s13+i,i = 1, s121 = 1, i = 1, 2, 3.
Dynamical Systems Singularly Perturbed Large-Scale Systems
In the matrix-function (5.3.2) the elements are taken as follows
(5.5.8)
vii (xi ) = xT
i Jxi , vij (xi , xj ) = xT 1
i 10 Jxj ,
MASTER IN MANAGEMENT
Moreover, the matrix G(M ) is negative definite for any i (0, 1) and
for i 0, i = 1, 2, 3.
STUDY IN THE CENTER OF MADRID AND TAKE ADVANTAGE OF THE UNIQUE OPPORTUNITIES
By Theorem state (xT, y T )T = 0 R12 of the Length: 1O MONTHS
THAT THE5.4.1
CAPITALthe equilibrium
OF SPAIN OFFERS
Av. Experience: 1 YEAR
PROPEL YOUR in
system determined EDUCATION BY EARNINGis
this example, A DOUBLE DEGREEasymptotically
uniformly THAT BEST SUITS YOUR
stable in Language: ENGLISH / SPANISH
the whole PROFESSIONAL
on M S,GOALS where M = {i : 0 < i 1, i = 1, 2, 3}. Format: FULL-TIME
STUDY A SEMESTER ABROAD AND BECOME A GLOBAL CITIZEN WITH THE BEYOND BORDERS
Intakes: SEPT / FEB
RemarkEXPERIENCE
5.5.2 In this example the independent degenerate subsystem
5 Specializations
dx1 0.1 0 #10 WORLDWIDE 55 Nationalities
=
Personalize your x1 MASTER IN MANAGEMENT
in class
dt program0 0.1 FINANCIAL TIMES
Moreover, the matrix G(M ) is negative definite for any i (0, 1) and
for i 0, i = 1, 2, 3.
By Theorem 5.4.1 the equilibrium state (xT, y T )T = 0 R12 of the
system determined in this example, is uniformly asymptotically stable in
the whole on M S, where M = {i : 0 < i 1, i = 1, 2, 3}.
Remark 5.5.2 In this example the independent degenerate subsystem
dx1 0.1 0
= x1
dt 0 0.1
5.5.2 Uniform time scaling In the case of uniform time scaling system
(5.5.1) is of the form
q r
dxi 1 2
= Ai xi + Si Ai x + Si Ai y , i = 1, 2, . . . , q,
dt =1 =1
q
dyj 1
(5.5.10) 1 = j Bj yj + + 1 Sq+j, Bj x
dt =1
r
2
+ j Sq+j, Bj y , j = 1, 2, . . . , r,
=1
r
1 1
ii (S) = Pi,q+ (Sq+,i Bi ) + (Sq+,i Bi )T Pi,q+
T
, i = 1, 2, . . . , q;
=1
q
c1ip (S) = PiiT (Sp
1 1
Ap ) + (Sp Ap )T Pii ,
=1
i1
c2ip (S) = Pip Ap + AT
p Pip +
T
Pi 1
(Sp 1
Ap ) + (Sp Ap )T Pi
=1
p1
1 1
+ Pi (Sp Aalp ) + (Sp Aalp )T Pi
=i+1
q
1 1
+ Pi (Sp Ap ) + (Sp Ap )T Pi
T
,
=p+1
r
1 1
ip (S) = Pi,q+ (Sq+,p Bp ) + (Sq+,p Bp )T Pi,q+ ,
=1
i, p = 1, 2, . . . , q, p > i;
q
2
q+j,q+j (S) = (Sj Aj )T P,q+j + P,q+j
T 2
(Sj Aj ) , j = 1, 2, . . . , r;
=1
Download free eBooks at bookboon.com
r
c1q+j,q+l (l , S) = T
Pq+j,q+j 2
196Bl ),
l (Sq+,l
=1
q
2
q+j,q+j (S) = (Sj Aj )T P,q+j + P,q+j
T 2
(Sj Aj ) , j = 1, 2, . . . , r;
=1
r
c1q+j,q+l (l , S) = T
Pq+j,q+j 2
l (Sq+,l
Bl ),
=1
j1
T 2
+ Pq+,q+j l (Sq+,l
Bl T
) + j (S 2 2q+,j Bj ) Pq+,q+l
=1
l1
2 2 T
+ Pq+j,q+ l (Sq+,l Bl ) + j (Sq+,j Bj ) Pq+,q+l
=j+1
r
2 2 T
+ Pq+j,q+ l (Sq+,l Bl ) + j (Sq+,j Bj ) Pq+l,q+ ,
=l+1
q
2
q+j,q+l (S) = (Sj Aj )T P,q+l + P,q+l
T 2
(Sj Aj ) ,
=1
j, l = 1, 2, . . . , r, l > j;
i1
T 2
ci,q+j (j , S) = Pi,q+j + Pi (Sj Aj )
=1
q
r
2 2
+ Pi (Sj Aj ) + Pi,q+ (Sq+,j
Bj )j ,
=i =1
q
i,q+j (S) = AT
i Pi,q+j +
1
(Si Ai )P,q+j
=1
j1
q
T 1 1
+ Pq+,q+i (Sq+,j Bj ) + Pq+i,q+ (Sq+,j Bj ),
=1 =j
i = 1, 2, . . . , q, j = 1, 2, . . . , 2, q + r = s.
Note that if the matrix C is negative definite, i.e. M (C) < 0 and
M (G) > 0, then the function DVM (x, y, 1 ) is negative definite for any
1 (0,
1 ) and for 1 0, where 1 = min {1, M (C)/M (G)}.
Note that if the matrix C is negative definite, i.e. M (C) < 0 and
Stability Theory of Large-Scale
M (G) > 0, then the function DVM (x, y, 1 ) is negative definite for any
DynamicalSystems Singularly Perturbed Large-Scale Systems
1 = min {1, M (C)/M (G)}.
1 (0, 1 ) and for 1 0, where
If M (C) < 0 and M (G) < 0, then
1 = 1.
i [ i , i ], 1 = min {1, 1 ,
1 }.
= {M : 0 < 1 <
M 1 , i = 1 i1 , i = 1, 2, . . . , r}.
vii (xi ) = xT
i Jxi ; v2+i,2+i (yi ) = yiT Jyi , i = 1, 2;
v12 (x1 , x2 ) = xT
1 10
1
Jx2 ; v34 (y1 , y2 ) = y1T 101 y2 ,
vi,2+j (xi , yj ) = xT
i 10
1
Jyj , i, j = 1, 2, J = diag {1, 1}.
CLICK HERE
to discover why both socially
and academically the University
of Groningen is one of the best
places for a student to be
www.rug.nl/feb/education
vii (xi ) = xT
i Jxi ; v2+i,2+i (yi ) = yiT Jyi , i = 1, 2;
v12 (x1 , x2 ) = xT
1 10 1
Jx2 ; v34 (y1 , y2 ) = y1T 101 y2 ,
210 vi,2+j (xi , y5. = xT
j ) SINGULARLY
i 10
1 PERTURBED SYSTEMS
Jyj , i, j = 1, 2, J = diag {1, 1}.
where
1 0.1 0.1 0.1
A11 = A22 = , A12 = ,
0.1 1 0.1 0.1
and
For the elements of the matrices C and G specified in such way we have
and
M (C)
1 = min 1, = min {1; 1.1553354} = 1.
M (G)
199
Stability Theory of Large-Scale
Dynamical Systems Singularly Perturbed Large-Scale Systems
5.6 GENERALIZATIONS AND APPLICATIONS 211
In this section general results of this Chapter are applied in two cases. In
Subsection 5.6.1 we present the results of stability analysis for singularly
perturbed large scale Lure-Postnikov systems under nonclassical structural
perturbations. A numerical example of the 12-th order system decomposed
into three interconnected singularly perturbed subsystems is considered as
illustration. In Subsection 5.6.2 we establish the conditions of a spacecraft
motion stabilization by means of three gyroframes. Here the possibility of
application of the matrix-valued function is shown and the obtained result
is compared with those obtained in terms of the vector Liapunov function.
