Dynamic Analysis of Structures

Dynamic Analysis of
Structures

John T. Katsikadelis
Professor Emeritus of Structural Analysis
School of Civil Engineering
National Technical University of Athens
Academic Press is an imprint of Elsevier
125 London Wall, London EC2Y 5AS, United Kingdom
525 B Street, Suite 1650, San Diego, CA 92101, United States
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
© 2020 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, recording, or any information storage and
retrieval system, without permission in writing from the publisher. Details on how to seek
permission, further information about the Publisher’s permissions policies and our
arrangements with organizations such as the Copyright Clearance Center and the Copyright
Licensing Agency, can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the
Publisher (other than as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience
broaden our understanding, changes in research methods, professional practices, or medical
treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating
and using any information, methods, compounds, or experiments described herein. In using
such information or methods they should be mindful of their own safety and the safety of others,
including parties for whom they have a professional responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors,
assume any liability for any injury and/or damage to persons or property as a matter of products
liability, negligence or otherwise, or from any use or operation of any methods, products,
instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library

ISBN 978-0-12-818643-5

For information on all Academic Press publications

visit our website at https://www.elsevier.com/books-and-journals

Publisher: Matthew Deans

Editorial Project Manager: Joshua Mearns
Production Project Manager: Nirmala Arumugam
Cover Designer: Greg Harris
Typeset by SPi Global, India
 Dedication

To my wife Efi
for her loving patience and support,
to my children Stefan and Christina,
and to
my granddaughter Katharina
for patiently enduring and sharing the years of
preparation of the book with me.

This work is also dedicated to my numerous students

whose attendance of the Structural Dynamics
course over the past 40 years encouraged an
international publication of the book.
Preface

The statement of the laws of motion by Newton 334 years ago (1686)a was a
milestone in the evolution of mechanics and modern engineering. The relation
between force (cause) and motion (effect) was quantified as a relation between
the linear momentum of the body and the force exerted on it. Thus, this relation
from a subject of philosophy up to that time turned out to be a valuable tool of
science for the study of the natural world. The subsequent developments in the
sciences were rapid. Astronomy, mathematics, mechanics of fluid and deform-
able bodies, and in general, mechanics of continuous media reached their peaks
in the centuries that followed, with immense applications to all engineering dis-
ciplines. Nevertheless, the laws of motion, which were stated as an axiom
(Axiomata sive Leges Motus) by Newton because, apparently, he could not jus-
tify their derivation, was a consequence of the discoveries of great scientists
who preceded him such as Galileo,b Kepler,c Hook, etc.
The implementation of the laws of motion leads to mathematical models
described by differential equations, ordinary or partial, whose solution effort
has given a great impetus to the development of mathematics. Unfortunately,
analytical solutions are limited to simple problems such as vibrations of discrete
systems with a few degrees of freedom; linear vibrations of beams, membranes,
plates, and shells with simple geometry; and simple support conditions made
from materials, mostly with a linear behavior. These solutions, while useful
for extracting qualitative conclusions about the dynamic response of structures,
are not capable of solving realistic problems in engineering, where the geometry
and loads are complicated while the response is generally nonlinear. Although it
has been shown that Newton’s law of motion is of an integer-order derivative,d
in recent years, the fractional derivatives have been proven more suitable for
modeling the actual structures. However, the use of fractional calculus has
not been employed in mathematical physics for three centuries because the

a. I. Newton, Philosophiae Naturalis Principia Mathematica, Royal Society Press, London, 1686.
b. J.T. Katsikadelis, Derivation of Newton’s law of motion using Galileo’s experimental data, Acta
Mech. 226 (2015) 3195–3204, doi:10.1007/s00707-015-1354.
c. J.T. Katsikadelis, Derivation of Newton’s law of motion from Kepler’s laws of planetary motion,
Arch. Appl. Mech. 88 (2018) (2017) 27–38, doi:10.1007/s00419-017-1245-x.
d. J.T. Katsikadelis, Is Newton’s law of motion really of integer differential form? Arch. Appl.
Mech. 89 (2019) 639–647, doi:10.1007/s00419-018-1486-3.

xv
xvi Preface

fractional derivative, contrary to the integer-order derivative, has no direct

physical or geometrical meaning. Thus, it was not adopted by the famous sci-
entists of earlier centuries who developed modern science and engineering, as
they could not visualize the significance of the fractional derivative. Thus, we
dare say that in a certain sense, Newton’s Principia has delayed the development
of mechanics and engineering.
Structural dynamics became a tool for the dynamic analysis of structures
only with the advent of computers in the middle of the last century. Nowadays,
the cheap, fast, and large-capacity computers as well as the development of effi-
cient numerical methods to solve the equations of motion allow the safe study of
complicated structures under dynamic loads. Aircrafts, large bridges, ships,
high-rise buildings, dams, fast trains, etc., are now being safely analyzed and
designed. The finite element method (FEM) permits the dynamic analysis of
continuous systems by modeling them with equivalent discrete systems. The
boundary element method (BEM) analyzes with great accuracy the dynamic
response of systems described by ordinary or partial differential equations.
Besides, the coupling of the two methods paves the way to solving very difficult
dynamic problems such as the interaction of the structure with soil or fluids as
well as the large deformation analysis of systems with nonlinear constitutive
equations. Recently, meshless methods, a numerical tool beyond FEM, have
been able to solve efficiently the equations describing the dynamic response
of continuous systems.
The dynamic analysis of structures today is mainly done by using profes-
sional computer codes such as NASTRAN, SAP2000, ETABS, ADINA, etc.,
which have relatively unlimited capabilities. While these programs are
undoubtedly a useful and indispensable tool in the hands of the engineer, they
do not allow for an understanding of the dynamic behavior of the structures or
the limits of the validity of the results. This can be achieved only with a deep
theoretical background, which is constantly expanding with the development of
the computational methods in mechanics. The engineer without theoretical
knowledge is unable to formulate the structural model. The modeling of struc-
tures cannot be done by the computer but only by the engineer, who is the only
one responsible for performing proper dynamic analysis. The belief that theo-
retical knowledge is not necessary today because there are ready-to-use com-
puter codes is, at the very least, dangerous. The engineer should never
accept the results of a computer program unless he/she can check their validity
and accuracy. Apparently, this is possible only if the engineer has a deep the-
oretical knowledge of the dynamics of structures. We emphasize that theoretical
knowledge today is more necessary than ever. Areas that a few decades ago
were covered by applied mathematics and theoretical or applied mechanics
have become the subject of the modern engineering praxis.
Dynamics of Structures for engineers, particularly for civil engineers, was
included as a regular course in the curricula of many universities long ago,
 Preface xvii

particularly after the advent of computers. Therefore, many books have appeared
on this subject. However, these books provide a means only to understand the
response of simple and mostly unrealistic structures when subjected to dynamic
loads, especially to ground motion. The principles of dynamics are illustrated by
applying them to very simple models, which cannot describe actual structures
and therefore cannot be employed for dynamic analysis and design. We should
have in mind that the dynamic analysis of actual structures requires their model-
ing, the formulation of the governing equations of motion, their solution under
any dynamic load, and the physical interpretation of the results.
In the last 30 years, almost all seismic codes have encountered earthquake
ground motion as an effective dynamic load. The advent of computers in the
early 1960s encouraged engineers to develop methods of dynamic analysis
of structures, modeled first by the FEM and later by other advanced numerical
methods. Today, these methods constitute a powerful tool for dynamic engi-
neering analysis. Thanks to the availability of cheap computer power, every
engineer can use them. The essential ingredients of a book on Dynamics of
Structures for Civil Engineers should be:
(a) The basic concepts and principles of structural dynamics as they are
applied to particles as well as rigid and deformable bodies, enabling the
student or the engineer to formulate the equations of motion of any struc-
ture, no matter how complex, once the dynamic model has been adopted.
(b) Realistic modeling of actual structures under dynamic loads.
(c) Analysis of the dynamic response of the structure represented by its
model under any specified load. The analysis should include single- and
multiple-degree-of-freedom systems for linear and nonlinear response
under any dynamic excitation.
(d) Approximation of real structures using computational methods such as the
FEM, which replaces the actual structure (distributed parameter system)
with an approximate discrete system for which analysis methods can be
applied.
(e) Effective present-day numerical methods for dynamic analysis, including the
numerical solution of eigenvalue problems and the direct solution of the equa-
tions of motion of large systems, namely, systems with a large number of
degrees of freedom such as those resulting from the employed discretization.
Students attending a course on Dynamics of Structures should be exposed at
least to the above subjects. However, not all of them can be found in a single
book. Therefore, people interested in structural dynamics should refer to more
than one book in order to retrieve the required knowledge. Apparently, these
books cannot be used as integrated textbooks in the sense described above.
The student should be acquainted with different symbols and approaches, which
complicate the acquisition of knowledge, an approach that is, at least, educa-
tionally inappropriate.
xviii Preface

Knowledge of the dynamics of structures is particularly necessary for engi-

neers, who are studying the response of their structures subjected to seismic
ground motion. Modern earthquake codes, which as mentioned treat the ground
motion as a dynamic load, require a deep theoretical understanding of the
dynamics of structures. The aim of this book is to give the student as well as
the professional engineer the minimum knowledge necessary to understand
the dynamic response of the structures. The author has taught structural dynam-
ics, both as an undergraduate and a graduate course, at the School of Civil Engi-
neering of the National Technical University of Athens (NTUA) continuously
since 1981, when it was first introduced as a formal course by the late Prof. A.E.
Armenákas. The author’s teaching experience together with his long research in
this area have contributed to the presentation of material in a simple and com-
prehensible way. The book contains numerous illustrative examples that help to
understand the theory. Particular emphasis is given to numerical methods,
which are presented methodically and given in the form of pseudo-codes so that
interested readers can write their own computer codes in the language of their
choice. The list of the programs in FORTRAN and MATLAB are given in elec-
tronic form on this book’s companion website. Nevertheless, the numerical
results in the Examples have been obtained using the MATLAB programs
because it is easy for the students to master this language. The programs can
be used to solve many problems of engineering praxis.
The book is divided into two parts.
The first part deals with the single-degree-of-freedom (SDOF) systems
and contains nine chapters. In particular, Chapter 1 presents the general
concepts and principles of dynamics as they apply to structural dynamics.
Chapter 2 deals with the free vibrations of the SDOF systems while
Chapter 3 deals with their forced vibrations. Particular emphasis is given
to the study of the resonance phenomenon. Chapter 4 presents the most effi-
cient numerical methods for solving the equation of motion. Chapter 5 deals
with the nonlinear response of the SDOF systems. Chapter 6 presents the
response of structures due to ground motion. Chapter 7 deals with the damp-
ing of structures while Chapter 8 approximates the continuous systems by
generalized SDOF systems and studies the dynamic response of beams trea-
ted as continuous systems. Finally, Chapter 9 describes the analysis of SDOF
systems in the frequency domain.
The second part deals with the multi-degree-of-freedom (MDOF) systems
together with the well-known computational methods for their analysis. It con-
tains seven chapters. In particular, Chapter 10 describes the various methods of
modeling of the MDOF systems, which are classified into categories that facil-
itate the selection of the appropriate method for formulating the equations of
motion. Chapter 11 presents in detail the FEM for the dynamic analysis of skel-
etal structures (trusses, frames, grids) as an extension of the static matrix struc-
tural analysis to dynamics. The reduction of degrees of freedom due to the
omission of axial deformations in specific structural members as well as the
dynamic response of flexible structures containing rigid bodies are also
 Preface xix

discussed. The derivation of the equivalent element nodal quantities, that is,
mass and stiffness matrices and forces, are derived by using the Lagrange equa-
tions instead of the principle of virtual works. Although the principle of vir-
tual works offers a handy tool for the derivation of these quantities, the use of
the Lagrange equations was preferred here. The reason is that the Lagrange
equations not only offer a straightforward method for the derivation of the
equivalent nodal quantities for all types of elements, especially in complex
systems with a nonlinear response, but they also allow understanding of their
physical significance. Chapter 12 studies the free vibrations of MDOF sys-
tems without and with damping. The linear eigenvalue problem is presented
from the mathematical point of view, aiming at drawing useful conclusions
about the eigenfrequencies and the mode shapes of the physical systems.
Chapter 13 presents the numerical methods for the computation of the eigen-
frequencies and mode shapes, especially for systems with a large number of
degrees of freedom such as those derived from the application of finite ele-
ments. Chapter 14 studies the forced vibrations of MDOF systems. A large
part of this chapter is devoted to the mode superposition method. It also dis-
cusses the use of Ritz vectors to reduce the degrees of freedom. Particular
emphasis is given to the response spectrum method. The response of linear
systems when they are subjected to a synchronous and an asynchronous
motion of the supports are also discussed. This chapter ends with the presen-
tation of the numerical methods, giving the respective pseudocodes for the
time integration of linear and nonlinear systems of equations of motion.
Chapter 15 discusses the approximation of multistory buildings by skeletal
structures and presents methods of formulating their equation of motion.
Finally, Chapter 16 discusses the response of seismically isolated buildings.
This chapter is introductory to the subject and aims primarily at understand-
ing the impact of base isolation on structures.
The book is supplemented by an appendix. Therein, the basic theory of rigid
body dynamics is presented for large and small displacements and the relevant
equations are derived, which are employed in the development of the material of
the book.
In closing, the author wishes to express his sincere thanks to his former stu-
dent and coworker Dr. A. J. Yiotis for carefully reading the manuscript as well
as for his suggestions, constructive recommendations and his overall contribu-
tion to minimizing the oversights in the text. Warm thanks also belong to Dr.
Nikos G. Babouskos, also a former student and coworker of the author, not only
for his careful reading of the manuscript and his apposite suggestions for
improvement of the book but also for his assistance in checking the computer
programs and in producing the numerical results of examples therein. Finally,
thanks belong to Dr. G. Dasios, professor of mathematics at the University of
Patras as well as to his former students, Dr. G. Tsiatas, associate professor of
mathematics at the University of Patras, and Dr. P. Tsopelas, associate professor
of mechanics at NTUA, for reading certain sections of the book and making
constructive suggestions.
xx Preface

It is a pleasure to make grateful acknowledgment of many helpful sugges-

tions contributed by my students who have attended the course over the past
39 years.

J.T. Katsikadelis
Athens
April 2020
Chapter 1

General concepts and principles

of structural dynamics
Chapter outline
1.1 Introduction 3 1.7 Hamilton’s principle 42
1.2 Types of dynamic loads 6 1.8 Lagrange’s equations 54
1.3 Dynamic degrees of freedom 9 1.8.1 Derivation of
1.4 Dynamic model and Lagrange’s equations 54
formulation of the equation 1.8.2 Lagrange multipliers 64
of motion of SDOF systems 11 1.8.3 Small displacements 68
1.5 Derivation of the equations 1.8.4 Raleigh’s dissipation
of motion using d’Alembert’s function 72
principle 15 1.9 Influence of the gravity loads 73
1.6 Principle of virtual 1.10 Problems 74
displacements 38 References and further reading 82

1.1 Introduction
Apart from static loads, engineering structures may be subjected to dynamic
loads, that is, loads whose magnitude as well as direction of action and/or posi-
tion vary with time. The analysis of stresses and deflections developed in a
given structure undergoing dynamic loads is the fundamental objective of the
dynamic analysis of structures. Between static and dynamic analysis of struc-
tures, there exist two substantial differences:
(a) In static analysis, the loads are assumed time-invariant, and the resulting
response is unique, at least in linear theory. On the other hand, in dynamic
analysis the loads are time-varying and the deformations and stresses
depend on time, that is, at each instant the response of the structure is
different.
(b) In dynamics analysis, the material points of the structure change position
with the time, hence they have velocity and acceleration. Inasmuch as
the structure has a mass, inertial forces are produced due to the accelerations
of the material points. These inertial forces constitute an additional loading
that cannot be ignored. To make it tangible, we consider the cantilever beam
of Fig. 1.1.1a. The beam has a mass per unit length m and a flexural rigidity
EI , both assumed constant along the length, and it is subjected to the time-

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00001-7

© 2020 Elsevier Inc. All rights reserved. 3
4 PART I Single-degree-of-freedom systems

varying position-dependent distributed loading pðx, t Þ (Fig. 1.1.1a). We can

derive the equation of motion by considering the dynamic equilibrium of a
beam element of length dx (Fig. 1.1.1b). In static consideration, the element
would be in equilibrium under the action of the external load pðx, t Þdx and
the internal forces, that is, the shear force Q, the bending moment M at the left
cross-section of the element, and the shear Q + dQ and moment M + dM at
its right cross-section.

(a)

(b)

(c)
FIG. 1.1.1 Vibrating cantilever beam.
 General concepts and principles of structural dynamics Chapter 1 5

The transverse deflection is a function not only of the spatial variable x but
also of time t, namely it is u ¼ u ðx, t Þ. As the element has a mass mdx, an inertial
€ arises, which, according to d’Alembert’s principle (see Section 1.5)
force m udx
opposes the motion, that is, if the positive transverse displacement u ðx, t Þ in the
beam is directed downward, the inertial force is directed upward (see Fig. 1.1.1b
and c). Similarly, due to angular acceleration ∂u€ðx, t Þ=∂x of the cross-section, an
inertial moment is also developed, which we may neglect [1]. Thus, referring to
Fig. 1.1.1b, we obtain the equation of dynamic equilibrium of the beam element
in the y direction as

€ ¼0
Q + Q + dQ + pðx, t Þdx  m udx

or
∂Q ∂2 u
¼ pðx, t Þ + m 2 (1.1.1)
∂x ∂t

From the Euler-Bernoulli beam theory we have

∂3 u
Q ¼ EI (1.1.2)
∂x 3

Substituting Eq. (1.1.2) into Eq. (1.1.1) yields

∂4 u ∂2 u
EI + m 2 ¼ pðx, t Þ (1.1.3)
∂x 4 ∂t

Eq. (1.1.3) is known as the equation of the dynamic equilibrium or the equa-
tion of motion of the vibrating beam. It is apparent that if we omit the inertial
term m∂2 u=∂t 2 in Eq. (1.1.3), we obtain the equation of the deflection of the
beam under static loading, that is,

d4u
EI ¼ pðx Þ (1.1.4)
dx 4

Fig. 1.1.1c shows the beam subjected to the inertial forces. These forces
resist the accelerations and they need to be accounted for in the solution. This
is the most important characteristic of the dynamic problem. Obviously, the
magnitude of the inertial forces depends on the magnitude of the acceleration.
When the produced accelerations are very small, as in the case of slow motion,
the inertial forces are very small too, and they can be neglected. In this case,
the time appears in the equation as a parameter and the response at any instant
6 PART I Single-degree-of-freedom systems

can be obtained by the static structural analysis, even though the load and the
response are time-varying. This response is pseudodynamic and is referred to
as quasistatic. The inertial forces appear in the equation of motion of the struc-
ture with the second derivatives of the displacements with respect to time.
Therefore, the equations that must be solved in dynamic analysis in order
to establish the deformations and stresses in the structure are differential
equations, contrary to static analysis where the governing equations are alge-
braic. For this reason, the solution procedure in dynamic analysis is essentially
different from that used in static analysis.

1.2 Types of dynamic loads

As already mentioned, dynamic loads are time-varying. Such loads are those
due to an unbalanced rotating machinery, the motion of vehicles on structures,
blast loads, wind loads, etc. The motion of the support of a structure, even in the
absence of external dynamic loads, produces a dynamic response, too. This is
the case of seismic ground motion (see Chapter 6).

FIG. 1.2.1 Periodic load.

Dynamic loads can be classified into two great groups that characterize the
approach of evaluating the structural response: The deterministic dynamic loads
and the nondeterministic or random dynamic loads. In the first group are the
dynamic loads whose time variation is fully determined, regardless of the
complexity of their mathematical presentation. They are also referred to as pre-
scribed dynamic loads. They can be represented by an analytic or a generalized
function (Dirac or Heaviside) as well as numerically by a set of their values at
discrete time instances. The second group includes the loads, whose time var-
iation is not completely known but it can be defined in a statistical sense. In this
book, the dynamic response of structures only under deterministic loads is
studied.
 General concepts and principles of structural dynamics Chapter 1 7

5
2
H = me sin t

2.5

0
H(t)

–2.5

T T T
–5
0 2 4 6 8 10
t
FIG. 1.2.2 Harmonic loading due to an unbalanced rotating mass.

From the analytical point of view, it is convenient to divide the deterministic

loads into two basic categories, periodic and nonperiodic loads. Periodic loads
are those whose time variation profile repeats continually at regular time inter-
vals T . Mathematically, they can be represented by a periodic function
pðt Þ ¼ pðt + nT Þ (1.2.1)
where n is a natural number. The time interval T is called the period of the load.
A general type of periodic load is shown in Fig. 1.2.1, which also identifies the
period of the load. A usual type of periodic load is the harmonic load caused by
an unbalanced rotating machine (Fig. 1.2.2); H ¼ mew2 cos wt is the horizontal
component of the centrifugal force. Loads that do not show any periodicity are
called nonperiodic loads. They may be of long duration, such as those resulting
from an earthquake. Nonperiodic loads of short duration are called impulsive
loads. A windblast striking a building and the pressure of a bomb explosion
on a structure are typical impulsive loads (Fig. 1.2.3). The earthquake produces
a special type of load, which is due to the excitation of the support of the
8 PART I Single-degree-of-freedom systems

structure and, as we will see later in this book (Chapter 6), it can be reduced to an
effective dynamic load if the accelerogram of the ground motion is known
(Fig. 1.2.4).

FIG. 1.2.3 Nonperiodic load due to explosion.

400

–400
0 10 20 30

FIG. 1.2.4 Effective dynamic load pðt Þ ¼ m u€g ðt Þ due to seismic ground motion.

If we examine static loading closer, we will see that even what we call static
loads are actually dynamic in nature. They are applied starting from a zero value
until the final prescribed value is reached within a time span. That is, they are
time-varying, thus dynamic. However, the duration of the application of the
static load is longer than the period of vibration of the structure. This produces
negligible accelerations and consequently the response under a “static load”
 General concepts and principles of structural dynamics Chapter 1 9

could be considered as a special case of the dynamic problem with negligible

accelerations, that is, quasistatic.

1.3 Dynamic degrees of freedom

The displacement method is the most suitable method for the dynamic analysis
of structures. In this method, the unknowns are the displacements. For structures
with distributed mass, the displacements are functions of the spatial coordinates
and of the time as well. Their dynamic response is described by partial differen-
tial equations of the hyperbolic type, which must be solved in order to determine
these displacements. The solutions of such equations belong to the most difficult
problems of mathematics. The available solutions refer to simple structures, for
example, beams with a constant cross-section, which are not adequate to study
the dynamic response of engineering structures. Thanks to the development of
modern computational methods such as FEM, BEM and Meshless Methods, the
actual structure is approximated by discrete models in which the mass is local-
ized at a finite number of points (nodal points). These models are adequate to
represent the effects of all significant inertial forces of a structure. At each
instant, the deformed configuration of the structure is determined from the dis-
placementsa of the nodal points, which are functions only of time. The response
of the discretized structure is governed by ordinary differential equations, which
are easy to solve analytically or at least numerically. The number of the indepen-
dent nodal displacements required to determine the deformed shape of the mov-
ing structure is called the number of degrees of freedom. It is apparent that
continuous systems have an infinite number of dynamic degrees of freedom.
Structures with one degree of freedom are called single-degree-of-freedom
(SDOF) systems. Accordingly, we have two-degree-of-freedom (2 DOF) sys-
tems, three-degree-of-freedom (3 DOF) systems, and generally multi-degree-
of-freedom (MDOF) systems. Fig. 1.3.1 shows SDOF systems. Fig. 1.3.1a rep-
resents the idealization of a silo. It consists of two massless columns and a square
rigid plate of mass m. With the assumption that the axial deformation of the col-
umns is negligible, the horizontal displacement u ðt Þ is adequate to completely
determine the motion of the system. Hence, the system has one degree of free-
dom. Likewise, under the same assumptions for the columns, the motion of
the water tower of Fig. 1.3.1b can be determined from the angle fðt Þ.
Fig. 1.3.1c represents the typical model of a SDOF. Fig. 1.3.2a represents the
model of the two-story shear frame. To determine its motion, it is necessary to
establish the two independent horizontal displacements u1 ðt Þ and u2 ðt Þ.
Fig. 1.3.2b represents a cantilever column with a mass atop. This may be
considered as the idealization of a water tower. The whole mass is lumped at
the top while the column is massless. During the motion, the mass undergoes
horizontal displacement u ðt Þ and rotation fðt Þ. These two geometrical quanti-
ties are independent. Hence, the system has two degrees of freedom and thus

a. The term displacent denotes both translation and rotation.

10 PART I Single-degree-of-freedom systems

two differential equations of motion are necessary to determine these displace-

ments. If, however, the mass m of the system is assumed to be concentrated at a
point, its rotational inertia I is equal to zero. Hence, the inertial moment I f€ðt Þ is
zero, too, and one of the equations of motion becomes algebraic. This permits
the elimination of the rotational displacement, leading to only one equation of
motion for u ðt Þ. Consequently, the system has only one dynamic degree of free-
dom, even though it has two static degrees of freedom. Apparently, the number
of static degrees of freedom is not necessarily equal to the number of dynamic
degrees of freedom. Concluding, we can say that in an MDOF system, the num-
ber of dynamic degrees of freedom is equal to the number of independent dif-
ferential equations of motion that must be formulated to establish the dynamic
response of the system.

Massless
Massless
columns
columns

(a) (b)

(c)
FIG. 1.3.1 Systems with one degree of freedom (SDOF).

Rigid

Rig

(a) (b)
FIG. 1.3.2 Systems with two degrees of freedom (2 DOF).
 General concepts and principles of structural dynamics Chapter 1 11

FIG. 1.3.3 Multi-degree-of-freedom (MDOF) system.

FIG. 1.3.4 System with infinite degrees of freedom. Continuous system.

The lumped mass idealization provides a simple means of reducing the num-
ber of degrees of freedom. Fig. 1.3.3 represents the discrete model of a canti-
lever column, whose mass has been localized at three points. Neglecting the
axial deformation of the column and considering plane motion, the system
has six degrees of freedom, the three transnational ui ðt Þ and the three rotational,
fi ðt Þ. If the masses are fully concentrated so that their rotational inertia can be
ignored, the inertial moments Ii f€i are zero and the number of dynamic degrees
of freedom reduces to three. Obviously, the number of degrees of freedom
increases with the number of nodal points, where the mass of the structure is
lumped. As the number of points becomes infinitely large, the discretized struc-
ture approaches the continuous system (Fig. 1.3.4).

1.4 Dynamic model and formulation of the equation

of motion of SDOF systems
The modeling of the real structure plays a fundamental role in the dynamic anal-
ysis of structures. It is the most difficult task in dynamic analysis because in this
stage of analysis, the experience and theoretical background of the engineer
intervene critically in approximating the structural response.
12 PART I Single-degree-of-freedom systems

Spring

Damper

Frictionless rollers
FIG. 1.4.1 Model of a SDOF system.

Fig. 1.4.1 shows a simple dynamic model of a SDOF system. It consists of a

rigid body of mass m constrained to move along the x axis in the plane of the
paper, a weightless spring connecting the mass to the firm support, and a damper.

Center of mass

FIG. 1.4.2 Forces applied to the free body.

The forces applied to the body at time t are shown in the free body diagram
of Fig. 1.4.2. These are
(a) The external load pðt Þ
(b) The elastic force fS
(c) The damping force fD
(d) The inertial force fI .
The spring force fS depends on the displacement u ðt Þ and it is generally
expressed by a nonlinear function, fS ¼ fS ðu Þ. For linear response of the struc-
ture, the force fS is proportional to the displacement and is given by
fS ¼ ku (1.4.1)
where k is the constant that represents the spring stiffness coefficient, that is, the
force required to change the length of the spring by a unit. The force fS repre-
sents the elastic force of the structure that resists the motion and tends to bring
the body to its initial undeformed position.
The damping force fD also resists the motion. It represents the energy loss
due to internal or external dissipative forces. Damping forces are complex in
nature. Their exact expression in terms of the parameters of motion and of
the geometrical and material properties of the structure is complicated and dif-
ficult to determine. The simplest form of damping is linear viscous damping.
This produces damping forces, which are the easiest to handle mathematically
 General concepts and principles of structural dynamics Chapter 1 13

and provide analytical results for the response of a system close to the experi-
mental ones. The linear viscous damping mechanism is indicated by a dashpot,
as shown in Fig. 1.4.1. In viscous damping, the resisting force is proportional to
the velocity
fD ¼ cu_ (1.4.2)
where c is a constant that can be established experimentally. Inasmuch as the
work done by this force is converted to heat, the damping force is a nonconser-
vative force. It is the force that makes the amplitude of a vibrating
structure decay.
The inertial force fI depends on the mass m of the body and its acceleration
€ It also resists the motion. It is given by Newton’s second law of motionb
u.
fI ¼ m u€ (1.4.3)
A simple example of a structure that can be modeled as SDOF is the
one-story, one-bay frame of Fig. 1.4.3. It consists of two identical weightless
columns fixed on the ground and having height h, cross-sectional moment of
inertia Ι, and modulus of elasticity E. The cross-sectional moment of inertia
of the horizontal beam is assumed infinitely large. This means that the beam
behaves like a rigid body of mass m and hence the cross sections of the columns
at the roof level cannot rotate when the frame deforms. The frame is subjected to
an external horizontal force pðt Þ, as shown in Fig 1.4.3a, which forces the frame
to move. Neglecting the axial deformation of the beam and columns, an allow-
able assumption for frames, the only possible movement is the displacement
u ðt Þ at the roof level. The rotation of the beam as a rigid body is excluded
because this would cause a change in the length of columns.

(a)

(b) (c)
FIG. 1.4.3 Two-column shear frame.

b. Actually, this form of Newton’s law of motion is attributed to L. Euler, who defined it indepen-
dently as a mechanical principle [2, 3]. This law was recently derived from Kepler’s laws of plan-
etary motion [4].
14 PART I Single-degree-of-freedom systems

Referring to Fig. 1.4.3b, we see that the elastic forces are the shear forces Q
at the top cross-sections of the columns. These forces are given by the known
relation of statics
12EI
Q¼ u ðt Þ (1.4.4)
h3
The quantity 12EI =h 3 represents the translational stiffness of the column.
This is the force required to produce a unit relative displacement between
the end cross-sections of the column. These shear forces tend to restore the
frame to the undeformed position. Therefore, they play the role of the spring
in the SDOF model with a stiffness coefficient
12EI
k ¼2 (1.4.5)
h3
The inertial force is given by fI ¼ m u€ while the damping force by fD ¼ cu. _
Another convenient model to represent the single-story frame is shown in
Fig. 1.4.3c. It consists of a mass m placed at the top of a column with transla-
tional stiffness equal to the sum of the translational stiffness coefficients of the
columns of the frame. During the motion, the top cross-sections of columns
undergo only the translational displacement u ðt Þ. Models of this type are also
suitable to idealize multistory shear frames (see Fig. 1.4.4), in which the masses
are placed at the floor levels and the girders are assumed rigid.

FIG. 1.4.4 Four-story shear frame and its model without damping.

(a) (b)
FIG. 1.4.5 Two-story, two-bay shear frame and its model without damping.
 General concepts and principles of structural dynamics Chapter 1 15

Fig. 1.4.5a shows another two-story shear frame. The columns are assumed
weightless. Fig. 1.4.5b shows its dynamic model. The column 1-2 is represented
by a spring of stiffness k ¼ 12EI =h 3 . The stiffness coefficients k1 and k2 include
only the stiffness of the columns with heights h1 and h2 , respectively.
Given the dynamic model of the structure, the equation of motion of the sys-
tem is formulated. For the SDOF system, the equation of motion can be formu-
lated using Newton’s second law of motion as it is applied for the motion of a
particle
m u€ ¼ F (1.4.6)
where
F ¼ pðt Þ  fS  fD (1.4.7)
is the resultant of the external forces. Using Eqs. (1.4.1), (1.4.2), (1.4.7),
Eq. (1.4.6) is written
m u€ + cu_ + ku ¼ pðt Þ (1.4.8)
Eq. (1.4.8) is the equation of motion of the SDOF system. The equation of
motion represents the dynamic equilibrium of the system. It is an ordinary dif-
ferential equation of the second order with respect to the unknown variable u ðt Þ.
The solution of this equation yields the displacement as a function of time. For
MDOF systems, the number of equations of motion that must be formulated is
equal to the number of dynamic degrees of freedom. The use of Newton’s law of
motion is not always well suited to formulate the equations, especially for
MDOF systems or complex SDOF systems. It requires advanced knowledge
of the dynamics of the rigid and deformable body as well as mastering various
special methods. Generally, the equations of motion can be formulated using:
(a) d’Alembert’s principle or method of equilibrium of forces.
(b) Principle of virtual work.
(c) Hamilton’s principle.
(d) Lagrange’s equations.
These methods will be presented in the following and will be demonstrated by
appropriate examples. The acquaintance with the application of these methods
constitutes a fundamental presupposition for the analysis of the dynamic
response of structures.

1.5 Derivation of the equations of motion using

d’Alembert’s principle
Actually, d’Alembert’s principle is a different interpretation of Newton’s
second law of motion. Suppose that we write it in the form
F  m€
u¼0 (1.5.1)
16 PART I Single-degree-of-freedom systems

where F is the resultant of all external forces acting on the particle of mass m and
€ is its acceleration with respect to an inertial frame of reference.c If we consider
u
that the term m€ u is another force, known as inertial force, then Eq. (1.5.1) states
that the vector sum of all forces, external and inertial, is zero during the motion.
But this is the necessary and sufficient condition for the static equilibrium of the
particle. Thus, in a sense, the dynamic problem is reduced to a problem of statics
according to the following statement, known as d’Alembert’s principle.
The laws of static equilibrium can be applied also to a dynamic system with
respect to an inertial frame of reference if the inertial forces are considered as
applied forces on the system together with the actual external forces.
The motion of a rigid body of mass m with respect to an inertial frame of
reference X, Y ,Z is decomposed into a translational motion of its center of
mass, where the whole mass is considered to be concentrated, and a rotational
motion about it (Fig. 1.5.1).

FIG. 1.5.1 Rigid body moving with respect to the inertial. frame X,Y , Z .

If R ¼ X ðt Þi + Y ðt Þj + Z ðt Þk is the position vector of a particle A of

the body with respect to the inertial system of axes XYZ and
r ¼ x ðt Þe1 + y ðt Þe2 + z ðt Þe3 the position of the same point with respect to
the nonrotating system of axes xyz through the center of mass C (see
Fig. 1.5.1), then the equations of motion of the body can be written as
€c
F ¼ mR (1.5.2a)
_c
Mc ¼ H (1.5.2b)
where F ¼ Fx i + Fy j + Fz k is the resultant of the external forces,
€ c ¼ X€ ci + Y€ c j + Z€c k is the acceleration of the center of mass,
R
Mc ¼ Mx e1 + My e2 + Mz e3 is the resultant moment of the external forces with

c. In classical dynamics, an inertial frame of reference is a frame of reference in which a body with
zero force acting upon it is not accelerating; that is, the body is at rest or it is moving at a constant
velocity in a straight line [5].
 General concepts and principles of structural dynamics Chapter 1 17

respect to the center to mass, and H_ c is the rate of change of the angular momen-
tum Hc of the body with respect to the same point given as
ZZZ
H_ c¼ r  r€rdV (1.5.3)
V

in which r ¼ rðx, y, z Þ is the mass density of the body.

Eq. (1.5.2a) is the equation of the translational motion while Eq. (1.5.2b) is
the equation of the rotational motion.
For a plane body moving in its plane, Eqs. (1.5.2a), (1.5.2b) become (see
Appendix)
Fx ¼ m X€ c (1.5.4a)
Fy ¼ m Y€ c (1.5.4b)
Mc ¼ Ic w_ (1.5.4c)
where w is the angular velocity of the rotational motion about the center of mass
and Ic the polar moment of inertia of the body about the same point.

Path of P

FIG. 1.5.2 Plane body moving in the XY plane. The system of xy axes moves with P without
rotating.

It is often convenient to study the motion with reference to an arbitrary point

P, which is not the center of mass of the body (see Fig. 1.5.2). Special attention
should be paid in this case because Eqs. (1.5.2a), (1.5.2b) take the form
€ p ¼ m€rc
F  mR (1.5.5a)
€p¼H
Mp  rc  m R _p (1.5.5b)
where Rp is the position vector of point P moving with the body and rc the posi-
_ p are the moment of the
tion vector of the center of mass with respect to P. Mp , H
external forces and the rate of change of the angular momentum with respect to
P, respectively.
18 PART I Single-degree-of-freedom systems

When small displacements are considered, as in the theory of linear vibra-

tions, Eqs. (1.5.5a), (1.5.5b) become (see Appendix).
 
Fx ¼ m X€ p  yc w_ (1.5.6a)
 
Fy ¼ m Y€ p + xc w_ (1.5.6b)
 
Mp ¼ Ip w_ + m xc Y€ p  yc X€ p (1.5.6c)
The kinetic energy of a plane body moving in its plane is given
(a) with respect to the center of mass

1  2  1
T ¼ m X_ c + Y_ c + Ic w2
2
(1.5.7)
2 2
(b) with respect to an arbitrary point P of the body (K€onig’s theorem)

1  2  1  
T ¼ m X_ p + Y_ p + Ip w2 + m xc Y_ p  yc X_ p w
2
(1.5.8)
2 2
We shall write now Eqs. (1.5.4a)–(1.5.4c) in terms of the displacement
vector. Apparently, the displacement vector from the beginning of the motion
is defined as
u ¼ Rðt Þ  Rð0Þ ¼ u ðt Þi + v ðt Þj (1.5.9)
where
u ¼ X ðt Þ  X ð0Þ, v ¼ Y ðt Þ  Y ð0Þ (1.5.10)
Hence, X€ ¼ u,
€ Y€ ¼ v€. Moreover, if fðt Þ represents the change of the rota-
_ w_ ¼ f,
tion in the same time interval and set w ¼ f, € Eqs. (1.5.4a)–(1.5.4c) are
written in terms of displacements as
Fx ¼ m u€c (1.5.11a)
Fy ¼ m v€c (1.5.11b)

Mc ¼ Ic f€ (1.5.11c)
or in matrix form
€c
Fc ¼ m c U (1.5.12)
where
8 9 8 9 2 3
< Fx = < uc = m 0 0
Fc ¼ Fx , Uc ¼ vc , mc ¼ 4 0 m 0 5 (1.5.13)
: ; : ;
Mc f 0 0 Ic
are the force vector, the displacement vector, and the mass matrix of the body,
respectively.
 General concepts and principles of structural dynamics Chapter 1 19

Similarly, Eqs. (1.5.6a)–(1.5.6c) are written

 
Fx ¼ m u€p  yc f€ (1.5.14a)
 
Fy ¼ m v€p + xc f€ (1.5.14b)
 
Mp ¼ Ip f€ + m xc v€p  yc u€p (1.5.14c)
or in matrix form

€p
Fp ¼ m p U (1.5.15)

where
8 9
>
< Fx >
=
Fp ¼ F x (1.5.16a)
>
: >
;
Mp
8 9
< up >
> =
Up ¼ vp (1.5.16b)
: >
> ;
f
2 3
m 0 my c
mp ¼ 4 0 m mx c 5 (1.5.16c)
my c mx c Ic
Note that the mass matrix is not diagonal when the point of reference is not
the center of mass.
Finally, Eqs. (1.5.7), (1.5.8) are written as

1   1
T ¼ m u_ 2c + v_ 2c + Ic f_
2
2 2
(1.5.17)
1_T _c
¼ U mc U
2 c

1   1  
T ¼ m u_ 2p + v_ 2p + Ip f_ + m xc u_ p  yc v_ p f_
2
2 2
(1.5.18)
1_T _p
¼ U mp U
2 p

The set of equations with reference to point P can also be derived from the
set of equations with reference to point C by transforming the displacements
and the forces from point C to P (see Section 10.7).
Example 1.5.1 Equation of motion of an elastically supported body
Consider the rigid plate of constant thickness and total mass m shown in
Fig. E1.1a. The plate is hinged at O and elastically supported at A. Formulate
20 PART I Single-degree-of-freedom systems

the equation of motion of the system for small amplitude motion using the
method of equilibrium of forces.
Solution
The only possible motion of the plate is the rotation in its plane about the point
O. Hence, the system has one degree of freedom. The motion can be described
either by the rotation fðt Þ about O or by the translational displacement of
a point, for example, the displacement u ðt Þ of point B, which is related
to fðt Þ as

u ðt Þ ¼ a tan fðt Þ  afðt Þ, u ðt Þ ¼ BB 0 cos f  BB 0 (1)

because we have assumed small displacements. Moreover, AA0 ¼ BB 0 =2 ¼ u=2

The applied forces are shown in the free body diagram in Fig. E1.1b. These
are:
The weight of the body:
W ¼ mg (2a)
The spring force:
2
fS ¼ k ðAA0 Þ ¼ ku (2b)
3
The inertial force at the center of mass along x

d2 1 b
fIx ¼ m ðCC 0 Þx ¼ m u€ (2c)
dt 2 2 a

The inertial force at the center of mass along y

d2 1
fIy ¼ m 2
ðCC 0 Þy ¼ m u€ (2d)
dt 2

The inertial moment about the center of mass

u€
MIc ¼ IC f€ ¼ IC (2e)
a

The external force

pðt Þ (2f)

The quantities ðCC 0 Þx and ðCC 0 Þy are the horizontal and the vertical dis-
placements of the center of mass C due to rotation, respectively. They are
obtained from Fig. E1.1b as
 General concepts and principles of structural dynamics Chapter 1 21

(a)

(b)

(c)
FIG. E1.1 Rigid plate in Example 1.5.1.

1b
ðCC 0 Þx ¼ ðOC Þf sin b ¼ u (3a)
2a
1
ðCC 0 Þy ¼ ðOC Þf cos b ¼ u (3b)
2
The equation of motion results from the dynamic equilibrium of moments
with respect to point O. Thus, we obtain

a 2a b a
W  fS  fIx  fIy  MIc + pðt Þa ¼ 0 (4)
2 3 2 2
22 PART I Single-degree-of-freedom systems

which by virtue of Eq. (2) becomes

" #
a 2 b2
m + + IC
2 2 4 W
u€ + ku ¼ + pðt Þ (5)
a2 9 2
Using Steiner’s formula, we have
" #
a 2 b2 a 2 + b2
IO ¼ m + + IC ¼ m (6)
2 2 3

Hence, Eq. (5) becomes

a 2 + b2 4 W
m 2
u€ + ku ¼ + pðt Þ (7)
3a 9 2
Eq. (7) can be also obtained if we consider the motion with reference to point
O and employ Eq. (1.5.14c) for
u€ 2a a
u€O ¼ v€O ¼ 0, f€ ¼ IO , MO ¼ fS + W + pðt Þa, MIo ¼ IO f€
a 3 2
The weight W can be eliminated from Eq. (7), if the total displacement u ðt Þ
is expressed as the sum of the static displacement ust caused by the weight plus
the additional dynamic displacement uðt Þ, as shown in Fig. E1.1c, that is,
u ðt Þ ¼ ust + uðt Þ (8)
The static equilibrium of moments with respect to point O, when the plate is
loaded only by the weight yields (see Fig. E1.1c)
4 a
kau st ¼ W (9)
9 2
Noting that u€st ¼ 0 because ust is a constant, and using Eqs. (8), (9), Eq. (7)
becomes
m ∗ u€
 + k ∗ u ¼ p ∗ ðt Þ (10)
where
a 2 + b2 4
m∗ ¼ m 2
, k ∗ ¼ k, p ∗ ðt Þ ¼ pðt Þ
3a 9
Eq. (10) has the form of Eq. (1.4.8) and represents the equation of motion of
the system. The quantities m ∗ , k ∗ , having dimensions of mass and translational
stiffness, respectively, are referred to as the generalized mass and the general-
ized stiffness of the SDOF system.
If the rotation fðt Þ, measured from the position of static equilibrium, is
taken as the parameter of motion in place of uðt Þ, the equation of motion results
from Eq. (10) using the relation u ¼ fa. Thus, we have
4
IO f€ + ka2 f ¼ apðt Þ (11)
9
 General concepts and principles of structural dynamics Chapter 1 23

Example 1.5.2 Equation of motion of a frame with a rigid column

Formulate the equation of motion of the plane frame shown in Fig. E1.2a for
small amplitude motion. The mass of the horizontal beam CD is negligibly small
while the column of height L and nonnegligible width h ¼ L=4 is assumed rigid
 The elastic stiffness of the ground is simulated by the
with total mass m ¼ mL.
spring CR while its damping by the two dashpots with damping parameters c.

(a)

(b)
FIG. E1.2 Frame with a rigid column in Example 1.5.2.

Solution
The only possible motion of the system is the rotation fðt Þ of the column as a
rigid body about the hinged support at point A of its base. Because the rotation is
small, we have:
sin f  f, cos f  1, f2  0
Hence
h h
u ¼ L sin f ¼ Lf, d ¼ sin f  f
2 2
24 PART I Single-degree-of-freedom systems

h h
ð1  2Þ ¼ sin f + L cos f  f + L,
2 2
h h
ð3  4Þ ¼ L sin f + cos f  Lf + ,
2 2
h h
ð5  6Þ ¼ cos f 
2 2
The forces applied on the column are shown in Fig. E1.2b. These are:
The elastic moment at the corner C MS ¼ ð1:5L
6EI
Þ2
4EI
d + 1:5L f
The elastic moment due to the rotational spring MR ¼ CR f ¼ EI
L f

MIA ¼ IA f€ ¼ mL €
2
The moment of inertia of the mass m 3 f

The elastic shear force at the beam end C QS ¼ ð1:5L

12EI
Þ3
6EI
d + ð1:5L Þ2
f
The damping forces fD ¼ c h2 f_
The external load pðt Þ
The equilibrium of moments with respect to point A yields

MIA + MS + MR + QS  ð3  4Þ + 2fD  ð5  6Þ  pðt Þ  ð1  2Þ ¼ 0

which after substituting their exressions becomes

 
mL2 € cL2 _ 79EI pL 28EI 2
f+ f+  f+ f ¼ Lpðt Þ (1)
3 32 18L 8 9L

Further, linearizing (f2  0) gives

 
mL2 € cL2 _ 79EI pL
f+ f+  f ¼ Lpðt Þ (2)
3 32 18L 8

If the displacement u ¼ Lf at the level of the beam is taken as the parameter

of the motion, the equation of motion becomes
 
m cL 79EI p
u€ + u_ +  u ¼ pðt Þ (3)
3 32 18L3 8L

Example 1.5.3 Equation of motion of a system of rigid bodies

The rigid body assemblage shown in Fig. E1.3a consists of the rigid bar AF of
total mass m hinged at A, and the rigid square plane body supported rigidly at F.
The dynamic excitation of the bar is due to the uniformly distributed transverse
load pðt Þ. The motion is constrained by a spring at B and the damper at G. For-
mulate the equation of motion of the system for small amplitude displacements
using the method of equilibrium of forces. The mass per unit length of the bar is
m ¼ m=3L and the surface mass density of the body is g ¼ 2m=L2
 General concepts and principles of structural dynamics Chapter 1 25

(a)

(b)
FIG. E1.3 System with two rigid bodies in Example 1.5.3.

Solution
As the bar AF is rigid, the only possible motion is its rotation about A. Hence,
the system has a SDOF. Its motion can be described either by the angle of rota-
tion fðt Þ about the hinge at A or by the transverse displacement of any point
along the axis of the bar. We choose the upward displacement u ðt Þ at point
B as the parameter of the motion. For small amplitude motion, the forces acting
on the system are shown in Fig. E1.3b. These are:
The elastic force fS at B: As it opposes the motion, it is directed downward
and is expressed as
fS ¼ ku (1)
The damping force fD at G: It is directed also downward and is expressed as

d d
fD ¼ c ðGG 0 Þ ¼ c ð1:625u Þ ¼ 1:625cu_ (2)
dt dt
The inertial force fIK and the inertial moment MIK at the center of mass K of
the bar due the distributed mass m  are

d2
 Þ
fIK ¼ ðm3L ðKK 0 Þ ¼ 0:75m u€ (3)
dt 2

MIK ¼ IK f€
26 PART I Single-degree-of-freedom systems

or taking into account that

ð3LÞ3  m  ð3LÞ3

IK ¼ m ¼ ¼ 0:75mL2
12 3L 12
u u u€
f ¼ ¼ 0:5 , f€ ¼ 0:5
2L L L
we obtain
MIK ¼ 0:375mLu€ (4)
The inertial force fIG and the inertial moment MIG at the center of mass G of
the rigid body due to the mass gL2 =2:
 
L d2 2m L2 d 2
fIG ¼ gL 2
ðGG 0 Þ ¼ 2 ð1:625u Þ ¼ 1:625m u€ (5)
2 dt L 2 dt 2
" # 
€ LðL=2Þ3 ðL=2ÞL3 u€
M I ¼ IG f ¼ g
G
+ 0:5 ¼ 0:052mLu€ (6)
12 12 L

The external load is 2Lpðt Þ.

The equilibrium of the moments about A yields the equation of motion of the
system. Thus, we have

fS  ð2LÞ  fD  ð3:25LÞ  fIK  ð1:5LÞ  fIG  ð3:25LÞ

(7)
MIK  MIG + pðt Þ  ð2LÞ  L ¼ 0

or inserting Eqs. (1)–(6) into Eq. (7) we obtain

m ∗ u€ + c ∗ u_ + k ∗ u ¼ p ∗ ðt Þ (8)

where
m ∗ ¼ 6:833m, c ∗ ¼ 5:281c, k ∗ ¼ 2k, p ∗ ðt Þ ¼ 2L
pðt Þ (9)
The quantities defined by Eq. (9) are referred to as the generalized mass, the
generalized damping, the generalized stiffness, and the generalized load,
respectively.
Once the dynamic displacement u ðt Þ is established from the solution of
Eq. (8), the vertical reaction RA can be evaluated from the dynamic equilibrium
of forces in the direction of the y axis. This yields

RA + pðt Þ2L  fIK  fIG  fS  fD ¼ 0

or using Eqs. (1)–(3), (5) we obtain

RA ¼ ku + 1:625cu_ + 2:375m u€  2L

pðt Þ
 General concepts and principles of structural dynamics Chapter 1 27

Example 1.5.4 Equation of motion of a single-story shear building

Formulate the equation of motion of the single-story building shown in Fig. E1.4a.
The damping is neglected. The columns are fixed on the ground, are inextensible,
and their mass is assumed to be lumped at their ends. Moreover, the roof plate
is assumed rigid. The material of the structure is reinforced concrete having spe-
cific weight g ¼ 24kN=m3 and modulus of elasticity E ¼ 2:1  107 kN=m2 . The
total load of the plate (dead and live) is 20kN=m2 . The force pðt Þ acts in the direc-
tion of the x axis and is given by pðt Þ ¼ 20sin 13t. The acceleration of gravity is
g ¼ 9:81m=s2 and the dimensions of the rectangular cross-sections of columns are
k1 : 30  30cm2 and k2 : 30  20cm2 .

(b)

(a)
FIG. E1.4 Single-story shear building in Example 1.5.4.

Solution
Taking into account that the structure is symmetric with respect to the x axis,
the columns are inextensible, and the load pðt Þ acts on the axis of symmetry, the
only possible motion of the plate is the horizontal displacement u ðt Þ in the
direction of the x axis. The SDOF model of the structure is shown in Fig. E1.4b.
The total mass of the system is due to the load of the plate and to half the
weight of the columns
5  10  20 + ð4  0:3  0:3 + 2  0:3  0:2Þ  2  24
m¼ ¼ 104:285
9:81
The stiffness of the system is equal to the sum of the translational stiffness
coefficients of all columns, which are given as
12EI i
ki ¼
hi3
where Ii is the moment of inertia of the cross-section of the i column with
respect to the y axis through its mass center and hi its height. Thus, we have
28 PART I Single-degree-of-freedom systems

Columns 30  30:
0:304
12  2:1  107 
k3030 ¼ 12 ¼ 2657:8kN=m
43
Columns 30  20:
0:303  0:20
12  2:1  107 
k3020 ¼ 12 ¼ 1771:9kN=m
43
Therefore the stiffness of the system is
k ¼ 4  2657:8 + 2  1771:9 ¼ 14175:0kN=m
The equation of motion results from the equilibrium of the forces shown in
Fig. E1.4b. This yields
fI  fS + pðt Þ ¼ 0
or
m u€ + ku ¼ pðt Þ
Substituting the numerical values for m, k and the expression for pðt Þ, the
above equation of motion becomes

5:21u€ + 708:75u ¼ sin 13t

Example 1.5.5 Equation of motion of a two-story shear frame

Formulate the equations of motion of the two-story shear frame shown in
Fig. E1.5a using the method of equilibrium of forces. The damping is ignored.
Solution
The system has two degrees of freedom because the girders are rigid and the
axial deformation of columns is ignored. The model of the structure is shown
in Fig. E1.5b. The masses are lumped at the story levels. The motion of the
system can be fully determined from the horizontal displacements u1 ðt Þ and
u2 ðt Þ of the masses m1 and m2 , respectively.

Rigid

Rigid

(a) (b) (c)

FIG. E1.5 Two-story shear frame in Example 1.5.5.
 General concepts and principles of structural dynamics Chapter 1 29

The equations of motion result from the dynamic equilibrium of forces

applied to the masses m1 and m2 . These forces are shown in the free body dia-
grams of Fig. E1.5c. Thus, we obtain

m1 u€1 + k1 ðu1  u2 Þ ¼ p1 ðt Þ (1)

m2 u€2  k1 ðu1  u2 Þ + k2 u2 ¼ p2 ðt Þ (2)

Eqs. (1), (2) are written in matrix form as

m€
u + ku ¼ pðt Þ (3)

where


 
u1 m1 0 k1 k1 p1 ðt Þ
u¼ , m¼ , k¼ , pðt Þ ¼
u2 0 m2 k1 k1 + k2 p2 ðt Þ

Example 1.5.6 Equation of motion of a general single-story shear building

The rigid horizontal plate is supported by K columns as shown in Fig. E1.6. The
columns are fixed on the ground as well as on the plate. Their principal axes have
arbitrary directions in the xy plane. Formulate the equation of motion of the
plate when the plate is loaded by the horizontal load P  ðt Þ through the point A.

Solution
We choose O xy as the system of reference of the motion, whose origin coin-
cides with point O at the beginning of motion. Let xi , yi represent the coordi-
nates of the center of mass of the cross-section of i column and fi the angle
between its principal x axis and the x axis. The axes xy will be referred to
as the global axes of the system while the axes xy as the local axes of the
column.
Inasmuch as the axial deformation of columns is ignored, the only possible
motion of the plate is inside its plane, which can be determined by the two
translational displacements of a point and the rotation of the plate. We study
the motion of the plate with reference to point O and let U , V represent its
translational components with respect to the global axes xy, which are
assumed fixed in the plane, and W  the rotation of the plate. As a consequence
of this motion, the cross-section of the i column at the level of the plate
undergoes the displacements u i , v i , wi , with respect to its base. These displace-
ments generate elastic forces X i , Y i , M i , which act on the plate. Thus,
we define the following vectors and matrices that will be used in the subse-
quent analysis.
30 PART I Single-degree-of-freedom systems

FIG. E1.6 General single-story shear building in Example 1.5.6.

(a) In global axes:

8 9
< U =
 ¼ V displacements of point O of the plate
U
:;
W
8 9
< ui =
 i ¼ vi displacements of i column
D
: i;
w
8 i9
< X =
 i ¼ Y i elastic forces of i column
F S
: i;
M
(b) In local axes
8 9
< ui =
D ¼ v i displacements of i column
i
: i;
w
8 9
< Xi =
FiS ¼ Y i elastic forces of i column
: i;
M
 General concepts and principles of structural dynamics Chapter 1 31

The transformation matrix for the vector quantities related to i column from the
global axes to the local axes is given as
2 3
cos fi sin fi 0
Ri ¼ 4  sin fi cos fi 0 5
0 0 1
Hence, the vectors are transformed from one system of axes to the other as
i
D i ¼ Ri D (1a)
 
 i ¼ Ri T D i
D (1b)
i
FiS ¼ Ri F (2a)
S
 
 i ¼ Ri T F i
F (2b)
S S

where
2 3
  cos fi  sin fi 0
R ¼ 4 sin fi
i T
cos fi 0 5
0 0 1
 i 1  i T
is the transpose of R . Note that R
i
¼ R because Ri is orthonormal.
The elastic forces X , Y , M are related to the displacements u i , v i , wi by
i i i

12EI y i
Xi ¼ u (3a)
h3
12EI x i
Yi ¼ v (3b)
h3
GI t i
Mi ¼ w (3c)
h
where Ix ,Iy are the principal moments of inertia of the column cross-section and
It is the torsional constant, E and G are the material constants, and h is the
height of the column.
Setting
12EI y 12EI x GI t
i
k11 ¼ i
, k22 ¼ i
, k33 ¼ (4)
h3 h3 h
Eqs. (3a)–(3c) can be written in matrix form as
8 9 2 i 38 9
< Xi = k11 0 0 < u i =
Y i ¼ 4 0 k22 i
0 5 vi
: i; i : i;
M 0 0 k33 w
or
Fi ¼ k i D i (5)
32 PART I Single-degree-of-freedom systems

The matrix
2 i
3
k11 0 0
6 7
ki ¼ 4 0 i
k22 0 5
i
0 0 k22

is the stiffness matrix of i column.

Eq. (5) is transformed in global axes using Eqs. (2b), (1a). Thus, we have
 
 i ¼ Ri T F i
F S S
 T
¼ Ri k i D i
 T
i
¼ Ri k i R i D

or
F i D
i ¼ k i (6)
where
 
 i ¼ Ri T k i R i
k (7)

is the stiffness matrix of the column in global axes, which becomes after per-
forming the matrix multiplications
2 i 3
k11 k12 0
i

6 7
i ¼ 6 ki ki 0 7
k (8)
4 21 22 5
0 0 k
i
33

where
9
k11 ¼ k11 >
i i
cos 2 fi + k22
i
sin 2 fi >
>
>
>
>
=
k22 ¼ k11
i i
sin 2 fi + k22
i
cos 2 fi
 i  (9)
k12 ¼ k21 ¼ k11 sin fi cos fi >
i i
 k22
i >
>
>
>
>
;
i
k 33 ¼ k33
i

Inasmuch as the plate is rigid, the displacements ui , vi , wi of the i column
 of point O. The geometrical rela-
depend on the plate displacements U , V , W
tions result from the following consideration.
 General concepts and principles of structural dynamics Chapter 1 33

FIG. E1.7 Displacements of point i due to rotation of plate.

The point i of the plate undergoes translational displacements due to

(a) The translational displacements of point O
 i
u t ¼ U
 i
v ¼ V
t

(b) The rotation of the plate about O. Referring to Fig. E1.7 and observing that
 ¼ wi , we obtain
cos ai ¼ xi =ri , sin ai ¼ yi =ri , W
 i
u r ¼ ri W  sin ai ¼  
yi W
 i
v r ¼ ri W cosai ¼ xi W


Thus, we have
   
ui ¼ ui t + ui r ¼ U  yi W (10a)
   

vi ¼ vi t + vi r ¼ V + xi W (10b)

wi ¼ W (10c)
The previous equations are written in matrix form as
8 9 2 38 9
< ui = 1 0  y i < U =
vi ¼ 4 0 1 xi 5 V (11)
: i; :;
w 0 0 1 W
or setting
2 3
1 0 
yi
ei ¼ 4 0 1 xi 5
0 0 1
34 PART I Single-degree-of-freedom systems

we can write Eq. (11) as

 i ¼ ei U
D  (12)
The matrix e defined by Eq. (12) is referred to as the translation matrix or
i

transformation matrix of the i column.

The equations of motion result from Eqs. (1.5.14a)–(1.5.14c) if they are
employed for point O, with P  O, up ¼ U , vp ¼ V , f ¼ 
 W. The external force
 ðt Þ ¼ P x ðt Þ, P y ðt Þ T . Thus,
F is equal to the sum of all elastic forces FiS plus P
we have
X
K  
P x ðt Þ  X ¼ m U€
i €
  yc W (13a)
i¼1

X
K  
P y ðt Þ  Y ¼ m V€
i €
 + xc W (13b)
i¼1

K 
X   
 ðt Þ 
M xi Y  yi X + M
i i €
 i ¼ m xc V€  yc U€ + Io W (13c)
i¼1

where M  ðt Þ with respect

 ðt Þ ¼ xA P y ðt Þ  yA P x ðt Þ is the moment of the force P
to O, m is the mass of the plate, and Io its moment of inertia with respect to O.
Eqs. (13a)–(13c) are written in matrix form as
8 9 2 38 i 9 2 32 32 38 9
€
< P x ðt Þ >
> = XK 1 0 0 > < X >= 1 0 0 m 0 0 yc >
1 0  < U >
=
6 7 6 76 76 7
P y ðt Þ  4 0 1 0 5 Y i
¼ 4 0 1 0 54 0 m 0 54 0 1 xc 5 V€
>
:  >
; i¼1 >
: i> ; >
: € >
;
M ðt Þ y i xi 1
 M y c xc 1
 0 0 Ic 0 0 1 W

or
X
K  
 ðt Þ 
P
T i
ei F €
 ¼ ðec ÞT mðec ÞU (14)
i¼1

where
8 9

< P x ðt Þ >
> =
 ðt Þ ¼ P y ðt Þ
P (15a)
:  >
> ;
M ðt Þ
2 3
m 0 0
m ¼ 40 m 0 5 (15b)
0 0 Ic
Finally, using Eqs. (6), (12), we obtain the equation of motion

M € + K
 U U ¼P
 ðt Þ (16)
 General concepts and principles of structural dynamics Chapter 1 35

where
2 3
m 0 m yc
 ¼ ðec ÞT mec ¼ 4 0
M m m xc 5 (17a)
m yc m xc Io
X
K    
¼
K ei
T T
Ri ki Ri ei (17b)
i¼1

are the mass and stiffness matrices of the structure, respectively.

The equation of motion (16) can be transformed with reference to the center
of mass by working as follows.
Using Eq. (12), we relate the displacements of the center of mass C to the
displacements of O. Hence, we have
U 
 c ¼ ec U (18)
which can be inverted to give
 ¼ ðec Þ1 U
U c (19)
We can readily show that
2 3
1 0 yc
ðec Þ1 ¼ 4 0 1 xc 5 (20)
0 0 1
Substituting Eq. (19) into Eq. (16) and premultiplying it byðec ÞT , we
obtaind
€ + K
c
mU  cU  ðt Þc
c ¼P (21)
where
 c ¼ ðec ÞT K
K  ðec Þ1 (22a)

 c ðt Þ ¼ ðec ÞT P
P  ðt Þ (22b)
Eq. (22a) represents the transformed stiffness matrix of the structure from
point O to the center of mass C .
The stiffness matrix with respect to an arbitrary point O is, in general, a full
3  3 matrix, namely
2 3
k11 k12 k13
 ¼6
K
7
4 k21 k22 k23 5 (23)
k31 k32 k33

T
d. The notation eT ¼ ðe1 Þ is employed.
36 PART I Single-degree-of-freedom systems

The off-diagonal terms cause coupling between the elastic force in one
direction and the displacement in another direction. For example, the element
k12 represents the force acting in the direction of the x axis when the plate
undergoes a unit displacement in the direction of the y axis. Similarly, the ele-
ment k31 represents the moment acting on the plate about the z axis, if the plate
undergoes a unit displacement, U ¼ 1, in the direction of the x axis. The elastic
center or center of resistance of the plate is defined as the point of the plate
where an applied force in any direction does not produce rotation. This implies
the vanishing of the elements k13 and k23 (hence also their symmetric k31 and
k32 ) in the stiffness matrix (23). This point can be established as follows.
The stiffness matrix K  is transformed from point O to the sought elastic cen-
ter E ðxE , yE Þ according to Eq. (22a), if ec is replaced by eE . Namely
2 32 32 3
    1 0 0 k11 k12 k13 1 0 yE
T 1
 E ¼ eE
K  eE
K ¼ 40 1 0 54 k21 k22 k23 54 0 1 xE 5
yE xE 1 k31 k32 k33 0 0 1
or after performing the matrix multiplications
2 E E E3
k11 k12 k13
6 7
6 E E E7

K ¼ 6 k21 k22 k23 7
E 6
7
4 5
E E E
k 31 k 32 k 33
2 3
k11 k12 k11 yE  k12 xE + k13
6 7
6 7
¼ 6 k21 k22 k21 yE  k22 xE + k23 7
4 5
k11 yE  k21 xE + k31 k12 yE  k22 xE + k32  2  2 
k 13 yE + k 23 xE + k 33
The vanishing of the elements k13 and k23 yields
E E

k11 yE  k12 xE + k13 ¼ 0

k21 yE  k22 xE + k23 ¼ 0
from which we establish the coordinates of E

  
k 11 k 13
k k
xE ¼
21 23  (24a)
k 11 k 12
k21 k22

 
k 12 k13
k k23
yE ¼
22  (24b)
k 11 k12
k21 k22
 General concepts and principles of structural dynamics Chapter 1 37

Thus, the stiffness matrix with respect to the elastic center takes the form
2 E E 32 3
k11 k12 0 k11 k12 0
6 7
 E ¼ 6 kE kE 0 76
K   7
4 21 22 54 k 21 k 22 0 5
  
0 0 k 13 yE + k 23 xE + k 33
0 0 k
E 2 2
33

The coupling is now limited between the translational displacements and

the corresponding elastic forces. They can also be decoupled if the matrix is
transformed into a new system of axes x 0 Ey 0 , resulting from xE
 y by rotation
through an angle q and demanding the off-diagonal elements to vanish. The
stiffness matrix in the new system is obtained using Eq. (7) for ki ¼ K
 E and
2 3
cos q sin q 0
R ¼ 4  sin q cos q 0 5
0 0 1
Thus, we have
K  ER
^ E ¼ RT K (25)
or after performing the matrix multiplications
2   3
k11  k22
6 k cos 2 q + k sin 2 q  k sin 2q 
sin 2q + k 12 cos2q 0 7
6 11 22 12 2 7
^ 6  7
K ¼ 6 k 11  k 22
E 7
6
4 sin 2q + k12 cos 2q k11 sin q + k22 cos q + k12 sin 2q 0
2 2 7
5
2
0 0 k13 y2 + k23 x 2 + k33
E E

The vanishing of the off-diagonal elements yields

2k12
tan 2q ¼  (26)
k 22  k11

The axes defined by angle q are referred to as the principal directions of stiff-
ness of the structure. The stiffness matrix becomes now diagonal
2 3
k^11 0 0
K^E ¼6 4 0 k^22 0 5
7

0 0 k^33

where
k^11 ¼ k11 cos 2 q + k22 sin 2 q  k12 sin 2q (27a)
k^22 ¼ k11 sin 2 q + k22 cos 2 q + k12 cos 2q (27b)
k^33 ¼ k13 y2E + k23 x2E + k33 (27c)
38 PART I Single-degree-of-freedom systems

The previous analysis allows us to draw the following conclusions:

(a) In static analysis, the concepts of the elastic center and the principal direc-
tions permit the uncoupling of the three equations of static equilibrium and
give a better insight into the deformation of the structure. These concepts,
however, do not have any advantage in the dynamic analysis because, in
general, the center of mass does not coincide with the elastic center and
therefore the equations of motion remain coupled through the second deriv-
atives of the displacements. Apparently, this fact complicates the study of
the dynamic analysis.
(b) The equations of motion can be decoupled with respect to the physical dis-
placements U , V , W only if the center of mass coincides with the elastic
center. This uncoupling should not be confused with that achieved via
modal coordinates (see Chapter 12).

1.6 Principle of virtual displacements

D’Alembert’s principle allows the application of the principle of virtual dis-
placements to formulate the equations of motion of structural systems, espe-
cially for complex assemblages comprising a number of interconnected
particles or rigid bodies of finite size. The principle of virtual displacements
or virtual works can be expressed as follows:
The necessary and sufficient condition for the dynamic equilibrium of a
system is the vanishing of the total work done by the set of all externally applied
forces (actual and inertial) when the system is subjected to a virtual displace-
ment, that is, a displacement pattern compatible with the geometrical con-
straints of the system.
Thus, the equations of motion of the dynamic system can be derived by first
identifying all forces (imposed external forces, elastic forces, damping forces,
and inertial forces) acting on the masses. Then by introducing a virtual displace-
ment corresponding to each degree of freedom, the equations of motion are
obtained by setting the virtual work produced by all forces equal to zero. A
major advantage of this approach is that the contribution of the work done
by the reactions of nonyielding bilateral supports as well as the internal inter-
action forces on the separated masses do not appear explicitly in the equations.
Moreover, the quantities we have to deal with are scalar and not vectorial, thus
they can be algebraically manipulated. An important provision for the applica-
tion of the principle of virtual displacements is that the masses of the system are
subjected to small displacements. This implies that the geometry of the structure
remains essentially unchanged after the action of the displacements.
Example 1.6.1 Equation of motion of a complex SDOF system
Formulate the equation of motion of the system shown in Fig. E1.8a by using the
principle of virtual displacements for small amplitude motion. It is assumed that
the cable can undertake compression.
 General concepts and principles of structural dynamics Chapter 1 39

(a)

(b)
FIG. E1.8 System in Example 1.6.1.

Solution
Because the cable is inextensible, the displaced configuration of the system can
be specified either by the angle of rotation of one of the bars or by the transverse
displacement of a point on it. Thus, the system has only one degree of freedom.
If the upward transverse displacement u ðt Þ of point C is taken as the basic
parameter of the motion, then all other displacements can be expressed in terms
of it. Fig. E1.8b shows the deformed system with all forces applied to it.
The elastic forces fS1 and fS2 are due to the deformation of the springs k1 and
k2 . They are directed downward as they oppose the motion. The force fD is
due to the viscous damping mechanism and is directed upward as it also
opposes the motion. The inertia moments MIA , MIO , and MIE are due to the rota-
tion of the masses about A, O , and E, respectively. All forces are expressed in
terms of the single displacement u ðt Þ
fS1 ¼ k1 ðBB 0 Þ ¼ ku=2, fS2 ¼ k2 ðCC 0 Þ ¼ 2ku
d
fD ¼ c ðDD 0 Þ ¼ cu_
dt
 ð2LÞ3 u€
m
MIA ¼ IA f€1 ¼  2 u€
¼ 1:333mL
3 2L
 ð1:5LÞ3 u€
m
MIE ¼ IE f€2 ¼  2 u€
¼ 0:750mL
3 1:5L
ð0:8LÞ2 u€
MIO ¼ IO f€3 ¼ mL
  2 u€
¼ 0:200mL
8 0:4L
40 PART I Single-degree-of-freedom systems

If point C is given a virtual displacement du, the forces ride the following
displacements
d ðCC 0 Þ ¼ du, dðBB 0 Þ ¼ du=2, dðDD0 Þ ¼ du
df1 ¼ du=2L, df2 ¼ du=1:5L, df3 ¼ du=0:4L
du ðx Þ ¼ xdf1 ¼ xdu=2L
The work done by the forces acting on the system due to the virtual displace-
ment should be set equal to zero, that is,
fS1 d ðBB 0 Þ  fS2 dðCC 0 Þ  fD dðDD 0 Þ  MIA df1
Z L
(1)
 MIE df2  MIO df3 + pðt Þdu ðx Þdx ¼ 0
0

Using the expressions for the forces and the displacements in terms of the
basic displacement derived previously, Eq. (1) yields

0:25ku  2ku  cu_  1:333mL
 2 u=2L
€  2 u=1:5L
 0:750mL €
0:200mL €
 2 u=0:4L + pðt ÞL=4du ¼ 0
or, inasmuch as du 6¼ 0, the expression within the square brackets should vanish.
This yields the equation of motion
m ∗ v€ + c ∗ v_ + k ∗ v ¼ p ∗ ðt Þ (2)
where

m ∗ ¼ 1:667mL, c ∗ ¼ c, k ∗ ¼ 2:25k, p ∗ ðt Þ ¼ 0:25
pðt ÞL
Example 1.6.2 Equation of motion of a rigid body assemblage
Formulate the equations of motion of the rigid body assemblage shown in-
Fig. E1.9a by using the principle of virtual displacements on the basis of small
amplitude motion.
Solution
Due to the spring k1 , the rigid bars can rotate independently from each other
about their hinged supports at A and F. Hence, the system has two degrees
of freedom. Its motion can be specified by the transverse downward displace-
ments u1 ðt Þ and u2 ðt Þ of points C and E, respectively. The forces applied to the
displaced system are shown in Fig. E1.9b. They are
The elastic force fS1 ¼ k1 ðCC 0 Þ ¼ k ðu2  u1 Þ
The elastic force fS2 ¼ k2 ðDD 0 Þ ¼ 4ku 2
The damping force fD ¼ c dtd ðBB 0 Þ ¼ c u_21
The inertial moment MIA ¼ IA f€1 ¼ IA 2a u€1  2
¼ 4ma
3 u €1
€
The inertial moment M ¼ IF f2 ¼ IF ¼
F u€2  2
8ma
u€2
I a 3

The system is given a virtual displacement pattern du1 and du2 corresponding
to the two degrees of freedom. The forces ride the following displacements

du1 du1
d ðBB 0 Þ ¼ , dðCC 0 Þ ¼ du1 , df1 ¼
2 2a
 General concepts and principles of structural dynamics Chapter 1 41

(a)

(b)
FIG. E1.9 System in Example 1.6.2.

du2
d ðDD 0 Þ ¼ 2du2 , d ðEE 0 Þ ¼ du2 , df2 ¼
a
According to the principle of virtual displacements, the work done by the
applied forces must be equal to zero, that is,

MIA df1  fD dðBB 0 Þ + fS1 dðCC 0 Þ  fS1 dðEE 0 Þ  fS2 dðDD 0 Þ

(1)
MIF df2 + pðt Þd ðCC 0 Þ ¼ 0

Introducing the expressions of the forces and virtual displacements into

terms of the basic displacements in Eq. (1) yields

 

2ma u_ 1
 u€1  c + k ðu2  u1 Þ + pðt Þ du1
3 4

 (2)
8ma
+  u€2  k ðu1  u2 Þ  8ku 2 du2 ¼ 0
3

Inasmuch as the quantities du1 and du2 are arbitrary, Eq. (2) is valid
only if


2ma u_ 1
u€1 + c  k ðu2  u1 Þ  pðt Þ ¼ 0 (3a)
3 4


8ma
u€2 + k ð9u2  u1 Þ ¼ 0 (3b)
3
42 PART I Single-degree-of-freedom systems

Eqs. (3a), (3b) are the equations of motion of the system. In matrix form they
are written as
2 3 2 c 3( ) "

2ma  #( ) ( )
6 3 0 7 u€1 0 u_ 1 k k u1 pðt Þ
4 44 5 ¼
 5 u€2 +
8ma u_ 2
+
k 9k u2 0
(4)
0 0 0
3

1.7 Hamilton’s principle

The development of dynamics and generally of mechanics has been accom-
plished through two different approaches. The first is based on Newton’s laws
of motion. These laws deal with the motion of a body under the action of forces
acting on it. The involved quantities are (i) the imposed forces, which may be
externally applied forces, forces of interaction between the masses, and reactions
of constraints and (ii) the momentum or the quantity of motion as defined by
Newton. Because both quantities are vector quantities, this approach of mechan-
ics is called vectorial mechanics. The analysis of complicated systems by direct
application of Newton’s laws of motion becomes increasingly difficult. The prin-
cipal reason is that the equations are vectorial in nature and the forces and accel-
erations are often difficult to determine. Moreover, the reactions of the constraints
and the interaction forces between bodies must be explicitly accounted in the
equations of motion and have to be evaluated even when there is no interest to
evaluate them. In addition, each problem seems to require its own particular insi-
ghts and there are no general procedures for obtaining the equations of motion.
The second approach is based mainly on the work of Lagrange and Hamilton
and is called analytical mechanics. In this approach, the involved quantities are
scalar functions, and therefore the fundamental equations, in contrast to vectorial
mechanics, do not depend on the choice of the coordinates. Also, it is not neces-
sary to include explicitly the forces of the constraints and the interaction forces. It
will be shown that this approach circumvents to some extent the difficulties found
in the direct application of Newton’s law of motion to complicated systems. Fur-
thermore, the equations of motion are formulated in a standard convenient form.
Analytical dynamics is based on Hamilton’s principle and Lagrange’s equations.
Hamilton’s principle is presented in this section. Lagrange’s equations are pre-
sented in the next section resulting directly from Hamilton’s principle.
One of the most important principles of dynamics is Hamilton’s principle,
named after the famous Irish mathematician and physicist Sir William Rowan
Hamilton (1805–65). Inertial and elastic forces are not explicitly involved in
this principle; instead, variations of the kinetic and potential energy are utilized.
This formulation has the advantage of dealing only with purely scalar quanti-
ties. In the procedure of virtual displacements, even though the works them-
selves are scalar quantities, vector quantities, displacements, and forces are
utilized to represent them. Hamilton’s principle is presented here for discrete
parameter systems.
 General concepts and principles of structural dynamics Chapter 1 43

FIG. 1.7.1 Particle moving in space.

