Exact Analysis of Biperiodic Structures

Exact Analysis of dic Cai - K Liu Exact Analysis of 6'-Periodic Structures This page is intentionally left blank ...

Author: C.W. Cai | J.K. Liu

31 downloads 847 Views 6MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Exact Analysis of

dic Cai -

K Liu

Exact Analysis of

6'-Periodic Structures

This page is intentionally left blank

E x a c t Analysis of

k

Periodic tructures C W Cai Department of Mechanics, Zhongshan University, China

J K Liu Department of Mechanics, Zhongshan University, China

H C Chan Department of Civil Engineering, The University of Hong Kong, Hong Kong

vg b

World Scientific

New Jersey. London .Singapore Hong Kong

Published by

World Scientific Publishing Co. Re. Ltd. P 0 Box 128, Farrer Road, Singapore 912805 USA once: Suite lB, 1060 Main Street, River Edge, NJ 07661 UK once: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-PublicationData A catalogue record for this book is available from the British Library.

EXACT ANALYSIS OF BI-PERIODIC STRUCTURES

Copyright O 2002 by World Scientific Publishing Co. Re. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, includingphotocopying, recording or any information storage and retrieval system now known or to be invented, without written permissionfrom the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.

ISBN 981-02-4928-4

Printed in Singapore.

PREFACE In the book "Exact Analysis of Structures with Periodicity using UTransformation" (World Scientific 1998), a comprehensive and systematic explanation has been given on the U-transformation method, its background, physical meaning and mathematical formulation. The book has demonstrated the application of the U-Transformation method in the analyses of many different kinds of periodic structures. As it has been rightly pointed out in the book, the method has a great potential for further development. With the research efforts by the authors and others in recent years, important advancement in the application of the U-transformation method has been made in the following areas: The static and dynamic analyses of bi-periodic structures Analysis of periodic systems with nonlinear disorder. The static and dynamic analyses of bi-periodic structures When the typical substructure in a periodic structure is itself a periodic structure, the original structure is classified as a bi-periodic structure: for example, a continuous truss supported on equidistant supports with multiple equal spans. As a singly periodic structure, the truss within each bay or span between two adjacent supports is a substructure. But there could be many degrees of freedom in such a substructure. If the U-transformation method is applied to analyze this structure as illustrated in the previous book, every uncoupled equation still contains many unknown variables, the number of which is equal to the number of degrees of freedom in each substructure. Therefore, it is not possible to obtain the explicit exact analytical solution yet. Though the substructure is periodic, it is not cyclic periodic. Hence, it is not possible to go any further to apply the same Utransformation technique directly to uncouple the equations. One of the main objectives for writing this new book is to show how to extend the U-transformation technique to uncouple the two sets of unknown variables in a bi-periodic structure to achieve an analytical exact solution. Through an example consisting of a system of masses and springs with bi-periodicity, this book presents a procedure on how to apply the U-transformation technique twice to uncouple the unknowns and get an analytical solution. The book also produces the static and dynamic analyses for certain engineering structures with bi-periodic properties. These include continuous truss with any number of spans, cable network and grillwork on supports with periodicity, and grillwork with periodic stiffening members or equidistant line supports. Explicit exact solutions are given for these examples. The availability of these exact solutions not only helps the checking of the convergence and accuracy of the numerical solutions for these structures, but also provides a basis for the

optimization design for these types of structures. It is envisaged that there may be a great prospect for the application of this technique in engineering.

Analysis of periodic systems with nonlinear disorder The study on the force vibration and localized mode shape of periodic systems with nonlinear disorder is yet another research area that has attained considerable success by the application of the U-transformation method. The localization of the mode shape of nearly periodic systems has been a research topic attracting enormous attention and concern in the past decade. In the same way, localization problem also exists in periodic systems with nonlinear disorder. This book illustrates the analytical approach and procedure for these problems together with the results. It looks that there are big differences in the physical and mechanical meaning of the problems in the above-mentioned two areas. But as a matter of fact there are similarities in the approaches to their analyses. It is appropriate to present them all together in this book. They are both good examples of the amazing successful application of the U-transformation method. The advantage of applying the U-transformation method is to make it possible for the linear simultaneous equations, either algebraic or differential equations, with cyclic periodicity to uncouple. The first chapter in this book will provide a rigorous proof for this significant statement and give the form of the uncoupled equations. The result will be used in the procedure to obtain the solutions for the example problems in this book. Many achievements in this new book are new results that have just appeared in international journals for the first time together with some which have not been published before. This book can be treated as an extension of the previous book "Exact Analysis of Structures with Periodicity using U-Transformation" with the latest advancement and development in the subject. Nevertheless, sufficient details and explanations have been given in this book to make it a new reference book on its own. However, it will be helpful if readers of this book have obtained some ideas of the mathematical procedures and the applications of the U-transformation method from the previous book.

Prof. H.C. Chan Oct. 30, 2001

CONTENTS Preface Chapter 1 U Transformation and Uncoupling of Governing Equations for Systems with Cyclic Bi-periodicity 1.1 Dynamic Properties of Structures with Cyclic Periodicity 1.1.1 Governing Equation 1.1.2 U Matrix and Cyclic Matrix 1.1.3 U Transformation and Uncoupling of Simultaneous Equations with Cyclic Periodicity 1.1.4 Dynamic Properties of Cyclic Periodic Structures 1.2 Uncoupling of Simultaneous Equations with Cyclic Periodicity for Variables with Two Subscripts 1.2.1 Double U Transformation 1.2.2 Uncoupling of Simultaneous Equations with Cyclic Periodicity for Variables with Two Subscripts 1.3 Uncoupling of Simultaneous Equations with Cyclic Bi-periodicity 1.3.1 Cyclic Bi-periodic Equation 1.3.2 Uncoupling of Cyclic Bi-periodic Equations 1.3.3 Uncoupling of Simultaneous Equations with Cyclic Bi-periodicity for Variables with Two Subscripts Chapter 2 Bi-periodic Mass-Spring Systems 2.1 Cyclic Bi-periodic Mass-Spring System 2.1.1 Static Solution 2.1.1a Example 2.1.2 Natural Vibration 2.1.2a Example 2.1.3 Forced Vibration 2.1.3a Example 2.2 Linear Bi-periodic Mass-Spring Systems 2.2.1 Bi-periodic Mass-Spring System with Fixed Extreme Ends 2.2.1a Natural Vibration Example 2.2. l b Forced Vibration Example 2.2.2 Bi-periodic Mass-Spring System with Free Extreme Ends 2.2.2a Natural Vibration Example 2.2.2b Forced Vibration Example 2.2.3 Bi-periodic Mass-Spring System with One End Fixed

...

MII

Contents

and the Other Free 2.2.3a Natural Vibration Example

Chapter 3 Bi-periodic Structures 3.1 Continuous Truss with Equidistant Supports 3.1.1 Governing Equation 3.1.2 Static Solution 3.1.2a Example 3.1.3 Natural Vibration 3.1.3a Example 3.1.4 Forced Vibration 3.1.4a Example 3.2 Continuous Beam with Equidistant Roller and Spring Supports 3.2.1 Governing Equation and Static Solution 3.2.2 Example Chapter 4 Structures with Bi-periodicity in Two Directions 4.1 Cable Networks with Periodic Supports 4.1.1 Static Solution 4.1.la Example 4.1.2 Natural Vibration 4.1.2a Example 4.1.3 Forced Vibration 4.1.3a Example 4.2 Grillwork with Periodic Supports 4.2.1 Governing Equation 4.2.2 Static Solution 4.2.3 Example 4.3 Grillwork with Periodic Stiffened Beams 4.3.1 Governing Equation 4.3.2 Static Solution 4.3.3 Example Chapter 5 Nearly Periodic Systems with Nonlinear Disorders 5.1 Periodic System with Nonlinear Disorders -Monocoupled System 5.1.1 Governing Equation 5.1.2 Localized Modes in the System with One Nonlinear Disorder 5.1.3 Localized Modes in the System with Two Nonlinear Disorders 5.2 Periodic System with One Nonlinear Disorder

Exact Analysis of Bi-periodic Structures

-Two-degree-coupling System 5.2.1 Governing Equation 5.2.2 Perturbation Solution 5.2.3 Localized Modes 5.3 Damped Periodic Systems with One Nonlinear Disorder 5.3.1 Forced Vibration Equation 5.3.2 Perturbation Solution 5.3.3 Localized Property of the Forced Vibration Mode References Nomenclature Index

ix

Chapter 1 U TRANSFORMATION AND UNCOUPLING OF GOVERNING EQUATIONS FOR SYSTEMS WITH CYCLIC BI-PERIODICITY 1.1 Dynamic Properties of Structures with Cyclic Periodicity

1.1.1 Governing Equation In general, the discrete equation for cyclic periodic structures without damping may be expressed as

where a superior dot denotes differentiation with respect to the time variable t, K and M are stiffness and mass matrices and X and F are displacement and loading vectors respectively. Generally they can be written as

and

where N represents the total number of substructures; the vector components xi and Fj ( j = 42, ...,N ) denote displacement and loading vectors for the j-th substructure, respectively. The numbers of dimensions of submatrices K , , M ,

2


( r ,s = 42,. ..,N ) and vector components x, and Fj ( j = 1,2,.

..,N ) are the same

as the degrees of freedom for a single substructure and let J denote the number of degrees of freedom of a substructure. The stiffness and mass matrices for the cyclic periodic structures possess cyclic periodicity as well as symmetry, namely

IT

where [ denotes the transposed matrix of [ ] . The simultaneous equation (1.1.1) with K, M having cyclic periodicity may be called a cyclic periodic equation. 1.1.2 U Matrix and Cyclic Matrix

Let

with the submatrices

U Transformation and Uncoupling of Governing Equations

3

in which y = 2 ; l r / ~ , i = f i and I, denotes the unit matrix of order J. It can be shown that

That leads to

where the superior bar denotes complex conjugation. U satisfying Eq. (1.1.10) is referred to as unitary matrix or U matrix. Eq. (1.1.10) indicates that the column vectors of U are a set of normalized orthogonal basis in the unitary space with N .J dimensions. The columns of U , are made up of the basis of the m-th subspace with J dimensions. An arbitrary vector, say U,xm ( x , is a J dimensional vector), in the m-th subspace possesses the cyclic periodicity. If

represents a vibration mode for a cyclic periodic structure with N substructures, then this mode is a rotating one, namely the deflection of one substructure has the same amplitude as, and a constant phase difference m y (= 2mnlN) from, the deflection of the preceding substructure. y is referred to as the period of the cyclic periodic structure. All of the rotating modes, the phase difference between two adjacent substructures must be 2 m / N ( m = 1,2,. .., N ) due to cyclic periodicity. As a result, all of the mode vectors lie in the N subspaces respectively.

4


A matrix with cyclic periodicity shown in Eq. (1.1.5) is referred to as cyclic matrix, such as the stiffness and mass matrices of structures with cyclic periodicity are cyclic matrices. The elementary cyclic matrices can be defined as

where the empty elements are equal to zero, Eo is a unit matrix and each element of matrix cj is a J dimensional square matrix. An arbitrary cyclic matrix can be expressed as the series of the elementary cyclic matrix, such as

and

where

E~

=

c0 and

denotes the quasi-diagonal matrix, i.e.,


Noting the cyclic periodicity of Urnand

it can be verified that

with

It is obvious that O j ( j

= 0,1,2,. .., N- 1) is

a diagonal matrix. One can now apply Eqs. (1.1.13) and (1.1.16) to derive the important formula:

5


6

The matrix k, is a Herrniltian one, i.e.,

k-T, = k , In the same way, we have

and

in, is also a Hermiltian matrix.

1.1.3

U Transformation and Uncoupling of Simultaneous Equations with Cyclic Periodicity

The U transformation can be defined as


7

where X , U are defined as Eqs. (1.1.3a) and (1.1.8) respectively and

and q , ( m = 1,2,. ..,N ) are vectors of dimension J. Because the coefficient matrix in Eq. (1.1.22) is an unitary matrix satisfying Eq. (1.1.10), the complex linear transformation (1.1.22) is referred to as U transformation. Recalling Eq. (1.1.1O), premultiplying both sides of Eq. (1.1.22) by ET,the inverse U transformation can be obtained as

The component forms of the U and inverse U transformations are

and

with y/=2n/N and i=n. Usually, the original variables x, ,x,, ,x, are real vectors representing the displacement vectors of substructures for a cyclic periodic structure and q, ,q2,-..,qN are a set of generalized displacement vectors. Noting Eq. (1.1.25b) and x j being real vector, it can be shown that

8

Exact Analysis ofBi-periodic Structures

and qN, qNIZ(if N is even) = real vector

(1.1.26b)

Applying the U transformation (1.1.22) to Eq. (1.1. I), namely substituting Eq. (1.1.22) into Eq. (1.1.1) and premultiplying both sides of Eq. (1.1.1) by U T, we have

where

Eq. (1.1.27) is made up of N independent equations, i.e., mrq, + krqr = f, , r

= 1,2,..., N

(1.1.29)

Noting the definitions of m,, k, and f, shown in Eqs. (1.1.21), (1.1.18) and (1.1.28) respectively, it is obvious that

and m,, m,,, (if N is even), k,, k,,, (if N is even) are real symmetric matrices, so 9,-, = ijr and q , , q,,, (if N is even) are real vectors. N N+l We need only consider -+ 1 (N is even) or - (N is odd) equations, i.e., 2 2 N N-1 r=1,2 ,...,-,N (Niseven)or r=1,2,...,-,N (N is odd) in Eq. (1.1.29). 2 2


9

1.1.4 Dynamic Properties of Cyclic Periodic Structures Consider now the natural vibration of rotationally periodic structures. The natural vibration equation can be expressed in terms of the generalized displacements as

where w denotes the natural frequency, q, represents the amplitude of the r-th generalized displacement and k , , m , denote generalized stiffness and mass matrices as shown in Eqs. (1.1.18) and (1.1.21) respectively. It is well known that the eigenvalues of the eigenvalue equation (1.1.3 1) with Hermiltian matrices are real numbers. The eigenvalues can be denoted as w,?,, 2 w : , ~ , , w : , ~ ( w,,, Ia:,+,,s = 1,2,. ..,J - 1) and the corresponding normalized orthogonal eigenvectors may be written as q,,, , q r , 2 , ..., ( I , , ~They . satisfy the

eigenvalue equation and the normalized orthogonal condition, i.e.,

and

leading to w:,, = ijrTskrq,,, = real number

Noting kN-,

-

= k, ,

and qN,S, q -,s

(1.1.34)

-

mN_, = m , , it is obvious that

(if N is even) s = 1,2,...,J are real eigenvectors.