(1) q
dxi
= Ai xi + Sil Ail yl + qi1 fi1 (i1 ),
dt
l=1
i1 = cT T
i1 x + ci2 y, i = 1, 2, . . . , q,
q
dyi (1)
i = i Sq+i,l Bil xl + Bi yi + qi2 l2 (i2 ) + qi3 fi3 (i3 ),
dt
(5.6.1) l=1
i2 = i cT T
i3 xi + ci4 yi ,
q
(2) T (3)
i3 = i cT
i5 Sq+i,l xl + cl6 Sq+i,l yl ,
l=1
l=i
i = 1, 2, . . . , q,
where
1
ij fij (ij ) [0, kij ] R+ , i = 1, 2, . . . , q, j = 1, 2, 3,
0 0
fi0 = Ai xi + qi1 fi1 (
i1 ),
i1 = cT i
i1 x ,
0 0
gi0 = Bi yi + qi2 fi2 (i2 ), i2 = cT i
i2 y ,
0
fi = Aii yi + qi1 [fi1 (
i1 ) fi1 (
i1 )],
0
gi = i Bii xi + qi2 [fi2 (i2 ) fi2 (i2 )],
q
(1)
fi = Sil Ail yl + qi1 [fi1 (i1 ) fi1 (
i1 )],
l=1
l=i
American online
LIGS University
is currently enrolling in the
Interactive Online BBA, MBA, MSc,
DBA and PhD programs:
q
(1)
gi = i Sq+i,l Bil xl + qi3 fi3 (i3 ),
l=1
l=i
i = 1, 2, . . . , q, 2q = s.
dxi
= fi0 + fi + fi , i = 1, 2, . . . , q,
(5.6.3) dt
dyi
= gi0 + gi + gi , i = 1, 2, . . . , q.
dt
Here the vector-functions fi0 and gi0 correspond to the independent degen-
erated subsystem
dxi 0 0
(5.6.4) = Ai xi + qi1 fi1 (
i1 ),
i1 = cT i
i1 x ,
dt
and subsystem describing the boundary layer
dyi 0 0
(5.6.5) i = Bi yi + qi2 fi2 (i2 ), i2 = cT
i2 yi .
dt
The vector-functions fi0 + fi and gi0 + gi correspond to the independent
singularly perturbed subsystems (5.6.2).
Alongside system (5.6.1) and subsystems (5.6.2), (5.6.4), (5.6.5) we shall
consider the matrix-function (5.3.2) with the elements (5.5.3) satisfying
estimates (5.5.4). As is known, for function (5.3.3) the inequality
is valid. Besides, the matrices A(M ), B(M ) and vector u are determined
as in Proposition 5.3.1.
Proposition 5.6.1 If for system (5.6.1) the matrix-function (5.3.2)
with elements (5.5.3) is constructed, then for the Dini derivatives of func-
tions (5.5.3)
i2 [Pii Ai + AT i T i T T
i Pii + Pii ki1 qi1 (ci1 ) + (ki1 qi1 (ci1 ) ) Pii ];
2
q+i [Pq+i,q+i Bi + BiT Pq+i,q+i + Pq+i,q+i ki2
qi2 (cii2 )T
+ (ki2 qi2 (cii2 )T )T Pq+i,q+i ];
i2 [Pii ki1
qi1 (cii1 )T + (ki1
qi1 (cii1 )T )T Pii ];
1 T T
i q+i (Pi,q+i Bii + Bii Pi,q+i + Pi,q+i ki2
qi2 cT T T T
i3 + (ki2 qi2 ci3 ) Pi,qi );
2
2
q+i
[Pq+i,q+i ki2 qi2 cT T T
i4 + (ki2 qi2 ci4 ) Pq+i,q+i ];
1
i q+i (Pi,q+i Aii + AT T i T i T T
ii Pi,q+i + Pi,q+i ki1 qi1 (ci2 ) + (ki1 qi1 (ci2 ) ) Pi,q+i );
2
i2 [Pii ki1
qi1 (cii1 )T + (ki1
qi1 (cii1 )T )T Pii ]
q
+
i j {(kj1 qj1 (cij1 )T )T Pji + Pji
T
kj1 qj1 (cij1 )T };
j=2
j>i
(2) (2)
i q+i [Pi,q+i ki3 qi3 cT T T T
i5 Sq+i,i + (ki3 qi3 ci5 Sq+i,i ) Pi,q+i ];
T
i q+i [Pi,q+i
ki1 qi1 (cii2 )T + (ki1
qi1 (cii2 )T )T Pi,q+i
T (3) (3)
+ Pq+i,q+i
ki3 qi3 cT T T
i6 Sq+i,i + (ki3 qi3 ci6 Sq+i,i ) Pq+i,q+i ]
Download free
respectively, eBooks
and 7i , at8ibookboon.com
, kij (S), k = 1, 2, . . . , 8, i, j = 1, 2, . . . , q, are
norms of the matrices 203
i2 Pii ki1
qi1 (cji1 )T + i j (Pij Aj + AT
i Pij )
i1
+ l j {PliT kl1
ql1 (cjl1 )T + (kl1
ql1 (cjl1 )T )T Plj }
216 5. SINGULARLY
l=1 PERTURBED SYSTEMS
j1
+
{i l Pil kl1 ql1 (cjl1 )T + l j (kl1
ql1 (cjl1 )T )T Plj }
l=i
q
+
{i l Pil kl1 ql1 (cjl1 )T + j l (kl1
ql1 (cjl1 )T )T PjlT };
l=j
(1) (2)
i q+i Pi,q+i (Sq+i,j Bij + ki3 qi3 cT
j5 Sq+i,j )
q
(1) (2)
+
i q+l Pi,q+l (Sq+l,j Blj + kl3 ql3 cT
j5 Sq+l,j );
l=1
l=i
T
Pi,q+i
ki1 qi1 (cji2 )T + i q+i (ki1
qi1 (cji2 )T )T Pi,q+i ;
(1)
q+i q+j Pq+i,q+j (Bj + kj2 qj2 cT T
j4 ) + i q+i Pi,q+i (Sij Aij )
i1
(3)
+
q+l q+i Pq+l,q+i (kl3 ql3 cT
j6 Sq+l,j )
l=1
q
(3)
+
q+i q+l Pq+i,q+l (kl3 ql3 cT
j6 Sq+l,j );
l=i
i1
(1)
q+i q+j (BiT +
(ki2 qi2 cT T
i4 ) )Pq+i,q+j + l i PliT Slj Alj
l=1
q
q
(1) T
+ i j Pil Slj Alj + i q+l Pi,q+l
(ki1 qi1 (clj2 )T )
l=i l=1
i1
T (3)
+ q+i q+l Pq+l,q+i
(kl3 ql3 cT
i6 Sq+l,j )
l=1
q
(3)
+
q+i q+l Pq+i,q+l (kl3 ql3 cT
j6 Sq+l,j );
l=i
(1)
i2 Pii (Sij Aij + ki1
qi1 (cji2 )T ) + i q+i (ki1
qi1 (cji1 )T )T Pi,q+i
+ i q+j Pi,q+j (Bj + kj2 qj2 cT
i4 )
i1
(1)
+ l i PliT (Sjl Alj + kl1
ql1 (cjl2 )T )
l=1
i1
(1)
q+i q+j (BiT + (ki2
qi2 cT T
i4 ) )Pq+i,q+j + l i PliT Slj Alj
l=1
q
q
(1) T
+ i j Pil Slj Alj + i q+l Pi,q+l
(ki1 qi1 (clj2 )T )
l=i l=1
i1
T (3)
+ q+i q+l Pq+l,q+i
(kl3 ql3 cT
i6 Sq+l,j )
l=1
q
(3)
+
q+i q+l Pq+i,q+l (kl3 ql3 cT
j6 Sq+l,j );
l=i
(1)
i2 Pii (Sij Aij + ki1
qi1 (cji2 )T ) + i q+i (ki1
qi1 (cji1 )T )T Pi,q+i
+ i q+j Pi,q+j (Bj + kj2 qj2 cT
i4 )
i1
(1)
+ l i PliT (Sjl AND
5.6 GENERALIZATIONS
Alj +
kl1 ql1 (cjl2 )T )
APPLICATIONS 217
l=1
q
(1)
+
i l Pil (Slj Alj + kl1 ql1 (cjl2 )T )
l=i+1
q
(3)
+
j q+l Pj,q+l (kl3 ql3 cT
l6 Sq+l,i );
l=1
2 T (1) (2)
q+j Pq+j,q+j (Sq+j,i Bji
+ kj3 qj3 cT T
j5 Sq+j,i ) + i q+j Ai Pi,q+j
q
+
l q+j (kl1 ql1 (cjl1 )T )T Pl,q+j ;
l=1
i1
T (1) (2)
q+l q+i Pq+l,q+i (Sq+l,j Blj + kl3 ql3 cj5 Sq+l,j )
l=1
q
T (1) (2)
+ q+i q+l Pq+i,q+l (Sq+l,j Blj + kl3 ql3 cj5 Sq+l,j )
l=i
+ q+i q+j [(ki2 qi2 cT T T
i3 ) Pq+i,q+j + Pq+i,q+j (ki2 qi2 ci3 )]
DV (x, y, M ) uT G(M )u
(5.6.7)
(x, y, M, S) Nx0 Ny0 M S,
Download free eBooks at bookboon.com
where
205
uT = (x1 , x2 , . . . , xq , y1 , y2 , . . . , yq ),
G(M cij +
) = [ ij (M )], i, j = 1, 2, . . . , s, s = 2q.