Consider a particle of mass m moving in space under the action of a force

Fðt Þ as shown in Fig. 1.7.1. If r ¼ rðt Þ ¼ x ðt Þi + y ðt Þj + z ðt Þk represents the
position vector of the particle at time t, then according to Newton’s second
law of motion, the Newtonian path of the particle is governed by the differential
equation

d 2r
m F¼0 (1.7.1)
dt 2
We confine our attention to an interval of time during which the particle
moves from point 1 at t ¼ t1 to point 2 at t ¼ t2 . We consider now a varied path,
specified by rðt Þ + drðt Þ, adjacent to the actual one. We will refer to the quan-
tity drðt Þ ¼ dx ðt Þi + dy ðt Þj + dz ðt Þk as the variation of r. The only restriction is
that the two paths coincide at time t ¼ t1 and t ¼ t2 . This implies that the var-
iation dr ¼ drðt Þ vanishes at these instants, that is,

drðt1 Þ ¼ drðt2 Þ ¼ 0 (1.7.2)

The first step to derive Hamilton’s principle is to take the inner product of
the left side of Eq. (1.7.1) with the vector dr and to integrate from time t1 to time
t2 . This gives
Z t2  
d2r
m  dr  F  dr dt ¼ 0 (1.7.3)
t1 dt 2

Integrating by parts the first term in the above integral and knowing that the
operator d acts like the differential operator [6], we obtain
Z
t2 Z t2  
t2
d 2r dr dr dr
m 2  drdt ¼ m dr  m  d dt
t1 dt dt t1 t1 dt dt
44 PART I Single-degree-of-freedom systems

The term outside the integral is equal to zero because of Eq. (1.7.2). More-
over, we can write the integrand as
      "   #
dr dr 1 dr dr 1 dr 2 1 dr 2
m d ¼m d  ¼m d ¼ d m ¼ dT
dt dt 2 dt dt 2 dt 2 dt

where
 2
1 dr
T¼ m ¼ x_ ðt Þ2 + y_ ðt Þ2 + z_ ðt Þ2 (1.7.4)
2 dt
is the kinetic energy of the particle. Hence, the integral (1.7.3) takes the form
Z t2
ðdT + F  drÞdt ¼ 0 (1.7.5)
t1

The variation dr is a virtual displacement that leads from the actual path to
the varied one. Hence the term F  dr in Eq. (1.7.5) is the virtual work done
by the force Fðt Þ. Eq. (1.7.5) is a statement of Hamilton’s principle as it is
applied to a particle. This equation can be transformed into a more convenient
form if the force Fðt Þ is separated in its conservative and nonconservative
components, that is
Fðt Þ ¼ Fc ðt Þ + Fnc ðt Þ (1.7.6)
A potential function A ¼ Aðx, y, z, t Þ exists from which the conservative
force Fc ðt Þ is derived as its minus gradient
 
∂A ∂A ∂A
Fc ¼  i+ j+ k (1.7.7)
∂x ∂y ∂z
Hence
 
∂A ∂A ∂A
Fc  dr ¼  dx + dy + dz
∂x ∂y ∂z
or
Fc  dr ¼ dA (1.7.8)
Hence, Hamilton’s principle, Eq. (1.7.5), can be written as
Z t2 Z t2
d ðT  AÞdt + dWnc dt ¼ 0 (1.7.9)
t1 t1

where
dWnc ¼ Fnc  dr
represents the virtual work of the nonconservative force.
 General concepts and principles of structural dynamics Chapter 1 45

In the absence of nonconservative forces, Fnc ¼ 0, Eq. (1.7.9) becomes

Z t2
d ðT  AÞdt ¼ 0 (1.7.10)
t1

The scalar quantity

L¼T A (1.7.11)
is termed the Lagrangian or the kinetic potential. We should emphasize that
Hamilton’s principle depends upon the energies of the system and is invariant
under the coordinate transformation.
Eq. (1.7.10) states that of all possible paths of motion of the particle during
an interval of time from t1 to t2 , the actual path is that for which the integral
Z t2
Ldt ¼ 0 (1.7.12)
t1

has a stationary value. In fact, it can be shown that this value is the minimum
value of the integral.
The derivation of Hamilton’s principle for a particle can be extended to
MDOF systems as well as to continuous systems. The potential energy usually
arises from the gravity field. However, it may also arise from other sources such
as electrical and magnetic fields. The strain energy U ðt Þ should be included as
an additional potential energy. Thus, we can write
Z t2 Z t2
d ðU  T + AÞdt  dWnc dt ¼ 0 (1.7.13)
t1 t1

Hamilton’s principle is rather utilized to derive the equations of motion of

continuous systems. The equations of motion of discrete parameter systems can
result directly from Lagrange’s equations.
Example 1.7.1 Equation of motion of the SDOF system
Formulate the equation of motion of the SDOF system shown in Fig. 1.4.1 using
Hamilton’s principle.
Solution
The potential energy is due to the strain energy stored in the spring during defor-
mation. It is expressed in terms of the spring stiffness coefficient k and the dis-
placement u as
1
U ¼ ku 2 (1)
2
The kinetic energy is due to the motion of the mass m and is given as
1
T ¼ m u_ 2 (2)
2
46 PART I Single-degree-of-freedom systems

The damping force fD ¼ cu,_ as a dissipative force, is nonconservative. The

virtual work of this force is
D
dWnc _
¼ fD du ¼ cudu (3)
The negative sign results from the fact that fD is opposite to the virtual dis-
placement du.
The external force is also treated as nonconservative and it does the
virtual work
p
dWnc ¼ pðt Þdu (4)
Because no conservative external forces act on the system, it is A ¼ 0.
The variations dU and dT are obtained from Eqs. (1), (2)
_ u_
dU ¼ kudu, dT ¼ m ud (5)
Introducing Eqs. (3)–(5) into Hamilton’s principle, Eq. (1.7.13), yields
Z t2 Z t2
_ u_ Þdt 
ðkudu  m ud _ + pðt Þdu dt ¼ 0
½cudu (6)
t1 t1

The next step is to remove the variation d u_ of the velocity u_ from Eq. (6).
This is achieved using integration by parts as follows:
Z t2 Z t2  
du
_ udt
m ud _ ¼ _
m ud dt
t1 t1 dt
Z t2
d
¼ m u_ ðdu Þdt (7)
t1 dt
Z t2
_ tt21 
¼ ½m udu €
m ududt
t1

According to Hamilton’s principle it holds

du ðt1 Þ ¼ du ðt2 Þ ¼ 0
Thus, the term outside the integral vanishes and Eq. (7) becomes
Z t2 Z t2
_ udt
m ud _ ¼ €
m ududt (8)
t1 t1

Eq. (6) by virtue of Eq. (8) is written as

Z t2
½m u€ + cu_ + ku  pðt Þdudt ¼ 0 (9)
t1

In order that the integral in Eq. (9) is equal to zero for any time interval
½t1 , t2 , its integrand should vanish, that is,
½m u€ + cu_ + ku  pðt Þdu ¼ 0
 General concepts and principles of structural dynamics Chapter 1 47

Moreover, because du is arbitrary, it must be

m u€ + cu_ + ku  pðt Þ ¼ 0
or
m u€ + cu_ + ku ¼ pðt Þ (10)
which is the equation of motion.
Example 1.7.2 Equation of motion of a two-story shear frame
Formulate the equations of motion of the frame in Example 1.5.5 using Hamil-
ton’s principle.
Solution
Referring to Fig. E1.5b and c in Example 1.5.5, we have
1 1
U ¼ k1 ðu1  u2 Þ2 + k2 u22
2 2
1 1
T ¼ m1 u_ 21 + m2 u_ 22
2 2
Their variations are
dU ¼ k1 ðu1  u2 Þðdu1  du2 Þ + k2 u2 du2
(1)
¼ k1 ðu1  u2 Þdu1  k1 ðu1  u2 Þdu2 + k2 u2 du2
dT ¼ m1 u_ 1 d u_ 1 + m2 u_ 2 d u_ 2 (2)
Integrating by parts the variation dT in the interval ½t1 , t2  yields
Z t2 Z t2
dTdt ¼ ðm1 u_ 1 d u_ 1 + m2 u_ 2 d u_ 2 Þdt
t1 t1
Z t2
¼ ½m1 u_ 1 du1 + m2 u_ 2 du2 tt21  ðm1 u€1 du1 + m2 u€2 du2 Þdt
t1

and taking into account that du1 ðt1 Þ ¼ du1 ðt2 Þ ¼ du2 ðt1 Þ ¼ du2 ðt2 Þ ¼ 0, we
obtain
Z t2 Z t2
dTdt ¼  ðm1 u€1 du1 + m2 u€2 du2 Þdt (3)
t1 t1

Moreover, it is
p
dWnc ¼ p1 ðt Þdu1 + p2 ðt Þdu2 and A ¼ 0 (4)
Introducing Eqs. (1), (3), (4) into Hamilton’s principle, Eq. (1.7.13), we
obtain Z t2
½k1 ðu1  u2 Þdu1  k1 ðu1  u2 Þdu2 + k2 u2 du2 + m1 u€1 du1
t1
+ m2 u€2 du2  p1 ðt Þdu1  p2 ðt Þdu2 dt ¼ 0
48 PART I Single-degree-of-freedom systems

or
Z t2
f½m1 u€1 + k1 ðu1  u2 Þ  p1 ðt Þdu1 + ½m2 u€2  k1 u1 + ðk1 + k2 Þu2
t1
p2 ðt Þdu2 gdt ¼ 0 (5)
Because Eq. (5) is valid for any interval ½t1 , t2 , its integrand must be equal
to zero, that is,
½m1 u€1 + k1 ðu1  u2 Þ  p1 ðt Þdu1 + ½m2 u€2  k1 u1 + ðk1 + k2 Þu2  p2 ðt Þdu2 ¼ 0
(6)
Inasmuch as the quantities du1 and du2 are arbitrary, Eq. (6) is valid only if
the quantities in the square brackets are equal to zero, that is,
m1 u€1 + k1 ðu1  u2 Þ  p1 ðt Þ ¼ 0 (7a)
m2 u€2  k1 u1 + ðk1 + k2 Þu2  p2 ðt Þ ¼ 0 (7b)
which give the equations of motion
m1 u€1 + k1 u1  k1 u2 ¼ p1 ðt Þ (8a)
m2 u€2  k1 u1 + ðk1 + k2 Þu2 ¼ p2 ðt Þ (8b)

Example 1.7.3 Equation of motion of a complex MDOF system

The system shown in Fig. E1.9 consists of the three rigid bars AB,BC , CD con-
nected by hinges at points B and C , and it is supported by a roller at point D and
a hinge at point A. The relative rotations of the bars at the hinges B and C are
restrained by moment-resisting rotational springs with stiffness coefficients
k3 ¼ k4 ¼ 4kL2 and by the rotational dashpots with damping coefficients
c3 ¼ c4 ¼ 2cL2 . In the transverse direction, the motion is restrained by the
two springs at points E and Q with stiffness coefficients k1 ¼ k, k2 ¼ 2k, and
the two dashpots at points F and G with damping coefficients c1 ¼ c and
c2 ¼ 3c. A constant axial force P is applied at point D. The system is set in
motion by the transverse load pðx, t Þ ¼ ðpx=LÞf ðt Þ, linearly distributed along
the bar CD. The mass per unit length of the bars AB and CD is m  while the
bar BC is massless and supports the rigid body S at H having surface mass

density g ¼ m=L. Assuming small amplitude displacements, formulate the
equations of motion of the system using Hamilton’s principle.

FIG. E1.10 System in Example 1.7.3.

 General concepts and principles of structural dynamics Chapter 1 49

Solution
Inasmuch as the bars are assumed rigid, this system has only two degrees of
freedom. The displaced configuration of the system can be determined from
the two transverse displacements u1 ðt Þ and u2 ðt Þ of points B and C . Referring
to Fig. E1.11, we have

FIG. E1.11 Deformed configuration of the system.

9
f1 ¼ u1 =4L =
f2 ¼ ðu2  u1 Þ=3L (1)
;
f3 ¼ u2 =3L
The displacements of points of application of the forces and the changes of
angles are expressed in terms of the basic quantities u1 and u2 as
9
EE 0 ¼ u1 =4, FF 0 ¼ u1 =2, GG 0 ¼ 3u1 =4 >>
=
HH 0 ¼ u1 + ðu2  u1 Þ=3, QQ 0 ¼ u2 =2
(2)
DfB ¼ f2  f1 ¼ ð4u2  7u1 Þ=12L >
>
;
DfC ¼ f3 + f2 ¼ ð2u2  u1 Þ=3L
The potential energy U due to the deformation of the springs is
1 1 1 1
U ¼ k1 ðEE 0 Þ + k2 ðQQ 0 Þ + k3 ðDfB Þ2 + k4 ðDfC Þ2
2 2
2 2 2 2
which by virtue of Eqs. (2) becomes
1 2 1 2 1 2
U¼ ku 1 + ku 2 + k ð4u2  7u1 Þ2 + k ð2u2  u1 Þ2
32 4 72 9
¼ 0:934ku 21 + 1:361ku 22  1:667u1 u2
Its variation is
dU ¼ k ð1:868u1  1:667u2 Þdu1 + k ð1:667u1 + 2:722u2 Þdu2 (3)
The kinetic energy consists of the kinetic energies T1 and T2 of the bars ΑΒ
and CD, and of the kinetic energy T3 of the rigid body S. Thus, we have

2
1 1 1 d 1
T ¼ IA f_ 1 + ID f_ 3 + m ðHH 0 Þ + IH f_ 2
2 2 2
(4)
2 2 2 dt 2
50 PART I Single-degree-of-freedom systems

where

ð4LÞ3 ð3LÞ3 L3

m ¼ mL, 
IA ¼ m 
, ID ¼ m 
, IH ¼ m (5)
3 3 6
Introducing Eqs. (1), (2), (5) into Eq. (4) yields

2 1 1 1
T ¼ m u_ 21 + m u_ 22 + m ðu_ 2 + 2u_ 1 Þ2 + m ðu_ 2  u_ 1 Þ2
3 2 18 108
¼ 0:898m u_ 21 + 0:565m u_ 22 + 0:204u_ 1 u_ 2

and its variation

dT ¼ m ð1:796u_ 1 + 0:204u_ 2 Þd u_ 1 + m ð0:204u_ 1 + 1:130u_ 2 Þd u_ 2

Rt
Using integration by parts in the integral t12 dTdt and taking into
account that
  d
du1 ðt1 Þ ¼ du1 ðt2 Þ ¼ du2 ðt1 Þ ¼ du2 ðt2 Þ ¼ 0 and d u_ ¼ d du
dt ¼ dt ðdu Þ

we obtain
Z t2 Z t2
dTdt ¼  ½m ð1:796u€1 + 0:204u€2 Þdu1 + m ð0:204u€1 + 1:130u€2 Þdu2 dt
t1 t1
(6)
The nonconservative forces include the loading pðx, t Þ and the damping
forces. Their virtual work is expressed in terms of the basic quantities as follows:
Z 3L  x
dWnc p
¼ pðx, t Þ 1  du2 dx
0 3L
Z 3L  (7)
x x
¼ p f ðt Þ 1  du2 dx ¼ 1:5
pLf ðt Þdu2
0 L 3L
d d d
D
dWnc ¼ c1 ðFF 0 Þd ðFF 0 Þ  c2 ðGG 0 ÞdðGG 0 Þ  c3 ðDfB ÞdðDfB Þ
dt dt dt (8)
d
c4 ðDfC Þd ðDfC Þ
dt
Using Eq. (2) and taking into account that c1 ¼ c, c2 ¼ 3c, c3 ¼ c4 ¼ 2cL2 ,
we can write
D
dWnc ¼ cð2:840u_ 1  0:833u_ 2 Þdu1 + cð0:833u_ 1  1:111u_ 2 Þdu2
Hence, we have
dWnc ¼ cð2:395u_ 1  0:833u_ 2 Þdu1 + cð0:833u_ 1  1:111u_ 2 Þdu2 + 1:5
pLf ðt Þdu2
(9)
 General concepts and principles of structural dynamics Chapter 1 51

Finally, the potential A of the external conservative forces is due to the con-
stant axial force P . Hence it is
A ¼ P ðDD 0 Þ ¼ Pe
and
dA ¼ Pde (10)
The variation de is evaluated as follows.
Referring to Fig. E1.11, we have
e ¼ ðAD Þ  ðAD0 Þ ¼ 10L  4L cos f1  3Lcos f2  3L cos f3
Therefore
de ¼ Lð4sin f1 df1 + 3 sin f2 df2 + 3sin f3 df3 Þ
(11)
¼ Lð4f1 df1 + 3f2 df2 + 3f3 df3 Þ
which is introduced into Eq. (10) to yield
   
7P P P 2P
dA ¼  u1 + u2 du1 + u1  u2 du2 (12)
12L 3L 3L 3L
Introducing the expressions for dU , dT , dWnc , and dA into Hamilton’s prin-
ciple, Eq. (1.7.13), we obtain the following equations of motion
 
7P
1:796m u€1 + 0:204m u€2 + 2:395cu_ 1  0:833cu_ 2 + 1:868k  u1
12L
 
P
+ 1:667k + u2 ¼ 0
3L
 
P
0:204m u€1 + 1:130m u€2  0:833cu_ 1 + 1:111cu_ 2 + 1:667k + u1
3L
 
2P
+ 2:722k  u2 ¼ 1:5
pLf ðt Þ
3L
or in the matrix form
" #( ) " #( )
1:796 0:204 u€1 2:395 0:833 u_ 1
m +c
0:204 1:130 u€2 0:833 1:111 u_ 2
" #( ) ( ) (13)
1:868  0:583l 1:667 + 0:333l u1 0
+k ¼
1:667 + 0:333l 2:722  0:667l u2 pLf ðt Þ
1:5
where l ¼ P=kL.
The elastic forces of the system are
fS1 ¼ k ð1:868  0:583lÞu1 + k ð1:667 + 0:333lÞu2
fS2 ¼ k ð1:667 + 0:333lÞu1 + k ð2:722  0:667lÞu2
52 PART I Single-degree-of-freedom systems

They may become zero if the system of equations

 
1:868  0:583l 1:667 + 0:333l u1 0
k ¼ (14)
1:667 + 0:333l 2:722  0:667l u2 0
has a nontrivial solution. This occurs if
 
 +1:868  0:583l 1:667 + 0:333l 
 
 1:667 + 0:333l +2:722  0:667l  ¼ 0 (15)

Expanding the determinant yields

0:27797l2  1:7227l + 2:3058 ¼ 0
from which we obtain
l1 ¼ 1:9555 l2 ¼ 4:2419
The obtained values of l specify two critical values, Pcr1 ¼ 1:9555kL and
Pcr2 ¼ 4:2419kL, of the compressive axial force for which the structure exhibits
no resistance to deformation, that is, it has no stiffness and the structure buckles.
Therefore, these critical loads are the buckling loads of the structure (first and
second). The condition for buckling is the vanishing of the determinant of the
stiffness matrix, Eq. (15). The resulting equation is called the buckling equation.
It is apparent that if the axial force is tensile, the determinant cannot vanish for
real values of the parameter l. Concluding, we can state that tensile axial forces
increase the stiffness of the structure while compressive axial forces reduce it
and may lead to buckling.
Example 1.7.4 Equation of motion of the elastic cantilever beam
Derive the equation of motion of the cantilever beam shown in Fig. E1.12.
Solution
The mass of the beam element is equal to mdx. Its kinetic energy is
mdx ½∂u ðx, t Þ=∂t 2 =2, which is integrated along the beam length to yield the
kinetic energy of the beam, namely

FIG. E1.12 Cantilever in Example 1.7.4.

Z  
1 L
∂u ðx, t Þ 2
T¼ m dx (1)
2 0 ∂t
 General concepts and principles of structural dynamics Chapter 1 53

The strain energy of the beam is obtained by integrating the strain energy
density over its volume V , namely
Z
1
U¼ sx ex dV (2)
2 V
From the beam theory we have
M ðx Þ sx ∂2 u ðx, t Þ
sx ¼ y, ex ¼ , M ðx Þ ¼ EI
I E ∂x 2
Substituting the previous equations into Eq. (2) and integrating over the
cross-section of the beam yield
Z  2 2
1 L ∂ u ðx, t Þ
U¼ EI dx (3)
2 0 ∂x 2
For the simplicity of the expressions, the differentiation with respect to time
t will be designated by an over-dot while that with respect to the spatial coor-
dinated x by a prime. Moreover, the arguments will be dropped for the same
reason. Hence, expressions (1) and (3) can be rewritten as
Z
1 L
T¼ m u_ 2 dx (4)
2 0
Z
1 L
EI ðu 00 Þ dx
2
U¼ (5)
2 0
Their variations are
Z L
dT ¼ _ udx
m ud _ (6)
0
Z L
dU ¼ EI u 00 du 00 dx (7)
0

Integrating twice by parts the integral representing dU yields

Z L
EI u 0000 dudx  ½EI u 000 du 0 + ½EI u 00 du 0 0
L L
dU ¼ (8)
0

The boundary conditions of the beam are

At x ¼ 0 u ¼ u 0 ¼ 0, hence du ¼ du 0 ¼ 0
At x ¼ L M ¼ EI u 00 ¼ 0 Q ¼ EI u 000 ¼ 0
Therefore, the quantities outside the integral vanish and Eq. (8) becomes
Z L
dU ¼ EI u 0000 dudx (9)
0
54 PART I Single-degree-of-freedom systems

Because no conservative loads act on the system, it is A ¼ 0. Moreover, the

virtual work of the external load is
Z L
dWnc ¼
p
pðx, t Þdudx (10)
0

Introducing Eqs. (5), (9), (10) into Hamilton’s principle, Eq. (1.7.13), we
obtain
Z t2
Z L Z L Z L 
EI u 0000 dudx  _ udx
m ud _  pðx, t Þdudx dt ¼ 0 (11)
t1 0 0 0

Interchanging the integration in the second term and performing integration

by parts with respect to time, yield
Z t2
Z L  Z L
Z t2 
_ udx
m ud _ dt ¼  €
½m ududt  + ½mudu tt21 dx
t1 0 0 t1
Z t
Z L 2
 (12)
¼ €
½m ududx  dt
t1 0

On the basis of Eq. (12), Eq. (11) becomes

Z t 2
Z L 
½EI u 0000 + m u€  pðx, t Þdudx dt ¼ 0 (13)
t1 0

Because Eq. (13) is valid for any interval ½t1 , t2 , the integrand must vanish,
namely
Z L
½EI u 0000 + m u€  pðx, t Þdudx ¼ 0 (14)
0

Moreover, because du is arbitrary, Eq. (14) is valid only if

EI u 0000 + m u€  pðx, t Þ ¼ 0 (15)
which yields the equation of motion of the cantilever
EI u 0000 + m u€ ¼ pðx, t Þ (16)
Apparently, Eq. 16 is identical to that obtained in Section 1.1.

1.8 Lagrange’s equations

1.8.1 Derivation of Lagrange’s equations
In a system with N degrees of freedom, the displaced configuration can be deter-
mined from a set of coordinates, which take a certain value at each instant. The
system of coordinates for the analysis of a given mechanical system is not nec-
essarily unique. Many coordinate systems are possible. Furthermore, the number
of coordinates may vary, but it cannot be less than N . Anyhow, if the number of
coordinates is greater than N , then additional equations, referred to as equations
 General concepts and principles of structural dynamics Chapter 1 55

of constraint, must relate the coordinates so that the number of coordinates is

equal to the number of degrees of freedom plus the independent equations of
constraint. The requirement that the equations of motion hold together with
the equations of constraint complicates the solution. For this reason, we seek,
if possible, to choose N independent coordinates, which can specify the config-
uration of the system. For example, we consider the simple pendulum shown in
Fig. 1.8.1a. The rod is rigid and weightless. Its length is L and it can rotate freely
about the hinge at O, such that the motion is confined in a single vertical plane.
The position of the mass can be specified by the angle q between the vertical
axis y and the rod. Hence, the system has a SDOF. However, the displaced
configuration can also be determined by the coordinates ðx, y Þ, which represent
the position of the mass m within the xy plane. These coordinates, however,
are not independent because they must satisfy the constraint equation
x 2 + y 2 ¼ L2

(a) (b)
FIG. 1.8.1 Simple (a) and double (b) pendulum.

Similarly, the configuration of the double pendulum of Fig. 1.8.1b can be

specified by the two angles q1 and q2 . Hence the system has two degrees of free-
dom. On the other hand, the position of the masses m1 , m2 can be determined by
the coordinates ðx1 , y1 Þ and ðx2 , y2 Þ, which, however, are not independent
because they must satisfy the following two constraint equations
x12 + y12 ¼ L21

ðx2  x1 Þ2 + ðy2  y1 Þ2 ¼ L22

The quantities q in the simple pendulum or q1 , q2 in the double pendulum,
which would determine the configuration of the system, could be considered as
coordinates in a more general sense. Any set of quantities that serves to specify
the configuration of the system is referred to as generalized coordinates. The
geometrical significance of the generalized quantities is not always cognizable.
For systems in motion, the generalized coordinates vary with time and are trea-
ted as algebraic variables. The process of obtaining one set of generalized coor-
dinates from another is known as a coordinate transformation.
56 PART I Single-degree-of-freedom systems

We consider now a transformation from a set of N generalized coordinates

q1 ðt Þ, q2 ðt Þ,…, qN ðt Þ to a set K of ordinary (for example, Cartesian) coordinates
x1 ,x2 ,…, xK ðK  N Þ. The transformation equations are of the form
x1 ¼ x1 ðq1 , q2 , …, qN Þ
x2 ¼ x2 ðq1 , q2 , …, qN Þ
(1.8.1)
⋯ ⋯
xK ¼ xK ðq1 , q2 , …, qN Þ
For example, the transformation equations of the generalized coordinates
q1 , q2 to the ordinary coordinates x1 , x1 , y2 , y2 of the double pendulum are
x1 ¼ L1 sin q1
y1 ¼ L1 cos q1
x2 ¼ L1 sin q1 + L2 sin q2
y2 ¼ L1 cos q1  L2 cos q2
The kinetic energy of a system with K degrees of freedom may also depend
on the generalized coordinates q1 , q2 , …, qN beside the generalized velocities
q_ 1 , q_ 2 ,…, q_ N , that is,
T ¼ T ðq1 , q2 , …, qN , q_ 1 , q_ 2 , …, q_ N Þ (1.8.2)
In conservative systems, the potential energy A depends only on the posi-
tion, namely, it is
A ¼ Aðq1 , q2 , …, qN Þ (1.8.3)
The work done by the forces derivable from the potential energy A, when the
generalized coordinates qi are given a virtual displacement dqi , is expressed as
dA ¼ Q1 dq1 + Q2 dq2 + ⋯ + QN dqN
where
∂A ∂A ∂A
Q1 ¼  ¼ , Q2 ¼  ¼ , …, QN ¼  (1.8.4)
∂q1 ∂q2 ∂qN
The quantity Qi dqi represents the work done through the displacement dqi .
Inasmuch as the quantity Qi may or may not represent a force, it is referred to as
generalized force. Hence, if qi represents a translational displacement then Qi is
a force, whereas if qi represents a rotation then Qi is a moment. In some prob-
lems, the quantities dqi may represent surfaces, volumes, etc. Therefore, the
nature of the corresponding Qi is defined so that the quantity Qi dqi has the phys-
ical dimension of work.
Lagrange’s equations may be derived by direct application of Hamilton’s
principle. Thus, when the applied forces are conservative, we write
Z t2
ðdT  dAÞdt ¼ 0 (1.8.5)
t1
 General concepts and principles of structural dynamics Chapter 1 57

The variations associated with the kinetic energy and the potential energy
defined by Eqs. (1.8.2), (1.8.3), respectively, are of the form
∂T ∂T ∂T ∂T
dT ¼ dq1 + ⋯ + dqN + d q_ + ⋯ + d q_
∂q1 ∂qN ∂q_ 1 1 ∂q_ N N
∂A ∂A
dA ¼ dq1 + ⋯ + dqN
∂q1 ∂qN
Substituting these expressions into Eq. (1.8.5), integrating by parts the terms
including d q_ i and taking into account dq1 ¼ dq2 ¼ ⋯ ¼ dqN ¼ 0 at instants t1
and t2 , we obtain
Z t 2 
  
  
∂T d ∂T ∂A ∂T d ∂T ∂A
  dq1 + ⋯ +   dqN dt ¼ 0
t1 ∂q1 dt ∂q_ 1 ∂q1 ∂qN dt ∂q_ N ∂qN

Because the time interval ½t1 , t2  as well as the virtual displacements dqi are
arbitrary, this previous equation results in the following equations
 
d ∂T ∂T ∂A
 + ¼ 0 ði ¼ 1, 2, …, N Þ (1.8.6)
dt ∂q_ i ∂qi ∂qi
which, using Eq. (1.8.4), become
 
d ∂T ∂T
 ¼ Qi ði ¼ 1, 2, …, N Þ (1.8.7)
dt ∂q_ i ∂qi
Eq. (1.8.6) or (1.8.7) are the Lagrange equations of motion.
When nonconservative forces act on the system in addition to the conserva-
tive forces, we can include them in Lagrange’s equations, if the work done by
the nonconservative forces riding the virtual displacements is expressed in
terms of the generalized forces, that is,
dWnc ¼ Q1 dq1 + Q2 dq2 + ⋯ + QN dqN (1.8.8)
Introducing Eq. (1.8.8) into Hamilton’s principle, Eq. (1.7.9), the Lagrange
equations (1.8.6) become
 
d ∂T ∂T ∂A
 + ¼ Qi ði ¼ 1, 2, …, N Þ (1.8.9)
dt ∂q_ i ∂qi ∂qi
The elastic force components, which are derivable from a potential U (strain
energy), can be also involved in Eq. (1.8.9). Noting that
U ¼ U ðq1 , q2 , …, qN Þ (1.8.10)
the associated variation is
∂U ∂U
dU ¼ dq1 + ⋯ + dqN
∂q1 ∂qN
Therefore, the components ∂U =∂qi express generalized elastic forces and
Lagrange’s equations become
58 PART I Single-degree-of-freedom systems

 
d ∂T ∂T ∂V
 + ¼ Qi ði ¼ 1, 2, …, N Þ (1.8.11)
dt ∂q_ i ∂qi ∂qi
where

V ¼U +A (1.8.12)

is the total potential energy of the system.

The N generalized forces Qi can be evaluated from the set of K actual forces
Fk associated with the set of Cartesian coordinates. For this purpose, we con-
sider the work done by the set of forces Qi when the coordinates qi are given an
increment, that is, a virtual displacement dqi

X
N
dW ¼ Qi dqi (1.8.13)
i¼1

If dxk represent the ensuing virtual displacements of the coordinates xk then

the set of forces Fk do the work
X
K
dW ¼ Fk dxk (1.8.14)
k¼1

From physical consideration, the work done by the two sets of forces is the
same. The only difference is that they are expressed in two different coordinate
systems. Therefore, we can write

X
N X
K
Qi dqi ¼ Fk dxk (1.8.15)
i¼1 k¼1

or in matrix form
QT dq ¼ FT dx (1.8.16)
where

Q ¼ fQ1 Q2 ⋯ QN gT , dq ¼ f dq1 dq2 ⋯ dqN gT (1.8.17a)

F ¼ f F1 F2 ⋯ FK gT , dx ¼ f dx1 dx2 ⋯ dxK gT (1.8.17b)

The relation between dxk and dqi results from the transformation equations
(1.8.1) by considering the variation of dxk . Thus, we have
∂xk ∂xk
dxk ¼ dq1 + ⋯ + dqN (1.8.18)
∂q1 ∂qN
or in matrix form
dx ¼ Jdq (1.8.19)
 General concepts and principles of structural dynamics Chapter 1 59

where J is the Jacobian matrix of the transformation (1.8.1), that is,

2 ∂x ∂x ∂x1 3
1 1
⋯
6 ∂q1 ∂q2 ∂qN 7
6 7
6 ∂x ∂x ∂x2 7
6 2 2
⋯ 7
6 7
J ¼ 6 ∂q1 ∂q2 ∂qN 7 (1.8.20)
6 7
6 ⋮ ⋮ ⋱ ⋮ 7
6 7
4 ∂xK ∂xK ∂xK 5
⋯
∂q1 ∂q2 ∂qN
Substituting Eq. (1.8.19) into (1.8.16) yields
QT dq ¼ FT Jdq (1.8.21)
from which we obtain
QT ¼ F T J (1.8.22)
or
Q ¼ JT F (1.8.23)
Apparently, Eq. (1.8.23) represents the sought relation between Q and F.
Example 1.8.1 Equation of motion of the double pendulum
Formulate the equations of motion of the double pendulum shown in
Fig. 1.8.1b.
Solution
Because the bars are inextensional, the displaced configuration of the moving
system can be specified by the generalized coordinates q1 and q2 . Referring to
Fig. 1.8.1b, the Cartesian coordinates of the masses m1 and m2 are expressed in
terms of q1 and q2 by the geometrical relations
x1 ¼ L1 sin q1
y1 ¼ L1 cos q1
(1)
x2 ¼ L1 sin q1 + L2 sin q2
y2 ¼ L1 cos q1  L2 cos q2
The kinetic energy of the system is
1   1  
T ¼ m1 x_ 21 + y_ 21 + m2 x_ 22 + y_ 22
2 2
which by virtue of Eqs. (1) becomes
1 1
T ¼ ðm1 + m2 ÞL21 q_ 1 + m2 L1 L2 q_ 1 q_ 2 cos ðq1  q2 Þ + m2 L22 q_ 2
2 2
(2)
2 2
The potential energy is
A ¼ m1 gy 1 + m2 gy 2
60 PART I Single-degree-of-freedom systems

or using Eqs. (1)

A ¼ ðm1 + m2 ÞgL1 cos q1  m2 gL2 cos q2 (3)

Differentiating Eqs. (2), (3) yields

∂T
¼ ðm1 + m2 ÞL21 q_ 1 + m2 L1 L2 q_ 2 cos ðq1  q2 Þ
∂q_ 1
 
d ∂T
¼ ðm1 + m2 ÞL21 q€1 + m2 L1 L2 q€2 cos ðq1  q2 Þ
dt ∂q_ 1
 
 m2 L1 L2 q_ 2 sin ðq1  q2 Þ q_ 1  q_ 2
∂T
¼ m2 L1 L2 q_ 1 q_ 2 sin ðq1  q2 Þ
∂q1
∂A
¼ ðm1 + m2 ÞgL1 sin q1
∂q1
∂T
¼ m2 L1 L2 q_ 1 cos ðq1  q2 Þ + m2 L22 q_ 2
∂q_ 2
 
d ∂T  
¼ m2 L1 L2 q€1 cos ðq1  q2 Þ  m2 L1 L2 q_ 1 sin ðq1  q2 Þ q_ 1  q_ 2 + m2 L22 q€2
_
dt ∂q 2

∂T
¼ m2 L1 L2 q_ 1 q_ 2 sin ðq1  q2 Þ
∂q2
∂A
¼ m2 gL2 sin q2
∂q2
Applying Eq. (1.8.6) for i ¼ 1, 2 and q1 ¼ q1 q2 ¼ q2 , we obtain the equa-
tions of motion of the double pendulum
 
ðm1 + m2 ÞL1 q€1 + m2 L2 q€2 cos a + q_ 2 sin a + ðm1 + m2 Þg sin q1 ¼ 0 (4a)
2

L1 q€1 cos a + L2 q€2  L1 q_ 1 sin a + g sin q2 ¼ 0

2
(4b)
where
a ¼ q1  q2

Example 1.8.2 Equation of motion of the “soft” pendulum

Formulate the equations of motion of the simple pendulum shown in Fig. E1.13,
taking into account the axial deformation of the rod (soft pendulum). The unde-
formed length of the rod is L, its cross-sectional area A, and the modulus of
elasticity of the material E.
 General concepts and principles of structural dynamics Chapter 1 61

FIG. E1.13 “Soft” pendulum in Example 1.8.2.

Solution
Because the rod is no more inextensional, the system has two degrees of freedom.
Its displaced configuration can be specified either by the orthogonal coordinates x
and y of the mass or by the angle of the q and the axial deformation of the rod.
The kinetic energy of the system is
1  
T ¼ m x_ 2 + y_ 2 (1)
2
The potential energy of the external force (gravitational force) is
A ¼ mgy (2)
and the potential of the elastic force
1
U ¼ ke2 (3)
2
where k ¼ EA=L is the axial stiffness of the rod and e its elongation. The latter
is expressed in terms of x and y as
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
e ¼ x 2 + y2  L (4)
Introducing Eq. (4) in the expression for the axial stiffness, Eq. (3), yields
1 EA pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi 2
U¼ x 2 + y2  L (5)
2 L
Differentiating the energies, we obtain
  !
d ∂T ∂A ∂U EA L
¼ m x€, ¼ 0, ¼ 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi x (6)
dt ∂x_ ∂x ∂x L x 2 + y2
  !
d ∂T ∂A ∂U EA L
¼ m y€, ¼ mg, ¼ 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi y (7)
dt ∂y_ ∂y ∂y L x 2 + y2
Introducing Eqs. (6), (7) into Lagrange’s equations (1.8.11), we obtain the
equations of motion of the soft pendulum
!
EA L
m x€ + 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi x ¼ 0 (8a)
L x 2 + y2
62 PART I Single-degree-of-freedom systems

!
EA L
m x€ + 1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi y ¼ mg (8b)
L x 2 + y2

Example 1.8.3 Equation of motion of a general one-story shear building

Formulate the equations of motion of the one-story building in Example 1.5.6
using the method of Lagrange’s equations.
Solution
The system has three degrees of freedom. We choose the displacements U , V of
 about the z axis as the gen-
point O in the plane of the plate and its rotation W
eralized coordinates of the system, namely

q1 ¼ U , q2 ¼ V , q3 ¼ W
Because the point O is not the center of mass of the plate, the kinetic energy
is given by the expression (1.5.18) (K€onig’s theorem)

1   1
_ 2  m y U_ W _
_ + m xc V W
T ¼ m U_ 2 + V 2 + Io W

c (1)
2 2
The potential energy U consists of the strain energy of all columns. For the i
column it is
1 h i  i 2  i 2  i 2 i
Ui ¼ k11 u + k22i
v + k33i
w
2
or using matrix notation
 i 8 9
 k 0 0 < u i =
1  i i i  11 i  1  T
Ui ¼ u v w  0 k22 0  v i ¼ Di ki Di (2)
2  0 0 k i  wi : ; 2
33

Taking into account that (see Eqs. 1a, 12 of Example 1.5.6)

 i T    
 ¼ Ri ei U,
D i ¼ Ri D  D ¼U T e i T Ri T
Eq. (2) is written as
1  T i 
Ui ¼ U kU
2
where
   
i ¼ ei T Ri T ki Ri ei
k
Thus, we have
XK
1
U¼ 
Ui ¼ UT KU
i¼1
2
(3)
1  + 2k23 V W
 2 + 2k12 U V + 2k13 U W 

¼ k11 U + k22 V + k33 W
2 2
2
 General concepts and principles of structural dynamics Chapter 1 63

ij are the elements of the matrix

where k
X
K X
K    
¼
K i ¼
k ei
T T
Ri k i Ri e i (4)
i¼1 i¼1

Because there are no external conservative forces, it is

A¼0 (5)
The generalized forces result from Eq. (1.8.20). In this case it is
 T
F ¼ P x ðt Þ P y ðt Þ 0
The transformation relations between the displacements u, v, w of a point
 y of the plate and the displacements of point O are
x,
 v ¼ V + xW,
u ¼ U  yW,  w¼W


which yields the Jacobian matrix, Eq. (1.8.20)

2 ∂u ∂u ∂u 3
6 ∂U ∂V ∂W
7 2 3
6 7 1 0 0
6 ∂v ∂v ∂v 7
J¼6
6 ∂U
7 ¼ 4 0 1 05
6 7
∂V ∂W 7
4 5 y x 1

∂w ∂w ∂w
∂U ∂V ∂W

Thus, for point A we have
 T
Q ¼ JT F ¼ P x ðt Þ P y ðt Þ 
y A P x ðt Þ + xA P y ðt Þ (6)
Differentiation of Eq. (1) yields
 
d ∂T €
¼ m U€
  m yc W (7a)
dt ∂U_
∂T
¼0 (7b)
∂U
!
d ∂T €
¼ m V€
 + mx c W (7c)
dt ∂ V

∂T
¼0 (7d)
∂V
 
d ∂T €  m y U€
 + m xc V€
¼ Io W (7e)
_
dt ∂W c

∂T
∂W ¼0 (7f)

Moreover, Eq. (3) yields

64 PART I Single-degree-of-freedom systems

∂U 
¼ k11 U + k12 V + k13 W (8a)
∂U
∂U 
¼ k21 U + k22 V + k23 W (8b)
∂V
∂U      
∂W ¼ k 31 U + k 32 V + k 33 W (8c)

Introducing Eqs. (6)–(8) into the Lagrange

 equations (1.8.11) with N ¼ 3,
 ¼ 0, yields the equation of motion
and taking into account that A U , V , W
of the structure

M € + K
 U U ¼P
 ðt Þ (9)
where
2 3 2 3
m 0 m yc k 11 k12 k13
6 7 6 7
 ¼6 0
M m mx c 7  ¼ 6 k21
K k22 k23 7
4 5, 4 5,
my c mx c Io k31 k32 k33
8  9
> P ðt Þ >
< x =

P ðt Þ ¼ 
P y ðt Þ
>
: >
;
y A P x ðt Þ + xA P y ðt Þ


As was anticipated, Eq. (9) is identical to Eq. (16) of Example 1.5.6.

1.8.2 Lagrange multipliers

Lagrange’s equations result as a direct application of Hamilton’s principle pro-
vided that the energies (kinetic and potential) as well as the virtual work of the
nonconservative forces can be expressed in terms of the generalized coordinates
and velocities, as is indicated in Eqs. (1.8.2), (1.8.3), (1.8.8), (1.8.10).
Lagrange’s equations apply to linear as well as to nonlinear systems.
In certain cases, it is impossible or it is not convenient to choose N indepen-
dent coordinates. Then, we choose K > N coordinates and we introduce
n ¼ K  N constraint equations, which in general have the form
g1 ðq1 , q2 , …, qK Þ ¼ 0
g2 ðq1 , q2 , …, qK Þ ¼ 0
(1.8.24)
… …
gn ðq1 , q2 , …, qK Þ ¼ 0
We distinguish two approaches to derive the equations of motion:
 General concepts and principles of structural dynamics Chapter 1 65

(a) Eq. (1.8.24) can be solved in terms of n ¼ K  N coordinates. Then the

redundant coordinates can be eliminated from Eqs. (1.8.2), (1.8.3),
(1.8.8), (1.8.10), and the equations of motion are formulated using
Lagrange’s equations (1.8.11). If the constraint equations are linear, the
technique presented in Section 1.8.1 can also be employed to reduce the
number of equations to N .
(b) Eq. (1.8.24) cannot be solved in terms of n ¼ K  N coordinates. In this
case, the equations of motion can be derived by using the method of
Lagrange multipliers.
The second approach preserves the symmetry of the problem because there are
no preferred coordinates while others are eliminated. Though the method of
Lagrange multipliers deals with more coordinates than the degrees of freedom
of the system, quite often this procedure results in simpler equations.
To illustrate the method of Lagrange multipliers, we consider the variations
of the constraint functions given by Eq. (1.8.24)
dg1 dg1
dg1 ¼ dq1 + ⋯ + dqK ¼ 0
dq1 dqk
dg2 dg2
dg2 ¼ dq1 + ⋯ + dqK ¼ 0 (1.8.25)
dq1 dqk
… …
dgn dgn
dgn ¼ dq1 + ⋯ + dqK ¼ 0
dq1 dqk
which we write as
X
K
aji dqi ¼ 0 ðj ¼ 1, 2, …, n Þ (1.8.26)
i¼1

where
dgj
aji ¼ (1.8.27)
dqi
If we assume that the constraints are frictionless, then no work is done by the
constraint forces Ri when they ride any virtual displacement dqi , that is,
X
K
Ri dqi ¼ 0 (1.8.28)
i¼1

Now we multiply Eq. (1.8.26) by a factor known as the Lagrange multiplier.

Thus, we obtain
XK
lj aji dqi ¼ 0 ðj ¼ 1, 2, …, n Þ (1.8.29)
i¼1
66 PART I Single-degree-of-freedom systems

where we note that a separate equation is written for each of the constraints.
Next, we subtract the sum of equations of the form (1.8.29) from
Eq. (1.8.28) and, interchanging the order of summation, we obtain
!
XK X n
Ri  lj aji dqi ¼ 0 (1.8.30)
i¼1 j¼1

from which, noting that dqi are arbitrary, we conclude that

X n Xn
dgj
Ri ¼ lj aji ¼ lj (1.8.31)
j¼1 j¼1
dqi

The constraint forces Ri constitute additional generalized forces, which

must be included in Lagrange’s equations. Thus Eq. (1.8.11) become
  X n
d ∂T ∂T ∂V dgj
 + ¼ Qi + lj ði ¼ 1, 2, …, K Þ (1.8.32)
dt ∂q_ i ∂qi ∂qi j¼1
dqi

What we have accomplished by this procedure is the inclusion of the con-

straint reactions in the equations of motion as additional generalized forces.
Therefore, the number of unknowns becomes K + n, namely the K generalized
coordinates Qi ðt Þ and the n functions lj ðt Þ. The available equations are also
K + n, that is Eq. (1.8.24) plus Eq. (1.8.32).
Eq. (1.8.32) can be derived from Hamilton’s principle if the potential energy
of the external forces is replaced by
Xn
A ¼A lj gj (1.8.33)
j¼1

The function A is referred to as the reduced potential energy. For a more

advanced formulation, including nonholonomic constraints and a dynamic
treatment of the Lagrange multipliers, see Ref. [7].
Example 1.8.4 Equation of motion of the simple pendulum using Lagrange
multipliers
Formulate the equations of motion of the simple pendulum shown in Fig. E1.14
in terms of the Cartesian coordinates x ðt Þ, y ðt Þ, assuming that the rod is weight-
less and rigid.

FIG. E1.14 Pendulum in Example 1.8.4.

 General concepts and principles of structural dynamics Chapter 1 67

Solution
The kinetic and the potential energies of the system are
1  
T ¼ m x_ 2 + y_ 2
2
A ¼ mgy
U ¼0
Because the rod is rigid, the coordinates must satisfy the constraint equation
g1 ¼ x 2 + y 2  L2 ¼ 0
Differentiating the quantities T and A we obtain
 
d ∂T ∂T ∂A ∂g1
¼ m x€, ¼ 0, ¼ 0, ¼ 2x, Q1 ¼ px
dt ∂x_ ∂qi ∂x ∂x
 
d ∂T ∂T ∂A ∂g1
¼ m y€, ¼ 0, ¼ mg, ¼ 2y, Q2 ¼ py
dt ∂y_ ∂y ∂y ∂y
Applying Eq. (1.8.32) for q1 ¼ x and q2 ¼ y we obtain the equations of
motion
m x€ ¼ px + 2xl (1a)
m y€ + mg ¼ py + 2yl (1b)
x 2 + y2  l 2 ¼ 0 (1c)
Eqs. (1) must be solved for the three unknowns x, y, and l. It should be
noted that two of these equations are differential and one algebraic and therefore
special care is required for their solution. A convenient method is to differen-
tiate the constraint equation twice with respect to time and then to solve the
€ y€ and the parameter
resulting linear system of equations for the accelerations x,
l. For the problem at hand, we obtain
2T
x x€ + y y€ ¼  (2)
m
Eqs. (1a), (1b), (2) are combined and written in matrix form
2 38 9 8 9
m 0 2x < x€ = < px =
4 0 m 2y 5 y€ ¼ py  mg (3)
: ; : ;
x y 0 l 2T =m
which are solved to yield
  px py  mg
L2 x€ + x x_ 2 + y_ 2 ¼ y 2  xy (4a)
m m
 2  px 2 py  mg
L y€ + y x_ + y_ ¼ xy + x
2 2
(4b)
m m
68 PART I Single-degree-of-freedom systems

 
xpx y py  mg
L l¼
2
 T (4c)
2 2
Eqs. (4a), (4b) are solved using techniques for nonlinear differential equa-
tions. Analytical solutions are in general out of the question. However, a numer-
ical solution is always feasible using the methods presented in Chapter 5. Once
the coordinates x ðt Þ, y ðt Þ and the Lagrange multiplier l have been established,
they are utilized in Eq. (1.8.31) to evaluate the constraint forces, which are the
components of the axial force of the rod. Thus, we have
∂g ∂g qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Rx ¼ l ¼ 2lx, Rx ¼ l ¼ 2ly, S ¼ R2x + R2y ¼ 2lL (5)
∂x ∂y

1.8.3 Small displacements

So far, we have not imposed any restrictions on the magnitude of the displace-
ments in our discussion on Lagrange’s equations. Therefore, the derived equa-
tions of motion hold equally for small and large displacements and they are in
general nonlinear differential equations. For small displacements, however,
about the position of static equilibrium, the equations of motions are highly sim-
plified as they become linear. This is very important in structural dynamics,
where we usually deal with small displacements and the resulting linear differ-
ential equations can be readily solved and predict the response of the structure.

1.8.3.1 Potential energy and stiffness matrix

We shall consider a system of particles that is in static equilibrium under the
action of a set of conservative forces. If its configuration is specified by ordinary
coordinates x1 ,x2 , …, x3 N , then the condition for static equilibrium is that the
virtual work done by the applied forces Fi is zero, that is,
X
3N
dW ¼ Fi dxi ¼ 0 (1.8.34)
i¼1

for all virtual displacements dxi consistent with the constraints, which are
assumed workless and bilateral.
Inasmuch as the forces are conservative, they are derivable from a potential
function V ¼ V ðx1 , x2 , …, x3N Þ, V ¼ U + A, according to the relation
∂V
Fi ¼  (1.8.35)
∂xi
Using Eq. (1.8.35), Eq. (1.8.34) is written as
X
3N
∂V
dW ¼  dxi ¼ 0 (1.8.36)
i¼1
∂xi
 General concepts and principles of structural dynamics Chapter 1 69

If the number of degrees of freedom is n < 3N , dx’s are not independent. It

is possible to find n independent generalized coordinates by considering trans-
formation equations, that is,
x1 ¼ x1 ðq1 , q2 , …, qn Þ
x2 ¼ x2 ðq1 , q2 , …, qn Þ
(1.8.37)
… …
x3N ¼ x3N ðq1 , q2 , …, qn Þ
Then we have
X
n
∂xi
dxi ¼ dqj (1.8.38)
j¼1
∂qj

Substituting the previous expression for dxi into Eq. (1.8.36), we obtain
3N X
X n
∂V ∂xi
dW ¼  dqj ¼ 0 (1.8.39)
i¼1 j¼1
∂xi ∂qj

Noting that

∂V X 3N
∂V ∂xi
¼ (1.8.40)
∂qj i¼1
∂xi ∂qj

and interchanging the order of summation, Eq. (1.8.39) becomes

X
n
∂V
dW ¼  dqj ¼ 0 (1.8.41)
j¼1
∂qj

Because dqj are assumed to be independent, the virtual work is zero only if
the coefficients of dqj are zero at the equilibrium condition, that is, if
 
∂V
¼ 0, j ¼ 1, 2, …, n (1.8.42)
∂qj 0

The subscript zero denotes that the derivatives refer to the equilibrium
position.
Let us expand now the potential energy function V ðq1 , q2 , …, qn Þ in a Tay-
lor series about the position of equilibrium
Xn   n  2 
∂V 1X n X
∂ V
V ¼ V0 + dqi + dqi dqj + ⋯ (1.8.43)
i¼1
∂qi 0 2 i¼1 j¼1 ∂qi ∂qj 0

We can arbitrarily set the potential energy at the reference position equal to
zero, that is,
V0 ¼ 0 (1.8.44)
70 PART I Single-degree-of-freedom systems

If we now assume that the displacements about the equilibrium position are
small, we can neglect terms of order higher than the second in Eq. (1.8.43).
Thus, using Eqs. (1.8.41), (1.8.44) the expression for the potential energy is sim-
plified as
n  2 
1X n X
∂ V
V¼ qi qj (1.8.45)
2 i¼1 j¼1 ∂qi ∂qj 0

or setting
 
∂2 V
kij ¼ kji ¼ (1.8.46)
∂qi ∂qj 0

we can write Eq. (1.8.45) in the form

1X n X n
V¼ kij qi qj (1.8.47)
2 i¼1 j¼1

The quantities kij defined by Eq. (1.8.46) are the stiffness coefficients of the
system. Thus we see that the potential energy is expressed by a homogeneous
quadratic function of the generalized coordinates qi if small motions about the
position of equilibrium are examined.
Eq. (1.8.47) is written in matrix form
1
V ¼ qT kq (1.8.48)
2
where
8 9 2 3
>
> q1 >
> k11 k12 ⋯ k1n
< = 6 k21 k22
q2 ⋯ k2n 7
q¼ , k ¼6
4 ⋮ ⋮
7 (1.8.49)
>
> ⋮> > ⋱ ⋮ 5
: ;
qn kn1 kn2 ⋯ knn
The matrix k is called the stiffness matrix of the system.
The expression for the potential energy given in Eq. (1.8.47) is an example
of a quadratic form. For a system whose reference equilibrium configuration is
stable, the potential energy V is positive for all possible values of qi , except
q1 ¼ q2 ¼ … ¼ qn ¼ 0. In this case, the function V is referred to as positive def-
inite. This condition, however, puts restrictions on the allowable values of kij . It
is clear that all diagonal elements must be positive. The necessary and sufficient
condition that V be positive definite is that
2 3
  k11 k12 ⋯ k1n
k k  6 k21 k22 ⋯ k2n 7
k11 > 0,  11 12  > 0, …, 6
4 ⋮ ⋮ ⋱ ⋮ 5>0
7 (1.8.50)
k21 k22
kn1 kn2 ⋯ knn
 General concepts and principles of structural dynamics Chapter 1 71

1.8.3.2 Kinetic energy and mass matrix

The kinetic energy is
 2
1X3N
∂xk
T¼ mk (1.8.51)
2 k¼1 ∂t

Differentiating Eq. (1.8.37) with respect to time yields

∂xk X n
∂xk
¼ q_ (1.8.52)
∂t j¼1
∂qj j

Consequently, it is about the position of the static equilibrium

 2 Xn X n    
∂xk ∂xk ∂xk
¼ q_ i q_ j (1.8.53)
∂t 0 i¼1 j¼1
∂qi 0 ∂qj 0

Introducing this expression into Eq. (1.8.51) we can write the kinetic energy
in the form
1X n X n
T¼ mij q_ i q_ j (1.8.54)
2 i¼1 j¼1

where it was set

X
3N    
∂xk ∂xk
mij ¼ mji ¼ mk (1.8.55)
k¼1
∂qi 0 ∂qj 0

The quantities mij defined by Eq. (1.8.55) are the inertia coefficients of the
system.
Eq. (1.8.54) is written in matrix form
1
T ¼ q_ T mq_ (1.8.56)
2
where
8 9 2 3
>
> q_ 1 >
> m11 m12 ⋯ m1n
< = 6 m21 m22
q_ 2 ⋯ m2n 7
q_ ¼ , m ¼6
4 ⋮
7 (1.8.57)
>
> ⋮> > ⋮ ⋱ ⋮ 5
: ;
q_ n mn1 mn2 ⋯ mnn

The matrix m is called the mass matrix of the system. The kinetic energy is a
positive definite quadratic function because it is the sum of positive quantities,
that is, the kinetic energies of the masses of the individual particles.
72 PART I Single-degree-of-freedom systems

The equations of motion are obtained by applying Eq. (1.8.11). Differenti-

ating the kinetic energy, Eq. (1.8.54), and the potential energy, Eq. (1.8.47),
yields
∂T
¼0
∂qi
  X n
d ∂T
¼ mij q€j
dt ∂q_ i j¼1

∂V X n
¼ kij qj
∂qi j¼1

Substituting into Lagrange’s equations, we obtain the following equations of

motion
X
n X
n
mij q€j + kij qj ¼ Qi ði ¼ 1, 2, …, n Þ (1.8.58)
j¼1 j¼1

or in matrix form
m€
q + kq ¼ pðtÞ (1.8.59)
where p(t)¼Q
The matrices m and k are symmetric. It is an advantage of the Lagrange for-
mulation of the equations of motion that it preserves the symmetry of the coef-
ficient matrices for those cases where T and V are represented by quadratic
functions of the velocities and displacements, respectively.

1.8.4 Raleigh’s dissipation function

The dissipative forces arising in a mechanical system are nonconservative
forces. Therefore, they are not derivable from a potential function. They are
involved in the Lagrange equations with their virtual work. When the dissipa-
tive forces are due to such sources as air resistance or internal friction, they are
usually assumed to depend linearly on the velocities along with the physical
coordinates and opposed to the motion, that is
X
n
fDj ¼  cij q_ j (1.8.60)
i¼1

where cij ¼ cji are the damping coefficients of the linear viscous damping.
Apparently, we can construct a quadratic function

1X n X n
R¼ cij q_ i q_ j (1.8.61)
2 i¼1 j¼1
 General concepts and principles of structural dynamics Chapter 1 73

which yields
∂R Xn
fDj ¼  ¼ cij q_ j (1.8.62)
∂q_ j i¼1

If these forces are introduced as generalized forces in Lagrange’s equations

(1.8.11) and shifted to the left side, we obtain
 
d ∂T ∂T ∂V ∂R
 + + ¼ Qi ði ¼ 1, 2, …, n Þ (1.8.63)
dt ∂q_ i ∂qi ∂qi ∂q_ i
The function R defined by Eq. (1.8.61) is known as Raleigh’s dissipation
function.
The equations of motion, Eq. (1.8.58), become now
X
n X
n X
n
mij q€j + cij q_ j + kij qj ¼ pi ði ¼ 1, 2, …, n Þ (1.8.64)
j¼1 j¼1 j¼1

where pi denotes the nonconservative external forces.

Eq. (1.8.64) is written in matrix form
q + cq_ + kq ¼ p
m€ (1.8.65)
The matrix c is called the damping matrix of the system.

1.9 Influence of the gravity loads

We consider the system of Fig. 1.9.1a, which can move in the vertical direction.
Apparently, the weight of the body must be added to the external forces. If the
vertical displacement from the undeformed position is designated by u ¼ u ðt Þ,
the equation of motion will read
m u€ + cu_ + ku ¼ pðt Þ + W (1.9.1)

(a) (b)
FIG. 1.9.1 Influence of the gravity load.
74 PART I Single-degree-of-freedom systems

The elongation ust of the spring under its own weight will be
ust ¼ W =k ¼ constant (1.9.2)
Further we set
u ¼ ust + uðt Þ (1.9.3)
where uðt Þ represents the vertical displacement measured from the position of
the static equilibrium.
Differentiating Eq. (1.9.3) yields
_ u€ ¼ u€
u_ ¼ u,  (1.9.4)
Using Eqs. (1.9.3), (1.9.4), the equation of motion (1.9.1) becomes
m u€
 + cu_ + ku st + k u ¼ pðt Þ + W
or using Eq. (1.9.2) we obtain
m u€
 + cu_ + k u ¼ pðt Þ (1.9.5)
The conclusion drawn from Eq. (1.9.5) states that, in the study of the
dynamic response of a system undergoing small displacements, the loads due
to gravity can be neglected. Of course, the total displacements will result as
the sum of the static plus dynamic displacements. That is, the superposition
principle is valid.

1.10 Problems
Problem P1.1 The plane square rigid body B of side length L and surface
mass density g is supported by two identical inclined columns having
cross-sectional moment of inertia I , modulus of elasticity E, and negligible
mass. Derive the equation of motion neglecting the axial deformation of
the columns (Fig. P1.1).

FIG. P1.1 Structure in problem P1.1

 General concepts and principles of structural dynamics Chapter 1 75

Problem P1.2 Consider the structure of Fig. P1.2a. The square plate of constant
thickness h ¼ a=10 and mass density g is supported at its center by a flexible
column having a circular cross-section with diameter d ¼ a=10, height a, and
material constants E, n. The plate is loaded by a force P acting in the plane
of the plate at point (Aða=8,  a=6Þ and in the direction ∡x, P ¼ b ¼ 30° as
shown in Fig. P1.2b. Derive the equations of motion of the plate when the mass
of the column is neglected.

(a) (b)
FIG. P1.2 Structure in problem P1.2

FIG. P1.3 Structural model in problem P1.3

Problem P1.3 The semicircular rigid plate of constant thickness and total mass
m is supported as shown in Fig. P1.3. Taking into account that the support at
point O is a hinge, formulate the equation of motion of the plate using (i)
the method of equilibrium of forces, (ii) the principle of virtual displacements,
and (iii) the method of the Lagrange equations.
Problem P.1.4 Consider the system shown in Fig. P1.4. The bars AD and EG
are rigid with masses m and m=3 , respectively. The mass at end D is concen-
trated. The elastic supports at points at B, E , and D are simulated by springs
with a stiffness k while the end G is supported by a viscous damper with a
damping coefficient c. The rod CE is weightless and rigid. Derive the equation
of motion using the principle of virtual displacements.
76 PART I Single-degree-of-freedom systems

FIG. P1.4 System in problem P1.4

Problem P1.5 Consider the system shown in Fig. P1.5. The mass m is sup-
ported at the top of the flexible and massless column 2  3, which is supported
on the ground by means of the rigid body 1  2 of mass 2a m.  The support 1 is
elastically restrained by the rotational spring CR . Formulate the equation of
motion of the structure using CR ¼ EI =2a, m  ¼ m=a.

FIG. P1.5 Structure in problem P1.5

Problem P1.6 Consider the two-story frame of Fig. P1.6. The columns 1  2,
10  20 , and the beam 3  30 are rigid while the columns 2  3, 20  30 , and the
beam 2  20 are massless and flexible with cross-sectional moment of inertia
I and modulus of elasticity E. The supports at 1 and 10 are elastically restrained
by rotational springs with a stiffness CR . Formulate the equation of motion of
the structure taking CR ¼ EI =2a and m  ¼ m=a.

FIG. P1.6 Two-story frame in problem P1.6

 General concepts and principles of structural dynamics Chapter 1 77

Problem P1.7 The system of Fig. P1.7 consists of the beam AB and the rigid
body S interconnected at B. The beam AB has a negligible mass, modulus of
elasticity E , and cross-sectional moment of inertia I . The beam is fixed at A
while the rigid body is elastically restrained at C by a rotational spring with
a stiffness CR ¼ EI =10L. The total mass m is uniformly distributed. The system
is loaded by the concentrated moment M ðt Þ at point B. Derive the equation of
motion of the system using Lagrange’s equations.

FIG. P1.7 System in problem P1.7

10 10

FIG. P1.8 Frame in problem P1.8

Problem P1.8 The frame of Fig. P1.8 consists of the rigid beam BD of total
mass m and the two massless and flexible columns AB and CD with a
cross-sectional moment of inertia I and modules of elasticity E. The two mass-
less cables FB and GD have cross-sectional area A and cannot undertake com-
pressive force. Derive the equation of motion of the structure taking
I =A ¼ a2 =25 and m  ¼ m=5a.
Problem P1.9 Consider the two-story frame of Fig. P1.9. The columns of the
frame are rigid and have a surface mass density g. Their elastic support on the
ground is simulated by the rotational springs with a stiffness CR ¼ EI =10a.
The horizontal beams are massless and flexible with a cross-sectional moment
of inertia I and modulus of elasticity E. Derive the equation of motion when the
structure is subjected to the horizontal loads pðt Þ at the beam levels.
78 PART I Single-degree-of-freedom systems

FIG. P1.9 Two-story frame in problem P1.9

Problem P1.10 The hinge O of the soft pendulum of Fig. P1.10 is elastically
restrained by the rotational spring with a stiffness CR ¼ EAL=10. The length of
the rod is L, its cross-sectional area A, and the modulus of elasticity E. Formu-
late the equation of motion of the pendulum.

FIG. P1.10 Soft pendulum in problem P1.10

Problem P1.11 The rigid bar AB of circular cross-section and mass density
m ¼ m=a is hinged at point A (Fig. P1.11). The cables DB,FB have cross-
sectional area A and modulus of elasticity E. They are assumed massless
and are prestressed so that they can undertake compressive forces. Formulate

FIG. P1.11 Structure in problem P1.11

 General concepts and principles of structural dynamics Chapter 1 79

the equation of motion of the structure taking into account that the load P is
removed suddenly at instant t ¼ 0. Evaluate the minimum prestressing force
of the cables DB, FB so that they can undertake compressive loads.
Problem P1.12 Consider the structure of Fig. P1.12. The column AC has a
circular cross-section and a mass per unit length m  ¼ m=a; it is supported by
a spherical hinge on the ground and is kept in place by three elastic cables of
cross-sectional area A and modulus of elasticity E. The cables are assumed mass-
less and are prestressed so that they can undertake compressive force. Derive the
equation of motion of the structure when it is loaded by the horizontal force P ðt Þ
acting at the top of the column in the direction ∡x,P ¼ b (Fig. P1.12b).

(a) (b)
FIG. P1.12 Structure in problem P1.12

Problem P1.13 The silo of Fig. P1.13 is supported on its fundament by four
identical columns of a square cross-section. The silo is full of material of density
g. The ground yields elastically with a subgrade constant Ks . The silo and the
fundament are rigid. Derive the equation of motion of the structure when it is
loaded by the horizontal force P ðt Þ acting at the top of the silo in the direction

(a) (b)
FIG. P1.13 Silo on elastic subgrade. (a) Vertical section. (b) Plan form.
80 PART I Single-degree-of-freedom systems

∡x,P ¼ b (Fig. P1.13b) using the following data: Side of the columns a=4;
thickness of the bottom and walls of the silo a=8; density of the material of
the silo 1:5g; and soil constant Ks ¼ EI =1500a 3 .
Problem P1.14 Consider the one-story building of Fig. P1.14. The rigid plate is
an equilateral triangular with a side a and it is supported by three columns of
height a, rectangular cross-section a=10  a=20, and modulus of elasticity E.
The columns are fixed at both ends. Derive the equation of motion of the plate
when a horizontal force P ðt Þ acts at point Að0, a=5Þ in the direction ∡x, P ¼ b.
The dead weight of the plate is included in p (kN=m2 ).

Rigid plate

FIG. P1.14 One-story building in problem P1.14

Problem P1.15 The two one-story buildings of Fig. P1.15 are connected with a
beam as shown in the figure. All columns have a square cross-section with a
moment of inertia Ic ¼ 2I . The connecting beam has a square cross-section with
moment of inertia Ib ¼ I . The structure is loaded by the horizontal force F(t) at the
level of the plates as shown in Fig. P1.15b. Formulate the equations of motion
using Lagrange’s equations. Assume: Torsion constant It ¼ 2:25d 4 =16, d ¼side
length of the square cross-section of the beam.

Rigid plate

(a)

beam

(b)
FIG. P1.15 Structure in problem P1.15. (a) vertical section, (b) plan form.

Problem P1.16 The system of Fig. P1.16 consists of the block of mass m1 ,
which can slide without friction on the inclined surface, and the pendulum of
 General concepts and principles of structural dynamics Chapter 1 81

length L and mass m2 , which is pivoted at the center of mass of the block. The
rod of the pendulum has a cross-sectional area A and modulus of elasticity E.
Assuming plane motion, derive the equation of motion of the system taking
EA=L ¼ 5k and m1 ¼ 5m2 .

.
FIG. P1.16 System in problem P1.16

Problem P1.17 Derive the equation of motion of the system shown in

Fig. P1.17. The mass of the case is m2 . Use m2 ¼ 5m1 , k2 ¼ 3k1 , c2 ¼ 1:5c1 .

FIG. P1.17 System in problem P1.17

Problem P1.18 Consider the crane of Fig. P1.18. The horizontal beam is
assumed rigid. The column is flexible with a cross-sectional moment of inertia
I and the cable axially deformable with cross-sectional area A. The mass of the
cable and column is negligible. Derive the equation of motion of the system
when it is loaded by the horizontal force pðt Þ in the plane of the structure using
I =A ¼ a 2 =100 and a common modulus of elasticity E.
82 PART I Single-degree-of-freedom systems

FIG. P1.18 Crane in Problem P1.18.

References and further reading

[1] S. Timoshenko, D.H. Young, W. Weaver Jr., Vibration Problems in Engineering, fifth ed., John
Wiley, New York, 1990.
[2] R.L. Coelho, On the deduction of Newton’s second law, Acta Mech. 229 (5) (2018) 2287–2290.
[3] Euler, L. (1912). Decouverte d’un Nouveau Principe de Mecanique, Memoires de l’academie
des sciences de Berlin 6, 185–217, in: Opera Omnia, Series II, vol. 5, 81–108, Leipzig.
[4] J.T. Katsikadelis, Derivation of Newton’s Law of Motion from Kepler’s Laws of Planetary
Motion, Arch. Appl. Mech. 88 (2018) 27–38, https://doi.org/10.1007/s00419-017-1245-x.
[5] D.T. Greenwood, Principles of Dynamics, Prentice-Hall, Englewood Cliffs, NJ, 1965.
[6] J.T. Katsikadelis, The Boundary Element Method for Engineers and Scientists, Academic Press,
Elsevier, Oxford, UK, 2016.
[7] S. Natsiavas, E. Paraskevopoulos, A set of ordinary differential equations of motion for con-
strained mechanical systems, Nonlinear Dyn. 79 (2015) 1911–1938.
[8] E.N. Strømmen, Structural Dynamics, Springer Series in Solid and Structural Mechanics,
Springer, New York, 2014.
Chapter 2

Single-degree-of-freedom
systems: Free vibrations
Chapter outline
2.1 Introduction 83 2.3.3 Overdamped system 96
2.2 Free undamped vibrations 83 2.4 Conservation of energy in an
2.3 Free damped vibrations 91 undamped system 97
2.3.1 Critically damped system 91 2.5 Problems 99
2.3.2 Underdamped system 92 References and further reading 103

2.1 Introduction
In this chapter, the free vibrations of a single-degree-of-freedom system
(SDOF) are studied, that is, its response when it is not subjected to any external
force, pðt Þ ¼ 0, but it is excited by an initial displacement and/or initial velocity.
The dynamic model of the system is shown in Fig. 1.4.1 and the equation of
motion (1.4.8) takes the form
m u€ + cu_ + ku ¼ 0 (2.1.1)
Eq. (2.1.1) is an ordinary linear homogeneous differential equation of the
second order with constant coefficients and its solution can be obtained using
known mathematical methods. Inasmuch as we are interested in the physical
response of the system described by this equation, it is advisable to analyze
the free vibration response in two stages, first for c ¼ 0 and then c 6¼ 0. In the
first case, we speak of free undamped vibrations while in the second case we
speak of free damped vibrations. Illustrative examples analyzing the free vibra-
tions of SDOF systems are presented. The pertinent bibliography with recom-
mended references for further study is also included.

2.2 Free undamped vibrations

Although systems without damping do not exist in the real world, the undamped
response is studied because it provides an insight into the free vibration response
of damped systems. In the absence of damping, c ¼ 0, Eq. (2.1.1) becomes
m u€ + ku ¼ 0 (2.2.1)
Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00002-9
© 2020 Elsevier Inc. All rights reserved. 83
84 PART I Single-degree-of-freedom systems

We look for a solution in the form [1,2]

u ¼ elt (2.2.2)
where l is an arbitrary constant to be determined. Substitution of Eq. (2.2.2)
into Eq. (2.2.1) gives

2 
ml + k elt ¼ 0 (2.2.3)

Because for t  0 it is elt 6¼ 0, Eq. (2.2.3) holds only if

ml2 + k ¼ 0 (2.2.4)

Eq. (2.2.4) is the characteristic equation of the differential equation (2.2.1).

Its solutions are
pffiffiffiffiffiffiffi
l1 ¼ iw, l2 ¼ iw, i ¼ 1 (2.2.5)
where
rffiffiffiffiffi
k
w¼ > 0: (2.2.6)
m
The general solution of Eq. (2.2.1) is

u ðt Þ ¼ A0 eiwt + B 0 eiwt (2.2.7)

where A0 and B 0 are arbitrary constants. Using Euler’s formula

eiwt ¼ cos wt  i sin wt (2.2.8)

Eq. (2.2.7) is written as

u ðt Þ ¼ ðA0 + B 0 Þcos wt + i ðA0  B 0 Þsin wt (2.2.9)

or introducing the new arbitrary constants A ¼ A0 + B 0 and B ¼ i ðA0  B 0 Þ
we can write

u ðt Þ ¼ A cos wt + B sin wt (2.2.10)

The velocity is obtained by the differentiation of Eq. (2.2.10)

u_ ðt Þ ¼ wA sin wt + wB cos wt (2.2.11)

The arbitrary constants A and B can be determined if the displacement u ðt Þ
and the velocity u_ ðt Þ of the moving system are known at a certain instant t0 .
Usually, it is taken t0 ¼ 0 and the quantities u ð0Þ ¼ u0 and u_ ð0Þ ¼ u_ 0 are
referred to as the initial conditions of the motion.
 Single-degree-of-freedom systems: Free vibrations Chapter 2 85

Eqs. (2.2.10), (2.2.11) for t ¼ 0 give

A ¼ u0 (2.2.12a)

u_ 0
B¼ (2.2.12b)
w
and Eq. (2.2.10) becomes
u_ 0
u ðt Þ ¼ u0 cos wt + sin wt (2.2.13)
w
Obviously, it is u ðt Þ ¼ 0, when u0 ¼ u_ 0 ¼ 0. Hence, the system is set
to motion only if it is given an initial displacement and/or an initial
velocity.
We set
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2ffi
u_ 0
r ¼ ðu 0 Þ2 + (2.2.14)
w

and we write Eq. (2.2.13) in the form

 
u0 u_ 0
u ðt Þ ¼ r cos wt + sin wt (2.2.15)
r rw

Inasmuch as it is
 2  2
u0 u_ 0 u0 u_ 0
 1, 1 and + ¼1 (2.2.16)
r rw r rw

we can set
u0 u_ 0
¼ cos q and ¼ sin q (2.2.17)
r rw
and Eq. (2.2.15) becomes
u ðt Þ ¼ rcos ðwt  qÞ (2.2.18)
where
 
u_ 0
q ¼ tan 1 (2.2.19)
wu0

Eq. (2.2.18) states that the motion of the system is a harmonic vibration
with amplitude ju ðt Þjmax ¼ r, angular velocity w, and phase angle q. The
86 PART I Single-degree-of-freedom systems

geometrical meaning of Eq. (2.2.18) is shown in Fig. 2.2.1. The displacement

u ðt Þ of the system can be visualized as the projection OA0 of the radius r ¼ OA
on the horizontal diameter,
pffiffiffiffiffiffiffiffiffi when it rotates counterclockwise with a constant
angular velocity w ¼ k=m starting off at an angle q equal to the phase angle.
The quantity w is referred to as the natural frequency of the system. Its phys-
ical dimension is s1 and it is measured in radians per second. The natural
frequency is also called eigenfrequency as it results as an eigenvalue of the
eigenvalue problem of linear algebra (see Chapter 12).

FIG. 2.2.1 Geometrical meaning of the free undamped vibration.

The time required for the undamped system to complete one cycle of free
vibration is referred to as the natural period of vibration of the system, which
is denoted by T and measured in seconds. It is related to the natural frequency of
vibration through
2p
T¼ (2.2.20)
w
The inverse of the period
1
f¼ (2.2.21)
T
expresses the number of cycles that the system performs in 1 s. This is referred
to as the natural cyclic frequency. The unit of f is the hertz (Hz) (cycles per
second, cps) and it is related to w through
w
f¼ (2.2.22)
2p
The displacement versus time for a system with w ¼ 8s1 , u0 ¼ 0:05m, and
u_ 0 ¼ 1m=s is shown in Fig. 2.2.2.
 Single-degree-of-freedom systems: Free vibrations Chapter 2 87

0.2
(du/dt)0 T = 2p /w
0.15

0.1

0.05
u0
0

–0.05
-

–0.1

–0.15
T = 2p /w
–0.2
0 0.5 1 1.5 2
FIG. 2.2.2 Response of an undamped SDOF system.

Example 2.2.1 Free undamped vibrations of an one-story shear building

Consider the one-story shear building shown in Fig. E2.1. The columns are
assumed massless, inextensible, and fixed on the base. Moreover, the slab is
assumed uniform and rigid. The material constants are E ¼ 2.1  107 kN/m2
and n ¼ 0:2. The total load of the plate (dead plus live) is 70kN=m2 . The dimen-
sions of the column cross-sections are k1 30  30cm2 and k2 30  20cm2 . The
acceleration of gravity is g ¼ 9:81m=s2 . Determine:
(i) The natural frequencies of the building.
(ii) The motion of the slab as well as of the top cross-section of the columns
ki , if the horizontal force P ¼ 1:0  103 kN at point Að2:0,  1:0Þ of
the slab acting in the direction b ¼ ∡x,P ¼ p=6 is removed suddenly
at t ¼ 0.