Let us consider the natural modes. Corresponding to the eigenvector qNSs ( s = 42,. ..,J ), the natural mode can be expressed as

10


X = U~4iv.s

leading to

The vibrating displacements of all substructures possess the same amplitude vector and vibrating phase. Corresponding to the eigenvector q N (if N is even), the natural mode is -,s

which leads to

so the displacement vectors for any two adjacent substructures are equal and opposite. Such a mode doesn't occur when N is odd. ..

N

For other natural frequencies w r , ( r # N,-;

2

repeated frequencies: w , , = u,-,,~ (r

N + N , -; 2

s = 1,2,. ..,J ), which are

s = 42,. ..,J ), the corresponding

modes take the form as

This is a rotating mode, the deflection of any substructure has a constant phase difference 2 m / N from that for the preceding substructure, i.e., the mode lies in the r-th mode subspace. The real and imaginary parts of each rotating mode are two independent standing modes corresponding to the same natural frequency.


11

The rotating mode corresponding to o,-,,, is

The phase difference between two adjacent substructures is -2mlN. A pair of rotating modes shown in Eqs. (1.1.38) and (1.1.39) corresponding to the same natural frequency are complex conjugate modes and their rotating directions are opposite. Generally, the natural frequencies for periodic structures are densely distributed in pass bands. The number of pass bands is in agreement with that of the degrees of freedom for a single substructure. For the case under consideration, there are J pass bands altogether. In fact, w,,( r = 1,2,...,N ) lie in the s-th pass band. There are N natural frequencies lying in a pass band. If N is a large number, the natural frequencies are densely distributed in every pass band. It is of interest to note that the natural frequencies obtained from Eq. (1.1.31) with a given r are dispersed, namely they lie in different pass bands. The upper and lower bounds of the pass bandscanbe foundby solving Eq. (1.1.31)with r = N ( r y l = 2 r ) a n d r = N l 2 ( r ry = r ), respectively. 1.2

Uncoupling of Simultaneous Equations with Cyclic Periodicity for Variables with Two Subscripts

1.2.1 Double U Transformation If a structure considered possesses cyclic periodicity in two directions, say x and y directions, the governing equation can be uncoupled by using U transformation

two times (in x and y directions respectively). Let x(,,,) denote the displacement vector of substructure ( j , k) ( j = 1,2,...,M ;

k = 1,2,...,N ) and the subscriptsj and k denote the ordinal numbers of substructure along x and y directions respectively. Firstly the substructure with cyclic periodicity in x and y directions can be regarded as cyclic periodic in x direction. The corresponding U transformation can be expressed as

12


where M and N denote the total numbers of substructures in x and y directions respectively and y , = 2 z / M denoting the period of the structure in x direction. The inverse U transformation is

The governing equation in terms of Q(,,,, ( r = 1,2,.

..,M

; k = 1,2,.

..,N ) also

possesses cyclic periodicity in y direction for the second subscript k. Secondly we can introduce again the U transformation into the governing equation with the unknown variables Q,,( ( r = 1,2,...,M ; k = 1,2,. ..,N ). The corresponding U and inverse U transformations can be expressed as

and

U Transformationand Uncoupling of Governing Equations

13

with y, = 2n/N denoting the period of the structure in y direction. Substituting Eq. (1.2.3a) into Eq. (1.2.lb) yields

with y, = 2n/M and y, = 2 n l N . Eq. (1.2.4a) is referred to as double U transformation, which may be regarded as an extension of the U transformation in two dimensional problems. Its inverse transformation can be obtained by introducing Eq. (1.2.2b) into Eq. (1.2.3b) as

1.2.2

Uncoupling of Simultaneous Equations with Cyclic Periodicity for Variables with Two Subscripts

Consider now the simultaneous equations having cyclic periodicity for variables with two subscripts. Such equations can be expressed as

possesses cyclic periodicity for subscriptsj, u and k, v respectively, where K(j,k,(u,v, namely

14


and

Usually the equilibrium equation of structures with cyclic periodicity in two directions takes the form as Eqs. (1.2.5) and (1.2.6) where x ( ~ , and ~ ) F(j,k) represent the displacement and loading vectors of substructure (j,k) and K(j,k)(u,v) denotes the stiffness coefficient matrix. satisfying Eq. (1.2.6) can be uncoupled by using the Eq. (1.2.5) with K(j,k)(u,v) double U transformation shown in Eqs. (1.2.4a) and (1.2.4b). Premultiplying both sides of Eq. (1.2.5) by the operator 1 e - i ( j - l ) r v-i(k-1)s" el gives

fiJN

7, j=l

t=l

and

Noting the cyclic periodicity of K(j,,,(,,v,shown in Eq. (1.2.6), we have K(j,k)(u.v)

- K(j',k')(l,O

(1.2.9a)

with j f =j - u + 1 ,

kf=k-v+l

Introducing Eqs. (1.2.9) and (1.2.4b) in Eq. (1.2.7) results in

(1.2.9b)


15

Thus the simultaneous equations (1.2.5) have been uncoupled into M times N independent equations.

1.3 Uncoupling of Simultaneous Equations with Cyclic Bi-periodicity*

1.3.1 Cyclic Bi-periodic Equation

If the coefficient matrix of simultaneous equations is a superposition of two matrices with cyclic periodicity and different periods, the simultaneous equations are referred to as cyclic bi-periodic. The simple form of cyclic bi-periodic equation to be considered can be expressed as

where K,., ( r, s = 1,2,...,N ) possess cyclic periodicity, i.e.,

and

with the other K ; , vanishing and N = np . In physics,

KO

represents the

* See L. Gao and J.K. Liu, Uncoupling of governing equations for cyclic bi-periodic structures, Advances in Structural Engineering, An International Journal, Vol. 4, No. 3, 137-146 (2001).

16


difference between the stiffness matrices of the [1 + (k - l)p] -th ( k = 1,2,. ..,n ) substructure and the other substructures. The j-th component equation of Eq. (1.3.1) is

Considering Eq. (1.3.2), Eq. (1.3.3) can be rewritten as

where x-,

,

= xN-, ( s = 0,1,2,. ..,N - 1) and the term ( - K Ox ) is treated as load as

well as F, . The coefficient matrices K,,, ( u = 42,. ..,N ) are the same in everyone of Eq. (1.3.4) ( j = 1,2,. ..,N ). It is a characteristic of cyclic periodic equations.

1.3.2 Uncoupling of Cyclic Bi-periodic Equations Applying the U transformation

and


17

with ly = 2n/N to Eq. (1.3.4), i.e., premultiplying both sides of Eq. (1.3.4) by the operator

L

JN

N

,and considering

e-i(j-l)m v j=,

r = +l,f2,...,fN

we have amqm= fm +f O , 3

m =1,2, ...,N

where

Eq. (1.3.7) leads to 4, = a i l ( f m+ fmO) Inserting Eq. (1.3.10) in the U transformation (1.3.5a), we obtain x j = x :I + x O ] ,

where

j = 1 , 2,..., N

18


Here x; denotes the solution of Eq. (1.3.1) with

KO

vanishing and x9 denotes

the influence of K O on the solution x j . Note that x; is dependent on x,+(,-,,~ ( u = 1,2,...,n ). Introducing the notations

and inserting j = 1+ (s - l)p into Eqs. (1.3.11) and (1.3.12b) gives

where

The coupled terms in simultaneous equations (1.3.14) possess cyclic periodicity, i.e.,

By using the U transformation once, the cyclic bi-periodic equation (1.3.1) with N unknown vectors becomes the cyclic mono-periodic equation (1.3.14) with n ( = N/p ) unknown vectors. In order to uncouple Eq. (1.3.14), the U transformation needs to be used again. For the present case, the corresponding U and inverse U transformation can be given as


19

and Q

1 =-

J;I

" ii(s-l)rc X s , r =172,...,n

(1.3.17b)

s=l

with q = 2 z / n = p y / . 1 " Premultiplying both sides of Eq. (1.3.14) by the operator - e-i's-')rcand & ,=I considering Eq. (1.3.16), we have

where

Noting that

and substituting Eq. (1.3.20) into Eq. (1.3.18) we obtain ArQr=br,

r = l , 2,...,n


20

and I denotes unit matrix. By means of U transformation twice the cyclic bi-periodic equation (1.3.1) has been uncoupled into a set of independent equations shown in Eq. (1.3.22) where everyone includes only one unknown vector. The number of unknowns in Eq. (1.3.22) with given r is in agreement with that of degrees of freedom for a single substructure. Recalling Eq. (1.3.13) and rp = p iy and making a comparison between Eqs. (1.3.9b) and (1.3.17b) gives

When the specific parameters in Eq. (1.3.1) are given, the solution for xi ( j = 1,2,...,N ) can be found by using the relevant formulas derived above.

1.3.3

Uncoupling of Simultaneous Equations with Cyclic Bi-periodicity for Variables with Two Subscripts

The simultaneous equations to be considered take the general form as


21

and K (j,k)(u,v) ( j , u = 1,2,. ..,M ; k, v = 1,2,. ..,N ) possess cyclic periodicity for two pairs of subscripts:j, u and k, v, respectively, namely

and K(j,k)(u,l)

- K(j,k+l)(u,Z)

- ". = K(j , N ) ( u , ~ - k + l ) - K(j,l)(u,N-k+2)

j,u=l,2. M ;

=

k=1,2,...,N

"

'

= K(j,k-l)(u,~)

(1.3.27b)

Introducing Eq. (1.3.27) and the notations X(-u,v)

-

= X(~-u,v)

-

,

=N - v )

u=0,1,2 ,..., M ;

X(-u,-v)

-

= X(M-~,N-~)

v=0,1,2,..., N

into Eq. (1.3.25), it can be rewritten as

j # 1 , 1 + p l , l + 2 p ,,...,l+(m-l)p, or k + l , l + ~ ~ , l ,..., + 2l +~( ~ n-l)p2 =

2. M ;

Applying the double U transformation

k=1,2,...,N

(1.3.29b)

22


u=1,2 ,...,M ;

v = 1 , 2 ,...,N

(1.3.30b)

with ry, = 2 z l M and ry, = 2 z / N to Eq. (1.3.29),i.e., premultiplying Eq. (1.3.29) by the operator

e-i(j-l)uK

fifi

e -i(k-I),,

and noting that

j=l

we have a(,,,)q(,,,)= f(u,v) + f ( : , v ) , where

u = 1,2,...,M ;

v = 1,2,...,N

(1.3.32)


23

From Eq. (1.3.32), q(,,,, can be formally expressed as

f(3

-1

(1.3.35)

9(u,v) = a(u,v)(f(u,v) +

Substituting Eq. (1.3.35) into the double U transformation (1.3.30a) yields

+

0

x ( ~ ,=~x, ; ~ , ~x)( , , ~ ),

j = 1,2,..., M ; k = 1,2,...,N

(1.3.36)

where

and

Here x ; , , ~ ,denotes the solution of Eq. (1.3.25) with represents the influence of

KO

KO

vanishing and xpjSk,

on the solution x ( ~ , ,., xpj,,, is dependent on

x ( ~ + ( ~ - ~ ) ~ (~r , = ~ 1,2,. + ( ~ ..,m - ~; ) ~s ~ = )1,2,. ..,n)

Introducing the notations

in Eqs. (1.3.36) and (1.3.37b), we obtain

to be determined.