Proposition 5.6.2 Let all conditions of Proposition 5.6.1 be satisfied.
Then for the total derivative of function (5.3.3) with elements (5.5.3)
Stability Theory of Large-Scale
Dynamical Systems
DV (x, y, M ) uT G(M )u Singularly Perturbed Large-Scale Systems
(5.6.7)
(x, y, M, S) Nx0 Ny0 M S,
where
uT = (x1 , x2 , . . . , xq , y1 , y2 , . . . , yq ),
G(M cij +
) = [ ij (M )], i, j = 1, 2, . . . , s, s = 2q.
The elements of the matrix G(M ) are
cii = 1i + 3i + 9i (S );
cji = 1ij (S ),
cij = i = j;
218 5. SINGULARLY PERTURBED SYSTEMS
cq+i,q+i = 2i + 5i + 11,i (S );
cq+j,q+i = 3ij (S ),
cq+i,q+j = i = j;
ci,q+i = 7i ;
ci,q+j = 6ij (S ),
i = j = 1, 2, . . . , q;
ij (M ) = i (4i + 10i (S ));
q+i,q+i (M ) = i 6i ;
ji (M ) = i 2ij (S ),
ij (M ) = i = j;
q+i,q+j (M ) = i 4ij (S ) + j 5ij (S ),
i = j;
i,q+i (M ) = i 8i ;
i,q+j (M ) = i 7ij (S ) + i j 8ij (S ),
i = j = 1, 2, . . . , q;
206
Stability Theory of Large-Scale
Dynamical Systems Singularly Perturbed Large-Scale Systems
5.6 GENERALIZATIONS AND APPLICATIONS 219
5.6.1.2 Uniform time scaling In the case of uniform time scaling system
(5.6.1) is (see Grujic et al. [1])
r
dxi (1)
= Ai xi + Si Ai y + qi1 fi1 (i1 ),
dt =1
cT
i1 = cT
i1 x + i2 y, i = 1, 2, . . . , q;
q
dyi (1)
i = Sq+i Bi x + i Bi yi
dt
=1
(5.6.8)
+ i qi2 fi2 (i2 ) + i qi3 fi3 (i3 ),
cT
i2 = i cT
i3 + i4 yi ,
q
r
(2) (3)
i3 = cT
i 5 Sq+i, x + cT
6 Sq+i, y
=1 =1
i = 1, 2, . . . , r, q + r = s,
where
1 i = 1, 2, . . . , q when j = 1,
ij fij (ij ) [0, kij ] R+ ,
i = 1, 2, . . . , r when j = 2, 3.
(1) (2) (3)
The structural matrices Sij , Sij , Sij , and S, and the set S are deter-
mined as in Section 5.5.1, i [ i , i ], , the numbers i and i are given.
For the analysis of asymptotic stability of large scale system of Lure
type (5.6.8) we construct matrix-function (5.3.7) with elements (5.5.11)
satisfying estimates (5.5.12). For function (5.5.13) the bilateral estimate
Economics!
(5.6.10)
(x, y, , S) N N M S,
1 x0 y0
2nd place: MSc Economics
2nd place: MSc Econometrics and Operations Research
2nd place: MSc Global Supply Chain Management and
Change
Sources: Keuzegids Master ranking 2013; Elsevier Beste Studies ranking 2012;
Financial Times Global Masters in Management ranking 2012
Maastricht
University is
the best specialist
university in the
Visit us and find out why we are the best! Netherlands
(Elsevier)
Masters Open Day: 22 February 2014
www.mastersopenday.nl
DV (x, y, ) uT C u + uT G u,
1
220
(5.6.10) 5. SINGULARLY PERTURBED 1SYSTEMS
(x, y, 1 , S) Nx0 Ny0 M S,
where
i = 1, 2, . . . , q;
cq+j,q+j = M (Cq+j,q+j (j , S)),
Cq+j,q+j (j ,S) = Pq+j,q+j j Bj + j BjT Pq+j,q+j + Pq+j,q+j j qj2 kj2
cj4 )T
(
(3)
+ (j qj2 kj2
cj4 )T )T Pq+j,q+j + Pq+j,q+j j qj3 kj3
(
cj6 )T S q+j,j
(
(3)
+ (j qj3 kj3
cj6 )T S q+j,j )T Pq+j,q+j
(
r
(3)
+ Pq+j,q+l l ql3 kl3
cj6 )T S q+l,j
(
l=2
l>j
(3)
+ (l ql3 kl3
cj6 )T S q+l,j )T Pq+j,q+l
( T
(3)
+ (j qj3 kj3
cl6 )T S q+j,l )T Pq+j,q+l
(
(3)
T
+ Pq+j,q+l j qj3 kj3
cl6 )T S q+j,l , j = 1, 2, . . . , r;
(
1/2
T
cip = M (Cip Cip ),
p T i T T
Cip =Pii qi1 ki1 (
ci1
) + (qp1 kp1 (
cp1 ) ) Ppp + AT
i Pip + Pip Ap
i
q
T p T p T T
+ Pp q1 k1 (
c1 ) + (q1 k1 (
c1 ) ) Pp
=1 =i+1
=p
p
q
i T i T T T
+ Pi q1 k1 (
c1 ) + (q1 k1 (
c1 ) ) Pi ,
=1 =p+1
=i
Download free eBooks at bookboon.com
208
Stability Theory of Large-Scale
Dynamical Systems Singularly Perturbed Large-Scale Systems
5.6 GENERALIZATIONS AND APPLICATIONS 221
i, p = 1, 2, . . . , q, p > i;
T1/2
cq+j,q+l = M (Cq+j,q+l (j , S) Cq+j,q+l (j , S)),
(3)
Cq+j,q+l (j , S) =Pq+j,q+l j qj3 kj3
cl6 )T S q+j,l
(
(3)
+ (l ql3 kl3 cj6 )T S q+l,j )T Pq+l,q+j
(
+ Pq+j,q+l l Bl + j BjT Pq+j,q+l
+ Pq+j,q+l l ql2 kl2 cl4 )T + (j qj2 kj2
(
cj4 )T )T Pq+j,q+l
(
j
T (3)
+ Pq+,q+l
q3 k3 cj6 )T S q+,j
(
=1
r
(3)
+ ( q3 k3
cj6 )T S q+,j )T Pq+,q+l
(
=j+1
=l
l
(3)
+ Pq+j,q+ q3 k3
cl6 )T S q+,l
(
=1
=j
r
(3)
+ ( q3 k3
cl6 )T S q+,l )T Pq+j,q+ ,
(
=l+1
j, l = 1, 2, . . . , r, l > j;
1/2
T
ci,q+j = M (Ci,q+j (j , S) Ci,q+j (j , S),
(1) j T
Ci,q+j (j , S) =Pii S ij Aij + Pii qi1 ki1
(
ci2 )
q
(1) j T
+ Pip S pj Apj + Pip qp1 kp1 (
cp2 )
p=2
p>i
i = 1, 2, . . . , q, j = 1, 2, . . . , r, q + r = s;
222 ii = M
5. SINGULARLY (ii (S)), SYSTEMS
PERTURBED
ii (S) =Pi,q+i qi2 ki2 ci3 )T + (qi2 ki2
(
ci3 )T )T Pi,q+i
(
r
(2) (2)
+
Pi,q+i qj3 kj3 ci5 )T S q+j,i + (qj3 kj3
(
cj5 )T S q+j,i )T Pi,q+j
( T
,
j=1
i = 1, 2, . . . , q;
q+j,q+j = M (q+j,q+j (S)),
q
(1) (1)
q+j,q+j (S) = (S ij Aij )T Pi,q+j + Pi,q+j S ij Aij
i=1
j T T T j T
+ (qi1 ki1 (
ci2 ) ) Pi,q+j + Pi,q+j qi1 ki1 (
ci2 ) ,
j = 1, 2, . . . , r;
1/2 T
Download free eBooks atbookboon.com
ip = M (ip (S) ip (S)),
ip (S) =Pi,q+p qp2 kp2 cp3 )T + (qi2 ki2
(
ci3 )T )T Pp,q+i209
(
r
(2) (2)
+
Pi,q+j qj3 kj3 cp5 )T S q+j,p + (qj3 kj3
(
ci5 )T S q+j,i )T Pp,q+j ,
(
j=1
q
(1) (1)
q+j,q+j (S) = (S ij Aij )T Pi,q+j + Pi,q+j S ij Aij
i=1
Stability Theory of Large-Scale
j T T T j T
Dynamical Systems + (qi1 ki1 (
ci2 ) ) Pi,q+j + Pi,q+j qi1 ki1 cSingularly
( i2 ) , Perturbed Large-Scale Systems
j = 1, 2, . . . , r;
T1/2
ip = M (ip (S) ip (S)),
ip (S) =Pi,q+p qp2 kp2 cp3 )T + (qi2 ki2
(
ci3 )T )T Pp,q+i
(
r
(2) (2)
+
Pi,q+j qj3 kj3 cp5 )T S q+j,p + (qj3 kj3
(
ci5 )T S q+j,i )T Pp,q+j ,
(
j=1
i, p = 1, 2, . . . , q, p > i;
T1/2
q+j,q+l = M (q+j,q+l (S) q+j,q+l (S)),
q
T (1) T
(1) T
q+j,q+l (S) = S ij Aij Pi,q+l + Pi,q+j S il Ail
i=1
j T T T l T
+ (qi1 ki1 (
ci2 ) ) Pi,q+l + Pi,q+j qi1 ki1 (
ci2 ) ,
j, l = 1, 2, . . . , r, l > j;
T 1/2
i,q+j = M (i,q+j (S) i,q+j (S)),
(1) T (2) T
i,q+j (S) = S q+j,i Bji Pq+j,q+j + qj3 kj3 ci5 )T S q+j,i Pq+j,q+j
(
r
(1) T T (1) T
+ S q+l,i Bli Pq+j,q+l + S q+l,i Bli Pq+j,q+l
l=2
l>j
(2) T T (2) T
+ ql3 kl3 ci5 )T S q+l,i Pq+j,q+l
(
+ ql3 kl3 ci5 )T S q+l,i Pq+j,q+l
(
T
+ ql2 kl2
cl3 )T Pq+j,q+l
( T
+ ql2 kl2 cl3 )T Pq+j,q+l
(
q
T
+ AT
i Pi,q+j +
q1 k1 ( i
cbe1 )T P,q+j ,
=1
i = 1, 2, . . . , q, j = 1, 2, . . . , r, q + r = s;
1 for l = i,
=
0 for l = i.