(a) (b)
FIG. E2.1 One-story shear building in Example 2.2.1. (a) Vertical section. (b) Plan form.
88 PART I Single-degree-of-freedom systems

Solution
The motion of the slab is described by the two displacements U ,V of its
center O along the x,y axes, respectively, and its rotation W about O. The
stiffness and mass matrices of the structure can be established using
Eqs. (17a), (17b) for the single-story building in Example 1.5.6. However,
taking into account that the structure is symmetric with respect to both axes,
the three components of the motion are uncoupled and an ad hoc solution can
be readily obtained.
(i) The stiffness of the columns in the x and y directions are:

12EI 1y
kx1 ¼ ky1 ¼ ¼ 1360:8kN=m (1a)
h3
12EI 2y
kx2 ¼ ¼ 907:2kN=m (1b)
h3
12EI 2x
ky2 ¼ ¼ 403:2kN=m (1c)
h3
Hence the respective stiffnesses of the structure are
Kx ¼ 4kx1 + 2kx2 ¼ 7257:6kN=m (2a)
Ky ¼ 4ky1 + 2ky2 ¼ 6249:6kN=m (2b)
The torsional stiffness KW is equal to the moment produced by the elastic
forces of the columns for unit rotation of the slab. Referring to Fig. E2.2,
we have

FIG. E2.2 Displacements of point i due to rotation of slab.

u i ¼ ri W sin ai ¼ yi W (3a)

v i ¼ ri W cos ai ¼ xi W (3b)
 Single-degree-of-freedom systems: Free vibrations Chapter 2 89

Hence, the elastic forces at the top cross-section are

fx i ¼ kxi yi W (4a)

fy i ¼ kyi xi W (4b)

Besides these forces, a torsional moment is produced due to the rotation

of the top cross-section of the column

GI it E
M i ¼ kWi W, kW i ¼ , G¼ (5)
h 2ð 1 + n Þ
The moment of the elastic forces of column i with respect O is


M i ¼ yi2 kxi + xi2 kyi + kWi W (6)

and its torsional stiffness results for W ¼ 1

kWi ¼ yi2 kxi + xi2 kyi + kWi (7)

Taking into account that [3]

It1  0:141  0:34 ¼ 1:142  103 m4
It2  0:196  0:3  0:23 ¼ 4:7040  104 m4

GI 1t GI 2
kW1 ¼ ¼ 1998:7kNm, kW 2 ¼ t ¼ 823:2kNm (8)
h h
x1 ¼ x6 ¼ 4:85m, x2 ¼ x5 ¼ 0, x3 ¼ x4 ¼ 4:85m

y1 ¼ y2 ¼ y3 ¼ 2:85m, y4 ¼ y5 ¼ y6 ¼ 2:85m

we obtain the torsional stiffness of the structure

X 
KW ¼ yi2 kxi + xi2 kyi + kWi ¼ 1:9663  105 kNm (9)
i¼1

The mass of the slab and its moment of inertia with respect to O are

m ¼ 10  5  70=9:81 ¼ 356:78kNm1 s2 (10)

10  53 + 103  5
Io ¼  70=9:81 ¼ 3716:44kNs2 (11)
12

Therefore, the three natural frequencies w1 ,w2 ,w3 are

90 PART I Single-degree-of-freedom systems

rffiffiffiffiffiffi 9
Kx 1 >
w1 ¼ ¼ 4:510s > >
>
>
m >
>
rffiffiffiffiffiffi >
=
Ky 1
w2 ¼ ¼ 4:185s > (12)
m >
>
rffiffiffiffiffiffiffi >
>
>
KW 1 >
w3 ¼ ¼ 7:274s > ;
I0

(ii) The components of the force P are

Px ¼ P cos b ¼ 866:02kN Py ¼ P sin b ¼ 500:0kN (13)

and its moment with respect to O
MO ¼ yA Px + xA Py ¼ 1866:02kNm (14)
which produces the static displacements
9
U0 ¼ Px =Kx ¼ 0:1193m >
=
V0 ¼ Py =Ky ¼ 0:0800m (15)
>
;
2
W0 ¼ MO =KW ¼ 0:949  10 rad
The slab performs free vibrations with initial conditions U ð0Þ ¼ U0 ,
U_ ð0Þ ¼ 0, V ð0Þ ¼ V0 , V ð0Þ ¼ 0, Wð0Þ ¼ W0 , W_ ð0Þ ¼ 0. Hence, the dis-


placements of the center of the slab are

U ðt Þ ¼ U0 cos w1 t, V ðt Þ ¼ V0 cos w2 t, Wðt Þ ¼ W0 cos w2 t (16)
while the displacements of the i column
ui ðt Þ ¼ U ðt Þ  yi Wðt Þ, vi ðt Þ ¼ V ðt Þ + xi Wðt Þ, wi ðt Þ ¼ Wðt Þ (17)

Fig. E2.3 presents the displacements of the column at the upper right
corner (x1 ¼ 4:85m, y1 ¼ 2:85m)

0.2
u1(t)
0.15 v1(t)
w1(t)
0.1

0.05

–0.05

–0.1

–0.15

–0.2
0 1 2 3 4 5
t

FIG. E2.3 Displacements u ðt Þ,v ðt Þ and rotation wðt Þ of the top cross-section of column at
x1 ¼ 4:85m, y1 ¼ 2:85m in Example 2.2.1.
 Single-degree-of-freedom systems: Free vibrations Chapter 2 91

2.3 Free damped vibrations

In this case, it is c 6¼ 0 and the equation of motion (2.1.1) becomes
m u€ + cu_ + ku ¼ 0 (2.3.1)
We look again for a solution of the differential equation (2.3.1) in the form
of Eq. (2.2.2). This produces the following characteristic equation
ml2 + cl + k ¼ 0
or
c
l2 + l + w2 ¼ 0 (2.3.2)
m
The two roots of Eq. (2.3.2) are evaluated from the expression
r
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c c 2
l1,2 ¼    w2 (2.3.3)
2m 2m
The type of the root depends on the discriminant of the quadratic polyno-
mial, Eq. (2.3.2)

c 2
D¼  w2 (2.3.4)
2m
Eq. (2.3.2) has for:
D > 0 two discrete real roots
D ¼ 0 one double real root
D < 0 two complex conjugate roots
The type of root of Eq. (2.3.2) determines the form of the solution of the dif-
ferential equation (2.3.1), and consequently the physical response of the system.
Therefore, we distinguish three types of motion with damping, according to
whether the discriminant is positive, negative, or zero. It is expedient to discuss
first the case when the discriminant vanishes, which expresses the condition for
critical damping.

2.3.1 Critically damped system

The value of the damping coefficient c for which the discriminant vanishes is
called critical damping and it is denoted by ccr . The system under critical damp-
ing is called a critically damped system. This designation is justified because, as
we shall see below, this value represents the transition threshold from oscilla-
tory to nonoscillatory motion and vice versa. Eq. (2.3.4) for D ¼ 0 determines
this value as
c ¼ ccr ¼ 2mw (2.3.5)
For this value of c, Eq. (2.3.3) gives the double root
l ¼ w (2.3.6)
92 PART I Single-degree-of-freedom systems

and the general solution of Eq. (2.3.1) is

u ðt Þ ¼ ðA + Bt Þewt (2.3.7)
The arbitrary constants are evaluated from the initial conditions u ð0Þ ¼ u0
and u_ ð0Þ ¼ u_ 0 . Thus, we obtain
A ¼ u0 , B ¼ wu0 + u_ 0 (2.3.8)
and Eq. (2.3.7) becomes
u ðt Þ ¼ ½u0 + ðwu0 + u_ 0 Þt ewt (2.3.9)
Fig. 2.3.1 shows the plot of the displacement given by Eq. (2.3.9) if
u0 ¼ 0:05m, u_ 0 ¼ 0:2ms1 , w ¼ 8s1 . We see that the motion of the critically
damped system is nonoscillatory, but the displacement vanishes exponentially,
that is, the system returns to static equilibrium in infinite time.

0.07

0.06 (du/dt)0

0.05

0.04
u(t)

0.03
u0
0.02

0.01

0
0 0.2 0.4 0.6 0.8 1
t
FIG. 2.3.1 Response of a system with critical damping.

2.3.2 Underdamped system

This case is the most interesting because in all structural systems and in the
majority of the mechanical systems, the damping is much less than the critical.
To facilitate the study of the underdamped systems, we introduce the damping
ratio x, which is defined as
c c
x¼ ¼ <1 (2.3.10)
ccr 2mw
 Single-degree-of-freedom systems: Free vibrations Chapter 2 93

Using Eq. (2.3.10), we write Eq. (2.3.3) as

qffiffiffiffiffiffiffiffiffiffiffiffi
l1,2 ¼ xw  w x 2  1 (2.3.11)

or
l1,2 ¼ xw  iwD (2.3.12)
where
qffiffiffiffiffiffiffiffiffiffiffiffi
wD ¼ w 1  x2 (2.3.13)

Further, the equation of motion (2.3.1) can be written as

u€ + 2xwu_ + w2 u ¼ 0 (2.3.14)
whose general solution is
u ðt Þ ¼ A0 eðxw + iwD Þt + B 0 eðxwiwD Þt
(2.3.15)
¼ exwt ðA0 eiwD t + B 0 eiwD t Þ
or using Eq. (2.2.8), we obtain
u ðt Þ ¼ exwt ðA cos wD t + B sin wD t Þ (2.3.16)
The arbitrary constants A,B are evaluated from the initial conditions
u ð0Þ ¼ u0 and u_ ð0Þ ¼ u_ 0 . These yield
u_ 0 + u0 xw
A ¼ u0 , B ¼ (2.3.17)
wD
and Eq. (2.3.16) is written as
 
u_ 0 + u0 xw
u ðt Þ ¼ exwt u0 cos wD t + sin wD t (2.3.18)
wD
We observe that in the absence of damping, that is, x ¼ 0, it is wD ¼ w and
Eq. (2.3.18), as anticipated, becomes identical to Eq. (2.2.13). Eq. (2.3.18) can
also be written in the form
u ðt Þ ¼ rexwt cos ðwD t  qÞ (2.3.19)
where the amplitude r and the phase angle q are evaluated from the relations
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
u_ 0 + u0 xw 2 2
r¼ + ðu0 Þ (2.3.20a)
wD
 
1 u_ 0 + u0 xw
q ¼ tan (2.3.20b)
wD u0
Fig. 2.3.2 shows the plot of the displacement given by Eq. (2.3.19) with
u0 ¼ 1m, u_ 0 ¼ 1ms1 , w ¼ 6s1 , x ¼ 0:07. Moreover, Fig. 2.3.3 presents the
94 PART I Single-degree-of-freedom systems

dynamic response of a SDOF system for various values of the damping ratio and
u0 ¼ 1, u_ 0 ¼ 7ms1 , w ¼ 6s1 . From the plot of Eq. (2.3.19), we conclude that
the motion of an underdamped system is harmonic vibration with frequency
wD and period T ¼ 2p=wD , whose amplitude, however, decays exponentially
with time and is bounded by the envelops rexwt . Hence, wD is called the
damped frequency and the respective period the damped period of the under-
damped system.

1.5 T = 2p /wD
T T
1

re−xw t
0.5 u0
u(t)

–0.5
−re-xw t
–1

–1.5

–2
0 1 2 3 4 5
t
FIG. 2.3.2 Response of an underdamped SDOF system.

2
x=0
1.5 x=0.03
x=0.06
x=0.1
1

0.5
u(t)

–0.5

–1

–1.5

–2
0 1 2 3 4 5
t
FIG. 2.3.3 Response of an underdamped SDOF system for various values of the damping ratio.
 Single-degree-of-freedom systems: Free vibrations Chapter 2 95

If we denote the displacement at time t by u ðt Þ, then the displacement at

time t + nT , that is, after n complete cycles, will be
u ðt + nT Þ ¼ rexwðt + nT Þ cos ½wD ðt + nT Þ  q (2.3.21)
Because
2p
T¼ (2.3.22)
wD
we have


2p
xw t + n w
u ðt + nT Þ ¼ re D
cos ½ðwD t + 2npÞ  q
 p2npx
ffiffiffiffiffiffiffiffi
(2.3.23)
¼ rexwt e 1x2 cos ðw
D t  qÞ

 pffiffiffiffiffiffiffi2ffi
2npx

¼ u ðt Þe 1x
from which we obtain
 
u ðt Þ 2npx
d ¼ ‘n ¼ pffiffiffiffiffiffiffiffiffiffiffiffi (2.3.24)
u ðt + nT Þ 1  x2
The quantity d defined by Eq. (2.3.24) is called the logarithmic decrement
and can be employed to determine the damping ratio x experimentally, when the
displacements u ðt Þ and u ðt + nT Þ between n consecutive cycles are measured.
Example 2.3.1 Free damped vibrations of a silo
Fig. E2.4 presents the idealization of a silo. It consists of two massless and inex-
tensional columns of cross-sectional area a  a and a square rigid plate of mass
m. At time t ¼ 0, the silo is displaced horizontally by a force P. Then the force is
suddenly removed and the system starts to vibrate. For P ¼ 200 kN, h ¼ 5m,
a ¼ 0:3m, E ¼ 2.1  107 kN/m2, m ¼ 100kN=ms2 determine:
(i) The damping ratio x of the system, if the horizontal displacement is reduced
to u1 ¼ 0:1u0 after n ¼ 5 oscillations.
(ii) Determine the displacement at time t1 ¼ 2.

FIG. E2.4 Model of the silo in Example 2.3.1.

96 PART I Single-degree-of-freedom systems

Solution
(i) The only possible motion of the structure is its horizontal displacement
u ðt Þ. Hence, the system has a single degree of freedom. The stiffness k
of the structure is due to the relative displacement of the column ends,
which yield the stiffness

12EI 12  2:1  107  0:34 =12

k ¼2 ¼ 2  ¼ 2721:6kN=m (1)
h3 53
The force P produces an initial displacement
u0 ¼ P=k ¼ 200=2721:6 ¼ 0:0735m (2)
Eq. (2.3.24) for n ¼ 5, u1 ¼ 0:1u0 gives x ¼ 0:0731.
pffiffiffiffiffiffiffiffiffi
(ii) The natural frequency is w ¼ k=m ¼ 5:2169s1 . The motion of the
silo is given by Eq. (2.3.18) with u0 ¼ 0:0735m, u_ 0 ¼ 0,
pffiffiffiffiffiffiffiffiffiffiffiffi
wD ¼ w 1  x2 ¼ 5:2029
u ðt Þ ¼ exp e0:3813t ½0:07349 cos 5:203t + 0:005386 sin 5:203t (3)
which for t ¼ t1 ¼ 2 gives u ðt1 Þ ¼ 0:0211m.

2.3.3 Overdamped system

When the damping is greater than the critical, that is, c > ccr , the system is said
to be overdamped. Hence
c c
x¼ ¼ >1 (2.3.25)
ccr 2mw
then Eq. (2.3.3) gives two distinct real roots
qffiffiffiffiffiffiffiffiffiffiffiffi
l1, 2 ¼ xw  W, W ¼ w x2  1 x > 1 (2.3.26)

and Eq. (2.3.1) has the general solution


u ðt Þ ¼ exwt A0 eWt + B 0 eWt (2.3.27)
Taking into account that
eWt + eWt eWt  eWt
cosh Wt ¼ , sinh Wt ¼ (2.3.28)
2 2
Eq. (2.3.27) can be written in the alternative form
u ðt Þ ¼ exwt ðA cosh Wt + B sinh Wt Þ (2.3.29)
 Single-degree-of-freedom systems: Free vibrations Chapter 2 97

The arbitrary constants A,B are evaluated from the initial conditions
u ð0Þ ¼ u0 and u_ ð0Þ ¼ u_ 0 . Then, Eq. (2.3.29) becomes
 
u_ 0 + u0 xw
u ðt Þ ¼ exwt u0 cosh Wt + sinh Wt (2.3.30)
W
Eq. (2.3.30) has been plotted in Fig. 2.3.4 with u0 ¼ 1m, u_ 0 ¼ 10ms1 , and
w ¼ 6s1 . It becomes evident that the motion of the overdamped system is
nonoscillatory.

1.5
x =1
x =1.5
x =2
x =2.5

1
u(t)

0.5

0
0 0.5 1 1.5 2
t
FIG. 2.3.4 Response of an overdamped SDOF system for various values of the damping ratio.

2.4 Conservation of energy in an undamped system

The free undamped vibrations of the SDOF system are governed by Eq. (2.2.1),
namely
m u€ + ku ¼ 0 (2.4.1)
Multiplication of the previous equation by the velocity u_ yields
m u€u_ + ku u_ ¼ 0 (2.4.2)
which may be written as
1
2 1
2
md u_ + kd u ¼ 0 (2.4.3)
2 2
98 PART I Single-degree-of-freedom systems

and after integration over the interval ½0, t gives

1 1
m u_ 2 + ku 2 ¼ E (2.4.4)
2 2
where E denotes the constant
1 1
E ¼ m ½u_ 0 2 + k ½u0 2 (2.4.5)
2 2
The first term on the left side of Eq. (2.4.4) represents the kinetic energy
T of the system while the second term represents the potential energy U , which
in this case is the elastic energy of the spring. Thus, we may write
T +U ¼E (2.4.6)
and make the following statement:
In a system that performs free undamped vibrations, the total energy, kinetic
plus potential, remains constant during the whole duration of the motion and is
equal to the sum of the kinetic and potential energy given to the system at the start
of the motion.
Eq. (2.4.4) for u_ ¼ 0 gives the maximum value of the potential energy
1
Umax ¼ ku 2max ¼ E (2.4.7)
2
hence
rffiffiffiffiffiffi
2E
umax ¼ (2.4.8)
k
Similarly, the maximum value of the kinetic energy is obtained from the
same equation for u ¼ 0
1
Tmax ¼ m u_ 2max ¼ E (2.4.9)
2
hence
rffiffiffiffiffiffi
2E
u_ max ¼ (2.4.10)
m
Eqs. (2.4.7), (2.4.9) imply
Umax ¼ Tmax (2.4.11)
or
1 2 1
ku ¼ m u_ 2max ¼ E (2.4.12)
2 max 2
 Single-degree-of-freedom systems: Free vibrations Chapter 2 99

Using Eq. (2.2.18) the potential energy is written

1 1
U ¼ ku 2 ¼ kr2 cos 2 ðwt  qÞ (2.4.13)
2 2
which takes its maximum value when cos ðwt  qÞ ¼ 1. Thus, we have
1
Umax ¼ kr2 (2.4.14)
2
Similarly, the kinetic energy is written
1 1
T ¼ m u_ 2 ¼ mw2 r2 sin 2 ðwt  qÞ (2.4.15)
2 2
which takes its maximum value when sin ðwt  qÞ ¼ 1. Thus, we have
1
Tmax ¼ mw2 r2 (2.4.16)
2
Substituting Eqs. (2.4.14), (2.4.16) into Eq. (2.4.11) yields
1 1
mw2 r2 ¼ kr2 (2.4.17)
2 2
which provides the following relation for computation of the natural frequency
of the system
rffiffiffiffiffi
k
w¼ (2.4.18)
m
The method we just described for the evaluation of the natural frequency of a
system based on the conservation of energy is known as the Rayleigh method.
Actually, Eq. (2.4.18) is identical to Eq. (2.2.6). Obviously, the present method
does not seem to provide any advantage over the direct method. This is gener-
ally true. The Rayleigh method is particularly useful for the approximate deter-
mination of the natural frequency of continuous systems, for which an analytical
solution is either very difficult or impractical to obtain (see Chapters 8 and 12).
Nevertheless, modern numerical methods have significantly limited the use of
the Rayleigh method. Of course, this method is very useful in cases where we
want to check the correctness of numerical approaches in problems for which
analytical or other numerical solutions are not available for comparison.

2.5 Problems
Problem P2.1 The structure of Fig. P2.1 consists of two identical rigid bars BA
and BC , both having line density m (mass/length). The support A is a hinge
while the support C is a simple support. The bracing rod DF has a
100 PART I Single-degree-of-freedom systems

cross-sectional area A, a modulus of elasticity E, and a negligible mass. A ver-

tical force P applied at point B is suddenly removed at t ¼ 0. Determine the
motion of the structure taking the horizontal displacement u ðt Þ of C as the
parameter of motion. Data: m ¼ 5kg=m, a ¼ 3m, E ¼ 2.1  108 kN/m2,
A ¼ 12cm2 .

FIG. P2.1 Structure in Problem 2.1.

Problem P2.2 An equipment of weight 15kN is supported on the horizontal

base by three identical systems, each consisting of a spring and a damper placed
at the three vertices of an equilateral triangle. The projection of the weight cen-
ter of the equipment coincides with that of the triangle. The weight of the equip-
ment produces a static deflection ust ¼ 2:5cm. The system performs free
vibrations. The dampers are regulated so that the amplitude of vibration reduces
to 1=10 of the initial deflection after five complete cycles. Determine the
damping coefficient c and compare the frequencies w and wD .
Problem P2.3 The wooden body of mass m1 is constrained by the spring k and
the damper c as shown in Fig. P2.3. A projectile of mass m2 ¼ 0:2m1 is fired
into the body and becomes implanted in it. If the speed of the projectile is v,
determine
(i) the maximum displacement if x ¼ 0.
(ii) the displacement u ðt Þ of the body if x ¼ 0:1.

FIG. P2.3 SDOF system in Problem P2.3.

Problem P2.4 A SDOF system of mass m and stiffness k performs free vibra-
tions. At the end of four complete cycles, the displacement is u ð0Þ=3. If the mass
 Single-degree-of-freedom systems: Free vibrations Chapter 2 101

of the system is increased by 50%, determine the amplitude of the vibrations

after four complete cycles.
Problem P2.5 The packing of a sensitive instrument is modeled by the system
of Fig. P2.5. During transportation, the box of mass m2 ¼ 10m1 falls vertically
from a height h. Assuming that the box does not bounce after the collision with
the ground, determine the motion of the instrument. What is its maximum
acceleration?

FIG. P2.5 Structural system in Problem P2.5.

Problem P2.6 The horizontal force P applied to the structure of Fig. P2.6 is
suddenly removed at t ¼ 0. Determine the motion of the structure when (i)
the cables are free of any pretension and (ii) have been prestressed to withstand
compression, and compute the minimum required pretension forces. In both
cases, the cables are assumed massless. Data: a ¼ 2m, P ¼ 100kN,
m ¼ 100 kNm1 s2 , E ¼ 2.1  107 kN/m2, I ¼ 880cm4 , A ¼ 5cm2

FIG. P2.6 Structure in Problem P2.6.

Problem P2.7 In the structure of Fig. P2.7, the rigid rod AB of circular cross-
section and line mass density m ¼ m=a is supported on the ground by a spher-
ical hinge and held in vertical position by three identical elastic massless cables
of cross-sectional area A and modulus of elasticity E. The cables are prestressed
102 PART I Single-degree-of-freedom systems

so that they can withstand compression. Their anchor points D,G,F form an
equilateral triangle. Determine the motion of the structure if the horizontal
force P applied at the top of the rod in the direction of the y axis is suddenly
removed at t ¼ 0. Evaluate the minimum values of the prestress forces, which
ensure the capability of the cables to withstand compression. Data: a ¼ 2m,
m ¼ 100 kN m1s2, A ¼ 5cm2 , E ¼ 2.1  108 kN/m2, P ¼ 100kN.

(a) (b)
FIG. P2.7 Structure in Problem P2.7. (A) Vertical view. (B) Plan form.

Problem P2.8 In the pendulum of Fig. P2.8, the rigid rod suspending the con-
centrated mass m has a line mass density m ¼ 2m=L. The hinge O is elastically
restrained by the rotational spring with stiffness CR ¼ kL2 =2. The rod is
displaced by an angle q0 from the vertical position and then is left to move.
Considering small displacements, derive the equation of motion and compute
its period.

FIG. P2.8 Pendulum in Problem P2.8.

Problem P2.9 The rigid silo of Fig. P2.9 is supported on its fundament by four
identical columns of square cross-sections with a side length a=8. The silo is full
of material with density g. The bottom and the walls of the silo have a thickness
a=8 and material density 1:5g. Compute the frequencies and the periods of the
structure.
 Single-degree-of-freedom systems: Free vibrations Chapter 2 103

(a) (b)
FIG. P2.9 Silo on four columns. (a) Vertical section. (b) Plan form.

References and further reading

[1] F.B. Hildebrand, Advanced Calculus for Applications, Prentice-Hall, Inc, Englewood Cliffs,
NJ, 1962.
[2] E. Kreyszig, Advanced Engineering Mathematics, fourth ed., John Wiley & Sons, New York,
1979.
[3] S. Timoshenko, J.N. Goodier, Theory of Elasticity, McGraw-Hill, New York, 1951.
[4] S. Graham Kelly, Mechanical Vibrations, Schaum’s Outline Series, McGraw-Hill, New York,
1996.
[5] W.T. Thomson, Theory of Vibration with Applications, fifth ed., Prentice Hall, Upper Saddle
River, NJ, 1998.
[6] S.S. Rao, Mechanical Vibrations, fifth ed., Prentice Hall, Upper Saddle River, NJ, 2011.
Chapter 3

Single-degree-of-freedom
systems: Forced vibrations
Chapter outline
3.1 Introduction 105 3.5.1 Rectangular pulse load 126
3.2 Response to harmonic loading 106 3.5.2 Triangular pulse load 128
3.2.1 Response of undamped 3.5.3 Asymmetrical triangular
systems to harmonic pulse load 131
loading 106 3.5.4 Response to piecewise
3.2.2 Response of damped linear loading 135
systems to harmonic 3.6 Response to a periodic loading 137
loading 110 3.6.1 Periodic loads 137
3.3 Response to arbitrary dynamic 3.6.2 Fourier series 138
loading—Duhamel’s integral 113 3.6.3 Response of the SDOF
3.3.1 Undamped vibrations 113 system to periodic
3.3.2 Damped vibrations 116 excitation 143
3.4 Analytical evaluation of the 3.7 Response to unit impulse 146
Duhamel integral-applications 117 3.7.1 The delta function or
3.4.1 Response to step Dirac’s delta function 146
function load 117 3.7.2 Response to unit impulse 148
3.4.2 Response to ramp 3.7.3 Response to arbitrary
function load 120 loading 151
3.4.3 Response to step function 3.7.4 The reciprocal theorem
load with finite in dynamics 151
rise time 121 3.8 Problems 152
3.5 Response to impulsive loads 125 References and further reading 157

3.1 Introduction
In this chapter, the forced vibrations of the SDOF system are studied. The
dynamic model of the system is shown in Fig. 1.4.1 and the motion of the system
is governed by Eq. (1.4.8), namely
m u€ + cu_ + ku ¼ pðt Þ (3.1.1)
where pðt Þ represents an arbitrary function of time. First, the response under a
harmonic load is examined. This type of loading is particularly important in the
dynamic analysis of structures because it allows understanding the major

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00003-0

© 2020 Elsevier Inc. All rights reserved. 105
106 PART I Single-degree-of-freedom systems

differences between the static and dynamic response and identifying phenom-
ena such as resonance that are not conceived by the static consideration. More-
over, the harmonic load analysis allows studying the response of SDOF systems
under a general periodic load using the Fourier series representation of a peri-
odic load. Then, the response of a SDOF system under an arbitrary load is stud-
ied using Duhamel’s integral. Finally, the response to a unit load is discussed by
exploiting the properties of the Dirac delta function. The chapter ends by pre-
senting the dynamic reciprocal theorem. Throughout the chapter, illustrative
examples analyzing the forced vibrations of SDOF systems are presented.
The pertinent bibliography with recommended references for further study is
also included. The chapter is enriched with problems to be solved aiming at
better understanding the theoretical issues.

3.2 Response to harmonic loading

In this section, we study the motion of SDOF systems subjected to harmonic
excitations whose magnitudes may be represented by a sine or cosine function
of time, that is,

pðt Þ ¼ p0 sin wt (3.2.1a)
or

pðt Þ ¼ p0 cos wt (3.2.1b)
where p0 is a constant representing the amplitude of the force and w is the fre-
quency known as the excitation frequency. As mentioned in Section 3.1, the study
of the response of a SDOF system under harmonic loads is particularly important
in the dynamic analysis of structures because it allows understanding the major
differences between the static and dynamic response and identifying phenomena
such as resonance that are not realized by the static consideration. In fact, any load
that is periodic can be treated by resolving it into its harmonic components using
the Fourier series representation of a periodic function. The response of the sys-
tem is obtained as the superposition of individual responses to the harmonic
components of the external excitation. We defer this discussion to Section 3.6.

3.2.1 Response of undamped systems to harmonic loading

We study first the response of the system to harmonic loading in absence of
damping. In this case, the equation of motion (3.1.1) for the sine load becomes

m u€ + ku ¼ p0 sin wt (3.2.2)
Eq. (3.2.2) is an ordinary linear nonhomogeneous differential equation of the
second order with constant coefficients and its solution can be obtained using any
established mathematical method [1, 2]. Here, we look for a solution in the form

u ¼ uh + up (3.2.3)
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 107

where uh is the solution of the homogeneous equation and up a particular solu-

tion of the nonhomogeneous equation, that is any function that satisfies
Eq. (3.2.2). The solution of the homogeneous equation was obtained in
Section 2.2, where the free undamped vibrations were discussed. It is given
by Eq. (2.2.10), namely
pffiffiffiffiffiffiffiffiffi
uh ¼ A cos wt + B sin wt, w ¼ k=m (3.2.4)
There are general methods for obtaining a particular solution [1, 2]. A simple
method that can be applied in this case is the method of undetermined coeffi-
cients. According to this method, we look for a solution up in the form

up ¼ C sin wt (3.2.5)
where C is a constant to be determined. Substituting Eq. (3.2.5) into Eq. (3.2.2)
yields
 
m w2 + k C  p0 ¼ 0
from which we obtain
p0 1
C¼ , (3.2.6a)
k 1  b2
w
b¼ (3.2.6b)
w
and the particular solution (3.2.5) becomes
p0 1
up ¼ 
sin wt (3.2.7)
k 1  b2
Hence the general solution of Eq. (3.2.2) reads
p0 1
u ðt Þ ¼ Acos wt + B sin wt + 
sin wt (3.2.8)
k 1  b2
The arbitrary constants A and B are evaluated from the initial conditions.
We examine the case u ð0Þ ¼ u_ ð0Þ ¼ 0. The expression for the velocity is
obtained by differentiation of Eq. (3.2.8) with respect to time t
p0 w
u_ ðt Þ ¼ Aw sin wt + Bw cos wt + 
cos wt (3.2.9)
k 1  b2
For the considered initial conditions, Eqs. (3.2.8), (3.2.9) give
p0 b
A ¼ 0 and B ¼ 
k 1  b2
and Eq. (3.2.8) becomes
p0 1
u ðt Þ ¼   b sin wt Þ
ð sin wt (3.2.10)
k 1  b2
108 PART I Single-degree-of-freedom systems

Obviously
ust ¼ p0 =k (3.2.11)
denotes the static displacement that would be produced by a load p0 , equal to
the amplitude of the harmonic excitation, if it were to be applied statically.
The time-dependent quantity
u ðt Þ 1
Rðt Þ ¼ ¼   b sin wt Þ
ð sin wt (3.2.12)
ust 1  b 2
is called the response ratio. It is dimensionless and expresses the number that
must multiply the static displacement at time t to obtain the respective dynamic
displacement. The response ratio provides a measure of the influence of the
dynamic loading.
The extreme value
D ¼ max jRðt Þj (3.2.13)
is referred to as the dynamic magnification factor (DMF). It is a very useful
quantity in dynamic analysis because, if it has been established for a given
loading, the extreme state of deformation and stress can be obtained by static
analysis.
Eq. (3.2.12) for b ¼ 1 takes the indeterminate form
0
Rðt Þ ¼
0
whose limit can be determined using the L’H^
opital rule. Thus, on the basis of
Eq. (3.2.6b) we obtain
  b sin wt
sin wt
‘im Rðt Þ ¼ ‘im
b!1 b!1 1  b2
sin bwt  b sin wt
¼ ‘im
b!1 1  b2
(3.2.14)
wt cos bwt  sin wt
¼ ‘im
b!1 2b
wt cos wt  sin wt
¼
2
From the latter relation, it is concluded that when b tends to 1, that is, when
the excitation frequency w of the harmonic force approaches the natural fre-
quency of the system, the dynamic displacement grows indefinitely with time,
although the amplitude of the harmonic loading is finite. This phenomenon is
called resonance. The growth of the amplitude of the displacement with time
due to resonance is shown in Fig. 3.2.1. The response is periodic with a period
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 109

2p=w. A measure of the growth rate can be obtained by taking the difference
of the amplitudes of two consecutive peaks.
The time peak occurs when

20
w t/2
15
2p /w 2p /w
10

5
R(t)

–5
p
–10 p

–15
–w t/2
–20
0 1 2 3 4 5
t
FIG. 3.2.1 Response ratio of an undamped system at resonance, D ¼ max jRðt Þj ! 1
when t ! 1 (w ¼ 7, b ¼ 1).

dRðt Þ w2 sin wt
¼ ¼0 (3.2.15)
dt 2
or
np
t¼ n ¼ 1, 2, … (3.2.16)
w
Hence the difference between consecutive peaks is
np p np
R + R ¼ p cos np
w w w (3.2.17)
¼ p
 the
When the excitation force is of the cosine type, pðt Þ ¼ p0 cos wt,
employed procedure yields the particular solution
p0 1
up ¼ 
cos wt (3.2.18)
k 1  b2
and the general solution for zero initial conditions, u ð0Þ ¼ u_ ð0Þ ¼ 0, is
obtained as
p0 1
u ðt Þ ¼   b cos wt Þ
ð cos wt (3.2.19)
k 1  b2
110 PART I Single-degree-of-freedom systems

3.2.2 Response of damped systems to harmonic loading

In this case, the equation of motion (3.1.1) for the sine load becomes
m u€ + cu_ + ku ¼ p0 sin wt
 (3.2.20)
The general solution of Eq. (3.2.20) can be sought again in the form
u ¼ uh + up (3.2.21)
where uh is the homogeneous solution and up a particular solution of the
nonhomogeneous equation. We will limit our discussion to underdamped sys-
tems. Thus, the homogeneous solution is given by Eq. (2.3.16), namely
uh ¼ exwt ðA cos wD t + B sin wD t Þ (3.2.22)
which was derived in Section 2.3.2. The particular solution is established using
the method of undetermined coefficients. According to this method, up is sought
in the form [1, 2]
 + C2 cos wt
up ¼ C1 sin wt  (3.2.23)
which is introduced into Eq. (3.2.20) to yield
p0 1  b2 p0 2xb
C1 ¼   , C2 ¼    (3.2.24)
k 1  b 2 2 + ð2xbÞ2 k 1  b2 2 + ð2xb Þ2

Hence, the general solution of Eq. (3.2.20) becomes

u ðt Þ ¼ exwt ðA sin wD t + B cos wD t Þ

|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Transient response

p0 1  
+   2
1  b2 sin wt   2xb cos wt  (3.2.25a)
k 1  b2 + ð2xb Þ2
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Steady-state response

 the previous procedure yields

If the excitation force is p0 cos wt,
u ðt Þ ¼ exwt ðA sin wD t + B cos wD t Þ
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Transient response

p0 1    
+   2
1  b2 2xb sin wt  + 1  b2 cos wt  (3.2.25b)
k 1  b2 + ð2xbÞ2
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Steady-state response

It is clear that the first term in Eq. (3.2.25a) decays rapidly with time because
of the exponential term exwt , so its contribution becomes negligible after a
short time, which, of course, depends on the damping ratio. Thus, the response
of the system is governed by the second term. For this reason, we say that the
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 111

first term expresses the transient response while the second term expresses the
steady-state response of the system. This is shown in Fig. 3.2.2.

FIG. 3.2.2 Response of an underdamped system to harmonic loading ðw ¼ 5, w ¼ 6, p0 =k ¼ 1,

x ¼ 0:1, u0 ¼ 1, u_ 0 ¼ 10Þ.

Referring to Eq. (3.2.25a) we see that the steady-state response can be writ-
ten in the form
  qÞ
u ðt Þ ¼ rsin ðwt (3.2.26)
where


p0 h 
2 2 2
i1=2
1 2xb
r¼ 1  b + ð2xbÞ , q ¼ tan (3.2.27)
k 1  b2
Therefore, the steady-state response of the underdamped system subjected
to a harmonic loading is an undamped free vibration.
The DMF is
r h 2 i1=2
D ¼ max jRðt Þj ¼ ¼ 1  b2 + ð2xbÞ2 (3.2.28)
p0 =k
We observe that D ¼ D ðbÞ. Consequently the maximum value of D is
obtained when
dD
¼0 (3.2.29)
db
This condition gives
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b ¼ 1  2x 2 (3.2.30)
112 PART I Single-degree-of-freedom systems

6
x=0

5
x=0.1

4
D (b)

3
x=0.2

2 x=0.3
ξ=0.7
x=0.5
1
ξ=1.0

0
0 0.5 1 1.5 2 2.5 3
Frequency ratio, b
FIG. 3.2.3 Variation of the amplitude of the response ratio Dðb, x Þ.

and

1
Dmax ¼ pffiffiffiffiffiffiffiffiffiffiffiffi (3.2.31)
2x 1  x2

Apparently, the maximum value of the dynamic factor does not occur for
b ¼ 1. Nevertheless, for a small value of x it is b  1, for example, for
x ¼ 0:05 Eq. (3.2.30) gives b ¼ 0:9975. Fig. 3.2.3 shows the variation of D ver-
sus the frequency ratio b for different values of the damping ratio x. Note that
if x ¼ 0, Dmax becomes infinite.

4
1/2x

0
R(t)

–2
–1/2x
–4

–6
0 5 10 15
t
FIG. 3.2.4 Response ratio of an underdamped system at resonance (w ¼ 5, x ¼ 0:1, b ¼ 1,
D ¼ max jRðt Þj ! 1=2x when t ! 1).
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 113

Let us study now the response of the system when b ¼ 1, which is the
conventional value for resonance. For this value of b it is w ¼ w and
Eq. (3.2.25a) becomes

p0 cos wt
u ðt Þ ¼ exwt ðAsin wD t + B cos wD t Þ  (3.2.32)
k 2x
If we assume zero initial conditions, u ð0Þ ¼ u_ ð0Þ ¼ 0, Eq. (3.2.32) gives

" ! #
1 xwt x
Rðt Þ ¼ e pffiffiffiffiffiffiffiffiffiffiffiffi sin wD t + cos wD t  cos wt (3.2.33)
2x 1  x2

The time variation of Rðt Þ is shown in Fig. 3.2.4.

We observe that the DMF D ¼ max jRðt Þj, for x 6¼ 0 tends asymptotically
to the value 1=2x. Namely, the presence of damping prevents the occurence
of infinite displacements when x 6¼ 0. The conclusion drawn from this analysis
is that damping limits considerably the consequences of resonance. However,
the resonance phenomenon should not be ignored in structural design because
it produces displacements much larger than the static ones.

3.3 Response to arbitrary dynamic loading—Duhamel’s

integral
3.3.1 Undamped vibrations
In this section, we study the response of undamped SDOF systems when the
external loading pðt Þ is neither harmonic nor periodic, but an arbitrary function
of time. In this case, the equation of motion is

m u€ + ku ¼ pðt Þ (3.3.1)

which we write in the form

1 pffiffiffiffiffiffiffiffiffi
u€ + w2 u ¼ pðt Þ, w ¼ k=m (3.3.2)
m
There are several methods to obtain the solution of Eq. (3.3.2). A con-
venient, straightforward, and rather simple method for solving ordinary dif-
ferential equations with constant coefficients is the Laplace transform
method [1, 3]. This method is based on the Laplace transform, which for a func-
tion u ðt Þ, t  0 is commonly denoted by L and defined as
Z 1
L½u ðt Þ ¼ U ðs Þ ¼ u ðt Þest dt (3.3.3)
0

where s is the variable of the transform.

114 PART I Single-degree-of-freedom systems

The inverse Laplace transform is defined as the function u ðt Þ, which results

as the solution of the integral equation (3.3.3), when the function U ðs Þ is given.
It is denoted by

u ðt Þ ¼ L1 ½u ðt Þ (3.3.4)

and it is given by the Fourier-Mellin integral.

A method to evaluate the Laplace transform of a simple function as well as
its inverse is to use the Tables of Laplace transform [4]. On the other hand,
ready-to-use computer applications such as Wolfram ALFA, etc., are also avail-
able. These applications can be employed to obtain the Laplace transform and
its inverse of complicated functions.
Using integration by parts, we can derive the Laplace transform of the
derivatives of a function u ðt Þ. Thus, we have

L½u_ ðt Þ ¼ sU  u ð0Þ (3.3.5a)

L½u€ðt Þ ¼ s2 U  ½su ð0Þ + u_ ð0Þ (3.3.5b)

Application of the Laplace transform to both sides of Eq. (3.3.2) yields

 1
L u€ + w2 u ¼ L½pðt Þ
m
or

1
L½u€ + w2 L½u  ¼ L½pðt Þ (3.3.6)
m

which by virtue of Eq. (3.3.5b) becomes

1
s2 U ðs Þ  ½su ð0Þ + u_ ð0Þ + w2 U ðs Þ ¼ P ðs Þ (3.3.7)
m
where P ðs Þ is the Laplace transform of pðt Þ.
Eq. (3.3.7) is an algebraic equation with respect to U ðs Þ giving

s 1 1 1
U ðs Þ ¼ u ð 0Þ + 2 u_ ð0Þ + P ðs Þ (3.3.8)
s2 +w 2 s +w 2 m s + w2
2

The function u ðt Þ is obtained by taking the inverse Laplace transform of

Eq. (3.3.8), namely
u ðt Þ ¼ L1 U ðs Þ
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 115

or
  
1 s 1 s 1 1 s
u ðt Þ ¼ u ð0ÞL + u_ ð0ÞL + L P ðs Þ
s 2 + w2 s 2 + w2 m s 2 + w2
(3.3.9)
From the table of the Laplace transforms we obtain

s
L1 2 ¼ cos wt (3.3.10a)
s + w2

1 1 sin wt
L 2 2
¼ (3.3.10b)
s +w w
Now we focus our attention on the last term on the right side of Eq. (3.3.9).
Its inverse Laplace transform can be obtained using the convolution theorem.
The convolution of two functions f ðt Þ and g ðt Þ denoted by f ðt Þ∗ g ðt Þ or
ðf ∗ g Þt is defined as
Z t
f ðt Þ∗ g ðt Þ ¼ f ðt  τÞg ðτÞdτ
0
Z t (3.3.11)
¼ g ðt  τÞf ðτÞdτ
0

The Convolution Theorem

Let f ðt Þ and g ðt Þ be functions with Laplace transforms F ðs Þ and G ðs Þ, respec-
tively, that is, L½f ðt Þ ¼ F ðs Þ and L½g ðt Þ ¼ G ðs Þ, then
L½ðf ∗ g Þðt Þ ¼ F ðsÞG ðsÞ (3.3.12a)
or equivalently
Z t
1
L ½F ðs ÞG ðs Þ ¼ ðf ∗g Þðt Þ ¼ f ðt  τÞg ðτÞdτ (3.3.12b)
0
Eq. (3.3.12b) can be employed to obtain the inverse Laplace of the last term
on the right side of Eq. (3.3.9). Thus, for F ðs Þ ¼ 1=ðs2 + w2 Þ and G ðs Þ ¼ P ðs Þ,
we obtain
 Z t
1 1
L1 2 P ðs Þ ¼ pðτÞ sin ½wðt  τÞdτ (3.3.13)
s + w2 mw 0
Hence, substituting Eqs. (3.3.10a), (3.3.10b), (3.3.13) into Eq. (3.3.9) gives
the solution of Eq. (3.3.1)
u_ ð0Þ
u ðt Þ ¼ sin wt + u ð0Þcos wt
w Z
t
1 (3.3.14)
+ pðτÞsin ½wðt  τÞdτ
mw 0
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Duhamel’s integral
116 PART I Single-degree-of-freedom systems

It obvious that the first two terms in Eq. (3.3.14) express the contribution of
the initial conditions to the motion of the system. The third term, which
expresses the contribution of the external loading, is known as the Duhamel
integral for the undamped system.

3.3.2 Damped vibrations

When damping is taken into account, the motion is governed by Eq. (3.1.1),
namely
m u€ + cu_ + ku ¼ pðt Þ (3.3.15)
which for an underdamped system is written as
1
u€ + 2xwu_ + w2 u ¼ pðt Þ (3.3.16)
m
For an arbitrary loading function pðt Þ, the solution of Eq. (3.3.16) is
obtained using the Laplace transform method following the procedure pre-
sented in the previous section. Thus taking the Laplace transform of
Eq. (3.3.16), we obtain
1
L½u€ + 2xwL½u_  + w2 L½u  ¼ L½pðt Þ (3.3.17)
m
Using Eqs. (3.3.5a), (3.3.5b) and solving for U ðs Þ, we obtain
s 1
U ðs Þ ¼ u ð 0Þ + 2 ½2xwu ð0Þ + u_ ð0Þ
s2+ 2xws + w 2 s + 2xws + w2
(3.3.18)
1 P ðs Þ
+
m s 2 + 2xws + w2
The evaluation of the inverse Laplace transform using the table of the
Laplace transforms requires the factorization of the denominator polynomial
f ðs Þ ¼ s 2 + 2xws + w2 (3.3.19)
Its roots are
l1 ¼ xw + iwD , l2 ¼ xw  iwD (3.3.20)
Hence
s 2 + 2xws + iw2 ¼ ðs  l1 Þðs  l2 Þ (3.3.21)
From the table of the Laplace transforms we obtain

1 1 1  l1 t 
L ¼ e  e l2 t
ðs  l1 Þðs  l2 Þ l1  l2
(3.3.22a)
1
¼ exwt sin wD t
wD
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 117


s 1  l1 t 
L1 ¼ l1 e  l2 el2 t
ðs  l 1 Þð s  l 2 Þ l1  l2 (3.3.22b)
xwt
¼e cos wD t
 Z t
P ðs Þ 1
L1 ¼ pðτÞexwðtτÞ sin wD ðt  τÞdτ (3.3.22c)
ðs  l 1 Þ ð s  l 2 Þ wD 0

The last expression was obtained using the convolution theorem.

By virtue of Eqs. (3.3.22a)–(3.3.22c), Eq. (3.3.18) gives
u ðt Þ ¼ L1 ½U ðsÞ

u_ ð0Þ + u ð0Þxw
¼ sin wD t + u ð0Þ cos wD t exwt
wD
Z t (3.3.23)
1
+ pðτÞexwðtτÞ sin ½wD ðt  τÞdτ
mwD 0
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
Duhamel’s integral

It is obvious that the first term in Eq. (3.3.23) expresses the contribution of
the initial conditions to the motion of the system. The last term represents the
Duhamel integral for the underdamped motion. It is obvious that Eq. (3.3.23)
for x ¼ 0 yields Eq. (3.3.14) as anticipated. The Duhamel integral can also be
derived using the method described in Section 3.7.3.

3.4 Analytical evaluation of the Duhamel integral-applications

The analytical evaluation of the Duhamel integral presents considerable diffi-
culties, especially when damping is taken into account. In practical terms, the
analytical integration is recommended when the excitation function is simple.
In recent years, however, the development of symbolic languages, such as
MATHEMATICA and MAPLE, has boosted interest in the use of analytical
integration. This paragraph presents some applications of the Duhamel integral
that help in understanding the response of an SDOF system under certain char-
acteristic forms of dynamic loading. The integration is performed analytically.

3.4.1 Response to step function load

A step load is a nonperiodic load applied suddenly at t ¼ t0 and remains constant
during the whole duration of the motion. Mathematically, this load can be repre-
sented using the Heaviside step function
pðt Þ ¼ p 0 H ðt Þ (3.4.1)
where

0, t < t0
H ðt  t0 Þ ¼ (3.4.2)
1, t > t0
118 PART I Single-degree-of-freedom systems

is the Heaviside step function and p0 the magnitude of the load. Fig. 3.4.1 shows
the step function load applied at t ¼ 0.

FIG. 3.4.1 The step function load.

For u ð0Þ ¼ u_ ð0Þ ¼ 0 the displacement is obtained from the Duhamel integral
in Eq. (3.3.14)
Z t Z t
1 p0
u ðt Þ ¼ p0 sin ½wðt  τÞdτ ¼  sin ½wðt  τÞd ½wðt  τÞ
mw 0 mw2 0
p0
¼ ½ cos wðt  τÞt0
mw2
or
p0
u ðt Þ ¼ ð1  cos wt Þ (3.4.3)
k
Taking into account that ust ¼ p0 =k represents the static displacement, the
response ratio is
u ðt Þ
R ðt Þ ¼ ¼ ð1  cos wt Þ (3.4.4)
ust
and the DMF
D ¼ max jRðt Þj ¼ 2 (3.4.5)
Eq. (3.4.5) shows that the suddenly applied load produces a maximum dis-
placement that is twice as large than the displacement that the load p0 would
produce if it were applied statically (slowly). This is an elementary but impor-
tant result that illustrates the difference between static and dynamic loading of a
structure.
When the damping is taken into account, the displacement for zero initial
conditions, u ð0Þ ¼ u_ ð0Þ ¼ 0, is obtained from Eq. (3.3.23)
Z t
p0
u ðt Þ ¼ exwðtτÞ sin ½wD ðt  τÞdτ (3.4.6)
mwD 0
The evaluation of the Duhamel integral is more complicated. Nevertheless,
using MAPLE we obtain
" ! #
p0 x
u ðt Þ ¼ 1  cos wD t + pffiffiffiffiffiffiffiffiffiffiffiffi sin wD t exwt (3.4.7)
k 1  x2
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 119

2.5
x=0
2
x=0.05

x=0.15
1.5

1
R(t)

0.5

0.

–0.5
0 0.5 1 1.5 2 2.5 3 3.5 4
t/TD (TD=2p /w D)
FIG. 3.4.2 Response ratio of a system under step load.

We observe that Eq. (3.4.7) for x ¼ 0 becomes identical to Eq. (3.4.3).

The response ratio results from Eq. (3.4.7)
" ! #
x xwt
Rðt Þ ¼ 1  cos wD t + pffiffiffiffiffiffiffiffiffiffiffiffi sin wD t e (3.4.8)
1  x2

Fig. 3.4.2 shows the plot of Eq. (3.4.8) for different values of the
damping ratio.
The extreme values of Rðt Þ occur when
" #
2
dR xwt ðwx Þ
¼e + wD sin wD t ¼ 0 (3.4.9)
dt wD

which gives
np
tn ¼ n ¼ 0,1, 2, … (3.4.10)
wD

Substituting tn into Eq. (3.4.8) yields

pffiffiffiffiffiffiffi2ffi
Rðtn Þ ¼ 1  ð1Þn enx= 1x
(3.4.11)
120 PART I Single-degree-of-freedom systems

The maximum value results for n ¼ 1. Hence

pffiffiffiffiffiffiffi2ffi
D ¼ 1 + ex= 1x
(3.4.12)

that is, the maximum displacement occurs at the first peak, where t1 ¼ p=wD .
This is shown in Fig. 3.4.2. Eq. (3.4.12) for x ¼ 0 gives D ¼ 2 as anticipated.

3.4.2 Response to ramp function load

The ramp function load is a load that increases linearly with time. Mathemat-
ically, it is defined by
pðt Þ ¼ lt t0 (3.4.13)

where l is a nonzero constant denoting the rate of increase of the load

(Fig. 3.4.3).

FIG. 3.4.3 Ramp function load; l ¼ p0 =t1 .

For zero initial conditions, u0 ¼ u_ 0 ¼ 0, the response of an undamped SDOF

system to a ramp function load is obtained by substituting the load pðt Þ ¼ p0 t=t1
into the Duhamel integral in Eq. (3.3.14)
Z
p0 =t1 t
u ðt Þ ¼ τ sin ½wðt  τÞdτ (3.4.14)
mw 0

which after integration gives


p0 t sin wt
u ðt Þ ¼  (3.4.15)
k t1 wt1

Fig. 3.4.4 shows the response ratio Rðt Þ ¼ u ðt Þ=ust , ust ¼ p0 =k, of the
undamped system to a ramp function load. We see that it oscillates about the
line p0 t=t1 .
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 121

30

25

20
R(t)

15

10

0
0 0.5 1 1.5 2 2.5 3
t/T
FIG. 3.4.4 Response ratio of a system under ramp function load (t1 ¼ T =10).

For damped systems, the analytical evaluation is rather complicated. How-

ever, it can be evaluated using a symbolic language.

3.4.3 Response to step function load with finite rise time. Static load
The step function load with finite rise time is a constant load that, however, is
not applied suddenly, but rises linearly up to a value p0 within a time t1 , the rise
time, and thereafter remains constant, as shown in Fig. 3.4.5. Mathematically it
is defined by
(p
0
t t  t1
pðt Þ ¼ t1 (3.4.16)
p0 t > t1

FIG. 3.4.5 Step function load with finite rise time.

For zero initial conditions, u0 ¼ u_ 0 ¼ 0, the response of the system is

obtained by studying the motion in two phases:
122 PART I Single-degree-of-freedom systems

Phase Ι: Forced vibration under a ramp function load.

This phase starts at t ¼ 0 and ends at t ¼ t1 . The response of the undamped sys-
tem is given by Eq. (3.4.15)


p0 t sin wt
u I ðt Þ ¼  , 0 < t  t1 (3.4.17)
k t1 wt1

Phase ΙI: Forced vibration under a step function load.

This phase starts at t ¼ t1 with initial conditions uII ð0Þ ¼ uI ðt1 Þ, u_ II ð0Þ ¼ u_ I ðt1 Þ
and loading pðt Þ ¼ p0 . The solution is given by Eq. (3.3.14). The value of the
Duhamel integral is given by Eq. (3.4.3). Thus, we have

u_ I ðt1 Þ p0
uII ðt Þ ¼ sin wt+ uI ðt1 Þcos wt+ ð1  cos wtÞ (3.4.18)
w k

where t¼ t  t1 > 0.

Note that for t1 ¼ nT , n ¼ 1, 2, …, Eq. (3.4.17) gives
p0
uI ðt1 Þ ¼ , u_ I ðt1 Þ ¼ 0 (3.4.19)
k
and substituting into Eq. (3.3.18) gives

p0
uII ðt Þ ¼ (3.4.20)
k

or

RII ðt Þ ¼ 1 (3.4.21)

which means that the motion in the constant load phase is not oscillatory but the
displacement is constant and equal to the static displacement.
Fig. 3.4.6 presents the response ratio Rðt Þ for different values of t1 .
We observe that for smaller values of t1 =T , the response is similar to that
of a step function load while for larger values, the response is similar to that of
a static load. Therefore, the loads in real structures should not be applied
suddenly but slowly rising, that is, in time much larger than the natural
period of the structure to avoid dynamic magnification effects. Due to this
property, the step function load with finite rise time is also called static
loading.
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 123

FIG. 3.4.6 Response ratio of a system under a step function load with finite rise time.

Example 3.4.1 Blast load on one-story building

A blast-induced pressure wave strikes the one-story building in Example 2.2.1
in the x direction. The time variation of the blast pressure is represented by
pðt Þ ¼ p0 ð1  t=t0 Þet=t0 (Fig. E3.1). Determine the stress resultants max Qx ,
max Mx of the columns. The peak positive pressure is p0 ¼ 12kN=m2 and
the duration of the positive phase t0 ¼ 1. The system is at rest at t ¼ 0.

FIG. E3.1 Blast-induced pressure in Example 3.4.1.

Solution
The mass of the structure, its stiffness in the x direction, and the corresponding
natural frequency were evaluated in Example 2.2.1
m ¼ 356:78kN m1 s2 , k ¼ Kx ¼ 7257:6kN=m, w ¼ w1 ¼ 4:51
124 PART I Single-degree-of-freedom systems

If A represents the area of the surface struck by the blast-induced pressure,

the peak force applied at the floor level of the building is
1 1
P ¼ A  p0 ¼  5  5  12 ¼ 150kN=m2 (1)
2 2
Hence
ust ¼ P=k ¼ 150=7257:6 ¼ 0:02067m (2)
Because the system is at rest at t ¼ 0, the response is given by Eq. (3.3.14)
with u0 ¼ u_ 0 ¼ 0
Z t
P
u ðt Þ ¼ ð1  τÞexp ðτÞsin ½wðt  τÞdτ (3)
mw 0
The analytical evaluation of the Duhamel gives
(
P 2w3 w2 ð w2  1 Þ
u ðt Þ ¼ sin wt  coswt
k ðw2 + 1Þ2 ð w2 + 1Þ 2
)
w2 ð w2  1 Þ w2
+ exp ð t Þ  t exp ðt Þ (4)
ð w 2 + 1Þ 2 ð w 2 + 1Þ

1.5
D = 1.265

0.5
R(t)

tmax = 0.563
–0.5

–1

–1.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
t
FIG. E3.2 Response ratio in Example 3.4.1.

Fig. E3.2 shows the plot of the response ratio Rðt Þ ¼ u ðt Þ=ðP=k Þ. The DMF
was found at D ¼ max jRðt Þj ¼ 1:1265. It was determined as the maximum
value of the array used to plot Rðt Þ. Thus, we have
umax ¼ Du st ¼ 1:1265  0:02067 ¼ 0:02328m (5)
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 125

The extreme values of the shear forces and bending moments are:
Columns 30  30
max Qx ¼ kx umax ¼ 1360:8  0:02328 ¼ 31:68kN
h
max Mx ¼ max Qx ¼ 31:68  5=2 ¼ 79:20kNm
2
Columns 30  20
max Qx ¼ kx umax ¼ 907:2  0:02328 ¼ 21:12kN
h
max Mx ¼ max Qx ¼ 21:12  5=2 ¼ 52:80kNm
2

3.5 Response to impulsive loads

In this section, a special group of dynamic loads with similar characteristics will
be considered: the impulsive loads or shock loads. The main characteristics of
such loads are their short duration and their high intensity (large magnitude).
Their duration is of the order of the natural period of the structure. Impulsive
loads are of great importance in the design of certain structural systems, for exam-
ple, buildings subject to aboveground wind blasts or explosions. Because of the
short duration, the maximum response is reached before damping starts to absorb
much energy. As a result, the maximum response is not significantly affected by
the presence of damping. Therefore, we can ignore damping in the study of the
SDOF systems under impulsive loads in the analysis that follows. The time func-
tion representing the impulsive load is arbitrary, e.g., Fig. 3.5.1. Nevertheless,
important conclusions can be drawn by considering impulsive loads with a reg-
ular shape. The response of a system to an impulsive load is studied in two phases.
The first phase is a forced vibration under the impulsive load with zero initial
conditions while the second phase is a free vibration with initial conditions being
the displacement and velocity that the system attains at the end of the first phase.
p(t)

t
FIG. 3.5.1 Time variation of an impulsive load.
126 PART I Single-degree-of-freedom systems

The response of the system to an impulsive load can be analyzed using the
methods presented in previous sections for the solution of the differential equa-
tion of motion under an arbitrary loading, that is, either by solving directly the
differential equation or by evaluating Duhamel’s integral. Another method to
obtain the response is to express the pulse as the superposition of two or more
simpler pulses for which the response solution is available or simple to deter-
mine (Fig. 3.5.2). Nevertheless, the analytical methods, especially for arbitrary
impulse loads, have lost their importance because of the development of effi-
cient numerical methods. For this reason, only two simple impulsive loads
are considered, the rectangular pulse load and the triangular pulse load.

p(t )

p(t ) p 0H (t ) if t t1
p0 p(t ) 0 if t t1

O t
t1

p(t )

p0 p(t ) p0 sin t if t t1
p(t ) 0 if t t1

O t
t1
FIG. 3.5.2 Impulsive loads represented by simple functions.

3.5.1 Rectangular pulse load

The rectangular pulse load is shown in Fig. 3.5.3. Mathematically, this load is
defined as 
p0 t  t1
pðt Þ ¼ (3.5.1)
0 t > t1

FIG. 3.5.3 Rectangular pulse load.

 Single-degree-of-freedom systems: Forced vibrations Chapter 3 127

Phase Ι: Forced vibration

The response of the system is a forced vibration due to the suddenly applied load
p0 at t ¼ 0 with zero initial conditions, u0 ¼ u_ 0 ¼ 0. The displacement is given
by Eq. (3.4.3), namely
p0
uI ðt Þ ¼ ð1  cos wt Þ, 0 < t  t1 (3.5.2)
k

Phase ΙΙ: Free vibration

The displacement is given by Eq. (2.2.13), namely
u_ I ðt1 Þ
uII ðt Þ ¼ sin wt+ uI ðt1 Þcos wt, t¼ t  t1  0 (3.5.3)
w
where
p0
uI ðt1 Þ ¼ ð1  cos wt1 Þ (3.5.4a)
k
p0 w
u_ I ðt1 Þ ¼ sin wt1 (3.5.4b)
k
We now evaluate the DMF D. First, we assume that the maximum displace-
ment occurs in Phase I. Eq. (3.5.2) has a maximum when
p0 w
u_ I ðt Þ ¼ sin wt ¼ 0 or wt ¼ np n ¼ 1, 2, …
k
from which we obtain
np n 2p nT
t¼
¼ ¼ (3.5.5)
w 2w 2
Therefore, the displacement can reach its maximum umax within Phase I
if t1  T =2. In this case, it is D ¼ 2.
We assume now that t1 < T =2, then the maximum displacement umax
occurs in Phase II. Hence, it will be

FIG. 3.5.4 Response spectrum for rectangular pulse load.

128 PART I Single-degree-of-freedom systems

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

u_ I ðt1 Þ 2
umax ¼ + ½ u I ðt1 Þ 2 (3.5.6)
w
Using Eqs. (3.5.4a), (3.5.4b) the previous relation becomes
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi




p0 2p 2p 2p
umax ¼ 1  2cos t1 + cos 2 t1 + sin 2 t1
k T T T
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

 ffi
p0 2p
¼ 2 1  cos t1
k T

or


2p0 t1
umax ¼ sin p (3.5.7)
k T
and the DMF D will be given by


umax t1 T
D¼ ¼ 2sin p , t1  (3.5.8)
p0 =k T 2
Eq. (3.5.8) shows that the maximum response depends only on the ratio
t1 =T . The plot of the function D ¼ D ðt1 =T Þ shown in Fig. 3.5.4 is referred
to as the displacement response spectrum or simply the response spectrum of
the impulsive load (see also Chapter 6). It is evident that the response spectrum
of a pulse load serves to determine the maximum response of the system under
this load without solving the differential equation of motion for the particular
pulse load.

3.5.2 Triangular pulse load

The next impulsive load is the decreasing triangular pulse load shown in
Fig. 3.5.5a.
Mathematically, this load is defined as
8

< t
p0 1  t  t1
pðt Þ ¼ t1 (3.5.9)
:
0 t > t1

(a) (b) (c)

FIG. 3.5.5 Decomposition of a triangular pulse load.
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 129

The response of the system is studied again in two phases:

Phase Ι: Forced vibration
Eq. (3.5.9) suggests that the pulse load can be considered as the superposition of
a rectangular pulse and a ramp function pulse (Fig. 3.5.5b and c) and use the
solutions we already found for the respective load functions, that is,
Eqs. (3.4.3), (3.4.15).
Thus, we obtain


p0 sin wt t
uI ðt Þ ¼ 1  cos wt +  , 0  t  t1 (3.5.10)
k t1 w t1

Phase ΙΙ: Free vibration

The displacement is given by Eq. (2.2.13)

u_ I ðt1 Þ
uII ðt Þ ¼ sin wt+ uI ðt1 Þcos wt, t¼ t  t1  0 (3.5.11)
w
where


p0 sin wt1
u I ðt1 Þ ¼  coswt1 (3.5.12a)
k t1 w


p0 w cos wt1 1
u_ I ðt1 Þ ¼ + sin wt1  (3.5.12b)
k wt1 wt1

We now evaluate the DMF D.

The maximum displacement in Phase I is obtained when


cos wt 1
u_ I ðt Þ ¼ w sin wt +  ¼0 (3.5.13)
t1 t1

which may be written as



t1 t t
2p sin 2p + cos 2p 1¼0 (3.5.14)
T T T
When t1 =T is specified, Eq. (3.5.14) is a nonlinear algebraic equation for the
ratio t=T . Solving this equation and introducing the obtained solution in
Eq. (3.5.10) yield the respective maximum displacement max uI . The curve
RI in Fig. 3.5.6 represents the function max uI =ust versus t1 =T .
130 PART I Single-degree-of-freedom systems

In Phase II, the maximum response is obtained from the expression

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

u_ I ðt1 Þ 2
max uII ¼ + ½uI ðt1 Þ2 (3.5.15)
w
where uI ðt1 Þ and u_ I ðt1 Þ are given by Eqs. (3.5.12a), (3.5.12b), which yield
when expressed in terms of t1 =T

2
 3
t1
sin 2p

p0 6 T t1 7
uI ðt1 Þ ¼ 6 4  cos 2p 7 (3.5.16a)
k t1 T 5
2p
T

2
 3
t1
cos 2p

p0 w 6
6 T t1 1 7 7
u_ I ðt1 Þ ¼ + sin 2p  (3.5.16b)
k 4 t1 T t1 5
2p 2p
T T

The curve RII in Fig. 3.5.6 represents the function max uII =ust versus t1 =T .
From this figure, we conclude that for t1 =T  0:4 the maximum response D of
the system to the triangular pulse load occurs in Phase II while for t1 =T > 0:4 it
occurs in Phase I.

R = maxu /ust
D(0.4)=1.0513
D(t1/T)

R = maxu /ust

t1/T
FIG. 3.5.6 DMF for the triangular pulse load.
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 131

3.5.3 Asymmetrical triangular pulse load

The next impulsive load is the asymmetrical triangular pulse load shown in
Fig. 3.5.7. The response can be determined in three phases using the solutions
obtained in Sections 3.4.2 and 3.5.2.
Phase Ι: Forced vibration


p0 sin wt
uI ðt Þ ¼ t , 0  t  t1 (3.5.17)
k t1 w

Phase II: Forced vibration

u_ I ðt1 Þ
uII ðt Þ ¼ sin wet + uI ðt1 Þ cos wet
w

 (3.5.18)
p0 e sin wet et
+ 1  cos wt +  , 0  et ¼ t  t1 < t2
k t2 w t2

Phase ΙΙΙ: Free vibration

u_ II ðt1 + t2 Þ
uIII ðt Þ ¼ sin wt+ uII ðt1 + t2 Þcos wt, t¼ t  ðt1 + t2 Þ  0
w
(3.5.19)

FIG. 3.5.7 Asymmetrical triangular pulse load.

132 PART I Single-degree-of-freedom systems

Table 3.5.1 summarizes the response ratios for various loadings.

TABLE 3.5.1 Response ratios Rðt Þ for various loading functions.

Loading pðt Þ Response ratio Rðt Þ
Rðt Þ ¼ 1  cos wt

Rðt Þ ¼ 1  cos wt, t  t1

Rðt Þ ¼ coswðt  t1 Þ  cos wt, t  t1


1 sin wt
Rðt Þ ¼ t , t  t1
t1 w
1
Rðt Þ ¼ 1 + ½ sinwðt  t1 Þ  sin wt , t  t1
wt1

sin wt t
Rðt Þ ¼ 1  cos wt +  , t  t1
wt1 t1
sinwt1  sinwðt  t1 Þ
Rðt Þ ¼  cos wt, t  t1
wt1


1 sin wt t1
Rðt Þ ¼ t , 0t 
t1 w 2
 
 
2 1 t1 t1
Rðt Þ ¼ t1  t + 2sin w t   sin wt ,  t  t1
t1 w 2 2


2 t1
Rðt Þ ¼ sinwðt  t1 Þ + 2sinw t   sinwt , t  t1
wt1 2
  b sinwt
sin wt
Rðt Þ ¼
1  b2

Example 3.5.1 One-story building under an impulsive moment

The rigid slab of the one-story building of Fig. E3.3a,b is loaded by the impul-
sive moment M ðt Þ, whose time variation is shown in Fig. E3.3c. The columns
are assumed massless and fixed on the ground. The pulse data are:
M0 ¼ 300kNm, M1 ¼ 500kNm, t1 ¼ T , where T is the period of the structure.
The other data are the same as in Example 2.2.1.
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 133

y
k1 k2 k1
M (t ) x
O 5m

k1 k2 k1
10 m
(a) (b)

(c)
FIG. E3.3 One-story building under impulsive load.

Determine the stress resultants max Qx , max Qy , max Mx , max My , and

max Mf of the corner columns using time history analysis.

Solution
Because the structure is symmetric with respect to both axes x and y, the only
possible motion of the slab due to the moment M ðt Þ is the rotation fðt Þ about its
center O. Hence the equation of motion of the slab is

IO f€ðt Þ + Kf fðt Þ ¼ M ðt Þ (1)

where IO is the moment of inertia of the mass of the plate with respect to O and
Kf the torsional stiffness of the structure. These quantities have been computed
in Example 2.2.1. Thus we have

IO ¼ 3:7164  103 kNm s2 (2)

Kf ¼ 1:9663  105 kNm (3)

rffiffiffiffiffiffi
Kf
w¼ ¼ 7:2738 (4)
IO

2p
T¼ ¼ 0:869s (5)
w

t1 ¼ T ¼ 0:869s (6)
134 PART I Single-degree-of-freedom systems

The motion is studied in two phases.

Phase Ι: 0  t  t1 . Forced vibration
The solution is given by the Duhamel integral
Z t
1
fI ðt Þ ¼ M ðτÞsin wðt  τÞdτ (7)
IO w 0
Taking into account that
M1  M0
M ðt Þ ¼ M0 + t (8)
t1
we have
 Z t Z
1 M1  M0 t
fI ðt Þ ¼ M0 sin wðt  τÞdτ + τ sin wðt  τÞdτ (9)
IO w 0 t1 0

Apparently, the response results as the superposition of a rectangular pulse

and a ramp function pulse. Hence, using Eqs. (3.4.2), (3.4.15), we obtain


1 M0 M1  M0 sin wt
fI ðt Þ ¼ ð1  cos wt Þ + t (10)
IO w w wt1 w
or


M0 M1  M0 sin wt
f I ðt Þ ¼ ð1  cos wt Þ + t (11)
Kf Kf t1 w
which for the adopted data becomes


sin wt
fI ðt Þ ¼ 1:525  103 ð1  cos wt Þ + 1:177  103 t  (12)
w

Phase ΙI: t1 < t. Free vibration

The solution is

f_ I ðt1 Þ
fII ðt Þ ¼ sin wðt  t1 Þ + fI ðt1 Þcos wðt  t1 Þ (13)
w
Eq. (12) for t ¼ t1 gives

fI ðt1 Þ ¼ 1:017  103 and f_ I ðt1 Þ ¼ 0 (14)

and Eq. (13) becomes

fII ¼ 1:017  103 cos wðt  t1 Þ (15)

 Single-degree-of-freedom systems: Forced vibrations Chapter 3 135

The response of the structure is shown in Fig. E3.4.

–3
4 ×10
fmax = 3.594e-3
3

2
f(t)

0
tmax = 0.462

–1 t1

–2
0 0.5 1 1.5 2 2.5 3 3.5 4
t
FIG. E3.4 Response of the structure in Example 3.5.1.

The stress resultants of the columns are computed using Eqs. (4a), 4(b), (6)
of Example 2.2.1. Thus, for the upper right column we have
kx1 ¼ ky1 ¼ 1360:8kN=m, kf1 ¼ 1:9985  103 kNm
x1 ¼ 4:85m, y1 ¼ 2:35m
max Qx1 ¼ kx1 y1 fmax ¼ 11:49kN (16a)
max Qy1 ¼ ky1 x1 fmax ¼ 23:72kN (16b)
h
max Mx1 ¼ max Qx1 ¼ 28:73kNm (16c)
2
h
max My1 ¼ max Qy1 ¼ 59:30kNm (16d)
2
max Mf1 ¼ kf1 fmax ¼ 7:183kNm (16e)

3.5.4 Response to piecewise linear loading

Certain types of loading can be represented by straight-line segments within
generally unequal time intervals, as shown in Fig. 3.5.8. The accelerogram, that
is, the employed recording of the acceleration of the ground motion during an
earthquake, is a representative example of such loading.
136 PART I Single-degree-of-freedom systems

FIG. 3.5.8 Piecewise linear loading.

The loading function in the i interval is given by the expression

pðt Þ ¼ ai + bi t, ti1  t  ti , i ¼ 1, 2, …, n (3.5.20)
where
pi  pi1
ai ¼ pi1 , bi ¼ , to ¼ 0 (3.5.21)
ti  ti1
The response is obtained in n phases, where n is the number of intervals.
The solution in the interval ti1  t  ti is given by Eq. (3.3.23). Hence
we have

u_ ðti1 Þ + u ðti1 Þxw 
ui ðtÞ ¼ sin wD t+ u ðti1 Þcos wD t exwt
wD
Z t (3.5.22)
1 xwðtτÞ 
+ ðai + bi τÞe sin ½wD ðt  τÞdτ
mwD 0
where t¼ t  ti1 , 0  t ti  ti1 .
The Duhamel integral can be evaluated as a sum of two integrals: one due to
the constant term ai of the loading and the other due to the linear term. The inte-
ð1Þ ð2Þ
grals are denoted by Di ðt Þ and Di ðt Þ, respectively.
The first integral is obtained from Eq. (3.4.7) for p0 ¼ ai
" ! #
ð1Þ ai x xwt
Di ðtÞ ¼ 1  
cos wD t + pffiffiffiffiffiffiffiffiffiffiffiffi sin wD t e (3.5.23a)
k 1  x2
The analytical evaluation of the second integral is rather complicated.
Therefore, the recourse to a symbolic language is inevitable. Thus, using
MAPLE, we obtain
("   # )
ð2Þ  bi 2x2  1 2x  2x
Di ð t Þ ¼ sin wD t+ cos wD t exwt + t (3.5.23b)
k wD w w

Substituting Eqs. (3.5.23a), (3.5.23b) into Eq. (3.5.22) gives

 
ui ðtÞ ¼ Ai0 + Ai1 t+ exwt Ai2 cos wD t+ Ai3 sin wD t (3.5.24)
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 137

in which

1 xbi
Ai0 ¼ ai  2
k w
bi
Ai1 ¼ , Ai2 ¼ u ðti1 Þ  Ai0 (3.5.25)
k
1 
A3 ¼
i
u_ ðti1 Þ + xwAi2  Ai1
wD
Differentiating Eq. (3.5.24) with respect to time gives the velocity
   
u_ i ðt Þ ¼ Ai1 + exwt xwAi2 + wD Ai3 cos wD t xwAi3 + wD Ai2 sin wD t
(3.5.26)
The presented method is exact. However, numerical methods are more con-
venient to compute the response to a piecewise linear loading (see Chapter 4).

3.6 Response to a periodic loading

3.6.1 Periodic loads
A periodic load is one whose time variation profile repeats continually at reg-
ular intervals T . It can be represented by a periodic function
pðt Þ ¼ pðt  nT Þ, n ¼ 0,1, 2, … (3.6.1)
The smallest constant T that satisfies Eq. (3.6.1) is called the period of the
periodic function. Fig. 3.6.1 shows examples of periodic functions. Many load-
ings in nature are periodic or can be approximated by periodic loads.