24


in which

with 9, = p l y l = 2z/m and 9, = p 2 y 2 = 2z/n. It is obvious that $ j , k ) ( r , s ) ( j , r = 1,2,...,m ; k, s = 1,2,...,n ) possess cyclic periodicity for the two pairs of subscripts 0, r and k, s), namely

and

The term Xij,,, on the right side of Eq. (1.3.39) can be found by means of the formulas (1.3.33), (1.3.34a), (1.3.37a) and (1.3.38). The set of simultaneous equations (1.3.39) possess cyclic periodicity for two subscripts. It takes the same form as that of Eq. (1.2.5). As a result Eq. (1.3.39) can be uncoupled by means of the double U transformation (see section 1.2.2). Let


with 9,= 2 r / m and 9,= 2 n l n . Premultiplying both sides of Eq. (1.3.39) by the operator 1 " " e-i(i-l)une - i ( k - l ) n results in

rJ m n

j=l

k=l

where

and

Substituting Eq. (1.3.46) into Eq. (1.3.44) and noting that ( 1 j =u,u+m;..,u +(p, -1)m and

25

26


results in

By applying the U transformation twice, the cyclic bi-periodic equation (1.3.25) with M x N unknown vectors is uncoupled into a set of independent equations shown in Eq. (1.3.43) where each includes only one unknown vector. Making a comparison between Eqs. (1.3.34b) and (1.3.42b), we have

where Q(,,+,,( u = 1,2,..., m ; v = 1,2,...,n ) can be found from Eq. (1.3.43). When the specific equation (1.3.25) is given, the solution for x ( ~ , ~ ) ( j= 1,2,...,M ; k = 1,2,...,N ) can be obtained from Eqs. (1.3.49), (1.3.34a), (1.3.35) and (1.3.30a).

Chapter 2 BI-PERIODIC MASS-SPRING SYSTEMS The physical meaning and mathematical formulation of the U-transformation have been described in reference [I]. Essentially the U-transformation method is a mode subspace method [2] for a cyclic periodic structure. In mathematics the Utransformation is an orthogonal linear transformation with complex coefficients. The linear simultaneous equations (algebraic or differential equations) with cyclic periodicity can be uncoupled by the U-transformation technique. The proof is given in the first Chapter. It is obvious that the governing equation of a cyclic periodic structure possesses the cyclic periodic property. Therefore the U-transformation can be applied to analyze cyclic periodic structures [3,4]. Furthermore the application of the U-transformation may be extended to analyze linear periodic structures [5-91, if its equivalent system with cyclic periodicity can be formed. What is referred to as a bi-periodic structure in this book is a structure consisting of two different sets of periodic properties. Certainly a bi-periodic structure may be regarded as a single periodic one but its substructure also possesses periodicity and many degrees of freedom. In order to utilize fully the property of biperiodicity, the proposed analysis method in this book requires the application of the U-transformation twice. As a result, the governing equation for a system with cyclic bi-periodicity may be fully uncoupled. That leads to the explicit analytical solution, which plays an important role in the optimal design or sensitivity analysis. 2.1

Cyclic Bi-periodic Mass-Spring System*

To illustrate the proposed method a simple model [lo] with cyclic bi-periodicity is analyzed. A general model, cyclic bi-periodic mass-spring system, is illustrated in Fig. 2.1.1 where all the coupling springs are of the same stiffness k ; K and K + AK denote the stiffness for two kinds of cantilever beams; M and M +AM denote two kinds of the lumped masses; and F, , x , denote the load, displacement for the r-th subsystem.

See L. Gao and J.K. Liu, Exact analytical solutions for static and dynamic analyses of cyclic biperiodic structures, Advances in Structural Engineering, An International Journal, Vol. 4, No. 3, 1471.58 (2001).

28


\ \ - ' p + l JP' \ M

k

Figure 2.1.1 Rotationally bi-periodic mass spring system 2.1.1 Static Solution

The equilibrium equation can be expressed as

j

+ 11+ p . 1+ (n - 1

and

j =1,2,. ..A

(2.1.1b)

where x,+, = x, ; X , = X, and 1+ (s - l)p (s=1,2,. ..,n) indicate the ordinal on the right numbers of the subsystems with stiffness K + AK . The term side of Eq. (2.1.la) may be treated as the load as well as Fj . One can now apply the U-transformation [3,5] to Eq. (2.1.1). The U- and inverse U-transformation may be defined as

Bi-periodic Mass-Spring Systems

29

and

in which y/ = 2z/N, i = f i and N = total number of subsystems, for present case N =pn. The equilibrium equation (2.1.l) may be expressed in terms of the generalized displacements q m (m= 1,2,...,N) as

where

The generalized displacement q m in Eq. (2.1.3) may be formally expressed as

Substituting Eqs. (2.1.5) and (2.1.4) into Eq. (2.1.2a) yields

30


1 x"-Cei(~-')m*q2

fi

M=I

=

AK ei(j-l)mye-i(u-l)mpy -NCF, n

N

U=I

M=I

K + 2k(l- cos m v)X1*'u-l'p

(2.1.7a)

where x; (j=1,2, ...,N) represents the solution for the perfect periodic system (i.e., AK = 0 ) subjected to the same loading as that acting on the bi-periodic system. When the specific loading condition is given, x; (j=1,2,. ..jV)can be obtained from Eqs. (2.1.4b), (2.1.5~)and(2.1.7b) Inserting j = 1+ (s - 1)p (s= 1,2,. ..,n) in Eqs. (2.1.6) and (2.1.7) gives

where

and

P,, denotes the influence coefficient for the single periodic system. By using the U-transformation once, the equilibrium equation (2.1.1) with N (= pn) unknowns becomes Eq. (2.1.8) with n unknowns. Note that the simultaneous equations (2.1.8) possess the cyclic periodicity, i.e.,

31


One can now apply the U-transformation again to Eq. (2.1.a). Introducing

and

with y, = 2n/n = py/ ,into Eq. (2.1.8)results in n

Q, = -MZps,le-i(s-l)rqQ, + b , ,

r=1,2 ,...,n

(2.1.13)

where

The governing equation (2.1.1) has been uncoupled and becomes a set of one degree of freedom equations by using the U-transformation twice. Substituting Eq. (2.1.10) into Eq. (2.1.13)and noting the identical relation

m = 1,2,...,N ( = pn) ;

we have

and

r

= 1,2

,..., n

(2.1.15)

Exnct Analysis of Bi-periodic Structures

32

Inserting Eqs. (2.1.16) and (2.1.17) into the later U-transformation (2.1.12a) yields

in which ty = 27r/N, p = 2n/n and N = pn . Making a comparison between Eqs. (2.1.4a) and (2.1.12b), we have

Finally the solution for xj of Eqs. (2.1.la) and (2.1 .lb) can be found by substituting Eqs. (2.1.9, (2.1.7), (2.1.19) and (2.1.4b) into Eq. (2.1.6). In order to explain the computational procedure and to verify the exactness of the formulas derived above, we need to consider a specific system and loading as an example. 2.1.la

Example

The systemic parameters and loads are given as n p=3, n=2 (as a result N=6, q~= n and y = - ) 3

and

F, = F, = P,

Fj = 0, j

Inserting Eq. (2.1.20) into (2.1.4b) yields

+ 1,4

(2.1.20a)


33

and then introducing Eqs. (2.1.21) and (2.1.20a) into Eq. (2.1.5~)gives 2P q:, = &K

1

+ 2k(l- cos rn n/3)

m = 2,4,6

(2.1.22a)

Substituting Eqs. (2.1.20a) and (2.1.22) into Eq. (2.1.7b), we have

It can be verified thatx; (j=1,2,...,6) is the exact displacement solution for the system with AK = 0 subjected to the loads shown in Eq. (2.1.20b). Recalling the definition shown in Eq. (2.1.9) andp=3 gives

Inserting Eqs. (2.1.24) and (2.1.20a) into Eq. (2.1.14) yields

and then introducing Eqs. (2.1.25), (2.1.20a) and (2.1.17) into Eq. (2.1.16) results in

From Eq. (2.1.19),

fi (m=1,2,.. .,6) can be obtained as


34

Finally, substituting Eqs. (2.1.5b), (2.1.27) and (2.1.20a) into Eq. (2.1.7a) we have 0

XI

0

X2

0

= X 40 = -

0

AK(K + k ) 2 ~ K(K + 3k)[K(K + 3k) + AK(K + k)]

0

= X 3 = X5 = X 6 = -

AK k(K + k)P K(K + 3k)[K(K + 3k) + AK(K + k)]

(2.1.28a)

(2.1.28b)

and then inserting Eqs. (2.1.28) and (2.1.23) into Eq. (2.1.6) results in

The displacement x j ( j = 1,2,. ..,6 ) shown in Eq. (2.1.29) satisfies the equilibrium equations (2.1.l) with the parameters shown in Eq. (2.1.20). 2.1.2

Natural Vibration

The natural vibration equation for the cyclic bi-periodic system shown in Fig. 2.1.1 may be expressed as

where w denotes the natural frequency, x j denotes the amplitude of j-th subsystem and the term - (AK - A M w 2 ) x j may be formally treated as the load. Applying the U-transformation (2.1.2) to Eq. (2.1.30) results in


( ~ + 2 k - ~ u ~ ) ~ ~ - 2 k c o s m l ym=1,2, q ~ =..., fN ~ ,

35

(2.1.31)

where

and then

Substituting Eqs. (2.1.32) and (2.1.33) into Eq. (2.1.2a) yields

x. =-

(AK - MU') N

Introducing the notation X , in Eq. (2.1.34) we have

ei(j-l)mve-i(u-l)mpy

2U=I2 K m =+~2 k - M u 2

-2kcosmy

= x,+(,-,,, and inserting j

XI+(~-~)~

= 1+ (s - l)p (s=1,2,.

..,n)

where

p,'

denotes the harmonic influence coefficient for the considered system with

AK=AM=o.

Obviously also possesses cyclic periodicity. Applying the Utransformation (2.1.12) to Eq. (2.1.35) results in

Exacr Analysis ofBi-periodic Structures

36

Substituting Eq. (2.1.36) Eq. (2.1.15), we have

and p y/ = p

into Eq. (2.1.37) and recalling

When Q, is non-vanishing, the frequency equation can be expressed as

,-,,,

.

= 0 (s=1,2,. .,n), When Q, (r=1,2,. ..,n) are identically equal to zero, i.e., x,+( the corresponding frequency equation can be obtained from Eqs. (2.1.31) and (2.1.32) as

2m =kn (niseven), k =1,2,..., p-1, 2m = 2kn (n is old), k = 1,2,..., (k < p 12)

(2.1.40)

where 2m is in agreement with the half wave number of the mode. In order that all mass points having mass M + Ah4 lie in the nodal points of the mode, the half wave number 2m must be equal to an integer times n and less than N. 2.1.2a

Example

The parameters are given as


p=3,

n=2,

AM=M,

AK=K

37

(2.1.41a)

That leads to

The frequency equation (2.1.39) becomes

r = 1,2 The solution for w of Eq. (2.1.42) can be found as

w

=

K+(2-fi)k M

7

~+(2+fi)k , (for r = 1 ) M

and K K + 2 k , (for r = 2 ) M' M

w =-

These natural frequencies are corresponding to the modes with x, and x, nonvanishing. Consider the other frequency equation (2.1.40), i.e.,

The square of frequency can be expressed as

x, and x, are identically equal to zero in the corresponding modes. Consider now the natural modes. Corresponding to the natural frequencies

38


shown in Eq. (2.1.43a), the modes can be expressed as

Substituting Eqs. (2.1.46) and (2.1.41) into Eq. (2.1.12a), we have

where an arbitrary constant factor is neglected. Introducing Eqs. (2.1.47), (2.1.41) and w

=

~+(2-fi)k into Eq. (2.1.34), M

the natural mode can be found as

Substituting Eqs. (2.1.471, (2.1.41) and

02=

+(2 +fi)k M

into Eq. (2.1.34)

results in

Similarly, corresponding to the natural frequencies shown in Eq. (2.1.43b), the modes in terms of the generalized displacements can be expressed as

That leads to XI

= 1,

X4

=1

Corresponding to w 2 = KIM ,the natural mode is

,

x . = l,

and corresponding to

02= (K

j=1,2 ,...,6

+ 2k)lM ,the natural mode is


39

Consider the other kind of modes with x, = x, = 0.The mode in terms of the generalized displacements can be expressed as

with the other qs vanishing. There are two independent modes corresponding to the same natural frequency for the cyclic periodic system. We are only interested in the mode with x , and x, vanishing. By substituting Eqs. (2.1.53) and (2.1.41) into Eq. (2.1.2a) and letting X, = 0,results in q, = imaginary number and

where an arbitrary constant factor is also neglected. K+k K+3k , the natural modes can be found Corresponding to w = - and M M by introducing m=l and 2 into Eq. (2.1.54) respectively as

and

2.1.3

Forced Vibration

The forced vibration equation for the system shown in Fig. 2.1.1 subjected to harmonic forces may be expressed as

40


with x, = x,, x, = x, due to cyclic periodicity. In Eq. (2.1.57), Fj and xj denote the amplitudes of the loading and displacement for the j-th subsystem and denotes the frequency of the harmonic loads. Applying the U-transformation (2.1.2) to Eq. (2.1.57), yields

where

From Eq. (2.1.58), q , can be formally expressed as

9; = f i / ( K + 2 k - M w 2 -2kcosrny)

(2.1.60~)

Substituting Eqs. (2.1.59), (2.1.60) and j = 1 + (s - 1)p into Eq. (2.1.2a), results in