Here kij
, i = j, and kii
are determined as in Section 5.6.1.2 of this
Chapter, j is determined as in Section 5.2.2, ckij is the kth component of
vector cij , S S is a constant matrix such that
or
224
(b) for any 15.SINGULARLY
(0, 1] and forPERTURBED SYSTEMS
1 0 respectively.
Theorem 5.6.2 Let the equations of singularly perturbed LSS of Lure
type (5.6.8) be such that for the system the matrix-function (5.3.7) is con-
structed with the elements (5.6.9) satisfying estimates (5.6.10) and for the
total derivative of function (5.3.3) by virtue of system (5.6.8) the correla-
tion (5.6.12) is satisfied and
(1) matrices A11 and A22 are positive definite;
(2) matrix C is negative definite;
(3) 1 (0, 1 ), i = i1 1 , i = 1, 2, . . . , r, i [ i , i ], 1 =
min {1, 1 ,
1 }.
Then the equilibrium state (xT, y T )T = 0 of system (5.6.8) is uniformly
S, , where
asymptotically stable on M
= {M : 0 < 1 <
M 1 }, i = i1 1 , i = 1, 2, . . . , r.
If all conditions of Theorem 5.6.2 are satisfied for Nix Njy = Rni nj ,
Download
then free eBooksstate
the equilibrium at bookboon.com
(xT, y T )T = 0 of system (5.6.8) is uniformly
S.
asymptotically stable in the whole on M 211
Proof The assertion of this theorem follows from Theorem 5.3.2.
Remark 5.6.1 If M (G) 0, then assertion of Theorem 5.6.2 remains
1 = min {1, 1 }.
valid for
(3) 1 (0, 1 ), i = i 1 , i = 1, 2, . . . , r, i [ i , i ], 1 =
min {1, 1 ,
1 }.
Then the equilibrium state (xT, y T )T = 0 of system (5.6.8) is uniformly
Stability Theory of Large-Scale
asymptotically stable on M S, , where
Dynamical Systems Singularly Perturbed Large-Scale Systems
M = {M : 0 < 1 < 1 }, i = 1 1 , i = 1, 2, . . . , r.
i
If all conditions of Theorem 5.6.2 are satisfied for Nix Njy = Rni nj ,
then the equilibrium state (xT, y T )T = 0 of system (5.6.8) is uniformly
asymptotically stable in the whole on M S.
The matrix
A11 A12 (M )
A1 (M ) = T ,
A12 (M ) A22 (M )
Download free eBooks at bookboon.com
where
212
0.2 0.01 0.01
A11 = 0.01 0.2 0.01 ,
0.01 0.01 0.2
vip (xi , xp ) = vpi (xi , xp ) 0.01 xi xp , i, p = 1, 2, 3, p > i;
v (yj , yl , M ) = v3+l,3+j (yj , yl , M ) 0.01j l yj yl ,
Stability3+j,3+l
Theory of Large-Scale
Dynamical Systems j, l = 1, 2, 3, l > j; Singularly Perturbed Large-Scale Systems
vi,3+j (xi , yj , j ) 0.01j xi yj , i, j = 1, 2, 3.
The matrix
A11 A12 (M )
A1 (M ) = T ,
A12 (M ) A22 (M )
where
0.2 0.01 0.01
A11 = 0.01 0.2 0.01 ,
0.01 0.01 0.2
0.011 0.012 0.013
A12 (M ) = 0.011 0.012 0.013 ,
226 5. SINGULARLY PERTURBED SYSTEMS
0.011 0.012 0.013
21 0.0112 0.0113
A22 (M ) = 0.011 2 22 0.0123 ,
0.011 3 0.0123 23
0.15383688 for ki1
= 2,
1i =
0.15278641 for ki1 = 0,
8.917237 for ki2
= 1,
2i =
12 for ki2
= 0,
0.00171 for ki1 = 2,
3i =
0 for ki1
= 0,
2.309017 105 for ki2
= 1,
4i =
10 5
for ki2
= 0,
3.2360679 for ki2
= 1,
5i =
0 for ki2
= 0,
for ki1
= 2,
= 0,
dissertation?
0.46264281 for ki1
= 2, ki2
0.38015581 for k = 0, k = 1,
= 1,
i1 i2
7i =
0.45199337 for = 2, in = 0,
Get in-depth k
feedback & advice from i1 experts i2k your
topic area. Find0.37
out what youfor i1 = i2 = 0,
can kdo to kimprove
the quality ofyour dissertation!