(a)

(b)

(c)
FIG. 3.6.1 Examples of periodic loads.
138 PART I Single-degree-of-freedom systems

3.6.2 Fourier series

The periodic function can be represented by the Fourier series, that is, a
trigonometric series of the form [5]
X
1
pðt Þ ¼ a0 + ðan cos nw0 t + bn sin nw0 t Þ (3.6.2)
n¼1

where an ,bn are constant coefficients to be determined and w0 ¼ T =2p is the

fundamental frequency of the periodic function. The coefficients an , bn ,
n ¼ 0, 1, 2, … are known as the Fourier series coefficients. They can be deter-
mined using the orthogonality property of the sine and cosine functions
presented below.
In general, a set of functions F : ffn ðt Þg is called orthogonal in a interval
½t1 , t2 , if for any two functions fm , fn F, m, n N holds
Z t2 
0 if m 6¼ n
fm ðt Þfn ðt Þdt ¼ (3.6.3)
t1 cn if m ¼ n
where cn is a constant.
The set F of functions fn is complete if no other function outside F exists
that satisfies the orthogonality condition (3.6.3).
It can be readily shown that the set
F ¼ f1, cos w0 t, cos 2w0 t, cos 3w0 t, …sin w0 t, sin 2w0 t, sin 3w0 t, …g (3.6.4)
is orthogonal in ½T =2, T =2 and complete. Indeed, have
Z T =2
1cos mw0 tdt ¼ 0 for each m (3.6.5a)
T =2

Z T =2
1sin mw0 tdt ¼ 0 for each m (3.6.5b)
T =2

Z T =2 
0 if m 6¼ n
cos mw0 t cos nw0 tdt ¼ (3.6.5c)
T =2 T=2 if m ¼ n
Z T =2
cos mw0 t sin nw0 tdt ¼ 0 for each m, n (3.6.5d)
T =2

Z T =2 
0 if m 6¼ n
sin mw0 t sin nw0 tdt ¼ (3.6.5e)
T =2 T =2 if m ¼ n

The coefficients an are evaluated by multiplying Eq. (3.6.2) by cos mw0 t,

integrating over the interval ½T =2, T =2, and using the orthogonality relations
(3.6.5a)–(3.6.5e). Thus, we obtain
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 139

Z T =2
1
a0 ¼ pðt Þdt (3.6.6a)
T T =2

Z T =2
2
an ¼ pðt Þcos nw0 tdt (3.6.6b)
T T =2

Similarly, the coefficients bn are evaluated by multiplying Eq. (3.6.2) by

sin mw0 t, integrating over the interval ½T =2, T =2, and using the orthogonal-
ity relations (3.6.5a)–(3.6.5e). Thus, we obtain
Z T =2
2
bn ¼ pðt Þ sin nw0 tdt (3.6.6c)
T T =2

The series (3.6.2) represents the function pðt Þ, that is, converges to pðt Þ for
n ! 1, provided that it satisfies the following conditions, known as Dirichlet
conditions:
(a) The function pðt Þ has a finite number of discontinuities in one period.
(b) The function pðt Þ has a finite number of maxima and minima in one period.
(c) The function pðt Þ is absolutely integrable over a period, that is,
Z T =2
jpðt Þjdt ¼ k < 1 (3.6.7)
T =2

We shall say that that the function pðt Þ is piecewise continuous in the finite
interval ½T =2, T =2, if it satisfies conditions (a) and (b). At the points of dis-
continuity, for example, point t1 in Fig. 3.6.1c, the Fourier series converges to
the mean value
1     
p t1 + p t1+ (3.6.8)
2
   
where p t1 and p t1+ are the left and right limits of pðt Þ at t1 .
In practice, the periodic function pðt Þ is approximated by a finite number of
terms of the Fourier series, that is, by a finite Fourier series.
Let
X
k
Sk ð t Þ ¼ a 0 + ðan cos nw0 t + bn sin nw0 t Þ (3.6.9)
n¼1

be the sum of the first k + 1 terms of the Fourier series, which will represent the
function pðt Þ in the interval ½T=2, T =2. Then we will have
X
k
pðt Þ ¼ a0 + ðan cosnw0 t + bn sin nw0 t Þ + ek ðt Þ (3.6.10)
n¼1
140 PART I Single-degree-of-freedom systems

where
ek ðt Þ ¼ pðt Þ  Sk ðt Þ (3.6.11)
is the error between pðt Þ and its approximation. The mean square error Ek is
given by
Z
1 T =2
Ek ¼ ½ek ðt Þ2 dt
T T =2
Z " #2 (3.6.12)
1 T =2 Xk
¼ pðt Þ  a0  ðan cos nw0 t + bn sin nw0 t Þ dt
T T =2 n¼1

The error Ek is a function of a0 , an and bn . Therefore, it is minimized when

∂Ek ∂Ek ∂Ek
¼ 0, ¼ 0, ¼ 0, ðn ¼ 1, 2, …, k Þ
∂a0 ∂an ∂bn
Differentiating Eq. (3.6.12) and interchanging differentiation with integra-
tion give
Z " #
∂Ek 2 T=2 Xk
¼ pðt Þ  a0  ðan cos nw0 t + bn sin nw0 t Þ dt (3.6.13a)
∂a0 T T=2 n¼1
Z " #
∂Ek 2 T =2 Xk
¼ pðt Þ  a0  ðan cos nw0 t + bn sin nw0 t Þ cos nw0 tdt
∂an T T =2 n¼1
(3.6.13b)
Z " #
∂Ek 2 T =2 X
k
¼ pðt Þ  a0  ðan cos nw0 t + bn sin nw0 t Þ sin nw0 tdt
∂bn T T =2 n¼1
(3.6.13c)
Using the orthogonality relations (3.6.5a)–(3.6.5e), the integrals (3.6.13a)–
(3.6.13c) become
Z
∂Ek 1 T =2
¼ a0  pðt Þdt ¼ 0 (3.6.14a)
∂a0 T T =2
Z
∂Ek 2 T =2
¼ an  pðt Þcos nw0 tdt ¼ 0 (3.6.14b)
∂an T T =2
Z
∂Ek 2 T =2
¼ bn  pðt Þ sin nw0 tdt ¼ 0 (3.6.14c)
∂bn T T =2
We observe that the values of the coefficients a0 , an , bn obtained by
Eqs. (3.6.14a)–(3.6.14c) are identical to those obtained by Eqs. (3.6.6a)–
(3.6.6c). Therefore, we may conclude that if a function is approximated by a
finite Fourier series, the mean square error is minimized.
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 141

Moreover, when a function is approximated by a finite Fourier series, the

error in the discontinuity region is considerable, even if the number of terms
of the series is very large. This is known as the Gibbs phenomenon. A technique
to overcome this problem is to adapt a suitable function at the region of discon-
tinuity that restores continuity (Fig. 3.6.2a). Similarly, a concentrated load can
be replaced with a bell-shaped function (Fig. 3.6.2b). This, of course, can be
done if the physical problem under consideration is not altered by this substi-
tution. For example, the functions
       
p t1 + p t1+ p t1  p t1+ pðt  t 1 Þ
pðt Þ ¼  sin , jt  t1 j  e (3.6.15a)
2 2 2e

P pðt  t1 Þ
pðt Þ ¼ 1 + cos (3.6.15b)
2e 2e
where e is a small number can play this role.

(a) (b)
FIG. 3.6.2 Functions pðt Þ replacing pðt Þ in the interval ½t1  e, t1 + e.

Example 3.6.1 Fourier series expansion of a periodic function. Gibbs

phenomenon
Expand the periodic function of Fig. 3.6.1b in Fourier series if t1 ¼ T =2.
Solution
The function pðt Þ within a period is defined as

2p0 t=T if 0  t < T =2
pðt Þ ¼ (1)
0 if T=2 < t  T
We readily prove that the Dirichlet conditions are satisfied. Hence, the func-
tion can be expanded in Fourier series. Eqs. (3.6.6a)–(3.6.6c) give
Z
1 T 1
a0 ¼ pðt Þdt ¼ p0 (2a)
T 0 4
Z
2 T cos np  1
an ¼ pðt Þcos nw0 tdt ¼ p0 (2b)
T 0 n 2 p2
142 PART I Single-degree-of-freedom systems

Z T
2 cos np
bn ¼ pðt Þ sin nw0 t ¼ p0 (2c)
T 0 np
Hence
p0 p0 X
N
cos np  1 p0 X
N
 cos np
pðt Þ ¼ + 2 cos nw 0 t + sin nw0 t (3)
4 p n¼1 n2 p n¼1 n
Fig. E3.5 shows the graphical representation of the finite Fourier series in
Eq. (3) for various values of N with p0 ¼ 1, T ¼ 1. We observe that the conver-
gence is very slow at point t ¼ T =2 due to the Gibbs phenomenon.

N = 20 N = 100
1.2 1.2

1 1

0.8 0.8

0.6 0.6
p(t)

p(t)

0.4 0.4

0.2 0.2

0 0

–0.2 –0.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
t t

N = 1000 N = 2000
1.2

1 1

0.8 0.8

0.6 0.6
p(t)

p(t)

0.4 0.4

0.2 0.2

0 0

–0.2 –0.2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
t t

FIG. E3.5 Gibbs’s phenomenon in Example 3.6.1.

Example 3.6.2 Fourier series expansion of a periodic load

Expand the periodic load of Fig. E3.6 in Fourier series if t1 ¼ T =3.

FIG. E3.6 Periodic load in Example 3.6.2.

 Single-degree-of-freedom systems: Forced vibrations Chapter 3 143

Solution
The function pðt Þ within a period T ¼ 3t1 can be represented using the delta
function.a Thus we obtain
pðt Þ ¼ P ½d ðt  T =3Þ  d ðt  2T =3Þ (1)
Eqs. (3.6.6a)–(3.6.6c) give
Z
1 T
a0 ¼ pðt Þdt ¼ 0 (2a)
T 0
Z 


2 T 2P 2np 4np
an ¼ pðt Þcos nw0 tdt ¼ cos  cos (2b)
T 0 T 3 3
Z 


2 T 2P 2np 4np
bn ¼ pðt Þ sin nw0 t ¼ sin  sin (2c)
T 0 T 3 3
Inserting these values of the coefficients in Eq. (3.6.2) we obtain
(
1 


2P X 2np 4np
pðt Þ ¼ cos  cos cos nw0 t
T n¼1 3 3


 )
2np 4np
+ sin  sin sin nw0 t (3)
3 3

3.6.3 Response of the SDOF system to periodic excitation

As was shown in the previous section, any periodic function can be expressed as
a sum of harmonic functions. Theoretically, an infinite number of harmonic
terms are required to represent accurately the function as a Fourier series. How-
ever, the contribution of the terms decreases with increasing n. Consequently, a
few terms are adequate to approximate the loading with acceptable accuracy.
For periodic loading, each term of the series represents harmonic excitation.
Thus, the total response of an SDOF system to a periodic loading can be
obtained as the sum of the responses of the individual harmonic terms, provided
that the system is linear, which allows the application of the superposition prin-
ciple. In subsequent analysis, we assume that the periodic excitation acts for a
sufficiently long time, by which the transient response due to the initial condi-
tions has decayed. Thus, only the steady-state response will be considered.
The steady-state response to the constant load pðt Þ ¼ p0 can be obtained
from Eq. (3.2.25a) by setting p0 ¼ a0 , x ¼ 0, w ¼ 0 and neglecting the transient
response. This yields
ao
u 0 ðt Þ ¼ (3.6.16a)
k

a. The delta function d ðt  t0 Þ and its properties are discussed in Section 3.7.1.
144 PART I Single-degree-of-freedom systems

For the sine term bn sin nw0 t the steady-state response results from
Eq. (3.2.25a) by setting p0 ¼ bn , w ¼ nw0 , and b ¼ b n ð¼ nw0 =wÞ
bn 1  
unsin ðt Þ ¼  2 1  b2n sin nw0 t  2xbn cos nw0 t
k 1b 2
+ ð2xbn Þ 2
n
(3.6.16b)
while for the cosine term an cos nw0 t the steady-state response is obtained from
Eq. (3.2.25b) by setting p0 ¼ an , w ¼ nw0 , and b ¼ b n ð¼ nw0 =wÞ
an 1   
uncos ðt Þ ¼   2xbn sin nw0 t + 1  b 2n cos nw0 t
k 1  b 2 2 + ð2xb Þ2
n n
(3.6.16c)
The steady-state response of the damped system to periodic loading results
as the superposition of responses to individual terms of the Fourier series
X
n X
n
u ðt Þ ¼ u0 ðt Þ + unsin ðt Þ + uncos ðt Þ (3.6.17)
n¼1 n¼1

Inserting Eqs. (3.6.16a)–(3.6.16c) into Eq. (3.6.17) gives

ao X 1
an 1   
u ðt Þ ¼ +  2 2xbn sin nw0 t + 1  b2n cos nw0 t
k n¼1
k 1b 2
+ ð2xb n Þ 2
n
X
1
bn 1  
+   1  b2n sin nw0 t  2xbn cosnw0 t
k 1b 2 2
n¼1 n + ð2xb n Þ2
(3.6.18)
or
ao X 1
1 1   
u ðt Þ ¼ +   2
an ð2xbn Þ + bn 1  b2n sin nw0 t
k n¼1
k 1  b 2 + ð2xb Þ2
n n
   
+ an 1  b n  bn ð2xbn Þ cos nw0 t
2

(3.6.19)
Theoretically, there is no transient response when x ¼ 0. However, for
small values of damping, which is the usual case in our structures, the
steady-state response can be obtained from Eqs. (3.6.16b), (3.6.16c) for
x  0. This yields
bn 1
unsin ðt Þ    sin nw0 t (3.6.20a)
k 1  b2n
an 1
uncos ðt Þ    cos nw0 t (3.6.20b)
k 1  b 2n
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 145

which are actually approximations of the particular solutions (3.2.7) and

(3.2.18) of the undamped motion due to harmonic loading.
On the basis of Eqs. (3.6.20a), (3.6.20b), (3.6.16a), the approximate steady-
state response becomes
ao X 1
an 1 X1
bn 1
u ðt Þ  + cos nw0 t + sin nw0 t (3.6.21)
k n¼1
k 1  b 2
n n¼1
k 1  b 2n

The response is a periodic function with a period T ¼ 2p=w0 . The contribu-

tion of each harmonic term in the series (3.6.19) depends on: (i) the amplitudes
an and bn of the harmonic components of the load, and (ii) the value of b n . The
terms with b n near 1 have the greatest influence on the response. It should be
noted that in the case bn ¼ 1, x ¼ 0 is meaningless.
Example 3.6.3 Response of an SDOF system subjected to periodic loading
Determine the response of a SDOF system subjected to the periodic loading of
Fig. 3.6.1b. Adopt t1 ¼ T =2, x ¼ 0, T ¼ 1, b ¼ 1:25.
Solution
The function pðt Þ within a period is defined as

2p0 t if 0 < t < 1=2
pðt Þ ¼ (1)
0 if 1=2 < t  1
The Fourier series coefficients are obtained using Eqs. (3.6.6a)–(3.6.6c).
They have already been evaluated in Example 3.6.1. Then the response to
the periodic loading is obtained from Eq. (3.6.19). Fig. E3.7 shows the response
ratio Rðt Þ ¼ u ðt Þ=ðp0 =k Þ (a) for different numbers of series terms and (b) for
n ¼ 50 and various values of damping ratio x. We observe that a small number
of terms gives an acceptable accuracy of the response while the response for
small values of x is very close to that with x ¼ 0, which justifies the
expression (3.6.21).

(a) (b)
FIG. E3.7 Response to periodic loading in Example 3.6.3.
146 PART I Single-degree-of-freedom systems

3.7 Response to unit impulse

3.7.1 The delta function or Dirac’s delta function
In problems of mechanics, we often come across concentrated loads, that is,
actions that are applied to a very small region, theoretically at a point in space
or an instant in time. In this section, we present a mathematical tool, the delta
function or Dirac’s delta function, that allows us to handle impulsive loads in a
simple way and understand better the dynamic response of SDOF systems to
such loads and in general to arbitrary dynamic loads.
Let P ¼ 1 be a unit force that acts on the SDOF system at time t ¼ 0. First,
we consider that the force is not imposed instantaneously but it is distributed
over a small time interval ½e, +e and let pðt Þ denote its distribution. This dis-
tribution, which most probably has the shape shown in Fig. 3.7.1, is not known.
Nevertheless, we know that it is sufficiently “concentrated” about t ¼ 0 and that
Z +1
pðt Þdt ¼ 1 (3.7.1)
1

FIG. 3.7.1 Distribution of a concentrated impulsive force over the interval ½e, +e.

namely, the total force is equal to unity. If we skip the problem of determining
analytically the function pðt Þ, we may assume a priori a prescribed shape for
this function, for example,

k=2, jt j < 1=k
pk ðt Þ ¼ (3.7.2a)
0, jt j  1=k
or
k
pk ðt Þ ¼ (3.7.2b)
pð1 + k 2 t 2 Þ
where k is a positive number.
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 147

Fig. 3.7.2a and b show both functions pk defined in Eqs. (3.7.2a), (3.7.2b),
respectively. Moreover, they satisfy Eq. (3.7.1), which means that they are
equivalent to pðt Þ and can represent the actual force P.

(a) (b)
FIG. 3.7.2 Functions pk sufficiently concentrated for large values of k.

The distribution of pðt Þ becomes more “concentrated” as the value of k in

Eqs. (3.7.2a), (3.7.2b) increases. The limiting case for k ! 1 yields a fictitious
distribution per unit time, which is denoted by d ðt Þ and defined by
d ðt Þ ¼ ‘im pk ðt Þ (3.7.3)
k!1

The function d ðt Þ is known as the delta function or Dirac’s delta function.

When the concentrated force acts at instant t ¼ t1 , the delta function is denoted
by dðt  t1 Þ. In mathematics, the delta function is treated in the theory of gen-
eralized functions [6, 7]. We give below the definition of the delta function as a
generalized function, and we mention some of its properties [8, 9].
For a point source applied at t ¼ 0, the one-dimensional delta function is
defined by the relation
Z +1
d ðt Þf ðt Þ dx ¼ f ð0Þ (3.7.4)
1

or by the relation
Z +1
dðt  t1 Þf ðt Þ dt ¼ f ðt1 Þ (3.7.5)
1

for a point source applied at t ¼ t1 . The function f ðt Þ is continuous in a finite

interval containing the source point t ¼ 0 or t ¼ t1 . The one-dimensional delta
function can also be described by the relations

0, t 6¼ 0
d ðt Þ ¼ (3.7.6a)
1, t ¼ 0
148 PART I Single-degree-of-freedom systems

and
Z +1 Z e
d ðt Þdt ¼ d ðt Þdt ¼ 1 (3.7.6b)
1 e

where e is a small positive number. According to this definition, the function

dðt Þ has zero value everywhere except at t ¼ 0, where it becomes infinite,
and satisfies Eq. (3.7.6b).
Eq. (3.7.4) may be obtained using the mean value theorem of integral cal-
culus. Thus, referring to Fig. 3.7.2a and choosing e ¼ 1=k, we obtain
Z 1 Z +e
 
1
f ðt Þd ðt Þdt ¼ ‘im f ðt Þpk ðt Þdt ¼ ‘im f ðt Þ
∗ 2e
1 e!0 e e!0 2e
(3.7.7)
¼ ‘im ½f ðt ∗ Þ ¼ f ð0Þ
e!0

where t* is a value of t in the interval (e, þe)

Moreover, using integration by parts we can show
Z b
d m d ðt  t1 Þ m d f ðt1 Þ
m
f ðx Þ dt ¼ ð 1 Þ , a < t1 < b (3.7.8)
a dt m dt m

3.7.2 Response to unit impulse

In dynamics, loads pðt Þ distributed over a very short interval ½e, +e are
referred to as impulsive loads (see Section 3.5). The integral of the distribution
of the impulsive load over the interval it acts is defined as the impulse of the load
and is denoted by I , namely
Z +e
I¼ pðt Þdt (3.7.9)
e

If the impulsive load P is concentrated and acts at instant t ¼ t1 , it can be

denoted by
pðt Þ ¼ Pdðt  t1 Þ (3.7.10)
Obviously, Eq. (3.7.5) for P ¼ 1 gives
Z +e
I¼ Pdðt  t1 Þdt ¼ 1 (3.7.11)
e

From Newton’s second law of motion, we have (see Appendix)

d ðm u_ Þ
¼ pðt Þ (3.7.12a)
dt
or if the mass is constant
d u_
m ¼ pðt Þ (3.7.12b)
dt
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 149

which after integration over the interval ½e, +e gives

Z +e
+e
m ½u_ e ¼ pðt Þdt (3.7.13)
e

or for e ! 0
mDu_ ¼ 1 (3.7.14)
where
   
Du_ ¼ u_ t1+  u_ t1 (3.7.15)
Hence the impulsive load produces an abrupt change (discontinuity) of the
 t1 . If the system is at rest before the action of the impulsive load,
velocity at time
then it is u_ t1 ¼ 0 and Eq. (3.7.14) gives
1
u_ ðt1 Þ ¼ (3.7.16)
m
where t1 designates t1+ . Therefore, Eq. (3.7.16) presents the initial velocity
given to the system by the impulsive load. However, the displacement remains
continuous, which means that
   
Du ¼ u t1+  u t1 ¼ 0 (3.7.17)
If the elastic and damping forces are taken into account, then Eq. (3.7.12b)
is written
d u_ du
m ¼ pðt Þ  c  ku (3.7.18)
dt dt
which after integration over the interval ½e, +e gives
Z +e Z +e
+e +e
m ½u_ e ¼ pðt Þdt  c½u e  k u ðt Þdt (3.7.19)
e e

Applying the mean value theorem of integral calculus to the integral of elas-
tic force, we write Eq. (3.7.19) as
Z +e
+e +e
m ½u_ e ¼ pðt Þdt  c½u e  ku ðt ∗ Þ2e,  e < t ∗ < e (3.7.20)
e

which by virtue of Eqs. (3.7.15), (3.7.17) becomes if e ! 0

mDu_ ¼ 1 (3.7.21)
This means that the elastic and damping forces do not influence the change
of the velocity
  when
  an impulsive load is applied.
If u t1 ¼ u_ t1 ¼ 0, then Eqs. (3.7.16), (3.7.17) imply that the unit
impulse produces free vibrations with initial conditions u ðt1 Þ ¼ 0,
u_ ðt1 Þ ¼ 1=m, where t1 designates t1+ .
150 PART I Single-degree-of-freedom systems

h(t-t)

t =1

t
FIG. 3.7.3 Response to unit impulse.

If τ denotes the instant t1 and h ðt  τÞ u ðt Þ the produced displacement,

then we obtain from Eq. (2.3.18)
1 xwðtτÞ
h ðt  τ Þ ¼ e sin ½wD ðt  τÞ, t  τ (3.7.22a)
mwD
Obviously, for the undamped system, x ¼ 0, it is
1
h ðt  τÞ ¼ sin ½wðt  τÞ, t  τ (3.7.22b)
mw
Eqs. (3.7.22a), (3.7.22b) express the response of the SDOF system to a unit
impulse acting at time τ. Note that h ðt  τÞ ¼ 0 if t < τ. Fig. 3.7.3 shows the
response to a unit impulse acting at t1 ¼ τ ¼ 1.
The derivation of the response function h ðt  τÞ presented previously was
achieved by physical consideration. Mathematically, it expresses the solution
of the equation of motion under the external loading pðt Þ ¼ d ðt  τÞ, that is,

1
u€ + 2xwu_ + w2 u ¼ d ðt  τ Þ (3.7.23)
m
The solution of the above equation can be obtained from Eq. (3.3.23) for
zero initial conditions and pðt Þ ¼ dðt  t1 Þ. Thus, we have
Z t
1
u ðt Þ ¼ dðτ  t1 ÞexwðtτÞ sin ½wD ðt  τÞdτ (3.7.24)
mwD 0
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 151

which by virtue of Eq. (3.7.5) gives

1 xwðtt1 Þ
u ð t Þ h ð t  t1 Þ ¼ e sin ½wD ðt  t1 Þ, t  t1 (3.7.25)
mwD
or by setting t1 ¼ τ, we recover Eq. (3.7.22a).
Eqs. (3.7.22a), (3.7.22b) represent the Green’s function of the equation of
motion of the damped and undamped SDOF systems, respectively.

3.7.3 Response to arbitrary loading

The arbitrary loading pðt Þ can be visualized as a sequence of pulse loads of
infinitesimal duration with magnitude P ¼ pðτÞdτ in the interval ½ 0, t . The
response of the system to this pulse load is
du ¼ ½pðτÞdτh ðt  τÞ (3.7.26)
Apparently, the response of the system at time t is the sum of all infinites-
imal responses from t ¼ 0 τ ¼ t. This sum is expressed by the integral
Z t Z t
u ðt Þ ¼ du ¼ pðτÞh ðt  τÞdτ (3.7.27)
0 0

The integral (3.7.27) is known as the convolution integral of the functions

pðt Þ and h ðt Þ, to which we referred as the Duhamel integral in Section 3.2.2.

3.7.4 The reciprocal theorem in dynamics

In statics, the reciprocity is expressed by Betti’s theorem, also known as the
Maxwell-Betti reciprocal work theorem, introduced by Enrico Betti in 1872.
This theorem is valid for systems with linear behavior and reads
If two sets of loads FI and FII act separately on a linearly elastic structure, the
work WI , II done by the first set of loads in acting through the displacements UII
produced by the second set of loads is equal to the work WII , I done by the second
set of loads in acting through the displacements UI produced by the first set of
loads, namely

WI , II ¼ WII , I (3.7.28)
In dynamics of linear systems, the reciprocity is expressed by the dynamic
reciprocal theorem [10], known also as the dynamic Betti-Rayleigh theorem.
This theorem for SDOF systems reads
If two loadings pI ðt Þ and pII ðt Þ act separately on a linear dynamic system and
produce the responses uI ðt Þ and uII ðt Þ then the convolution CI , II of the loading
pI ðt Þ with the response uII ðt Þ is equal to the convolution CII , I of the loading pII ðt Þ
with the response uI ðt Þ
152 PART I Single-degree-of-freedom systems

CI , II ¼ CII , I (3.7.29)
or using definition (3.3.11) for the convolution, we may write
pI ðt Þ∗ uII ðt Þ ¼ pII ðt Þ∗ uI ðt Þ (3.7.30a)
or in integral form
Z t Z t
pI ðτÞuII ðt  τÞdτ ¼ pII ðτÞuI ðt  τÞdτ (3.7.30b)
0 0

The dynamic reciprocal theorem is readily proved by taking into account

that the convolution satisfies the following properties [8]:
ðiÞ f ðt Þ ∗g ðt Þ ¼ g ðt Þ ∗f ðt Þ (3.7.31a)
ðiiÞ f ðt Þ∗ ½g ðt Þ∗ q ðt Þ ¼ ½g ðt Þ ∗f ðt Þ∗ q ðt Þ (3.7.31b)
where f ðt Þ, g ðt Þ, and q ðt Þ are arbitrary functions.
If h ðt Þ represents the response to the unit impulse, Eqs. (3.7.22a), (3.7.22b),
we may write on the basis of Eq. (3.7.25)
uI ðt Þ ¼ pI ðt Þ ∗h ðt Þ (3.7.32a)
uII ðt Þ ¼ pII ðt Þ∗ h ðt Þ (3.7.32b)
We may further write
pI ðt Þ ∗uII ðt Þ ¼ pI ðt Þ∗ ½pII ðt Þ∗ h ðt Þ
¼ pI ðt Þ∗ ½h ðt Þ ∗pII ðt Þ
¼ ½pI ðt Þ ∗h ðt Þ∗ pII ðt Þ (3.7.33)
¼ uI ðt Þ ∗pII ðt Þ
¼ pII ðt Þ∗ uI ðt Þ
which proves the dynamic reciprocal theorem.
The dynamic reciprocal theorem has many applications in mechanics, for
example, moving loads, the boundary integral equation method for dynamic
problems [10], etc.

3.8 Problems
Problem P3.1 A machine carrying a mass m0 is placed on the slab of the build-
ing of Fig. P3.1. The mass rotates eccentrically about the point ð2:5, 2:0Þ at a
distance s ¼ 1:0m with a frequency f ¼ 4Hz. The columns are massless, inex-
tensible, and fixed on the base while the slab is assumed uniform and rigid.
Determine the time history of the shear forces Qx , Qy , the bending moments
Mx , My , and the torsion moment Mw at the top of the columns. The material
constants are E ¼ 1:2  107 kN=m2 , n ¼ 0:2. The total load of the plate (dead
plus live) is q ¼ 12kN=m2 ; m0 ¼ m=5, where m is the total mass of the slab.
The acceleration of gravity is g ¼ 9:81m=s2 .
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 153

FIG. P3.1 Building in problem P3.1

Problem P3.2 A SDOF system with parameters m, k, x is subjected to

harmonic loading p ¼ p0 cos wt.  Determine the expression of the steady-state
response and show that its amplitude is the same with that produced by the load-

ing p ¼ p0 sin wt.
Problem P3.3 An undamped SDOF system is subjected to the loading
 When w ¼ w1 , the system is set in resonance. Subsequently, a mass
p ¼ p0 sin wt.
Dm ¼ 0:20kN m1 s is added and the system is set in resonance, when
w ¼ 0:6w1 . Determine the mass m of the system.
Problem P3.4 The one-story building of Fig. P3.4a is supported by three col-
umns placed at the vertices of an equilateral triangle. The slab is rigid and the
columns are fixed at both ends and have a rectangular cross-section, a modulus
of elasticity E, and negligible mass. The slab is subjected to the load
 for 0  t  t1 and pðt Þ ¼ 0 for t1 < t (Fig. P3.4b) acting in
pðt Þ ¼ p0 cos 3 wt

(a) (b)
FIG. P3.4 One-story building in problem P3.4
154 PART I Single-degree-of-freedom systems

the x direction. Plot the response ratio Rðt Þ of the structure and the stress resul-
tants at the base of the columns 1 and 3 using the following data: Height of col-
umns a ¼ 4m, cross-section of columns a=10  a=20, side length of the
triangular slab a, p0 ¼ 10kN, E ¼ 2:1  107 kN=m2 , load of the plate, including
the dead load, q ¼ 20kN=m2 , t1 ¼ p=2w,  x ¼ 0:05, w ¼ 8s1 , acceleration of
gravity g ¼ 9:81m=s . 2

Problem P3.5 The structure of Fig. P3.5a consists of the rigid girder BC and
the two flexible columns AB and CD having a cross-sectional moment of inertia
I and a modulus of elasticity E. The cables FB and GC have cross-sectional area
A, cannot withstand compression, and are assumed massless. The structure is sub-
jected to the impulsive loads pðt Þ shown in Fig. P3.5b and c. Study the response of
the structure and determine the maximum error when the impulsive loads are
substituted by equivalent concentrated forces. Plot the function D ðt1 =T Þ for the
two load cases. For which value of the ratio t1 =T is the maximum error less than
2%? Data: a ¼ 1:5m, I ¼ 33,740cm4 (IPE450), E ¼ 2:1  108 kN=m2 ,
A ¼ 3cm2 , t1 ¼ 0:1, m  ¼ 1:0kNm1 s2 =m, p10 ¼ 10kN. The value p20 is deter-
mined so that both loads have the same impulse.

(a)

(b) (c)
FIG. P3.5 Structure in problem P3.5
 Single-degree-of-freedom systems: Forced vibrations Chapter 3 155

Problem P3.6 Consider the structure of Fig. P3.6. The rigid column AC of circu-
lar cross-section and mass per unit length m is supported by the three elastic cables
of cross-sectional area A and modulus of elasticity E. The support on the ground is
a spherical hinge. The cables have been prestressed so that they can withstand com-
pression. Three advertising panels are massless fixed at the top of the columns, as
shown in the figure. The structure is subjected to the wind blast load of Fig. P3.6c in
the y direction (see Fig. P3.6b). Determine the minimum prestress force of the
cables. The cables and the panels have negligible mass. Use the data a ¼ 5:0m,
p0 ¼ 4kN=m2 , m  ¼ 0:5kNm1 s2 =m, E ¼ 2:1  108 kN=m2 , A ¼ 4cm2 ,
t1 ¼ 0:1s.

(a) (b)

(c)
FIG. P3.6 Structure in problem P3.6
156 PART I Single-degree-of-freedom systems

Problem P3.7 The one-story building in Example 2.2.1 is subjected to a blast

pressure in the x direction. The time variation of the blast pressure is approx-
imated as shown in Fig. P3.7. Determine the stress resultants max Qx and
max Mx of the columns. The peak positive pressure is p0 ¼ 10kN and the
peak negative pressure is p1 ¼ 0:2p0 ; t1 ¼ 0:2s and t2 ¼ 5t1 . The system is
at rest at t ¼ 0.

10

p
0
4

2
t1 t2
0
p1

–2

0 0.5 1 1.5 2
FIG. P3.7 Blast pressure in problem P3.7

Problem P3.8 Show that an impulsive load generated by the load pðt Þ
and acting over the interval ½t1 , t2  can be represented as pI ¼ ½H ðt  t1 Þ
H ðt  t2 Þpðt Þ, where H ðt  ti Þ is the Heaviside step function. Write a
MATLAB program that constructs the impulsive load.
Problem P3.9 Determine the dynamic response of a SDOF system subjected to
the sine periodic loading of Fig. P3.9. Assume: T ¼ p=w,  w ¼ 1:1w,
 x ¼ 0, and
x ¼ 0:1.

FIG. P3.9 Periodic load in problem P3.9

 Single-degree-of-freedom systems: Forced vibrations Chapter 3 157

Problem P3.10 Determine the dynamic response of a SDOF system subjected

 T ¼ p=w,
to the periodic loading of Fig. P3.10. Assume w ¼ 1:1w,  x ¼ 0,
and x ¼ 0:1.

FIG. P3.10 Periodic load in problem P3.10

References and further reading

[1] F.B. Hildebrand, Advanced Calculus for Applications, Prentice Hall, Englewood Cliffs, New
Jersey, 1962.
[2] E. Kreyszig, Advanced Engineering Mathematics, fourth ed., John Wiley & Sons, New York,
1979.
[3] R.V. Churchill, Operational Mathematics, second ed., MacGraw-Hill Book Company, Inc.,
New York, 1958.
[4] F. Oberhettinger, L. Badii, Tables of Laplace Transforms, Springer-Verlag, Berlin/Heidelberg/
New York, 1973.
[5] H.P. Hsu, Fourier Analysis, Simon and Schuster, New York, 1967.
[6] G.F. Roach, Green’s Functions, Van Nostrand Reinhold Company, London, 1970.
[7] Μ. Greenberg, Application of Green’s Functions in Science and Engineering, Prentice Hall,
Englewood Cliff, NJ, 1971.
[8] T. Myint-U, L. Debnath, Linear Partial Differential Equations for Scientists and Engineers,
fourth ed., Birkh€auser, Boston, 2007.
[9] J.T. Katsikadelis, The Boundary Element Method for Engineers and Scientists, Academic
Press, Elsevier, Oxford, UK, 2016.
[10] J. Dominguez, Boundary Elements in Dynamics, Computational Mechanics Publications,
Southampton, Boston, 1993.
Chapter 4

Numerical integration of the

equation of motion
Chapter outline
4.1 Introduction 159 4.5.4 Stability of the central
4.2 The central difference method 160 difference method 180
4.3 The average acceleration 4.5.5 Stability of the average
method 164 acceleration method 182
4.4 The analog equation method 169 4.5.6 Stability of the analog
4.5 Stability of the numerical equation method 185
integration methods 174 4.6 Accuracy of the numerical
4.5.1 Errors in the numerical integration 185
integration 174 4.7 Problems 188
4.5.2 Difference equations 175 References and further reading 189
4.5.3 Difference equations and
stability of the numerical
integration methods 178

4.1 Introduction
The previous analysis shows that an analytical solution of the equation of
motion for an single-degree-of-freedom (SDOF) system is possible only if
the external force is described by a simple function. If the excitation force varies
arbitrarily with time or is given by a set of its values, an analytical solution is out
of the question. However, such problems can be tackled numerically by time
step integration methods for differential equations. The literature about these
methods is vast. Extensive chapters and whole books cover this subject. They
present the mathematical development of these methods, their computer imple-
mentation, and their accuracy, convergence, and stability. Several computer
packages include ready-to-use subroutines for the solution of the differential
equation of motion.
Some of these methods have been specially developed for the study of the
dynamic response of systems. A survey of these methods is given in [1, 2]. The
central difference method (CDM), Houbolt’s method, Wilson’s q-Method, and
Newmark’s method are the most well known among them [3, 4]. Nevertheless,
with the increase of cheap computer power, some of them have lost their impor-
tance while others have taken dominating places in the computational arena.

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00004-2

© 2020 Elsevier Inc. All rights reserved. 159
160 PART I Single-degree-of-freedom systems

These methods determine the displacement u ðt Þ by numerical integration of

the equation of motion
m u€ + cu_ + ku ¼ pðt Þ (4.1.1)
The displacement u ðt Þ is computed step by step. The numerical methods
start from the values u ð0Þ and u_ ð0Þ, which are known at t ¼ 0, and subsequently
marching in the integration time, these quantities are computed at instants
Dt, 2Dt, 3Dt,… where Dt denotes a small time interval. The development of
these methods is based on two assumptions. The first assumption is that the equation
of motion (4.1.1) is satisfied at discrete time instants Dt apart, contrary to the ana-
lytical methods in which the equation is satisfied at any instant t. This means that the
equilibrium of all forces, namely inertial, damping, elastic, and external, is satisfied
only at discrete instants within the time interval of the solution. The second assump-
tion is that the parameters of the solution, namely the displacement, velocity, and
acceleration, vary within the time interval Dt according to a prescribed law. The
accuracy, stability, and efficiency of the solution method depend on the numerical
integration scheme as well as on the magnitude of Dt. In the following sections, only
some of these methods, which are usually employed for the study of the dynamic
response of systems, will be briefly discussed. We confine our presentation to the
central difference method (CDM), Newmark’s average acceleration method
(AAM) [5], and the analog equation method (AEM), a new efficient method based
on the principle of the analog equation [6].
The presentation that follows is intended to discuss only the basic concepts
underlying these methods and provide a few computational algorithms together
with the computer programs based on them. While this material would be suf-
ficient for many practical problems and research applications, the reader should
recognize that a wealth of knowledge exists on the subject.

4.2 The central difference method

In this method, the values of u ðt + Dt Þ and u ðt  Dt Þ of the function u ðt Þ at
instants t + Dt and t  Dt (Fig. 4.2.1) are approximated by their Taylor series
1 1
u ðt + Dt Þ ¼ u ðt Þ + Dt u_ ðt Þ + Dt 2 u€ðt Þ + Dt 3 u___ðt Þ + ⋯ (4.2.1)
2 6
1 1
u ðt  Dt Þ ¼ u ðt Þ  Dt u_ ðt Þ + Dt 2 u€ðt Þ  Dt 3 u
___ðt Þ + ⋯ (4.2.2)
2 6
Subtracting Eq. (4.2.2) from Eq. (4.2.1) yields
2
u ðt + Dt Þ  u ðt  Dt Þ ¼ 2Dt u_ ðt Þ + Dt 3 u
___ðt Þ + ⋯ (4.2.3)
6
For small values of Dt, the terms of order higher than two can be neglected
and Eq. (4.2.3) yields the following relation for the approximation of the first
derivative at time t
u ðt + Dt Þ  u ðt  Dt Þ
u_ ðt Þ  (4.2.4)
2Dt
 Numerical integration of the equation of motion Chapter 4 161

FIG. 4.2.1 Discretization of the interval ½0, T  into N equal intervals h ¼ T =N .

Further, adding Eqs. (4.2.1), (4.2.2) and neglecting the terms of order higher
than three, we obtain the following expression to approximate the second deriv-
ative of u ðt Þ at time t

u ðt + Dt Þ  2u ðt Þ + u ðt  Dt Þ
u€ðt Þ  (4.2.5)
Dt 2
Substitution of the derivatives u_ ðt Þ and u€ðt Þ into Eq. (4.1.1) with their
approximations (4.2.4) and (4.2.5) gives

u ðt + Dt Þ  2u ðt Þ + u ðt  Dt Þ u ðt + Dt Þ  u ðt  Dt Þ
m +c + ku ðt Þ ¼ pðt Þ
Dt 2 2Dt
which is solved for u ðt + Dt Þ to yield

m  
c  2m
m c 
+ u ð t + Dt Þ ¼ p ð t Þ  k  u ðt Þ   u ðt  Dt Þ
Dt 2 2Dt Dt 2 Dt 2 2Dt
(4.2.6)

The previous equation may be written as

K^ u ðt + Dt Þ ¼ P^ (4.2.7)
where
m c
K^ ¼ 2 + (4.2.8)
Dt 2Dt
 

2m m c 
P^ ¼ pðt Þ  k  2 u ðt Þ   u ðt  Dt Þ (4.2.9)
Dt Dt 2 2Dt

The quantities K^ and P^ are referred to as the effective stiffness and the
effective load, respectively. Obviously, Eq. (4.2.7) allows the evaluation of
the displacement at instant t + Dt, if the displacements at the two preceding
instants t and t  Dt are known. Because u ð0Þ is known from the initial
162 PART I Single-degree-of-freedom systems

conditions, the procedure starts at t ¼ Dt. Obviously, this requires the value of
u ðDt Þ, which is unknown in the first instance, but it can be determined from
Eq. (4.2.2) for t ¼ 0. Thus neglecting terms of order higher than two, we have
1
u ðDt Þ  u ð0Þ  Dt u_ ð0Þ + Dt 2 u€ð0Þ (4.2.10)
2
In the above equation, the quantities u ð0Þ and u_ ð0Þ are known from the ini-
tial conditions while u€ð0Þ can be computed from the equation of motion,
Eq. (4.1.1), for t ¼ 0. Thus, we obtain
u€ð0Þ ¼ ½pð0Þ  cu_ ð0Þ  ku ð0Þ=m (4.2.11)
The stability of the CDM requires that time step Dt is less than a certain
critical value, that is, qffiffiffiffiffiffiffiffiffiffiffiffi
Dt  Dtcr ¼ T 1  x2 =p (4.2.12)

where T is the period of the system (see Eq. 4.5.30). Otherwise, the procedure
“blows up” with time and the solution makes no sense. This is discussed in
Section 4.5.
Because T is usually a small number, Dt should be small, which implies
that a large number of time steps are required to solve the equation of motion.
This has been a major drawback of the method, especially in older times when
the computer capabilities in terms of memory and speed were restricted. This
fact has led researchers to develop integration methods in which the size of
the time step is not restricted by a critical value. Table 4.2.1 presents the

TABLE 4.2.1 Central difference method (CDM).

A. Data
w (or k), m, x, u0 , u_ 0 , ttot , pðt Þ
B. Initial computations
pffiffiffiffiffiffiffiffiffi
B.1. k ¼ mw2 (or w ¼ k=m ), c ¼ 2mwx, T ¼ 2p=w, ust ¼ pmax =k
B.2. u€0 ¼ ðp0  cu_ 0  ku 0 Þ=m
pffiffiffiffiffiffiffiffiffiffiffiffi
B.3. Choose Dt < Dtcr ¼ T 1  x2 =p and compute
B.4. ao ¼ 1=Dt 2 , a1 ¼ 1=2Dt, a2 ¼ 2ao , a3 ¼ 1=a2
B.5. uDt ¼ u0  Dt u_ 0 + a3 u€0 , K^ ¼ ao m + a1 c
C. For each time step compute:
C.1. P^t ¼ pt  ðk  a2 m Þut  ðao m  a1 cÞutDt
C.2. ut + Dt ¼ P^t =K^ , Rt + Dt ¼ ut + Dt =ust
C.3. u_ t ¼ a1 ðut + Dt  utDt Þ, u€ðt Þ ¼ ao ðut + Dt  2ut + utDt Þ
C.4. Increase time t ¼ t + D and check: If t  ttot end. Else set utDt ¼ ut , ut ¼ ut + Dt
and go to C.1
 Numerical integration of the equation of motion Chapter 4 163

algorithm for the numerical implementation of the CDM in pseudocode-type

notation so that the reader can write a computer code in the language of his/
her preference.
Adhering to the steps of Table 4.2.1, a computer program called centr_
diff_lin.m has been written in MATLAB for the numerical integration of the
equation of motion using the CDM. The program is available on this book’s
companion website. It computes the displacement u ðt Þ, the velocity u_ ðt Þ, the
acceleration u€ðt Þ, and the response ratio Rðt Þ ¼ u ðt Þ=ðpmax =k Þ and gives their
graphical representation. Moreover, it computes the dynamic magnification
factor D ¼ max jRðt Þj and the time tmax it occurs. The user of the program is
responsible for providing the function representing the excitation force.
Example 4.2.1 Response of a SDOF using the central difference method (CDM)
Using the CDM, determine the response of a SDOF system with
m ¼ 100kN m1 s2 , k ¼ 2500kN=m, x ¼ 0:05, and u ð0Þ ¼ u_ ð0Þ ¼ 0, and sub-
jected to the load pðt Þ ¼ po exp ð1  0:5t Þ, po ¼ 100kN.

Solution
The solution is obtained using the program centr_diff_lin.m with Dt ¼ 0:01.
Fig. E4.1 gives the response of the SDOF system. Moreover, Fig. E4.2 shows

Displacement Velocity
0.2 0.5

0.1
u,t(t)
u(t)

0
0

–0.1 –0.5
0 5 10 0 5 10
t t
Acceleration Response ratio
4 2

2
1
u,tt (t)

R(t)

0
0
–2

–4 –1
0 5 10 0 5 10
t t
FIG. E4.1 Response of the SDOF system in Example 4.2.1.
164 PART I Single-degree-of-freedom systems

the displacement u ðt Þ as compared with the exact one together with the error
u ðt Þ  uex ðt Þ. The exact solution was obtained by analytical evaluation of
Duhamel’s integral giving

 

4po ð1  2xwÞ
u ðt Þ ¼ e 10:5t
+  cos w D t + sin w D t e 1xwt
m ð1  4xw + 4w2 Þ 2wD
(1)

x 10 -4
0.2 1
u(t) computed
u(t) exact 0.8 u-uex

0.15 0.6

0.4

0.1 0.2
u(t)

0
0.05 –0.2

–0.4
0 –0.6

–0.8
-0.05 –1
0 2 4 6 8 10 0 2 4 6 8 10
t
FIG. E4.2 Computed solution and error in Example 4.2.1.

4.3 The average acceleration method

In 1959, N. M. Newmark developed a family of time step methods for the
numerical integration of the equation of motion [5] based on the approxima-
tion of the acceleration in each time step. In the following, we describe
only the well-known AAM (Average Acceleration Method), which is also
known as Newmark’s method with b ¼ 1=4 or Newmark’s trapezoidal rule
method. The AAM is the most widely employed numerical method in struc-
tural dynamics as it is simple to implement numerically as well as being
unconditionally stable and accurate (see Section 4.5.5). It can also be
employed to solve nonlinear equations of short duration motion [6, 7].
In this method, the acceleration in the interval t to t + Dt is assumed con-
stant and equal to its mean value (Fig. 4.3.1)
1
u€ðt + τÞ ¼ ½u€ðt Þ + u€ðt + Dt Þ 0  τ  Dt (4.3.1)
2
Integrating with respect to τ yields
τ
u_ ðt + τÞ ¼ ½u€ðt Þ + u€ðt + Dt Þ + C1
2
 Numerical integration of the equation of motion Chapter 4 165

FIG. 4.3.1 Variation of the acceleration, velocity, and displacement in the average acceleration
method.

where C1 is an arbitrary constant. For τ ¼ 0 we obtain C1 ¼ u_ ðt Þ, hence

τ
u_ ðt + τÞ ¼ u_ ðt Þ + ½u€ðt Þ + u€ðt + Dt Þ 0  τ  Dt (4.3.2)
2

Integrating once more with respect to τ yields

τ2
u ðt + τÞ ¼ τu_ ðt Þ + ½u€ðt Þ + u€ðt + Dt Þ + C2
4

which for τ ¼ 0 yields C2 ¼ u ðt Þ. Thus we have

τ2
u ðt + τÞ ¼ u ðt Þ + τ u_ ðt Þ + ½u€ðt Þ + u€ðt + Dt Þ 0  τ  Dt (4.3.3)
4
Eqs. (4.3.2), (4.3.3) for τ ¼ Dt give
Dt
u_ ðt + Dt Þ ¼ u_ ðt Þ + ½u€ðt Þ + u€ðt + Dt Þ (4.3.4)
2
Dt 2
u ðt + Dt Þ ¼ u ðt Þ + Dt u_ ðt Þ + ½u€ðt Þ + u€ðt + Dt Þ (4.3.5)
4
166 PART I Single-degree-of-freedom systems

Setting
Du ¼ u ðt + Dt Þ  u ðt Þ (4.3.6a)
Du_ ¼ u_ ðt + Dt Þ  u_ ðt Þ (4.3.6b)
Du€ ¼ u€ðt + Dt Þ  u€ðt Þ (4.3.6c)
Eqs. (4.3.4), (4.3.5) are written as
Dt
Du_ ¼ ½2u€ðt Þ + Du€ (4.3.7)
2
Dt 2
Du ¼ Dt u_ ðt Þ + ½2u€ðt Þ + Du€ (4.3.8)
4
€ we obtain
Solving Eq. (4.3.8) for Du,

4
Du€ ¼ ½Du  Dt u_ ðt Þ  2u€ðt Þ (4.3.9)
Dt 2

Moreover, substituting Eq. (4.3.9) into Eq. (4.3.7) yields

2
Du_ ¼ Du  2u_ ðt Þ (4.3.10)
Dt

We shall now express Du in terms of u_ ðt Þ and u€ðt Þ. For this purpose, we

apply Eq. (4.1.1) at time t + Dt and t. This yields

m u€ðt + Dt Þ + cu_ ðt + Dt Þ + ku ðt + Dt Þ ¼ pðt + Dt Þ

m u€ðt Þ + cu_ ðt Þ + ku ðt Þ ¼ pðt Þ

which after subtracting give

mDu€ + cDu_ + kDu ¼ Dp (4.3.11)
where
Dp ¼ pðt + Dt Þ  pðt Þ
Substituting Eqs. (4.3.9), (4.3.10) into Eq. (4.3.11) yields
K^ Du ¼ P^ (4.3.12)
where
2c 4m
K^ ¼ k + + (4.3.13)
Dt Dt 2
 
^ 4m
P ¼ Dp + + 2c u_ ðt Þ + 2m u€ðt Þ (4.3.14)
Dt
 Numerical integration of the equation of motion Chapter 4 167

The quantities K^ and P^ are referred to as effective stiffness and effective

load, respectively.
Eq. (4.3.12) allows the computation of Du when u_ ðt Þ and u€ðt Þ are known at
time t. Then Du€ðt Þ and Du_ ðt Þ can be computed using Eqs. (4.3.9), (4.3.10).
Subsequently, the values of u ðt + Dt Þ, u_ ðt + Dt Þ, u€ðt + Dt Þ result from the
expressions
u ðt + Dt Þ ¼ u ðt Þ + Du (4.3.15a)
u_ ðt + Dt Þ ¼ u_ ðt Þ + Du_ (4.3.15b)
u€ðt + Dt Þ ¼ u€ðt Þ + Du€ (4.3.15c)
It may be more convenient to compute the acceleration u€ðt Þ using the
equation of motion (4.1.1) than Eq. (4.3.15c), that is,
u€ðt + Dt Þ ¼ ½pðt + Dt Þ  cu_ ðt + Dt Þ  ku ðt + Dt Þ=m (4.3.16)
Table 4.3.1 presents the algorithm for the numerical implementation of the
AAM in pseudocode-type notation so that the reader can write a computer code
in the language of his/her preference.

TABLE 4.3.1 Average acceleration method (AAM).

A. Data
w (or k), m, x, u0 , u_ 0 , ttot , pðt Þ
B. Initial computations
pffiffiffiffiffiffiffiffiffi
B.1. k ¼ mw2 (or w ¼ k=m ), c ¼ 2mwx, ust ¼ pmax =k
B.2. u€0 ¼ ðp0  cu_ 0  ku 0 Þ=m
B.3. Choose Dt (usually Dt ¼ 0:1T =p) and compute
B.4. K^ ¼ k + 2c=Dt + 4m=Dt 2
C. For each time step compute
C.1. Dp ¼ pðt + Dt Þ  pðt Þ, P^ ¼ Dp + ð4m=Dt + 2cÞu_ ðt Þ + 2m u€ðt Þ
^ K^
C.2. Du ¼ P=
Du_ ¼ 2Du=Dt  2u_ ðt Þ
Du€ ¼ 4½Du  Dt u_ ðt Þ=Dt 2  2u€ðt Þ
C.3. u ðt + Dt Þ ¼ u ðt Þ + Du
u_ ðt + Dt Þ ¼ u_ ðt Þ + Du_
u€ðt + Dt Þ ¼ u€ðt Þ + Du€
C.4. Increase time t ¼ t + Dt and check: If t  ttot end. Else set u_ ðt Þ ¼ u_ ðt + Dt Þ,
u€ðt Þ ¼ u€ðt + Dt Þ and go to C.1
168 PART I Single-degree-of-freedom systems

As will be shown, contrary to the CDM, the stability of the AAM does not
demand any restriction on the size of the time step Dt. The time step, however,
is influenced by the accuracy of the method and its capability to describe an
oscillatory motion. Therefore, it must be small enough. The selection of Dt
equal to 1/10 of the period of the system or of the period of the excitation force
produces accurate results.
Adhering to the steps of Table 4.3.1, a computer program called av_acc_lin.m
has been written in MATLAB for the numerical integration of the equation of
motion. The program is available on this book’s companion website. It computes
the displacement u ðt Þ, the velocity u_ ðt Þ, the acceleration u€ðt Þ, and the response
ratio Rðt Þ ¼ u ðt Þ=ðpmax =k Þ and gives their graphical representation. Moreover, it
computes the dynamic magnification factor D ¼ max jRðt Þj and the time tmax it
occurs. The user of the program is responsible for providing the function of the
excitation force.

Example 4.3.1 Response of an SDOF using the average acceleration

method (AAM)
Determine the response the SDOF system in Example 4.2.1 using the AAM.

Solution
The solution was evaluated using the program av_acc_lin.m with Dt ¼ 0:01.
Fig. E4.3 gives the graphical representation of the displacement together with
the error u ðt Þ  uex ðt Þ. The computed error by the AAM is almost double the
error of the CDM. Moreover, Fig. E4.4 shows the response of the system under
the harmonic load p ¼ 2 sin 5t. Obviously, this excitation produces resonance
(w ¼ w ¼ 5).

x 10−4
0.2 2
u(t) computed
u(t) exact 1.5 u-uex
0.15
1

0.5
0.1
u(t)

0.05
–0.5

–1
0
–1.5

−0.05 –2
0 2 4 6 8 10 0 2 4 6 8 10
t t
FIG. E4.3 Computed solution and error in Example 4.3.1.
 Numerical integration of the equation of motion Chapter 4 169

Displacement Velocity
0.01 0.04

0.005 0.02

u,t(t)
u(t)

0 0

–0.005 –0.02

–0.01 –0.04
0 5 10 15 0 5 10 15
t t
Acceleration Response ratio
0.2 10

0.1 5
u,tt (t)

R(t)
0 0

–0.1 –5

–0.2 –10
0 5 10 15 0 5 10 15
t t
FIG. E4.4 Response of the system in Example 4.3.1 under harmonic load.

4.4 The analog equation method

The AEM (Analog Equation Method) is a general method for solving differen-
tial equations, ordinary or partial, linear or nonlinear, of the elliptic, parabolic,
or hyperbolic type. The AEM is based on the principle of the analog equation,
according to which a differential equation can be replaced by another equation,
the so-called analog equation, provided that the substitute equation retains the
principal mathematical symbol of the differential equation, that is, the highest
order derivative of the original differential equation. Thus, a differential equa-
tion whose solution cannot be obtained can be reduced to an equation with a
known solution. The AEM, which was first presented in 1994 [8], has been
employed as a computational method for the solution of numerous difficult
problems in engineering science and mathematical physics [9–11]. Here, the
AEM is presented for the solution of the equation of motion in structural dynam-
ics as developed in [6]. It is self-starting, unconditionally stable, accurate, and
conserves energy. It performs well when large deformations and long-time
durations are considered and it can be used as a practical method for the inte-
gration of the equations of motion in cases where widely used time integration
procedures, for example, Newmark’s AAM, become unstable [6, 7].
We consider the IVP (initial value problem) for the SDOF system
m u€ + cu_ + ku ¼ pðt Þ t  ½0, ttot , ttot > 0 (4.4.1)
170 PART I Single-degree-of-freedom systems

u ð0Þ ¼ u0 , u_ ð0Þ ¼ u_ 0 (4.4.2)

Let u ¼ u ðt Þ be the sought solution. Then, if the operator d =dt is applied to
2 2

it, we have
u€ ¼ q ðt Þ (4.4.3)
where q ðt Þ is a fictitious source, unknown in the first instance. Eq. (4.4.3) is the
analog equation of Eq. (4.4.1). It indicates that the solution of Eq. (4.4.1) can
be obtained by solving Eq. (4.4.3) with the initial conditions (4.4.2), if the q ðt Þ
is first established. This is achieved as follows.
Taking the Laplace transform of Eq. (4.4.3) we obtain
1 1 1
U ðs Þ ¼ u ð0Þ + 2 u_ ð0Þ + 2 Q ðs Þ (4.4.4)
s s s
where U ðs Þ,Q ðs Þ are the Laplace transforms of u ðt Þ,q ðt Þ, respectively. The
inverse Laplace transform of Eq. (4.4.4) gives the solution in integral from
Z t
u ðt Þ ¼ u ð0Þ + u_ ð0Þt + q ðτÞðt  τÞdτ (4.4.5)
0

Thus, the IVP of Eqs. (4.4.1), (4.4.2) is transformed into the equivalent
Volterra integral equation for q ðt Þ.
Eq. (4.4.5) is solved numerically within a time interval ½0, T . The interval
½0, T  is divided into N equal intervals, Dt ¼ h, h ¼ T =N , in which q ðt Þ is
assumed to vary according to a certain law, for example, constant, linear,
etc. In this analysis, q ðt Þ is assumed to be constant and equal to the mean value
in the interval h (Fig. 4.4.1). That is
qr1 + qr
qrm ¼ (4.4.6)
2

FIG. 4.4.1 Discretization of the interval ½0, T  into N equal intervals h ¼ T=N .

Hence, Eq. (4.4.5) at instant t ¼ nh can be written as

un ¼ u0 + nh u_ 0
" Z Z Z #
h 2h nh
+ q1 m
ðnh  τÞdτ + q2
m
ðnh  τÞdτ + ⋯ + qnm ðnh  τÞdτ
0 h ðn1Þh

(4.4.7)
 Numerical integration of the equation of motion Chapter 4 171

which after evaluation of the integrals yields

X
n 1
un ¼ u0 + nh u_ 0 + c1 ½2ðn  r Þ + 1qrm + c1 qnm
r¼1
(4.4.8)
X
n 1
¼ un1 + h u_ 0 + 2c1 qrm + c1 qnm
r¼1

where
h2
c1 ¼ (4.4.9)
2
The velocity is obtained by direct differentiation of Eq. (4.4.5) using
Leibnitz’ rule for integrals [12]. Thus, we have
Z t
u_ ðt Þ ¼ u_ ð0Þ + q ðτÞdτ (4.4.10)
0

Using the same discretization for the interval ½0, T  to approximate the inte-
gral in Eq. (4.4.10), we have
X
n1
u_ n ¼ u_ 0 + c2 qrm + c2 qnm
r¼1 (4.4.11)
¼ u_ n1 + c2 qnm
where
c2 ¼ h (4.4.12)
P
n1
Solving Eq. (4.4.11) for qrm and substituting into Eq. (4.4.8) gives
r¼1

un ¼ un1 + h u_ n  c1 qnm (4.4.13)

By virtue of Eq. (4.4.6), Eqs. (4.4.13), (4.4.11) are written as
c1 c1
qn  h u_ n + un ¼  qn1 + un1 (4.4.14)
2 2
c2 c2
 qn + u_ n ¼ u_ n1 + qn1 (4.4.15)
2 2
Moreover, Eq. (4.4.1) at time t ¼ nt is written as
mq n + cu_ n + ku n ¼ pn (4.4.16)
Eqs. (4.4.14), (4.4.15), (4.4.16) can be combined as
2 3 2 3
m c k 8 9 0 0 0 8 9 8 9
6 1 q
7< n = 6 1 7< qn1 = < 1 =
6 c h 1 7 u_ 6 c 0 1 7 u_
6 2 1 7 n ¼6 2 1 7 n1 + 0 p (4.4.17)
4 5: ; 4 5: ; : ; n
1 un 1 u 0
 c2 1 0 c2 1 0 n1
2 2
172 PART I Single-degree-of-freedom systems

Because m 6¼ 0, the coefficient matrix in Eq. (4.4.17) is not singular for suf-
ficient small h and the system can be solved successively for n ¼ 1, 2, … to yield
the solution un and the derivatives u_ n , u€n ¼ qn at instant t ¼ nh  T . For n ¼ 1,
the value q0 appears in the right side of Eq. (4.4.17). This quantity can be readily
obtained from Eq. (4.4.1) for t ¼ 0. Thus, we have
q0 ¼ ðp0  cu_ 0  ku 0 Þ=m (4.4.18)
Eq. (4.4.17) can be also written as
Un ¼ AUn1 + bpn , n ¼ 1, 2, …,N (4.4.19)
in which
8 9
< qn =
Un ¼ u_ n (4.4.20a)
: ;
un
2 31 2 3
m c k 0 0 0
6 1 7 6 1 7
6 c1 h 1 7 6  c1 0 1 7
6
A¼6 2 7 6 7 (4.4.20b)
7 6 2 7
4 1 5 4 1 5
 c2 1 0 c2 1 0
2 2
2 31
m c k 8 9
6 1 7 <1=
6 c h 1 7
b¼66 2
1 7
7 :0; (4.4.20c)
4 1 5 0
 c2 1 0
2
The recurrence formula (4.4.19) can be employed to construct the solution
algorithm. However, the solution procedure can be further simplified. Thus,
applying Eq. (4.4.19) for n ¼ 1, 2, … we have
U1 ¼ AUo + bp1
U2 ¼ AU1 + bp2
¼ AðAUo + bp1 Þ + bp2
(4.4.21)
¼ A2 Uo + Abp1 + bp2
⋯ ¼ ⋯⋯⋯⋯⋯⋯⋯⋯⋯
 
Un ¼ An Uo + An1 p1 + An2 p2 + ⋯A0 pn b
Obviously, the last of Eq. (4.4.21) gives the solution vector Un at instant
tn ¼ nh using only the known vector U0 at t ¼ 0. The matrix A and the vector
b are computed only once.
Table 4.4.1 presents the algorithm for the numerical implementation of
AEM in pseudocode-type notation so that the reader can write a computer code
in the language of his/her preference.
 Numerical integration of the equation of motion Chapter 4 173

Adhering to the steps of Table 4.4.1, a computer program called aem_lin.m

has been written in MATLAB for the numerical integration of the equation of
motion using the AEM. The program is available on this book’s companion
website. It computes the displacement u ðt Þ, the velocity u_ ðt Þ, the acceleration
u€ðt Þ, and the response ratio Rðt Þ ¼ u ðt Þ=ðpmax =k Þ and gives their graphical rep-
resentation. Moreover, it computes the dynamic magnification factor
D ¼ max jRðt Þj and the time tmax it occurs. The user is responsible for provid-
ing the function of excitation force.

TABLE 4.4.1 The analog equation method (AEM).

A. Data
w (or k), m, x, u0 , u_ 0 , ttot , pðt Þ, ttot
B. Initial computations
Choose: h ð¼ Dt Þ and compute ntot ,
Compute: k ¼ mw2 , c ¼ 2mwx
q0 ¼ ðp0  cu_ 0  ku 0 Þ=m, c1 ¼ h 2 =2, c2 ¼ h
Formulate: U0 :¼ fq0 u_ 0 u0 gT
2 31 2 3 2 31
m c k 0 0 0 m c k 8 9
6 c1 7 6 1 7 6 c1 7 <1=
6 h 1 7 6  c1 0 1 7 6 h 1 7
6
Compute: A ¼ 6 2 7 6 7 6 7
7 6 2 7, b ¼ 6 2 7 :0;
4 c2 5 4 5 4 c2 5 0
 1 0 1  1 0
2 c2 1 0 2
2
C. Compute solution
for n :¼ 1 to ntot
Un ¼ AUn1 + bpn
end

Example 4.4.1 Response of an SDOF system using the analog equation method
Determine the response the SDOF system in Example 4.2.1 using the AEM.

Solution
The solution is obtained using the program aem_lin.m with Dt ¼ 0:01.
Fig. E4.5 gives the graphical representation of the displacement together with
the error u ðt Þ  uex ðt Þ. Moreover, Fig. E4.6 shows the response of the system
under the static load p ¼ t, if 0  t  1 and p ¼ 1, if 1 < t.
174 PART I Single-degree-of-freedom systems

x 10−4
0.2 2
u(t) computed
u-uex
u(t) exact 1.5
0.15
1

0.1 0.5
u(t)

0
0.05
–0.5

–1
0

–1.5

–0.05
0 2 4 6 8 10 –2
0 2 4 6 8 10
t

FIG. E4.5 Computed solution and error in Example 4.4.1.

× 10
-4 Displacement × 10
-4 Velocity
6 10

4 5
du(t)
u(t)

2 0

0 –5
0 5 10 0 5 10
t t
× 10
-3 Acceleration Response ratio
2 1.5

0 1
ddu(t)

R(t)

–2 0.5

–4 0
0 5 10 0 5 10
t t
FIG. E4.6 Response of the system in Example 4.4.1 under static load.

4.5 Stability of the numerical integration methods

4.5.1 Errors in the numerical integration
The numerical integration of the differential equations, hence of the equation of
motion, introduces errors, which influence not only the accuracy of the obtained
solution but also its capability to produce a reliable solution, that is, a solution
that within certain accuracy represents the actual solution. Factors that may
 Numerical integration of the equation of motion Chapter 4 175

contribute to errors in the results obtained from well-defined loadings are of

three types [13]:
(i) Round-off errors: They result from calculations being done using numbers
expressed by too few digits.
(ii) Truncation errors: They are introduced in representing u ðt + Dt Þ or
u_ ðt + Dt Þ by a finite number of terms in the Taylor series expansion.
(iii) Instability errors: They are introduced by amplification of the errors from
one step to the subsequent one during the computations. The stability of
any method is improved by reducing the length of the time step.
Round-off errors are random in nature and therefore must be treated by statis-
tical methods. In computer calculations, it is often possible to reduce the round-
off errors by using a higher precision in the computations. Truncation errors are
accumulated locally at each step. These errors provide a useful criterion for
measuring the accuracy of the various methods of numerical integration, pro-
vided that the employed numerical scheme is stable.
To study the stability of the numerical method, it is important to know the
influence of the error introduced at one step on the computations at the next step.
If the error tends to increase, then the solution soon becomes unbounded and
ceases to make sense. In this case, we say that the method is unstable. The study
of the stability of numerical integration methods of the differential equation of
motion and general of IVPs of differential equations is facilitated by the use of
difference equations, elements of which are discussed below.

4.5.2 Difference equations

If tn represents the time nh ðh ¼ Dt Þ, n ¼ 0, 1, 2… and un the displacement at
the same time, then we may write Eq. (4.2.6) as
un + 1 ¼ 2a1 un  a2 un1 + gn (4.5.1)
where
2  h 2 w2 1  hwx pn h 2
a1 ¼ , a2 ¼ , gn ¼
2ð1 + hwx Þ 1 + hwx m ð1 + hwx Þ
Eq. (4.5.1) is a recursive relation, which permits the computation of un + 1
from its previous two values. Applying this relation successively for
n ¼ 0, 1, … we have
9
For n ¼ 0 u1 ¼ 2a1 uo  a2 u1 + g0 >
>
>
>
For n ¼ 1 u2 ¼ 2a1 u1  a2 u0 + g1 =
For n ¼ 2 u3 ¼ 2a1 u2  a2 u1 + g2 (4.5.2)
>
⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯⋯ > >
>
;
For n ¼ n un + 1 ¼ 2a1 un  a2 un1 + gn
176 PART I Single-degree-of-freedom systems

We observe that un + 1 can be computed if the two initial values u1 and uo
are known. Eq. (4.5.1) may also be written as

un + 1  2a1 un + a2 un1 ¼ gn (4.5.3)

Eq. (4.5.3) is a difference equation of the second order [14] whose solution
can yield un + 1 using the recursive procedure (4.5.2). Because Eq. (4.5.3) is
linear with respect to un1 , un , and un + 1 , it is called a linear difference equa-
tion. If gn ¼ 0 the difference equation is called homogeneous while if gn 6¼ 0 it is
called nonhomogeneous. In general, an equation of the form

un + k + a1 un + k1 + a2 un + k2 + … + ak un ¼ gn (4.5.4)

where a1 ,a2 , …,ak are constants is a linear difference equation of order k

with constant coefficients. The solution of the difference equation is a
sequence of uk values for which it is true. The order of the difference equation
is the difference between the largest and the smallest value of the argument k
appearing in it.
First, we examine the homogeneous linear difference equation ðgn ¼ 0Þ. The
solution of this equation is sought in the form un ¼ rn , which is inserted in
Eq. (4.5.4) to give

rn + k + a1 rn + k1 + a2 rn + k2 + … + ak rn ¼ 0

or dividing by rn we obtain the characteristic equation of the difference

equation (4.5.4)

pðrÞ ¼ rk + a1 rk1 + a2 rk2 + … + ak ¼ 0 (4.5.5)

which is polynomial of order k.

If we assume that all roots r1 ,r2 ,…,rk of pðrÞ are distinct, then
rn1 ,rn2 ,…, rnk are solutions of Eq. (4.5.4). Moreover, because of the linearity
of the equation

un ¼ c1 rn1 + c2 rn2 + … + ck rnk (4.5.6)

is also a solution for all n for which the difference equation is defined with
c1 ,c2 ,…, ck being arbitrary constants. Eq. (4.5.6) is the general solution of
the difference equation (4.5.4). The arbitrary constants c1 ,c2 ,…, ck are deter-
mined from the k initial conditions.
If two roots of the polynomial (4.5.5) are complex conjugate, say r1 ¼
a + ib and r2 ¼ a  ib, then we can write them in exponential form

r1 ¼ reiq , r2 ¼ reiq (4.5.7)

where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r¼ a2 + b 2 , q ¼ tan 1 ðb=aÞ (4.5.8)
 Numerical integration of the equation of motion Chapter 4 177

Using Euler’s identity (2.2.8), we obtain

rn1 ¼ r n einq ¼ r n ð cos nq + i sin nqÞ:
rn2 ¼ r n einq ¼ r n ð cos nq  i sin nqÞ
thus we can write
c1 rn1 + c2 rn2 ¼ r n ½ðc1 + c2 Þcos nq + iðc1  c2 Þsin nq
 (4.5.9)
¼ r n c10 cos nq + c20 sin nq
where c10 ¼ c1 + c2 and c20 ¼ iðc1  c2 Þ. We conclude from Eq. (4.5.9) that the
contribution of the complex roots to the solution is oscillatory.
Finally, we examine the case where the polynomial (4.5.5) has a double root,
say r1 . Then, the second solution will be nrn1 . This is verified by direct substi-
tution of un ¼ nrn1 into Eq. (4.5.4). Thus we obtain
ðn + k Þrn1 + k + a1 ðn + k  1Þrn1 + k1 + … + ak1 ðn + 1Þrn1 + 1 + ak nrn1
 
¼ nrn1 rk1 + a1 rk1 + … + ak1 r1 + ak
 1

+ rn1 + 1 krk1
1 + a1 ðk  1Þrk2 + … + ak1
which yields
nrn1 pðr1 Þ + rn1 + 1 p0 ðr1 Þ ¼ 0
The latter relation results because not only pðr1 Þ ¼ 0 but also p0 ðr1 Þ ¼ 0.
The solution of the nonhomogeneous equation, gn 6¼ 0, is obtained as the
sum of the homogeneous solution uno and a particular solution of the nonhomo-
geneous equation unp , namely
un ¼ uno + unp (4.5.10)
In the special case where gn ¼ g ¼ constant, the particular solution is read-
ily obtained. Indeed, introducing un ¼ G ¼ constant in Eq. (4.5.4) yields
ð1 + a1 + a2 + … + ak ÞG ¼ g
from which we obtain
g
G¼ (4.5.11)
1 + a1 + a2 + … + ak
provided that 1 + a1 + a2 + … + ak does not vanish.
Example 4.5.1 Solution of a second-order difference equation
Find the solution of the IVP
un + 2  2un + 1 + 2un ¼ 0 (1)
u0 ¼ 0, u1 ¼ 1 (2)
Solution
The characteristic equation of (1) is
r2  2r + 2 ¼ 0 (3)
178 PART I Single-degree-of-freedom systems

Its roots are r1 ¼ 1 + i, r2 ¼ 1  i and Eq. (4.5.8) give

pffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffi
r ¼ 12 + 12 ¼ 2, q ¼ tan 1 ð1Þ ¼ p=4
Hence, Eq. (4.5.9) gives

np np
un ¼ 2n=2 c1 cos + c2 sin (4)
4 4
Applying the initial conditions gives c1 ¼ 0, c2 ¼ 1 and the solution becomes
np
un ¼ 2n=2 sin (5)
4

Example 4.5.2 Solution of a third-order difference equation

Find the solution of the IVP
un + 3  2un + 2  un + 1 + 2un ¼ 0 (1)
u0 ¼ 0, u1 ¼ 1, u2 ¼ 1 (2)
Solution
The difference equation is of the third order. Its characteristic equation is
r3  2r2  r + 2 ¼ 0 (3)
Its roots are þ1, 1, 2. Hence the general solution of (1) is
un ¼ c1 ð1Þn + c2 ð1Þn + c3 ð2Þn
(4)
¼ c1 + ð1Þn c2 + 2n c3
Applying the initial conditions gives the following system for the arbitrary
constants c1 ,c2 , c3
c1 + c2 + c3 ¼ 0
c1  c2 + 2c3 ¼ 1 (5)
c1 + c2 + 4c3 ¼ 1
which is solved to give c1 ¼ 0, c2 ¼ 1=3, c3 ¼ 1=3. Therefore, we obtain the
closed form solution for the difference equation
1 2n
un ¼  ð1Þn + (6)
3 3

4.5.3 Difference equations and stability of the numerical

integration methods
Along with the widespread use of computers for solving differential equations,
it was observed that some known integration schemes lead to errors in the
solution that are much greater than expected due to discretization. Moreover,
these errors for a certain value of the variable t increased, although the time step
 Numerical integration of the equation of motion Chapter 4 179

Dt ¼ h was reduced. To understand this behavior, we examine the solution of

the differential equation
u_ ¼ 2u + 1, u ð0Þ ¼ 1 (4.5.12)
Using the central difference to approximate the derivative, we obtain the
difference equation
un + 1 + 4hu n  un1 ¼ 2h (4.5.13)
where h is the constant time step of the integration.
The solution of the previous difference equation is
1
un ¼ c1 rn1 + c2 rn2 + (4.5.14)
2
where r1 and r2 are the roots of the characteristic equation
r2 + 4hr  1 ¼ 0
These roots are
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r1, 2 ¼ 2h  1 + 4h 2
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Expanding 1 + 4h 2 in the Taylor series and keeping only the linear terms,
the roots are expressed as
r1 ¼ 1  2h (4.5.15a)
r2 ¼ ð1 + 2h Þ (4.5.15b)
and Eq. (4.5.14) is written
1
un ¼ c1 ð1  2h Þn + c2 ð1Þn ð1 + 2h Þn + (4.5.16)
2
It is known from differential calculus that

‘im ð1 + eÞ1=e ¼ e
e!0

Using this relation and taking into account that n ¼ tn =h, we obtain for a
given tn

‘im ð1 + 2h Þn ¼ ‘im ð1 + 2h Þð1=2hÞð2tn Þ ¼ e2tn

h!0 h!0

In the same way, we obtain

‘im ð1  2h Þn ¼ e2tn
h!0

Consequently, for h ! 0 the solution (4.5.16) becomes

 
2tn 1
u n ¼ c1 e + + c2 ð1Þn e2tn (4.5.17)
2
180 PART I Single-degree-of-freedom systems

On the other hand, the exact solution of Eq. (4.5.12) is

1
un ¼ c1 e2tn + (4.5.18)
2
Evidently, the first term in Eq. (4.5.17) is the exact solution. The second
term is spurious (extraneous) and results from the fact that the first-order dif-
ferential equation is substituted by a second-order difference equation. The
application of the initial conditions would give c2 ¼ 0 if the computations were
exact. In practice, however, errors are introduced, which are mainly due to the
rounding of numbers or the inaccuracy of the starting value. Therefore, the con-
stant c2 is not exactly zero and consequently, a small error is introduced in each
integration step. This is magnified because it is multiplied by the factor
ð1Þn e2tn , which increases exponentially. Because the first term of the solution
(4.5.17) diminishes exponentially, the introduced error due to the spurious solu-
tion dominates the exact solution and leads to a totally wrong result. We
describe a method as unstable if the error increases exponentially with tn .
For the first order differential equations, the one-step integration methods do
not exhibit instability for small values of h. The multistep methods, however,
which lead to difference equations of order greater than one, introduce spurious
solutions and they may be unstable either for all values of h or for a certain
region of the values of h. In order to decide whether a multistep method is sta-
ble, we work as follows.
If the multistep method leads to a difference equation of order k, we find the
roots of the characteristic equation. If ri (i ¼ 1,2, …,kÞ are these roots, the gen-
eral solution will be

un ¼ c1 rn1 + c2 rn2 + … + ck rnk

One of the solutions, say rn1 , will tend to the actual solution of the differential
equation. The remaining roots are spurious. We will say that a multistep method
is strongly stable if for h ! 0 the spurious roots satisfy the condition

jri j < 1, i ¼ 2, 3, …, k (4.5.19)

Because we do not know which is the actual solution, the above condition
should apply to all roots ri . Apparently, this condition ensures that the error
diminishes as n increases. On the contrary, the error increases exponentially
if jri j > 1.

4.5.4 Stability of the central difference method

As was shown in Section 4.5.2, the CDM leads to the difference equation
(4.5.3), namely
un + 1  2a1 un + a2 un1 ¼ gn (4.5.20)
 Numerical integration of the equation of motion Chapter 4 181

where
2  h 2 w2 1  hwx pn h 2
a1 ¼ , a2 ¼ , gn ¼ (4.5.21)
2ð1 + hwx Þ 1 + hwx m ð1 + hwx Þ
The characteristic equation of Eq. (4.5.20) is
r2  2a1 r + a2 ¼ 0 (4.5.22)
whose roots are
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r1, 2 ¼ a1  a21  a2

The type of root depends on the sign of the discriminant D ¼ a21  a2 , which
may be written as

pffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffi
s2 s  2 1  x2 s + 2 1  x2
D ðs Þ ¼ , s ¼ hw
4ð1 + sxÞ2

pffiffiffiffiffiffiffiffiffiffiffiffis > 0, the sign of the discriminant depends only on the factor
Because
s  2 1  x2 . Hence, we distinguish the following two cases
qffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffi
2 1  x2 T
ðiÞ Dðs Þ > 0 s > 2 1  x 2
or h > ¼ 1  x2 (4.5.23)
w p
qffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffi
2 1  x2 T
ðiiÞ Dðs Þ  0 s  2 1  x 2
or h  ¼ 1  x 2 (4.5.24)
w p
In case (i), the characteristic equation has two real roots, r1 , r1 , The stability
of the solution requires that jr1 j < 1 and jr2 j < 1. But it can be shown that
jr2 j > 1, hence jr1 j > 1. Consequently, the solution is unstable in this case.
In case (ii), the characteristic equation has two complex conjugate roots
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r1, 2 ¼ a1  i a2  a21 (4.5.25)

This gives
sffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi 1  sx
j r1 j ¼ j r2 j ¼ a2 ¼ 1 (4.5.26)
1 + sx

Consequently, the solution is stable in case (ii).