41

where

Applying the U-transformation (2.1.12) to Eq. (2.1.61) yields a,(u)Q,=b,

r=1,2 ,..., n

(2.1.65)

where

Recalling Eqs. (2.1.59a) and (2.1.12b), we have

fP+(u-l)n

=-

(AK - AMu2) Q, r

fi

=,

. . . ,n

u = 1,2,..., p

(2.1.69)


42

Finally substituting Eqs. (2.1.60), (2.1.69) into Eq. (2.1.2a), the exact solution can be expressed as

0 qr+(u-1)n

-

--

(AK - AMw2)

Qr

K + 2k - M

W ~ - 2k cos[r

+ (u - l)n]y

and x; can be found from Eqs. (2.1.64), (2.1.60~)and (2.1.59b) if the loading is given. 2.1.3a

Example

The same system shown in Fig. 2.1.1 is considered. The structural and loading parameters are given as

and F, = F, = P,

F2 = F, = F,

= F, = 0

Inserting Eq. (2.1.71) into (2.1.59b) gives

and then substituting Eqs. (2.1.60c), (2.1.72) and (2.1.71a) into Eq. (2.1.64) results in


43

leading to

Inserting Eqs. (2.1.74) and (2.1.71a) into Eq. (2.1.67) gives

Substituting Eqs. (2.1.66), (2.1.71a) and (2.1.75) into (2.1.68) yields

That leads to

Finally substituting Eqs. (2.1.60b), (2.1.77) and (2.1.7la) into Eq. (2.1.70b), we have

44


and then superposing this solution on x; ( j = 1,2,...,6) shown in Eq. (2.1.73), results in

This solution represents the steady state response. When w approaches zero the solution shown in Eq. (2.1.79) approaches the static one shown in Eq. (2.1.29) with K K+2k AK = K . When w 2 approaches - -, xj ( j = 1,2,...,6) will approach M' M K K+2k infinity. Corresponding to w = - and -, two natural modes possess the

M

M

property of x, = x4 # 0 . However, there are six natural frequencies for the considered system. It can be shown that, when x, approaches a finite value at one natural frequency, the work done by the external forces due to the displacement of the corresponding natural vibration is identically equal to zero, namely x, = -x4 including x, = x4 = 0 .

2.2 Linear Bi-periodic Mass-Spring Systems The U-transformation method is applicable to static and dynamic analyses of cyclic mono-periodic and bi-periodic systems. We can not directly apply the Utransformation method to analyze linear periodic systems. If the equivalent system with cyclic periodicity can be formed, the U-transformation method can be applied to the analysis of the equivalent system instead of the original linear periodic one. The following sections will illustrate how this can be done.

2.2.1 Bi-periodic Mass-Spring Systems with Fixed Extreme Ends Consider a bi-periodic mass spring system with fixed extreme ends and np - 1 mass points as shown in Fig. 2.2.1(a), where M and M + AM denote the masses of

Bi-periodic Moss-Spring Systems

45

two kinds of periodic particles respectively and k denotes the stiffness constant of all coupling springs. Assuming that the mass points can move only along the longitudinal direction. x, and F, denote the longitudinal displacement and load for mass point j. The fixed extreme end condition can be expressed as x, = 0 and x,+, = 0. The equivalent system with cyclic bi-periodicity must satisfy these restrained conditions. Such an equivalent system can be achieved by the following procedures. First, the mass-spring system is extended by its symmetrical image and the symmetric loading is applied on the corresponding extended part as shown in Fig. 2.2.l(b). In order to form a cyclic bi-periodic system, it is necessary that two fixed ends of the original system can be replaced by the particles with mass M + AM without any restriction and two extreme ends of the extended system may be imaginarily regarded as the same particle with mass M +AM , namely the first mass point is next to the 2np-th one. Because of the symmetry of the extended system and corresponding loading, the displacements of the symmetric centers must be equal to zero, i.e., x, = 0 and x,, = 0 . As a result the fixed end conditions of the original system can be satisfied automatically in its extended system. Therefore both systems shown in Fig. 2.2.l(a) and (b) are equivalent. Such an extended system with cyclic bi-periodicity is called equivalent system which can be analyzed by the U-transformation method. Consider now the harmonic vibration for the equivalent system shown in Fig. 2.2.l(b). The governing equation can be expressed as

j

+ 1,1+p,...,1+(2n -1)p

and j = 1,2,...,2np

(2.2.lb)

where xj and F, denote the amplitudes of the longitudinal displacement and load for the j-th mass point; w denotes the vibration frequency; 1+ (m - l)p ( m = 1,2, ...,2n ) is the ordinal number of the subsystem with mass M +AM and the loads F, ( j = 1,2,...,2np) must satisfy the symmetric condition, i.e.,

(a) Original system

symmetric line

,.-.+

FJ

I

-Fj

--l

(b) Equivalent system Figure 2.2.1 Bi-periodic mass spring system with fixed extreme ends


47

where N = np and F i ( j = 2,3,..., N) denote the real loads acting on the original system. The term AMw2xj acts as a kind of loading. The expressions on the left sides of Eqs. (2.2.la) and (2.2.lb) possess cyclic periodicity. We can now apply the U-transformation to Eq. (2.2.1). Let

7T

where y = -, i = f i and 2N denotes the total number of subsystems. N 1 ZN Prernultiplying Eq. (2.2.1) by -- e*"-')"* ,results in

Juv

j=l

( 2 k - ~ u ~ ) q-2kcosmyq, , = f; + f;

m = 1,2,...,2N

where f-" -

AMw2 - 2np

J2N

- " m p *

.=I

Introducing Eq. (2.2.2) into Eq. (2.2.5b) gives

xl+(U-~)~

(2.2.4)

48


As a result

f; = O and

fiN= O

From Eq. (2.2.4), q, can be expressed as 0

I

4, = 4 , + 4 ,

(2.2.6a)

q: = f : / ( 2 k - ~ w ~ -2kcosmy)

(2.2.6b)

q: = fi/(2k - M

(2.2.6~)

W ~- 2k cos m y )

Substituting Eqs. (2.2.5), (2.2.6) and j = 1+ (s - l)p into the U-transformation (2.2.3a), we have

where

x;(j = 1,2,...,2N) represents the solution of the perfect periodic system with AM vanishing. It can be expressed as

Because

p,,

possesses cyclic periodicity, as

shown in Eq. (2.1.11), the


49

simultaneous equations (2.2.7) can be uncoupled by using the U-transformation. Let

and

with

(D

= z/n and i =

fi. Because of

,,-,= 2 .

X, being real value, Q

1 2n Premultiplying two sides of equation (2.2.7) by -Ce-"'-""

J2n

results in

=.I

where

X j must satisfy the symmetric condition, i.e., X;,+,-, = -X: and X,' = X:,, = 0 . As a result

leading to b,, = b2,,= 0 . Substituting Eq. (2.2.9) with u = 1 into Eq. (2.2.12) yields


50

Q2n-r= Q,

and

Q,

= Q2,

=0

where

Malung a comparison between Eqs. (2.2.5a) and (2.2.1 1b) gives

Finally substituting Eq. (2.2.6) into the U-transformation (2.2.3a), the solution for x j can be expressed as

and xi is defined in Eqs. (2.2.5c), (2.2.6~)and (2.2.10). 2.2.la

Natural Vibration Example

Letting F j =O ( j= 1,2,...,2 N ) as a result xl = O and b, =O ( r = 1,2,...,2n), the independent frequency equation can be obtained from Eqs. (2.2.14) and (2.2.15) as

if X , (s = 1,2,...,2n) are not identically equal to zero. Consider the case of X , (s = 1,2,...,2n) vanishing. As a result the terms on the


51

right side of Eq. (2.2.4) are equal to zero. Corresponding frequency equation can be obtained as

where rn denotes the half wave number of the mode for the original system. Taking a specific example as shown in Fig. 2.2.2.

Figure 2.2.2 Bi-periodic mass spring system with fixed ends, p=3 and n=2

The parameters are

leading to

Substituting Eq. (2.2.20)into Eq. (2.2.18)yields

A nondirnensional frequency parameter may be defined as

52


The frequency equation (2.2.21) can be rewritten, in term of Ro as

The roots of Eq. (2.2.23a) are

Ro = 0.198062264, 1.55495813,3.24697960

(2.2.23b)

Noting Eq. (2.2.20), the other frequency equation (2.2.19) becomes

That leads to

The two roots are

R, = l ( m = 2 ) ,

3(m=4)

There are five natural frequencies altogether. The total number of the natural frequencies is in agreement with the number of degrees of freedom for the original system. Consider now the natural modes. Corresponding to the frequency equation (2.2.21) with SZ, shown in Eq. (2.2.23b) the modes can be expressed in terms of the generalized displacements, as

with the other Q, vanishing. Introducing Eqs. (2.2.20) and (2.2.25) into (2.2.11a) and letting X, = 0 yields


Ql = imaginary number = -i

Q, = Ql = i

53

(2.2.26a)

neglecting an arbitrary constant factor in the expression of the mode. Inserting Eqs. (2.2.20a) and (2.2.26a) into Eq. (2.2.16) gives

Substituting Eqs. (2.2.20) and (2.2.27) into Eq. (2.2.17b), the natural mode can be obtained as

54


Obviously the modes depend on the nondimensional frequency parameter R, . For three values of R, shown in Eq. (2.2.23b), the numerical results of modes are given in Table 2.2.1.

x,lT Table 2.2.1 Natural modes [ x , x , x 3 for the system shown in Fig. 2.2.2 a - -

Mode X1

0

0

0

0

0

Corresponding to the frequency equation (2.2.24a) with R, shown in Eq. (2.2.24c), the natural modes can be expressed as

Introducing Eqs. (2.2.20b) and (2.2.29) into Eq. (2.2.3a), and letting x , results in

q , = imaginary number

=0

(2.2.30a)


55

where an arbitrary constant factor is neglected. The results for the modes shown in Eq. (2.2.30b) are also given in Table 2.2.1. All modes shown in Table 2.2.1 satisfy the fixed end conditions (x, = x, = 0) of the original system.

2.2.lb Forced Vibration Example Consider now the system shown in Fig. 2.2.2, subjected to a harmonic force Pei" acting at the center mass point. The parameters of the system and loading are given as z 7r n = 2 , p = 3 , A M = M , N = 6 , ty=- q = i 6'

(2.2.3 la)

and F, = P , I;I. = 0

j= 2,3,5,6

(2.2.31b)

Introducing Eq. (2.2.31) into Eq. (2.2.5~)yields

and then substituting Eqs. (2.2.6~)and (2.2.32) into Eq. (2.2.10) results in xi = X; = 0

(2.2.33a)

56


j = 2,3,...,6

x ; ~= - ~-x;.

with Ro = ~ w ~ /That k .leads to

XI3 -- 7 -0,

X ~ = X ~ ~ = - X ~

(2.2.34b)

Inserting Eqs. (2.2.31a) and (2.2.34a) into Eq. (2.2.13b) gives

Substituting Eqs. (2.2.15), (2.2.31a) and (2.2.35) into Eq. (2.2.14) we have

Q,=a,Q 2 = Q 4 = 0 From Eqs. (2.2.16), (2.2.3la) and (2.2.36),

fi can be obtained as

(2.2.36b)


57

Finally inserting Eq. (2.2.37) (2.2.17b) (2.2.33) and (2.2.31a) into Eq. (2.2.17a), the frequency response hnctions for the amplitudes of displacements can be found as

When R, approaches zero the solution xj shown in Eq. (2.2.38) will approach the static displacement, namely

Obviously the resonance frequencies are the roots of Eq. (2.2.23a). When R, approaches each value shown in Eq. (2.2.23b), the amplitudes of displacement will approach infmity. It can be shown that, when each amplitude of displacement approaches a finite value at one natural frequency, the harmonic force is acting at a nodal point of the corresponding mode.