0.02407149 for ki1
= 2, ki2
= 1,
0.02279776 for k = 2, k = 0,
i1 i2
8i =
0.02408377 for k
i1 = 0, ki2 = 1,
Get Help0.02280625
Now for ki1
= ki2
= 0,
0.02497321 for k11
= 2,
91 =
0 for k11 = 0,
0.0149 for k21
= 2,
92 =
0 for k21 = 0,
2.309017 10 5
for ki2
= 1,
4i =
10 5
for ki2 = 0,
3.2360679 for ki2
= 1,
5i =
0 for ki2
= 0,
1.4828427 102 for ki1
= 2,
6i =
10 2
for ki1 = 0,
0.46264281 for ki1
= 2, ki2
= 1,
0.38015581 for ki1 = 0, ki2
= 1,
7i =
0.45199337 for ki1 = 2, ki2 = 0,
0.37 for ki1
= ki2
= 0,
0.02407149 for ki1
= 2, ki2
= 1,
0.02279776 for ki1 = 2, ki2
= 0,
8i =
0.02408377 for ki1 = 0, ki2 = 1,
0.02280625 for ki1
= ki2
= 0,
0.02497321 for k11
= 2,
91 =
0 for k11 = 0,
0.0149 for k21
= 2,
92 =
5.6 GENERALIZATIONS0 AND APPLICATIONS 227
for k21 = 0,
0.00828427 for k31
= 2,
93 =
0 for k31 = 0,
4 105 for
ki3 = 1,
10,i (S ) =
0 for ki3 = 0,
4.8626044 103 for ki1
= 2,
ki3 = 1,
3.2360679 103 for ki1
= 2,
ki3 = 0,
11,i (S ) =
4 105 for ki1
= 0, ki3
= 1,
0 for ki1 = ki3 = 0,
0.020000022 for ki1
= 2,
1ij (S ) =
0 for ki1
= 0,
3 105 for ki3
= 1,
2ij (S ) =
1.5 10 5
for ki3
= 0,
106 for ki1
= 2,
3ij (S ) =
0 for ki1 = 0,
0.05121 for ki2
= ki3
= 1,
0.0499996 for ki2
= 0,
ki3 = 1,
4ij (S ) =
0.051161
for ki2
= 1, ki3
= 0,
0.049949 for ki2 = ki3 = 0,
0.064213 for ki1
= ki2
= ki3
= 1,
0.0614868 for ki1 = 0, = ki3 = 1,
Download free eBooks at bookboon.com ki2
0.050653 for ki1 = 2,
ki2 = 0, ki3
= 1,
214
0.06158 for ki1
= 2, ki2 = 1, ki3 = 0,
5ij (S ) =
0.050612 for ki1 = ki2
= 0, ki3
= 1,
0.0611818 for ki1 = 0,
ki2 = 1, ki3
= 0,
0.05121 for ki2
= ki3
= 1,
0.0499996 for ki2
= 0,
ki3 = 1,
Stability
4ijTheory
(S ) = of Large-Scale
Dynamical Systems
0.051161 for ki2
= 1, ki3
= 0, Singularly Perturbed Large-Scale Systems
0.049949 for ki2
= ki3
= 0,
0.064213 for ki1
= ki2
= ki3
= 1,
0.0614868 for ki1
= 0,
ki2 = ki3
= 1,
0.050653 for ki1 = 2,
ki2 = 0, ki3
= 1,
0.06158 for ki1
= 2,
ki2 = 1,
ki3 = 0,
5ij (S ) =
0.050612 for ki1
= ki2
= 0, ki3
= 1,
0.0611818 for k
= 0, k
= 1,
ki3 = 0,
i1 i2
0.050414 for ki1 = 2, ki2 = ki3 = 0,
228 0.050015 for ki1
5. SINGULARLY
= ki2
= ki3
PERTURBED
= 0,
SYSTEMS
0.04989 for
ki1 = ki2
= ki3
= 0,
0.04995 for
ki1 = 2,
ki2 = ki3
= 0,
0.049960962 for ki1 = 0,
ki2 = 1, ki3
= 0,
0.049891 for
ki1 = ki2
= 0, ki3
= 1,
6ij (S ) =
0.04998 for ki1 = 2, ki2 = 1, ki3
= 0,
0.04997 for
ki1 = 2,
ki2 = 0, ki3
= 1,
0.049965 for ki1 = 0,
ki2 = ki3 = 1,
0.04999 for
ki1 = 2,
ki2 = ki3
= 1,
0.021251 for
ki1 = 2,
ki3 = 1,
0.020212 for
ki1 = 0,
ki3 = 1,
7ij (S ) =
0.023521
for
ki1 = 2, ki3
= 0,
0.023454 for
ki1 = ki3
= 0,
0.0018061 for ki2
= ki3
= 1,
0.001798 for ki2
= 0,
ki3 = 1,
8ij (S ) =
0.001124 for ki2
= 1, ki3
= 0,
0.0011 for ki2
= ki3
= 0.
Let = (1, 1, 1, 1, 1, 1), then the elements of the matrix G(M ) are de-
termined as
cii = 0.1261032,
i = 1, 2, 3; cij =
cji = 1012, i, j = 1, 2, 3, i = j;
c3+i,3+i = 5.676308,
i = 1, 2, 3;
c3+j,3+i = 106 ,
c3+i,3+j = i, j = 1, 2, 3, i = j;
ci,3+i = 0.4626428,
i = 1, 2, 3; ci,q+j = 0.04999,
i, j = 1, 2, 3, i = j;
ii = 6.309017 105 i ,
i = 1, 2, 3;
5
ij =
ji = 3 10 i , i, j = 1, 2, 3, i = j;
3+i,3+i = 0.0148284i,
i = 1, 2, 3;
3+i,3+j =
3+j,3+i = 0.051161i + 0.064213j , i, j = 1, 2, 3, i = j;
i,q+i = 0.02408377i,
i = 1, 2, 3;
i,q+j = 5.6
GENERALIZATIONS AND APPLICATIONS
0.023521i + 0.0018061ij , i, j = 1, 2, 3, i = j. 229
For such definition of the elements the matrix G(M ) is negative definite
for j (0, 1] and for j 0, j = 1, 2, 3.
By Theorem 5.6.2 the equilibrium state (xT, y T )T = 0 R12 of the
system specified in this example, is absolutely stable on [0, K] M S,
where
Download free eBooks at bookboon.com
215
K = diag {2, 1, 1, 2, 1, 1, 2, 1, 1},
= {M : M = diag {1 , 2 , 3 }, j (0, 1], j = 1, 2, 3}.
M
Stability Theory 5.6
of Large-Scale
GENERALIZATIONS AND APPLICATIONS 229
Dynamical Systems Singularly Perturbed Large-Scale Systems
For such definition of the elements the matrix G(M ) is negative definite
for j (0, 1] and for j 0, j = 1, 2, 3.
By Theorem 5.6.2 the equilibrium state (xT, y T )T = 0 R12 of the
system specified in this example, is absolutely stable on [0, K] M S,
where
Brain power
of each gyroshroud (with a rotor) about the axis of self-rotation, Bi be
By 2020,
electricity
moment of inertia of each gyroscope about the precession axis, Ci be
equatorial moment of inertia of each gyroshroud (with the rotor),
the
wind
needs.
could provide one-tenth of our planets
the Already today, SKFs innovative know-
how is crucial to running a large proportion of the
and H be
worlds windi turbines.
the intrinsic moment of momentum of the gyroscopes of the gyroframe Up to 25 %of i . the generating costs relate to mainte-
Assuming that Ai = Ci , we obtain for the SCGS system the system of be reduced dramatically thanks to our
nance. These can
systems for on-line condition monitoring and automatic
equations of motion (for the details see Abdullin et al. [1] and the references
lubrication. We help make it more economical to create
cleaner, cheaper energy out of thin air.
By sharing our experience, expertise, and creativity,
industries can boost performance beyond expectations.
Therefore we need the best employees who can
meet this challenge!
I1 = J1 + B2 + A1 + A3 (123),
Bi = 2Bi , Ai = 2Ai , Hi = 2Hi ,
where 0i is the precession angle at which unloading starts, k are the start-
up times (k = 0, 1, 2, . . . ), and tk = k1 + T (k 1) are the unloading
217
The necessity of allowing for GS unloading arises when a prescribed
orientation of a spacecraft should be maintained for a long time. It is
accepted that unloading is relized by the law
Stability Theory of Large-Scale
Dynamical Systems0, for t [tk , k ); Singularly Perturbed Large-Scale Systems
M (i (t)) =
M0 sign i (k ), for t [k , tk+1 ), |i (k )| = 0i ,
Ti Myi + Myi = Fi (i ),
i = fi (i )(F1i (i ) + F2i (pi )) F3i (i ),
The values of x0ji (x01i = i0 , x02i = p0i , x03i = 0i = qi0 ) are determined
by the dead zones, the noise of both the most sensitive elements and sig-
nal amplifiers, quantization, and other nonlinearities of the characteristics.
The value of i0 is determined by the TS hysteresis, the dead zone, and
other nonlinearities of the static characteristics of both the sensor itself
and amplifiers, noise in the amplifiers, errors and quantization in computa-
tional devices, etc.
232
The quantity i addresses SYSTEMS
5. SINGULARLY PERTURBED
the saturation of the TS or
amplifying devices.
In the monograph by Abdullin et al. [1], the following notation was
introduced
2 = p2 (cos 3 1) + p1 sin 3 ;
3 = (p2 sin 3 p1 cos 3 ) cos 2 ;
mi = Mi /Ii ; m = M /Ii ; mi = Mi /Bi ; nji = kji /(Bi Ti );
gi = Hi /Bi ; hi = bi /Bi ; di = 1/Ti ; i = i /(Bi Ti );
i (i ) = Fi (Bi Ti i )/(Bi Ti ) i ;
ri = i fi (i )(n1i i + n2i pi ) + n3i gi .
x1i = x2i + i ,
x2i = ai x3i cos i + mi + m i ,
(5.6.13) x3i = hi x3i + gi x2i cos i + x4i + mi ,
x4i = di x4i + fi (i )(n1i x1i + n2i x2i ) n3i x3i + ri + i i ,
= x3i , i = 1, 2, 3.
2 x1 = S1 x2 ,
x2 = S2 x2 M Ax3 + m + mp ,
(5.6.14) x3 = Hx3 + M Gx2 + x4 + m ,
M x4 = M Dx4 + N1 x1 + N2 x2 M N3 x3 + M r,
= x3 .