The stability criterion (4.5.26) is also explained by writing the complex roots
in exponential form
r1 ¼ reiqt , r2 ¼ reiqt (4.5.27)
where
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffi 1  hwx a2
r ¼ a2 ¼  1 and q ¼ tan 1 1 (4.5.28)
1 + hwx a21
182 PART I Single-degree-of-freedom systems

Then solution (4.5.20) is written as

pffiffiffiffiffi n
un ¼ ð a2 Þ ðc1 cos nq + c2 sin nqÞ + unp (4.5.29)
pffiffiffiffiffi
which is bounded because a2  1.
The previous investigation leads to the conclusion that the CDM is stable if
the time step satisfies the condition.
qffiffiffiffiffiffiffiffiffiffiffiffi
T T
Dt < 1  x2  (4.5.30)
p p
This condition imposes a constraint on the size of the time step. In such a
case, we say that the numerical integration method is conditionally stable. It
should be noted that this condition is never a restriction for SDOF systems
because the accuracy of the results requires choosing a much smaller time step,
usually Dt ¼ 0:1T =p is adequate.

4.5.5 Stability of the average acceleration method

We consider the equation of motion (4.4.1) at time t ¼ nt
u€n + 2xwu_ n + w2 un ¼ pn , pn ¼ pn =m (4.5.31)
and the approximate expressions (4.3.5) and (4.3.4) for the displacement and
velocity at the same instant, namely
h2
un ¼ un1 + h u_ n + ½u€n1 + u€n  (4.5.32)
4
h
u_ n ¼ u_ n1 + ½u€n1 + u€n  (4.5.33)
2
The previous three equations can be combined as
2 3 2 3
1 2xw w2 8 9 0 0 0 8 9 8 9
6 h2 7< u€n = 6 h 2 7< u€n1 = < 1 =
6 7 6 7
6 0 1 7 u_ 6 h 1 7 u_ n1 + 0 pn
6 4 7: n ; ¼ 6 4 7: (4.5.34)
4 h 5 un 4h 5 un1 ; : 0 ;
 1 0 1 0
2 2
Solving for the vector Un ¼ f u€n u_ n u_ n gT , we obtain
Un ¼ AUn1 + b
pn , n ¼ 1, 2, …,N (4.5.35)
where
2 31 2 3
1 2xw w2 0 0 0
6 h2 7 6 h2 7
6 7 6 7
6
A¼6 4 0 1 7 6 h 1 7 (4.5.36a)
7 64 7
4 h 5 4h 5
 1 0 1 0
2 2
 Numerical integration of the equation of motion Chapter 4 183

2 31
1 2xw w2 8 9
6 7 >
> 1>
6 h2 7 < > =
6 1 7
b ¼ 6 4 0 7 0 (4.5.36b)
6 7 >
> >
4 h 5 : > ;
0
 1 0
2
Applying Eq. (4.5.35) for n ¼ 1, 2, … we have
U1 ¼ AUo + b
p1
U2 ¼ AU1 + b
p2
¼ AðAUo + b
p1 Þ + b
p2
(4.5.37)
¼ A2 Uo + Ab
p1 + b
p2
⋯ ¼ ⋯⋯⋯⋯⋯⋯⋯⋯⋯
 
Un ¼ An Uo + An1 p1 + An2 p2 + …A0 pn b
The matrix A is known as the amplification matrix. The stability of the
method requires that An is bounded. This is true if the spectral radius of A sat-
isfies the condition [15]
rðAÞ ¼ max fjr1 j, jr2 j, jr3 jg  1 (4.5.38)
where ri (i ¼ 1, 2, 3) are the eigenvalues of the matrix A.
Using a symbolic language (here MATLAB) we find
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
4  s2 + 4s x2  1 4  s 2  4s x2  1
r1 ¼ , r2 ¼ , r3 ¼ 0 (4.5.39)
4 + 4sx + s 2 4 + 4sx + s 2

where s ¼ xw.
The type of the roots r1 , r2 depends on the sign of the discriminant
DðxÞ ¼ x2  1. Hence, we distinguish the following two cases
(i) If D ðx Þ > 0, both eigenvalues are real. It can be shown that jr1 j < 1, hence
jr2 j < 1. Therefore, the method is stable
(ii) If DðxÞ  0, the eigenvalues are complex conjugate
pffiffiffiffiffiffiffiffiffiffiffiffi
4  s 2  i4s 1  x2
r1,2 ¼
4 + 4sx + s 2

and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð4  s 2 Þ2 + 16s2 1  x 2 ðs 2 + 4Þ2  16s 2 x2
j r1 j ¼ j r2 j ¼ ¼  1 (4.5.40)
ð4 + 4sx + s 2 Þ ½ðs 2 + 4Þ + 4sx 2
The equality is valid for x ¼ 0.
184 PART I Single-degree-of-freedom systems

The conclusion is that the AAM is stable without imposing any constraint on
the size of the time step. We say in this case that the integration method is
unconditionally stable.
Essentially, the procedure based on the condition (4.5.38) to prove the sta-
bility of the AAM is not different from that presented in Section 4.5.3, where the
stability results from the response of the difference equation. This is shown in
what it follows.
We write Eq. (4.5.35) for tn ,tn1 , tn2
Un  AUn1 ¼ b
pn
Un1  AUn2 ¼ b
pn1 (4.5.41)
Un2  AUn3 ¼ b
pn2
or in matrix form
8 9
2 3> Un > 8 9
I A 0 >
0 < >
= < pn =
4 0 I A 0 5 Un1
¼ b pn1 (4.5.42)
> Un2 > : ;
0 0 I A >
: >
; pn2
Un3
Eq. (4.5.42), beside the displacements un ,un1 ,un2 , un3 , contains the
velocities u_ n , u_ n1 , u_ n2 , u_ n3 and the accelerations u€n , u€n1 , u€n2 , u€n3 . Reor-
dering these equations and eliminating the velocities and accelerations yield the
equation
un  2a1 un1 + a2 un2 + a3 un3 ¼ c1 pn + c2 pn1 + c3 pn2 (4.5.43)
where
4  w2 h 2 4  4sx + w2 h 2
a1 ¼ , a2 ¼ , a3 ¼ 0 (4.5.44)
4a 4a
1 + xwh 1  xwh
c1 ¼ , c2 ¼ , c3 ¼ 0 (4.5.45)
a a
where
w2 h 2
a ¼ 1 + xwh +
4
Eq. (4.5.43) is a difference equation whose characteristic equation is
r3  2a1 r2 + a2 r ¼ 0 (4.5.46)
The roots of Eq. (4.5.46) are
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
4  s 2 + 4s x 2  1 4  s 2  4s x2  1
r1 ¼ , r2 ¼ , r3 ¼ 0 (4.5.47)
4 + 4sx + s 2 4 + 4sx + s 2
that is, they are identical to the eigenvalues of the matrix A.
 Numerical integration of the equation of motion Chapter 4 185

4.5.6 Stability of the analog equation method

The stability of the AEM is studied by considering the eigenvalues of the ampli-
fication matrix A given by Eq. (4.4.20b). Hence, the stability requires the valid-
ity of Eq. (4.5.38). Using a symbolic language (here MATLAB) we obtain the
three eigenvalues
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
4  s2 + 4s x2  1 4  s 2  4s x2  1
r1 ¼ , r2 ¼ , r3 ¼ 0 (4.5.48)
4 + 4sx + s 2 4 + 4sx + s 2
where s ¼ xw. Obviously, they are identical to those given by Eq. (4.5.39).
Consequently, we can state that the AEM is unconditionally stable.

4.6 Accuracy of the numerical integration

The methods for numerical integration of the equation of motion found in the
literature are numerous. They are either unconditionally stable or conditionally
stable. In the latter case, wDt must be less than a certain value, which is greater
pffiffiffiffiffiffiffiffiffiffiffiffi
than one, for example, in the method of central differences this value is
2 1  x2 . However, the accuracy of the results requires values of wDt much
smaller than one. Therefore, the limitation of ensuring the stability of the solu-
tion is not decisive for the choice of the integration step, at least for the SDOF
systems. Hence, the choice of method should be based on the relative accuracy
of the numerical results. Actually, the accuracy of the numerical method
depends on the truncation error in the Taylor series representing the derivatives
(for example, CDM) or the approximation of the second derivative within the
time step (for example, AAM, AEM).
For free vibrations, as a measure of the relative accuracy, we define the
quantities
T  T
PE ¼ (4.6.1)
T
and
rn  rn + 1
AD ¼ (4.6.2)
rn
where T and T are the exact and the approximate periods, respectively, and
rn ,rn + 1 are the amplitudes at the consecutive times tn ,tn + 1 .
The first relation expresses the period elongation (PE) while the second the
amplitude decay (AD) over time.
The numerical solution obtained by the AAM or the AEM can be written in
terms of the eigenvalues
un ¼ c10 rn1 + c20 rn2 ð r 3 ¼ 0Þ (4.6.3)
186 PART I Single-degree-of-freedom systems

or
un ¼ r n ðc1 sin nq + c2 cos nqÞ
(4.6.4)
 n + c2 cos wt
¼ r n ðc1 sin wt  nÞ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where r ¼ a 2 + b2 , q ¼ tan 1 ðb=a Þ, a ¼ Reðr1 Þ, b ¼ Imðr1 Þ, w ¼ q=h,
tn ¼ nh.
The corresponding exact solution is
qffiffiffiffiffiffiffiffiffiffiffiffi
un ¼ exwtn ðc1 sin wD tn + c2 cos wD tn Þ, wD ¼ 1  x2 , tn ¼ nh (4.6.5)

Comparison of Eqs. (4.6.4), (4.6.5) could show the accuracy of the numer-
ical scheme. Thus, the period elongation is
pffiffiffiffiffiffiffiffiffiffiffiffi
T  T h 1  x2
PE ¼ ¼ 1 (4.6.6)
T q
For the amplitude decay, we can define an equivalent damping ratio x from
the relation

n 
r n ¼ exwt ¼ exqn (4.6.7)
which gives
x ¼  ln r=q (4.6.8)
The difference Dx ¼ x  x can be employed as a measure for the amplitude
decay. The dependence of the period elongation and amplitude decay on h=T is
shown in Figs. 4.6.1 and 4.6.2, respectively. Obviously, for small values of h=T
the scheme is accurate. Note that for x ¼ 0 it is jr2 j ¼ jr3 j ¼ r ¼ 1 and Eq. (4.6.8)
yields x ¼ 0. That is, there is no amplitude decay.

0.25
x=0
0.2 x = 0.1
Period elongation %

x = 0.2
0.15

0.1

0.05

0
0 0.05 0.1 0.15 0.2 0.25 0.3
h/T
FIG. 4.6.1 Period elongation versus h=T for different values x.
 Numerical integration of the equation of motion Chapter 4 187

−3
x 10
2
x=0
0 x = 0.1
x = 0.2

−2
Δx

−4

−6

−8
0 0.05 0.1 0.15 0.2 0.25 0.3
h/T
FIG. 4.6.2 Amplitude decay Dx ¼ x  x versus h=T for different values of x.

Example 4.6.1 Stability of Houbolt’s method

Houbolt’s method for the numerical integration of the equation of motion in the
absence of damping requires the solution of the difference equation [4]
 
2 + s 2 un + 1  5un + 4un1  un2 ¼ 0, s ¼ wh (1)

Investigate the stability of the method.

Solution
The characteristic equation of Eq. (1) is
 
2 + s 2 r3  5r2 + 4r  1 ¼ 0 (2)

Its roots are

 
1 b 12a  25
2
r1 ¼ +5 (2)
3a 2 b
  pffiffiffi  
1 b 12a  25 3 b 12a  25
r2,3 ¼  + +5 i +2 (3)
3a 4 b 6a 2 b
where
 
a ¼ 2 + s2 (4)

pffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1=3
b ¼ 720a + 108a 2 + 1000 + 12 3a 104a + 100 + 27a 2 (5)

The stability condition requires that jr1 j < 1 and jr2 j ¼ jr3 j < 1. This
condition is satisfied, as is shown in Fig. E4.7.
188 PART I Single-degree-of-freedom systems

1
|r1|
0.9
|r2|
0.8 |r3|
0.7

0.6

0.5
|r2|=|r3|
0.4

0.3

0.2

0.1
| r1| (spurious)
0
1 2 3 4 5 6 7 8 9 10
h/t
FIG. E4.7 Houbolt’s method. Roots of the characteristic equation.

4.7 Problems
Problem P4.1 Find the solution of the IVPs for the difference equations
un + 1  3un ¼ 5, u0 ¼ 1
un + 2  4un + 1 + 3un ¼ 2n , u0 ¼ 0, u1 ¼ 1
Hint: The particular solution of (ii) will be sought in the form gn ¼ 2n c,
where c is a constant.
Problem P4.2 A SDOF system is subjected to pulse the load
 
t

pðt Þ ¼ p0 1  1 + et=t1 , 0  t  t1
t1
pðt Þ ¼ 0, t > t1
Study the response of the system using
(i) The central difference method.
(ii) The average acceleration method.
(iii) The analog equation method.
Compare the results with the exact solution.
Data: u ð0Þ ¼ u_ ð0Þ ¼ 0, m ¼ 50kNm1 s2 , x ¼ 0:05, w ¼ 5s1 , p0 ¼ 100kN,
t1 ¼ 0:5s and ttot ¼ 10s.
Problem P4.3 A SDOF system is subjected to the piecewise linear load pðt Þ
shown in Fig. P4.3. Use the three discussed numerical methods to establish
its response.
 Numerical integration of the equation of motion Chapter 4 189

Compare the results with the exact solution. Data: u ð0Þ ¼ u_ ð0Þ ¼ 0,
m ¼ 50kN m1 s2 , x ¼ 0:05, w ¼ 5s1 , ti ¼ 0:01i, pi ¼ 20 1 + ð1Þi ði + 5Þ=

ði + 1Þ , i ¼ 0, 2, …,100.

FIG. P4.3 Piecewise linear load in Problem P4.3.

References and further reading

[1] M.A. Dokainish, K. Subbaraj, A survey of direct time-integration method in computational
structural dynamics. I. Explicit methods, Comput. Struct. 32 (1989) 1371–1386.
[2] K. Subbaraj, M.A. Dokainish, A survey of direct time-integration methods in computational
structural dynamics. II. Implicit methods, Comput. Struct. 32 (1989) 1387–1401.
[3] C.H. Norris, Structural Design for Dynamic Loads, McGraw-Hill, New York, 1959.
[4] K.J. Bathe, E.L. Wilson, Numerical Methods in Finite Elements, Prentice-Hall, Englewood
Cliffs, NJ, 1976.
[5] N.M. Newmark, A method of computation of structural dynamics, J. Eng. Mech. 85 (1959)
67–94.
[6] J.T. Katsikadelis, A new direct time integration method for the equations of motion in struc-
tural dynamics, ZAMM Z. Angew. Math. Mech. 94 (9) (2014) 757–774, https://doi.org/
10.1002/zamm.20120024.
[7] K.J. Bathe, Conserving energy and momentum in nonlinear dynamics: a simple implicit time
integration scheme, Comput. Struct. 85 (2007) 437–445.
[8] J.T. Katsikadelis, The analog equation method–a powerful BEM-based solution technique
for solving linear and nonlinear engineering problems, in: Transactions on Modelling and
Simulation, vol. 7, WIT Press, pp. 167–182, www.witpress.com, 1743-355X.
[9] J.T. Katsikadelis, The analog boundary integral equation method for nonlinear static and
dynamic problems in continuum mechanics, J. Theor. Appl. Mech. 40 (4) (2002).
[10] J.T. Katsikadelis, The Boundary Element Method for Plate Analysis, Academic Press,
Elsevier, Oxford, UK, 2014.
[11] J.T. Katsikadelis, The Boundary Element for Engineers and Scientists, Academic Press,
Elsevier, Oxford, UK, 2016.
[12] F.B. Hildebrand, Advanced Calculus for Applications, Prentice-Hall, Inc., Englewood Cliffs,
NJ, 1962.
[13] J.L. Humar, Dynamics of Structures, second ed., A.A. Balkema Publishers, Lisse, NL, 2002.
[14] F. Scheid, Numerical Analysis, second ed., Schaum’s Outline Series, McGraw-Hill, New
York, 1998.
[15] U.N. Faddeeva, Computational Methods in Linear Algebra, Dover Publications, New York,
1959.
Chapter 5

Nonlinear response:
Single-degree-of-freedom
systems
Chapter outline
5.1 Introduction 191 5.4 The analog equation
5.2 The central difference method 195 method 203
5.3 The average acceleration 5.5 Problems 211
method 197 References and further reading 215

5.1 Introduction
The equation of motion of a vibrating system expresses the equilibrium condi-
tion of all forces applied to the system, namely the external excitation force, the
inertial force, the damping force, and the elastic force. The equilibrium condi-
tion reads
fI ðt Þ + fD ðt Þ + fS ðt Þ ¼ pðt Þ (5.1.1)
The forces fI , fD , and fS depend on the physical properties of the system.
In the systems we analyzed, the physical properties are not time-dependent
and the dependence of these forces on the cause that produces them is linear,
that is,
fI ¼ m u€ (5.1.2a)
fD ¼ cu_ (5.1.2b)
fS ¼ ku (5.1.2c)
where m, c,k are constant quantities. Systems with such a physical response are
referred to as linear systems.
In general, however, Eq. (5.1.2a)-(5.1.2c) may be of the form
fI ¼ m ðt Þu€ (5.1.3a)
_ tÞ
fD ¼ fD ðu, u, (5.1.3b)
_ tÞ
fS ¼ fS ðu, u, (5.1.3c)

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00005-4

© 2020 Elsevier Inc. All rights reserved. 191
192 PART I Single-degree-of-freedom systems

That is, the mass may vary with time and the damping and elastic forces may
be nonlinear functions of u, u, _ t Þ, fS ðu, u,
_ and t, that is, fD ðu, u, _ t Þ. In this case,
Εq. (5.1.1) takes the form
_ t Þ + fS ðu, u,
m ðt Þu€ + fD ðu, u, _ t Þ ¼ pðt Þ (5.1.4)
which is a nonlinear differential equation of the second order. A dynamic sys-
tem, whose response is described by Eq. (5.1.4), is referred to as nonlinear.
Although systems with variable mass are not unusual [1], our discussion will
be limited to systems with constant mass. Besides, the forces fD and fS will
be considered of the form fD ðu_ Þ and fD ðu Þ. Thus, Eq. (5.1.4) becomes
m u€ + fD ðu_ Þ + fS ðu Þ ¼ pðt Þ (5.1.5)
We distinguish two types of nonlinearity: the geometric nonlinearity, which
is due to large displacements implying large deformations of the structure, and
the material nonlinearity, which is due to nonlinear constitutive equations (e.g.,
hyperelastic or elastoplastic materials). Of course, both types of nonlinearity
can simultaneously characterize the response of a system.
The analytical solution of the nonlinear equations of motion is a difficult and
complicated mathematical problem. Exact solutions are available only for a few
cases and for differential equations of a specific form [2, 3]. The existing solu-
tions aim rather at a qualitative study of the response of the system described by
a nonlinear equation than at offering a computational means for practical ana-
lyses. The knowledge of the nonlinear response of the single-degree-of-freedom
(SDOF) systems comes from approximate methods, and mainly from numerical
methods. Therefore, the recourse to numerical methods to solve the nonlinear
equations of motion is inevitable. The step-by-step methods play a dominant
role. The Runge-Kutta methods, usually employed for the solution of nonlinear
equations, belong to these methods [4].
The dynamic response of nonlinear systems can be studied effectively by
demanding the fulfillment of equation motion (5.1.5) at discrete time instants
Dt apart by the use of the step-by-step integration methods we discussed in
Chapter 4. These methods as developed for nonlinear equations of motion
are presented directly below while for the analytical methods, the reader is
advised to look in the vast related literature [2, 3, 5].
Example 5.1.1 Systems with a geometrical nonlinearity
Derive the equation of motion of the system shown in Fig. E5.1. The supports at
A,B and the interconnection at C are hinges. The mass m at C is concentrated.
The system is set to motion by the initial conditions u0 , u_ 0 and/or the vertical
external force pðt Þ. The bars are assumed massless.
 Nonlinear response: Single-degree-of-freedom systems Chapter 5 193

l l

C
A φ B
u(t )
E, A E, A
m
p(t )
fI
S fS S

p(t )
FIG. E5.1 SDOF system with nonlinear response.

Solution
During the motion, the force pðt Þ, the inertial force fI , and the elastic forces of
the bars are in equilibrium
The inertial force is given by the relation

fI ¼ m u€ (1)

The total elastic force is caused by the elongation of the bars and is given by

fS ¼ 2S sin f (2)

where
EA
S¼ d (3)
l
E is the modulus of elasticity of the material of the bar, A its cross-sectional area,
and d the elongation of the bars.
The elongation of the bars at time t is
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
d¼ l 2 + u2  l (4)

and
u
sin f ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi (5)
l 2 + u2
194 PART I Single-degree-of-freedom systems

Hence, the elastic force in Eq. (2) by virtue of Eqs. (3)–(5) is expressed as
2 3
u6 1 7
fS ðu Þ ¼ 2EA 41  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5 (6)
l 2
1 + ðu=l Þ

and the equation of motion (5.1.1) in the absence of damping becomes

2 3
u6 1 7
m u€ + 2EA 41  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5 ¼ pðt Þ (7)
l
1 + ðu=l Þ2

Obviously, the equation of motion is nonlinear. Fig. E5.2 shows the graph
of fS ðu Þ. Because d 2 fS =du 2 > 0 the curve is concave upward. In this system,
the slope kT increases continually, which implies that the elastic force
increases with increasing u. In this case, we say that the system exhibits hard-
ening, in contrast to other systems that exhibit softening. In the latter systems,
the curve fS ðu Þ is concave downward (d 2 fS =du 2 < 0), that is, the slope
kT decreases continually, which implies that elastic force decreases with
increasing u, for example, a system with stiffness fS ¼ 40u  u 3 exhibits soft-
ening; see Fig. E5.3.

FIG. E5.2 System with hardening.

 Nonlinear response: Single-degree-of-freedom systems Chapter 5 195

FIG. E5.3 System with softening.

5.2 The central difference method

We consider the case where the damping force depends linearly on the velocity
(viscous damping) and the elastic force depends nonlinearly on the displace-
ment, then Eq. (5.1.5) becomes
m u€ + cu_ + fS ðu Þ ¼ pðt Þ (5.2.1)
The central difference method (CDM) developed for linear equations can be
readily adjusted to solve this equation explicitly. Thus, using Eqs. (4.2.4),
(4.2.5) to replace the derivatives in Eq. (5.2.1), we obtain
un + 1  2un + un1 un + 1  un1
m 2
+c + fS ðun Þ ¼ pn (5.2.2)
h 2h
or solving for un + 1 gives
K^ un + 1 ¼ P^ (5.2.3)
where
m c
K^ ¼ 2 + (5.2.4)
h 2h
2m m c 
P^ ¼ pn  fS ðun Þ + 2 un  2  un1 (5.2.5)
h h 2h
Eq. (5.2.3) is used to compute un + 1 from the displacements at the two preceding
instants. The value of u1 is computed from Eq. (4.2.10), namely
1
u ðDt Þ  u ð0Þ  Dt u_ ð0Þ + Dt 2 u€ð0Þ (5.2.6)
2
196 PART I Single-degree-of-freedom systems

in which u€0 results now from Eq. (5.2.1) for t ¼ 0

u€0 ¼ ½p0  cu_ 0  fS ðu0 Þ=m (5.2.7)
Obviously, the numerical algorithm presented in Table 4.2.1 can be employed
to write a computer code for the nonlinear response of the SDOF systems after
modifying appropriately to include the nonlinear restoring force fS ðu Þ. On the
basis of this algorithm, a computer program called centr_diff_nlin.m has been
written in MATLAB for the numerical integration of the nonlinear equation of
motion using the CDM. The program is available on this book’s companion
website.
Example 5.2.1 Simple pendulum. CDM solution
Solve the initial value problem (IVP) describing the motion of the simple pen-
dulum (Fig. E5.4) under its own weight, namely
g
q€ + sin q ¼ 0 (1)
l
qð0Þ ¼ q0 (2a)
q_ ð0Þ ¼ q_ 0 (2b)
where qðt Þ represents the angle of the pendulum from the vertical position, l is
its length, and g the acceleration of gravity. The equation of motion (1) can be
readily derived using the method of Lagrange’s equations (see Section 1.8).
Eq. (1) admits an exact solution [6]
npffiffiffi hpffiffiffiffiffiffiffi io
qðt Þ ¼ 2 sin1 k sn g=l ðt + T0 Þ; k (3)

where k ¼ sin 2 ðq0 =2Þ and T0 is the quarter of the period; sn represents the
sn-Jacobean elliptic function [7].
Solution
The response of the pendulum for l ¼ g, q0 ¼ 0:40p, and q_ 0 ¼ 0 is obtained using
the program centr_diff_nlin.m with Dt ¼ 0:01. It is shown in Fig. E5.5 as com-
pared with the exact one.

FIG. E5.4 Simple pendulum.

 Nonlinear response: Single-degree-of-freedom systems Chapter 5 197

t t

FIG. E5.5 Response of the simple pendulum in Example 5.2.1.

5.3 The average acceleration method

The average acceleration method (AAM) presented in Section 4.3 can be mod-
ified to solve the nonlinear equation of motion, Eq. (5.2.1). Although this
method becomes unstable for long-duration motions [9], it is presented here
because it is still widely used for the solution of the nonlinear equation of
motion.
At time t + Dt this equation is written
m u€ðt + Dt Þ + fD ðt + Dt Þ + fS ðt + Dt Þ ¼ pðt + Dt Þ (5.3.1)
Subtracting Eq. (5.2.1) from the foregoing equation yields
mDu€ + DfD + DfS ¼ Dp (5.3.2)

(a) (b)
FIG. 5.3.1 Tangent and secant of (a) damping and (b) stiffness curve.

Expanding fD ðt + Dt Þ and fS ðt + Dt Þ in the Taylor series gives

DfD ¼ fD ðt + Dt Þ  fD ðt Þ
¼ fD ðu_ + Du_ Þ  fD ðu_ Þ
(5.3.3)
df 1 d 2 fD
¼ D Du_ + ðDu_ Þ2 + ⋯
d u_ 2 d u_ 2
198 PART I Single-degree-of-freedom systems

DfS ¼ fS ðt + Dt Þ  fS ðt Þ
¼ fS ðu + Du Þ  fS ðu Þ
(5.3.4)
df 1 d 2 fS
¼ S Du + ðDu Þ2 + ⋯
du 2 du 2
For small values of Dt, the quantities Du, Du_ are also small. Thus, neglect-
ing the nonlinear terms in Eqs. (5.3.3), (5.3.4), we obtain
DfD  cT Du_ (5.3.5)
DfS  kT Du (5.3.6)
Obviously, cT and kT express the slope of the tangent to the curves fD ðu_ Þ
and fS ðu Þ, respectively, at time t.
Referring to Fig. 5.3.1a, we have
fD ðt + Dt Þ ¼ fD ðt Þ + DfD
However, the exact value of DfD is
DfD ¼ Du_ tan fc
where fc is the angle of the secant. Hence, approximating DfD by Eq. (5.3.5)
introduces the error (see Fig. 5.3.1a)
eD ¼ ðcT  tan fc ÞDu_ (5.3.7)
because cT is the slope of the tangent.
Similarly, the use of kT to approximate DfS introduces the error (see
Fig. 5.3.1b)
eS ¼ ðkT  tanfk ÞDu (5.3.8)
The errors eD and eS cannot be avoided because u_ ðt + Dt Þ and u ðt + Dt Þ
are not known at instant t + Dt. However, as we will show, they can be kept
under a given bound, which specifies the accuracy of the solution procedure.
By virtue of Eqs. (5.3.5), (5.3.6), Eq. (5.3.2) is written in incremental form
mDu€ + cT Du_ + kT Du ¼ Dp (5.3.9)
The previous equation is of the form (4.3.11). Hence, the AAM is suitable to
solve it. Thus, using Eqs. (4.3.9), (4.3.10) to express Du€ and Du,
_ we obtain
2
Du_ ¼ Du  2u_ ðt Þ (5.3.10)
Dt
4
Du€ ¼ ½Du  Dt u_ ðt Þ  2u€ðt Þ (5.3.11)
Dt 2
Then Eq. (5.3.9) becomes
   
2cT 4m 4m
kT + + 2 Du ¼ Dp + + 2cT u_ ðt Þ + 2m u€ðt Þ (5.3.12)
Dt Dt Dt
 Nonlinear response: Single-degree-of-freedom systems Chapter 5 199

or
k ∗ Du ¼ Dp∗ (5.3.13)

where
2cT 4m
k ∗ ¼ kT + + 2 (5.3.14)
Dt Dt
and
 
4m
Dp∗ ¼ Dp + + 2cT u_ n + 2m u€n (5.3.15)
Dt

The value of Du obtained from Eq. (5.3.13) is used in Eq. (5.3.10) to eval-
_ Then we obtain
uate Du.
un + 1 ¼ un + Du (5.3.16a)

u_ n + 1 ¼ u_ n + Du_ (5.3.16b)

The acceleration u€n + 1 is evaluated directly from Eq. (5.3.1)

1
u€n + 1 ¼ ½pn + 1  fD ðu_ n + 1 Þ  fS ðun + 1 Þ (5.3.16c)
m
Apparently, the value u€n + 1 obtained from the above equation reduces the
error introduced by cT and kT . The study of the stability of the time step inte-
gration schemes for nonlinear differential equation of motion is an issue beyond
the scope of this book.
The described method gives good results if the slope of the tangent is close
to that of the secant. For this reason, it is recommended to check the following
errors at the end of each step
   
kT  ksec  cT  csec 

ek ¼    a, ec ¼   a (5.3.17)
kT  cT 

where
DfS DfD
ksec ¼ , csec ¼ (5.3.18)
Du Du_

and a is a small specified number defining the upper bound of the error, for
example, a ¼ 0:01. If it is

min fek , ec g > a (5.3.19)

then we must reduce the time step Dt and repeat the computations. Table 5.3.1
presents the algorithm for the numerical solution of the nonlinear equation of
motion (5.1.5) using the AAM.
200 PART I Single-degree-of-freedom systems

TABLE 5.3.1 Average acceleration method. Nonlinear equation of motion.

A. Data
m, fD ðu_ Þ, fS ðu Þ, pðt Þ, u ð0Þ, u_ ð0Þ, ttot , a
B. Initial computations
u€ð0Þ ¼ fpð0Þ  fD ½u_ ð0Þ  fS ½u ð0Þg=m
Select: Dt and set t ¼ 0 and
C. In each step compute
1. kT ¼ ðdf S =du Þt cT ¼ ðdf D =d u_ Þt
2. k ∗ ¼ kT + 2cT =Dt + 4m=Dt 2 , Dp ¼ pðt + Dt Þ  pðt Þ
3. p∗ ¼ Dp + ð4m=Dt + 2cT Þu_ ðt Þ + 2m u€ðt Þ
4. Du ¼ p∗ =k ∗ , Du_ ¼ 2Du=Dt  2u_ ðt Þ
5. u_ ðt + Dt Þ ¼ u_ ðt Þ + Du_
u€ðt + Dt Þ ¼ fpðt + Dt Þ  fD ½u_ ðt + Dt Þ  fS ½u ðt + Dt Þg=m
   
 S =Du  c Df =Du_ 
ek ¼ kT Df
kT , ec ¼  T cTD 
6. Check: if max fec , ek g < a go to the next time step. Else set t ¼ t + Dt and check
if t > ttot end. Else set u ðt Þ ¼ u ðt + Dt Þ, u_ ðt Þ ¼ u_ ðt + Dt Þ, u€ðt Þ ¼ u€ðt + Dt Þ, and
go to C.1

When the damping force depends linearly on the velocity, then the response
is governed by Eq. (5.2.1). In this case, the previously presented incremental
method can be improved by employing an iterative procedure within each step,
which minimizes the error introduced by the tangent stiffness kT .
The starting point is Eq. (5.3.13), which we write as
k ∗ du ð1Þ ¼ Dp∗ (5.3.20)
The quantity du ð1Þ is the first approximation to Dun within the time step
from tn to tn + 1 . That is
Dunð1Þ ¼ du ð1Þ (5.3.21)

FIG. 5.3.2 Graph of the function p*(u).

The index n denotes the number of the step that brings us from the displace-
ment un to un + 1 . Fig. 5.3.2 presents the graph of the function p∗ ðu Þ. Apparently,
 Nonlinear response: Single-degree-of-freedom systems Chapter 5 201

the displacement du ð1Þ resulting from Eq. (5.3.20) decreases Dp∗ by dpð1Þ .
Hence, there is a remaining force
DF ð2Þ ¼ Dp∗  dpð1Þ (5.3.22)
which must be equilibrated. The change dpð1Þ is computed using Eq. (5.3.20), if
the tangential slope kT is replaced with the slope of the secant
"  #
ð1Þ fS un + du ð1Þ  fS ðun Þ 2c 4m
dp ¼ + + du ð1Þ
du ð1Þ Dt Dt 2
   
2c 4m
¼ fS un + du ð1Þ  fS ðun Þ + + 2 du ð1Þ
Dt Dt
which by virtue of Eq. (5.3.14) becomes
 
dpð1Þ ¼ fS un + du ð1Þ  fS ðun Þ + ðk ∗  kT Þdu ð1Þ (5.3.23)

The remaining force DF ð2Þ produces an additional displacement du ð2Þ ,

which is computed from the relation
k ∗ du ð2Þ ¼ DF ð2Þ (5.3.24)
Hence, the new approximation to Dun is
Dunð2Þ ¼ du ð1Þ + du ð2Þ
(5.3.25)
¼ Dunð1Þ + du ð2Þ
Then, it is used to compute the new remaining force
DF ð3Þ ¼ DF ð2Þ  dpð2Þ (5.3.26)
where
   
dpð2Þ ¼ fS un + Dunð2Þ  fS un + Dunð1Þ + ðk ∗  kT Þdu ð2Þ (5.3.27)

The force DF ð3Þ produces the additional displacement du ð3Þ , which is com-
puted from the relation
k ∗ du ð3Þ ¼ DF ð3Þ (5.3.28)
Thus, the new approximation of Dun is
Dunð3Þ ¼ Dunð2Þ + du ð3Þ (5.3.29)
Consequently, for the i + 1 approximation it is
DF ði + 1Þ ¼ DF ðiÞ  dpðiÞ (5.3.30)
   
dpðiÞ ¼ fS un + DunðiÞ  fS un + Dunði1Þ + ðk ∗  kT Þdu ðiÞ (5.3.31)
202 PART I Single-degree-of-freedom systems

k ∗ du ði + 1Þ ¼ DF ði + 1Þ (5.3.32)
Dunði + 1Þ ¼ DunðiÞ + du ði + 1Þ (5.3.33)
Note that for i ¼ 1 it must be set Dunð0Þ ¼ 0 and DF ð1Þ ¼ Dp∗ .
The iteration procedure is terminated after I iterations, if
 
du ði + 1Þ 
 
 e (5.3.34)
 DunðI Þ 

where e is a specified small number. Then we assume that the convergence has
been achieved. The value Dun ¼ DunðI Þ is considered exact and it is used to
compute un + 1 and u_ n + 1 , u€n + 1 . Subsequently, the procedure continues to
the next step. The iterative procedure within the time step from tn to tn + 1 is
summarized in Table 5.3.2. This procedure is known as the modified
Newton-Raphson method.

TABLE 5.3.2 Modified Newton-Raphson method for the minimization

of the error in the average acceleration method.
A. Initial values
ð0Þ ð0Þ
un + 1 ¼ un , fS ¼ fS ðun Þ, DRð1Þ ¼ Dp∗n
B. In each iteration i ¼ 1,2,3,… compute:
1. du ðiÞ ¼ DRðiÞ =k ∗
ði Þ ði1Þ
2. un + 1 ¼ un + 1 + du ðiÞ
ði Þ ði1Þ
3. DpðiÞ ¼ fS  fS + ðk ∗  kT Þdu ðiÞ
ði + 1Þ
4. DR ¼ DR  DpðiÞ
ði Þ

5. du ði + 1Þ ¼ DRði + 1Þ =k ∗
 
6. If du ði + 1Þ =DunðI Þ  > e set i ¼ i + 1 and go to B.2

Adhering to the steps in Table 5.3.1, a computer program called av_acc_

nlin.m has been written in MATLAB for the solution of the nonlinear equation
of motion. The electronic version of the program is available on this book’s
companion website.
Example 5.3.1 Simple pendulum. AAM solution
Solve the IVP describing the motion of the simple pendulum (Fig. E5.4) under
its own weight. The motion is described by Eqs. (1), (2a), (2b) of Example 5.2.1.
Solution
The solution is obtained using the program av_acc_nlin.m with Dt ¼ 0:01. The
response is shown in Fig. E5.5 as compared with the exact one.
 Nonlinear response: Single-degree-of-freedom systems Chapter 5 203

5.4 The analog equation method

The solution procedure developed in Section 4.4 for the linear equation of
motion can be straightforwardly extended to the nonlinear equation. The non-
linear IVP treated here is described by the equations
m u€ + fD ðu_ Þ + fS ðu Þ ¼ pðt Þ (5.4.1)
u ð0Þ ¼ u0 , u_ ð0Þ ¼ u_ 0 (5.4.2)
where fD ðu_ Þ and fS ðu Þ are nonlinear functions of their arguments.
Applying Eq. (5.4.1) for t ¼ nDt, we obtain
mq n + fD ðu_ n Þ + fS ðun Þ ¼ pn , qn ¼ u€n (5.4.3)
Moreover, Eqs. (4.4.14), (4.4.15) hold also, that is,
2 c1 3 2 c1 3
" #( ) " #( ) + 
h 1 u_ n 0 1 u_ n1 6 27 6 2 7
¼ +4 5qn + 4
c2 5 n1
q (5.4.4)
1 0 un 1 0 un1 c2
 +
2 2

TABLE 5.4.1 The analog equation method. Nonlinear equation of motion.

A. Data
Read: m, fD ðu_ Þ, fS ðu Þ, u0 , u_ 0 , pðt Þ, ttot
B. Initial computations
1. Select: h ¼ Dt and compute ntot
2. Compute: c1 ¼ h 2 =2, c2 ¼ h, q0 ¼ ½pðt Þ  fD ðu_ 0 Þ + fS ðu0 Þ=m
C. Compute solution
for n ¼ 1 to ntot solve for fqn u_ n un gT the system of the nonlinear algebraic equations:
mq n + fD ðu_ n Þ + fS ðun Þ ¼ pn
2 c 3 2 c 3


 +
1

1
h 1 u_ n 0 1 u_ n1 6 27 6 2 7
¼ + 4 c 5qn + 4 c 5qn1
1 0 un 1 0 un1 2 2
 +
2 2

Eqs. (5.4.3), (5.4.4) constitute a system of three algebraic equations, one

nonlinear and the other two linear. They can be solved successively for
n ¼ 1, 2, … to yield the solution un and the derivatives u_ n , u€n ¼ qn at instant
t ¼ nh  T . For n ¼ 1, the value q0 appears in the right side of Eq. (5.4.4). This
quantity is readily obtained from Eq. (5.4.1) for t ¼ 0. This yields
q0 ¼ ½pð0Þ  fD ðu_ 0 Þ + fS ðu0 Þ=m (5.4.5)
Several procedures can be applied to solve this system of algebraic equa-
tions. For example, a simple procedure is to substitute un and u_ n from
Eq. (5.4.4) into Eq. (5.4.3) and solve the resulting nonlinear equation for qn .
This can be achieved by employing any ready-to-use subroutine for nonlinear
algebraic equations, for example, MATLAB function fsolve. The steps of the
204 PART I Single-degree-of-freedom systems

solution procedure are presented in Table 5.4.1. Adhering to the steps of this
table, a computer program called aem_nlin.m has been written in MATLAB
for the numerical integration of the nonlinear equation of motion using the
AEM. The program is available on this book’s companion website. It computes
the displacement u ðt Þ, the velocity u_ ðt Þ, and the acceleration u€ðt Þ.
Example 5.4.1 The Duffing equation. AEM solution
Use the AEM to solve the IVP for the Duffing equation
u€ + 0:2u_ + u + u 3 ¼ pðt Þ (1)
u ð 0Þ ¼ 0 (2a)
u_ ð0Þ ¼ 1 (2b)
For
pðt Þ ¼ e0:1t ½ð0:01sin t  0:2cost  sin t Þ  0:2ð0:1 sin t  cos t Þ
+ sin t + e0:2t ð sin t Þ3 
Eq. (1) admits an exact solution uexact ðt Þ ¼ e0:1t sin t.
Solution
The solution is obtained using the program aem_nlin.m with Dt ¼ 0:01. The
graph of the solution is shown in Fig. E5.6 as compared with the exact one.

1
Computed
0.8 Exact
error X103
0.6

0.4

0.2

−0.2

−0.4

−0.6

−0.8
0 5 10 15 20 25
t
FIG. E5.6 Solution u and error u  uexact in Example 5.4.1.
 Nonlinear response: Single-degree-of-freedom systems Chapter 5 205

Example 5.4.2 Response of a nonlinear system with hardening

Use the AEM to study the dynamic response of the SDOF system shown in
Fig. E5.1, when it is subjected to:
(i) An initial displacement u0 ¼ 0:05m from the position of static equilibrium.
(ii) External load p0 ¼ mg suddenly applied at time t ¼ 0.
(iii) External load pðt Þ ¼ 10 sin Wt.
Data: m ¼ 3:8722kN m1 s2 , E ¼ 2:1  108 kN=m2 , A ¼ 3:142  104 m2 ,
g ¼ 9:81m=s2 , l ¼ 3:00m.
Solution
(i) The static displacement ust produced by the weight of the body is calculated
from Eq. (7) of Example 5.1.1 for u€ ¼ 0 and pðt Þ ¼ mg, namely

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2EAðust =l Þ 1  1= 1 + ðust =l Þ2 ¼ mg (1)

which is solved for the data of the problem to yield ust ¼ 0:25m.
The external force is the constant weight of the body, that is, pðt Þ ¼ mg,
while the elastic force
fS ðust + u Þ (2)
where u ¼ u ðt Þ denotes the additional displacement due to the dynamic
response. Thus, the equation of motion becomes
m u€ + fS ðust + u Þ ¼ mg (3)
with initial conditions
u0 ¼ 0:05m, u_ 0 ¼ 0 (4)
It should be noted that due to the nonlinearity of the elastic force, the super-
position of the displacements does not apply. The computed response of the sys-
tem with Dt ¼ 0:01 is shown in Fig. E5.7.
(ii) In this case the equation of motion reads
m u€ + fS ðu Þ ¼ mg (5)
with initial conditions
u0 ¼ 0, u_ 0 ¼ 0 (6)
where u ¼ u ðt Þ denotes the total dynamic displacement from the undeformed
position. The computed response of the system with Dt ¼ 0:01 is shown in
Fig. E5.8.
(iii) In this case, the IVP becomes
m u€ + fS ðu Þ ¼ mg + 10sin Wt (7)
u0 ¼ 0, u_ 0 ¼ 0 (8)
The computed response with Dt ¼ 0:01 is shown in Fig. E5.9 for the ratio
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
W=w ¼ 0:1, where w ¼ EA=lm . In all cases, the results are compared with
those obtained by the CDM
206 PART I Single-degree-of-freedom systems

u(t)

t
FIG. E5.7 Response in Example 5.4.2 (i).
u(t)

t
FIG. E5.8 Response in Example 5.4.2 (ii).

FIG. E5.9 Response in Example 5.4.2 (iii).

 Nonlinear response: Single-degree-of-freedom systems Chapter 5 207

Example 5.4.3 Response of a nonlinear system with softening

Study the dynamic response of a SDOF system with the following data:
m ¼ 1kN m1 s2 , u ð0Þ ¼ 1m, u_ ð0Þ ¼ 1m=s, fS ¼ 2u 3 , and ttot ¼ 1s. Because
d 2 fS=du 2 ¼ 12u < 0 the system exhibits softening. The integration of the
equation of motion will be performed using all three methods, that is, CDM,
AAM, and AEM. The problem admits an exact solution u ðt Þ ¼ 1=ð1 + t Þ.
Solution
The IVP is
u€  2u 3 ¼ 0 (1)
u0 ¼ 1, u_ 0 ¼ 1 (2)
The solution in the interval 0  t  1 for Dt ¼ 0:001 is shown in Fig. E5.10.

FIG. E5.10 Response of the SDOF system in Example 5.4.2.

208 PART I Single-degree-of-freedom systems

Example 5.4.4 Response of an elastoplastic system

In this example, the dynamic response of a system with material nonlinearity is
studied. The one-story building (Fig. E5.11a) is modeled as a SDOF system
with mass m. The structure is subjected to the horizontal rectangular pulse load
pðt Þ of Fig. E5.11b. The material is elastoplastic in nature. This yields the force-
displacement relation shown in Fig. E5.12 due to the elastoplastic hinges pro-
duced by the horizontal load. Neglecting damping, study the response of the
system using both the AAM and AEM. Data: t1 ¼ 0:25s, fSy ¼ 69:68kN,
k ¼ 5360kN=m, m ¼ 36kN m1 s2 , and p0 ¼ 45kN.

(a) (b)
FIG. E5.11 System with a material nonlinearity in Example 5.4.4.

FIG. E5.12 Restoring force in Example 5.4.4.

Solution
It is
fSy 69:68
uy ¼ ¼ ¼ 0:013m ¼ yield displacement
k 5360:00
um ¼ maximum displacement where the velocity changes sign
uR ¼ um  uy ¼ remaining plastic deformation
The equation of motion is
m u€ + fS ðu Þ ¼ pðt Þ
 Nonlinear response: Single-degree-of-freedom systems Chapter 5 209

where fS is the stiffness of the structure. It is

fS ¼ ku if u < uy ¼ 0:013m
fS ¼ fSy  y  u  u m
if u
fS ¼ fSy  k ðum  u Þ if um  2uy < u < um , u_ < 0
Moreover, we have
rffiffiffiffiffi
k 2p
w¼ ¼ 12:20s1 , T¼ ¼ 0:515s
m w

(i) Solution by the AAM.

We employ the AAM following the displacement step by step.
At instant t we have
pðt Þ  fS
u€ ¼ ¼ 0:0278½pðt Þ  fS 
m
4m 4  36
k ∗ ¼ kT + 2 ¼ k + ¼ kT + 1, 440, 000, kT ¼ 0, u > uy
Dt 0:012
The numerical results obtained by the AAM with Dt ¼ 0:01 < T =10 are
shown in Table E5.1 for two different time steps.
(ii) Solution by AEM.
A computer program has been written in MATLAB for the evaluation of the
elastoplastic response of the SDOF system. The obtained numerical results
are shown in Table E5.1 for two different time steps.
Finally, the time history of response of the system as well as of the restoring
force are shown in Fig. E5.13. and Fig. E5.14.

TABLE E5.1 Numerical solution of elastoplastic system in Example 5.4.4.

u ðt Þ (cm)
Dt ¼ 0:01 Dt ¼ 0:001
t (s) AEM AAM AEM AAM

0.00 0.000 0.000 0.000 0.000

0.05 0.124 0.151 0.149 0.151
0.10 0.503 0.550 0.546 0.551
0.15 1.003 1.053 1.050 1.055
0.20 1.446 1.483 1.480 1.484
0.25 1.721 1.745 1.743 1.745

0.30 1.643 1.709 1.677 1.677

Continued
 TABLE E5.1 Numerical solution of elastoplastic system in
Example 5.4.4.—cont’d

u ðt Þ (cm)
Dt ¼ 0:01 Dt ¼ 0:001
t (s) AEM AAM AEM AAM
0.35 1.134 1.234 1.177 1.175
0.40 0.378 0.487 0.423 0.421
0.45 0.353 0.262 0.312 0.313

0.50 0.796 0.744 0.763 0.763

0.55 0.791 0.786 0.768 0.766
0.60 0.341 0.374 0.324 0.321
0.65 0.393 0.343 0.408 0.412
0.70 1.147 1.107 1.164 1.167
0.75 1.649 1.642 1.671 1.673

0.80 1.718 1.757 1.747 1.747

0.85 1.330 1.410 1.363 1.362
0.90 0.624 0.726 0.659 0.658
0.95 0.145 0.049 0.111 0.113
1.00 0.701 0.637 0.663 0.663

FIG. E5.13 Displacement in Example 5.4.4.

 Nonlinear response: Single-degree-of-freedom systems Chapter 5 211

FIG. E5.14 Restoring force in Example 5.4.4.

5.5 Problems
Problem P5.1 Show that the motion of the simple pendulum is governed by
the IVP
g
q€ + sin q ¼ 0, qð0Þ ¼ q0 , q_ ð0Þ ¼ q_ 0 (1)
l
Solve the equation of motion numerically when g=l ¼ 1, q0 ¼ 0:1p, q_ 0 ¼ 0
and compute the period T of the pendulum. Give the graphical representation
of the function T ¼ T ðq0 Þ, 0:1  q0  1 for the two time steps Dt ¼ 0:1 and
Dt ¼ 0:0001. Compare with the exact expression
sffiffiffi Z
l p=2 df sin ðq=2Þ
T ðq 0 Þ ¼ 4 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi , k ¼ sin ðq0 =2Þ, sin f ¼ (2)
g 0 1  k 2 sin 2 f sin ðq0 =2Þ

Use numerical integration to evaluate the Legendre elliptic integral of first

kind in Eq. (2).
Problem P5.2 The support O of the simple pendulum in Fig. P5.2 is subjected
 The rotation about the
to the horizontal harmonic motion u ðt Þ ¼ u0 sin wt.
support is elastically restrained by the spring CR . Study the motion of the
system if g=l ¼ 1, u0 w2 =l ¼ 5, CR ¼ kl 2 =2, q0 ¼ p=6, q_ 0 ¼ 0, u0 ¼ 0:1m,
w ¼ 5s1 .
212 PART I Single-degree-of-freedom systems

FIG. P5.2 Pendulum in problem P5.2.

Problem P5.3 Study the response of the system shown in Fig. P5.3, when
(i) u ð0Þ ¼ 0:05m, u_ ð0Þ ¼ 0, pðt Þ ¼ 0. The initial displacement will be taken
from the position of the static equilibrium.
(ii) u ð0Þ ¼ u_ ð0Þ ¼ 0, pðt Þ ¼ mg ð1  t=t1 ÞH ðt1  t Þ:pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
(iii) u ð0Þ ¼ u_ ð0Þ ¼ 0, P ðt Þ ¼ 15sin wt,
 where w ¼ kmin =m .
(iv) Plot of the dependence of the period T as a function of u0 , T ¼ T ðu0 Þ, if
u_ ð0Þ ¼ 0 and pðt Þ ¼ 0. The initial displacement will be taken from the
position of the static equilibrium.
Data: t1 ¼ 1s, ttot ¼ 5s, m ¼ 10kN m1 s2 , E ¼ 2:1  108 kN=m2 , A ¼ 1cm2 ,
L ¼ 4:0m, and g ¼ 9:81 ms2 . The cables are assumed massless to avoid the
sag due to self-weight [10–12].

FIG. P5.3 System in problem P5.3.

Problem P5.4 Study the response of the system shown in Fig. P5.3 if the elastic
force is approximated by the first three terms of its Tailor series. Compare the
results with those in Problem P5.3 when pðt Þ ¼ mg ð1  t=t1 ÞH ðt1  t Þ and
u ð0Þ ¼ u_ ð0Þ ¼ 0.
Problem P5.5 The response of a nonlinear system is governed by the following
equation, known as the van der Pol equation

u€  m 1  u 2 u_ + u ¼ 0
Plot the solution u ðt Þ, if u ð0Þ ¼ 0, u_ ð0Þ ¼ 0:1 and for (i) m ¼ 0:2 (ii) m ¼ 1:2.
Problem P5.6 The horizontal beams of the frame in Fig. P5.6 have negligible
mass and are flexible while the shear walls having a uniform mass density with
 Nonlinear response: Single-degree-of-freedom systems Chapter 5 213

specific weight g are assumed rigid. Their support on the ground is elastic and it
is expressed by the relation
 
1
MR ¼ CR f  f2 (a)
4
The system is set in motion by the initial conditions u0 ¼ 5cm, u_ 0 ¼ 0. Plot
the dependence of the ratio T=T0 (T0 is the period of the structure resulting
when the nonlinear term in Eq. a is ignored) on the initial displacement and dis-
cuss the influence of the nonlinearity of the elastic supports on the response of
the structure. Data: a ¼ 5m, g ¼ 24kN=m3 , E¼2.1107kN/m2, CR ¼ EI =5a,
and cross-sectional dimensions of the beam 0:20  0:40m2 .

FIG. P5.6 Frame in problem P5.6.

Problem P5.7 The buoy of Fig. P5.7 consists of two massless cones with a base
dimeter 2R and a height h. A concentrated mass m attached at the bottom of the
body keeps the buoy floating at the position shown in the figure. Study the

FIG. P5.7 Floating buoy in problem P5.7.

214 PART I Single-degree-of-freedom systems

vertical motion of the buoy if it is displaced vertically downward from the equi-
librium position by u0 . Data: m ¼ 10kNm1 s2 h ¼ 5m, R ¼ 2m,
u ð0Þ ¼ 0:30m, u_ ð0Þ ¼ 0, and specific weight of the liquid g ¼ 2kN=m3 .
Problem P5.8 The water tower of Fig. P5.8a is subjected to the blast load of
Fig. P5.8b. The response of the structuren is elastoplastic.
h The
iorestoring force in
the elastic branch is given by fS ¼ ku 1 + 1= 1 + ð10u Þ6 =2 (Fig. P5.8c).
The structure is modeled by a SDOF system. Study the response of the structure
in the interval of t 0 ¼ 1s. Data: m ¼ 50kNm1 s2 , k ¼ 2000kN=m, x ¼ 0:07,
p0 ¼ 25kN, t1 ¼ 0:1s, and yield displacement uy ¼ 0:1m.

(b)
(a)

{ }

(c)
FIG. P5.8 Water tower in problem P5.8.
 Nonlinear response: Single-degree-of-freedom systems Chapter 5 215

References and further reading

[1] J.T. Katsikadelis, Derivation of Newton’s law of motion using Galileo’s experimental data,
Acta Mech. 226 (9) (2015) 3195–3204, https://doi.org/10.1007/s00707-015-1354-y.
[2] H. Nayfeh, D.T. Mook, Nonlinear Oscillations, John Wiley & Sons, Inc, New York, 1995.
[3] F. Verhulst, Nonlinear Differential Equations and Dynamical Systems, Springer, Springer-
Verlag, Berlin Heidelberg, 1996.
[4] J.C. Butcher, Numerical Methods for Ordinary Differential Equations, second ed., John Wiley
& Sons Ltd, 2008.
[5] H.T. Davis, Introduction to Nonlinear Differential and Integral Equations, Dover Publications
Inc, New York, 2010.
[6] A. Belendez, C. Pascual, D.I. Mendez, T. Belendez, C. Neipp, Exact solution for the nonlinear
pendulum, Rev. Bras. Ensimo Fis. 29 (4) (2007) 645–648.
[7] M. Abramowitz, I.A. Stegun (Eds.), Handbook of Mathematical Functions, Dover Publica-
tions, New York, 1970.
[8] K.J. Bathe, Conserving energy and momentum in nonlinear dynamics: a simple implicit time
integration scheme, Comput. Struct. 85 (2007) 437–445.
[9] J.T. Katsikadelis, A new direct time integration method for the equations of motion in struc-
tural dynamics, Angew. Math. Mech. 94 (9) (2014) 757–774, https://doi.org/10.1002/
zamm.20120024.
[10] J.W. Leonard, Tension Structures, McGraw-Hill, New York, 1988.
[11] J.T. Katsikadelis, Finite deformation of cables under 3-D loading: an analytic solution, in:
D.E. Beskos, D.L. Karabalis, A.N. Kounadis (Eds.), Proc. of the 4th National Congress on
Steel Structures, Patras, May 24–25, vol. II, 2002, pp. 526–534.
[12] C.G. Tsiatas, J.T. Katsikadelis, Nonlinear analysis of elastic cable-supported membranes, Eng.
Anal. Bound. Elem. 35 (2011) 1149–1158.
Chapter 6

Response to ground motion

and vibration isolation
Chapter outline
6.1 Introduction 217 6.4 Vibration isolation 235
6.2 Equation of motion: Relative 6.4.1 Transmission of force 235
displacement 217 6.4.2 Transmission of motion 237
6.2.1 Response spectra 220 6.5 Problems 240
6.3 Equation of motion in terms References and further reading 243
of the total displacement 230

6.1 Introduction
Structural systems are often excited by the motion of their support. The
response of a structure to support excitation is dynamic even though no external
dynamic loads act on it. The seismic motion of the ground represents a typical
example of support excitation of structures. The study of the response of struc-
tures to earthquake-induced motion is a specific but very important subject of
structural dynamics. It is discussed in depth in books on earthquake engineering
as well as in books on structural dynamics, preparing engineers to design struc-
tures for earthquake-induced motion [1,2]. This book treats the dynamic
response of structures when the excitation force is known. Therefore, the dis-
cussion in this chapter is limited only to the study of the dynamic response
of the SDOF system due to support excitation. Besides, some basic concepts such
as the response spectrum concept, which facilitates the dynamic analysis of struc-
tures due to ground motion, are presented. The general problem of the support
excitation of structures will be examined later when the MDOF (multi-degree-
of-freedom) systems are studied. The transmission of vibrations from the
structure to the fundament and vice versa are also discussed. Illustrative examples
analyzing the response of SDOF systems due to ground motion are presented. The
pertinent bibliography with recommended references for further reading is also
included. The chapter is enriched with problems to be solved.

6.2 Equation of motion: Relative displacement

Fig. 6.2.1 shows a simplified dynamic model of a one-story shear building
whose support moves horizontaly according to a known law described by the

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00006-6

© 2020 Elsevier Inc. All rights reserved. 217
218 PART I Single-degree-of-freedom systems

function ug ðt Þ. The usual method to analyze the motion of the system is to

decompose the total displacement, namely the displacement with respect to a
fixed reference frame, utot , in two components (Fig. 6.2.1)
utot ðt Þ ¼ ug ðt Þ + u ðt Þ (6.2.1)
where u ðt Þ is the relative displacement of the top of the column with respect to
its base.
The deformation of the structure is caused by the relative displacement u ðt Þ.
Therefore, the elastic force fS ðt Þ and the damping force fD ðt Þ depend only on
u ðt Þ and u_ ðt Þ, respectively. However, the inertial force fI ðt Þ depends on the
total acceleration u€tot ðt Þ. Thus, we will have
fI ðt Þ ¼ m u€tot ðt Þ, fS ðt Þ ¼ ku ðt Þ, fD ðt Þ ¼ cu_ ðt Þ (6.2.2)

utot

m ug u
Fixed axis of reference

c m c

k k

(a) ug (t ) (b)

m
fI mutot fD cu

fS ku
(c)
FIG. 6.2.1 SDOF system subjected to ground motion (a), Dynamic model (b), Forces on the
free body (c).

The dynamic equilibrium of the system (see Fig. 6.2.1c) requires

fI ðt Þ + fD ðt Þ + fS ðt Þ ¼ 0
which by virtue of Eq. (6.2.2) becomes
m u€tot + cu_ + ku ¼ 0 (6.2.3)
or using Eq. (6.2.1) we have
m u€ + cu_ + ku ¼ m u€g (6.2.4)
 Response to ground motion and vibration isolation Chapter 6 219

Eq. (6.2.4) is an equation of forced motion, where the excitation function

is m u€g . In this case, the excitation function is called effective load and is
denoted by
pðt Þ ¼ m u€g ðt Þ (6.2.5)
We observe that the influence of the ground motion on the structure does not
depend directly on the ground displacement ug ðt Þ, but on its acceleration u€g ðt Þ.
Therefore, the dynamic response of the structure due to ground motion can be
studied if the accelerogram of the seismic motion is known, namely if there is a
record of the function u€g ðt Þ during the earthquake (see Fig. 6.2.2). It must be
emphasized that the initial conditions u ð0Þ and u_ ð0Þ for the relative displace-
ment do not vanish if ug ð0Þ 6¼ 0 or u_ g ð0Þ 6¼ 0, even though the system is at rest
at time t ¼ 0. Indeed, if
utot ð0Þ ¼ 0, u_ tot ð0Þ ¼ 0

max ug,tt=258.59
ug,tt (cm/s2)

t
FIG. 6.2.2 Accelerogram from Athens earthquake, Sept. 7, 1999 (Recorded by ITSAK, Α399-1.
V2, longitudinal component, max u€g ¼ 258:59cm=s2 ).

then Eq. (6.2.1) gives

u ð0Þ ¼ ug ð0Þ, u_ ð0Þ ¼ u_ g ð0Þ (6.2.6)
Eq. (6.2.4) can also be written as
m u€ + 2mwx u_ + ku ¼ m u€g ðt Þ (6.2.7a)
or
u€ + 2xwu_ + w2 u ¼ u€g ðt Þ (6.2.7b)
220 PART I Single-degree-of-freedom systems

Apparently, Eq. (6.2.7b) states that the deformation of the system due to
given ground acceleration u€g ðt Þ depends only on the natural frequency w, hence
on the natural period T ¼ 2p=w, and on the damping ratio x, that is,
u ¼ u ðt, T , x Þ. Consequently, two systems with the same natural period T
and the same damping ratio x will undergo the same displacement u ðt Þ under
the same ground motion, in spite of the fact that the two systems may have dif-
ferent masses or different stiffnesses.
The negative sign in the effective load pðt Þ ¼ m u€g ðt Þ affects only the
direction of the displacement and not its magnitude. In practice, this has little
significance inasmuch as the engineer is usually interested in the maximum
absolute value of u ðt Þ. Therefore, the sign can be omitted in this case. This
assumption allows us to write the Duhamel integral in the forma
1
u ðt Þ ¼ U ðt Þ (6.2.8)
wD
where
Z t
U ðt Þ ¼ u€g ðτÞexwðtτÞ sin wD ðt  τÞdτ (6.2.9)
0
For a given ground motion and a fixed damping ratio, we can evaluate the
largest absolute value of the function U ðt Þ, hence of u ðt Þ, for an interval of
values of the natural period T of the damped SDOF system and plot the curves
u ðT , xÞ ¼ max t ju ðt, T , xÞj for discrete values of the damping ratio x. Fig. 6.2.3
shows the curves u ðT , xÞ ¼ U ðT , x Þ=wD , 0 < T  2 for different values of x,
when the ground motion is induced by the Athens earthquake, Sept. 7, 1999.
The respective accelerogram is shown in Fig. 6.2.2. The curves u ðT , x Þ were
obtained by direct solution of Eqs. (6.2.7a), (6.2.7b). The solution can be
obtained using either the analytic solution presented in Section 3.5.4 or numer-
ically using any of the methods presented in Chapter 4.

6.2.1 Response spectra

The curve u ðT , x Þ ¼ max t ju ðt, T , x Þj is called the response spectrum of the
relative displacement or the deformation response spectrum. The response
spectrum, introduced by M. A. Biot in 1932 [3], can be used as a practical means
to study the effect of the ground motion on structures. It is clear that the defor-
mation response spectrum permits the direct evaluation of the absolute maxi-
mum value of the relative displacement of the SDOF system for a specified
excitation without solving the equation of motion, if its natural period T and
damping ratio x are given.

pffiffiffiffiffiffiffiffiffiffiffiffi
a. In Eqs. (6.2.8) and (6.2.9), we can set wD ¼ w 1  x 2  w, because in real structures the value of
the damping ratio is small (x ¼ 3%  15%, hence x2 ≪1) and the error due to this approximation is
much smaller than that due to the uncertainty of the determination of u€g ðt Þ.
 Response to ground motion and vibration isolation Chapter 6 221

u T

T (s)
FIG. 6.2.3 Deformation response spectrum u ðT , x Þ for Athens earthquake, Sept. 7, 1999.

Similarly, we define the relative velocity spectrum and the acceleration

response spectrum as the graphs of the curves u_ ðT , xÞ ¼ max t ju_ ðt, T , x Þj
and u€tot ðT , xÞ ¼ max t ju€tot ðt, T, x Þj, respectively.
The expression of the relative velocity results by direct differentiation of
Eq. (6.2.9) with respect to time t. Because t appears as a parameter in the limits
of the integral, the derivative can be obtained using Leibnitz’s rule [4].
Thus, we obtain
Z t
u_ ðt Þ ¼ wxu ðt Þ + u€g ðτÞexwðtτÞ cos wD ðt  τÞdτ (6.2.10)
0

which by virtue of Eq. (6.2.8) is written

x
u_ ðt Þ ¼  pffiffiffiffiffiffiffiffiffiffiffiffi U ðt Þ + U ðt Þ (6.2.11)
1  x2
where it was set
Z t
U ðt Þ ¼ u€g ðτÞexwðtτÞ cos wD ðt  τÞdτ (6.2.12)
0

Differentiation of the expression (6.2.10) gives the relative acceleration

and the total acceleration is obtained from Eq. (6.2.1), which gives
u€tot ðt Þ ¼ u€g ðt Þ + u€ðt Þ. However, the total acceleration can be obtained directly
from Eq. (6.2.7b), that is
u€tot ðt Þ ¼ 2xwu_ ðt Þ  w2 u ðt Þ
222 PART I Single-degree-of-freedom systems

which by virtue of Eqs. (6.2.8), (6.2.11) becomes

w  
u€tot ðt Þ ¼ pffiffiffiffiffiffiffiffiffiffiffiffi 2x2  1 U ðt Þ  2xwU ðt Þ (6.2.13)
1  x2
The graphs of the absolute maximum of the functions (6.2.11) and (6.2.13),
that is, u_ ðT , xÞ ¼ max t ju_ ðt, T , x Þj and u€tot ðT , xÞ ¼ max t ju€tot ðt, T , x Þj, in an
interval of the undamped natural period T for a fixed value of the damping ratio
x of the SDOF systems give the response spectra of the relative velocity and the
(total) acceleration. Figs. 6.2.4 and 6.2.5 show these response spectra for the
accelerogram of Fig. 6.2.2. The curves were obtained by direct solution of
Eq. (6.2.7b). It should be noted that the quantities u ðT , xÞ, u_ ðT , xÞ, and
u€ðT , xÞ do not occur at the same time instant.
The shear force at the base of the SDOF system of Fig. 6.2.1 is equal to the
elastic force, that is,
Qo ¼ fS ¼ ku
or taking into account that k ¼ mw2 and using Eq. (6.2.8) we obtain
w
Qo ¼ m pffiffiffiffiffiffiffiffiffiffiffiffi U ðt Þ (6.2.14)
1  x2
Obviously, the maximum value of the base shear force is
w
max Qo ¼ m pffiffiffiffiffiffiffiffiffiffiffiffi U ðT , xÞ (6.2.15)
1  x2

T (s)
FIG. 6.2.4 Response spectrum of the relative velocity u_ ðT , xÞ for the Athens Earthquake,
Sept. 7, 1999.
 Response to ground motion and vibration isolation Chapter 6 223

T (s)
FIG. 6.2.5 Response spectrum of the total acceleration u€tot ðT , xÞ for the Athens earthquake,
Sept. 7, 1999.

For small values of the damping ratio (say, 0  x  0:15), we may set x 2  0.
Thus, the last two equations become
Qo ¼ mwU ðt Þ (6.2.16)
max Qo ¼ mwU ðT , x Þ (6.2.17)
We observe that the quantity wU ðT , x Þ in Eq. (6.2.17) has dimensions of
acceleration. In earthquake engineering, this quantity is designated by
Spa ðT , x Þ and it is known as spectral pseudoacceleration. It is an important
quantity because it allows direct evaluation of the maximum elastic force (base
shear force) in the SDOF system from the graph of Spa ðT , xÞ.
The quantity Spv ðT , x Þ ¼ U ðT , x Þ has dimensions of velocity and it is
known as spectral pseudovelocity. Actually, the quantities Spv ðT , x Þ,
Spa ðT , x Þ are different from u_ ðT , xÞ and u€tot ðT , x Þ. Therefore, they should
not be confused. Nevertheless, it is Spa ðT , xÞ ¼ u€tot ðT , xÞ, if x ¼ 0. Indeed,
Eq. (6.2.3) for x ¼ 0 becomes
u€tot ¼ w2 u ¼ wU ðt Þ (6.2.18)
from which we obtain
u€tot ðT , x Þ ¼ Spa ðT , x Þ (6.2.19)
The deviation of the pseudoacceleration from the extreme value of the total
acceleration is small for small values of x (say 0  x  0:1). Thus we may set
Spa ðT , x Þ  u€tot ðT , x Þ. This is shown in Fig. 6.2.6.
224 PART I Single-degree-of-freedom systems

(utot),tt(T,0)
Spa(T,0)

T (s)
FIG. 6.2.6 Response spectra Spa ðT , x Þ and u€tot ðT , x Þ, ðx ¼ 0Þ.

For the sake of uniform notation, we set Sd ðT , x Þ ¼ u ðT , xÞ. It is obvious

that if one of the quantities Sd ðT , x Þ, Spv ðT , xÞ, Spa ðT , xÞ is known, the remain-
ing quantities result immediately. Thus, if Sd ðT , xÞ is known then we obtain
Spv ðT , xÞ ¼ wSd ðT , x Þ (6.2.20)
Spa ðT , x Þ ¼ wSpv ðT , xÞ (6.2.21)
The above relations are valid because the resulting values refer to the same
time and even at the instant where the maximum displacement Sd ðT , xÞ occurs.
Eq. (6.2.17) can be written as
max Qo ¼ mS pa ðT , x Þ
Spa ðT , xÞ (6.2.22)
¼W
g
where W is the weight of the SDOF system and g the acceleration of gravity.
The ratio
Spa ðT , x Þ
e¼ (6.2.23)
g
referred to as the elastic seismic coefficient is a quantity specified in the earth-
quake codes and it is used to evaluate the maximum equivalent shear force at the
base of a SDOF shear building.
 Response to ground motion and vibration isolation Chapter 6 225

The response spectra of the displacement, pseudovelocity, and pseudoacce-

leration contain the same information for the dynamic response of the system
because each of them results from another by multiplying or dividing it by a
constant number. The reason for using the three spectra is that each of them
refers to a different natural quantity. We saw that Sd ðT , x Þ expresses the
extreme value of the relative displacement while Spa ðT , x Þ is employed to
evaluate the shear force at the base of the one-story shear building. Finally,
Spv ðT , x Þ is used to calculate the extreme value of the energy of the
system during the earthquake. Indeed, the maximum value of the energy of
deformation is
1
E ðT , xÞ ¼ kSd ðT , x Þ2
2
1 Spv ðT , xÞ2
¼ k (6.2.24)
2 w2
1
¼ mS pv ðT , xÞ2
2
Fig. 6.2.7 presents the graphs of Sd ðT , x Þ, Spv ðT , x Þ, and Spa ðT , xÞ for the
Athens earthquake, Sept. 7, 1999. In earthquake engineering, the three plots are
usually presented in a single graphical representation using a logarithmic scale
for the four axes [2].
All previous results have been obtained by direct solution of the equation of
motion. For this reason, a computer program has been written in MATLAB and
given the name response_spectrum_aem.m. The program evaluates the spec-
tra of the deformation, relative velocity, and acceleration and makes their
graphs. The program uses the numerical method presented in Section 4.4.
The user should provide the one-dimensional array containing the values of
the accelerogram at the respective times. The electronic version of the program
is given on this book’s companion website.
Example 6.2.1 Fig. E6.1a shows a cement silo of square plan form supported on
four identical columns. The silo is made of reinforced concrete. When the silo is
empty, a horizontal load P in the x direction is applied, which causes a horizon-
tal displacement of the silo equal to 5cm. Then, the force is suddenly removed
and the system starts to vibrate. After time to the silo has performed six com-
plete vibrations and the amplitude was reduced to 1cm. Determine:
(i) The damping ratio x.
(ii) The damping coefficient c, the damped natural frequency wD , and the
time to
(iii) The displacement after 40 and 60s.
(iv) The maximum shear force max Q and the maximum bending moment
max M of the columns, when the silo is full and is subjected to the support
 of duration t1 ¼ 3s in the absence of damping.
excitation ug ¼ uo sin wt
Determine also the dynamic magnification factor D.
226 PART I Single-degree-of-freedom systems

Sd

(a)

(b)
Spa

(c)
FIG. 6.2.7 Response spectra for the Athens earthquake, Sept. 7, 1999 (x ¼ 0:1).
 Response to ground motion and vibration isolation Chapter 6 227

5m
z
2.5 m
x utot
1m P ug u
m c
10 m
k
x

(a) (b)
FIG. E6.1 Silo in Example 6.2.1

Data: Specific weight of cement g c ¼ 17kN=m3 , specific weight of reinforced

concrete g b ¼ 24kN=m3 , uo ¼ 1:7mm, w ¼ 2:5s1 , modulus of elasticity of
the reinforced concrete E ¼ 2:1  107 kN=m2 , cross-sectional area of columns
0:35  0:35m2 , and thickness of the silo walls and bottom 0:20m. The axial defor-
mation of the columns is ignored while their mass is assumed lumped at their ends.
Solution
The walls of the silo do not deform inside their plane. Therefore, the silo
behaves in actual fact as a rigid body. Because the axial deformation of the col-
umns is neglected, the only possible motion of the structure in the xz plane is the
horizontal one along the x axis. This is shown in the dynamic model of the sys-
tem in Fig. E6.1b.
(i) Τhe damping ratio x is computed from the relation (2.3.24) for n ¼ 6, t ¼ 0,
u ð0Þ ¼ 5cm, u ðnT Þ ¼ 1cm. This yields
 
x 1 5
pffiffiffiffiffiffiffiffiffiffiffiffi ¼ ‘n
1x 2 2  6p 1
which gives x ¼ 4:265%.
(ii) The computation of c and wD requires the determination of the mass m and
the stiffness k of the structure
 2   24
m¼ 5:0  4:62  2:5 + 4:62  0:2=0:928 + 4  0:352  10=2 
9:81
¼ 40:6kNm1 s2

12EI 12  2:1  107  0:354 =12

k ¼4 ¼ 4 ¼ 1260:5kN=m
h3 103
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
w ¼ 1260:5=40:6 ¼ 5:57s1
228 PART I Single-degree-of-freedom systems

Hence
c ¼ 2mwx ¼ 2  40:6  5:57  0:04265 ¼ 19:3kNm1 s
qffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
wD ¼ w 1  x 2 ¼ 5:57 1  0:042652 ¼ 5:56s1
2p
to ¼ 6T ¼ 6 ¼ 6:78s
wD

(iii) The system performs free damped vibrations. The displacement u ðt Þ is

computed from Eq. (2.3.18) for u ð0Þ ¼ 0:05m, u_ ð0Þ ¼ 0, x ¼ 0:04265,
w ¼ 5:57s1 , wD ¼ 5:56s1 . Thus we have
u ðt Þ ¼ e0:238t ð5cos 5:56t + 0:214 sin 5:56t Þ  102 (1)
which yields
u ð40Þ ¼ 0:282  103 cm, u ð60Þ ¼ 0:268  105 cm
(iv) When the silo is filled with cement, we have
  17
m ¼ 40:6 + 4:62  2:5 + 4:62  0:8=3  ¼ 142:1kNm1 s2
9:81
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
w ¼ 1260:5=142:1 ¼ 2:97834s1

The motion is examined in two phases.

Phase Ι (0  t  3). The motion is forced vibration. The equation of motion
is given by Eq. (6.2.4) for c ¼ 0, that is
m u€ + ku ¼ pðt Þ (2)
The effective force is
 ¼ 142:1  1:7  103  2:52 sin wt
pðt Þ ¼ mu o w2 sin wt  ¼ 1:5098 sin wt

The displacement is given by Eq. (3.2.8). Because the system is at rest at
t ¼ 0, it is utot ð0Þ ¼ u_ tot ð0Þ ¼ 0. Hence
u ð0Þ ¼ utot ð0Þ  ug ð0Þ ¼ 0
and
 
 o ¼ 2:5  1:7  103 ¼ 4:25  103 m=s
u_ ð0Þ ¼ u_ tot ð0Þ  u_ g ð0Þ ¼ wu
These initial conditions yield
u_ ð0Þ Po b
A¼  and B ¼ 0
w k 1  b2
Hence
u_ ð0Þ po 1
u I ðt Þ ¼ sin wt +   b sin wt Þ
ð sin wt
w k 1  b2
 Response to ground motion and vibration isolation Chapter 6 229

 ¼ 2:5=2:98 ¼ 0:8389, po ¼ 1:5098kN,

The above relation for b ¼ w=w
k ¼ 1260:5kN=m, and u_ ð0Þ ¼ 4:25  103 m=s yields
uI ðt Þ ¼ 103  ð4:8303 sin wt  4:0545sin wt
Þ
u_ Ι ðt Þ ¼ 102  ð1:4386 cos wt  1:0136 cos wt
Þ
Phase II (t  3). The motion is free vibration and the displacement is given
by Eq. (2.2.13). Thus we have
u_ I ð3Þ
uII ðt Þ ¼ sin wet + uI ð3Þ cos wet , et ¼ t  3  0 (3)
w
The initial conditions for this phase are
uI ð3Þ ¼ 1:5309  103 m, u_ I ð3Þ ¼ 1:620884630  102 ms1
which give
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ffi
u_ ð3Þ 2
max uII ¼ r ¼ + u_ ð3Þ2 ¼ 5:65341  103 m (4)
w
Fig. E6.2 shows the time history of the response ratio Rðt Þ ¼ u ðt Þ=ðpo =k Þ.
We observe that the max u takes place in phase II. Hence the dynamic load fac-
tor is
max uII
D¼ ¼ 4:7199 (5)
p0 =k
R(t)

t (s)
FIG. E6.2 Time history of the response ratio Rðt Þ in Example 6.2.1.
230 PART I Single-degree-of-freedom systems

and
12EI k
max Q ¼ 3
max u ¼ max uII ¼ 1:781kN (6a)
h 4
h
max M ¼ max Q ¼ 8:907kNm (6b)
2

6.3 Equation of motion in terms of the total displacement

In certain cases, it is convenient to formulate the equation of motion in terms of
the total displacement utot ðt Þ. Then, the elastic force and damping force will be
expressed as
   
fS ¼ k utot  ug , fD ¼ c u_ tot  u_ g (6.3.1)
The dynamic equilibrium yields the following equation of motion
   
m u€tot + c u_ tot  u_ g + k utot  ug ¼ 0 (6.3.2)
or
m u€tot + cu_ tot + ku tot ¼ pðt Þ (6.3.3)
where now
pðt Þ ¼ cu_ g + ku g (6.3.4)
Eq. (6.3.3) is suitable to study the dynamic response when the u€g does not
exist in the classical sense or the supports of the structure undergo different
excitations (see Example 6.3.2).
Example 6.3.1 The support of a SDOF system, whose dynamic model is shown
in Fig. E6.3a, undergoes a sudden constant displacement ug ¼ u0 . Determine the
dynamic response of the system when utot ð0Þ ¼ u_ tot ð0Þ ¼ 0.