2.2.2 Bi-periodic Mass-Spring Systems with Free Extreme Ends Consider a bi-periodic mass-spring system with free extreme ends and 2np @=2d+l) particles altogether as shown in Fig. 2.2.3(a) where M, AM ,k, xj , Fj and N have the same meanings as those in Fig. 2.2.1 .The bi-periodic system can also be regarded as single periodic. Each subsystem is made up of 2d particles with mass M and one particle having mass M + AM . The bi-periodic system shown in

(a) Original system

2N=2np

center line

!

p=2d+ I

FZN+I-j=Fj M

k

(b) Equivalent system Figure 2.2.3 Bi-periodic mass spring system with free extreme ends

a-

i


59

Fig. 2.2.3(a) has n subsystems and np @=2d+l) mass points altogether. In order to form an equivalent system having cyclic bi-periodicity, it is necessary to extend the original system by its symmetrical image and apply antisymmetric loading on the corresponding extended particles. In such an extended system the vibration displacements will be antisymmetric about the center line if the initial displacements and velocities are also antisymmetric. For each pair of symmetric particles, both longitudinal displacements are the same. As a result x, = x,, and x,, = x, which indicate that the connecting spring between both particles np and np+l dose not transmit any force and both extreme ends of the extended system can be imaginarily connected by the same spring as the other one, i.e., the 2np-th particle can be regarded as the preceding one of the first particle. Therefore the extended system shown in Fig. 2.2.3 (b) can be regarded as cyclic bi-periodic. The harmonic vibration equation takes the form as

j i t d +1,d + l + p,..., d +1+(2n-1)p and j =1,2,...,2N(= 2np) (2.2.40b) where 2N denotes the total number of particles for the equivalent system, p=2d+l, xi, Fj denote the amplitudes of the longitudinal displacement and loading for the j-th particles and w denotes the frequency of the external excitations. For the equivalent system, the loads must satisfy the antisymmetric condition, i.e.,

where N=np and Fj (j=1,2,. ..,np) are real loads acting on the original system . Applying the U transformation (2.2.3) to Eq. (2.2.40) we have

60


with i y = n / N and i = f i . Introducing Eq. (2.2.41) into Eq. (2.2.44a) yields

That leads to

f; G O The solution for q m of Eq. (2.2.42) can be formally expressed as

Substituting Eqs. (2.2.43), (2.2.45) and j U-transformation (2.2.3a), we have

where

=d

+ 1 + (s - l)p

(s=1,2,.. .,2n) into the


61

1 e"s-u)pmv/(2k- M C B-~2k cos rn ~y)] a,. =-C[ 2N 2N

,=I

and

Eq. (2.2.46) is similar to Eq. (2.2.7). Applying the U-transformation (2.2.11) to Eq. (2.2.46) results in

where y, = n/n and (2.2.5 la)

Xj must satisfy the antisymmetric condition, i.e., X;,+,-, = X: (s=1,2,. ..,n). As a result

that leads to

-

bn = O , b2,-, =b,

r=1,2,...,n-1

Introducing Eqs. (2.2.51~)and (2.2.48) into Eq. (2.2.50) gives

(2.2.5 1c)


62

Q,

E

0,

-

Q2,-,

= Q,

r = 1,2,...,n - 1

where

Making a comparison between Eqs. (2.2.43) and (2.2.11b) and noting the definition of X , shown in Eq. (2.2.47) yields

2.2.2a Natural Vibration Example Let Fj = 0 (j=1,2,. ..,2N). That leads to x; = 0 (j=1,2,...,2N) and b, = 0 (r=1,2,. ..,2n). If X , (s=1,2,...2n) are not identically equal to zero, the frequency equation can be expressed as a, ( w ) = 0 , namely

If X, (s=1,2,. ..,2n) are identically equal to zero the frequency equation can be obtained from Eq. (2.2.42) with f : = fi = 0 as 2k - ~

a

- 2k) cos~ m y

=0

m=n,3n,5n,. ..,@-2)n

(2.2.56)

It is interesting to note that Eqs. (2.2.55) and (2.2.56) formally are the same as Eqs. (2.2.18) and (2.2.19) respectively, but the parameters r and m take different values in corresponding Eqs. (2.2.55) and (2.2.18) and Eqs. (2.2.56) and (2.2.19) respectively. Consider the system with the following parameters


63

That leads to

The system considered is shown in Fig. 2.2.4 .

pe iot

pe iot

Figure 2.2.4 Bi-periodic mass spring system with free extreme ends, p=3 and n=2 Introducing Eq. (2.2.57) into Eq. (2.2.55) yields Mw2

1--z(2k 3 "=I

- MU'

-2kcos[r +4(u -1)lE)-' = 0 6

MU When r = 1 , the roots for Ro(= -) k shown in Eq. (2.2.23b), i.e.,

Ro = 0.198062264,

r = 1,4

(2.2.58)

of Eq. (2.2.58) are the same as those

1.55495813, 3.24697960

When r = 4 , the root for R, is

Inserting Eq. (2.2.57) into Eq. (2.2.56) gives

(2.2.59a)

64


That leads to

The corresponding mode possesses the property of x, = x, = 0. Let us now pay attention to the natural modes. Firstly consider the modes with x, and x, non-vanishing. Corresponding to the natural frequencies shown in Eq. (2.2.59a), the mode, in terms of the generalized displacement, is

Because of the antisymmetry of displacement for the equivalent system, namely X,,+,-, = X, (s =1,2,..., n),Eq.(2.2.11b)canberewrittenas

In order to make the mode antisymmetric. Ql must have a complex constant factor I.Z -

e

for the present case. Without loss of generality, let .?I

Q, = & ' a = l + i ,

e3= I - i ,

Q, =Q, = O

(2.2.63)

Substituting Eqs. (2.2.63) and (2.2.57) into Eq. (2.2.1 la) results in XI = X , = 1 ,

X, = X 3 =-1

Introducing Eqs. (2.2.57) and (2.2.63) into Eq. (2.2.54) gives

(2.2.64)


f:=O

miseven

65

(2.2.65~)

Inserting Eqs. (2.2.65), (2.2.45), (2.2.57) and f; = 0 into Eq. (2.2.3a) we have

which includes three modes corresponding to three values of 52, shown in Eq. (2.2.59a). The numerical results are given in Table 2.2.2. Table 2.2.2 Natural frequencies noand modes [ x, x, x, x, x , x, ] for the system shown in Fig. 2.2.4

no

0

0.198062264

1

1.55495813

2

3.24697960

mode

Corresponding to Ro=2 (see Eq. (2.2.59b)), the natural mode can be expressed as

66


In view of Eq. (2.2.62), Q4 does not include any complex factor. Substituting Eqs. (2.2.67a) and (2.2.57) into Eq. (2.2.11a) and letting X, = 1 , results in

Substituting Eqs. (2.2.67) and (2.2.57) into Eq. (2.2.54) gives

Finally substituting Eqs. (2.2.69), (2.2.57) and (2.2.45) with Eq. (2.2.3a) results in

q: = 0

into

with Q, = 2 . That leads to

Because the system considered is not subjected to any constraint, there is a mode of rigid body motion which corresponding to zero frequency. Introducing o = 0 and F j ( j= 1,2,...,2N) = 0 into Eq. (2.2.42), the nontrivial solution can be expressed as


67

Substituting Eqs. (2.2.72) and (2.2.57b) into Eq. (2.2.3a) and letting x, = 1 (normalization) yields

fi

corresponding to q,, = . The rigid body mode can also be obtained from Eq. (2.2.70) with a, vanishing. Moreover, let us consider the mode with x, and x, vanishing. Corresponding to the natural frequency no= 1 (see Eq. (2.2.60b)), the nontrivial solution for generalized displacement can be expressed as

In order to make the mode antisymrnrtic for equivalent system, i.e., 1

x N + = x (j= 1 2 , .N) , q ,

must have a complex factor elTv . Letting

.Z

q, = 2e1%nd substituting Eqs. (2.2.74) and (2.2.57b) into Eq. (2.2.3a) we have

The results for all natural frequencies and modes are summarized in Table 2.2.2.

2.2.2b Forced Vibration Example Consider the system having the same parameters shown in Eq. (2.2.57) and subjected to two harmonic forces at the second and fifth mass points as shown in Fig. 2.2.4, where P and w denote the amplitude and frequency of the external excitation. The loading condition can be expressed as

Introducing Eqs. (2.2.76) and (2.2.57) into Eq. (2.2.44b) yields

68


with the other f: vanishing. Substituting Eqs. (2.2.77), (2.2.45~)and (2.2.57b) into Eq. (2.2.49) results in

P = X; =-

(Qo-1) k Qo(3-Qo)

= X;

=x;

= X = ~

--P

(2.2.78a)

1

(2.2.78b)

k Qo(3-Qo)

~ l = XJ ~ - j ~= 172?...,6

(2.2.78~)

That leads to

x,' = x2 l--xr -x: 3 -

p =-

(Qo -1) k Qo(3-Qo)

(2.2.79)

Introducing Eqs. (2.2.79) and (2.2.57) into Eq. (2.2.5 1) results in

P 2(R, -1)

b4 = -

b,

k Qo(3-Qo)

=0

r = 1,2,3

Inserting Eqs. (2.2.80), (2.2.53) and (2.2.57) into Eq. (2.2.52) yields

Q4

p (Qo-1) Q, = o k ~~(2-a,)'

=-

r = 1,2,3

(2.2.81)

Substituting Eqs. (2.2.81) and (2.2.57) into Eq. (2.2.1la), we have

x

"

That leads to

(Qo-1) k 2Q0(2- a o )

=-p

s = 1,2,3,4

(2.2.82a)


69

This function represents the frequency response for displacement of the loaded mass point. Introducing Eqs. (2.2.81) and (2.2.57) into Eq. (2.2.54) gives f,"=-

p (Qo - 1)

J7(2-00)

-i:x

e

Y

p (00-1) f,"=j,", f l02 --- & (2 - 0 0 )

(2.2.83a)

Finally, substituting Eqs. (2.2.83), (2.2.77), (2.2.45) and (2.2.57) into Eq. (2.2.3a) we have

2.2.3

Bi-periodic Mass-Spring Systems with One End Fixed and the Other Free

Consider a bi-periodic mass spring system with fixed and free extreme ends and (2n+l)d+n particles as shown in Fig. 2.2.5(a) where n denotes the number of the particles with mass M +AM and 2d denotes the number of particles with mass M between two adjacent particles having mass M + AM . The symbols Fj and x j denote the amplitudes of the longitudinal load and displacement for the j-th particle and k denotes the stiffness of the coupling spring. In order to apply the U-transformation approach, we have to form a cyclic biperiodic system that is equivalent to the original one. It can be achieved by the following procedures. Firstly, the system is extended at the free end by its symmetrical image and an antisymmetric loading is applied on the corresponding extended part as shown in Fig. 2.2.5(b). Secondly, the new system can be treated as a bi-periodic mass spring system with fixed extreme ends as stated in section 2.2.1. The equivalent system with cyclic bi-periodicity and (4n + 2)p ( p = 2d + 1)

(a)

Original system

syaunetrical

p=2d+l N=(2n+ l )p

(b) Transitional system

M t

M+AM

pj

1

I

Fxtz-j=Fj

symmetrical line 2

XN+Z-j=X j

i

F

I

O - o . . . ~ . - . o ~ . - ~ . O O ~ - ~ . - u ~ O . . . C ) ? r r C ) . . ~ ~ I 21t2d l+p 1t2p j l+np i lyn+l)~N+Z-j 1+2n~ N+j

First particle =(1+2N)-th particle,

x, r x

,+,

(c) Equivalent system with cyclic bi-periodicity Figure 2.2.5 Bi-periodic mass spring system with fixed and free ends

72


degrees of freedom is shown in Fig. 2.2.5(c), where the first and last particles are imaginarily regarded as the same particle. The loads and displacements of the equivalent system must satisfy the following relations:

and

where N = (2n + l ) p , p = 2d + 1 and n, d are the parameters of the original system. The harmonic vibration equation for the equivalent system with 2N particles can be expressed as

where w denotes the vibration frequency and 1 + ( m - l ) p ( m = 1,2,...,4 n + 2) is the ordinal number of the particle having mass M +AM . Applying the U transformation (2.2.3) to Eq. (2.2.87) results in

where

77

+ IY

=-

N

and


73

Substituting Eq. (2.2.85) into Eq. (2.2.89b) yields n -m F

f; =- - 4i i

Jzni

m is odd

j=2

fi = 0

m is even

Inserting Eq. (2.2.86) into Eq. (2.2.3b) gives

qN

= q Z N = 0,

qm= O

m is even

that indicates q m must be an imaginary number, From Eq. (2.2.88), q m can be expressed as

= f;/(2k

- M U *- 2k cos m y/)

Substituting Eqs. (2.2.89), (2.2.92) and j (2.2.3a) we have

=1

+ (s - 1)p

(2.2.92~)

into the U-transformation


74

where

xl.(j = 1,2,...,2 N ) denotes the solution for displacement of the mono-periodic system (i.e., AM = 0 ) under the same loading as that acting on the considered system, namely XI

'

=

-c JWV 1

2N

ei"-"mY 4,'

=,,

Let

where p = ~ / ( 2 n + l ) = ~ yand / i=&. We can now apply the U-transformation (2.2.97) to Eq. (2.2.93). The uncoupling equation can be obtained as

where


75

Because X: ( s = 1,2,...,4n + 2 ) possess the property shown in Eq. (2.2.86), namely

XZ+,-, = Xi,+,+, = -X;,,-,

= -XJ

j = 2,3,..., n + 1

(2.2.100a) (2.2.100b)

X ( = Xi,,, = 0 Eq. (2.2.99b) becomes

b, =- - 4i

2

J4n+2 ,=2

sin(s - 1)rp X:

bzn+,= b4n+2= 0 , b, = 0

r is odd

r is even

That leads to

Q2n+l

= Q4n+2 = 0 ,

Q, = 0

Q4n+2-r = Qr

r is even

Making a comparison between Eqs. (2.2.89a) and (2.2.97b) yields

The final results for x j can be found by introducing Eqs. (2.2.103), (2.2.90) and (2.2.92) into Eq. (2.2.3a). 2.2.3a Natural Vibration Example There are two sets of frequency equations corresponding to X, non-vanishing

76


and vanishing respectively , namely

and

m = 2n + 1,3(2n+ 1),5(2n+ 1), ...(p- 2)(2n + 1)

(2.2.105)

with p = 2d + 1 and y/ = 7r/(2n+ l ) p . When the specific parameters n, d and AM are given, the natural frequency can be found from Eqs. (2.2.104) and (2.2.105). Taking a specific example as shown in Fig. 2.2.6.