S 1 S1 S 1 , S 2 S2 S 2 , j (0, 1],
where
0 1 0 1 0 0
S 1 = 1 0 0, S1 = 0 1 0 ,
1 0 0 0 1 1
0 0 a1 p2 a12
S 2 = a2 p3 a23 0 0 ,
0 a3 p1 a31
0
0 a13 a1 p2
S 2 = a2 p3
0 a21 .
234 5. SINGULARLY
a32 PERTURBED
a3 p1
SYSTEMS
0
11 (x1 , 2 ) = 2 xT
1 B11 x1 , 12 (x1 , x2 , 2 ) = 2 xT
1 B12 x2 ,
13 = 0; 14 (x1 , x4 , 2 , M ) = 2 xT
1 B14 (M x4 ), 15 = 0,
22 (x2 ) = xT
2 B22 x2 , 23 (x2 , x3 ) = xT
2 B23 x3 ,
(5.6.16) 24 (x2 , x4 , M ) = xT
2 B24 (M x4 ), 25 (x2 , ) = xT
2 B25 ,
T
33 (x3 ) =
Download freeeBooks at x3 B33 x3 ,
bookboon.com34 (x3 , x4 , M ) = xT
3 B34 (M x4 ),
35 (x3 , ) = xT
3 B35 , = (M x4 )T B44 x4 ,
44 (x4 , M )220
45 = 0, 55 () = T B55 .
11 of
Stability Theory , 2 ) = 2 xT
(x1Large-Scale 1 B11 x1 , 12 (x1 , x2 , 2 ) = 2 xT
1 B12 x2 ,
Dynamical Systems
13 = 0; 14 (x1 , x4 , 2 , M ) = 2 xT B14 (M x4 ), Singularly
15 = 0,
Perturbed Large-Scale Systems
1
22 (x2 ) = xT
2 B22 x2 , 23 (x2 , x3 ) = xT
2 B23 x3 ,
(5.6.16) 24 (x2 , x4 , M ) = xT
2 B24 (M x4 ), 25 (x2 , ) = xT
2 B25 ,
33 (x3 ) = xT
3 B33 x3 , 34 (x3 , x4 , M ) = xT
3 B34 (M x4 ),
35 (x3 , ) = xT
3 B35 , 44 (x4 , M ) = (M x4 )T B44 x4 ,
45 = 0, 55 () = T B55 .
uT Bu (x1 , x2 , x3 , x4 , , 2 , M )T uT Bu,
where
u = (x1 , x2 , x3 , x4 , )T ,
B = [bij ]5i,j=1 , bij = bji , B = [bij ]5i,j=1 , bij = bji ,
b11 = 2 m (B11 ), b22 = m (B22 ), b33 = m (B33 ), b44 = m (B44 ),
b55 = m (B55 ), b11 = 2 M (B11 ), b22 = M (B22 ), b33 = M (B33 ),
1/2
T
b44 = M (B44 ), b55 = M (B55 ), b12 = b12 = 2 M (B12 B12 ),
236 5. bSINGULARLY PERTURBED 1/2
SYSTEMS T
b13 = 13 = 0, b14 = b14 = 2 M (B14 B14 ),
1/2T
b15 = b15 = 0, b23 = b23 = M (B23 B23 ),
1/2
T 1/2
T
b24 = b24 = M (B24 B24 ), b25 = b25 = M (B25 B25 ),
1/2
T 1/2
T
b34 = b34 = M (B34 B34 ), b35 = b35 = M (B35 B35 ),
b45 = b45 = 0.
5 5 5 4
d dii dij
(x1 , x2 , x3 , x4 , , 2 , M ) = +2 xT
i Kii xi
dt i=1
dt i=1 j=2
dt i=1
4
4 4
4
4
+2 xT
i Kij xj + xT
i Ki1 + xT
i Ki2 (m + mp ) + xT
i Ki3 m
i=1 j=2 i=1 i=1 i=1
4
+ xT T T
i Ki4 r + m B35 + (m + mp ) B25 ,
i=1
where
5 5 5 4
d dii dij
(x1 , x2 , x3 , x4 , , 2 , M ) = +2 xT
i Kii xi
dt i=1
dt i=1 j=2
dt i=1
4
4 4
4
4
+2 xT
i Kij xj + xT
i Ki1 + xT
i Ki2 (m + mp ) + xT
i Ki3 m
i=1 j=2 i=1 i=1 i=1
4
+ xT T T
i Ki4 r + m B35 + (m + mp ) B25 ,
i=1
where
K11 = 0;
K21 = 2(S2T B25 + (M G)T B35 );
K31 = 2(B55 (M A)T B25 H T B35 );
K41 = 2B35 ;
T
K12 = 22 B22 ;
K22 = 2B22 ;
K23 = 2B23 ;
K42 = 2M B24 ;
T
K13 = 0; K23 = 2B23 ; K33 = 2B33 ; K43 = 2M B34 ;
K14 = 22 B14 M ;
K24 = 2B24 M ;
K34 = 2B34 M ;
K44 = 2B44 M ;
cos 2 if the corresponding factors are negative,
2 =
1 if the corresponding factors are positive,
M = diag (4 , 5 , 6 ),
cos k3 if the corresponding factors are negative,
k =
1 if the corresponding factors are positive,
Download free eBooks at bookboon.com
k = 4, 5, 6;
223
3 3 0
S1 = 2
3 2 3 0 ,
3 = 1 (3 )2 ;
(2 )2 3 (2 )2
3 2
Stability Theory of Large-Scale
Dynamical Systems Singularly Perturbed Large-Scale Systems
238 5. SINGULARLY PERTURBED SYSTEMS
w = (x1 , x2 , x3 , x4 )T ,
K = [ij ]4i,j=1 , ij = ji ,
T
ii = M (Kii ), ij = M (Kij Kij ), i, j = 1, 2, 3, 4, i = j,
j = (1j , 2j , 3j , 4j )T , j = 1, 2, 3, 4,
ij is the norm of the matrices Kij
.
Assume that m m , m + mp m + mp , and r r; then
estimate (5.6.19) takes the following form in domain (5.6.12)
d
(5.6.20) (x1 , x2 , x3 , x4 , , 2 , M ) M w2 + lw + f,
dt
where
l = 1 + 2 (m + mp ) + 3 m + 4 r,
1/2 T T 1/2
f = M (B35 B35 )m + M (B25 B25 )m + mp .
d
From (5.6.20), it follows that the expression (x1 , x2 , x3 , x4 , , 2 , M )
dt
is negative definite if and only if the following conditions are satisfied
M (K) < 0,
l + l2 + 4f M (K)
(5.6.21) w > .
2M (K)
However, the inequality below holds in domain (5.6.12)
3
2 2 2 2
1/2
(5.6.22) w (i ) + (pi ) + (qi ) + (Ui ) ,
i=1
Proof From the condition m (B) > 0, it follows that scalar function
(5.6.18) is positive-definite in the sense of Liapunov. If the condition
M (K) < 0 and inequality (5.6.21) are satisfied, then the total derivative
of function (5.6.18) will be negative definite by virtue of system (5.6.14). As
is known, these conditions are sufficient for the motion of system (5.6.14)
to be asymptotically stable.
Section 5.1 The results of stability analysis of solutions for the given class of
systems in terms of vector functions are presented in Grujic et al. [1]. In this
chapter we use some results from the above monograph and set out the results
of development of a new method of qualitative analysis of singularly perturbed
systems dynamics in terms of auxiliary matrix-valued functions. In the frame-
work of this approach we succeed in reducing the requirement to the individual
subsystems of system F and extending the boundaries of the admissible upper
values of small parameters as compared to those obtained or/and applied in terms
of the vector Liapunov function.
Section 5.2 In this section we use the same models of singularly perturbed
system under nonclassical structural perturbations as in Grujic et al. [1], but
in the mathematical composition of large scale system for the given model of
connectedness.
Sections 5.3 5.4 These two sections are based on the results by Martynyuk and
Miladzhanov [4, 5, 9], and Miladzhanov [4, 5].
Section 5.5 Some results of this section are presented by Martynyuk and Mi-
ladzhanov [9].
Section 5.6 Subsection 5.6.1 is based on the papers by Martynyuk and Mi-
ladzhanov [9, 10]. Also some estimates from Section 5.5 are used.
Section 5.6.2 is based on the paper by Martynyuk and Miladzhanov [8]. Be-
sides, we employ some results by Voronov and Matrosov (Eds.) [1] who applied
the vector Liapunov function in the solution of the problem.
225
Stability Theory of Large-Scale
Dynamical Systems References
References
Coppel, W. A.