(a) (b)
FIG. E6.3 SDOF system in Example 6.3.1
 Response to ground motion and vibration isolation Chapter 6 231

Solution
The function representing the ground motion is shown in Fig. E6.3b. The prob-
lem can be solved by the following two ways:
(a) Formulating the equation of motion in terms of the total displacement, that
is, Eq. (6.3.3). This yields
m u€tot + ku tot ¼ ku 0 (1)
This implies that an effective constant excitation force pðt Þ ¼ ku 0 is
suddenly applied. The solution of Eq. (1) is obtained from Eq. (3.4.3)
for p0 ¼ ku 0 . Hence
pffiffiffiffiffiffiffiffiffi
utot ðt Þ ¼ u0 ð1  cos wt Þ, w ¼ k=m (2)
We observe that
ðutot Þmax ¼ 2u0
The relative displacement is
u ðt Þ ¼ utot  ug ¼ u0 cos wt (3)
which yields an elastic force
fS ¼ ku ¼ ku 0 cos wt (4)
and
max jfS j ¼ ku 0 (5)
(b) Formulating the equation of motion in terms of the relative displacement
u ðt Þ. In this case, we have
ug ð0Þ ¼ u0 , u_ g ð0Þ ¼ 0, u€g ð0Þ ¼ 0
Consequently
pðt Þ ¼ 0, u ð0Þ ¼ utot ð0Þ  ug ð0Þ ¼ u0 , u_ ð0Þ ¼ u_ tot ð0Þ  u_ g ð0Þ ¼ 0
and the equation of motion becomes
m u€ + ku ¼ 0 (4)
with initial conditions u ð0Þ ¼ u0 , u_ ð0Þ ¼ 0.
The solution of Eq. (4) is obtained from Eq. (2.2.13) as
u ðt Þ ¼ u0 cos wt (5)
which is identical with that given by Eq. (3).
Example 6.3.2 The supports of the columns of the one-story frame of Fig. E6.4
are subjected to the displacements ug1 ¼ uo sin ðwt   qÞ and ug2 ¼ uo sin wt. 
Determine the equation of motion of the structure and give the expressions
of the relative displacement u ðt Þ and the stress resultants Q ðt Þ, M ðt Þ at the
top cross-sections of the columns. The dead load of the rigid beam is included
232 PART I Single-degree-of-freedom systems

utot (t ) utot (t )
p

EI
h
35 × 70 40 × 80

ug 1 ug 2
L
FIG. E6.4 One-story frame in Example 6.3.2

in the load p. The material of the columns is reinforced concrete. Data: specific
weight of concrete g ¼ 24kN=m3 , u0 ¼ 0:02m, x ¼ 0:05, w ¼ 2:5rad=s,
u ð0Þ ¼ 0, u_ ð0Þ ¼ 0, L ¼ 15m, h ¼ 7m, E ¼ 2:1  107 kN=m2 , and
p ¼ 200kN=m. The mass of the columns is assumed lumped at their ends.
Solution
The system has one degree of freedom. The equation of motion with respect to
the total displacement utot ðt Þ is
       
m u€tot + c1 u_ tot  u_ g1 + c2 u_ tot  u_ g2 + k1 utot  ug1 + k2 utot  ug2 ¼ 0 (1)
or
m u€tot + cu_ tot + ku tot ¼ pðt Þ (2)
where
c ¼ c1 + c2 , k ¼ k1 + k2 (3)
p ¼ c1 u_ g1 + c2 u_ g2 + k1 ug1 + k2 ug2 (4)
The mass of the system is
200  15 + ð0:35  0:70 + 0:40  0:80Þ  ð7=2Þ  24
m¼ ¼ 310:6kNm1 s2
9:81
The stiffness of the columns
12EI 1 12  2:1  107  0:70  0:353
k1 ¼ ¼ ¼ 1837:5kN=m
h3 12  73
3EI 2 3  2:1  107  0:80  0:403
k2 ¼ ¼ ¼ 783:7kN=m
h3 12  73
Hence the stiffness of the system is
k ¼ k1 + k2 ¼ 2621:2kN=m
 Response to ground motion and vibration isolation Chapter 6 233

The natural frequency is

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
w ¼ 2621:2=310:6 ¼ 2:905s1
and the natural period
T ¼ 2p=w ¼ 2:163s
The damping coefficient c is
c ¼ 2mwx ¼ 2  310:6  2:905  0:05 ¼ 90:23kNm1 s
and the damped frequency
qffiffiffiffiffiffiffiffiffiffiffiffi
wD ¼ w 1  x2 ¼ 2:901s1
The damping coefficients c1 and c2 that contribute to the effective force can-
not be directly determined. However, they can be estimated by setting
c1 ¼ ac and c2 ¼ ð1  a Þc, 0  a  1
Apparently, the cases a ¼ 0 and a ¼ 1 are not realistic.
For the data of the problem, Eq. (2) becomes
310:6u€tot + 90:23u_ tot + 2621:2utot ¼ pðt Þ (5)
where
pðt Þ ¼ 36:75sin ðwt
  qÞ + 15:67sin wt

(6)
+ 1:80w½a cos ðwt
  qÞ + ð1  a Þ cos wt

Eq. (5) can be solved using the analytic method presented in Chapter 3.
Figs. E6.5 and E6.6 show the graphs of utot ðt Þ for various values of the phase
angle q and the allocation coefficient a. Fig. E6.7 shows the influence of a on
the maximum displacement max |utot ðt Þ|.
utot(t) (m)

t
FIG. E6.5 Total displacement for different values of the phase angle q in Example 6.3.2 (a ¼ 0:2).
234 PART I Single-degree-of-freedom systems

utot(t) (m)

t
FIG. E6.6 Total displacement for different values of the allocation coefficient a in Example 6.3.2
(q ¼ 3p=4).
max|utot(t)|

a
FIG. E6.7 Extreme value max|utot ðt Þ| versus the allocation coefficient a in Example 6.3.2
(x ¼ 0:05).

From the study of the numerical results, we may draw the following
conclusion:
The percentage allocation of the damping to the two columns does not affect
significantly the dynamic response of the structure. Consequently, an arbitrary
but reasonable allocation, for example, 0:4 < a < 0:6, allows treating practical
cases of asynchronous support excitations.
 Response to ground motion and vibration isolation Chapter 6 235

The stress resultants Q ðt Þ and M ðt Þ at the top cross-sections of the columns

are evaluated from the relations.
12EI 1    
Q1 ðt Þ ¼ 3
utot  ug1 ¼ 1837:5 utot  ug1
h
3EI 2    
Q2 ðt Þ ¼ 3 utot  ug2 ¼ 783:7 utot  ug2
h
6EI 1    
M1 ðt Þ ¼ 2 utot  ug1 ¼ 6431:2 utot  ug1
h
3EI 2    
M2 ðt Þ ¼ 2 utot  ug2 ¼ 5485:9 utot  ug2
h

6.4 Vibration isolation

Very often, machines that generate vibrations are mounted on our structures.
These vibrations, which are transferred to the supporting structure (soil, foun-
dation), are annoying or even harmful and need to be minimized. This is
achieved by inserting spring and damping devices between the machine and
its foundation. An inverse problem arises when vibrations from the environment
(ground support) are transferred to the structure. This problem occurs when sen-
sitive instruments are placed on top or inside moving structures (earthquake
motion of a building, motion of a car on a rough road, spacecrafts, airplanes,
ships, etc.) or special structures near vibration generation sources (monuments).
In the first situation, we talk about the transmission of force while the second is
about the transmission of motion.

6.4.1 Transmission of force

Let us consider first the force isolation problem. Fig. 6.4.1 presents a SDOF
system consisting of a mass m mounted on the fundament by means of a system
of springs and dampers having total stiffness k and total damping c. A vertical
 is applied to the mass. The transfer of force is
harmonic force p ¼ p0 sin wt

FIG. 6.4.1 Force transmitted to the foundation.

236 PART I Single-degree-of-freedom systems

examined when the motion is in the steady state phase. Hence, the vertical dis-
placement is given by Eq. (3.2.26), that is,
  qÞ
u ðt Þ ¼ rsin ðwt (6.4.1)
where

p0 h 2 i 1
2
r¼ 1  b2 + ð2xb Þ2 (6.4.2a)
k
 
2xb
q ¼ tan 1 
, b ¼ w=w (6.4.2b)
1  b2
The total force transmitted to the foundation is
f ¼ fS + fD
¼ ku + cu_
(6.4.3)
  qÞ + crw cos ðwt
¼ krsin ðwt   qÞ
  q  fÞ
¼ fT sin ðwt
where
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
fT ¼ r k 2 + ðcwÞ2 (6.4.4)
 
1 cw 
f ¼ tan (6.4.5)
k
Using Eqs. (6.4.2a), (6.4.2b) for the expression of r and c=k ¼ 2x=w, the
previous relations are written as
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u 1 + ð2xbÞ2
fT ¼ p0 t 2 (6.4.6)
1  b2 + ð2xb Þ2

f ¼ tan 1 ð2xb Þ (6.4.7)

The ratio TR ¼ fT =p0 is called the transmission ratio or transmissibility. It
provides a measure of the force transmitted from the vibrating machine to the
fundament. Fig. 6.4.2 shows the graph of the function TRðb, x Þ for various
values ofpffiffix.ffi We observe that the reduction of the transmitted force is achieved
for b > 2. It is also apparent that the reduction of the ratio TR increases with
decreasing x. Theoretically, the value x ¼ 0 gives the minimum value of the
transmission ratio. However, this is not realistic for two reasons: first, because
the actual isolation systems have damping and second, some damping is desir-
able to avoid adverse results at the startup of the machine as it passes through the
resonance frequency.
 Response to ground motion and vibration isolation Chapter 6 237

TR=fT /p0

FIG. 6.4.2 Transmissibility for harmonic excitation.

6.4.2 Transmission of motion

In this case, the problem is to determine the motion that is transmitted from the
ground to the machine and in general to the structure supported on the ground.
Fig. 6.4.3 shows a mass supported on the ground by means of a system of
springs and dampers with total stiffness k and damping c. The vertical ground
motion is harmonic

ug ¼ u0 sin wt (6.4.8)

FIG. 6.4.3 Ground motion transmitted to the structure.

The equation of motion with respect to the relative displacement is


m u€ + cu_ + ku ¼ m w2 u0 sin wt (6.4.9)
The relative displacement in the steady state is given by Eq. (3.2.25) for
p0 ¼ m w2 u0 . That is
238 PART I Single-degree-of-freedom systems

m w2 u0 1   
u ðt Þ ¼  2   2xb cos wt
1  b2 sin wt 
k 1  b + ð2xbÞ
2 2

   (6.4.10)
u0
¼    2xb3 cos wt
1  b2 b 2 sin wt 
2 2 2
1  b + ð2xbÞ
and the total displacement
utot ¼ ug + u
u0   
 +
¼ u0 sin wt  2
  2xb3 cos wt
1  b2 b 2 sin wt 
1  b2 + ð2xbÞ2
u0 nh i o
¼  2
1  b2 + ð2xb Þ2 sin wt 
  2xb 3 cos wt
1  b2 + ð2xbÞ2
  Þ
¼ uT sin ðwt
(6.4.11)
where
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u 1 + ð2xbÞ2
uT ¼ u0 t 2 (6.4.12)
1  b2 + ð2xb Þ2
!
1 2xb3
¼ tan   (6.4.13)
1  b 2 + ð2xb Þ2
From Eqs. (6.4.8), (6.4.12), we obtain
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
ðu€tot Þmax ðutot Þmax uT u 1 + ð2xbÞ2
TR ¼   ¼   ¼ ¼ t 2 (6.4.14)
u€g max ug max u0 1  b2 + ð2xb Þ2

We observe that the transmissibility of the amplitude of the ground har-

monic motion is the same as that of the force transmissibility. The same is valid
for the extreme values of the accelerations.
Example 6.4.1 A sensitive instrument of weight W ¼ 30kN is installed at a
place where the environment exerts vertical harmonic motion with a cyclic fre-
quency f ¼ 15Hz and maximum amplitude u0 ¼ 0:02m. For the insulation of
the motion, the instrument is placed on an elastic layer with stiffness
k ¼ 28kN=m and damping x ¼ 0:1. Determine (i) the acceleration transmitted
to the instrument and (ii) If the smooth functioning of the instrument requires
that the acceleration applied to it is less than 0:1g, indicate how to ensure that by
using the same elastic layer.
Solution
(i) Computation of the transmission ratio TR
The mass and the natural frequency of the instrument are
 Response to ground motion and vibration isolation Chapter 6 239

W 30kN
m¼ ¼ ¼ 3:058kNm1 s2
g 9:81m=s2
rffiffiffiffiffi rffiffiffiffiffiffiffiffiffiffiffi
k 28
w¼ ¼ ¼ 3:026s1
m 3:058
The circular frequency of the harmonic motion and the maximum
acceleration are
w ¼ 2pf ¼ 94:248s1
   
u€g max ¼ w2 u0 ¼ 177:653 ¼ 18:109g g ¼ 9:81ms2
hence
w
b ¼ ¼ 31:147
w
and
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u 1 + ð2xbÞ2
TR ¼ t 2 ¼ 0:0065
1  b 2 + ð2xbÞ2

Consequently, the maximum acceleration transmitted to the

instrument is
 
ðu€tot Þmax ¼ TR  u€g max ¼ 0:118g

(ii) Modification of the system so that ðu€tot Þmax < 0:1g

It must be
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u 1 + ð2xbÞ2 ðu€tot Þ 0:1g
TR ¼ t  ¼   max < ¼ 0:0055
2 2
1  b + ð2xbÞ 2 u€g max 18:109g

For the same elastic layer, the stiffness k and the damping coefficient c
are unaltered. Consequently, the transmission ratio TR can be modified if
the ratio b is changed. But because w is prescribed, this ratio can be chan-
ged, only if the natural frequency w is changed. This is possible if the mass
of the system (instrument) is changed by an increment Dm.
Let m 0 ¼ m + Dm be the new mass and x 0 , w0 the new damping ratio and
the frequency of the system (instrument), respectively. Then we will have
rffiffiffiffiffiffi
w0 m
¼ (1)
w m0
c ¼ 2mwx ¼ 2m 0 w0 x0
240 PART I Single-degree-of-freedom systems

which yield
rffiffiffiffiffiffi
0 mw m
x ¼x 0 0 ¼x (2)
mw m0
Moreover, we have
rffiffiffiffiffiffi
0 w m0
b ¼ 0¼b (3)
w m
From Eqs. (2), (3) we obtain
x0 b0 ¼ xb ¼ 3:115
Consequently, it must be
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u u
u 1 + ð2x 0 b0 Þ
2
u 2
TR ¼ u
¼ u 1 +
ð2  3:115Þ < 0:0055
t 2 2
t 2 2
1  b0 + ð2x0 b 0 Þ 1  b0
2
+ ð2  3:115Þ2

which holds if
pffiffiffiffiffiffiffiffiffiffiffiffi
33:885 < b0 ¼ b m 0 =m or m 0 > 3:619kNm1 s2
Hence Dm  0:561kNm1 s2 .

6.5 Problems
Problem P6.1 The supports 1 and 2 of the structure shown in Fig. P6.1 are sub-
 and ug2 ðt Þ ¼ u0 sin ðwt
jected to the motions ug1 ðt Þ ¼ u0 sin wt   p=3Þ, respec-
tively. A plane square rigid body of side a and density g is attached to
node 3. The mass of the column and the beam is neglected. Determine the
response of the structure and the reactions at the supports as well as the forces
that produce the support excitations. Data: L ¼ 8m, a ¼ 2m, cross section of the

FIG. P6.1 Structure in problem P6.1

 Response to ground motion and vibration isolation Chapter 6 241

beam and the column IPE360, w ¼ 10s1 , u0 ¼ 3cm, g ¼ 10kN=m2 , and

E ¼ 2:1  108 kN=m2 .
Problem P6.2 A SDOF system with parameters m, x,k is subjected
to ground motion whose accelerogram is given by u€gi ¼ 100ð1Þi
½ði + 5Þ=ði + 1Þ  ði + 1Þ=ði + 5Þ with duration t1 ¼ 2:5s (Fig. P6.2). Compute
and plot the response spectra of the displacement, velocity, and acceleration
for the interval of the natural period 0 < T  2 and x ¼ 0:05, 0:1.
Data: u ð0Þ ¼ u_ ð0Þ ¼ 0, u€g ð0Þ ¼ 0, m ¼ 4, and k ¼ 100.
ug,tt

(ug)i

(ug)i

t
FIG. P6.2 Accelerogram in problem P6.2

Problem P6.3 The rigid vertical column AC (Fig. P6.3) of a circular cross-
section, line density m is supported on the ground by means of a spherical hinge
at A and three elastic cables of cross-sectional area A and modulus of elasticity E.

(a) (b)
FIG. P6.3 Structure in problem P6.3
242 PART I Single-degree-of-freedom systems

The cables have been prestressed so that they can undertake compressive forces.
The column carries three advertising panels of total mass 10ma,  which are
arranged as in Fig. P6.3b. Their support on the column extends to a length
1:25a. The structure is subjected to ground motion in the y direction, whose accel-
erogram u€g ðt Þ is given in Problem P6.2. Determine the minimum prestressing
force of the cable GB using the results of Problem P6.2. The cables are assumed
massless. Data: a ¼ 5m, m  ¼ 0:5kNm1 s2 =m, E ¼ 2:1 108 kN=m2 , and
A ¼ 4cm . 2

Problem P6.4 The one-story building of Fig. P6.4 is subjected to ground

motion whose accelerogram u€g ðt Þ is given in Problem P6.2. The motion takes
place in the direction of angle b with respect to the x axis. Determine the
dynamic response of the structure and plot the relative displacement u ðb Þ
of the top cross-section of columns 1 and 2 as a function of the angle b,
0  b  2p. Moreover, compute the extreme normal stresses of the same
cross-sections due to bending. Data: x ¼ 0:07 and E ¼ 2:1  107 kN=m2 . The
mass of the columns is neglected. The load q includes also the dead weight
of the slab.

FIG. P6.4 One-story building in problem P6.4

Problem P6.5 The vertical columns of the frame in Fig. P6.5 have specific
weight g b and are assumed rigid. The elastic support on the ground is simulated
by the rotational springs CR . The horizontal beams are flexible with cross-
sectional moment of inertia I and modulus of elasticity E while their mass
and axial deformation are assumed negligible. The structure is subjected to
 of total duration ttot ¼ 3s. Give
the horizontal ground motion ug ðt Þ ¼ u0 sin wt
the graph of the response spectrum of the rotation of the structure and compute
the extreme values of the shear force and the bending moment of the beams.
Data: Cross-sectional area of the beams A ¼ a=10  a=10, CR ¼ EI =a,
E ¼ 2:1  107 kN=m, g b ¼ 24kN=m3 , u0 ¼ 0:03m, w ¼ 3, 5 and 7s1 .
 Response to ground motion and vibration isolation Chapter 6 243

=∞

FIG. P6.5 Frame in problem P6.5

Problem P6.6 A vehicle traveling with a velocity v¼80km/h on a multispan

bridge (Fig. P6.6) is idealized by the SDOF system of Fig. P6.6a. The length
of each span is L ¼ 20m. The deck of the bridge has been permanently
deformed due to creep so that each span can be simulated by a sinusoidal curve
with amplitude h ¼ 5cm. The stiffness of the system is k ¼ 10kN=m and its
damping ratio x ¼ 0:15. The tires of the wheels are assumed undeformable.
The total weight of the vehicle is W ¼ 20kN. (i) Determine the vertical motion
of the vehicle, (ii) Compute the maximum acceleration applied to the passenger
of the vehicle, and (iii) Compute the velocity that causes resonance to the vehi-
cle and gives its extreme amplitude.

(a)
h
h
(b) L

FIG. P6.6 Vehicle traveling on a bridge in problem P6.6

References and further reading

[1] R.W. Clough, J. Penzien, Dynamic of Structures, second ed., McGraw-Hill, Inc., New York, 1993.
[2] A.K. Chopra, Dynamics of Structures: Theory and Applications to Earthquake Engineering,
Prentice Hall, Englewood Cliffs, NJ, 1995.
[3] M.A. Biot, Vibrations of buildings during earthquake, in: Transient Oscillations in Elastic
System, Aeronautics Department, Calif. Inst. of Tech., Pasadena, CA, 1932. Chapter II in
Ph.D. Thesis No. 259.
244 PART I Single-degree-of-freedom systems

[4] F.B. Hildebrand, Advanced Calculus for Applications, Prentice Hall, Englewood Cliffs, NJ,
1962.
[5] R.R. Craig Jr., J. Andrew, A.J. Kurdila, Fundamentals of Structural Dynamics, second ed., John
Wiley, New Jersey, 2006.
[6] J.W. Leonard, Tension Structures, McGraw-Hill, New York, 1988.
[7] J.T. Katsikadelis, Finite deformation of cables under 3-D loading: an analytic solution, in:
D.E. Beskos, D.L. Karabalis, A.N. Kounadis (Eds.), Proc. of the 4th National Congress on Steel
Structures, Patras, May 24–25, vol. II, 2002, pp. 526–534.
Chapter 7

Damping in structures
Chapter outline
7.1 Introduction 245 7.6.1 Introduction 257
7.2 Loss of energy due to damping 246 7.6.2 The fractional derivative 258
7.3 Equivalent viscous damping 249 7.7 Measurement of damping 260
7.4 Hysteretic damping 250 7.7.1 Free vibration decay
7.5 Coulomb damping 252 method 261
7.5.1 Free vibrations with 7.7.2 Resonance amplitude
Coulomb damping 252 method 262
7.5.2 Forced vibrations with 7.7.3 Width of response curve
Coulomb damping 255 method 263
7.6 Damping modeling via 7.8 Problems 265
fractional derivatives 257 References and further reading 267

7.1 Introduction
In this chapter, the damping of structures is discussed. Damping appears in all
mechanical systems that perform vibrations. It is the dissipation of energy in a
vibrating structure. The type of energy into which the mechanical energy is
transformed depends on the system and the physical mechanism that causes
the dissipation. The energy is lost either in the form of heat or is radiated into
the environment. For example, the loss of energy in the form of heat is perceived
when an iron rod is subjected to alternating bending. The sound produced by a
body that is hit represents the loss of energy dissipated into the environment. In
the study of vibrations, we are interested in the damping related to the response
of the structure. Damping is due to different energy dissipation mechanisms act-
ing simultaneously. In spite of the age-long detailed studies on the damping of
structures, the understanding of damping mechanisms is quite primitive. A
well-known method to get rid of this problem is to use so-called viscous damp-
ing. This approach was first introduced by Rayleigh [1] via his famous dissipa-
tion function (see Section 1.8.4).
The loss of energy of a vibrating system reduces the amplitude of the free
vibration. When a system undergoing forced vibrations reaches the phase of
the steady-state response, the loss of energy is balanced by the energy input into
the system by the excitation force.
In vibrating systems, we distinguish different types of damping forces,
which may be due to the internal molecular friction, the sliding friction, or the

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00007-8

© 2020 Elsevier Inc. All rights reserved. 245
246 PART I Single-degree-of-freedom systems

resistance of the fluid in which the system moves. Sophisticated mathematical

models of damping forces are complex and are not convenient to study the vib-
rations. For this reason, simplified damping models have been developed that
allow an adequate study of the dynamic response of structures. We have already
used the model of viscous damping, which leads to equations that can be solved
analytically. The presented examples facilitate the comprehension of all con-
cepts. The pertinent bibliography with recommended references for further
study is also included. The chapter is enriched with problems to solved.

7.2 Loss of energy due to damping

The energy loss is determined in the phase of the steady-state response. The graph-
ical representation of the relationship fD ðu Þ between the damping force and the
displacement varies greatly for different types of damping. In all cases, however,
the curve is closed in a complete oscillation and includes a region called the hys-
teresis loop, whose area is equal to the energy lost per cycle (see also Refs. [2–5]).
In general, the loss of energy is expressed by the integral
Z T
WD ¼ fD ðu Þdu (7.2.1)
0

where T ¼ 2p=
w the period of the vibration in the steady-state response. The
quantity WD depends on several factors such as temperature, frequency, or
amplitude of the vibration.
In this section, we consider the simplest form of energy loss, that is, the loss
due to viscous damping. The damping force, in this case, is given by the relation
_ Moreover, the displacement in the steady-state phase, due to the
fD ¼ cu.

t, is given by Eq. (3.2.26), namely
harmonic force pðt Þ ¼ p0 sin w

t  qÞ
u ðt Þ ¼ rsin ðw (7.2.2)
where
p0 1
r¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (7.2.3)
k  
2 2
1  b + ð2xbÞ2
2xb
tan q ¼ (7.2.4)
1  b2
Consequently
u_ ðt Þ ¼ r

t  qÞ
w cos ðw (7.2.5)
The integral (7.2.1) yields
Z T
WD ¼ cu_ 2 dt
0
Z 2p=

w
(7.2.6)
w2 r 2
¼ c
cos 2 ðw

t  qÞdt
0
wr2
¼ pc

 Damping in structures Chapter 7 247

Of particular
pffiffiffiffiffiffiffiffiffi interest is the
pffiffiffiffiffiffienergy
ffi loss at resonance. Then it is

¼ w ¼ k=m , c ¼ 2mxw ¼ 2x km and Eq. (7.2.6) becomes
w
WD ¼ 2xpkr2 (7.2.7)
Moreover, the energy input into the system is due to the work that the force

t produces in a complete oscillation. Namely
pðt Þ ¼ p0 sin w
Z T
Wp ¼ pðt Þdu
0
Z 2p=

w
(7.2.8)
¼
t  qÞ sin w
w cos ðw
p0 r

tdt
0

¼ p0 rp sin q
Taking into account that (see Eq. 3.2.27)
2xb
sin q ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 ¼ 2xbrk=p0 (7.2.9)
1  b2 + ð2xbÞ2

we can write Eq. (7.2.8) as

wr2 ¼ WD
Wp ¼ pc
(7.2.10)
Indeed, during the steady-state response, the input into the system by the
excitation force is equal to the dissipated energy due to damping. We come
to the same conclusion if we examine the work produced by the elastic and
inertial forces.
The work produced by the elastic force in a cycle is
Z T
WS ¼ fS ðu Þdu
0
Z 2p=

w
¼ _
ku udt (7.2.11)
0
Z 2p=

w
2

r
¼ kw
t  qÞ cos ðw
sin ðw
t  qÞdt ¼ 0
0

Similarly, the work produced by the inertial force is

Z T
WI ¼ fI ðu Þdu
0
Z 2p=

w
¼ m u€udt
_ (7.2.12)
0
Z 2p=

w
w3 r2
¼ m

t  qÞ cos ðw
sin ðw
t  qÞdt ¼ 0
0
248 PART I Single-degree-of-freedom systems

Thus, it must be
WI + WD + W S ¼ Wp (7.2.13)
which by virtue of Eqs. (7.2.11), (7.2.12) yields Eq. (7.2.10).
Eq. (7.2.5) yields
u_ ðt Þ2 ¼ w

2 r2 cos 2 ðw

t  qÞ
 
¼w
r 1  sin 2 ðw
2 2

t  qÞ (7.2.14)
 
¼w
2 r2  u 2
hence
fD2 ¼ c2 u_ ðt Þ2
  (7.2.15)
¼ c2 w

2 r2  u 2
which is readily transformed into
   2
fD 2 u
+ ¼1 (7.2.16)
wr
c
r

FIG. 7.2.1 Hysteresis loop in a system with viscous damping.

The above relation represents an ellipse with semiaxes r and c
wr in the

plane u, fD , Fig. 7.2.1. The area of the ellipse is pc
wr2 , that is, it is equal to
the energy lost in a cycle. Therefore, the hysteresis loop of a system with viscous
damping is an ellipse.
The properties of damping in the material systems are expressed in
different ways depending on the scientific area where they are applied. Here
we mention two units for measuring the relative dissipated energy that are most
commonly encountered: the specific damping capacity and the specific damp-
ing factor.
The specific damping capacity is defined as
WD
wD ¼ (7.2.17)
U
where U ¼ kr2 =2 is the elastic energy of the system.
 Damping in structures Chapter 7 249

The specific damping factor is defined as

WD
¼ (7.2.18)
2pU
For viscous damping the above quantities result as

wr2 2pc

pc
w 2pð2mwx Þ

w
wD ¼ ¼ ¼ ¼ 4pxb (7.2.19)
kr2 =2 mw2 mw2
 ¼ 2xb (7.2.20)

7.3 Equivalent viscous damping

When the damping of the structure is not inherently viscous, the curve fD ðu Þ is
not generally an ellipse. It is, however, possible to determine an equivalent vis-
cous damping coefficient ceq by equating the area WD of the experimentally
obtained damping loop in a cycle of harmonic vibration (Fig. 7.3.1) with the
theoretical value of the area of the hysteresis loop representing dissipation of
energy due to viscous damping.

FIG. 7.3.1 Energy dissipated in a cycle of harmonic vibration determined experimentally.

On the basis of Eq. (7.2.6) we write

wr2
WD ¼ ceq p
(7.3.1)

from which we define the equivalent damping coefficient

WD
ceq ¼ (7.3.2)
wr2
p

and the respective damping ratio
ceq
xeq ¼ pffiffiffiffiffiffiffi (7.3.3)
2 km
250 PART I Single-degree-of-freedom systems

7.4 Hysteretic damping

When the structure is subjected to cyclic loading, the dissipation of energy
occurs in the interior of the material of the structure. Experiments performed
by many researchers have shown that in most structural materials, such as steel
and aluminum, the energy loss per cycle does not depend on the frequency of the
external excitation force, at least for a wide range of frequencies, but it is pro-
portional to the square of the amplitude of the oscillation. This damping is called
structural damping or hysteretic damping.
The equation of motion for forced vibrations with hysteretic damping is of
the form
m u€ + f ðu Þ ¼ pðt Þ (7.4.1)
where f ðu Þ is in general a nonlinear function of the displacement. The solution
of Eq. (7.4.1) can be obtained numerically using any of the methods presented in
Chapter 5.
The force f ðu Þ can be expressed as the sum of two forces, an elastic force
with a mean value fS ¼ ku and a damping force
k
fD ¼ u_ (7.4.2)
w

where  is the constant of damping. This is the simplest model that can be used
to represent the hysteretic damping.
The loss of energy in a cycle due to hysteretic damping results from
Eq. (7.2.6), if the coefficient c is replaced with k=

w. This yields
WD ¼ kpr2h (7.4.3)
where rh is the amplitude of the vibration with hysteretic damping. We observe
that the dissipation of energy per cycle does not depend on the excitation fre-
quency w
, but on the square of the amplitude of the vibration.
Apparently, the function f ðu Þ representing the resisting force can be written
k
f¼ u_ + ku (7.4.4)
w

and the equation of motion (7.4.1) for harmonic excitation takes the form
k
m u€ + u_ + ku ¼ p0 sin w

t (7.4.5)
w

We define the damping coefficient ch ¼ k=

w, which yields
x h ¼ ch =2mw ¼ =2b (7.4.6)
that is, the damping ratio changes with b.
The solution of Eq. (7.4.5) for the steady-state response is obtained from
Eq. (7.2.2), if x is replaced with xh . Thus
 Damping in structures Chapter 7 251

t  qh Þ
u ðt Þ ¼ rh sin ðw (7.4.7)
where
p0 1
rh ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi (7.4.8)
k  2
1  b2 + 2

tan qh ¼ (7.4.9)
1  b2
Working as in Section 7.2, we can determine the geometry of the hysteresis
loop for this type of damping. Thus, we obtain from Eq. (7.4.7)
u_ ðt Þ2 ¼ r2h w

2 cos 2 ðw

t  qh Þ
2 2
 
¼ rh w
1  sin 2 ðw
t  qh Þ (7.4.10)
 
¼w
2 r2h  u 2
Moreover, it is
 2
k
fD2 ¼ ðf  ku Þ2 ¼ u_ 2 (7.4.11)
w

which by virtue of (7.4.10) is written as
 
ðf  ku Þ2 ¼ 2 k 2 r2h  u 2 (7.4.12)
or
   2
f  ku 2 u
+ ¼1 (7.4.13)
krh rh
Therefore, in the plane u, f , Eq. (7.4.13) represents a rotated ellipse
(Fig. 7.4.1). The area of the ellipse is pkr2h and expresses the loss of energy
in a cycle. We observe that Eq. (7.4.13) does not involve the excitation fre-
quency. This implies that the hysteresis loop can be determined experimentally
using a low excitation frequency, that is, quasistatic, by plotting the load-
displacement curve.
For f ¼ 0, we obtain the abscissa of the ellipse on the u axis

d ¼ rh pffiffiffiffiffiffiffiffiffiffiffiffiffi (7.4.14)
1 + 2
which can be used to evaluate the damping coefficient . Hence
d
 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (7.4.15)
r2h  d 2
252 PART I Single-degree-of-freedom systems

FIG. 7.4.1 Hysteresis loop in a system with hysteretic damping.

7.5 Coulomb damping

As mentioned in Section 7.1, damping is due to various energy dissipation
mechanisms acting simultaneously. Although the model of viscous damping
or hysteretic damping in most cases can adequately approximate the behavior
of the structure, it is not suitable to describe the response of the structure when
forces due to Coulomb friction are present. These forces arise on the dry contact
interface of two bodies with a relative sliding. The friction force is expressed by
the relation F ¼ mN , where N is the force normal to the contact surface and m
the friction coefficient, which is determined experimentally and is greater
before the sliding. The direction of the frictional force is opposite to that of
motion, that is, its sign is opposite to the sign of the velocity. Therefore, we
can set

F ¼  signðu_ ÞmN (7.5.1)

where
signðu_ Þ ¼ u=
_ ju_ j (7.5.2)

7.5.1 Free vibrations with Coulomb damping

We consider the SDOF system of Fig. 7.5.1a, which is sliding on a dry surface.
When the body is moving to the right (Fig. 7.5.1b) the equilibrium of forces
yields the equation of motion
m u€ + ku ¼ F (7.5.3)

while when the body is moving to the left (Fig. 7.5.1c) the equilibrium of forces
yields the equation of motion

m u€ + ku ¼ F (7.5.4)
 Damping in structures Chapter 7 253

(a)

(b) (c)
FIG. 7.5.1 Motion with Coulomb friction.

On the basis of Eq. (7.5.1), Eqs. (7.5.3), (7.5.4) can be combined as

m u€ + ku ¼ F signðu_ Þ (7.5.5)
or
m u€ + F signðu_ Þ + ku ¼ 0 (7.5.6)
Eq. (7.5.6) is nonlinear because of the term of damping and can be solved by
any of the methods developed in Chapter 5. In the following, an analytical solu-
tion of Eq. (7.5.6) is presented with zero initial conditions, u ð0Þ ¼ u0 , u_ ð0Þ ¼ 0.
The body is displaced from the equilibrium position to the right by u0 and we
study the subsequent motion. The body will move to the left according to
Eq. (7.5.4). A particular solution to this equation is
F
up ðt Þ ¼ (7.5.7)
k
Hence the general solution of Eq. (7.5.4) is
F
u ðt Þ ¼ A cos wt + B sin wt + (7.5.8)
k
Applying the initial conditions
u ð0Þ ¼ u0 , u_ ð0Þ ¼ 0 (7.5.9)
gives
F
A ¼ u0  , B ¼ 0 (7.5.10)
k
and the displacement (7.5.8) becomes
 
F F
u ðt Þ ¼ u 0  cos wt + (7.5.11)
k k
254 PART I Single-degree-of-freedom systems

Eq. (7.5.11) holds until the velocity vanishes, namely t ¼ p=w. At that
instant, the body is at the extreme left position, where the displacement is
u ðp=wÞ ¼ ðu0  2F=k Þ. The body will now start moving to the right with ini-
tial conditions
p
  p

F
u ¼  u0  2 , u_ ¼0 (7.5.12)
w k w

The motion is now described by Eq. (7.5.3), which has a particular solution

F
up ðt Þ ¼  (7.5.13)
k
and general solution

F
u ðt Þ ¼ A cos wt + B sin wt  (7.5.14)
k
The initial conditions (7.5.12) give

3F
A ¼ u0  , B¼0 (7.5.15)
k
and Eq. (7.5.14) becomes
 
F F
u ðt Þ ¼ u 0  3 cos wt  (7.5.16)
k k

FIG. 7.5.2 Free vibrations with Coulomb damping.

 Damping in structures Chapter 7 255

Eq. (7.5.16) holds until the body reaches the extreme right position, namely
until the instant t ¼ 2p=w. At that time, the body has completed a full oscillation
and the displacement is u ð2p=wÞ ¼ ðu0  4F=k Þ. This solution procedure con-
tinues to obtain the response of the next oscillations. The graphical representa-
tion of the displacement versus time is shown in Fig. 7.5.2. The curve was
obtained by numerical integration of the equation of motion (7.5.6) with
m ¼ 10kNm1 s2 , T ¼ 0:5s, F ¼ 23:685kN, u0 ¼ 0:55m, u_ 0 ¼ using the pro-
gram aem_nlin.m developed in Chapter 5. The numerical results coincide with
those obtained by the above-presented analytical solution. The motion is a
vibration with a period T ¼ 2p=w, which means that the Coulomb friction does
not affect the frequency or the period of vibration. The amplitude of vibration is
reduced in each cycle by 4F=k. A consequence of this is that the envelopes of
the curve are straight lines, unlike in the cases of viscous or hysteretic damping
where the envelopes are exponential functions. The motion of the system con-
tinues until the elastic force ku becomes smaller than the force F of the friction.
Until now, no difference was made between static friction Fs ¼ ms N and
dynamic friction Fd ¼ md N . The first occurs when the body is stationary and
the second when the body moves. Generally, it is md < ms , hence the dynamic
friction coefficient md will be used in the equation of motion while the static
friction coefficient ms is used for the control of the motion. In viscous or hys-
teretic damping, theoretically, the body does not stop moving because the
amplitude of the vibration reduces exponentially. Nevertheless, real structural
systems stop after a finite time. This is due to the fact that the Coulomb friction
coexists with other forms of damping and forces the moving systems to stop.

7.5.2 Forced vibrations with Coulomb damping

In this case, the equation of motion becomes
m u€ + F signðu_ Þ + ku ¼ pðt Þ (7.5.17)
An analytical solution can be achieved by splitting the above equation into
two equations. Thus the equation describing the motion to the right is
m u€ + ku ¼ F + pðt Þ (7.5.18)
whose general solution is
Z
F 1 t
u ðt Þ ¼ Acos wt + B sin wt  + pðτÞ sin wðt  τÞdτ (7.5.19)
k mw 0
On the other hand, the equation describing the motion to the left is
m u€ + ku ¼ F + pðt Þ (7.5.20)
whose general solution is
Z
F 1 t
u ðt Þ ¼ A cos wt + B sin wt + + pðτÞsin wðt  τÞdτ (7.5.21)
k mw 0
256 PART I Single-degree-of-freedom systems

t the general solution is (see Eq. 3.2.8)

For a harmonic load pðt Þ ¼ p0 sin w
F p0 1

w
u ðt Þ ¼ A cos wt + B sin wt  +
t, b ¼
sin w (7.5.22)
k k 1  b2 w

Apparently, the derivation of the solution for forced vibrations becomes

quite complicated if we follow the procedure applied for the free oscillations,
that is, by splitting the equation of motion into two equations. Nevertheless, for
small damping, we can approximate the solution in the phase of the steady-state
response. Thus, the solution is given by Eq. (7.2.2).
The graph of the friction force versus displacement in one cycle takes the
form of the rectangle of Fig. 7.5.3. Hence, the work produced by the friction
force in a complete oscillation is

WD ¼ 4Fr (7.5.23)
We can determine an equivalent coefficient of viscous damping by equating
the loss of energy (Eq. 7.5.23) with that of the viscous damping given by
Eq. (7.3.1). Namely,

FIG. 7.5.3 Hysteresis loop in a system with Coulomb damping.

wr2 ¼ 4Fr
ceq p
(7.5.24)

which gives
4F
ceq ¼ (7.5.25)
p

wr
and an equivalent damping ratio
ceq 2F
xeq ¼ ¼ (7.5.26)
2mw pkrb
 Damping in structures Chapter 7 257

Substituting this value of the damping ratio in Eq. (7.2.3) yields

p0 1
r¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi (7.5.27)
k  2
 2 2
1  b + ð4F=pkrÞ

which is solved for r to give

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
p0 u1  ð4F=pp0 Þ2
r¼ t  2 (7.5.28)
k 1  b2

Eq. (7.5.28) holds if 1  ð4F=pp0 Þ2 > 0, that is, when F=p0 < p=4. Obvi-
ously, for F=p0 > p=4 r becomes imaginary and this method for determining
an equivalent damping coefficient does not apply.
The phase angle results from Eq. (7.2.4) by setting x ¼ xeq and taking the
value of r from Eq. (7.5.28). Thus, we have

4F=pp0
tan q ¼  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (7.5.29)
1  ð4F=pp0 Þ2

The plus sign is valid when b < 1 while the minus sign is valid when b > 1.

7.6 Damping modeling via fractional derivatives

7.6.1 Introduction
In the modeling of the linear elastic behavior of structures, damping has been
classically modeled as being linearly dependent on velocity. This damping
model is adequate for the dynamic analysis of lightly damped structures. How-
ever, this model, while mathematically straightforward, fails to describe the
damping response of heavily damped structures. This drawback is more pro-
nounced in materials used in devices aimed at increasing the damping of struc-
tures, such as seismic isolators.
Historically, the need for more refined models to control damping has
pushed researchers to use viscoelastic models in the dynamic analysis of struc-
tures. Applicable viscoelastic models are expressed as a series of ordinary time
derivatives relating time-dependent stress to strain fields. These models yield
acceptable plots of material properties, but they have drawbacks. Normally,
these models contain many terms, making them mathematically cumbersome
and increasing the order of the differential equations describing the response
of the system. An alternative representation of these models is realized via
fractional derivatives, that is, derivatives of noninteger order [6]. The real
strength of this approach is that the fractional derivative models describe
258 PART I Single-degree-of-freedom systems

damping effects using fewer material parameters than the integer order differ-
ential models but with equal precision. This approach involves fractional dif-
ferential equations, which needs acquaintance with fractional calculus.

7.6.2 The fractional derivative

The fractional (noninteger order) derivatives are as old as calculus. The theory
of derivatives of noninteger order goes back to G. W. Leibnitz. After Leibnitz
defined the derivative of integer order, d n y=dx n with n being an integer,
L’H^ opital asked: “What if n is a fraction, say n ¼ 1=2?” Leibnitz gave an
answer to this question (Sept. 30, 1695) concluding “It will be a paradox”
and added prophetically “From this apparent paradox, one day useful conse-
quences will be drawn” [7]. For three centuries, the fractional derivative
inspired pure theoretical mathematical developments useful only for mathema-
ticians. The integer order derivative allows giving geometrical interpretations to
the proposed physical models resulting from Newton’s law of motion. Appar-
ently, this made the then-revolutionary concepts accessible to contemporary
scientists, who were well experienced in geometry. In a sense, the long delay
to apply fractional calculus may be attributed to this fact. We recall that frac-
tional derivatives lack the straightforward geometrical interpretation of their
integer counterparts. However, research carried out in recent years has pointed
out that fractional order derivatives provide an effective tool to reliably model
many complex physical and engineering systems. In recent years, this fact gave
a great boost to the study of fractional differential equations. In this context,
many books on fractional calculus as well as numerous publications on the
study of physical and engineering systems via fractional derivatives have been
written; see for example [8] and the references therein.
There are several definitions of the fractional derivative [8]. The Riemann-
Liouville and Caputo fractional derivatives are among the most widely used.
Nevertheless, as the scope of this section is not to give a detailed account of
this subject, we will restrict our discussion to the Caputo derivative, defined as
Z t
1 u ðmÞ ðτÞ
DCa u ðt Þ ¼ dτ, m  1 < a < m (7.6.1)
Gðm  aÞ 0 ðt  τÞa + 1m
where Gðz Þ is the gamma function with argument z, a is the order of the frac-
tional derivative, and m an integer. It can be shown that
lim DCa u ðt Þ ¼ u ðmÞ ðt Þ lim DCa u ðt Þ ¼ u ðm1Þ ðt Þ  u ðm1Þ ð0Þ (7.6.2)
a!m a!m1

Obviously, for m ¼ 1, the fractional derivative interpolates the first-order

derivative [9]. Fig. 7.6.1 shows the fractional derivative of the function
u ¼ t  t 3 =6 + t 5 =120 for various orders of the fractional derivative.
 Damping in structures Chapter 7 259

1.2

1 D 1u
c

0.8 D 0.99999u
c
D 0.9u
0.6 c
D 0.5u
0.4 c
D 0.1u
c
0.2
D 0.00001u
c
0 u
−0.2

−0.4

−0.6
0 0.5 1 1.5 2 2.5 3
FIG. 7.6.1 The fractional derivative of the function u ¼ t  t 3 =6 + t 5 =120 for various orders.

If damping is modeled via the fractional derivative, the equation of motion

of a SDOF system is written as

m u€ + cD aC u + ku ¼ pðt Þ (7.6.3)

Obviously, for a ! 1, we obtain the equation of motion with viscous

damping

m u€ + cu_ + ku ¼ pðt Þ (7.6.4)

while for a ! 0 the equation of motion becomes

m u€ + ðc + k Þu ¼ pðt Þ + cu 0 (7.6.5)

that is, it yields the elastic response with stiffness k ∗ ¼ c + k and excitation
force p∗ ¼ pðt Þ + cu 0 . The Caputo derivative is employed because, contrary
to other types of fractional derivatives, it allows the application of initial con-
ditions having a direct physical significance.
Eq. (7.6.3) represents a fractional differential equation. Analytical solutions
of such equations are difficult or impossible to obtain. This reason has recently
boosted the development of efficient numerical methods for solving fractional
differential equations [10, 11].
Fig. 7.6.2 shows the free vibration response of an DOF system for various
values of the order a
260 PART I Single-degree-of-freedom systems

FIG. 7.6.2 Free vibration response of a SDOF system for various values of the order a.

The solution of Eq. (7.6.3) has been obtained using the method presented
in [11]. The developed MATLAB program has been given the name
three_term_FD.m and is included on this book’s companion website. Note
that for a  1 and a  0 the solutions are identical to the corresponding
analytical ones.
The fractional calculus has allowed the definition of any order of fractional
derivative, real or imaginary. This fact enables us to consider the fractional
derivative to be a function of time (explicit variable-order fractional derivative)
or of some other time-dependent variable (implicit variable-order fractional
derivative). Thus, the variable-order Caputo derivative for m ¼ 1 reads
Z t
aðt Þ 1 u_ ðτÞ
DC u ð t Þ ¼ dτ ðExplicitÞ (7.6.6a)
Gð1  aðt ÞÞ 0 ðt  τÞaðt Þ
Z t
aðu Þ 1 u_ ðτÞ
DC u ð t Þ ¼ dτ ðImplicitÞ (7.6.6b)
Gð1  aðu ÞÞ 0 ðt  τÞaðu Þ
Z t
aðu_ Þ 1 u_ ðτÞ
DC u ð t Þ ¼ dτ ðImplicitÞ (7.6.6c)
Gð1  aðu_ ÞÞ 0 ðt  τÞaðu_ Þ
The concept of a variable-order fractional derivative exhibits notable advan-
tages over the constant order derivative and it has been recently used to model
the dynamic response of actual structures [12, 13].

7.7 Measurement of damping

As stated in Chapter 1, the mass and stiffness of a dynamic system can be deter-
mined from its physical characteristics. The mass can be determined from the
geometry and the mass density of the structural elements. The stiffness is also
 Damping in structures Chapter 7 261

determined if the geometry and material properties of the structural elements

are known. However, it is difficult or at least impractical to relate damping
to known or measurable physical characteristics of the system. Therefore, the
damping of a given structure cannot be measured precisely during the design
phase, but only experimentally after its construction. There are several exper-
imental techniques to measure the damping of a structure. Most of them are
based on the assumption that the damping is viscous. When damping is not vis-
cous in nature, an equivalent viscous damping is usually determined.

7.7.1 Free vibration decay method

This is the simplest and most frequently used method of determining the viscous
damping ratio x through experimental measurements. This method was already
presented in Chapter 2 when we discussed the free damped vibrations. The sys-
tem is excited by means of an appropriate experimental instrumentation and is
left to perform free oscillations. Subsequently, the peak amplitudes ui and ui + n
over n consecutive cycles are measured. As shown in Section 2.3.2, the damp-
ing ratio can be calculated using Eq. (2.3.24), namely
 
2npx ui
pffiffiffiffiffiffiffiffiffiffiffiffi ¼ ln (7.7.1)
1  x2 ui + n
which yields
dn
x ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (7.7.2)
4p n 2 + d n 2
2

where
 
ui
dn ¼ ln (7.7.3)
ui + n
This method was illustrated in Example E6.2.1.
When the damping is hysteretic, the damping force is given by Eq. (7.4.2). In

 w, provided that the damping is small. Hence,
free vibrations, it can be set as w
the equivalent damping coefficient ch and the damping ratio are obtained from
the relations
k
ch ¼ (7.7.4a)
w

xh ¼ (7.7.4b)
2
The hysteretic damping coefficient  is obtained by using Eq. (7.7.2), if it is
set x ¼ xh . This yields
262 PART I Single-degree-of-freedom systems

2dn
 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (7.7.5)
4p n 2 + dn 2
2

Example 7.7.1 Evaluation of damping in a SDOF system

In a test, the silo in Example 2.3.1 is subjected to harmonic excitation at a fre-
quency 15s1 and the force-displacement relation is obtained at the steady state.
The amplitude of the displacement is measured as 10 cm, and the energy loss per
cycle 50kNm. If damping is considered to be viscous, determine c and x.
Solution
The mass of the system is m ¼ 100kNm1 s2 , the stiffness k ¼ 2721:6kN=m,
and frequency w ¼ 5:2169s1 , taken from Example 2.3.1.
(i) For viscous damping, the energy loss per cycle is given by Eq. (7.2.6),
namely
wr2
WD ¼ pc
(1)
1

¼ 15s , r ¼ 0:1m, WD ¼ 50kNm gives
which for w
WD 50
c¼ ¼ ¼ 106:10 (2)
wr2 p  15  0:12
p

c 106:10
x¼ ¼ ¼ 0:102 (3)
2mw 2  100  5:2169s1

7.7.2 Resonance amplitude method

t. In the phase of
The system is subjected to harmonic excitation pðt Þ ¼ p0 sin w
the steady-state response, the amplitude and the phase angle are given by
Eq. (3.2.27), namely

p0 h i 1
2 2
r¼ 1  b2 + ð2xb Þ2 (7.7.6a)
k
 
2xb
q ¼ tan 1 (7.7.6b)
1  b2
For b ¼ 1 Eq. (7.7.6b) yields q ¼ 90° regardless of the value of x. If the
employed experimental instrumentation allows the measurement of the phase
angle q, then we adjust the excitation frequency so that q ¼ 90° and we measure
the amplitude of vibration r. Besides, setting b ¼ 1 in Eq. (3.2.28) we obtain
r 1
D ¼ max jRðt Þj ¼ ¼ (7.7.7)
p0 =k 2x
which yields
 Damping in structures Chapter 7 263

p0 =k
x¼ (7.7.8)
2r
This previous method requires knowledge of the stiffness k of the structure,
which is determined either from the physical characteristics of the structure or
experimentally, for example, by imposing a load and measuring the resulting
displacement.
If the measurement of the phase difference is not easy, then we measure
experimentally the amplitude of the vibration in the range of resonance. Sub-
sequently, we plot the curve Dðb Þ ¼ rðb Þ=ðp0 =k Þ (see Fig. 7.7.1) and determine
its maximum. Then Eq. (3.2.31), namely
1
Dmax ¼ pffiffiffiffiffiffiffiffiffiffiffiffi (7.7.9)
2x 1  x 2

FIG. 7.7.1 Graphical representation of the curve D ðbÞ in the range of resonance.

allows the evaluation of x. For small values of x, it is x 2  0, hence

x  1=2Dmax . This method also requires knowledge of the stiffness k of the
structure.

7.7.3 Width of response curve method

The width of the response curve DðbÞ in the range of resonance can be used to
determine the damping of the structure. In this method, the frequencies corre-
sponding to the phase angles 45° are measured. One of these frequencies is
264 PART I Single-degree-of-freedom systems

below the resonance frequency while the other is above it (see Fig. 7.7.2). The
respective values of b are obtained from Eq. (7.7.6b). Thus, we have
2xb1
¼1 (7.7.10a)
1  b21
2xb2
¼ 1 (7.7.10b)
1  b22
which can be written as
1  b21  2xb1 ¼ 0 (7.7.11a)
1  b22  2xb2 ¼0 (7.7.11b)

FIG. 7.7.2 Response curve DðbÞ in the neighborhood of resonance to determine bI and b II .

Subtracting the previous equations gives

1 1w
1

2  w
x ¼ ðb 2  b 1 Þ ¼ (7.7.12)
2 2 w

1 and w
Eq. (7.7.12) can be used to calculate x when the frequencies w
2 have
been measured and the natural frequency w is known or can be determined.
The previously described method is based on the capability of measuring the
phase angle, which may not be simple because it requires complex instrumen-
tation. However, another property of the response curve may be used to deter-
mine the damping ratio. This method is not based on the measurement of the
phase angle, but on experimental measurements. The response curve D ðb Þ is
plotted in the range of resonance and the values b corresponding to
 Damping in structures Chapter 7 265

pffiffiffi
1= 2Dmax are determined. It is apparent from Fig. 7.7.2 that these values are
two, which are denoted by bI and bII . For small values of the damping ratio, it is
Dmax  1=2x. Besides, using Eq. (3.2.28) we can write
1 1 1
pffiffiffi ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  (7.7.13)
2 2x 2
1  b2 + ð2xb Þ2

The solution of the above equation gives two values of b

  qffiffiffiffiffiffiffiffiffiffiffiffi
2
b I ¼ 1  2x  2x 1 + x2
2
(7.7.14a)

  qffiffiffiffiffiffiffiffiffiffiffiffi
b 2II ¼ 1  2x2 + 2x 1 + x2 (7.7.14b)
pffiffiffiffiffiffiffiffiffiffiffiffi
Using the binomial theorem to expand 1 + x 2 and neglecting terms of
order higher than the second, the above relations reduce to

b2I  1  2x  2x 2 (7.7.15a)

b2II  1 + 2x  2x 2 (7.7.15b)

or after expanding in Taylor series

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3
bI  1  2x  2x2  1  x  x 2 (7.7.16a)
2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3
b II  1 + 2x  2x2  1 + x  x 2 (7.7.16b)
2
which give
1
x ¼ ðbII  b I Þ (7.7.17)
2
Eq. (7.7.17) is similar to Eq. (7.7.12), but avoids the measurement of the
phase angles.

7.8 Problems
Problem 7.1 The damping force applied to a system moving in a fluid is given
by the relation fD ¼ lu_ a , where l and a are real constants. Give:
(i) The graphical representation of the displacement u ðt Þ when the system
performs free vibrations with initial conditions u ð0Þ ¼ 0:02m, u_ ð0Þ ¼ 0.
(ii) The graphical representation of the displacement u ðt Þ when the system is

t with zero initial conditions.
subjected to the harmonic load pðt Þ ¼ p0 sin w
Data: m ¼ 10kNm1 s2 , k ¼ 1500kN=m, l ¼ 100, a ¼ 3, p0 ¼ 300kN, and

¼ 2s1 .
w
266 PART I Single-degree-of-freedom systems

Problem 7.2 The system of Fig. P7.2 is subjected to Coulomb damping. Give:
(i) The graphical representation of the displacement u ðt Þ when the system
performs free vibrations with initial conditions (a) u ð0Þ ¼ 0:40m,
u_ ð0Þ ¼ 0 and (b) u ð0Þ ¼ 0, u_ ð0Þ ¼ 2:8ms1 .
(ii) The graphical representation of the displacement u ðt Þ when the system is

t with zero initial conditions.
subjected to the harmonic load pðt Þ ¼ p0 sin w
(iii) Calculate the equivalent damping ratio xeq of the equivalent viscous
damping.
Data: m ¼ 10:132kNm1 s2 , k ¼ 1600kN=m, b ¼ w

=w ¼ 0:4, N ¼ 70kN,
m ¼ 0:32, and p0 ¼ 1:57F.

FIG. P7.2 System subjected to Coulomb damping.

Problem P7.3 Two bodies B1 and B2 with masses m1 and m2 , respectively, are
placed on two inclined planes whose angles are f1 and f2 , as shown in
Fig. P7.3. The bodies are connected by a massless cable of length L and axial
stiffness k ¼ EA=L. The friction coefficient between the bodies and the inclined
planes is m while between the cable and the pulley it is zero. Determine the
equation of motion of the system.

FIG. P7.3 System in problem P7.3.

Problem P7.4 The damping of a SDOF system is expressed by the Caputo frac-
tional derivative of order a ¼ 0:5. Compare the response of the system with that
of viscous damping. Data: m ¼ 10kNm1 s2 , x ¼ 0:1, k ¼ 500kN=m,
p ¼ p0 sin 5t, and p0 ¼ 2kN. Hint: Use the program three_term_FD.m avail-
able on this book’s companion website.
 Damping in structures Chapter 7 267

References and further reading

[1] L. Rayleigh, Theory of Sound, second ed., Dover Publications, New York, 1877; reissued
1945.
[2] R.W. Clough, J. Penzien, Dynamics of Structures, second ed., McGraw-Hill, New York, 1993.
[3] A.K. Chopra, Dynamics of Structures: Theory and Applications to Earthquake Engineering,
Prentice Hall, Englewood Cliffs, NJ, 1995.
[4] J.L. Humar, Dynamics of Structures, second ed., A.A. Balkema Publishers, Lisse, NL, 2002.
[5] S. Adhikari, Damping Models for Structural Vibration, Dissertation submitted to the Univer-
sity of Cambridge for the Degree of Doctor of Philosophy, Trinity College, Cambridge, 2000.
[6] R.L. Bagley, P.J. Torvik, Fractional calculus—a different approach to the analysis of viscoe-
lastically damped structures, AIAA J. 27 (1998) 1412–1417.
[7] G.W. Leibniz, S€amtliche Schriften und Briefe, 3te Reihe, Mathematischer, Naturwis-
senschaftlicher und Technischer Briefwechsel, Band VI, Letter 163, 2004, p. 510. Berlin-
Brandenburgischen Akademie der Wissenschaften und Akademie der Wissenschaften in
G€ottingen.
[8] I. Podlubny, Fractional Differential Equations, Academic Press, New York, 1999.
[9] J.T. Katsikadelis, Generalized fractional derivatives and their applications to mechanical sys-
tems. Arch. Appl. Mech. 85 (2015) 1307–1320, https://doi.org/10.1007/s00419-014-0969-0.
[10] K. Diethelm, N.J. Ford, A.D. Freed, Y. Luchko, Algorithms for fractional calculus: a selection
for numerical Methods, Comput. Methods Appl. Mech. Eng. 194 (2005) 743–773.
[11] T. Katsikadelis, Numerical Solution of multi-term fractional differential equations, ZAMM,
Z. Angew. Math. Mech. 89 (7) (2009) 593–608.
[12] C.F.M. Coimbra, Mechanics with variable-order differential operators, Ann. Phys. (Leipzig)
12 (11–12) (2003) 692–703, https://doi.org/10.1002/andp.200310032.
[13] J.T. Katsikadelis, Numerical solution of variable order fractional differential equations,
arXiv:1802.00519 [math.NA] (2018).
Chapter 8

Generalized single-degree-of-
freedom systems—Continuous
systems
Chapter outline
8.1 Introduction 269 8.3.3 Free vibrations of beams 286
8.2 Generalized single-degree-of- 8.3.4 Orthogonality of the
freedom systems 275 free-vibration modes 291
8.3 Continuous systems 284 8.3.5 Forced vibrations of
8.3.1 Introduction 284 beams 293
8.3.2 Solution of the beam 8.4 Problems 295
equation of motion 285 References and further reading 297

8.1 Introduction
In this chapter, the method of global shape functions is employed to approx-
imate the response of continuous systems by SDOF systems, which we call
generalized SDOF systems. The example that follows helps in understanding
the basic ingredients of the method as well as the error introduced by the
lumped mass assumption.
In order to study the dynamic response of the frame shown in Fig. 8.1.1a, we
approximate it by the model of Fig. 8.1.1b. As mentioned, in formulating this
model, it was assumed that the mass of the columns is concentrated at their ends,
that is, half at the top and half at the foot of the column. The consequence of this
assumption is that the elastic curve of the columns has the form of an unloaded
beam fixed at both ends, whose end cross-sections undergo a relative displace-
ment u ðt Þ. This assumption is, however, not entirely correct because the mass of
the columns is actually distributed along their length. This fact, apparently, pro-

u€
duces inertial forces fI ðx, t Þ ¼ m
ðx, t Þ, where u
ðx, t Þ is the actual elastic curve
of the columns (see Fig. 8.1.1c). Thus, the problem would be correctly
addressed if the columns were treated as systems of infinite degrees of freedom,
that is, continuous systems whose top cross-sections are connected by the rigid
beam of mass m. Such an approach, however, would be quite difficult because
the analysis leads to partial differential equations (see Section 1.1). Neverthe-
less, it is possible to approximate the system by another model, which is closer

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00008-X

© 2020 Elsevier Inc. All rights reserved. 269
270 PART I Single-degree-of-freedom systems

to the actual system than that of Fig. 8.1.1b. In this model, we consider that the
mass is distributed along the length of columns, but their deformed shape is cho-
sen so that it is close to the actual one. We observe that the end cross-sections of
the columns in Fig. 8.1.1a do not rotate during the motion. Hence the shape func-
tion ðx Þ ( see Eq. (8.1.1)) representing the shape of the elastic curve should be
chosen so that it satisfies the geometrical boundary conditions at the ends of the
columns, which demand ð0Þ ¼ 0, 0 ð0Þ ¼ 0, ðh Þ 6¼ 0, and 0 ðh Þ ¼ 0. The func-
tions ðx Þ are not unique. They constitute an infinite set of functions called
geometrically admissible functions.

(a) (b)

(c)
FIG. 8.1.1 Model of a two-column frame with rigid beam.

After this consideration, we may set

u
ðx, t Þ ¼ ðx Þu ðt Þ (8.1.1)
where u ðt Þ is a time-dependent function denoting the displacement of a
certain cross-section of the column, say at x ¼ a, 0 < a  h. This yields
u
ða, t Þ ¼ ða Þu ðt Þ ¼ u ðt Þ, which implies that ða Þ ¼ 1. If it is taken a ¼ h, then
u ðt Þ ¼ u
ðh, t Þ represents the displacement at the top of the column.
Eq. (8.1.1) states that the shape of the elastic curve remains the same
during motion while its values are multiplied by the common factor u ðt Þ.
The function ðx Þ, which approximates the elastic curve, is referred to as shape
function.
In the example of Fig. 8.1.1, a shape function for the columns is the elastic
curve of a beam with constant stiffness EI , whose ends are subjected to a rel-
ative displacement equal to one. Therefore, the shape function is obtained from
the solution of the differential equation of the elastic curve of the beam in the
absence of loading (Fig. 8.1.2a)
d 4 ðx Þ
EI ¼0 (8.1.2)
dx 4
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 271

(a) (b)
FIG. 8.1.2 Shape function (a) and and deflection curve (b) in Example 8.1.1

with boundary conditions

0 0
ð0Þ ¼ 0, ð0Þ ¼ 0, ðh Þ ¼ 1, ðh Þ ¼ 0 (8.1.3)
Integrating Eq. (8.1.2) yields
x3 x2
ðx Þ ¼ c 1 + c2 + c3 x + c4 (8.1.4)
6 2
where ci (i ¼ 1, 2, 3, 4) are arbitrary constants.
Applying now the boundary conditions (8.1.3), we obtain
12 6
c1 ¼  , c2 ¼ , c3 ¼ 0, c4 ¼ 0 (8.1.5)
h3 h2
Then substituting these values for the arbitrary constants into Eq. (8.1.4)
yields the shape function

x 2
x 3
ðx Þ ¼ 3 2 ¼ 3x2  2x3 , x ¼ x=h (8.1.6)
h h
After the selection of the shape function, the motion of the system is deter-
mined by the time-dependent function u ðt Þ. The equation of motion can be
derived using different methods such as Hamilton’s principle, Lagrange equa-
tions, or the principle of virtual work. To implement these methods, it is nec-
essary to determine the elastic energy U and the kinetic energy T of the system
as well as the virtual work of the external force pðt Þ.
(a) Evaluation of the elastic energy U . The elastic energy is due to the strain
energy of the two columns because of bending. The strain energy per unit
volume is
1
U0 ¼ s x e x (8.1.7)
2
272 PART I Single-degree-of-freedom systems

where sx is the normal stress on the cross-section of the column and ex the cor-
responding strain. For a linearly elastic material with modulus of elasticity E,
it is ex ¼ sx =E. Taking into account that the bending stress in a beam is
sx ¼ Mx y=I ðx Þ, we obtain
 
1 Mx y 2
U0 ¼ (8.1.8)
2E I ðx Þ
where I ðx Þ is the moment of inertia of the, in general, variable cross-section
and Mx the bending moment. Hence, the total elastic energy of the one
column is
Z
U
¼ U0 dV
2 V
Z   (8.1.9)
1 Mx y 2
¼ dV
V 2E I ðx Þ

where V is the volume of the column. It is known from the beam theory
that
Mx ¼ EI ðx Þu
00 ðx, t Þ (8.1.10)
which is introduced into Eq. (8.1.9) to yield
Z
U 1 h
EI ðx Þ½u
00 ðx, t Þ dx
2
¼ (8.1.11)
2 2 0
For I ðx Þ ¼ I ¼ constant and u
ðx, t Þ ¼ ðx Þu ðt Þ, the previous equation
becomes
Z h
U EI
½ 00 ðx Þ dx
2 2
¼ ½u ðt Þ (8.1.12)
2 2 0

Differentiating Eq. (8.1.6) twice with respect to x gives

00 6
ðx Þ ¼ ð1  2xÞ (8.1.13)
h2
Substituting Eq. (8.1.13) into Eq. (8.1.12) and integrating give
U 1 12EI 2
¼ u (8.1.14)
2 2 h3
Therefore, the elastic energy of the two columns is
12EI 2
U¼ u (8.1.15)
h3
and its variation
12EI
dU ¼ 2 udu (8.1.16)
h3
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 273

(b) Evaluation of the kinetic energy T . The kinetic energy is due to the velocity
u_ ðt Þ of the mass m of the horizontal beam as well as to the velocity u
_ ðx, t Þ
of the mass m
at points x of the column axis. Thus, we have
Z h
1 1

½u
_ ðx, t Þ dx
2
T ¼ m u_ + 2
2
m (8.1.17)
2 0 2
By virtue of Eq. (8.1.1), the above relation is written as
Z h
1
T ¼ m u_ 2 + m
u_ 2 ½ ðx Þ2 dx (8.1.18)
2 0

Then using Eq. (8.1.6) and integrating give

1
T ¼ m u_ 2 + 0:371mh

u_ 2 (8.1.19)
2
and

Þud
dT ¼ ðm + 0:742h m _ u_ (8.1.20)

(c) Evaluation of the virtual work dWnc . This is due to the nonconservative
external force pðt Þ. This is
dWnc ¼ pðt Þdu (8.1.21)
Now we proceed to the derivation of the equation of motion using:
1. Hamilton’s principle. Substituting Eqs. (8.1.16), (8.1.20), (8.1.21) into
Eq. (1.7.13) and taking into account that A ¼ 0, we obtain
Z t2  
EI
24 3 udu  ðm + 0:742h m
Þud
_ u_  pðt Þdu dt ¼ 0 (8.1.22)
t1 h
which after elimination of du_ using integration by parts yields the equa-
tion of motion
m ∗ u€ + k ∗ u ¼ p∗ ðt Þ (8.1.23)
where
24EI


m ∗ ¼ m + 0:741mh, k∗ ¼ , p∗ ðt Þ ¼ pðt Þ (8.1.24)
h3

2. Lagrange equations. The equation of motion will result from

Eq. (1.8.11) by setting qi ¼ u, A ¼ 0 and Qi ¼ pðt Þ. This gives
 
d ∂T ∂T ∂U
 + ¼ pðt Þ (8.1.25)
dt ∂u_ ∂u ∂u
which by virtue of Eqs. (8.1.15), (8.1.19) yields Eq. (8.1.23).
274 PART I Single-degree-of-freedom systems

3. The principle of virtual work. The principle of virtual work stated in

Section 1.6 for rigid bodies can also be extended to deformable bodies,
but now we must take account of the work done by the internal forces in
riding through the virtual deformation. Thus, for a deformable body, the
principle of virtual work becomes

dWex ¼ dWin (8.1.26)

where dWex and dWin denote the virtual work of the external and inter-
nal forces, respectively.
For the system of Fig. 8.1.1, the external forces are the excitation
force pðt Þ and the inertial forces. Hence
Z h
€ 2
dWex ¼ m udu
u€
m
ðx, t Þd u
ðx, t Þdx + pðt Þdu
0

which by virtue of Eqs. (8.1.1), (8.1.6) becomes

€  0:742mhdu
dWex ¼ m udu
+ pðt Þdu (8.1.27)

The virtual work of the internal forces for the two columns is
obtained from the relationa
Z
dWin ¼ 2 sx dex dV (8.1.28)
V

where V is the volume of the column.

Taking into account that sx ¼ Mx y=I , sx ¼ Eex , Mx ¼ EI u
00 ðx, t Þ,
we obtain from Eq. (8.1.28)

12EI
dWin ¼ 2 udu (8.1.29)
h3
Substituting Eqs. (8.1.27), (8.1.29) into Eq. (8.1.26) yields the equation
of motion (8.1.23).

a
The expression for the strain energy of a beam due to bending is derived from the strain energy
of the beam by taking its variation ([1], Chap. 1):
Z Z
1 1
Win ¼ sx ex dV ¼ Es2 dV (a)
2 V 2 V x
which by taking the variation gives
Z Z
dWin ¼ Eex dex dV ¼ sx dex dV (b)
V V
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 275

FIG. 8.1.3 Dependence of T*/T on the ratio of the column mass over total beam mass in the frame
of Fig. 8.1.1

If the mass of the column is assumed concentrated at the ends of the col-
umns, the coefficients of the corresponding equation of motion result as
24EI


m ∗ ¼ m + mh, k∗ ¼ , p∗ ðt Þ ¼ pðt Þ (8.1.30)
h3
We observe that the generalized mass is less by 0:259mh.
Fig. 8.1.3


shows the variation of the ratio T ∗ =T versus the ratio mh=m, where T ∗
is the natural period of the generalized single-degree-of-freedom system
and T the the period of the model in Fig. 8.1.1b. We observe that the lumped
mass assumption has a small influence on the natural period when the mass
of the columns with respect to the mass of the horizontal beam is small.
Illustrative examples facilitating the comprehension of all concepts are pre-
sented and the pertinent bibliography with recommended references for fur-
ther study is also included. The chapter is enriched with problems to be
solved.

8.2 Generalized single-degree-of-freedom systems

After the introductory example in the previous section, we can extend the dis-
cussed method to approximate the response of a more complex system with a gen-
eralized SDOF system. When the members of the system are undeformed, the
substitute SDOF system represents the actual response of the system. Such exam-
ples were discussed in Chapter 1. Example 1.7.3 is a representative case. The
shape function of that example is shown in Fig. E1.11. However, when the mem-
bers connecting the lumped masses are deformable, then the generalized SDOF
system simulating the actual one can be used to approximate its response. The
accuracy of the approximation depends on the selection of the shape function.
276 PART I Single-degree-of-freedom systems

Nevertheless, the method allows us to approximate easily the dynamic character-

istics of a complex system, thus circumventing exact solution methods.
To illustrate the application of the method to complex systems, we consider
the beam of Fig. 8.2.1a, which is fixed at point 1 and simply supported at point 2.
Moreover, the beam rests on the nonhomogeneous Winkler’s type elastic foun-
dation with variable reaction modulus k ðx Þ. The foundation reacts also with dis-
tributed linear damping having coefficient cðx Þ. The cross-section of the beam
is variable, hence I ¼ I ðx Þ and m ¼ m ðx Þ. The beam caries the lumped masses
mi with rotational inertia Ii at points x ¼ xi , i ¼ 1, 2, …,N . The beam is loaded
by the distributed load pðx, t Þ and the axial load P at the end cross-section 2.
The beam may also be subjected to concentrated loads and/or concentrated and
distributed moments. However, for the convenience of the illustration of the
method, these loads are not considered.
To treat the structure as a generalized SDOF, we will seek the transverse
deflection in the form of Eq. (8.1.1), that is, u
ðx, t Þ. The shape function ðx Þ
will be selected so that
0 0
ð0Þ ¼ 0, ð0Þ ¼ 0, ðLÞ ¼ 0, ðLÞ 6¼ 0 (8.2.1)
The deflection curve of the beam fixed at one end and simply supported at
the other is a geometrically admissible function and thus can be used as a shape
function. Taking ðL=2Þ ¼ 1, we may write
u
ðx, t Þ ¼ ðx Þu ðt Þ (8.2.2)
Obviously, it is u
ðL=2, t Þ ¼ u ðt Þ, hence u ðt Þ represents the displacement at
the middle of the beam.
We will use Hamilton’s principle to derive the equation of motion of
the generalized SDOF system. For this purpose, we evaluate the quantities
U , K , dWnc , and A.