Figure 2.2.6 Bi-periodic mass spring system with fixed and free ends d = n = 1

The parameters are given as d=l,

n=l,

AM=M

That leads to

Introducing Eq. (2.2.106)into the frequency equation (2.2.104)yields

(2.2.106a)


77

Applying the relations

to Eq. (2.2.107),the frequency equation becomes - 2Ri

+ 1oR;- 12Q0 + 1 = 0

MU '

with R, = -. The roots for R, of Eq. (2.2.109)are

k

0,= 0.08995531, 1.77031853, 3.13972616

(2.2.110)

Inserting Eq. (2.2.106)into the later frequency equation (2.2.105)gives

That leads to

Consider now the natural modes. Corresponding to the frequency equation

(2.2.109), the natural mode, in terms of the generalized displacement Q, ( F1,2,. ..,6), can be expressed as

78


with the other Q, vanishing. From Eqs. (2.2.101a) and (2.2.102a), Q, must be an imaginary number. Introducing Eqs. (2.2.106) and (2.2.112a) into Eq. (2.2.97a) and letting x, = X 2 = 1 ,yields

Inserting Eqs. (2.2.106) and (2.2.112b) into Eq. (2.2.103) results in

&

f o1 = f o7 = f o1 3 -- - - M &

3

with the other f: vanishing. Substituting Eqs. (2.2.106), (2.2.114) and (2.2.92) with Eq. (2.2.3a), we have

That leads to

(2.2.114a)

qk = 0

into


79

Because the natural frequency R, is a root of Eq. (2.2.109), Eq. (2.2.117a) may be rewritten as

Inserting Eq. (2.2.117b) into Eq. (2.2.116) the natural modes can be expressed as

where Ro may be an arbitrary root of Eq. (2.2.109). Therefore Eq. (2.2.118) represents three natural modes. Their numerical results are given in Table 2.2.3 . Table 2.2.3 Natural frequency R, and mode [ x , x , ... x , ] for the system shown in Fig. 2.2.6 n o

Mode

0.0899553 1

1.77031853

3.13972616

1

80


Corresponding to the frequency equation (2.2.11 la), the mode can be expressed, in terms of q, ,as

In view of Eq. (2.2.91) q, must be an imaginary number. Substituting Eqs. (2.2.106) and (2.2.119) into Eq. (2.2.3a), the natural mode can be obtained as

3fi i and SZ, = 1 . The result is also given in Table 2.2.3. corresponding to q, = - 2 Obviously the particle with mass 2M lies at the nodal point of the mode, i.e., x, = 0.

In conclusion of this chapter we must show clearly that the key to the settlement of the question lies in forming a cyclic bi-periodic system which is equivalent to the considered one. The necessary conditions are as follows: Firstly the system extended by the symmetric image must possess the cyclic biperiodicity and secondly the boundary condition of the original system must be satisfied automatically in its extended system by means of applying the symmetric or antisymmetric loading on the corresponding extended part.

Chapter 3 BI-PERIODIC STRUCTURES 3.1

Continuous Trusses with Equidistant Supports

The plane truss to be considered is the Warren truss. The transverse vibration of the Warren truss with two simply supported ends was investigated by using the Utransformation technique [7], where the truss is regarded as a mono-periodic structure. Recently the static and dynamic analyses of the continuous Warren truss with equidistant roller supports was performed by Cai et a1 [11,12] where the truss is treated as a bi-periodic structure and the U-transformation is also used.

3.1.1 Governing Equation

Figure 3.1.1 Warren truss with equidistant roller supports Consider the continuous truss resting on equidistant roller supports as shown in Fig. 3.1.1. The truss is subjected to transverse loads acting at the nodes. The truss is made up of four sets of bars pin-jointed at the nodes so that only axial forces but no bending moments and shear forces act on the cross-sections of the bars. The bars in the longitudinal direction have modulus of elasticity E, , cross-sectional area A,

82


and length L, . The inclined bars have modulus of elasticity E,, cross-sectional area A, and length L, . In Fig. 3.1.1 N and n denote the total numbers of substructures and spans, respectively and p denotes the number of substructures between two adjacent supports. A typical substructure is shown in Fig. 3.1.2(a). Each substructure consists of four nodes and four bars. In order to avoid repetition, we consider only the nodal loads of two nodes on the left of every substructure. The serial number of both the node and bar is made up of two integer numbers in which the first one is the ordinal number of the node or bar in the substructure and the second one indicates the ordinal number of the substructure. In order to avoid ambiguity the serial numbers of the nodes are given in round brackets.

Bar 4,

(a) Serial numbers of nodes and bars

@)

Displacements and internal forces

Figure 3.1.2 Substructure At the outset, we must create a cyclic bi-periodic system which is equivalent to the original one. The considered truss is extended by its symmetrical image and apply the antisymmetric loading on the corresponding extended part as shown in Fig. 3.1.3 where two bars denoted by dotted lines are additional ones. Such an extended system may be regarded as a cyclic bi-periodic, when two pairs extreme nodes a , a' and b , b' are imaginarily jointed by hinges respectively, i.e., the first substructure is next to the last (2Nth) one. Two additional bars are not subjected to any load for antisymmetric deformation. Consequently the antisymmetrical deformation for the extended truss is not affected by such additional bars which are necessary in order to form the cyclic periodic system. If and only if the displacements

Symmetric line

Figure 3.1.3 Equivalent system with cyclic bi-periodicity subjected to antisymmetric loads

84


of the extended truss possess antisymmetry, the extended truss with cyclic periodicity is equivalent to the original one. When the supports are replaced by the supporting reactions, the equivalent truss may be regarded as a cyclic mono-periodic structure which can be analyzed by using the U-transformation technique [7]. The total potential energy of the equivalent truss considered may be expressed as

where J3 denotes the potential energy of the j-th substructure and 2N denotes the total number of substructure for the equivalent structure. In general, the potential energy may be defined as

,

where [K],,, denotes the stiffness matrix of the substructure; { S ) ,, { F ) denote the displacement and loading vectors for the j-th substructure, respectively and superior bar indicates complex conjugation. It is necessary to have the superior bar in Eq. (3.1.2) at deriving the variational equation with complex variables. For the present case, the displacement vector (61, may be defined as

and

with its components shown in Fig. 3.1.2(b) where u and v denote the longitudinal and transverse displacement components, respectively. By using the conventional

assembly process the stiffness matrix for every substructure can be obtained from the stiffness matrices of four bar elements as

where

I

K I + ~ , c o s 2 a K2sinacosa K2sinacosa K , sin2a [Kill = -K2 c o s 2 a -K2sinacosa -K2sinacosa -K2sin2a

- ~ , c o s ~ a -K2sinacosa - K , sina cosa - K2sin2a K, + 2 K 2 c o s 2 a 0 0 2K2 sin2a

1

in which K, = E,A, /L, , K2 = E2A, /L2 and a denotes the inclination as shown inFig. 3.1.1. The loading vector should include both the external load and supporting reaction. Noting the longitudinal load vanishing, the loading vector takes the form as

86


where p indicates the number of substructures between two adjacent supports; Pk indicates the supporting reaction at k-th support and F(l,j),F(,,,)denote the external loads acting at the nodes (1, j ) and (2, j ) respectively, which must satisfy the antisymmetric conditions, i.e.,

qaj)( j = 1,2; -,N ) are real loads acting on the original truss. The continuity condition between two adjacent substructures may be expressed as

in which F(,,,,( j = 2,3, - ..,N ) and

where (6,),,+, = {S,), due to cyclic periodicity. Eq. (3.1.8) shows the coupling of the energy of substructures. Obviously Eq. (3.1.8) possesses cyclic periodicity. One can now apply the U-transformation to Eqs. (3.1. I), (3.1.2) and (3.1.8). The U- and inverse U-transformations may be expressed as

and

Bi-periodic Structures

. The generalized displacement vector

with y = z / N and i = defined as

87

{q),,, may be

and

By means of the generalized displacement, the potential energy shown in Eqs. (3.1.1) and (3.1.2) can be written as

where

and the continuity condition shown in Eq. (3.1.8) becomes {q ) ~m

=e'"'W

{q~)m,

m = 1,2...,2N

which may be rewritten as

with

in which I is the unit matrix of fourth order. In order to eliminate the non-independent variables {qR},(rn = 1,2-..,2N ) in

Emct Analysis of Bi-periodic Structures

88

Eq. (3.1.1 I), substituting Eq. (3.1.14) into Eq. (3.1.11) yields

where

In Eq. (3.1.16), the real and imaginary parts of {q,), are independent variables. =I 0 , the Substituting Eq. (3.1.16) into the first order of variation equation & equilibrium equation can be obtained as

By inserting Eqs. (3.1.4), (3.1.5) and (3.1.15) into Eq. (3.1.17), results in

- K , c o ~ ~ a ( l + e - " ~ ) -K2sinacosa(l-e-'mv) - K2 sinacosa(1- e-imv) - K2sin2'a(l+ e-"v) 2Kl (1 - cos m y ) + 2K2 cos2a 0 0 2K, sin2a Eq. (3.1.19) is equivalent to the nodal equilibrium equation. Introducing Eq. (3.1.6) into Eq. (3.1.12) yields

in which &,,,,,

&,,,,

1

(3.1.20)

and f&, denote the generalized loads corresponding to the


89

external loading and supporting reaction respectively, namely

Introducing the anti-symmetric condition for nodal loads shown in Eq. (3.1.7) into Eqs. (3.1.21b) and (3.1.21c) results in

Substituting Eqs. (3.1.15) and (3.1.21a) into Eq. (3.1.18) yields

Recalling the definition of {q, J, Eq. (3.1.19) for q (,,,, and q(,,,, gives

(1 - e-imY)[P(l - cosm p) + cos2 alsin a c o s a

-isinm ~ y s i n a c o a s~ 9(1,m)

=

Dm

q(2.m)

shown in Eq. (3.1.lob) and solving

+

Dm

q(4.m)

90


q(3,m)=

(1 - eimv)[P(l- cos my) + cos2a ] sin a cos a Dm

q(2,m)

+ i sin my sin a cos3a q(4,m) Dm

where Dm= (1 - cosm y ) [ ~ / 3 ~ ( 1cosm y ) + 4flcos2 a + cos4a]

(3.1.24~)

Substituting Eqs. (3.1.24a) and (3.1.24b) into the second and fourth component equations in Eq. (3.1.19) results in

where

-

K I Z ,=~K Z I ,=~3.1.2

2Kl sin2a ( l + e-imv)[2cos2a + P(1- cos m ry)] (3.1.26b) 2~2(1-cosm~)+4~cos2a+cos4a

Static Solution [11]

Consider the governing equation (3.1.25). The solution for q(,,,, and q(,,m, of Eq. (3.1.25) may be expressed as

and


91

(3.1.28) Substituting Eq. (3.1.27) into the U-transformation (3.1.9a) yields

where f(;,,,is dependent on the unknown supporting reactions which can be determined by the compatibility condition at supports, i.e.,

Substituting Eqs. (3.1.21d), (3.1.29a) with j = (s - l)p + 1 into the above equation, the restraint condition can be expressed as

where

Here V, denotes the transverse displacement at the s-th supported node caused by the external force for the equivalent system without supports. The compatibility equation (3.1.31) is linear simultaneous equations with unknown Pk ( k = 1,2,...,2n ). The coefficients P,,, ( s, k = 1,2, - ,2n ) of Eq. (3.1.31) possess cyclic periodicity, i.e.,


92

The independent coefficients are Pk,, ( k = 1,2;..,2n ). Eq. (3.1.34) indicates the simultaneous equations have the cyclic periodicity. One can now apply the U-transformation to Eq. (3.1.31). Let

with q

=z/n =p y

. 2n

Premultiplying Eq. (3.1.31) by the operator (I/&)

e-"'l"*

results in

s=l

where

By using the U-transformation twice, the governing equation becomes a set of one degree of freedom equations as shown in Eq. (3.1.36). Obviously the solution for Q, of Eq. (3.1.36) is


93

where

Consider now the denominator on the right side of Eq. (3.1.38). Note that

Substituting Eq. (3.1.39) into Eq. (3.1.38) results in

When the specific structure parameters and external loads are given, the generalized supporting reactions can be calculated from Eq. (3.1.41). Then the supporting reactions and the displacements for all nodes can be found from the related formulas given above. Recalling the definitions of both generalized supporting reactions f&, and

Q, shown in Eqs. (3.1.21d) and (3.1.35b) respectively, there is a simple relation, i.e.,

Consequently, when we are only interested in the nodal displacements, it is not necessary to find the supporting reactions. In order to explain the procedure of the calculation and verify the exactness of the formulas given in the present section, we need to consider a specific truss with loading.

Esnct Analysis of Bi-periodic Srrlrcftrres

94

3.1.2a

Example

Consider a Warren truss having six substructures and four supports subjected to a concentrated load of magnitude F a t the center node as shown in Fig. 3.1.4.