[1] Stability and Asymptotic Behaviour of Differential Equations. Boston: Heath,
1965.
Cronin, J.
[1] Recurrent solutions of some autonomous equations. J. Differential Equations 3
(1976) 595 600.
De Baggis, H.
[1] Dynamical sytems with stable structure. In: Contributions to the theory of non-
linear oscillations. Vol. 2. Princeton: Princeton University Press, 1952 (Ann.
of Math. Studies, No 29).
Diamond, P.
[1] Discrete Liapunov function with V > 0. J. Austral. Math. Soc. 20B (1978)
280 284.
Djordjevic, M. Z.
[1] Stability analysis of large scale systems whose subsystems may be unstable.
Large Scale Systems 5 (1983) 252 262.
Eckhaus, W.
[1] Asymptotic Analysis of Singular Perturbations. Amsterdam: North-Holland,
1979.
Furasov, V. D.
[1] Stability and Stabilization of Discrete Processes. Moscow: Nauka, 1982. [Rus-
sian]
Download free eBooks at bookboon.com
Cronin, J.
[1] Recurrent solutions of some autonomous equations. J. Differential Equations 3
(1976) 595 600.
De Baggis, H.
[1] Dynamical sytems with stable structure. In: Contributions to the theory of non-
linear oscillations. Vol. 2. Princeton: Princeton University Press, 1952 (Ann.
of Math. Studies, No 29).
Diamond, P.
[1] Discrete Liapunov function with V > 0. J. Austral. Math. Soc. 20B (1978)
280 284.
Djordjevic, M. Z.
[1] Stability analysis of large scale systems whose subsystems may be unstable.
Large Scale Systems 5 (1983) 252 262.
Eckhaus, W.
[1] Asymptotic Analysis of Singular Perturbations. Amsterdam: North-Holland,
1979.
Furasov, V. D.
[1] Stability and Stabilization of Discrete Processes. Moscow: Nauka, 1982. [Rus-
244 REFERENCES
sian]
Garofalo, F. and Glielmo, L. (Eds).
[1] Robust Control via Variable Structure and Lyapunov Techniques. Berlin: Sprin-
ger-Verlag, 1996.
Grujic, Lj. T.
[1] Uniform asymptotic stability of non-linear singularly perturbed general and large
scale systems. Int. J. Control 33 (1981) 481 504.
Grujic, Lj. T., Martynyuk, A. A. and Ribbens-Pavella, M.
[1] Large-Scale Systems Stability under Structural and Singular Perturbations.
Berlin: Springer-Verlag, 1987.
Hahn, W.
[1] Theorie and Anwendung der Direkten Methode von Liapunov. Berlin: Springer-
Verlag, 1959.
[2] Stability of Motion. Berlin: Springer-Verlag, 1967.
Halanay . and Wexler D.
[1] Qualitative Theory of Impulsive Systems. Moscow: Mir, 1971. [Russian]
Hale, J. K.
[1] Topics in Dynamic Bifurcation Theory. Regional Conference Series in Mathe-
matics, No. 47, AMS, Providence, 1981.
[2] Asymptotic Behaviour of Dissipative Systems. Math. Surveys and Monographs,
Providence: Amer. Math. Soc., Vol. 25, 1988.
Hirsch, M. free
Download W. eBooks at bookboon.com
[1] Systems of differential equations that are competitive or cooperative V: Struc-
228
tural stability in 3-dimensional systems. SIAM J. Math. Anal. 21 (1990) 1225
1234.
Hoppensteadt, F.
[1] Asymptotic stability in singular perturbation problem. Diff. Eqns. 4 (1968)
[2] Stability of Motion. Berlin: Springer-Verlag, 1967.
Halanay . and Wexler D.
[1] Qualitative Theory of Impulsive Systems. Moscow: Mir, 1971. [Russian]
Stability Theory of Large-Scale
Dynamical
Hale, J. K.Systems References
[1] Topics in Dynamic Bifurcation Theory. Regional Conference Series in Mathe-
matics, No. 47, AMS, Providence, 1981.
[2] Asymptotic Behaviour of Dissipative Systems. Math. Surveys and Monographs,
Providence: Amer. Math. Soc., Vol. 25, 1988.
Hirsch, M. W.
[1] Systems of differential equations that are competitive or cooperative V: Struc-
tural stability in 3-dimensional systems. SIAM J. Math. Anal. 21 (1990) 1225
1234.
Hoppensteadt, F.
[1] Asymptotic stability in singular perturbation problem. Diff. Eqns. 4 (1968)
510 521.
Ikeda, M. and Siljak, D. D.
[1] Hierarchical Liapunov functions. J. Math. Anal. Appl. 112(1) (1985) 110 128.
Iooss, G., and Joseph, D.D.
[1] Elementary Stability and Bifurcation Theory. New York: Springer- Verlag, 1990.
Kaplan, W.
[1] Dynamical systems with indeterminacy. Amer. J. Math. 72 (1950) 573 594.
Kokotovic, P. and Khalil, H.
[1] Singular Perturbation Methods in Control: Analysis and Design. London: Aca-
REFERENCES 245
demic Press, 1986.
Kou, Chunhai, Zhang Shunian and Wu, Shujin
[1] Stability analysis in terms of two measures for impulsive differential equations.
Bull. London Math. Soc. (to appear)
Leela, S.
[1] Stability of differential systems with impulsive perturbations in terms of two
Download free eBooks at bookboon.com
measures. Nonlin. Anal. 1 (1977) 667 677.
Lefschetz, S. 229
[1] Differential Equations: Geometric Theory. New York: Interscience Publishers,
1957.
Levins, R.
J. Math. Anal. Appl. 140 (1989) 107 114.
LaSalle, J. P.
Stability Theory of Large-Scale
[1] Stability theory for difference equations. In.: Studies in Ordinary Differential
Dynamical Systems References
Equations. (Ed.: J. Hale), The Mathematical Association of America, 1977,
1 31.
Leela, S.
[1] Stability of differential systems with impulsive perturbations in terms of two
measures. Nonlin. Anal. 1 (1977) 667 677.
Lefschetz, S.
[1] Differential Equations: Geometric Theory. New York: Interscience Publishers,
1957.
Levins, R.
[1] Complex systems. In: Towards a Theoretical Biology. (Ed.: C. H. Waddington),
246 REFERENCES
Edinburg: Edinburd Univ. Press, 1969, 73 88.
Luca, N. and Talpalaru, P.
[1] Stability and asymptotic behaviour of a class of discrete systems. Ann. Mat.
Pure Appl. 112 (1977) 351 382.
Lukyanova, T.A. and Martynyuk, A.A.
[1] Robust stability: Three approaches for discrete-time systems. Nonlinear Dy-
namics and Systems Theory 2 (2002) 45 55.
Lyapunov, A. M.
[1] The General Problem of Stability of Motion. Kharkov, Kharkov Mathemati-
cal Society, 1892. [Russian] (see also French translation in Ann. Fac Toulouse
9 (1907) 203 474, and English translation in: Ann. of Mathematical Study,
No. 17, Princeton: Princeton University Press, 1949, and London: Taylor and
Francis, 1992)
Malkin, I.G.
[1] To the question of inverse Liapunov theorem on asymptotic stability. Prikl. Mat.
Mekh. 18(2) (1954) 129 138. [Russian]
OMalley, R.E.
[1] Introduction to Singular Perturbations, New York, Academic Press, 1974.
Markus, L.
[1] Structurally stable differential systems. Ann. Math. 73 (1961) 1 19.
Martynyuk, A. A. Challenge the way we run
[1] The Lyapunov matrix function. Nonlin. Anal. 8 (1984) 1223 1226.
[2] Extension of the state space of dynamical systems and the problem of stabi-
lity. Colloquia Mathematica Societaties Janas Bolyai 47. Differential Equations:
Qualitative Theory, Szeged (hungary), 1984, 711 749.
[3] The Lyapunov matrix function and stability of hybrid systems. Nonlin. Anal.
10 (1986) 1449 1457.
EXPERIENCE THE POWER OF
[4] Uniform asymptotic stability of a singularly perturbed system via the Liapunov
FULL ENGAGEMENT
matrix-function. Nonlin. Anal. 11 (1987) 1 4.
[5] Hierarchical matrix Lyapunov function. Diff. and Integ. Eqns 2(4) (1989) 411
417.
[6] The hierarchical Lyapunov matrix function and stability under structural per-
turbations. Dokl. Akad. Nauk SSSR 305(1) (1989) 41 44. [Russian]
RUN FASTER.
[7] Stability analysis of nonlinear systems via Liapunov matrix function. Prikl.
RUN LONGER..