(a)

(b)

(c)

(d)
FIG. 8.2.1 Beam resting on foundation with nonhomogeneous elastic and damping reaction.
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 277

(a) Evaluation of the elastic energy U . This energy consists of the strain energy
of the beam due to bending and the elastic energy of the springs of Winkler’s
model. That is,
Z Z
1 L 00 2 1 L
U¼ EI ðx Þ½u
ðx, t Þ dx + k ðx Þ½u
ðx, t Þ2 dx (8.2.3)
2 0 2 0
which yields
Z L Z L
dU ¼ EI ðx Þu
00 ðx, t Þd u
00 ðx, t Þdx + k ðx Þu
ðx, t Þd u
ðx, t Þdx (8.2.4)
0 0

Eq. (8.1.1) gives du
ðx, t Þ ¼ ðx Þdu, du
00 ðx, t Þ ¼ 00

ðx Þdu. Thus,
Eq. (8.2.4) becomes
dU ¼ k ∗ udu (8.2.5)
where
Z L Z L
00 2
k∗ ¼ EI ðx Þ½ ðx Þ dx + k ðx Þ½ ðx Þ2 dx (8.2.6)
0 0

(b) Evaluation of the kinetic energy T . This energy consists of the kinetic
energy of the distributed mass m ðx Þ and the kinetic energy of the lumped
masses mi .That is
Z h i2
1 L 1X N
1X N
0
m ðx Þ½u
_ ðx, t Þ dx + mi ½u
_ ðxi , t Þ + Ii u
_ ðxi , t Þ
2 2
T¼
2 0 2 i¼1 2 i¼1
(8.2.7)
0
where u
_ ðx, t Þ and u
_ ðx, t Þ are the transverse velocity and the angular veloc-
ity, respectively, at the cross-section x. Eq. (8.2.7) gives
Z L X
N
dT ¼ m ðx Þu
_ ðx, t Þdu
_ ðx, t Þdx + mi u
_ ðxi , t Þdu
_ ðxi , t Þ
0 i¼1
(8.2.8)
X
N
0 0
+ Ii u
_ ðxi , t Þdu
_ ðxi , t Þ
i¼1

which by virtue of Eq. (8.1.1) results in

_ u_
dT ¼ m ∗ ud (8.2.9)
where
Z L X
N X
N
Ii ½ 0 ðxi Þ
2
m∗ ¼ m ðx Þ½ ðx Þ2 dx + mi ½ ðxi Þ2 + (8.2.10)
0 i¼1 i¼1
278 PART I Single-degree-of-freedom systems

(c) Evaluation of the virtual work dWnc of the nonconservative forces. This is
due to the damping force fD ðx, t Þ ¼ cðx Þu
_ ðx, t Þ and the external force
pðx, t Þ. Hence we have
Z L Z L
dWnc ¼  cðx Þu
_ ðx, t Þd u
ðx, t Þdx + pðx, t Þd u
ðx, t Þdx (8.2.11)
0 0

which by virtue of Eq. (8.1.1) gives

dWnc ¼ ½c∗ u_ + p∗ ðt Þdu (8.2.12)
where
Z L
c∗ ¼ cðx Þ ðx Þdx (8.2.13a)
0

Z L
p∗ ðt Þ ¼ pðx, t Þ ðx Þdx (8.2.13b)
0

(d) The potential A of the conservative forces. This is due to the work of the
axial force P. Obviously, if the axial deformation is neglected, it is
A ¼ 0. But if we consider shortening of the beam due to bending, that is,
if we adopt large displacements, then the shortening is expressed by the
nonlinear term of the strain-displacement relation. Thus, according to the
nonlinear theory of elasticity, we have [1, 2]
 
1 ∂u
2
ex ¼ (8.2.14)
2 ∂x

and the total shortening is

Z L
1
½u
0 ðx, t Þ dx
2
DL ¼ (8.2.15)
2 0

Consequently
Z L
1
½u
0 ðx, t Þ dx
2
A ¼ PDL ¼  P (8.2.16)
2 0

and
Z L
dA ¼ P u
0 ðx, t Þd u
0 ðx, t Þdx (8.2.17)
0

which by virtue of Eq. (8.1.1) gives

dA ¼ k
∗ udu (8.2.18)
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 279

where
Z L
k
∗ ¼ P ½ 0 ðx Þ dx
2
(8.2.19)
0

Substituting Eqs. (8.2.5), (8.2.9), (8.2.12), (8.2.18) into Hamilton’s prin-

ciple, Eq. (1.7.13), results in
Z t2 Z t2

_ u_  k
∗ udu dt 
k ∗ udu  m ∗ ud ½c∗ u_ + p∗ ðt Þdudt ¼ 0 (8.2.20)
t1 t1

or after integration by part to eliminate the derivative from du_

Z t2
 
m ∗ u€ + c∗ u_ + k ∗  k
∗ u  p∗ ðt Þ dudt ¼ 0 (8.2.21)
t1

from which we obtain the equation of motion of the generalized SDOF

system

m ∗ u€ + c∗ u_ + k ∗  k
∗ u ¼ p∗ ðt Þ (8.2.22)
Looking at the equation of motion (8.2.22), we draw a useful conclusion
regarding the stability of the structure. It is evident that the stiffness of the
system vanishes when the axial load P takes the critical value
Z h
½ 0 ðx Þ dx
2
Pcr ¼ k ∗ = (8.2.23)
0

This value of the axial force is the buckling load of the structure.
A consequence of this is the vanishing of the natural frequency,
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

w¼ k ∗  k
∗ =m ∗ ¼ 0. Therefore, a method of determining the buckling
load is to find the value of the compressive axial load, which annuls the nat-
ural frequency. This method is known as the dynamic criterion for buckling.
The method of approximating the continuous systems by a generalized
SDOF system can be successful with regard to the displacement. However,
we should be careful when we need to determine the stress resultants from
the obtained deflection curve using the known relations
M ðx, t Þ ¼ EI u
00 ðx, t Þ (8.2.24)
000
Q ðx, t Þ ¼ EI u
ðx, t Þ (8.2.25)
The stress resultants resulting from the above relations may deviate con-
siderably from the actual ones. This is illustrated by the following example.
We consider the cantilever beam of Fig. 8.2.2. The function
px
ðx Þ ¼ 1  cos (8.2.26)
2L
280 PART I Single-degree-of-freedom systems

is a geometrically admissible function, that is,

0 0
ð0Þ ¼ 0, ð0Þ ¼ 0, ðLÞ ¼ 1, ðLÞ 6¼ 0 (8.2.27)

FIG. 8.2.2 Cantilever beam as generalized SDOF system

Therefore, it can be used as a shape function for the cantilever beam and
the resulting displacement is

px 
u
ðx, t Þ ¼ 1  cos u ðt Þ (8.2.28)
2L
Eq. (8.2.25) gives

p 2 px
Q ðx, t Þ ¼ EI sin u ðt Þ (8.2.29)
2L 2L
which results in
Q ð0, t Þ ¼ 0 (8.2.30)
This result is absurd. Nevertheless, this problem can be circumvented
if the stress resultant is evaluated using the procedure described in
Example 8.2.1.

Example 8.2.1 The industrial chimney of length L ¼ 75m shown in Fig. E8.1
consists of the outer reinforced concrete shell, which supports the linings. The
thickness of the thermal insulation layer is ti ¼ 0:10m and that of the refractory
layer tr ¼ 0:10m. The chimney is fixed on the ground.
1. Determine the natural frequency of the structure using two different shape
functions: (i) the elastic curve of a cantilever with constant cross-section
under a uniformly distributed load, and (ii) the first vibration mode of a
cantilever with constant cross-section.
2. Study the dynamic response of the chimney subjected to the impulsive wind
pressure shown in Fig. E8.1b. The analysis will be done using the shape
function that produces more accurate results.
3. Compute the bending moment and shear force at the base of the chimney and
give their expressions as a function of time.
4. Compute the dynamic magnification factor D ¼ max jRðt Þj for the
displacement.
Data:
Specific weight of reinforced concrete: g b ¼ 24kN=m3
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 281

Specific weight of thermal insulation: g i ¼ 0:7kN=m3

Specific weight of refractory bricks: g r ¼ 20kN=m3
Peak value of wind pressure: pw ¼ 1kN=m2
Time duration of wind pressure: t1 ¼ 2s

(a) (b)
FIG. E8.1 Industrial chimney and loading.

Modulus of elasticity of reinforced concrete: E ¼ 2:1  107 kN=m2

Solution
Computation of m ∗ ,k ∗ ,p∗ ðt Þ.
Mean radius of reinforced concrete shell: r ðx Þ ¼ 3:05  0:0243x
Mean radius of thermal insulation: ri ðx Þ ¼ 2:85  0:0233x
Mean radius of the refractory lining: rr ðx Þ ¼ 2:75  0:0233x
Thickness of reinforced concrete: t ðx Þ ¼ 0:30  0:002x
The mass and the moment of inertia
m ðx Þ ¼ ½2pr ðx Þt ðx Þg b + 2pri ðx Þti g i + 2prr ðx Þtr g r =g (1)
I ðx Þ  pr ðx Þ3 t ðx Þ (2)

1. The natural frequency of the chimney

(i) The shape function is the elastic curve of a cantilever with constant
cross-section under a uniformly distributed load p ¼ 8EI =L4 , that is,
282 PART I Single-degree-of-freedom systems

1 2
ðx Þ ¼ 6x  4x3 + x4 , x ¼ x=L (3)
3
Computation of the integrals using MATLAB gives
Z L
EI ðx Þ½ 00 ðx Þ dx ¼ 2977:6218
2
k ¼
∗ (4)
0
Z L
m∗ ¼ m ðx Þ½ ðx Þ2 dx ¼ 120:9955 (5)
0

Hence
pffiffiffiffiffiffiffiffiffiffiffiffiffi
w¼ k ∗ =m ∗ ¼ 4:9608 (6)

(ii) The shape function is the first vibration mode of a cantilever with a
uniform cross-section (see Section 8.3.3.2)
1
ðx Þ ¼ ½ cosh lx  cos lx  0:7341ð sinh lx  sinh lx Þ, l ¼ 1:8751=L (7)
3
Z L
EI ðx Þ½ 00 ðx Þ dx ¼ 2715:2000
2
k∗ ¼ (8)
0
Z L
m∗ ¼ m ðx Þ½ ðx Þ2 dx ¼ 116:8332 (9)
0
pffiffiffiffiffiffiffiffiffiffiffiffiffi
w¼ k ∗ =m ∗ ¼ 4:8208 (10
Obviously, case (ii) is more accurate because it yields a smaller
natural frequency (see Chapter 12).
2. The dynamic response of the chimney
The outer diameter of the chimney is
d ðx Þ ¼ 6:40  0:0507x (11)
hence the wind load per unit length is given by
pðx, t Þ ¼ pw d ðx Þf ðt Þ (12)
where
8
> t
>
> 2 if 0  t  t1 =2
>
<  t1 
f ðt Þ ¼ t (13)
>
> 2 1 if t1 =2  t  t1
>
> t
: 1
0 if t1  t
hence the peak load of the generalized SDOF system is
Z L
p∗w ¼ pw d ðx Þ ðx Þdx ¼ 106:81 (14)
0
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 283

The dynamic response will be obtained from the solution of the equation
of motion
m ∗ u€ + k ∗ u ¼ p∗w f ðt Þ (15)
The solution of Eq. (15) is obtained analytically in three phases.
Phase Ι. 0  t  1. The solution is given by Eq. (3.4.15)
 
2p∗w sin wt
uI ¼ t ¼ 0:039337t  0:008160 sin 4:8208t (16a)
k ∗ t1 w
Phase ΙΙ. 1  t  2, et ¼ t  1. The solution is given by Eqs. (3.3.14),
(3.5.10)
 
u_ I ð1Þ p∗w sin wet e
uII ðt Þ ¼ sin wet + uI ð1Þcos wet + 1  cos wet + t (16b)
w k∗ w
where
uI ð1Þ ¼ 0:04745m, u_ I ð1Þ ¼ 0:03508m=s
Phase IΙΙ. 2  t, ^t ¼ t  2. The solution is given by Eq. (2.2.13)
u_ II ð2Þ
uIII ðt Þ ¼ sin w^t + uII ð2Þcos w^t (16c)
w
where
uII ð2Þ ¼ 0:01447m, u_ II ð2Þ ¼ 0:00759m=s
Fig. E8.2 shows the graphical representation of the response ratio Rðt Þ,
from which we conclude that max jRðt Þj occurs in phase ΙI. The numerical
solution gives D ¼ 1:2673 occurring at t ¼ 1:12s.

FIG. E8.2 Graphical representation of the response ratio in Example 8.2.1

284 PART I Single-degree-of-freedom systems

The stress resultants at the base of the chimney are obtained by considering
the equilibrium of all external forces, that is, Q ð0, t Þ, M ð0, t Þ, the wind pres-
sure pðx, t Þ, and the inertia force fI ðx, t Þ. Thus, referring to Fig. E8.3, we have

FIG. E8.3 External forces in Example 8.2.1

Z L Z L
Q ð0, t Þ + fI ðx, t Þdx  pðx, t Þdx ¼ 0
0 0
Z L Z L
M ð0, t Þ  xf I ðx, t Þdx + xpðx, t Þdx ¼ 0
0 0

By setting in the previous equilibrium equations

fI ðx, t Þ ¼ m ðx Þu€

ðx, t Þ ¼ m ðx Þ ðx Þu€ðt Þ and pðx, t Þ ¼ pw d ðx Þf ðt Þ
we obtain after the evaluation of the integrals
Z L Z L
Q ð0, t Þ ¼ u€ðt Þ m ðx Þ ðx Þdx + pw f ðt Þ d ðx Þdx
0 0
Z L Z L
M ð0, t Þ ¼ u€ðt Þ m ðx Þ ðx Þxdx  pw f ðt Þ d ðx Þxdx
0 0

which yield
Q ð0, t Þ ¼ 211:6454u€ðt Þ + 337:4063f ðt Þ (17)
M ð0, t Þ ¼ 10421:7751u€ðt Þ  10870:3125f ðt Þ (18)

8.3 Continuous systems

8.3.1 Introduction
The modeling of a structure as a continuous system can accurately express its
real response, provided that the employed constitutive relations represent
the actual physical law that relates the stresses and strains. As we saw in
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 285

Section 1.1, this modeling leads to partial differential equations of the

hyperbolic type, namely, for the vibrations of the beam [3]
∂4 u ∂2 u
EI +

m ¼ pðx, t Þ (8.3.1)
∂x 4 ∂t 2
where u ¼ u ðx, t Þ represents the transverse displacement, EI the bending stiff-
ness of the beam, m
the mass per unit length, and pðx, t Þ the transverse load
density.
Similarly, for the vibrations of the thin plate [1]
∂4 u ∂4 u ∂4 u r ∂2 u pðx, y, t Þ
+ 2 + + ¼ (8.3.2)
∂x 4 ∂x 2 ∂y 2 ∂y 4 D ∂t 2 D
where u ¼ u ðx, y, t Þ represents the transverse displacement, D the bending
stiffness of the plate, r the surface mass density, and pðx, y, t Þ the transverse
load density. The above equations are more complicated for the shell and
become even more complicated if the thickness of the structural element is var-
iable or the response is nonlinear [1, 4]. The effort to obtain analytical solutions
of the equations governing the response of continuous systems has inspired
mathematicians and applied physicists to develop the theory of partial differen-
tial equations [5, 6]. This, however, did not satisfy people who wanted to use the
derived solutions in engineering praxis because these solutions treat a very lim-
ited number of problems. The development of approximate and especially of
modern numerical solutions (e.g., FEM, BEM, MM) has provided us with
efficient tools to solve problems of mathematical physics and engineering
described by complicated ordinary and partial differential equations. The ques-
tion that arises nowadays and demands investigation is how reliably these equa-
tions describe the actual response of physical systems. The system identification
based on experimental data can answer this question. Though the available pro-
fessional computer codes efficiently analyze engineering problems, they cannot
give adequate insight into the dynamic response of structures. Therefore, the use
of analytical methods is necessary to understand the response of structures and
to develop simple and desired-accuracy solutions, which are very useful to
check new numerical methods. Because the scope of this book is the analysis
of structures consisting of straight-line structural elements, the discussion will
be limited to the solution of the equation of the vibrating beam. Results of this
analysis are used in SDOF generalized systems.

8.3.2 Solution of the beam equation of motion

The transverse flexural vibrations of the beam with constant stiffness are gov-
erned by Eq. (8.3.1) subjected to the specified boundary and initial conditions.
The solution u ¼ u ðx, t Þ can be obtained in the form
u ¼ uh + up (8.3.3)
286 PART I Single-degree-of-freedom systems

where uh is the solution of the homogeneous equation

∂4 uh ∂2 uh
EI
2 ¼0
+m (8.3.4)
∂x 4 ∂t
and up a particular solution of the nonhomogeneous equation
∂4 up ∂2 up
EI +

m ¼ pðx, t Þ (8.3.5)
∂x 4 ∂t 2

8.3.3 Free vibrations of beams

For pðx, t Þ ¼ 0, Eq. (8.3.1) yields the homogeneous equation, which represents
the equation of free vibrations of the beam. Assuming constant mass,
m
¼ constant, and dropping the subscript for convenience in Eq. (8.3.4),
we write this equation as
∂4 u ∂2 u
EI + m
¼0 (8.3.6)
∂x 4 ∂t 2
Eq. (8.3.6) is solved using the method of separation of variables for partial
differential equations [6], that is, the solution is sought in the form
u ðx, t Þ ¼ fðx ÞY ðt Þ (8.3.7)
This equation indicates that the free vibration motion is the product of a
specific deflection shape fðx Þ and an amplitude Y ðt Þ depending on time.
Introducing the above expression for the deflection into Eq. (8.3.6) gives

ðx ÞY€ ðt Þ ¼ 0
EI fðivÞ ðx ÞY ðt Þ + mf (8.3.8)
which is written as
EiI fðivÞ ðx Þ Y€ ðt Þ
¼ (8.3.9)
m
fðx Þ Y ðt Þ
For the obvious problem associated with the possibility of the vanishing of
the product fðx ÞY ðt Þ, we refer to [7, 8]. Here, we just state that the roots of the
denominator in Eq. (8.3.9) coincide with those of the nominator, an assumption
that permits this division.
Because the left side of Eq. (8.3.9) is independent of t and the right side
independent of x, this equation is valid only if both sides are equal to a
constant l. Hence
EI fðivÞ ðx Þ Y€ ðt Þ
¼ ¼l
m
fðx Þ Y ðt Þ
or
Y€ ðt Þ + lY ðt Þ ¼ 0 (8.3.10a)
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 287

m
fðivÞ ðx Þ  l
f ðx Þ ¼ 0 (8.3.10b)
EI
If l  0, the solution of Eq. (8.3.10a) does not represent an oscillatory
motion. Therefore, l must be a positive constant, l ¼ w2 . Thus, the solution
of Eq. (8.3.10a) is (see Section 2.2)
Y ðt Þ ¼ A cos wt + B sin wt (8.3.11)
where A,B are arbitrary constants depending on the initial conditions Y ð0Þ and
Y_ ð0Þ. Thus we have

Y_ ð0Þ
Y ðt Þ ¼ Y ð0Þcos wt + sin wt (8.3.12)
w
Apparently, w is the natural frequency of the vibration, unknown in the first
instance.
Similarly, Eq. (8.3.10b) is written as

fðivÞ ðx Þ  b 4 fðx Þ ¼ 0 (8.3.13)

where
m

b 4 ¼ w2 (8.3.14)
EI
The solution of Eq. (8.3.13) is sought in the form

fðx Þ ¼ C 0 ekx (8.3.15)

which is introduced into Eq. (8.3.13) to give the characteristic equation

k 4  b4 ¼ 0 (8.3.16)
whose roots are
k1, 2 ¼ ib, k3, 4 ¼ b (8.3.17)
Using each of these roots in Eq. (8.3.15) yields four terms, which are added
to give the general solution

fðx Þ ¼ C10 eibx + C20 eibx + C30 ebx + C40 ebx (8.3.18)
where C 0 i ði ¼ 1, 2, 3, 4Þ are arbitrary complex constants.
Using Euler’s formula (2.2.8) and the expressions of the hyperbolic sine and
cosine, Eq. (8.3.18) becomes
fðx Þ ¼ C1 cos bx + C2 sin bx + C3 cosh bx + C4 sinh bx (8.3.19)
in which Ci ði ¼ 1, 2, 3, 4Þ are new arbitrary constants related to
C 0 i ði ¼ 1, 2, 3, 4Þ, and can be determined from the boundary (support) condi-
tions of the one-span beam.
288 PART I Single-degree-of-freedom systems

In the following, the above-described procedure for free-vibration analysis

is illustrated by analyzing the simply supported and the cantilever beams.
For further reading, the reader is advised to look in Refs. [9–13].

8.3.3.1 The simply supported beam

For a simply supported beam of length L (Fig. 8.3.1) the support conditions are
realized as:
u ð0, t Þ ¼ 0, M ð0, t Þ ¼ EI u 00 ð0, t Þ ¼ 0 (8.3.20a)
u ðL, t Þ ¼ 0, M ðL, t Þ ¼ EI u 00 ðL, t Þ ¼ 0 (8.3.20b)
which by virtue of Eq. (8.3.7) become
fð0ÞY ðt Þ ¼ 0, f00 ð0ÞY ðt Þ ¼ 0 (8.3.21a)
fðLÞY ðt Þ ¼ 0, f00 ðLÞY ðt Þ ¼ 0 (8.3.21b)

FIG. 8.3.1 Vibration modes and natural frequencies of a uniform simply supported beam.

Inasmuch as Eqs. (8.3.21a), (8.3.21b) are valid for all values of t, they are
satisfied only if
fð0Þ ¼ 0, f00 ð0Þ ¼ 0 (8.3.22a)
fðLÞ ¼ 0, f00 ðLÞ ¼ 0 (8.3.22b)
Introducing Eq. (8.3.19) into Eq. (8.3.22a) yields
C1 + C3 ¼ 0 (8.3.23a)
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 289

C1 + C3 ¼ 0 (8.3.23b)
from which we obtain
C1 ¼ C3 ¼ 0 (8.3.24)
Further, introducing Eq. (8.3.19) into Eq. (8.3.22b), we obtain
C2 sin bL + C4 sinh bL ¼ 0 (8.3.25a)

b2 ½C2 sin bL + C4 sinh bL ¼ 0 (8.3.25b)

Eqs. (8.3.25a), (8.3.25b) provide the system of two homogeneous algebraic

equations for the evaluation of C2 ,C4 , that is,
   
sin bL sinh bL C2 0
¼ (8.3.26)
 sin bL sinh bL C4 0
The system has a nontrivial solution if its determinant vanishes, that is,
sin bL sinh bL ¼ 0 (8.3.27)
Nevertheless, because sinh bL 6¼ 0, it must be
sin bL ¼ 0 (8.3.28)
which is satisfied if
np
bn ¼ , n ¼ 1, 2, … (8.3.29)
L
Introducing this value of b n into Eq. (8.3.14) gives the corresponding natural
frequencies
rffiffiffiffiffiffiffiffiffiffi
EI
wn ¼ n p
2 2
(8.3.30)

4
mL
Substituting Eq. (8.3.28) into Eq. (8.3.26), we obtain
" #( ) ( )
0 sinh bL C2 0
¼ (8.3.31)
0 sinh bL C4 0

which yields C4 ¼ 0 and C2 ¼ arbitrary, therefore

fðx Þ ¼ C2 sin bx (8.3.32)
On the basis of Eq. (8.3.29), we obtain the mode shapes

np 
fn ðx Þ ¼ C2 sin x , n ¼ 1, 2, … (8.3.33)
L
The first three of these mode shapes are shown in Fig. 8.3.1 along with their
natural frequencies
290 PART I Single-degree-of-freedom systems

8.3.3.2 The cantilever beam

For a cantilever beam of length L the support conditions are realized as:
u ð0, t Þ ¼ 0, u 0 ð0, t Þ ¼ 0 (8.3.34a)
00 000
M ðL, t Þ ¼ EI u ðL, t Þ ¼ 0, Q ðL, t Þ ¼ EI u ðL, t Þ ¼ 0 (8.3.34b)
which by virtue of Eq. (8.3.7) become
fð0ÞY ðt Þ ¼ 0, f0 ð0ÞY ðt Þ ¼ 0 (8.3.35a)
f00 ðLÞY ðt Þ ¼ 0, f000 ðLÞY ðt Þ ¼ 0 (8.3.35b)
Because Eqs. (8.3.35a), (8.3.35b) are valid for all values of t, they are
satisfied only if
fð0Þ ¼ 0, f0 ð0Þ ¼ 0 (8.3.36a)
f00 ðLÞ ¼ 0, f000 ðLÞ ¼ 0 (8.3.36b)
Introducing Eq. (8.3.19) into Eqs. (8.3.36a), (8.3.36b) yields the system of
equations for the evaluation of the coefficients Ci
2 38 9 8 9
1 0 1 0 > C1 >
> > > >0> >
6 0 1 0 1 7 < C2 = < 0 =
6 7 ¼ (8.3.37)
4  cos bL  sin bL coshbL sinh bL 5> C3 > > 0 >
: >
> ; > : > ;
sin bL  cos bL sinh bL cosh bL C4 0
Eq. (8.3.37) has a nontrivial solution if the determinant of the coefficient
matrix vanishes, that is,
 
 1 0 1 0 
 
 0 1 0 1 
 ¼0 (8.3.38)
  cos bL  sin bL cosh bL sinh bL 

 sin bL  cos bL sinh bL cosh bL 
or after evaluation of the determinant
cos bL cosh bL + 1 ¼ 0 (8.3.39)
The roots bn , n ¼ 1, 2, … of Eq. (8.3.39) are used in Eq. (8.3.14) to obtain
the natural frequencies of the vibrating cantilever beam.
Substituting Eq. (8.3.39) into Eq. (8.3.37) and solving the resulting
homogeneous system of equations, we obtain
cos b n L + cosh bn L
C3 ¼ C1 , C4 ¼ C2 , C2 ¼  C1 (8.3.40)
sin b n L + sinh bn L
which are introduced into Eq. (8.3.19) to give the mode shapes
 
cos b n L + cosh bn L
fn ðx Þ ¼ C1 cos bn x  cosh b n x  ð sin bn x  sinh bn x Þ
sin b n L + sinh bn L
(8.3.41)
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 291

Table 8.3.1 gives the first five roots of bn L of Eq. (8.3.39). They have been
obtained numerically using the function fsolve of MATLAB. Note that for n > 3
they can be obtained from the relation
p
bn L  ð2n  1Þ (8.3.42)
2
The first three of these mode shapes are shown in Fig. 8.3.2 along with their
natural frequencies

TABLE 8.3.1 First five roots of Eq. (8.3.39).

n bn L
1 1.8751040688
2 4.6940911329
3 7.8547574382
4 10.995540734
5 14.137168391

FIG. 8.3.2 Vibration modes and natural frequencies of a uniform cantilever beam.

8.3.4 Orthogonality of the free-vibration modes

The infinite set of free-vibration mode shapes F : ffn ðx Þg, n ¼ 1, 2, … has a
nice property. They are orthogonal in the interval ½0, L, that is, they satisfy
the orthogonality condition (see Section 3.6.2)
292 PART I Single-degree-of-freedom systems

Z L
0 if r 6¼ n
fr ðt Þfn ðt Þdt ¼ (8.3.43)
0 cn if r ¼ n
for any two functions fn ,fr F, m, nN . In the language of partial differential
equations, the free-vibration modes are called the eigenfunctions of the eigenvalue
problem described by the differential equation (8.3.13) and its boundary condi-
tions. The orthogonality condition is readily proved by proceeding as follows.
The mode shapes fn ,fr satisfy Eq. (8.3.13), that is,
fðnivÞ ðx Þ  b4n fn ðx Þ ¼ 0 (8.3.44a)

fðrivÞ ðx Þ  b 4r fr ðx Þ ¼ 0 (8.3.44b)
Multiplication of Eq. (8.3.44a) by fr ðx Þ and integrating over the interval
½0, L gives
Z L Z L
b4n fn ðx Þfr ðx Þdx ¼ fðnivÞ ðx Þfr ðx Þdx (8.3.45)
0 0

Further, integrating the right side of the above equation twice by parts gives
Z Z L
L
00 00

00 0
L
000 L
fn ðx Þfr ðx Þdx  fn ðx Þfr ðx Þ 0 + fn ðx Þfr ðx Þ 0 ¼ bn
4
fn ðx Þfr ðx Þdx
0 0
(8.3.46)
Obviously, the terms in square brackets in the above equation vanish if
either end of the beam is simply supported, fixed, or free. Thus, we have
Z L Z L
4
bn fn ðx Þfr ðx Þdx ¼ f00n ðx Þf00r ðx Þdx (8.3.47a)
0 0

Similarly, multiplying Eq. (8.3.44b) by fn ðx Þ, integrating over the interval

½0, L, and performing the integrations by parts yields the symmetric relation
Z L Z L
b4r fn ðx Þfr ðx Þdx ¼ f00n ðx Þf00r ðx Þdx (8.3.47b)
0 0

Subtracting Eq. (8.3.47b) from Eq. (8.3.47a) gives

Z
 L
0 ¼ b 4n  b4r fn ðx Þfr ðx Þdx (8.3.48)
0

which for b 4n 6¼ b4r results in the orthogonality condition for the mode shapes
Z L
fn ðx Þfr ðx Þdx ¼ 0 (8.3.49)
0

It can also be shown that the set F : ffn ðx Þg is complete, that is there is no
other function outside the set F, which satisfies the condition (8.3.49).
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 293

8.3.5 Forced vibrations of beams

The method of separation of variables can be employed to study the forced
vibrations of one-span beams by solving Eq. (8.3.1). The beam may be
subjected to initial conditions
u ðx, 0Þ ¼ f ðx Þ, u_ ðx, 0Þ ¼ g ðx Þ
The total flexural displacement of the beam is obtained as the superposition
of all modal contributions, namely
X
∞
u ðx, t Þ ¼ fn ðx ÞYn ðt Þ (8.3.50)
n¼1

Substituting the previous equation into Eq. (8.3.1), multiplying by fr ðx Þ,

and integrating over the interval ½0, L, we obtain
X∞  Z L  X ∞  Z L 
Yn ð t Þ ðivÞ
EI fn ðx Þfr ðx Þdx + €
Y n ðt Þ
n ðx Þfr ðx Þdx
mf
n¼1 0 n¼1 0
Z L
¼ fr ðx Þpðx, t Þ (8.3.51)
0

Using now the orthogonality condition (8.3.49) and taking into account
Eqs. (8.3.45), (8.3.14), we obtain
Mn Y€ n ðt Þ + Kn Yn ðt Þ ¼ Pn ðt Þ (8.3.52)
where
Z L Z L


Mn ¼ m f2n ðx Þdx, Kn ¼ w2n Mn , Pn ¼ fn ðx Þpðx, t Þdx (8.3.53)
0 0

denote the modal mass, the modal stiffness, and the modal force. These quan-
tities are also referred to as the generalized mass, the generalized stiffness, and
the generalized force, respectively.
The solution of Eq. (8.3.52) is given by Eq. (3.3.14), that is,
Z t
Y_ n ð0Þ 1
Y n ðt Þ ¼ sin wn t + Yn ð0Þcos wn t + Pn ðτÞsin ½wn ðt  τÞdτ
wn M n wn 0
(8.3.54)
The initial conditions Yn ð0Þ, Y_ n ð0Þ for the time function result from
Eq. (8.3.50). This yields
X
∞
u ðx, 0Þ ¼ fn ðx ÞYn ð0Þ ¼ f ðx Þ (8.3.55)
n¼1

Multiplying the previous equations by fn ðx Þ, integrating over the interval

½0, L, and using the orthogonality condition (8.3.49), we obtain
294 PART I Single-degree-of-freedom systems

Z L
f ðx Þfn ðx Þdx
Yn ð 0Þ ¼ 0
Z L
(8.3.56a)
f2n ðx Þdx
0

Similarly, we obtain
Z L
g ðx Þfn ðx Þdx
Y_ n ð0Þ ¼ 0
Z L
(8.3.56b)
f2n ðx Þdx
0

Example 8.3.1 A simply supported beam of length L is subjected to a suddenly

applied uniform p0 under zero initial conditions, f ðx Þ ¼ g ðx Þ ¼ 0. Determine the
expressions of the displacement, bending moment, and shear force.
Solution
1. Determine the natural frequencies and mode shapes of the vibration.
They are obtained from Eqs. (8.3.30), (8.3.33):
rffiffiffiffiffiffiffiffiffiffi
np 
EI
wn ¼ n p
2 2
, f ð x Þ ¼ sin x , n ¼ 1, 2, … (1)

4
mL n
L

2. Determine the modal mass, modal force, and initial conditions. They are
obtained from Eqs. (8.3.53), (8.3.56a), (8.3.56b):
Z L
np  mL

Mn ¼ m
sin 2 x dx ¼ (2)
0 L 2
Z L
np  L
Pn ¼ p 0 sin x dx ¼ p0 ½1  ð1Þn  (3)
0 L np
Yn ð0Þ ¼ 0, Y_ n ð0Þ ¼ 0 (4)

3. Determine the time-varying amplitudes Yn ðt Þ. Substituting Eqs. (2)–(4)

into Eq. (8.3.54) and using Eq. (3.4.3) give
4p0 L4
Yn ð t Þ ¼ ð1  cos wn t Þ, n ¼ 1, 3, 5, … (5)
n 4 p4 EI
Hence
X∞
4p0 L4 X
∞
1
np 
u ðx, t Þ ¼ fn ðx ÞYn ðt Þ ¼ 4 4
sin x ð1  cos wn t Þ,
n¼1
p EI n¼1 n L (6)
n ¼ 1, 3, 5, …
Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 295

The bending moment and the shear force are evaluated from the
expressions
M ðx, t Þ ¼ EI u 00 ðx, t Þ ¼ EI f00 ðx ÞY ðt Þ (7)
000 000
Q ðx, t Þ ¼ EI u ðx, t Þ ¼ EI f ðx ÞY ðt Þ (8)

8.4 Problems
Problem 8.1 The television tower of Fig. P8.1 is subjected to seismic ground
motion ug ðt Þ. Derive the equation of motion if the structure is approximated by
a SDOF system. The reaction moment
 of the elastic ground is represented by the
nonlinear expression MR ¼ CR f + 14 f2 , where CR ¼ KI f ; If is the moment of
inertia of the planform of the fundament and K ¼ E=10h the foundation mod-
ulus with E being the modulus of elasticity of the material of the structure. The
cross-section of the flexible column, the planform of the fundament, and the
body B are circular with diameters D, Df ¼ 8D, and DB ¼ 5D, respectively.
The density of the material is r. The fundament and the body B are
assumed rigid.

FIG. P8.1 Television tower in problem P8.1

Problem P8.2 The continuous beam of Fig. P8.2 rests on Winkler’s elastic
foundation with variable modulus k ðx Þ. The beam is axially subjected to the
load P. Determine the value of the load that produces buckling.

FIG. P8.2 Continous beam on nonhomogeneous Winkler’s elastic foundation in problem P8.2
296 PART I Single-degree-of-freedom systems

Problem P8.3 A vehicle of weight W ¼ 30kN is traveling on a simply sup-

ported bridge with velocity v ¼ 80km=h as shown in Fig. P8.3a. The vehicle
is simulated by a single wheel of negligible mass. Determine the dynamic mag-
nification factor for the bridge when the beam is approximated by a generalized
SDOF system. The material of the bridge is prestressed concrete with modulus
of elasticity E ¼ 2:1  107 kN=m2 and specific weight g ¼ 25kN=m3 . The
cross-section of the bridge is shown in Fig. P8.3b.

(a)

(b)
FIG. P8.3 Simple supported bridge in Problem P8.3

Problem P8.4 Use the method of separation of variables to solve the equation
of free flexural vibrations of the:
(a) Fixed-fixed beam.
(b) Fixed-simply supported beam
In both cases, determine the natural mode shapes and the frequency equation.
Compute the first three natural frequencies.
Problem P8.5 Analyze the free flexural vibrations of the two-span continuous
beam of Fig. P8.5 by solving the equation of motion of the beam. Hint: Consider
the continuity condition at support 2, that is, u 0 I ðL, t Þ ¼ u 0 II ð0, t Þ.

FIG. P8.5 Two-span continuous beam in problem P8.5

Generalized single-degree-of-freedom systems—Continuous systems Chapter 8 297

References and further reading

[1] J.T. Katsikadelis, The Boundary Element Method for Plate Analysis, Academic Press,
Elsevier, Oxford, UK, 2014.
[2] D.O. Brush, B.O. Almroth, Buckling of Bars, Plates and Shells, McGraw-Hill, New York,
1975.
[3] S. Timoshenko, D.E. Young, W. Weaver Jr., Vibration Problems in Engineering, fourth ed.,
John Wiley & Sons, New York, 1974.
[4] J.T. Katsikadelis, G.C. Tsiatas, Nonlinear dynamic analysis of beams with variable stiffness,
J. Sound Vib. 270 (2004) 847–863.
[5] E. Zauderer, Partial Differential Equations of Applied Mathematics, second ed., John Wiley &
Sons, Singapore, 1989.
[6] T. Myint-U, L. Debnath, Linear Partial Differential Equations for Scientists and Engineers,
fourth ed., Birkh€auser, Boston, 2007.
[7] E.C. Titchmarsh, Eigenfunction Expansions, Part I (1958) & Part II (1962), Oxford University
Press, Oxford, UK, 1958.
[8] G.L. Cain, G.H. Meyer, Separation of Variables for Partial Differential Equations: An Eigen-
function Approach, in: Studies in Advanced Mathematics, CRC Press, London, 2006.
[9] R.W. Clough, J. Penzien, Dynamics of Structures, McGraw-Hill, New York, 1993.
[10] J.L. Humar, Dynamics of Structures, second ed., A.A. Balkema Publishers, Lisse, NL, 2002.
[11] A.K. Chopra, Dynamics of Structures: Theory and Applications to Earthquake Engineering,
Prentice Hall, Englewood Cliffs, NJ, 1995.
[12] S.S. Rao, Vibration of Continuous Systems, John Wiley & Sons, New Jersey, 2007.
[13] L. Meirovitch, Analytical Methods in Vibration, Macmillan, London, 1967.
Chapter 9

Analysis in the frequency

domain
Chapter outline
9.1 Introduction 299 9.6 Discrete Fourier transform 310
9.2 Complex form of the Fourier 9.7 Application of the discrete
series 301 Fourier transform
9.3 Complex dynamic response to dynamic analysis 313
to periodic load 303 9.8 Fast Fourier transform 314
9.4 Fourier integral 9.8.1 The Sande-Tukey
representation of a algorithm 315
nonperiodic load 304 9.9 Problems 320
9.5 Response to a nonperiodic References and further reading 322
load 307

9.1 Introduction
The methods we discussed in the previous chapters for solving the equation of
motion of a SDOF system were accomplished using time as an independent var-
iable or, said differently, the solution was obtained in the time domain. These
methods are either analytical, which in the general case lead to the evaluation of
the Duhamel integral, or numerical, implemented by the step-by-step integra-
tion methods. Occasionally, simpler or more convenient analytical or numerical
solutions may be possible for certain types of dynamics problems, using integral
transforms such as the Laplace transform or the Fourier transform. The integral
converts the linear differential equation into a linear algebraic equation, from
which the integral transform of the unknown function is obtained. Then the
inverse transform results in the solution in the time domain.
We have already discussed the Laplace transform in Section 3.3, where it
was employed to solve the equation of motion of a SDOF system under an arbi-
trary external excitation. The Laplace transform uses a parameter that does not
have a direct physical meaning. Instead, the parameter in the Fourier transform
has the physical meaning of frequency [1–3]. The method of analyzing dynam-
ical systems using the Fourier transform is known as the analysis in the fre-
quency domain. It plays an important role in studying the dynamic response
of linear systems, that is, systems described by linear differential equations.

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00009-1

© 2020 Elsevier Inc. All rights reserved. 299
300 PART I Single-degree-of-freedom systems

The analysis in the frequency domain is employed in many fields of engineering

science and physics. Therefore, the available related literature is extensive and
reference to it is made with regard to the specific scientific discipline under
treatment.
As we have seen in Section 3.4, the drawback of time-domain analysis is the
evaluation of the Duhamel integral, which can be achieved analytically only for
very simple loading functions. Instead, the Fourier transform converts the con-
volution integral, such as the integral of Duhamel, into the product of the trans-
forms of the functions involved in the convolution. Then the inverse transform
of the product gives the response of the system in the time domain.
The efficiency of the method depends on the possibility of finding the Fou-
rier transform of the functions of the convolution as well as the inverse of their
product. There are tables [4] that give the Fourier transform and its inverse for a
variety of functions. Unfortunately, these functions are rather simple while the
use of tables requires tedious analytical work with the risk of a possible error.
Certainly, the symbolic languages such as MAPLE, MATHEMATICA, and
WOLFRAM ALPHA offer new possibilities for the application of the method.
Nevertheless, the Fourier transform method cannot always meet the practi-
cal needs, especially when the load function is specified by a set of values, as for
example the accelerogram of an earthquake. For this reason, the Fourier trans-
form for a long time has not been equally applied to the dynamic analysis of
structures as it has to other areas of engineering science. The use of numerical
methods of computing the Fourier transform has brought to light this method in
its discrete form, referred to as the discrete Fourier transform. Of course, the
discrete Fourier transform is not simpler or more functioning than the numerical
methods of evaluating the Duhamel integral. Therefore, its use would have been
limited if a specific algorithm for the numerical computation of the Fourier
transform had not been developed, the so-called Fast Fourier transform, which
limits the numerical computations by several orders and makes dynamic anal-
ysis in the frequency domain quite effective.
Besides its effectiveness, the analysis in the frequency domain has other
important advantages. It shows clearly the frequencies of the loading, providing
thus the possibility of detecting those frequencies that unfavorably excite the
structure. It is also suitable for the study of systems involving large or infinite
regions, such as the interaction of dams in large reservoirs with waves, soil
structure interaction in earthquake engineering, and, in general, problems where
the physical characteristics of the system such as stiffness or damping depend
on the frequency of the oscillation. For these reasons, it was considered appro-
priate to include a brief discussion of the analysis in the frequency domain in
this book. Apparently, this would enable the engineers involved in earthquake
analysis to understand the capabilities of the method.
The method could be presented starting from the mathematical development
of the Fourier transform [4], but in order to understand its physical significance,
the complex Fourier series and the response of the SDOF system to periodic
 Analysis in the frequency domain Chapter 9 301

loading are preceded. Then, the method extends to nonperiodic loads by intro-
ducing the Fourier integral and the Fourier transform. The chapter closes with
the presentation of the discrete Fourier transform and the fast Fourier transform.

9.2 Complex form of the Fourier series

In Section 3.6, the dynamic response of a structure under a general periodic load
was obtained in the time domain by expanding the periodic load in Fourier
series, that is, by representing it through an infinite sum of harmonic loads. This
procedure was discussed in detail there. However, the study of the response of a
system in the frequency domain is facilitated by representing the Fourier series
(3.6.2) in exponential form. We recall that
X
1
pðt Þ ¼ a0 +
t + bn sin n w
ðan cos n w
t Þ (9.2.1)
n¼1
Z T =2 Z
1 1 T
a0 ¼ pðt Þdt ¼ pðt Þdt (9.2.2a)
T T =2 T 0

Z T =2
2
an ¼
tdt, n ¼ 1, 2, 3…
pðt Þcos n w (9.2.2b)
T T =2

Z T =2
2
bn ¼
tdt,
pðt Þ sin n w n ¼ 1, 2, 3… (9.2.2c)
T T =2

From Euler’s formula (2.2.8), we have

einw
t + einw
t

t ¼
cos n w (9.2.3a)
2
einw
t  einw
t

t ¼
sin n w (9.2.3b)
2i
Substituting Eq. (9.2.3a) into Eq. (9.2.2b) gives
Z
1 T =2  
an ¼ pðt Þ einw
t + einw
t dt (9.2.4a)
T T =2

Replacing n with n in the previous equation gives

Z
1 T =2  
an ¼ pðt Þ einw
t + einw
t dt ¼ an (9.2.4b)
T T =2

Similarly, substituting Eq. (9.2.3b) into Eq. (9.2.2c) yields

Z
1 T =T ein w
t  einw
t
bn ¼ pðt Þ dt: (9.2.5a)
T T =2 i
302 PART I Single-degree-of-freedom systems

Hence
Z T =2
1 einw
t  einw
t
bn ¼ pðt Þ dt ¼ bn (9.2.5b)
T T =2 i
The Fourier series (9.2.1) by virtue of Eqs. (9.2.3a), (9.2.3b) becomes
X1
einw
t + einw
t X 1
ein w
t  einw
t
pðt Þ ¼ a0 + an + bn
n¼1
2 n¼1
2i
(9.2.6)
1X 1
1X 1
¼ ao + ðan  ibn Þeinw
t + ðan + ibn Þeinw
t
2 n¼1 2 n¼1

Further, using Eqs. (9.2.4a), (9.2.5a), the previous equation becomes

1X 1
1 X1
pðt Þ ¼ ao + ðan  ibn Þeinw
t + ðan  ibn Þein w
t
2 n¼1 2 n¼1
(9.2.7)
X
1

t
¼ cn e in w

n¼1

in which the series coefficients are given as

1
cn ¼ ðan  ibn Þ
2
Z T =2
1    
¼ pðt Þ einw
t + einw
t  ein w
t  einw
t dt (9.2.8a)
2T T =2
Z
1 T =2
¼ pðt Þeinw
t dt
T T =2
and
1
cn ¼ ðan + ibn Þ
Z2 (9.2.8b)
1 T =2
¼ pðt Þeinw
t dt
T T =2
Z
1 T =2
c0 ¼ pðt Þdt ¼ a0 (9.2.8c)
T T =2
Eq. (9.2.7) represents the complex form or exponential form of the Fourier
series.
From Eq. (9.2.7), the reasonable question arises as to how it is possible for a
real function to be expressed as a sum of complex terms. It is easy, however, to
prove that the right side of this equation is a real function as long as we think that
to each term cn einw
t the term cn ein w
t corresponds, whose sum yields a real
function.
 Analysis in the frequency domain Chapter 9 303

9.3 Complex dynamic response to periodic load

In Section 3.6.3, the method of determining the steady-state response of a SDOF
system subjected to a periodic load was presented. In this method, the periodic
load was analyzed into harmonic terms, both sine and cosine, by expanding it in
a Fourier series. After establishing the steady-state response for each term, the
total response was obtained as the sum of all responses. Following a similar pro-
cedure, it is easy to determine the steady-state response to a periodic load when
it is has been expanded in a complex Fourier series. To this end, we must first
determine the steady-state response of the system when subjected to the load
pðt Þ ¼ p0 eiw
t . In this case, the equation of the SDOF system is written
m u€ + cu_ + ku ¼ p0 eiw
t (9.3.1)
The solution will be obtained as a sum of the homogeneous solution uh and a
particular solution up of the nonhomogeneous equation. The homogeneous
solution is given by Eq. (3.2.22), namely
uh ¼ exwt ðA cos wD t + B sin wD t Þ (9.3.2)
The particular solution is sought in the form
up ¼ Ceiw
t (9.3.3)
Substituting the previous expression into Eq. (9.3.1) gives
p0
C¼ (9.3.4)
ðm w
+ icw
2
+ kÞ
which is inserted into Eq. (9.3.3) to yield
p0
up ¼ eiw
t (9.3.5)
m w
2 + icw

+k
Hence, the general solution of Eq. (9.3.1) is
p0
u ðt Þ ¼ exwt ðA cos wD t + B sin wD t Þ + eiw
t (9.3.6)

2
m w + icw
+k
The first term in Eq. (9.3.6) becomes negligible with increasing time and it
represents the transient response of the system. The second term expresses the
steady-state response and can be written as

Þp0 eiw
t
u ðt Þ ¼ H ðw (9.3.7)
where it was set
1

Þ ¼
H ðw
m w
2 + icw
+k
1 (9.3.8)
¼  , b ¼ w

=w
k 1  b 2 + 2ixb
304 PART I Single-degree-of-freedom systems

Þ is called the complex frequency response function, also

The function H ðw
known as the transfer function. In summary, setting p0 ¼ cn we may write the
steady-state response to a periodic load as
X
1
u ðt Þ ¼
Þeinw
t
c n H ðw (9.3.9)
n¼1

where
Z T =2 Z
1 1 T
cn ¼ pðt Þeinw
t dt ¼ pðt Þein w
t dt (9.3.10)
T T =2 T 0

9.4 Fourier integral representation of a nonperiodic load

When the load is nonperiodic, it cannot be represented as a Fourier series. But it
is possible to express the nonperiodic function in the form of an integral over the
interval ð1, 1Þ, which one might regard as a Fourier series whose period is
infinitely large. In Section 3.6.2, we saw that a function periodic in the interval
½0, 1Þ or ð1, 1Þ is expanded in a Fourier series, Eq. (9.2.1). The coefficients
in this equation are written as
Z
1 T =2
a0 ¼ pðτÞdτ (9.4.1a)
T T =2
Z
2 T =2
an ¼ pðτÞcos n w
τdτ, n ¼ 1, 2, 3… (9.4.1b)
T T =2
Z
2 T =2
bn ¼
τdτ, n ¼ 1, 2, 3…
pðτÞ sin n w (9.4.2)
T T =2
The previous equations resulted from Eq. (9.2.2a), Eq. (9.2.2b), Eq. (9.2.2c)
by substituting t with τ in the integrand. Note that it is allowed because τ is a
dummy variable.
Substituting Eqs. (9.4.1a), (9.4.1b), (9.4.2) in Eq. (9.2.1) yields
Z " #
1 Z T =2
1 T =2 2X
pðt Þ ¼ pðτÞdτ +
τ cosn w
pðτÞ cos n w
tdτ
T T =2 T n¼1 T =2
" # (9.4.3)
1 Z T =2
2X
+
τ sin n w
pðτÞ sin n w
tdτ
T n¼1 T =2

or
Z "Z #
1 T =2
2X 1 T=2
pðt Þ ¼ pðτÞdτ +
ðτ  t Þdτ
pðτÞcos n w (9.4.4)
T T =2 T n¼1 T =2
 Analysis in the frequency domain Chapter 9 305

If the function pðt Þ is not periodic, we may set

( Z " #)
1 Z T =2
1 T=2 2X
pðt Þ ¼ lim pðτÞdτ +
ðτ  t Þdτ
pðτÞcos n w
T !1 T T=2 T n¼1 T =2
(9.4.5)
in which t does not change in passing to the limit. It is assumed that the function
pðt Þ satisfies the Dirichlet conditions (see Section 3.6.2). Therefore, it is abso-
lutely integrable over the interval ½T =2, T =2, that is,
Z 
1  T =2  1 Z T =2
 k
j a0 j ¼  pðτÞdτ  jpðτÞjdτ < (9.4.6)
T  T =2  T T =2 T

which vanishes for T ! 1. Hence

( " #)
1 Z T =2
2X
pðt Þ ¼ lim
ðτ  t Þdτ
pðτÞ cos n w (9.4.7)
T !1 T T =2
n¼1

If we set now
2p

¼w
nw
n , D

n + 1  w
w¼w
n ¼ (9.4.8)
T
we may write Eq. (9.4.7) as
X
1
pðt Þ ¼ lim
n ÞD

P ðw w (9.4.9)
T !1
n¼1

n Þ is the value of the function

where P ðw
Z
1 T =2

Þ ¼
P ðw
ðτ  t Þdτ
pðτÞcos w (9.4.10)
p T =2

¼w
for w
is understood as a continuous variable (Fig. 9.4.1)

n , where w

FIG. 9.4.1 Fourier transform of p(t).

306 PART I Single-degree-of-freedom systems

Þ, we readily deduce that the sum

From the graphical representation of P ðw
X
1

n ÞD

P ðw w (9.4.11)
n¼1

Þ. When T ! 1, then D

approaches the area under the curve y ¼ P ðw w ! 0 and
the sum (9.4.11) becomes a definite integral. Consequently, we may write
Eq. (9.4.9) as
Z 1
pðt Þ ¼
Þd w
P ðw
(9.4.12)
0

or by virtue of Eq. (9.4.10)

Z 1 Z 1 
1
pðt Þ ¼
ðτ  t Þdτ d w
pðτÞcos w
(9.4.13)
0 p 1
The integral (9.4.13) is known as the Fourier integral. The Fourier integral con-
verges to the function if it is piecewise continuous in every finite interval and
absolutely integrable over ð1, 1Þ. At points of discontinuity, the Fourier
integral converges to the mean value.
The Fourier integral can be expressed in a complex form using Euler’s for-
mula, Thus, writing
1h i
cos w
ðτ  t Þ ¼ eiw
ðτt Þ + eiw
ðτt Þ (9.4.14)
2
and inserting into Eq. (9.4.13) splits it into two integrals, that is,
Z Z 1  Z Z 1 
1 1
ðτt Þ 1 1
ðτt Þ
i w
pðt Þ ¼ pðτÞe iw

+
dτ d w pðτÞe

dτ d w
2p 0 1 2p 0 1

Changing the integration variable from w
to 
w in the second integral, we

obtain
Z Z 1 
1 1
pðt Þ ¼ pðτÞeiw
τ dτ eiw
t d w

(9.4.15)
2p 1 1
The forgoing relation allows writing
Z 1

Þ ¼
P ðw pðτÞeiw
τ dτ
1

or returning to the variable t from τ

Z 1
P ðw
Þ ¼ pðt Þeiw
t dt (9.4.16)
1

and inserting it into Eq. (9.4.15), gives

Z
1 1
pðt Þ ¼
Þeiw
t d w
P ðw
(9.4.17)
2p 1
 Analysis in the frequency domain Chapter 9 307

The function P ðw
Þ defined by Eq. (9.4.16) is called the (direct) Fourier
transform of pðt Þ while the function pðt Þ resulting from Eq. (9.4.17) is called
the inverse Fourier transform of P ðw
Þ.
In the time domain, a function will be denoted by a small letter while its
Fourier transform is by the same capital letter. The relationship between them
will be symbolized by

Þ
pðt Þ , P ðw (9.4.18)
Usually, we denote the Fourier transform of a function pðt Þ by F ½pðt Þ while
its inverse is by F 1 ½pðt Þ, namely

Þ ¼ F ½pðt Þ
P ðw (9.4.19)
pðt Þ ¼ F 1 ½P ðw

Þ (9.4.20)
The Fourier transform of the derivative of a function pðt Þ is readily estab-
lished by applying integration by parts to Eq. (9.4.16). Generally, for a function
of which the ðn  1Þ order derivatives are continuous and the nth order deriv-
ative is piecewise continuous, it can be shown that
Z 1
pðnÞ ðt Þeiw
t dt ¼ ðiw

Þn P ðw

Þ (9.4.21)
1

Example 9.4.1 The Fourier transform of a function

Find the Fourier transform of the function pðt Þ ¼ ejt j .
Solution Z 1

Þ ¼
P ðw ejt j eiw
t dt
1
Z 0 Z 1
¼ eð1iw
Þt dt + eð1 + iw
Þt dt
1 0
1 1
¼ +
1  iw
1 + iw


2
¼
1+w
2

9.5 Response to a nonperiodic load

In Section 3.7.3, we have seen that the response of a SDOF system to an arbi-
trary load is given by the convolution integral
Z t
u ðt Þ ¼ pðt Þ∗ h ðt Þ ¼ pðτÞh ðt  τÞdτ (9.5.1)
0

where h ðt  τÞ is the response to the unit impulse given by

exwðtτÞ
h ðt  τ Þ ¼ sin wD ðt  τÞ, t > τ (9.5.2)
mwD
308 PART I Single-degree-of-freedom systems

for x 6¼ 0 and
1
h ðt  τ Þ ¼ sin wðt  τÞ, t > τ (9.5.3)
mw
for x ¼ 0.
The integral (9.5.1) can be also written as
Z 1
u ðt Þ ¼ pðt Þ∗ h ðt Þ ¼ pðτÞh ðt  τÞdτ (9.5.4)
1

because pðτÞ ¼ 0, when τ < 0 and h ðt  τÞ ¼ 0, when τ > t.

The establishment of the response to an arbitrary load in the frequency
domain is achieved by taking the Fourier transform of the convolution
(9.5.4). Thus, we have
Z 1 Z 1 

Þ ¼
U ðw pðt Þh ðt  τÞdτ eiw
t dt (9.5.5)
1 1

If we set t  τ ¼ s, then t ¼ τ + s and Eq. (9.5.5) gives

Z 1 Z 1 

Þ ¼
U ðw h ðs Þeiw
s ds pðτÞeiw
τ dτ
Z1
1
1

¼
ÞpðτÞeiw
τ dτ
H ðw
1 (9.5.6)
Z 1
¼ H ðw
Þ pðτÞeiw
τ dτ
1

ÞH ðw
¼ P ðw
Þ
From Eq. (9.5.6) we deduce that the Fourier transform of the response to an
arbitrary load, namely of the convolution integral, is equal to the product of the
Fourier transforms of the functions in the convolution. Hence, we may write
symbolically

Þ
pðt Þ , P ðw (9.5.7)

Þ
h ðt Þ , H ðw (9.5.8)

Þ
u ðt Þ ¼ pðt Þ ∗h ðt Þ , U ðw (9.5.9)
The Fourier transform of the function u ðt  t0 Þ is obtained as
Z 1
F½uðt  t0  ¼ u ðt  t0 Þeiw
t dt
1
Z 1
¼ u ðxÞeiw
ðx + t0 Þ dt
1 (9.5.10)
Z 1
¼ eiw
t0 u ðx Þeiw
x dt
1
¼ eiw
t0 F ½u ðt Þ
 Analysis in the frequency domain Chapter 9 309

The forgoing equation represents the shifting property of the Fourier

transform.
The Fourier transform H ðw
Þ of the response h ðt  τÞ to the unit impulse can
be obtained as follows.
The equation of motion of the response h ðt  τÞ to the unit impulse results by
setting pðt Þ ¼ d ðt  τÞ in the equation of the SDOF system. That is,
m h€ + ch_ + kh ¼ dðt  τÞ (9.5.11)
By virtue of Eqs. (9.5.10), (9.4.21), (3.7.6b), Eq. (9.5.11) gives
 
m w
2 + ci w

+ k H ðw
Þ ¼ 1 (9.5.12)
from which we obtain
1

Þ ¼
H ðw (9.5.13)

2
ðm w
+ kÞ
+ ci w
or
1

Þ ¼  2
H ðw  (9.5.14)
k b  1 + 2ixb
Eq. (9.5.14) is identical to Eq. (9.3.8), that is, the complex response function
is the Fourier transform of the response of the SDOF system to the unit impulse.
Example 9.5.1 The Fourier transform method for the dynamic response of a
SDOF system
Determine the response of a SDOF (k,w,x) system subjected to the load
pðt Þ ¼ p0 , 0 < t using the frequency domain analysis.
Solution
The Fourier transform of the load is
Z Z
1 1
t 1 1 p0

Þ ¼
P ðw p0 e i w
dt ¼ p0 eiw
t dt ¼ (1)
2p 1 2p 0

2pi w
The Fourier transform of the response to the unit impulse is
1

Þ ¼  2
H ðw , b ¼ w

=w (2)
k b  1 + 2ixb
Consequently, by virtue of Eq. (9.5.9), we obtain
p0
U ðw
Þ ¼ P ðw

ÞH ðw

Þ ¼  2  (3)

k b  1 + 2ixb
2pi w
The response in the time domain results as the inverse Fourier transform of

Þ, that is,
U ðw
Z 1
p0 eiw
t
u ðt Þ ¼  2  dw

(4)
2pik 1 w

b + 2ixb  1
310 PART I Single-degree-of-freedom systems

which is further written as

Z 1
p0 eiwbt
u ðt Þ ¼ db (5)
2pikw 1 b ðb  b 1 Þðb  b 2 Þ

where b 1 , b2 are the roots of the polynomial b2 + 2ixb  1, namely

qffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffi
b 1 ¼ ix + 1  x2 , b 2 ¼ ix  1  x 2 (6)

The integral (5) is evaluated using the method of closed line integrals in the
complex domain of b. This method yields
u ðt Þ ¼ 0, t  0
" !#
p0 x
u ðt Þ ¼ 1e xwt
cos wD t + pffiffiffiffiffiffiffiffiffiffiffiffi sin wD t , 0 < t
k 1  x2
From the last example, we observe that the dynamic analysis in the fre-
quency domain requires the evaluation of complicated integrals, even for the
simplest load cases. This problem is circumvented by applying numerical
methods for the evaluation of the Fourier transform, such as the discrete Fourier
transform (DFT), and the fast Fourier transform (FFT). These methods are dis-
cussed in the next sections.

9.6 Discrete Fourier transform

In the previous paragraph, we developed the method of dynamic analysis in the
frequency domain. The steps we followed are summarized as follows:
1. We take the Fourier transforms of the loading function, P ðw
Þ ¼ F ½pðt Þ and
the response function to the unit pulse H ðw
Þ ¼ F ½h ðt Þ.
2. We evaluate the product of the two transforms U ðw
Þ ¼ P ðw

ÞH ðw
Þ obtained
in the previous step.
3. We evaluate the inverse Fourier transform, which gives the response of the
system u ðt Þ ¼ F 1 ½U ðw

Þ in the time domain.
The success of the method is based on the capability of finding the Fourier trans-
form of a given function and its inverse. However, this is not always easy. With
the exception of simple functions whose Fourier transforms are obtained from
tables [4], their establishment requires the computation of complicated inte-
grals, a task that is difficult, tedious, or even impossible.
Very often, the loading function is specified by a set of values at distinct time
instants, for example, the accelerogram of an earthquake. As we mentioned, the
previous difficulties can be overcome by developing methods for the numerical
computation of the Fourier transform (direct and inverse). The discrete
Fourier transform (DFT) is among them. Below, we present the numerical
 Analysis in the frequency domain Chapter 9 311

implementation of the DFT and its application for dynamic analysis in the fre-
quency domain.
We consider the function pðt Þ of Fig. 9.6.1, which is defined in the interval
0  t  ttot .

tot

FIG. 9.6.1 Nonperiodic function expanded perodically.

Then we assume that the function is periodically extended from 1 to +1

with a period T0  ttot to include a number of zero values of pðt Þ in T0 .
If this function is expanded in Fourier series, we obtain by virtue of
Eqs. (9.2.7), (9.2.8a), (9.2.8b), (9.2.8c)
X
1
pðt Þ ¼ cn ein w
t (9.6.1)
n¼1
Z T0 =2
1
cn ¼ pðt Þeinw
t dt
T0 T0 =2
Z T0 , n ¼ 0,  1,  2, … (9.6.2)
1
t
in w
¼ pðt Þe dt
T0 0

The integral (9.6.2) can be evaluated numerically. For this purpose, we

divide the interval T0 into N equal subintervals of length Dt ¼ T0 =N and
approximate the integral with a sum of N rectangles.
If we set
s ðt Þ ¼ pðt Þeinw
t (9.6.3)
then we have

1 NX1
cn ¼ s ðtk ÞDt (9.6.4)
T0 k¼0

where s ðtk Þ is the value of the integrand at instants tk ¼ kDt. That is,
s ðtk Þ ¼ pðtk Þeinw
kDt (9.6.5)
312 PART I Single-degree-of-freedom systems

or taking into account that

T0 ¼ N Dt,
¼ 2p=T0 ¼ 2p=N Dt
w (9.6.6)
we obtain
s ðtk Þ ¼ pðtk Þe2pikn=N (9.6.7)
and Eq. (9.6.4) becomes

1 NX
1
cn ¼ pðtk Þe2pikn=N (9.6.8)
N k¼0

As was shown in Section 3.6, the Fourier series can be approximated by a

sum of finite terms, Eq. (3.6.9), with very good accuracy even in cases of a dis-
continuous function, as demonstrated by Example 3.6.2. Therefore, on the basis
of Eq. (9.6.6) we can write Eq. (9.6.1) as
X
M
pðtk Þ ’ cn e2pikn=N (9.6.9)
n¼M

In the foregoing equation, we split the sum into two sums

X
1 X
M
pðtk Þ ¼ cn e2pikn=N + cn e2pikn=N (9.6.10)
n¼M n¼0

then taking into account that the function pðt Þ is periodic with a period
T0 ¼ N Dt, we can write the first sum as
X
1 X
1
cn e2pikn=N ¼ cn + N e2pik ðn + N Þ=N (9.6.11)
n¼M n¼M

Subsequently, setting M ¼ ðN  1Þ=2, n~ ¼ n + N and taking into account

the periodicity of pðt Þ, we can write the right side of the previous equation as
X
2M X
M
cn~ e2pik n~ =N ¼ cn e2pikn=N (9.6.12)
n~ ¼M + 1 n¼1

Substituting Eq. (9.6.11) into Eq. (9.6.10), and taking into account
Eq. (9.6.12), we obtain
X
2M
pðtk Þ ¼ cn e2pikn=N (9.6.13)
n¼0

or because 2M ¼ N  1, it follows:
X
N 1
pðtk Þ ¼ cn e2pikn=N (9.6.14)
n¼0
 Analysis in the frequency domain Chapter 9 313

In order to match the expressions between the continuous Fourier transform

and the discrete Fourier transform, we set cn ¼ Pn and Eqs. (9.6.8), (9.6.14) are
written

1 NX
1
Pn ¼ pk e2pikn=N , k ¼ 0, 1, 2, …,N  1 (9.6.15)
N k¼0

X
N 1
pk ¼ Pn e2pikn=N , k ¼ 0, 1, 2, …,N  1 (9.6.16)
n¼0

The foregoing relations express the discrete Fourier transform (DFT), direct
and inverse, respectively.
The DFT approximates numerically the continuous Fourier transform,
defined by Eqs. (9.4.16), (9.4.17). The accuracy of the DFT is very good if
Dt is selected small. However, there is a fundamental difference between the
continuous Fourier transform and the discrete Fourier transform. The first pro-
vides the exact transform of the actual function while the second assumes a peri-
odic extension of the function. This means that the discrete transform is
applicable when the interval T0 is finite. It holds only within the period. Outside
it, the two transforms are completely different unless the function happens to be
periodic.

9.7 Application of the discrete Fourier transform

to dynamic analysis
As we mentioned in Section 9.5, the response in the time domain of a SDOF
system subjected to an arbitrary load is given by the convolution integral of
the load function pðt Þ and the response function to the unit impulse h ðt Þ.
Namely,
u ðt Þ ¼ pðt Þ ∗h ðt Þ (9.7.1)
Eq. (9.7.1) can be used as the basis to obtain the response in the frequency
domain by the DFT adhering to the following steps:
1. We compute the DFT of the function pðt Þ, 0  t  ttot , which is assumed
extended periodically with a period T0  ttot so that pðtN Þ ¼ 0.

1 NX
1
Pn ¼ pðtk Þe2pikn=N (9.7.2)
N k¼0

2. We compute the DFT of the response function h ðt Þ. This requires the con-
finement of h ðt Þ in an interval equal or smaller than T0 .

1 NX
1
Hn ¼ h ðtk Þe2pikn=N (9.7.3)
N k¼0
314 PART I Single-degree-of-freedom systems

3. We compute the DFT of the product

U n ¼ Pn H n (9.7.4)
4. We compute the inverse DFT of the product
X
N 1
u ðtk Þ ¼ Un e2pikn=N , k ¼ 0, 1, 2, …,N  1 (9.7.5)
n¼0

which yields the response in the time domain.

Details about the application of the DFT method to dynamic analysis can be
found in the relevant literature, for example, Refs. [3, 5].

9.8 Fast Fourier transform

The discrete convolution of two functions pðt Þ, h ðt Þ is defined as
X
N 1
u ðtk Þ ¼ pðtm Þh ðtkm ÞDt, tk ¼ kDt, tkm ¼ ðk  m ÞDt (9.8.1)
m¼0

where both functions pðt Þ and h ðt Þ are periodic, that is,

pðtm + rN Þ ¼ pðtm Þ (9.8.2a)
h ðtm + rN Þ ¼ h ðtm Þ, r ¼ 0,  1,  2, … (9.8.2b)
Eq. (9.8.1) gives the dynamic response u ðtk Þ directly in the time domain at
time tk if pðt Þ represents the load and h ðt Þ the response to the unit
impulsive load.
We observe that the time-domain analysis based on Eq. (9.8.1) requires N 2
multiplications between real numbers. Instead, for the dynamic analysis in the
frequency domain, the required multiplications are N 2 between real and com-
plex numbers as dictated by each of Eqs. (9.7.2), (9.7.3), N of complex numbers
as dictated by Eq. (9.7.4), and N 2 multiplications of complex numbers as dic-
tated by Eq. (9.7.5).
It is clear that the dynamic analysis in the frequency domain using the DFT
requires significantly more computations, a fact that does not encourage its use.
However, it is possible to reduce the number of computations drastically by tak-
ing advantage of the harmonic properties of the involved functions. The algo-
rithm that reduces the calculations is referred to as the fast Fourier transform
(FFT). The appearance of the FFT gave an impetus to dynamic analysis in the
frequency domain. The number of operations (multiplications) required by the
FFT decreases from N 2 to N log 2 N . Fig. 9.8.1 illustrates the advantage of FFT
 Analysis in the frequency domain Chapter 9 315

over DFT. The first FFT algorithm was developed by Gauss in the early 19th
century [6]. Also, the contributions of Runge, Danielson, Lanczos, and others
in the early 20th century were significant. However, its use did not attract the
interest of many researchers because the calculations had to be performed by
hand. It was only with the advent of computers that the FFT came to the fore-
ground. In 1965, J. W. Cooley and J. W. Tukey published an algorithm for cal-
culating the FFT [7, 8]. This algorithm is similar to that of Gauss and others and
is named after them as the Cooley-Tukey algorithm. Today, there are several
algorithms for FFT based on this algorithm. Below we present the Sande-Tukey
algorithm that is a variation of the Cooley-Tukey.

FIG. 9.8.1 Number of operations in DFT and FFT.

9.8.1 The Sande-Tukey algorithm

In this algorithm, we assume that N is a power of 2, that is,
N ¼ 2M (9.8.3)
where M is an integer. This constraint is introduced to simplify the algorithm.
In general, the DFT can be represented as
X
N 1
Pn ¼ ~k e2pikn=N , p
p ~k ¼ pk =N , n ¼ 0, 1, 2, …,N  1 (9.8.4)
k¼0
316 PART I Single-degree-of-freedom systems

Eq. (9.8.4) can also be written in the form

X
N 1
Pn ¼ ~k W nk
p (9.8.5)
k¼0

where W is the complex weight function defined as

W ¼ e2pi=N (9.8.6)

We divide now the interval into two subintervals, and we express

Eq. (9.8.4) as

ðNX
=2Þ1 X
N 1
Pn ¼ pk e2pikn=N +
~ ~k e2pikn=N , n ¼ 0, 1, 2, …,N  1 (9.8.7)
p
k¼0 k¼N =2

Then we introduce a new variable m ¼ k  N =2, so that the total number

of the indices is the same in both sums. Thus, we may write the foregoing
equation as

ðNX
=2Þ1 ðNX
=2Þ1
Pn ¼ ~k e2pikn=N +
p ~m + N =2 e2pinðm + N =2Þ=N
p (9.8.8)
k¼0 m¼0

or

=2Þ1

ðNX 
Pn ¼ ~k + epin p
p ~k + N =2 e2pikn=N (9.8.9)
k¼0

n
We observe that eipn ¼ ðeip Þ ¼ ð1Þn . Consequently, for points with
even n this factor is equal to one while with odd n it is equal to 1. The next
step is to separate the terms of Eq. (9.8.9) into two sums corresponding to the
even and odd values of n. Hence, for even values, we have

=2Þ1

ðNX 
P2n ¼ p ~k + N =2 e2pik ð2nÞ=N
~k + p
k¼0
(9.8.10)
=2Þ1

ðNX 
¼ p ~k + N =2 e2pikn=ðN =2Þ
~k + p
k¼0
 Analysis in the frequency domain Chapter 9 317

while for odd values

=2Þ1

ðNX 
P2n + 1 ¼ p ~k + N =2 e2pik ð2n + 1Þ=N
~k  p
k¼0
(9.8.11)
=2Þ1

ðNX 
¼ p ~k + N =2 e2pik=N e2pikn=ðN =2Þ
~k  p
k¼0

where n ¼ 0, 1, 2, …, ðN =2Þ  1.
By virtue of Eq. (9.8.6), Eqs. (9.8.10), (9.8.11) may be written

=21

NX 
P2n ¼ p ~k + N =2 W 2kn
~k + p (9.8.12)
k¼0

=21

NX 
P2n + 1 ¼ p pk + N =2 W k W 2kn
~k  ~ (9.8.13)
k¼0

We can now make an important observation, which is the key to the method.
The even and the odd expressions can be considered as two DFTs of N =2 points
each. We further set
~k + p
gk ¼ p ~k + N =2 (9.8.14)


hk ¼ p~k  p
~k + N =2 W k , k ¼ 0, 1, 2, …, ðN =2Þ  1 (9.8.15)

Hence
)
P2n ¼ Gn
, n ¼ 0, 1, 2, …, ðN =2Þ  1 (9.8.16)
P2n + 1 ¼ Hn

In other words, a computation at N points has been replaced by two com-

putations at N =2 points each. Because each of the latter computations requires
ðN =2Þ2 complex calculations against N 2 required by the initial computation,
the number of multiplications is reduced to 2ðN =2Þ2 ¼ N 2 =2. It is obvious,
therefore, that the process of bisecting the transforms can be repeated in a sec-
ond stage. We can thus compute 4 DFT with N =4 points each using the first and
last point of each sequence of N =4 points. This technique continues and ends in
318 PART I Single-degree-of-freedom systems

DFT with two points (see Fig. 9.8.2). The total number of complex multiplica-
tion for a given DFT reduces to N log 2 N . The importance of FFT over DFT is
demonstrated in Fig. 9.8.1.

FIG. 9.8.2 Flow chart of the first substitution stage of the DFF with N points by two DFF of N =2
points each when N ¼ 8.

On the basis of the previous analysis, a computer program has been

written in MATLAB that evaluates the dynamic response of a SDOF. The pro-
gram uses the MATLAB functions fft.m and ifft.m. This program, which is
given the name FFT_dymamic.m, is available on this book’s companion
website.
Example 9.8.1 The Fourier transform for the dynamic response of a SDOF
system
Determine the response of SDOF system subjected to the Athens earthquake
in 1999 using the FFT method. Data: m ¼ 1, x ¼ 0:1, k ¼ 25, and pðt Þ ¼ u€g ðt Þ
where u€g ðt Þ is the accelerogram of the earthquake.
Solution
The solution is obtained using the program FFT_dymamic.m. The response
of the system is shown in Fig. E9.1 as compared with the numerical solution
in the time domain using the program aem.lin.m given in Section 4.4. It
is obvious that the computed responses by both methods are graphically
identical.
 Analysis in the frequency domain Chapter 9 319

u(t) (m)

t
u,t(t) (m/s)

t
u,tt(t) (m/s2)

t
FIG. E9.1 Dynamic response of the SDOF system using FFT in Example 9.8.1.
320 PART I Single-degree-of-freedom systems

Example 9.8.2 Amplitude spectrum of an accelerogram

Compute and plot the amplitude spectrum of the 1999 Athens earthquake
using the FFT method.
Solution
The amplitude spectrum is established by computing the magnitude of the
values of the Fourier transform of the accelerogram and it is plotted versus
the frequency (Hz) in the interval of the duration of the excitation. Fig. E9.2
shows the amplitude spectrum of the Athens 1999 earthquake as computed
using the program FFT_Ampl_Spectrum.m, which utilizes the function fft.m
of MATLAB. The program FFT_Ampl_Spectrum.m is available on this
book’s companion website.

FIG. E9.2 Amplitude spectrum of the 1999 Athens earthquake in Example 9.8.2.

9.9 Problems
Problem P9.1 Write a computer program for the evaluation of DFT and com-
pute the DFT of the function shown in Fig. P9.1. Then compute the inverse DFT
and compare the results with the exact function.

FIG. P9.1 Function p(t) in problem P9.1

 Analysis in the frequency domain Chapter 9 321

Problem P9.2 The SDOF system m, x,k is subjected to a ground motion ug ðt Þ.

Study its dynamic response using:
(a) The exact analytic method in the time domain.
(b) The numerical computation of the convolution integral in the time domain.
(c) The DFT in the frequency domain.
h i
Data: x ¼ 0:1, u€gi ¼ 20 1 + ð1Þi ði + 5Þ=ði + 1Þ , ti ¼ 0:05i, i ¼ 1, 2, …,20,
u ð0Þ ¼ u_ ð0Þ ¼ 0, and w ¼ 4ps1 .
Problem P9.3 Give the graph of the Fourier transform Aðw
Þ versus the fre-
quency (Hz) of the accelerogram of the Mexico City earthquake using the
FFT. The file Mexico_Earthquake.txt including the values of the accelerogram
is available on this book’s companion website.
Problem P9.4 The water tower of Fig. P9.4a is subjected to the load
pðt Þ ¼ 150sin pt=t1 , t1 ¼ 0:5 (Fig. P9.4b), which acts at the center of mass of
the tank. Approximating the structure by a generalized SDOF system, study
the dynamic response in the frequency domain for ttot ¼ 1s. The material of
the structure is reinforced concrete with a specific weight g ¼ 25kN=m3 . The col-
umn and the tank have a circular cross-section. The interior of the tank is divided
into small compartments to eliminate the liquid-structure interaction phenomena.
Use the first mode shape of the cantilever with a constant cross-section as a shape
function, that is, ψðx Þ ¼ ½ coshlx  cos lx  0:7341ð sinh lx  sinhlx Þ=3,
l ¼ 1:8751=L, where L is the length of the cantilever.