Figure 3.1.4

Plane truss with six substructures and four supports subjected to a concentrated force of magnitude F at the center node

The structural parameters are given as K, = K, = K ,

N = 6 , n=3, p=2,

a =n/3

(3.1.43a)

which lead to w=z/6,

p=lr/3,

p=l

(3.1.43b)

The nodal loads can be expressed as =F

51.4)

62.j)

9

=O *

6l.j)= O J

7

j

= 1,2;..,6

#

4

(3.1.44a) (3.1.44b)

Introducing Eqs. (3.1.43) and (3.1.44) into Eq. (3.1.22), the generalized loads can be obtained as


f(l,rn)

1 - --

6

rnz sin-. F, 2

95

rn = 1,2,...,12

The stiffness coefficients of Eq. (3.1.25) can be found by substituting Eq. (3.1.43) into Eq. (3.1.26) as

3K rnz K12,, = K ~ I , = , -(3 - 2 cos -)(I 40, 6 49 Dm =-16

+ e-imY)

rnz 2 cos 6

Inserting Eqs. (3.1.45) and (3.1.43) into Eq.(3.1.33) gives

where K,,,, and A, can be calculated from Eqs. (3.1.46) and (3.1.28) if m is given. Substituting Eqs. (3.1.46) and (3.1.28) into Eq. (3.1.47) results in

Introducing Eqs. (3.1.48) and (3.1.43) into Eq. (3.1.37) yields

96


Now the generalized supporting reaction can be found by substituting Eqs. (3.1.49), (3.1.46), (3.1.28) and (3.1.43) into Eq. (3.1.4I), as

Inserting Eqs. (3.1SO) and (3.1.43) into the U-transformation (3.1.35a) results in

Since we consider the equivalent truss subjected to the antisymmetric loads instead of the original one, the supports at the symmetric line are not subjected to any loads, i.e., 4 = 0 and P,, = 0 . The real supporting reactions at two extreme ends of the original truss can be found easily by solving the equilibrium equation for 2 the whole truss, i.e., P, = P4 = -F . 35 Introducing Eq. (3.1.50) and p = 2 into Eq. (3.1.42) gives

with the other components vanishing. Substituting Eqs. (3.1.45), (3.1.52), (3.1.43), (3.1.46) and (3.1.28) into Eq. (3.1.29), the transverse displacements for all nodes can be found as


97

The results show that the restraint condition Eq. (3.1.30) is satisfied. The longitudinal displacements for all nodes also can be obtained by inserting Eqs. (3.1.24), (3.1.27),(3.1.28), (3.1.46), (3.1.45), (3.1.52) and (3.1.43) into the first and third component equations in Eq. (3.1.9a), i.e.,

The results are summarized as follows

98


Eqs. (3.1.53b), (3.1.53d), (3.1.55b) and (3.1.55d) indicate that the nodal displacements possess antisymmetry. The axial force on the cross section of every bar can be found by using the Hook's law that may be specifically expressed as

N 3 j = K , ( U ( ~ ,cOSa ~ + ~ -) v ( , , ~ + sina , ) - u ( , , ~cosa ) + v ( ~ sina) ,~) N 4 j = K , ( U ( , , ~C)O

+

S ~v

( ~sina , ~ )- u ( , , ) cosa - v(,,,,sina)

(3.1.56~) (3.1.56d)

,

with j = 1,2, ,2N , where N,, on the left side of Eq. (3.1.56) denotes the axial force and its two subscripts denote the serial numbers of the bar and substructure respectively, as shown in Fig. 3.1.2. Introducing Eqs. (3.1.53), (3.1.55) and (3.1.43a) into Eq. (3.1.56) results in 2J;; N,, =-F, 105

N,, =-- l g A F , 210

N,, =--F,4 4 7 105

9J;; NZ3= -F, 35

447 N,, =-F 105

N3, = --FJ;;

3

, N,, =--F4 J 5

(3.1.57a)

105

J5

, N

= -F 43

3

(3.1.57~)


99

in which N,, = 0 and N,,,, = 0 indicate the axial forces vanishing for two additional bars. It can be verified easily that the equilibrium equation for every node is satisfied and then the solution for displacement and axial force is an exact one for the truss shown in Fig. 3.1.4. 3.1.3 Natural Vibration [12] Consider now the natural vibration of the continuous truss with equidistant supports. The natu~alvibration equation can be obtained easily from the equilibrium equation (3.1.25) by using the inertia force instead of the static loading. The masses of the bars are assumed to be lumped at the nodes. Two lumped masses denoted by M I and M 2 are attached to each of the lower and upper nodes respectively as shown in Fig. 3.1.1. We also assume that the inertia forces in the longitudinal direction may be neglected. By using the w 2 ~ , v ( , , and j ) w~M,v(,,~) instead of F(,,,) and F(,,,) ,namely

and substituting Eq. (3.1.59) into Eq. (3.1.25), the natural vibration equation can be expressed as

100


and f(:,,, have been defined as shown in where K ,,,, , K,,, , K12,,, K,,,, Eqs. (3.1.26) and (3.1.21d) respectively; w denotes the natural frequency; q(,,,, , q(,,,, represent the amplitudes of the transverse generalized displacements. Noting Eq. (3.1.2 1d) and the antisymmetry of P, ,the generalized force has the property

That leads to

The solution for q(,,,, and q(,,,, of Eq. (3.1.60) can be expressed as

rn = 1,2,...,2N and rn

#

N,2N

(3.1.63)

where Am(@)= (K11.m-w2M1)(K22,m - w ~ M ~ ) - K I ~ , ~ K ~ (3.1.64) I,~ For the original truss with simply supported ends and without the internal supports (as a result f(:,,, = 0 ), the frequency equation can be expressed as

If A, ( 0 ) z 0 , substituting Eq. (3.1.63) into the U-transformation (3.1.9a), i.e.,


101

we have

where A;,,, is a function of P, which can be determined by the constraint condition at supports shown in Eq. (3.1.30). Substituting Eqs. (3.1.67a) and (3.1.21d) into Eq. (3.1.30) ,we have

where

Eq. (3.1.68) is the linear simultaneous equations with unknown Pk (k = 1,2,...,2n) . The coefficients pSxkpossesses cyclic periodicity shown in Eq. (3.1.34). As a result, Eq. (3.1.68) can be uncoupled by means of the Utransformation. One can now apply the U-transformation (3.1.35) to Eq. (3.1.68). The uncoupling equation can be expressed as

where y, = ls/n and

102


In view of Eq. (3.1.35b) and antisymmetry of Pk,it can be shown that

Recalling Eq. (3.1.40) and introducing Eq. (3.1.70b) into Eq. (3.1.70a) we have

and

The frequency equation can be expressed as

corresponding to P, non-vanishing. When w is a root of Eq. (3.1.73) the nontrivial solution for Qr of Eq. (3.1.72a) exists. If all supported nodes are located at the nodal points of the natural mode, the supporting reactions are identically equal to zero. The corresponding frequency equation can be obtained from Eq. (3.1.65) with rn = n,2n, ...,pn ,namely

where rn is in agreement with the half wave number of the natural mode for the original system. 3.1.3a Example

Consider a continuos Warren truss having 2n substructures and n spans. The


103

structural parameters are

and

where n is an arbitrary positive integer. Introducing Eqs. (3.1.79, (3.1.26) and (3.1.64) into the frequency equation (3.1.73), it yields

where

K2 ( m ) =

288(1+ cos m y ) ( 3 - 2 cos m y ) (49-32c0smy)~

and Q denotes the non-dimensional frequency parameter. Eq. (3.1.76) depends on the positive integer n, i.e., the total number of spans of the continuous truss. For several values of n, the natural frequencies are calculated from Eq. (3.1.76)by using numerical method. The results are shown in Table 3.1.1. There are 3(n-1) natural frequencies altogether. One can now consider the other frequency equation corresponding to the case of Q, ( r = 42;-.,2n ) vanishing. Substituting Eqs. (3.1.75)and (3.1.26)into Eq. (3.1.74),one can show that

104


Table 3.1.1 Natural frequency w of the continuous truss with p = 2 and n spans

JG

Multiplier: a k denotes the ordinal number of the pass bands [ W L ,WU )represents W L < W < WU .


105

The analytical solution for Q in Eq. (3.1.78) is

Introducing Eq. (3.1.77) into Eq. (3.1.79a), it yields

The corresponding frequencies are

These natural frequencies are independent of n and in agreement with those for the sectional truss between two adjacent supports, i.e., the case n = 1. The numerical results are also shown in Table 3.1.1. They represent the lower limits of the three pass bands. The total number of the natural frequencies is equal to 3n, which is in agreement with the number of the degrees of freedom for the truss considered. The results in Table 3.1.1 show that all natural frequencies lie in the three pass bands. If n approaches inflnity, each pass band is full of natural frequencies. The upper and lower limits of the pass bands can be obtained by solving Eq. (3.1.76) with r = 0, n . The explicit solution for Q can be expressed as Q = L ( 7 1 - a ) , 51 and

It can be shown readily that

4

3

1 -(71+m) 51

for r = O

(3.1.80a)

106


where w, and w,, ( k = 1,2,3) denote the lower and upper limits of the k-th pass band, respectively. In Table 3.1.1, the numerical results show that the natural frequencies of the continuous truss with arbitrary number of spans do not include any upper limit of the three pass bands, but they can approach every upper bound of the pass bands as a limit when n approaches infinity. When n is a large number, the natural frequencies are densely distributed in each pass band. It is not easy to find the dense natural frequencies by numerical methods. In the present approach, the frequency equation of the continuous truss with n spans is uncoupled to form n frequency equations. The natural frequencies obtained from each one are dispersed, namely, they lie in different pass bands. Consequently, accurate natural frequencies can be easily found no matter how large n is. Consider now the natural modes. If R, denotes a root for R of the r-th equation in Eq. (3.1.76) for a given r, the non-trivial solution for R m ( m = 1,2,. . ,2n ) can be expressed as

with the other Qm vanishing where A indicates an arbitrary real constant with the unit of force. Because the supporting reactions have antisymmetry for the extended truss, i.e., P2n+2-r =-Ps ( ~ = 2 , 3 , . . . , n )and P, = P,,, = 0 , according to Eq. (3.1.35b), Q, must be an imaginary number similar to in Eq. (3.1.22a). Substituting Eq. (3.1.8 1) into Eq. (3.1.42) with p = 2 , one can show that

A,,,)


with the other

f(:,,,

107

vanishing.

Substituting Eqs. (3.1.82), (3.1.75), (3.1.26) and (3.1.64) into Eq. (3.1.67) gives

(3.1.83b) where K, (m) and K , (m) have been defined in Eqs. (3.1.77a) and (3.1.77b) and 24(3 - 2 cos m y ) K3(m)= 49-32cosmy with y/ = 7r/2n. It can be proved that when r is odd, the mode shown in Eq. (3.1.83) is symmetric, i.e., v ( , , , ~ + , - j ) - v(,,,) and v ( ~ , ~ , + , - - v ( , , ) ; when r is even, the mode is

,,

-

antisymmetric, i.e., v ( , , ~ , , + ~ - j ) satisfies the constraint condition

and

'(2,2n+l-j)

= -'(2,j)

. Also, the mode

3.1.4 Forced Vibration [l2]

The continuous truss subjected to transverse harmonic loads acting at the nodes is considered. The forced vibration equation can be expressed as

108


amplitudes; w denotes the excitating frequency and K,,,, , K 2 , K 2 , K22,m are the same as those shown in Eq. (3.1.26). Recalling Eq. (3.1.22), the generalized force has the property

In view of Eqs. (3.1.87a) and (3.1.61), the solution for q(,,,,, and q(,,,,, in Eq. (3.1.86) with m = 2N will vanish; namely

For the case of m = 2N (my/ = 2 n ) , all of the substructures having the same amplitude vector and phase, there is no antisymmetric mode for the equivalent system. Introducing Eqs. (3.1.87b), (3.1.61) and K,,,, = K,,,, = 0 into Eq. (3.1.86) with m = N ,we have

For m = N ( my/ = n ), two adjacent substructures have the same amplitude vector and opposite phase, i.e., the nodes (1, m) (m = 1,2-..,2N) are located at nodal points of the corresponding mode. Consequently, the generalized displacement q(,,,, corresponding to v(,,])( j = 1,2- - - ,2N ) is equal to zero. In general, the solution for q(,,,, and q(,,,, in Eq. (3.1.86) can be expressed as

( i f A,(@) where

#

0),

m = 1,2..-,2N


109

One can show that

Substituting Eq. (3.1.89) into the U-transformation (3.1.66) results in

where v ( , , ~and ) v ( , , ~denote ) the amplitudes of the transverse displacements for

; is the function of P, which can be nodes (1, j) and (2, j), respectively, and )4 determined by the constraint condition at supports, i.e., Eq. (3.1.30). Substituting Eqs. (3.1.21d) and (3.1.92a) into Eq. (3.1.30), we have

in which

Here Vs represents the transverse amplitude of the s-th supported node caused by the external harmonic excitation for the equivalent system without supports. It can


110

be verified that Vs ( s = 1,2,...,2n ) possess antisymrnetry, i.e.,

One can now apply the U-transformation (3.1.35) to Eq. (3.1.93). To this end, 2n

premultiplying Eq. (3.1.93) by the operator (I/&)

e-i(s-')r'

,we have

s=l

where

Therefore, Eq. (3.1.93) is uncoupled into a set of single degree of freedom equations as shown in Eq. (3.1.96) by using the U-transformation. Introducing Eq. (3.1.95) into Eq. (3.1.97b) yields

and

b2,-,

= b,

,

b, = b2, = 0

Considering Eq. (3.1.98b), Eq. (3.1.96) can be rewritten as ar(w)Qr +br = 0 ,

r =1,2;-.,n-l


111

where

Recalling Eq. (3.1.40) and substituting Eq. (3.1.97a) into Eq. (3.1.1OO), one can show readily

If a r ( o ) # O (r=1,2;-.,n-I), expressed as

the solution for Qr in Eq. (3.1.99a) can be

When the specific structure and loading parameters are given, Qr can be calculated from Eqs. (3.1.101), (3.1.102), (3.1.94b) and (3.1.98). Then the supporting reactions and the nodal displacements can be found from the related formulas given above. Recalling the definitions of both generalized supporting reactions h;,,, and Qr shown in Eqs. (3.1.21d) and (3.1.35b), respectively, a simple relation between

A:,,,

and Q, can be established

Consequently, when we are only interested in the nodal displacements, it is not necessary to find the supporting reactions Pk. It will save considerable computing effort.