Mekh. 27(8) (1991) 3 15. [Russian]
READ MORE & PRE-ORDER TODAY
RUN EASIER
[8] Stability analysis with respect to two measures of system under structural per-
WWW.GAITEYE.COM
turbations. Differential Equations and Dynamical Systems 1(3) (1993) 257
266.
Pontryagin, L. S.
[1] Asymptotic behaviour of systems of differential equations solutions with a small
parameter under the highest derivatives. Izv. AN SSSR, Ser. Matem. 31 (1957)
605 626 [Russian]
Download
Rama free eBooks
Mohana at bookboon.com
Rao, M.
[1] Ordinary Differential Equations. Theory and234
Applications. New Delhi Madras:
Affiliated East-West Press Pvt Ltd, 1980.
Ribbens-Pavella, M.
[1] Critical survey of transient stability studies of multimachine power systems by
[1] Introduction to Structurally Stability of the Systems of Differential Equations.
Boston: Birkhauser, 1988.
Stability Theory of Large-Scale
Piontkovskii, A. A., and Rutkovskaya, L. D.
Dynamical Systems References
[1] Investiagation of certain stability theory problems by the vector Lyapunov func-
tion method. Avtomatika i Telemekhanika 10 (1967) 23 31. [Russian]
Pontryagin, L. S.
[1] Asymptotic behaviour of systems of differential equations solutions with a small
parameter under the highest derivatives. Izv. AN SSSR, Ser. Matem. 31 (1957)
605 626 [Russian]
Rama Mohana Rao, M.
[1] Ordinary Differential Equations. Theory and Applications. New Delhi Madras:
Affiliated East-West Press Pvt Ltd, 1980.
Ribbens-Pavella, M.
[1] Critical survey of transient stability studies of multimachine power systems by
Liapunovs direct method. Ninth Allerton Conference on Circuit and System
Theory, October 6 8, 1971. (Manuscript)
Ribbens-Pavella, M., Grujic, Lj.T., Sabatel, J.
[1] Direct methods for stability analysis of large scale power systems. IFAC, New
REFERENCES 251
Dehli, August 16-18, 1979. (Manuscript)
Rouche, N., Habets, P., and Laloy, M.
[1] Stability Theory by Liapunovs Direct Method. New York: Springer-Verlag, 1977.
Samarskii, A. A., and Gulia, A. V.
[1] Stability of Difference Configurations. Moscow: Nauka, 1973.
Samoilenko, A. M., and Perestyuk, N. A.
[1] Differential Equations with Pulsed Action. Kiev: Vishcha Shkola, 1987 [Rus-
sian].
Sell, G .R.
[1] References on dynamical systems. University of Minnesota, Preprint 91129,
1993.
Sezer, M.E. and Siljak, D. D.
[1] Robust stability of discrete systems. Int. J. Control 48 (1988) 2055 2063.
Shaaban, H., and Grujic, Lj.T.
[1] Transient stability analysis of large scale power systems with speed governor via
vector Liapunov functions. IEEE Proc. 132(2) (1985) 45 52.
Shub, M.
[1] Global Stability of Dynamical Systems. Berlin: Springer-Verlag, 1987.
Siljak, D.D.
[1] On large-scale system stability. In: Proceedings Ninth Annual Allerton Confer-
ence on Circuit and System Theory, October 6 8. Allerton House, Monticallo,
Illinois, 1971, 731 741.
[2] Stability of large-scale systems under structural perturbations. IEEE Trans.
Syst., Man, Cybern. SMC-2 (1972) 657 663.
[3] On stability of large-scale systems under structural perturbations. IEEE Trans.
Syst., Man, Cybern. SCM-3 (1973) 415 417.
[4] Large-Scale Dynamic Systems: Stability and Structure. New York: North Hol-
land, 1978.
[5] Decentralized Control of Complex Systems. Boston: Academic Press, Inc., 1991.
Simeonov, P. S. and Bainov, D. D.
[1] The second method of Liapunov for systems with an impulse effect. Tamkang
J. Math. 16 (1986) 247 263.
Simon, H. A.
[1]Download free eBooksofatcomplexity.
The architecture bookboon.com
Proc. Amer Philos. Soc. 106 (1962) 467 482.
Skowronski, J. M. 235
[1] Nonlinear Liapunov Dynamics. Singapore, etc.: World Scientific, 1990.
Syst., Man, Cybern. SMC-2 (1972) 657 663.
[3] On stability of large-scale systems under structural perturbations. IEEE Trans.
Syst., Man, Cybern. SCM-3 (1973) 415 417.
Stability Theory of Large-Scale
[4] Large-Scale Dynamic Systems: Stability and Structure. New York: North Hol-
Dynamical Systems References
land, 1978.
[5] Decentralized Control of Complex Systems. Boston: Academic Press, Inc., 1991.
Simeonov, P. S. and Bainov, D. D.
[1] The second method of Liapunov for systems with an impulse effect. Tamkang
J. Math. 16 (1986) 247 263.
Simon, H. A.
[1] The architecture of complexity. Proc. Amer Philos. Soc. 106 (1962) 467 482.
Skowronski, J. M.
[1]252Nonlinear Liapunov Dynamics. REFERENCES
Singapore, etc.: World Scientific, 1990.
Smale, S.
[1] Structurally stable systems are not dense. Amer. J. Math. 88 (1966) 491 496.
Thom, R.
[1] Structural stability, catastrophe theory, and applied mathematics. SIAM Review
19(2) (1977) 189 201.
Tikhonov, A. N.
[1] Systems of differential equations containing small parameters in the derivatives.
Math. Sborn. 73 (1952) 576 586. [Russian]
Tsypkin, Ya. Z. and Melikyan, A. A.
[1] Theory of Nonlinear Pulse Systems. Moscow: Nauka, 1973. [Russian]
Vasileva, A. B. and Butuzov, V. F.
[1] Asymptotic Expansions of Solutions of Singularly Perturbed Equations. Moscow:
Nauka, 1973. [Russian]
Voronov, A. A., and Matrosov, V.M. (Eds.)
[1] Method of Vector Lyapunov Functions in Stability Theory. Moscow: Nauka,
1987. [Russian]
Volosov, V. M.
[1] Averaging in systems of ordinary differential equations. Russian Math. Surveys
17 (1962) 1 126. [Russian]
Vorotnikov, V. I.
[1] Stability of dynamical systems with respect to part of variables. Moscow: Nauka,
1991. [Russian]
Walter, W.
[1] Differential and Integral Inequalities. Berlin: Springer-Verlag, 1970.
Zeeman, E. C.
[1] Stability of dynamical systems. Nonlinearity 1 (1988) 115 155.
Zubov, V. I.
[1] Methods of A. M. Liapunov and its Applications. Leningrad: Izdat. Leningrad.
Gos. Universitet, 1957. [Russian]
[2] Mathematical Methods of Investigations of Automatic Control Systems. Lenin-
grad: Mashinostroeniye, 1979. [Russian]
[3] Control Processes and Stability. St-Petersberg: St-Petersberg Gos. Universitet,
1999. [Russian]
236
Stability Theory of Large-Scale
Dynamical Systems Index
Index
-condition, 20 stable, 14
-condition, 20 asymptotically (in the whole), 4
B-topology, 19 connectively, 26
-neighborhood of function x(t), 19 equi-asymptotically, 14
Asymptotically contractive set, 45 exponentially, 14
Compact differentiable manifold, 24 in the whole, 14
Comparison function of class PC, 154 quasi-uniformly asymptotically, 16
of class PCK, 154 uniformly, 14
of class M , 154 uniformly asymptotically, 14
Continuous large-scale system, 5, 30 uniformly in the whole, 3
Dini derivative, 23, 33 unstable, 15
Discrete-time large-scale system, 6, 91 Equilibrium state (xT, y T )T = 0 uniformly
Equations of the i-th degenerate independent asymptotically stable, 184, 192
subsystem Si0, 22 in the whole, 176, 182, 198
Equilibrium state attractive, 16 unstable, 185, 186
x0 -uniformly, 15 structurally uniformly asymptotically stable in the
t0 -uniformly, 15 whole, 186
uniformly, 15 Impulsive large-scale system, 18, 91
www.sylvania.com
We do not reinvent
the wheel we reinvent
light.
Fascinating lighting offers an infinite spectrum of
possibilities: Innovative technologies and new
markets provide both opportunities and challenges.
An environment in which your expertise is in high
demand. Enjoy the supportive working atmosphere
within our global group and benefit from international
career paths. Implement sustainable ideas in close
cooperation with other specialists and contribute to
influencing our future. Come and join us in reinventing
light every day.
Light is OSRAM
238