(b)

(a)
FIG. P9.4 Water tower (a) and load (b) in problem P9.4
322 PART I Single-degree-of-freedom systems

References and further reading

[1] R. Bracewell, The Fourier Transform and Its Applications, McGraw-Hill, New York, 1978.
[2] T. Myint-U, L. Debnath, Linear Partial Differential Equations for Scientists and Engineers,
fourth ed., Birkh€auser, Boston, 2007.
[3] J.L. Humar, Dynamics of Structures, second ed., A.A. Balkema Publishers, Lisse, NL, 2002.
[4] F. Oberhettinger, Tabellen zur Fourier Transformation, 1957 Sringer-Verlag, Berlin, 1957.
[5] R.R. Craig Jr., A.J. Kurdila, Fundamentals of Structural Dynamics, second ed., John Wiley,
New Jersey, 2006.
[6] M.T. Heideman, D.H. Johnson, C.S. Burrus, Gauss and the history of the fast Fourier transform,
IEEE ASSP Mag. 1 (4) (1984) 14–21.
[7] J.W. Cooley, J.W. Tukey, An algorithm for the machine calculation of complex Fourier series,
Math. Comput. 19 (1965) 297–301.
[8] J.F. Hall, A FFT algorithm for structural dynamics, Earthq. Eng. Struct. Dyn. 10 (1982)
797–811.
[9] W.H. Press, B.P. Flannery, S.A. Teukolsky, W.T. Vetterlin, Numerical Recipes in FORTRAN,
second ed., Cambridge University Press, New York, 1992.
Chapter 10

Multi-degree-of-freedom
systems: Models and equations
of motion
Chapter outline
10.1 Introduction 325 10.5 Systems with distributed
10.2 Systems with localized mass mass and distributed stiffness 341
and localized stiffness 327 10.5.2 The method of global
10.3 Systems with distributed shape functions 342
mass and localized stiffness 328 10.6 Mixed systems 347
10.4 Systems with localized mass 10.7 Transformations of the
and distributed stiffness 330 equations of motion 351
10.4.1 The method of 10.8 Problems 354
influence coefficients 334 References and further reading 358

10.1 Introduction
So far, we have studied the dynamic response of SDOF systems. We have also
shown how a system with infinite degrees of freedom can be approximated by
a SDOF system. The trustworthiness of this approximation depends on various
issues. If the actual distribution of the physical properties of the structure, that
is, mass and stiffness, and that of the external force produce deformation during
the motion similar to the assumed, then the approximation with a SDOF system
gives acceptable results. A key shortcoming of this approximation is the diffi-
culty in determining the degree of reliability of the obtained results. In general,
however, the study of the dynamic response of structures requires their model-
ing with MDOF systems, especially when the deformation shapes are compli-
cated. In engineering structures, the mass, though distributed to all its members,
is usually lumped at certain points or regions. For example, in buildings the
mass is lumped at the levels of the stories or in a water tower at the top of
the column that supports the tank. This fact allows describing the motion of
a structure with that of a MDOF system with deformation parameters the dis-
placements of the points where the dynamic characteristics (mass and moment
of inertia) are concentrated.
Fig. 10.1.1a shows a three-story frame whose horizontal beams are virtually
rigid. In this structure, the mass of the columns is negligible compared to that of

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00010-8

© 2020 Elsevier Inc. All rights reserved. 325
326 PART II Multi-degree-of-freedom systems

the beams. Hence, the mass is lumped at the level of the beams and the structure
can be approximated by the model of Fig. 10.1.1b. The deformation shape dur-
ing motion is shown in Fig. 10.1.1c. Obviously, its motion can be determined by
establishing the displacements u1 ðt Þ, u2 ðt Þ, and u3 ðt Þ, that is, the system has
three degrees of motion.

(a) (b) (c)

FIG. 10.1.1 Three-story frame (a), dynamic model (b) and deformation shape (c).

(a) (b)
FIG. 10.1.2 Water tower (a) and its deformed dynamic model (b).

Fig. 10.1.2a shows a water tower. With the assumption that the mass of the
column is negligible compared with that of the tank, we can model the water
tower with the system of Fig. 10.1.2b. That is, the water tower is simulated
by a flexible column, which is fixed at the ground and has a mass m at its
top with a moment of inertia Io . The mass m can move horizontally and rotate
within the plane. The determination of the motion requires the establishment of
the displacement u ðt Þ and the rotation fðt Þ of the top cross-section of the col-
umn, that is, the system has two degrees of freedom.
The MDOF systems can be categorized as follows:
1. Systems with localized mass and localized stiffness.
2. Systems with distributed mass and localized stiffness.
 MDOF systems: Models and equations of motion Chapter 10 327

3. Systems with localized mass and distributed stiffness.

4. Systems with distributed mass and distributed stiffness.
5. Mixed systems resulting from a combination of the systems of previous
categories.
In what it follows, the term localized mass denotes a lumped mass that may have
a moment of inertia while the term concentrated mass denotes a mass that is
concentrated at a point and has a zero moment of inertia. The above classifica-
tion is useful as it dictates a convenient method to formulate the equations of
motion of the MDOF system. Its effectiveness is explained by illustrative exam-
ples. The pertinent bibliography with recommended references for further study
is also included. The chapter is enriched with problems to be solved.

10.2 Systems with localized mass and localized stiffness

The system of Fig. 10.2.1a consists of two rigid bodies with masses m1 and m2 ,
whose horizontal motion is elastically constrained by two massless springs k1 and
k2 . The system has localized masses and stiffnesses. The degrees of freedom are
two, the displacements u1 ðt Þ and u2 ðt Þ. The equations of motion can be derived
using the method of the equilibrium of the forces (D’Alembert’s principle) pre-
sented in Section 1.5. The forces applied to the masses are shown in Fig. 10.2.1b.
The equilibrium of mass m1 gives
fI 1 + fD1 + fS1 ¼ p1 (10.2.1)
or
m1 u€ 1 + c1 ðu_ 1  u_ 2 Þ + k1 ðu1  u2 Þ ¼ p1 ðt Þ (10.2.2)
Similarly, the equilibrium of mass m2 gives
fI 2 + fD2  fD1 + fS2  fS1 ¼ p2 (10.2.3)
or
m2 u€2  c1 u_ 1 + ðc2 + c1 Þu_ 2  k1 u1 + ðk2 + k1 Þu2 ¼ p2 ðt Þ (10.2.4)
Eqs. (10.2.2), (10.2.4) are written in matrix form

 
 
 
m1 0 u€1 c1 c1 u_ 1 k1 k1 u1
+ +
0 m2 u€2 c1 c2 + c1 u_ 2 k1 k2 + k1 u2
 
p1
¼ (10.2.5)
p2
or
u + Cu_ + Ku ¼pðt Þ
M€ (10.2.6)
where




m1 0 c1 c1 k1 k1
M¼ , C¼ , K¼ (10.2.7)
0 m2 c1 c1 + c2 k1 k1 + k2
328 PART II Multi-degree-of-freedom systems

are the mass, the damping, and stiffness matrices of the system, respec-
tively, and
   
u1 p1 ð t Þ
u¼ , pðt Þ ¼ (10.2.8)
u2 p2 ðt Þ

the displacement and load vectors.

(a)

(b)
FIG. 10.2.1 System with localized mass and localized stiffness (a). Forces applied to the masses
m1 and m2 (b).

10.3 Systems with distributed mass and localized stiffness

The system of Fig. 10.3.1a consists of the rigid bar AC , whose mass is uni-
formly distributed along its length with line mass density m, and the circular
rigid body of total mass m ¼ mL,  also uniformly distributed. The motion of
the system is elastically constrained by the two massless springs, k1 ¼ k and
k2 ¼ 2k. A damper with damping coefficient c acts at point C . The system is
loaded by the moment M ðt Þ and the force pðt Þ. This structure, which repre-
sents a system with distributed mass and localized stiffness, has two degrees
of freedom. We choose the vertical displacement u ðt Þ of the end C of the bar
and the rotation fðt Þ of the circular body to determine the motion of the sys-
tem. Referring to Fig. 10.3.1b, the equilibrium of the rigid bar with respect to
point A gives:
 MDOF systems: Models and equations of motion Chapter 10 329

(a)

(b)
FIG. 10.3.1 System with distributed mass and localized stiffness (a). Forces acting on the two
bodies (b).

MIA + 2Lf D + Lf S1 + 2Lf S2 ¼ pðt ÞL (10.3.1)

Obviously, we have
u€ ð2LÞ3 u€ 4
MIA ¼ IA 
¼m ¼ mL _ fS1 ¼ 0:5ku
 2 u€ , fD ¼ cu,
2L 3 2L 3
fS2 ¼ 2k ðu  0:5LfÞ
which are substituted into Eq. (10.3.1) to yield
2 9 1
 u€ + cu_ + ku  kLf ¼ pðt Þ
mL (10.3.2)
3 4 2
Similarly, the equilibrium of the circular body with respect to point O gives
MIO  0:5Lf S2 ¼ M ðt Þ (10.3.3)
Taking into account that

mL 1  3€
mL
MIO ¼ IO f€ ¼ p ð 0:5L Þ 4€
f ¼ f, fS2 ¼ 2k ðu  0:5LfÞ
pð0:5LÞ2 2 8
and substituting into Eq. (10.3.3), we obtain
 3€
mL
f  kLu + 0:5L2 kf ¼ M ðt Þ (10.3.4)
8
330 PART II Multi-degree-of-freedom systems

Eqs. (10.3.2), (10.3.4) are the equations of motion, which in matrix form
become
u + Cu_ + Ku ¼pðt Þ
M€ (10.3.5)
where
2 3
2
 " #
 0
mL 9
63 7 c 0 k kL
M¼4 , C¼ , K¼
 35 4 (10.3.6)
mL 0 0
0 kL 0:5L2 k
8
represent the mass, damping, and stiffness matrices of the system and
8 9
  < 1 pðt Þ =
u
u¼ , pðt Þ ¼ 2 (10.3.7)
f : ;
M ðt Þ
the displacement and load vectors.

10.4 Systems with localized mass and distributed stiffness

The system of Fig. 10.4.1 a represents an example of this case. It consists of the
flexible column AO, whose mass is assumed negligible, and the plane square
body of side length a and mass m uniformly distributed.

(a) (b)
FIG. 10.4.1 System with localized mass and distributed stiffness (a). Deformed dynamic model (b).

The equations of motion will results from the motion of the plane rigid body in
its plane. We examine the motion with respect to the point O, which does not coin-
cide with the center of mass of the body. In general, the system has three degrees of
freedom, namely the horizontal displacement, the vertical displacement, and the
rotation about O. Because the column is flexible, the horizontal displacement and
the rotation are due to the bending deformation. The vertical displacement is
caused by (i) the axial deformation of the column, which is very small and thus
 MDOF systems: Models and equations of motion Chapter 10 331

neglected, and (ii) the shortening of the chord of the deflection curve, which is
also neglected in the linear theory. Therefore, the parameters of the motion are
the horizontal displacement u ðt Þ and the rotation fðt Þ, Fig. 10.4.1b. In the follow-
ing, the equations of motion are derived using two different methods.
1. The method of the Lagrange equations
(i) Elastic energy: This is due to the bending deformation of the column. The
deflection curve can be set in the form

uðx, t Þ ¼ 1 ðx Þu ðt Þ + 2 ðx Þfðt Þ (10.4.1)

where 1 ðx Þ and 2 ðx Þ are the elastic curves of the column for u ðt Þ ¼ 1,

fðt Þ ¼ 0 and fðt Þ ¼ 1, u ðt Þ ¼ 0, respectively. They can be obtained from
the solution of the following two boundary value problems

d4 1 0 0
¼ 0, 1 ð0Þ ¼ 0, 1 ð0Þ ¼ 0, 1 ðh Þ ¼ 1, 1 ðh Þ ¼ 0 (10.4.2a)
dx 4
and

d4 2 0 0
¼ 0, 2 ð0Þ ¼ 0, 2 ð0Þ ¼ 0, 2 ðh Þ ¼ 0, 2 ðh Þ ¼ 1 (10.4.2b)
dx 4
Integrating of the differential equation (10.4.2a) gives

1 3 1 2
1 ðx Þ ¼ c1 x + c2 x + c3 x + c4 (10.4.3)
6 2
After evaluation of the arbitrary constants by applying the boundary
conditions, we obtain
 x 2  x 3
1 ðx Þ ¼ 3 +2 ¼ 3x2 + 2x3 , x ¼ x=h (10.4.4)
h h
Similarly, we obtain
 x 2  x 
2 ð x Þ ¼ h  1 ¼ hx 2 ðx  1Þ, x ¼ x=h (10.4.5)
h h
The elastic energy is given (see Eq. 8.1.11)
Z
1 h
EI ½u00 ðx, t Þ dx
2
U¼
2 0
which by virtue of Eq. (10.4.1) becomes
Z
1 h  00 2
U¼ EI 1 ðx Þu ðt Þ + 002 ðx Þfðt Þ dx (10.4.6)
2 0
Differentiating Eq. (10.4.6) with respect to u ðt Þ and fðt Þ gives
332 PART II Multi-degree-of-freedom systems

Z h
∂U 00 00
00
¼ EI 1 ðx Þu ðt Þ + 2 ðx Þfðt Þ 1 ðx Þdx
∂u 0
Z h Z
00
2 h
00 00
¼ EIu ðt Þ (10.4.7a)
1 ðx Þ dx + EI fðt Þ 2 ðx Þ 1 ðx Þdx
0 0
12EI 6EI
¼ u+ 2 f
h3 h
Z h
∂U  00 00
00
¼ EI 1 ðx Þu ðt Þ + 2 ðx Þfðt Þ 2 ðx Þdx
∂f 0
Z h Z h
00 00
 00
2
¼ EIu ðt Þ (10.4.7b)
1 ðx Þ 2 ðx Þdx + EI fðt Þ 2 ðx Þ dx
0 0
6EI 4EI
¼ u+ f
h2 h

(ii) Kinetic energy: The kinetic energy with respect to point O is evaluated
from Eq. (1.5.8). Taking the origin of the coordinates at point O we have:
XP ¼ u, YP ¼ 0, w ¼ f, _ xc ¼ 0, yc ¼ a=2, IP ¼ IO and Eq. (1.5.8)
becomes
1 1 a
T ¼ m u_ 2 + IO f_ 2  m u_ f_ (10.4.8)
2 2 2
Differentiating Eq. (10.4.8) with respect to u_ and f_ gives
∂T a
¼ m u_  m f_ (10.4.9a)
∂u_ 2
∂T a
¼ IO f_  m u_ (10.4.9b)
∂f_ 2

(iii) Generalized forces: They result from Eq. (1.8.23) by setting q1 ¼ u,

q2 ¼ f. Thus, we obtain
Q1 ¼ pðt Þ, Q2 ¼ pðt Þa=2 (10.4.10a,b)

(iv) The potential of the external conservative forces: Because there are no
conservative forces, it is A ¼ 0.
Substituting Eqs. (10.4.7a), (10.4.7b), (10.4.9a) (10.4.9b), (10.4.10a,b)
into the Lagrange equation (1.8.11) gives
a 12EI 6EI
m u€  m f€ + 3 u + 2 f ¼ pðt Þ (10.4.11a)
2 h h
a 6EI 4EI a
IO f€  m u€ + 2 u + f ¼  pðt Þ (10.4.11b)
2 h h 2
 MDOF systems: Models and equations of motion Chapter 10 333

or in matrix form
M€
u + Ku ¼pðt Þ (10.4.12)
where
2 3
2 ma 3 12EI 6EI
m  6 h3 h2 7
M ¼ 4 ma 2 5, K¼6 4 6EI 4EI 5
7 (10.4.13a,b)
 IO
2 h2 h
  ( )
u pðt Þ
u¼ , pðt Þ ¼ a (10.4.13c,d)
f pðt Þ
2

2. The method of equilibrium of forces

The forces acting on the square rigid body during the motion are shown in
Fig. 10.4.2. The equilibrium of forces may be considered either with respect
to point O or to the center of mass C .

FIG. 10.4.2 Free body diagram.

a. Equilibrium with respect to point O.

Because point O does not coincide with the mass center, the equations
of motion will be obtained from Eqs. (1.5.14a), (1.5.14b), (1.5.14c) by
taking the origin of axes at the point O and setting: P  O, uP ¼ u,
vP ¼ 0, xc ¼ 0, yc ¼ a=2, Fx ¼ pðt Þ  fS , Fy ¼ 0, MO ¼ pðt Þa=2  MS .
Thus, we obtain
a
m€u  m f€ + fS ¼ pðt Þ (10.4.14a)
2
a a
m u€ + IO f€ + MS ¼ pðt Þ (10.4.14b)
2 2
The elastic forces fS and MS are the shear force and the bending
moment of a beam fixed at end A and subjected to the displacement u
and rotation f at end O. Thus, we have
12EI 6EI
fS ¼ 3
u+ 2 f (10.4.15a)
h h
334 PART II Multi-degree-of-freedom systems

6EI 4EI
MS ¼ 2
u+ f (10.4.15b)
h h
Obviously, they are identical to those given by Eqs. (10.4.7a),
(10.4.7b).
Substituting the previous expressions for fS and MS in
Eqs. (10.4.14a), (10.4.14b) yields
a 12EI 6EI
u  m f€ + 3 u + 2 f ¼ pðt Þ
m€ (10.4.16a)
2 h h
a 6EI 4EI a
IO f€  m u€ + 2 u + f ¼  pðt Þ (10.4.16b)
2 h h 2
which are identical to Eqs. (10.4.11a), (10.4.11b).
b. Equilibrium with respect to the mass center C .
The equations of motion will be obtained from Eqs. (1.5.11a),
(1.5.11b), (1.5.11c) by setting uc ¼ u  af=2, Fx ¼ fS + pðt Þ,
MC ¼ MS  fS a=2. Thus, we obtain
a
m€u  m f€ + fS ¼ pðt Þ (10.4.17a)
2
a
Ic f€ + MS + fS ¼ 0 (10.4.17b)
2
The first of the above equations is identical to (10.4.14a). The second
equation, however, looks different from (10.4.14b). Nevertheless, multi-
plying Eq. (10.4.17a) by a=2 and adding it to (10.4.17b) gives

a 2 
a a
 m€ u + IC + m f€ + MS ¼  pðt Þ (10.4.18a)
2 2 2
Because IC + m ða=2Þ2 ¼ IO (Steiner’s formula), the previous equa-
tion becomes
a a
 m€ u + IO f€ + MS ¼  pðt Þ (10.4.18b)
2 2
which is identical to (10.4.14b).
The above transformation of the equations of motion from the center
of mass to point O is rather occasional. A formal method to transform the
equations of motion when we change the point of reference is presented in
Section 10.7.

10.4.1 The method of influence coefficients

The equations of motion of linear systems with localized masses and distributed
stiffnesses can be derived using the method of influence coefficients, which is
based on the superposition principle. A frame structure can be idealized as an
assemblage of beam elements interconnected at nodal points. We illustrate this
 MDOF systems: Models and equations of motion Chapter 10 335

method with the plane frame of Fig. 10.4.3. The masses are localized at the
nodes, where the external loads are applied. In the general case in which the
axial deformation of the beams and columns is not neglected, each node i
has three degrees of freedom with respect to the global system of axes X Y ,
 Y axes, respectively, and a rota-
two translations ui , vi in the directions of the X,
 
tion fi about the Z axes. Hence, the frame of Fig. 10.4.3 with n ¼ 6 free nodes
has in total N ¼ 3n ¼ 18 degrees of freedom.

FIG. 10.4.3 Frame with n ¼ 6 nodes.

Referring to Fig. 10.4.4, we represent the displacement vector by


 T ¼ u1 v1 f1 u2 v2 f2 ⋯ ui vi fi ⋯ un vn fn
u (10.4.19)
or if they are labeled from 1 to N ¼ 3n, we may write
uT ¼ f u1 u2 u3 u4 u5 u6 ⋯ u3i2 u3i1 u3i ⋯ uN g (10.4.20)
where
u3i2 ¼ ui , u3i1 ¼ vi , u3i ¼ fi , i ¼ 1, 2, …,n (10.4.21)

FIG. 10.4.4 Nodal displacements of the frame. N ¼ 3n ¼ 18.

336 PART II Multi-degree-of-freedom systems

Similarly, we formulate the vector of the external nodal loads (Fig. 10.4.5)

FIG. 10.4.5 External loads acting on the nodes of the frame.

pðt ÞT ¼ f px1 py1 M1 px2 py2 M2 ⋯ pxi pyi Mi ⋯ MN g

(10.4.22)
¼ f p1 p2 p3 p4 p5 p6 ⋯ p3i2 p3i1 p3i ⋯ pN g
where
p3i2 ¼ pxi , p3i1 ¼ pyi , p3i ¼ Mi , i ¼ 1, 2, …,n (10.4.23)
The number of equations will be equal to the number of degrees of freedom.
The equations of motion can be formulated using the method of equilibrium of
forces. If fIi , fDi , fSi represent the inertia, damping, and elastic forces, respec-
tively, and pi ðt Þ the external load in the direction of the displacement ui , then
it should be (see Fig. 10.4.6)
fIi + fDi + fSi ¼ pi ðt Þ, i ¼ 1, 2, …,N (10.4.24)

(a) (b)
FIG. 10.4.6 Equilibrium of forces (a) or moments (b) in the direction of the displacement ui .

Eq. (10.4.24) holds in the direction of all displacement components

u1 , u2 ,…,uN , N ¼ 3n. The expressions of the forces fIi , fDi , fSi can be deter-
mined using the following method, which is referred to as the method of influ-
ence coefficients.
 MDOF systems: Models and equations of motion Chapter 10 337

10.4.1.1 Elastic forces

In a SDOF system, the elastic force depends only on one displacement. Appar-
ently, in a MDOF system this force will depend on all the displacements. Thus,
for a linear system, the superposition principle allows writing
fSi ¼ ki1 u1 + ki2 u2 + … + kij uj + … + kiN uN (10.4.25)

FIG. 10.4.7 Stiffness influence coefficients ki4 .

where kij are constants expressing the stiffness influence coefficients or simply
the stiffness coefficients. They relate the elastic force fSi to the displacement uj .
The physical meaning of the coefficient kij results by setting uj ¼ 1 and uk6¼j ¼ 0
in Eq. (10.4.25), that is, by applying a unit displacement along the degree of
freedom j, holding all other displacements zero, as shown in Fig. 10.4.7. This
yields
fSi ¼ kij (10.4.26)
Eq. (10.4.26) states that the stiffness coefficient kij expresses the elastic
force applied along the degree of freedom i for a unit displacement along the
degree of freedom j, that is, the displacement uj , while all other displacements
are zero.
Fig. 10.4.7 shows the deformation of the frame when the displacement u4 ¼ 1
is applied at node 4 while all other displacements are zero, that is, uj ¼ 0
( j ¼ 1, 2, …, N , j 6¼ 4). Obviously, the stiffness influence coefficients are equal
to the forces required to maintain the deformed shape of the frame when sub-
jected to the displacement u4 ¼ 1.
For i ¼ 1, 2, …, N , Eq. (10.4.25) yields N equations, which we write in
matrix form
8 9 2 38 9
> f > k k ⋯ k1N > > u1 >
< S1 >
> = 6 11 12 7< u2 =
>
fS2 k k ⋯ k
¼6 21 22
4 ⋯ ⋯ ⋯ ⋯ 5> ⋮ >
2N 7 (10.4.27)
> ⋮ >
>
>
: ; : >
> ;
fSN kn1 kn2 ⋯ kNN uN
338 PART II Multi-degree-of-freedom systems

or
f S ¼ ku (10.4.28)
where
8 9 8 9 2 3
> fS1 > > u1 > k11 k12 ⋯ k1N
>
> >
> > >
> >
< fS1 = < u2 = 6k k ⋯ k 7
6 21 22 2N 7
fS ¼ , u¼ , k¼6 7 (10.4.29)
> ⋮ >
> > > ⋮ >
> > 4⋯ ⋯ ⋯ ⋯ 5
>
: >
; >
: ; >
fSN uN kn1 kn2 ⋯ kNN
The vector f S with dimensions N  1 represents the vector of the elastic
forces, the vector u with dimensions N  1 represents the vector of the displace-
ments, and the matrix k with dimensions N  N represents the stiffness matrix
of the structure. Obviously, this method, which explains the physical meaning
of the stiffness coefficients, is by no means suitable for their evaluation because
it requires N static analyses of the fixed structure successively for u1 ¼ 1,
u2 ¼ 1, …, uN ¼ 1. The stiffness matrix, however, can be established using
other methods of structural analysis, for example, the flexibility method or
the direct stiffness method. Anyhow, the establishment of the stiffness matrix
of a structure is a subject of the static structural analysis and it is evaluated using
the most suitable method for a particular structure.

10.4.1.2 Damping forces

The concept of the influence coefficients can be employed to express the damp-
ing forces fDi . Thus, assuming viscous damping we may write
fDi ¼ ci1 u_ 1 + ci2 u_ 2 + … + cij u_ j + … + ciN u_ N (10.4.30)
Applying Eq. (10.4.30) for i ¼ 1, 2, …, N yields N equations, which we
write in matrix form
8 9 2 38 9
> fD1 > c11 c12 ⋯ c1N > u_ 1 >
> >
> > >
>
< f = 6 c c ⋯ c 7> < u_ > =
D2 6 21 22 2N 7 2
¼6 7 (10.4.31)
>
>
> ⋮ >>
>
4 ⋯ ⋯ ⋯ ⋯ 5> >
> ⋮ >>
>
: ; : ;
fDN cn1 cn2 ⋯ cNN u_ N
or
f D ¼ cu_ (10.4.32)
where
8 9 8 9 2 3
> fD1 > > u_ 1 > c11 c12 ⋯ c1N
>
> >
> > >
> >
< fD2 = < u_ 2 = 6c c ⋯ c
6 21 22 2N
7
7
fD ¼ , u_ ¼ , c¼6 7 (10.4.33)
> ⋮ >
> > > ⋮ >
> > 4⋯ ⋯ ⋯ ⋯ 5
>
: >
; >
: ; >
fDN u_ N cn1 cn2 ⋯ cNN
 MDOF systems: Models and equations of motion Chapter 10 339

FIG. 10.4.8 Damping influence coefficients ci4 .

The vector f D with dimensions N  1 represents the vector of the damping

forces, the vector u_ with dimensions N  1 represents the vector of the veloc-
ities, and the matrix c with dimensions N  N represents the damping matrix of
the structure. The elements cij of the damping matrix are the damping influence
coefficients. Their physical meaning is analogous to that of the stiffness influ-
ence coefficients, namely cij expresses the damping force applied along the
degree of freedom i for unit velocity along the degree of freedom j, that is,
u_ j ¼ 1, while all other velocities are zero. Fig. 10.4.8 show the damping influ-
ence coefficients at the nodes of the frame, when u_ 4 ¼ 1, u_ j ¼ 0 (j ¼ 1, 2, …, N ,
j 6¼ 4). Note that the damping coefficients, contrary to the stiffness coefficients,
cannot be established from the geometrical data of the structure and the physical
properties of its elements. In Section 12.11, we will show how the damping
matrix of the structure can be established by assuming known damping ratios
for each mode shape estimated from experimental data obtained from similar
structures.

10.4.1.3 Inertial forces

Similarly, we can relate the inertial force fIi to the acceleration u€ j . Thus, we may
write
fIi ¼ mi1 u€ 1 + mi2 u€ 2 + … + mij u€ j + … + miN u€ N (10.4.34)
For i ¼ 1, 2, …, N , Eq. (10.4.34) yields N equations, which we write in
matrix form
8 9 2 38 9
> fI 1 >
> m11 m12 ⋯ m1N > u€ 1 >
>
< = 6 7 < >
> =
fI 2 m m ⋯ m u€2
¼6 21 22 2N 7
(10.4.35)
> ⋮ > 4 ⋯ ⋯ ⋯ ⋯ 5> ⋮ >
: >
> ; : >
> ;
fIN mn1 mn2 ⋯ mNN u€ N
or
f I ¼ m€
u (10.4.36)
340 PART II Multi-degree-of-freedom systems

where
8 9 8 9 2 3
> fI 1 > > u€ 1 > m11 m12 ⋯ m1N
< >
> = < >
> = 6 m21 m22
fI 2 u€ 2 ⋯ m2N 7
fI ¼ , €¼
u , m¼6
4 ⋯ ⋯
7 (10.4.37)
> ⋮ > > ⋮ >
> ⋯ ⋯ 5
: >
> ; : >
;
fIN u€ N mn1 mn2 ⋯ mNN
The vector f I with dimensions N  1 represents the vector of the inertial
forces, the vector u € with dimensions N  1 represents the vector of the accel-
erations, and the matrix m with dimensions N  N represents the mass matrix
or inertial matrix of the structure. The elements mij of the mass matrix are the
mass influence coefficients. Their physical meaning is analogous to that of the
stiffness and damping influence coefficients, namely mij expresses the inertial
force applied along the degree of freedom i for unit acceleration along the
degree of freedom j, that is, u€ j ¼ 1, while all other accelerations are zero.
Fig. 10.4.9 shows the mass influence coefficients at the nodes of the frame,
when u€ 4 ¼ 1, u€ j ¼ 0 (j ¼ 1, 2, …, N , j 6¼ 4).

FIG. 10.4.9 Mass influence coefficients mi4 .

In actual structures, the mass is distributed. The model that considers the
mass lumped at certain points of the structure, for example, at the nodes of a
frame, approximates adequately the dynamic response of the structure. When
it is assumed that the lumped mass has no geometrical dimensions, that is, it
is simulated by a material particle, then its rotational inertia is zero and the
respective influence coefficients vanish.
Writing now Eq. (10.4.24) for all directions i ¼ 1, 2, …, N , we obtain
fI 1 + fD1 + fS1 ¼ p1 ðt Þ
fI 2 + fD2 + fS2 ¼ p2 ðt Þ
(10.4.38)
… … … …
fIN + fDN + fSN ¼ pN ðt Þ
or
f I + f D + f S ¼ p ðt Þ (10.4.39)
 MDOF systems: Models and equations of motion Chapter 10 341

which by virtue of Eqs. (10.4.28), (10.4.32), (10.4.36) gives the equation of

motion of the structure
u + cu_ + ku ¼ pðt Þ
m€ (10.4.40)

10.5 Systems with distributed mass and distributed stiffness

Systems with distributed mass and distributed stiffness are also referred to as
continuous systems or distributed parameter systems. Theoretically, they have
infinite degrees of freedom and their motion is described by partial differential
equations. The systems of Fig. 10.5.1 belong to this category.

(a) (b)
FIG. 10.5.1 Continuous systems: (a) Cantilever beam, (b) Chimney fixed on the ground.

Continuous systems can be approximated by MDOF systems, even if the

mass distribution does not necessarily have a particular concentration at some
points. This approximation can be realized using two methods. The first method
analyzes the structure by expressing its deformed shape as a superposition of a
series of global shape functions of the spatial coordinates, each multiplied by its
own generalized coordinate. This method is known as the method of global
shape functions. On the contrary, the second method treats the structure as a
discrete MDOF system in which the mass and stiffness are concentrated at cer-
tain points, but with interacting displacements. The discretization can be per-
formed either by the flexibility method or the stiffness method. The latter is
realized by the finite element method (FEM) [1], in which the structure is
approximated by a set of discrete elements with known deformation and inertial
properties. Both methods, that is, the method of the global shape functions and
the FEM, convert the governing partial differential equation into a system of
ordinary differential equations of motion, which can be solved by well-
established numerical methods. Consequently, the dynamic analysis of contin-
uous systems through a direct analytical solution of partial differential
equations, a very difficult and in most cases insurmountable mathematical
342 PART II Multi-degree-of-freedom systems

problem, is circumvented. Nevertheless, the research is ongoing and new effi-

cient methods have been developed for solving time-dependent structural prob-
lems, for example, the boundary element method (BEM) [2] or the meshless
methods beyond the element methods [3]. The method of global shape functions
is presented below while the FEM is presented in this Chapter 11.

10.5.1 The method of global shape functions

In this method, it is assumed that the continuous function u ðx, t Þ, which repre-
sents the displacement of the points of the structure, can be approximated by the
finite superposition series
X
N
uðx, t Þ ¼ 1 ðx Þu1 ðt Þ + 2 ðx Þu2 ðt Þ + … + N ðx ÞuN ðt Þ ¼ i ðx Þui ðt Þ
i¼1
(10.5.1)
The functions i ðx Þ express shapes of deformation of the structure and are
referred to as global shape functions while ui ðt Þ are functions of time that
express the generalized coordinates, for example, displacements or rotations
of certain points of the structure. The minimum requirements for the
expression (10.5.1) to approximate the actual deformation u ðx, t Þ are: (i) the
shape functions must be at least geometrically admissible, that is, satisfy
the geometric (essential) boundary (support) conditions of the structure, and
(ii) must be linearly independent, that is, any of them cannot result as a linear
combination of two or more of the others.
The generalized coordinates are the unknown time functions, which must be
determined by the solution of equations of motion. The number of generalized
coordinates is equal to the degrees of freedom of the substitute MDOF system,
which approximates the actual continuous system. The equations of motion are
usually derived using the principle of virtual work, the Lagrange equations, or
the Hamilton principle. Here, the latter approach, as presented for the general-
ized SDOF system in Section 8.2, is employed to illustrate the method of global
shape functions.

FIG. 10.5.2 Cantilever beam.

 MDOF systems: Models and equations of motion Chapter 10 343

Suppose we are going to apply the method of global shape functions to

approximate the dynamic response of the cantilever beam of Fig. 10.5.2. The
cantilever has variable stiffness I ðx Þ and mass m ðx Þ. It is loaded by the trans-
verse load pðx, t Þ and the axial force P at the free end.
(i) Elastic energy: The elastic energy is equal to the strain energy of the can-
tilever and it is given (see Eq. 8.1.11)
Z
1 L
EI ðx Þ½u00 ðx, t Þ dx
2
U¼ (10.5.2)
2 0
Its variation is
Z L
dU ¼ EI ðx Þu00 ðx, t Þd u00 ðx, t Þdx (10.5.3)
0

or using Eq. (10.5.1)

Z L ! !
X
N X
N
00 00
dU ¼ EI ðx Þ i ui j duj dx
0 i¼1 j¼1
Z !
L N X
X N
00 00
¼ EI ðx Þ i j ui duj dx (10.5.4)
0 i¼1 j¼1

X
N X
N
¼ kij ui duj
i¼1 j¼1

where
Z L
00 00
kij ¼ EI ðx Þ i j dx (10.5.5)
0

(ii) Kinetic energy: The kinetic energy is due to the transverse displacements
and rotations of the mass elements. Thus, we have
Z Z h i2
1 L 1 L 0
m ðx Þ½u_ ðx, t Þ dx + I ðx Þ u_ ðx, t Þ dx
2
T¼ (10.5.6)
2 0 2 0
The second term in the above expression is due to the rotation of the
cross-sections and its contribution is small. In the following development,
without limiting the generality, we omit this term for the sake of simplicity.
The variation of the kinetic energy is
Z L
dT ¼ m ðx Þu_ ðx, t Þd u_ ðx, t Þdx (10.5.7)
0

or using Eq. (10.5.1)

344 PART II Multi-degree-of-freedom systems

Z ! !
L X
N X
N
dT ¼ m ðx Þ i u_ i j d u_ j dx
0 i¼1 j¼1
Z !
L X
N X
N
¼ m ðx Þ i j u_ i d u_ j dx (10.5.8)
0 i¼1 j¼1

X
N X
N
¼ mij u_ i d u_ j
i¼1 j¼1

where
Z L
mij ¼ m ðx Þ i j dx (10.5.9)
0

(iii) Virtual work of the nonconservative forces:

The nonconservative forces are the load pðx, t Þ and the damping force.
The damping may be external or internal. As in the case of generalized
SDOF systems (see Fig. 8.2.1), we assume that the external damping force
is viscous and is distributed along the length of the beam, that is,
fD ðx, t Þ ¼ cðx Þu_ ðx, t Þ, where cðx Þ is the damping coefficient. The internal
damping is due to the deformation of the element of the beam, resists it, and
depends on the velocity of the strain. Hence, if the produced stress is denoted
by sD , we may write
∂ex
sD ¼ cs (10.5.10)
∂t
where cs is the coefficient of the internal damping and ex the stain. Then,
the virtual work of the internal nonconservative forces is
Z
dWnc ¼  sD dex dV
in
(10.5.11)
V
Taking into account that
sx M
ex ¼ , sx ¼ y, M ¼ EI ðx Þu00 ðx, t Þ (10.5.12)
E I ðx Þ
Eq. (10.5.11) becomes
Z L
00
dWncin
¼ cs I ðx Þu_ ðx, t Þd u00 ðx, t Þdx (10.5.13)
0
or using Eq. (10.5.1)
Z L ! !
X
N X
N
00 00
¼ cs I ðx Þ i u_ i
in
dWnc j duj dx
0 i¼1 j¼1
Z !
L X
N X
N
00 00
¼ cs I ðx Þ i j u_ i duj dx (10.5.14)
0 i¼1 j¼1
N X
X N
¼ cijin u_ i duj
i¼1 j¼1
 MDOF systems: Models and equations of motion Chapter 10 345

where
Z L
00 00
cijin ¼ c s I ðx Þ i j dx (10.5.15)
0

The virtual work of the external damping force is

Z L
dWncex
¼ cðx Þu_ ðx, t Þd uðx, t Þdx (10.5.16)
0

or using Eq. (10.5.1)

Z L ! !
X
N X
N
dWnc ¼  c ðx Þ i u_ i
ex
j duj dx
0 i¼1 j¼1
Z !
L N X
X N
¼ c ðx Þ i j u_ i duj dx (10.5.17)
0 i¼1 j¼1

X
N X
N
¼ cijex u_ i duj
i¼1 j¼1

where
Z L
cijex ¼ c ðx Þ i j dx (10.5.18)
0

Finally, the virtual work due the external nonconservative load pðx, t Þ is
Z L
dWnc ¼
p
pðx, t Þd uðx, t Þdx (10.5.19)
0

or using Eq. (10.5.1)

Z L X
N
p
dWnc ¼ pðx, t Þ j duj dx
0 j¼1
(10.5.20)
X
N
¼ pj ðt Þduj
j¼1

where
Z L
pj ðt Þ ¼ pðx, t Þ j dx (10.5.21)
0

Consequently, the virtual work due to all nonconservative forces is

X
N X
N X
N
dWnc ¼  cij u_ i duj + pj ðt Þduj (10.5.22)
i¼1 j¼1 j¼1
346 PART II Multi-degree-of-freedom systems

where it was set

cij¼ cijin + cijex (10.5.23)

(iv) The potential energy of the conservative forces: The potential energy of the
conservative forces is due to the constant axial force P. Thus, we have
A ¼ Pe (10.5.24)
where e is the shorting of the elastic curve due to bending and is given as
Z L Z
1 L 0 2
e¼ ex dx ¼ ½u ðx, t Þ dx (10.5.25)
0 2 0

The variation of the potential energy is

dA ¼ Pde
Z L (10.5.26)
¼ P u0 ðx, t Þd u0 ðx, t Þdx
0

or using Eq. (10.5.1)

Z ! !
L X
N X
N
0 0
dA ¼ P i ui j duj dx
0 i¼1 j¼1
Z !
L X
N X
N
0 0
¼ P i j ui duj dx (10.5.27)
0 i¼1 j¼1

X
N X
N
¼ P kGij ui duj
i¼1 j¼1

where
Z L
0 0
kGij ¼ i j dx (10.5.28)
0

Substituting the previous expressions for dU ,dK , dWnc , and dA in

Hamilton’s principle, Eq. (1.7.13), produces the following equations of
motion
X
N X
N X
N 
mij u€ j + cij u_ j + kij  Pk Gij uj ¼ pi , i ¼ 1, 2…, N (10.5.29)
j¼1 j¼1 j¼1

or in matrix form
u + Cu_ + ðK  PKG Þu ¼pðt Þ
M€ (10.5.30)
The matrices M, C, K represent the mass, damping, and stiffness matri-
ces, respectively. Their elements are evaluated from Eqs. (10.5.9),
 MDOF systems: Models and equations of motion Chapter 10 347

(10.5.23), (10.5.5). The matrix KG , whose elements are evaluated from

Eq. (10.5.28), is referred to as the geometric stiffness matrix. It expresses
here the influence of the axial force on the dynamic response of the struc-
ture. It is obvious that the compressive axial force reduces the stiffness of the
system while the tensile force increases it. The vanishing of the determinant
of the total stiffness matrix K  PKG , namely
det ðK  PKG Þ ¼ 0 (10.5.31)
yields the N values of the buckling load.
The method of global shape functions is also known as the Ritz method. The
shape functions are, as we have already mentioned, geometrically admissible
and linearly independent functions representing deformation patterns of the
entire structure. They are chosen arbitrarily but appropriately, based on the
experience of the user of the method. The success of the method depends on
the choice of shape functions, but this is still a difficult problem [4]. The use
of the Ritz method has been substantially reduced with the emergence of mod-
ern computational methods, such as the finite element method (FEM) [1], the
boundary element method (BEM) [2], and the meshless methods [3]. However,
it is useful either as the theoretical background for developing new computa-
tional methods or for testing them.

10.6 Mixed systems

Mixed systems are those resulting from the combination of the previously dis-
cussed categories. As an example of this case, we refer to the structure of
Fig. 10.6.1a. Indeed, this system has localized masses (m1 ,m2 ), distributed
masses (mass of bars), localized stiffness (stiffness kR of the rotational spring),
and distributed stiffnesses (stiffness of bars). We assume: m1 ¼ m2 ¼ m,
m ¼ m=a, kR ¼ EI =4a.

(a)

(b)
FIG. 10.6.1 Mixed system (a) and motion parameters (b).
348 PART II Multi-degree-of-freedom systems

For the determination of motion, we take as degrees of freedom the displace-

ments of the masses, namely the rotation f1 of the mass m1 , the displacement u,

and the rotation f2 of the mass m2 ; see Fig. 10.6.1b. The equations of motion
will be derived using the Lagrange equations.
(i) Elastic energy: If u1 , u2 , u3 , u4 denote the displacements and the rotations at
the ends of the beam element as in Fig. 10.6.2, its elastic curve can be set in
the form
uðx, t Þ ¼ 1 ðx Þu1 + 2 ðx Þu2 + 3 ðx Þu3 + 4 ðx Þu4 (10.6.1)
The shape functions i express the elastic curves of the beam for ui ¼ 1,
when uj ¼ 0, j 6¼ i. Following the procedure presented in Section 10.4, we
obtain.

FIG. 10.6.2 Degrees of freedom of the beam element.

 x 2  x 3
1 ðx Þ ¼ 1  3 +2 ¼ 1  3x 2 + 2x 3 (10.6.2a)
L L
 x  x 2
2 ðx Þ ¼ L  1 ¼ Lx ðx  1Þ2 (10.6.2b)
L L
 x 2  x 3
3 ðx Þ ¼ 3 2 ¼ 3x2  2x3 (10.6.2c)
L L
 x 2  x 
4 ð x Þ ¼ L  1 ¼ Lx2 ðx  1Þ (10.6.2d)
L L
where
x ¼ x=h (10.6.3)
For the element 2–3 holds: u1 ¼ af1 , u2 ¼ f1 , u3 ¼ u  a f2 =2, u4 ¼ f2 ,
L ¼ 3a.
For the element 4–5 holds: u1 ¼ u + a f2 =2, u2 ¼ f2 , u3 ¼ 0, u4 ¼ 0,
L ¼ 3a.
The elastic energy is given by the expression
Z Z
1 3 1 5 1
EI ½u00 ðx, t Þ dx + EI ½u00 ðx, t Þ dx + kR f1
2 2 2
U¼ (10.6.4)
2 2 2 4 2
or using Eq. (10.6.1), we obtain
 MDOF systems: Models and equations of motion Chapter 10 349

Z
 1 3a  
 f1 , f2 ¼
U u, EI 00 
1 ðx Þa f1 +
00 
2 ðx Þf1 +
00
3 ðx Þ u  a f2 =2 + 00  2
4 ðx Þf2 dx
2 0
Z
1 3a   1
00
u + a f2 =2 + 00  2 + kR f1
2
+ EI 1 ðx Þ 2 ðx Þf2 dx
2 0 2
(10.6.5)
Differentiating the above relation yields after evaluation of the integrals
∂U
¼ k11 u + k12 f1 + k13 f2 (10.6.6a)
∂u
∂U
¼ k21 u + k22 f1 + k23 f2 (10.6.6b)
∂f1
∂U
¼ k31 u + k32 f1 + k33 f2 (10.6.6c)
∂f2
where
8 EI 10 EI
k11 ¼ , k12 ¼ k21 ¼  , k13 ¼ k31 ¼ 0
9 a3 9 a2
121 EI 17 EI 38 EI
k22 ¼ , k23 ¼ k32 ¼ , k33 ¼
36 a 9 a 9 a
Therefore the stiffness matrix of the system is
2 3
32 40a 0
EI 4
k¼ 40a 121a2 17a 2 5 (10.6.7)
36a 3
0 17a 2 152a 2

(ii) Kinetic energy: The kinetic energy results from the expression
Z Z
1 3 1 5
 ½u_ ðx, t Þ dx +  ½u_ ðx, t Þ dx
2 2
T¼ m m
2 2 2 4
(10.6.8)
1 1 1
+ I1 f_ 12 + m u_ 2 + IC f_ 22
2 2 2
or using Eq. (10.6.1), we obtain
  1 Z 3a h   i2
_ f_ 1 , f_ 2 ¼
T u,  1 ðx Þa f_ 1 + 2 ðx Þf_ 1 + 3 ðx Þ u_  a f_ 2 =2 +
m _
4 ð x Þf 2 dx
2 0
Z
1 3a h   i2
+  1 ðx Þ u_ + a f_ 2 =2 + 2 ðx Þf_ 2 dx
m
2 0
2 2
1 1 1
+ I1 f_ + m u_ + IC f_
2
2 1 2 2 2
(10.6.9)
350 PART II Multi-degree-of-freedom systems

Differentiating the above relation yields after evaluation of the integrals

 
d ∂T
 + m12 f€1 + m13 f€2
¼ m11 u€ (10.6.10a)
dt ∂u_
!
d ∂T
 + m22 f€1 + m23 f€2
¼ m21 u€ (10.6.10b)
dt ∂f_
1
!
d ∂T
 + m32 f€1 + m33 f€2
¼ m31 u€ (10.6.10c)
dt ∂f_
2

where
2712 558
m11 ¼ m, m12 ¼ m21 ¼ ma, m13 ¼ m31 ¼ 0
840 840
2294 2 675 2 1832 2
m22 ¼ ma , m23 ¼ m32 ¼  ma , m33 ¼ ma
840 840 840
Therefore, the mass matrix of the system is
2 3
2712 558a 0
m 4
m¼ 558a 2294a 2 675a 2 5 (10.6.11)
840
0 675a 2 1832a 2

(iii) Generalized forces: The virtual work of the nonconservative forces is due
to the load pðt Þ. Thus, we have

Q1 d u + Q2 df1 + Q3 df2 ¼ pðt Þd u + a f2 =2
(10.6.12)
¼ pðt Þd u  pðt Þadf2 =2
from which we obtain
Q1 ¼ pðt Þ
Q2 ¼ 0
Q3 ¼ pðt Þa=2
Hence the equation of motion of the mixed system is
2 38 9
€
m6
2712 558a 0 < u >
> =
2 7 f €

4 558a 2294a 675a 5
2
1
840 : € >
> ;
0 675a 2 1832a 2 f2
2 38 9 8 9 (10.6.13)
32 40a 0 >
< u >= >
< pðt Þ >
=
EI 6 2 7 
+ 4 40a 121a 2
17a 5 f ¼ 0
36a3 >
: >
1
; > : >
;
0 17a 2 152a 2 f2 pðt Þa=2
 MDOF systems: Models and equations of motion Chapter 10 351

10.7 Transformations of the equations of motion

In the previous sections, we presented various methods for the formulation of
the equations of motions of a MDOF system with N degrees of freedom. In each
case, a set of independent displacements is selected, which, as stated in
Section 1.8, expresses the generalized coordinates or simply the coordinates,
as we will now be referring to, of the system and are denoted by the vector
of the displacements u in the equation of motions. It is often necessary or con-
venient to express the equations in a system of also independent coordinates u 

other than that originally selected. The transformation of the coordinates from u
to u is a linear relation of the form
u ¼R
u (10.7.1)
where R is the N  N square matrix referred to as the transformation matrix. If
M, C, and K are the mass, damping, and stiffness matrices, respectively, and
pðt Þ the vector of the external loads in the original system of coordinates u,
we need to determine the matrices M,  C, K and the vector pðt Þ in the trans-
formed system of coordinates u . This can be achieved using the principle of vir-
tual work as follows.
The elastic forces in the coordinates u are given by the relation
f S ¼Ku (10.7.2)
 by the relation
while in the coordinates u
f S ¼K
u (10.7.3)
We give u a virtual displacement du. Then the corresponding virtual dis-
placement du results from the relation (10.7.1)
du ¼Rd
u (10.7.4)
The virtual work done by the elastic forces f S due to the virtual displacement
du is
dWS ¼duT f S
(10.7.5)
¼ duT Ku
or using Eqs. (10.7.1), (10.7.4)
dWS ¼d
uT RT KR
u (10.7.6)
Similarly, the virtual work of the elastic forces in the system of the coordi-
 is
nates u
 S ¼d
dW uT f S
(10.7.7)
¼ d
uT Ku
352 PART II Multi-degree-of-freedom systems

 S ¼ dWS , which gives

Obviously, it must be d W
uT RT KR
d u ¼ d u
uT K
or

uT K
d   RT KR u¼0 (10.7.8)
 and d
Because the vectors u u are nonzero, it must be
 ¼ RT KR
K (10.7.9)
Similarly, by considering the virtual work of the inertial forces f I ¼ M€
u and
f I ¼ Mu
€ in the two systems of coordinates u and u , we obtain
 ¼ RT MR
M (10.7.10)
Further, by considering the virtual work of the damping forces f D ¼ Cu_ kai
f D ¼ C
u_ in the two systems of coordinates, we prove that
 ¼ RT CR
C (10.7.11)
Finally, by expressing the virtual work of the external forces in the two sys-
tems of coordinates, we have

uÞT pðt Þ
dWp ¼ duT pðt Þ ¼ ðRd
(10.7.12)
uT RT pðt Þ
¼ d

and
 p ¼ d
dW  ðt Þ
uT p (10.7.13)
 p , we obtain
Because these two virtual works are equal, dWp ¼ d W
ðt Þ ¼ RT pðt Þ
p (10.7.14)

Example 10.7.1 The equation of motion of the system shown in Fig. 10.4.1 with
respect to point O is (see Eqs. 10.4.11a, 10.4.11b):
2 ma 3 

m    ( pðt Þ )
6 2 7 u€ + k11 k12 u
¼
4 ma 5 € a (1)
 IO f k21 k22 f pðt Þ
2 2

where
12EI 6EI 4EI
k11 ¼ , k12 ¼ k21 ¼ , k22 ¼ (2)
h3 h2 h
Transform Eq. (1) with respect to the mass center C of the body.
 MDOF systems: Models and equations of motion Chapter 10 353

Solution
If u and f denote the displacement of the mass center and its rotation about it,
then the transformation relations result by geometrical consideration as
a
u ¼ u + f (3)
2
f ¼ f (4)
or
  " a # 
u 1 u
¼ 2 (5)
f 0 1 f

Hence
"
a# " #
1 1 0
R¼ 2 , RT ¼ a (6)
1
0 1 2
Then Eq. (10.7.10) gives
2 a 32 a 3
" # m m
1 0
¼ a 6 2 74 1 2 5
M 4 5
1 a
2 m IO 0 1
2 (7)
" #
m 0
¼
0 IC
where
a2
IC ¼ m (8)
6
is the moment of inertia with respect to the center of mass C .
Further, Eq. (10.7.9) gives
" #" #2 3
1 0 k11 k12 1 a
¼ a
K 4 25
1 k21 k22
2 0 1
" # (9)
k11 k12
¼
k21 k22
where
12EI
k11 ¼ k11 ¼ 3
h
  a 6EI
k 12 ¼ k 21 ¼ k11 + k12 ¼ 3 ða + h Þ (10)
a 2
2 h  
 12EI a 2 ah h 2
k 22 ¼ k11 + ak 12 + k22 ¼ 3 + +
2 h 4 2 3
354 PART II Multi-degree-of-freedom systems

Finally, Eq. (10.7.14) gives

2 38 9
1 0 < pðt Þ =
 ðt Þ ¼ 4 a 5
p a
1 : pðt Þ ;
2 2 (11)
( )
pðt Þ
¼
0
Therefore, the transformed equation of motion is
" #( ) 2 3( ) ( )
2 a+h
m 0 u€ 6EI 4 u pðt Þ
+ 3 3a 2 + 6ah + 2h 2 5  ¼ (12)
0 IC f€ h a+h f 0
6
Note that we obtain the same equations if the equations are derived directly
with respect to center of mass C .

10.8 Problems
Problem P10.1 Formulate the equations of motion of the system shown in
Fig. P10.1. Data: k1 ¼ 3k, k2 ¼ 2k, k3 ¼ k, c1 ¼ c3 ¼ c, c2 ¼ 2c, m1 ¼ m2 ¼ m,
and m3 ¼ 2m.

FIG. P10.1 System in problem P10.1.

Problem P10.2 A square plate with a side length a, constant thickness

d ¼ a=10, and mass density g is supported by three columns of height
h ¼ a=2 and square cross-section of a side length b ¼ a=20 (Fig. P10.2).
Neglecting the axial deformation and the mass of the columns, formulate the
equations of motion of the structure using the Lagrange equations. The module
of elasticity is E and the shear modulus G ¼ 0:4E.
 MDOF systems: Models and equations of motion Chapter 10 355

FIG. P10.2 Structure in problem P10.2.

Problem P10.3 A rigid bar of total mass m is supported by the three springs
k1 , k2 , k3 and the damper c as shown in Fig. P10.3. The circular disc of mass
0:5m at the end D of the bar has a diameter 0:2L. Formulate the equation of
motion. Data: k2 ¼ 1:5k1 and k3 ¼ 2k1 .

FIG. P10.3 System in problem P10.3.

Problem P10.4 Formulate the equations of motion of the plane frame of

Fig. P10.4 when the supports 1 and 2 are subjected to the horizontal displace-
ments ug1 ðt Þ and ug2 ðt Þ, respectively. Give the expressions for the evaluation of
the support reactions. The beams are assumed rigid (I ¼ 1) while the columns
have the same cross-section and modulus of elasticity E.

FIG. P10.4 Frame in problem P10.4.

356 PART II Multi-degree-of-freedom systems

Problem P10.5 The beam of Fig. P10.5 consists of the flexible and massless
part AB and the rigid part BC of mass m  per unit length. The beam is clamped
at A while the hinged support at C is elastically restrained by the rotational
spring CR ¼ EI =10L. The beam is loaded by a concentrated moment M ðt Þ
applied at B. Formulate the equation of motion of the structure and give the
expressions for the evaluation of the support reactions.

FIG. P10.5 Beam in problem P10.5.

Problem P.10.6 The structure of Fig. P10.6 consists of the two bars AB and
BC . The bar AB is massless and flexible and is simply supported at A while
 per unit length and is elastically restrained at
the bar BC is rigid with mass m
the hinged end C by the rotational spring CR ¼ EI =10L. The structure is
loaded by a concentrated moment M ðt Þ applied at B. Formulate the equation
of motion of the structure and give the expressions for the evaluation of the
support reactions.

FIG. P10.6 Structure in problem P10.6.

Problem P10.7 Formulate the equation of motion of the plane frame of

Fig. P10.7. The massless cables FB and GC have cross-sectional area A and
are prestressed so that they can undertake compressive forces. Data:
 ¼ m=a (kg=m), and CR ¼ EI =10L.
I =A ¼ a2 =25, m
 MDOF systems: Models and equations of motion Chapter 10 357

FIG. P10.7 Plane frame in problem P10.7.

Problem P10.8 Formulate the equations of motion of the system shown in

 per unit length while the cable
Fig. P10.8. The bar AB is rigid and has mass m pffiffiffi
and the pulley are assumed massless. The total length of the cable is 1:5a 2.
The hinged support at A is elastically restrained by the rotational spring
CR ¼ EAa=250. The hanging body of mass m ¼ ma  is moving downward with
a velocity v ¼ v0 sin wt.

FIG. P10.8 System in problem P10.8.

Problem P10.9 The thin spherical tank of Fig. P10.9 has a diameter R ¼ a=3
and is full with a liquid of density g. It is supported on the ground through nine,
hinged at both ends, massless rods of cross-sectional area A and modulus of
elasticity E. The points 10 ,20 ,30 lie on a horizontal circular ring of the spherical
tank at a height a from the ground and at a distance R=3 from the center of the
sphere. The mass of the tank is neglected.
358 PART II Multi-degree-of-freedom systems

Formulate the equation of motion when

1. The structure performs free vibrations.
2. The structure is subjected to ground motion ug ðt Þ in the direction b with
respect to the x axis.

FIG. P10.9 Spherical tank in problem P10.9.

Problem P10.10 A vehicle traveling along a bridge is idealized as shown in

Fig. P10.10. The unsprung mass mc is connected to the sprung mass through
the suspension system (k,c). The bridge has a span L and is simply supported.
Using the Ritz method, formulate the equation of motion of the system bridge-
vehicle when the vehicle is traveling with a constant velocity v. The bridge is
simulated by a two-degree-of-freedom system with global shape functions
n ¼ sin ðnpx=LÞ, n ¼ 1, 2. The mass mc maintains its contact to the bridge dur-

ing the motion. Data: mc ¼ m0 =10, m0 ¼ mL=20, and k ¼ EI =100L.

FIG. P10.10 Vehicle traveling on a bridge in problem P10.10.

References and further reading

[1] O. Zienkiewicz, R. Taylor, The finite element method, seventh ed., Butterworth-Heinemann,
Oxford, UK, 2013.
[2] J.T. Katsikadelis, The boundary element method for engineers and scientists, second ed.,
Academic press, Elsevier, Oxford, UK, 2016.
[3] G.R. Liu, Meshfree methods: moving beyond the finite element method, second ed., CRC Press,
Boca Raton, FL, 2010.
[4] J.T. Katsikadelis, A generalized ritz method for partial differential equations in domains of arbi-
trary geometry using global shape functions. Eng. Anal. Bound. Elem. 32 (5) (2008) 353–367,
https://doi.org/10.1016/j.enganabound.2007.001.
Chapter 11

The finite element method

Chapter outline
11.1 Introduction 359 11.8 The finite element method
11.2 The finite element method for the plane grid 453
for the plane truss 360 11.8.1 Properties of the
11.2.1 Properties of the plane plane grid element 453
truss element 360 11.8.2 Transformation of the
11.2.2 Transformation of the nodal coordinates
nodal coordinates of of the plane grid
the truss element 374 element 463
11.2.3 Equation of motion of 11.9 The finite element method
the plane truss 377 for the space frame 476
11.2.4 Steps to formulate the 11.9.1 Properties of the
equations of motion space frame element 476
for a plane truss by the 11.9.2 Transformation of the
finite element method 381 nodal coordinates of
11.2.5 Modification of the the space frame
equations of motion element 482
due to the supports 11.10 The finite element method
of the structure 382 for the space truss 494
11.3 The finite element method 11.10.1 Properties of the
for the plane frame 392 space truss element 494
11.3.1 Properties of the 11.10.2 Transformation of
plane frame element 392 the nodal coordinates
11.3.2 Transformation of the of the space truss
nodal coordinates of the element 498
plane frame element 410 11.11 Rigid bodies within flexible
11.4 Static condensation: Guyan’s skeletal structures 503
reduction 424 11.11.1 Rigid bodies in
11.5 Flexural vibrations of a spaces frames 503
plane frame 429 11.11.2 Rigid bodies in
11.6 Reduction of the degrees spaces trusses, plane
of freedom due to grids, plane frames,
constraints 439 and plane trusses 510
11.7 Axial constraints in the 11.12 Problems 517
plane frame 442 References and further reading 521

Dynamic Analysis of Structures. https://doi.org/10.1016/B978-0-12-818643-5.00011-X

© 2020 Elsevier Inc. All rights reserved. 359
360 PART II Multi-degree-of-freedom systems

11.1 Introduction
The distributed mass is the usual case we encounter in actual structures, that is,
the actual structures should be modeled as continuous systems. This modeling
leads to partial differential equations with prescribed initial and boundary con-
ditions. The solution of such equations is a difficult mathematical problem, even
for individual components of structures (e.g., beams, plates, walls, etc.). On the
other hand, the method of global shape functions, which can provide an accept-
able solution, has limited capabilities due to the difficulty in choosing the global
shape functions [1]. In practice, this method cannot cope with conventional
structures such as a multistory frame or a grid. Besides, it cannot be pro-
grammed for automatic use on a computer. Therefore, neither method is
suitable for the dynamic analysis of realistic structures in engineering. At this
dead end, the finite element method (FEM) method gives a way out.
The FEM represents a major breakthrough in the field of computational
mechanics. Extensive literature on the general formulation of the FEM [2] as
well as on its application to dynamic problems is now available [3].
In the finite element approach, the structure being analyzed (e.g., frame,
wall, plate, shell, three-dimensional deformable body, or a combination of
them) is divided into a finite number of small subregions, elements, which
are interconnected at discrete points, the nodes, where the compatibility
condition for the displacements and the equilibrium of equivalent nodal forces
are ensured. After the discretization, we assume that the field function (the dis-
placement in this case) within each element varies according to a known law,
which is expressed as a superposition of shape patterns, the shape functions,
corresponding to unit values of the element nodal displacements. After that,
the elastic and kinetic energy as well as the virtual work of the nonconservative
nodal forces in terms of the nodal displacements, viewed as generalized coor-
dinates, are established. The equivalent nodal forces (elastic, inertial, damping,
and external forces) may result by the use of the Lagrange equations or the prin-
ciple of virtual work. Finally, the compatibility of the nodal displacements and
the equilibrium of the nodal forces give the differential equations of motion of
the structure. The FEM is presented below only for skeletal structures, that is,
structures consisting of straight-line elements (trusses, frames, grids) because
surface structures (walls, plates, shells) as well as three-dimensional (3D)
deformable bodies do not fall within the scope of this book.

11.2 The finite element method for the plane truss

11.2.1 Properties of the plane truss element
We consider a plane truss element with a variable cross-section Aðx Þ, mass per
unit length m ðx Þ, and modulus of elasticity E, whose local axis in the unde-
formed state coincides with the x axis. The ends j, k of the element are called
nodes of the element, thus any quantity (force, displacement) associated with
them is referred to as nodal quantity (e.g., nodal force, nodal displacement).
 The finite element method Chapter 11 361

Fig. 11.2.1 Degrees of freedom of a plane truss element.

In the deformed state at time t, the ends of the elements are displaced to points
j 0 , k 0 and the element occupies the position j 0 k 0 (Fig. 11.2.1).
The element has four degrees of freedom, the displacements u1 ðt Þ, u2 ðt Þ of
node j and the displacements u3 ðt Þ, u4 ðt Þ of node k. These quantities constitute
the nodal coordinates of the plane truss element. A typical point x of the ele-
ment axis undergoes two displacements, the axial displacement u ðx, t Þ and the
transverse one v ðx, t Þ. We consider the element as a generalized system and
apply the Ritz method presented in Section 10.5.1. Thus, we set the axial
displacement in the form
u ðx, t Þ ¼ u1 ðt Þ 1 ðx Þ + u 3 ðt Þ 3 ðx Þ (11.2.1)
where 1 ðx Þ and 3 ðx Þ are shape functions expressing the axial deformation for
u1 ¼ 1, u3 ¼ 0 and u1 ¼ 0, u3 ¼ 1, respectively. The functions 1 ðx Þ and 3 ðx Þ
can be established as follows.
The equilibrium of the element dx (see Fig. 11.2.2) yields
dN
¼0 (11.2.2)
dx
or taking into account that
N ¼ Aðx Þsx
¼ EAðx Þex (11.2.3)
d i
¼ EAðx Þ
dx
we write Eq. (11.2.2) as


d d i
EAðx Þ ¼0 (11.2.4)
dx dx
and after integration
Z
dx
i ¼ c1 + c2 (11.2.5)
Aðx Þ

Fig. 11.2.2 Beam element dx.

362 PART II Multi-degree-of-freedom systems

The arbitrary constants c1 and c2 are evaluated from the boundary condi-
tions. It is obvious that for a variable cross-section, the determination of the
shape functions requires the evaluation of the integral in Eq. (11.2.5). This
relation, although it gives shape functions expressing the exact static axial
deformation, is not suitable for the automation of the method because different
shape functions have to be determined for elements with a different law of
variation of the cross-section. This difficulty is surpassed if we accept the
same shape functions for all elements, regardless of the variation law of the
cross-section and, indeed, those resulting from Eq. (11.2.4) for a constant
cross-section. In that case, we obtain
i ¼ c1 x + c2
which for 1 ð0Þ ¼ 1 and 1 ð LÞ ¼ 0 gives
1 ðx Þ ¼ 1  x, x ¼ x=L (11.2.6a)
and for 3 ð 0Þ ¼ 0 and 3 ð LÞ ¼ 1

3 ðx Þ ¼ x, x ¼ x=L (11.2.6b)
The shape functions (11.2.6a), (11.26b) can also result if we arbitrarily
accept a linear law of variation of axial displacement within the element.
The assumption of linear variation of the displacement is at the expense of accu-
racy. However, the error is acceptable when the element is small and the cross-
sectional variation is not intense. This becomes clear from Example 11.2.1.
Taking into account that the element is a straight line in the deformed state,
we can write the transverse displacement asa
v ðx, t Þ ¼ u2 ð1  x Þ + u4 x
(11.2.7)
¼ u2 2 ðx Þ + u4 4 ðx Þ

a. Due to the axial deformartion we should write

v ðx, t Þ ¼ u2 ð1  x 0 Þ + u4 x 0 , x0 ¼ x 0 =L0 (a)

But it is
x0 x + u ðx, t Þ x=L + u ðx, t Þ=L u ðx, t Þ
x0 ¼ ¼ ¼ x+ (b)
L0 L + ðu3  u1 Þ 1 + ðu3  u1 Þ=L L

because ðu3  u1 Þ=L≪1:

Eq. (a) by virtue of Eq. (b) becomes



u ðx, t Þ u ðx, t Þ
v ðx, t Þ  u2 1  x  + u4 x +
L L
u4  u2
 u2 ð1  x Þ + u4 x  u ðx, t Þ
L
 u2 ð1  x Þ + u4 x

because the nonlinear term u4 u

L u ðx, t Þ is much smaller compared to the linear ones and thus it can
2

be neglected.
 The finite element method Chapter 11 363

Obviously, it is 2 ðx Þ ¼ 1 ðx Þ and 4 ðx Þ ¼ 3 ðx Þ.
The equivalent nodal forces in the directions of the degrees of freedom can
be established using the method of the Lagrange equations or the principle of
virtual work. Both methods are presented in the following.

11.2.1.1 The method of the Lagrange equations

We refer to Eq. (1.8.11), which we write for the truss element
 
d ∂T ∂T ∂U ∂A
 + + ¼ Qi , i ¼ 1, 2, 3, 4 (11.2.8)
dt ∂u_ i ∂ui ∂ui ∂ui
The forces ∂A=∂ui , derived from the potential A, are gravitational forces. As
shown in Section 1.9, they are ignored in the linear dynamic analysis but the
deformation produced by them should be superimposed to that resulting from
the dynamic analysis in order to obtain the total deformation of the structure.
Obviously, we deduce from Eq. (11.2.8) that the terms
 
d ∂T ∂T
fIi ¼  (11.2.9)
dt ∂u_ i ∂ui
∂U
fSi ¼ (11.2.10)
∂ui
express the nodal inertial force and the nodal elastic force, respectively, in the
direction of the generalized coordinate ui . Finally, the force Qi in Eq. (11.2.8)
includes the nonconservative generalized forces. In our case, these are the
damping forces fDi and the external forces pi . Regarding the sign convention
for the nodal forces, they are positive when their direction coincides with the
positive direction of the axes they refer.
(i) Nodal elastic forces and stiffness matrix of the truss element
Eq. (11.2.7) expresses the displacement of the element as a rigid body.
Therefore, the elastic energy is due only to the axial deformation of the element
given by
Z
1
U¼ sx ex dV (11.2.11)
2 V
Taking into account that sx ¼ Eex , ex ¼ ∂u=∂x, the previous relation
becomes
Z Z  2
1 L ∂u
U¼ E dxdydz
2 0 A ∂x
Z L  2 (11.2.12)
1 ∂u
¼ EAðx Þ dx
2 0 ∂x
or using Eq. (11.2.1)
364 PART II Multi-degree-of-freedom systems

Z
1 L  0 0
2
U ðu1 , u3 Þ ¼ EAðx Þ u1 1 ðx Þ + u3 3 ðx Þ dx (11.2.13)
2 0

The nodal elastic forces result from Eq. (11.2.10) for i ¼ 1, 2, 3, 4. Thus,
we obtain
∂U
fS1 ¼ ¼ k11 u1 + k13 u3 (11.2.14a)
∂u1
∂U
fS2 ¼ ¼0 (11.2.14b)
∂u2
∂U
fS3 ¼ ¼ k31 u1 + k33 u3 (11.2.14c)
∂u3
∂U
fS4 ¼ ¼0 (11.2.14d)
∂u4
where
Z L
kij ¼ EAðx Þ 0i ðx Þ 0j ðx Þdx, i, j ¼ 1, 3 (11.2.15)
0

Eqs. (11.2.14a)–(11.2.14d) are written in matrix form as

8 9 2 38 9
> fS1 > k11 0 k13 0 > u1 >
> >
>
> > 6
> 7>> >
> >
>
> >
< fS2 = 6 0 0 0 07 < >
> =
6 7 u2
¼6 7 (11.2.16)
> > 6 07 > u3 >
> fS3 >
>
> > 4 k31 0 k33
> 5>> >
> >
>
>
: ; > : >
> ;
fS4 0 0 0 0 u4
or
f eS ¼ ke ue (11.2.17)
where u are the vectors of the nodal elastic forces and the nodal displace-
f eS , e

ments, respectively, and ke the stiffness matrix of the e truss element. Hence, the
element matrix ke for the plane truss element is defined as
2 3
k11 0 k13 0
6 7
60 0 0 07
6 7
k ¼6
e
7 (11.2.18)
6 k31 0 k33 0 7
4 5
0 0 0 0
Obviously, we deduce from Eq. (11.2.15) that kij ¼ kji . Hence, the stiffness
matrix is symmetric.
For an element with a constant cross-section Aðx Þ ¼ Ae and a length Le ,
Eq. (11.2.15) is integrated analytically and Eq. (11.2.18) yields
 The finite element method Chapter 11 365

2 3
1 0 1 0
e6
EA 0 0 0 07
ke ¼ e 6 7 (11.2.19)
L 4 1 0 1 05
0 0 0 0
It should be noted that Eq. (11.2.10) expresses the Castigliano theorem.
Therefore, the previous method for establishing the stiffness matrix is identical
to the so-called energy method.
(ii) Nodal inertial forces and mass matrix of the truss element
In the FEM, the equivalent inertial forces are obtained by two different assump-
tions of the mass distribution on the element: the consistent mass assumption,
which assumes a continuous distribution of the mass on the element, and the
lumped mass assumption, which lumps the mass at its nodes. The inertial mass
matrices resulting from both assumptions are derived below.
(a) Consistent mass matrix
During the motion, the infinitesimal mass m ðx Þdx undergoes the two displace-
ments u ðx, t Þ and v ðx, t Þ. Therefore, the kinetic energy of the truss element will
be given by the expression
Z n o
1 L
T¼ m ðx Þ ½u_ ðx, t Þ2 + ½v_ ðx, t Þ2 dx (11.2.20)
2 0
or using Eqs. (11.2.1), (11.2.7)
Z n o
1 L 2 2
T ðu_ 1 , …, u_ 4 Þ ¼ m ðx Þ ½u_ 1 1 ðx Þ + u_ 3 3 ðx Þ + ½u_ 2 1 ðx Þ + u_ 4 3 ðx Þ dx
2 0
(11.2.21)
The inertial forces result from Eq. (11.2.9) for i ¼ 1, 2, 3, 4. Thus, after
performing the differentiations we obtain
 
d ∂T ∂T
fI 1 ¼  ¼ m11 u€1 + m13 u€3 (11.2.22a)
dt ∂u_ 1 ∂u1
 
d ∂T ∂T
fI 2 ¼  ¼ m22 u€2 + m24 u€4 (11.2.22b)
dt ∂u_ 2 ∂u2
 
d ∂T ∂T
fI 3 ¼  ¼ m31 u€1 + m33 u€3 (11.2.22c)
dt ∂u_ 3 ∂u3
 
d ∂T ∂T
fI 4 ¼  ¼ m42 u€2 + m44 u€4 (11.2.22d)
dt ∂u_ 4 ∂u4
where
Z L
mij ¼ mi + 1, j + 1 ¼ m ðx Þ i ðx Þ j ðx Þdx, i, j ¼ 1, 3 (11.2.23)
0
366 PART II Multi-degree-of-freedom systems

Eqs. (11.2.22a)–(11.2.22d) are written in matrix form

8 9 2 38 9
fI 1 > u€
> > 1>
>
m11 0 m13 0
>
> > >
> >
>
>
< fI 2 = 6 7>> >
60 m22 0 m24 7 u€2 =
<
6
¼6 7 (11.2.24)
>
> fI 3 >
>
> 4 m31 0 m33 0 7 5>> u€3 >
>
>
> ;
: > > >
>
: >
;
fI 4 0 m42 0 m44 u€4
or
€e
f eI ¼ me u (11.2.25)
€e are the vectors of the nodal inertial forces and the nodal acceler-
where f eI , u
ations, respectively, and me the mass matrix of the e truss element. Therefore,
the element mass matrix me is defined as
2 3
m11 0 m13 0
60 m22 0 m24 7
me ¼ 64 m31 0
7 (11.2.26)
m33 0 5
0 m42 0 m44
Obviously, we deduce from Eq. (11.2.23) that mij ¼ mji . Hence, the mass
matrix is symmetric.
 Eq. (11.2.23) is integrated
For an element with a constant mass, m ðx Þ ¼ m,
analytically and Eq. (11.2.26) gives
2 3
2 0 1 0
e6
m 60 2 0 17 7
me ¼ (11.2.27)
6 41 0 2 05
0 1 0 2
 e Le is the total mass of the e element.
where m e ¼ m
(b) Lumped mass matrix
According to this assumption, the mass of the element is concentrated at its
nodes by static considerations, that is, they are obtained as the reactions of a
simply supported beam under the load m ðx Þ (see Fig. 11.2.3). Thus, we have
Z L
m1 ¼ m ðx Þð1  x Þdx (11.2.28a)
0
Z L
m2 ¼ m ðx Þxdx (11.2.28b)
0

Therefore, the kinetic energy of the truss element is given by the expression
1  1 
T ¼ m1 u_ 1 2 + u_ 2 2 + m2 u_ 3 2 + u_ 4 2 (11.2.29)
2 2
 The finite element method Chapter 11 367

Fig. 11.2.3 Lumped mass assumption.

The inertial forces result from Eq. (11.2.9) for i ¼ 1,2, 3, 4. Hence, after
performing the differentiations, we obtain
 
d ∂T ∂T
fI 1 ¼  ¼ m11 u€1 (11.2.30a)
dt ∂u_ 1 ∂u1
 
d ∂T ∂T
fI 2 ¼  ¼ m22 u€2 (11.2.30b)
dt ∂u_ 2 ∂u2
 
d ∂T ∂T
fI 3 ¼  ¼ m33 u€3 (11.2.30c)
dt ∂u_ 3 ∂u3
 
d ∂T ∂T
fI 4 ¼  ¼ m44 u€4 (11.2.30d)
dt ∂u_ 4 ∂u4
where
m11 ¼ m22 ¼ m1 , m33 ¼ m44 ¼ m2 (11.2.31)
Eqs. (11.2.30a)–(11.2.3d) are written in matrix form as
8 9 2 38 9
fI 1 > u€
> > 1>
>
m11 0 0 0
>
> > >
> >
>
>
< fI 2 = 6 7>> >
60 m22 0 0 7 u€2 =
<
6
¼6 7 (11.2.32)
>
>
> fI 3 >
>
> 40 0 m33 0 7 5>>
> u€3 >
>
>
> ;
: > : >
> ;
fI 4 0 0 0 m44 u€4
Therefore, the element mass matrix me is defined as
2 3
m11 0 0 0
60 0 7
6 m22 0 7
me ¼ 6 7 (11.2.33)
40 0 m33 0 5
0 0 0 m44
We observe that the lumped mass assumption results in a diagonal mass matrix.
For an element with constant mass, m ðx Þ ¼ m  Eqs. (11.2.28a),
(11.2.28b) give
 e Le =2
m11 ¼ m22 ¼ m33 ¼ m44 ¼ m (11.2.34)
368 PART II Multi-degree-of-freedom systems

and the mass matrix, Eq. (11.2.33), becomes

2 3
1 0 0 0
1 e6 0 1 0 07
m ¼ m 6
e 7 (11.2.35)
2 40 0 1 05
0 0 0 1
 e Le is the total mass of the truss element e.
where m e ¼ m
(iii) Nodal damping forces and damping matrix of the truss element
We consider only internal damping. It is therefore due to the deformation of the
truss element; it resists it and depends on the velocity of the strain. If we denote
with sD the axial stress due to damping, we can write
∂ex
sD ¼ cs (11.2.36)
∂t
where cs is the coefficient of the internal damping and ex the axial strain.
The virtual work of the internal damping force in the truss element is
Z
dWnc ¼  sD dex dV
in
(11.2.37)
V

Taking into account that ex ¼ ∂u=∂x, the previous relation becomes

Z L
∂2 u ðx, t Þ ∂u ðx, t Þ
dWncin
¼ c s Að x Þ d dx (11.2.38)
0 ∂x∂t ∂x
or using Eq. (11.2.1)
Z L
 0 0
  0 0

dWncin
¼ cs Aðx Þ u_ 1 1 ðx Þ + u_ 3 3 ðx Þd u1 1 ðx Þ + u 3 3 ðx Þ dx
0
¼ ðc11 u_ 1 + c13 u_ 3 Þdu1  ðc31 u_ 1 + c33 u_ 3 Þdu3
(11.2.39)
where
Z L
cij ¼ cs Aðx Þ 0i ðx Þ 0j ðx Þdx, i, j ¼ 1, 3 (11.2.40)
0

The damping forces are obtained as generalized forces from Eqs. (11.2.8),
(1.8.8) for i ¼ 1, 2, 3, 4. Thus we have
fD1 ¼ Q1 ¼ c11 u_ 1 + c13 u_ 3 (11.2.41a)
fD2 ¼ Q2 ¼ 0 (11.2.41b)
fD3 ¼ Q3 ¼ c31 u_ 1 + c33 u_ 3 (11.2.41c)
fD4 ¼ Q4 ¼ 0