3.1.4a

Example

Consider now a Warren truss having six substructures and four roller supports subjected to a harmonic loading Fei" acting at the center node as shown in Fig. 3.1.5.

Figure 3.1.5

Plane truss with six substructures and four supports subjected to a harmonic force at the center node

The structural parameters are given as

and

The amplitudes of the nodal loads can be expressed as

51,.4) =F 0 (hardening

K K 4% and y < 0 (softening spring) w < spring) w 2 > -+. Consequently M M M we can come to the conclusion that if y > 0 there is one localized mode with w greater than w, and if y < 0 there is one localized mode with w less than w, . For the localized mode the amplitudes of all subsystems may be obtained from Eqs. (5.1.21) and (5.1.30) as

Nearly Periodic Systems with Nonlinear Disorders

21 1

It indicates that the amplitudes decay exponentially on either side of the nonlinear disorder. The attenuation rate 6 of localized mode may be found by substituting Eqs. (5.1.27) and (5.1.32) into Eq. (5.1.29) as

5 5

is the odd hnction of nondimensional parameter 77. When 17 approaches zero, also approaches zero. It indicates that if kc decreases or y increases or increases, the mode will approach strongly localized state. The localized level of the modes is dependent on not only the structural parameters kc and y but also the amplitude Ajl, related to the initial condition or the total energy of system. Ail ,o can be determined by applying the energy conservation to the considered system. Ignoring the O(E) term it can be proved that the following equation

is a necessary condition for the energy conservation. p2 is an energy-like quantity 1 which is related to the total energy E of the system by -~ p +' O(E) = E . 2 Inserting Eqs. (5.1.34), (5.1.35) and (5.1.33) into Eq. (5.1.36) results in

21 2


which indicates that when kc/y is given, in order that a localized mode would occur, however small the amplitude, the energy constant p2 must be greater than 18kC/3yl.If the system energy is below the required level it is impossible to realize any localized mode. This property is different from that for the linear periodic system. 5.1.3 Localized Modes in the System with Two Nonlinear Disorders

For the system under consideration,the governing equation (5.1.22) becomes

1

1

in which n = j, - j, and

6 has been defined by Eqs. (5.1.29) and (5.1.27).

Eq. (5.1.38a) minus and plus Eq. (5.1.38b) yield

The above equations are equivalent to


21 3

respectively. The simultaneous equations (5.1.40) may be divided into three sets of equations, i.e,

(ii)

(iii)

'jlo

-fl,(l- 2 f i , the

nonsymmetric localized mode does not exist. 5.3

Damped Periodic Systems with One Nonlinear Disorder [23]

Consider the system shown in Fig. 5.3.l(a) which consists of n number of subsystems connected to each other by means of a linear spring having stiffness &kc. Each subsystem is made up of a mass M connected to both a dashpot with a nondimensional damping coefficient €6, and a spring with linear stiffness K (for x ~ one), where ordered subsystems) or nonlinear stifmess K + ~ ~ (for, disordered E is a positive small parameter. In Fig. 5.3.l(a), s denotes the ordinal number of the disordered subsystem and x j denotes the longitudinal displacement of the j-th mass. In order to apply the U-transformation to uncouple the linear terms of the governing equation, an equivalent system with cyclic periodicity must be created. It is necessary to extend the original system by its symmetrical image and apply the antisymmetric loading on the corresponding extended part as shown in Fig. 5.3.l(b) in which the first and last (2n-th) masses are imaginarily jointed by a spring with stiffness &kc. This imaginary spring is not subjected to any load for antisymmetric vibration. If and only if the dynamic response of the extended system is antisymmetric, two extreme end conditions of the original system are satisfied in the extended one, i.e., the extended system is equivalent to the original one. The response of the first half (i.e., substructures 1 n ) of the equivalent system is the same as that of the original system.

-

5.3.1

Forced Vibration Equation

Applying Newton's second law to every mass in the equivalent system, one can write the differential equations of motion as follows

(a) Original system

Centre line

(b) Equivalent system

Figure 5.3.1 Damped periodic system with a nonlinear disorder


241

and

where the superior dot denotes the derivative with respect to the time variable t, w, denotes the natural frequency for the single ordered subsystem and x ~ , , +=~xI,xO= xZn due to cyclic periodicity, EFjo denotes the amplitude of the harmonic force acting on the j-th mass and R denotes the driving frequency, &yo is the coefficient of the cubic term of the nonlinear stiffness in the disordered subsystem. The external excitation for the equivalent system must satisfy the antisymmetry condition, i.e.,

-

where F,,, F,,, indicate the real excitation acting on the original system. If the initial conditions are antisymmetric, then the dynamic displacements are also antisymmetric, i.e.,

One can now apply the U-transformation to the governing equation (5.3.1). The U- and inverse U- transformations may be expressed as

and

242


with y = - and i = f i , where 2n denotes the total number of subsystems for n the equivalent system. Noting that the displacements are always real variables, it can be proved that the generalized displacements q, (m = 1,2;-.,2n) have the following property iT

and qn , qzn must be real variables, in which the superior bar denotes complex conjugation. By using the U-transformation, i.e., premultiplying both sides of Eq. (5.3.1) by 1 2n the operator - e'"-""* , Eq. (5.3.1) becomes

4%

j=I

where

w m ( m = 0,1,2,. n - 1) is the (m+l)-th natural frequency for the undamped periodic system without any disorder. The lower and upper bounds of the pass band can be expressed respectively as a * ,

Nearly Periodic @stems with Nonlinear Disorders

243

Introducing the time substitution

Rt=r+v into Eq. (5.3.7) results in

and V, =

2 wm -

+

n2 a2

E

2k,(1- cosm y) M R ~

where the prime symbol designates differentiation with respect to the new time variable z and y, is an unhown phase angle. Consider now the case of primary resonance, i.e., R = wo . By letting

Eq. (5.3.13) can be written as

where

77,=770+

2k, (1 - cos m v / ) MR2

Inserting Eq. (5.3.15) in Eq. (5.3.12), gives

4: + q m= G m

m = 1,2,...,2n

244


in which 1

I

cos(r + a,) -cos(r --)myyox:(q1 ,q2,...,q2n) 2

Gm=

According to the perturbation method, we seek a solution of Eq. (5.3.17) in the form of power series in & not only for q m ( r ),but also for a,. Hence, let q m ( r )= q m 0 ( r + ) E q m l ( ~+)& 7 , z ( r ) +...

(5.3.19)

p = p o + & P I+ E2 p2 +.-.

(5.3.20)

Eq. (5.3.19) is equivalent to x j ( r )=

x ~ ~ ( ~ ) + E x ~ ~ ( ~ ) + E ~ x ~ ~ ( ~ ) + (5.3.21) . . .

with

Cei(i-~mrqmp

1 x ( r )= -

J2n

2n

r

= 0,1,2,--.

(5.3.22)

,=I

Substituting Eqs. (5.3.19) and (5.3.20) into Eqs. (5.3.17) and (5.3.18), the coefficients of equal powers of & on both sides of Eq. (5.3.17) must be equal, i.e.,


245

5.3.2 Perturbation Solution

The solution of Eq. (5.3.23a) may be expressed as qmo= a,, cos 7 + b,, sin 7

m

-

-

= 1,2;..,2n

(5.3.24)

-

with a2,-,,, = urn,,and b2,-,,, =b,,, due to q2,-,,,,o= q,,,, where a,, and b,, ( m = 1,2,. ..,2n ) are complex constants to be determined. The physical displacements corresponding to q,, shown in Eq. (5.3.24) can be obtained from Eq. (5.3.22) with ~0 as

xj,

= Ajo cos r

+ Bjosin 7

j

= 1,2,-..,2n

(5.3.25a)

where

Ajo and Bjo are real numbers and A,,-,,, = A,, , B,,-,,, = B j o , which lead to -

.

X 2 n - j , ~- n j ~

Without loss of generality, we can assume that the initial velocity for the

246


disordered subsystem is zero besides the antisymmetry for both initial displacement and velocity, which leads to

Bso = 0

(5.3.26)

xS0 = Aso cos r

(5.3.27)

and

In order to prevent secular terms, the coefficients of cos T and sin T on the right side of Eq. (5.3.23b) must be zero. Introducing Eqs. (5.3.24) and (5.3.27) into Eq. (5.3.23b), letting the coefficients of cos r and sin r be equal to zero, give 2eiimv

I

[(g

MD2G

1 Fjo cos(j - -)m y cos 9, - cos(s - -)m y 2A; 2 " 4 2

I

Consider now a specific loading condition as that there is no excitation acting on each subsystem except the disordered one, i.e.,

Fjo = 0

j +s

and

j=1,2;-.,n

(5.3.29a)

Inserting Eq. (5.3.29) into Eq. (5.3.28), the solution for am, and bmo of simultaneous equations (5.3.28a7b)can be expressed as


1 2 a,,,, = --e 2kc

I

i-mv

fi

b m"

- 1 2kc

2 -e2

J2n

i'mv

1 cos(s - $rn

( D + l - c ~ s m y ) I ,+CI,

247

(5.3.30a)

( ~ + 1 - c o s m y )+~c 2

1 CI, -(D+l-cosmy)I, COS(S---)my (5.3.30b) 2 (D+I-cos~~)~+c~

in which

C and D are two nondimensional parameters. They are dependent on the nondirnensional frequency, stifiess and damping constant, i.e.,

fi 3 and w, ' K

EC, . In Eqs. (5.3.31a) and (5.3.31b), A,, and p, are unknown variables. Substituting Eq. (5.3.30) into Eqs. (5.3.25b) and (5.3.25~)results in

where

248


Consider now the s-th set of simultaneous equations in Eq. (5.3.32). Inserting j=s and Eq. (5.3.26) in Eq. (5.3.32), yields

Noting the definitions of I, and I , shown in Eqs. (5.3.31a,b), Eqs. (5.3.34a,b) may be rewritten as

F,, sin po = 2kcPsAs, a,'+ P,' From the above equation, we can find the phase angle with zero-order approximation as 2kc Ps

po = tan-'

3

%as + q ~ o ~ : o(a: +

a:

and the frequency response curve as

in which a, and

P,

are dependent on 0 . They can be expressed as


249

where C and D are dependent on R l w o besides the structural parameters as shown in Eqs. (5.3.31~)and (5.3.31d). If the parameters of the system and loading are given, the response As, for the loaded subsystem can be calculated from Eq. (5.3.37), and the other Ajo and Bjo can be obtained by substituting Eq. (5.3.34) into Eq. (5.3.32) as

R The characteristic of the frequency response ( (Aso--) curve is similar to

I

W0

that for the single nonlinear subsystem, i.e., the jump phenomenon may occur. For = 30 the specific case of &yo = 0.2 , E

Exact Analysis of Biperiodic Structures

Exact Analysis of Biperiodic Structures

Exact Analysis of Discrete Data

Exact Analysis of Discrete Data

Exact Analysis of Discrete Data

Exact Analysis of Discrete Data

Matrix Analysis of Structures

Seismic Analysis of Structures

Inelastic Analysis of Structures

Inelastic analysis of structures

Analysis of Composite Structures

Nonlinear Analysis of Structures

Asymptotic analysis of periodic structures

Exact Revenge

Exact Revenge

Exact Revenge

Exact Revenge

Exact Revenge

Exact Revenge

Exact Revenge

Plastic Analysis and Design of Steel Structures

Programming the Dynamic Analysis of Structures

Analysis and design of plated structures

Modelling of Steel Structures for Computer Analysis

Exact algorithms for exact satisfiability problems

Inelastic Analysis of Solids and Structures

Analysis and design of flight vehicle structures

Matrix Analysis of Structures , Second Edition

Asymptotic Analysis for Periodic Structures

Asymptotic Analysis for Periodic Structures

Asymptotic Analysis for Periodic Structures

Exact Analysis of Biperiodic Structures