Contents

1. Cover
2. Preface
3. About the Authors
4. Chapter 1: Introduction
a. 1.1 Overview of Structural Dynamics
b. 1.2 Dynamic Loads
c. 1.3 Characteristics of a Dynamic Problem
d. 1.4 Application of Structural Dynamics
e. Exercises
f. References
5. Chapter 2: Establishment of the Structural Equation of Motion
a. 2.1 General
b. 2.2 Formulation of the Equations of Motion
c. 2.3 Theory of Total Potential Energy Invariant Value of Elastic System Dynamics
d. 2.4 Influence of Gravitational Forces
e. 2.5 Influence of Support Excitation
f. Exercises
g. References
6. Chapter 3: Single Degree of Freedom Systems
a. 3.1 Response of Free Vibrations
b. 3.2 Response to Harmonic Loading
c. 3.3 Periodic Load Response
d. 3.4 Impulsive Loading Response
e. 3.5 Response of Arbitrary Load
f. 3.6 Energy in Vibration
g. 3.7 Structural Vibration Test
h. 3.8 Vibration Isolation Principle
i. 3.9 Structural Vibration Induced Fatigue
j. Exercises
k. References
7. Chapter 4: Multi-Degree of Freedom System
a. 4.1 Two Degrees of Freedom System
b. 4.2 Free Vibrations of Undamped System
c. 4.3 Practical Calculation Method of Dynamic Characteristics
d. 4.4 Mode Superposition Method for Damped System
e. 4.5 Numerical Analysis of Damping System
f. 4.6 Stability and Accuracy Analysis of Stepwise Integration Method
g. Exercises
h. References
8. Chapter 5: Distributed-Parameter System
a. 5.1 Overview
b. 5.2 Establish Differential Equations for Motion
c. 5.3 Free Vibration of a Beam
d. 5.4 Orthogonality Relationships
e. 5.5 Modal Decomposition
f. References
9. Chapter 6: Stochastic Structural Vibrations
a. 6.1 Overview
b. 6.2 Stochastic Process
c. 6.3 Stochastic Response of Linear SDOF System
d. 6.4 Stochastic Response of Linear MDOF System
e. 6.5 Nonlinear Structural Stochastic Response Analysis
f. 6.6 State Space Method for Structural Stochastic Response Analysis
g. Exercises
h. References
10. Chapter 7: Research Topics of Structural Dynamics
a. 7.1 Analysis of Structural Seismic Response
b. 7.2 Structural Vibration Control
c. 7.3 Modal Analysis and Theory
d. 7.4 Structural Dynamic Damage Identification
 e. 7.5 Nonlinear Problems of Dynamic Analysis
f. 7.6 Sub-Structure Method
g. 7.7 Dynamics of Offshore Structures
h. Exercises
i. References
11. Chapter 8: Structural Dynamics of Marine Pipeline and Riser
a. 8.1 Overview
b. 8.2 Environmental Conditions
c. 8.3 Hydrodynamic Loads
d. 8.4 Structural Response Analysis
e. 8.5 Vortex Induced Vibrations
f. Exercises
g. References
12. Answers to Exercises
13. Index
14. End User License Agreement

List of Illustrations
1. Chapter 1
a. Figure 1.1 Classification of dynamic loads.
b. Figure 1.2 Types of dynamic load.
c. Figure 1.3 The difference between static and dynamic.
d. Figure 1.4 Schematics of lumped mass procedure.
e. Figure 1.5 Discretization of a simple beam by generalized coordinate procedure.
2. Chapter 2
a. Figure 2.1 Definition of degree of freedom.
b. Figure 2.2 Single body with SDOF dynamic system.
c. Figure 2.3 Dual body with 2DOF dynamic system.
d. Figure 2.4 Inverted Pendulum system with SDOF.
e. Figure 2.5 Inverted Pendulum system with 1DOF.
f. Figure 2.6 Single body with 1DOF dynamic system.
g. Figure 2.7 Influence of gravity on SDOF equilibrium.
h. Figure 2.8
3. Chapter 3
a. Figure 3.1 Commonly used single degree of freedom systems.
b. Figure 3.2 To simplify a complex structure.
c. Figure 3.3 Single degree of freedom system.
d. Figure 3.4 The response of undamped free vibration system.
e. Figure 3.5 Influence of different damping on free vibration motion.
f. Figure 3.6 Influence of different damping on free vibration motion.
g. Figure 3.7 Free vibration decay curve.
h. Figure 3.8 The curve of the dynamic amplification factor Rd of the undamped…
i. Figure 3.9 The dynamic response of undamped system increases with time during resonance.
j. Figure 3.10 A non-compact beam with a motor.
k. Figure 3.11 Resonance reaction time-history curve of damping system.
l. Figure 3.12 Amplitude-frequency characteristic curve of damping system.
m. Figure 3.13 Phase-frequency characteristic curve of damping system.
n. Figure 3.14 Simple basic dynamic system.
o. Figure 3.15 Calculating damping ratio on amplitude-frequency characteristic curve.
p. Figure 3.16 Simple dynamical system.
q. Figure 3.17 Displacement time-history diagram.
r. Figure 3.18 Sine-wave impulse.
s. Figure 3.19 Rectangular impulse.
t. Figure 3.20 Triangular impulse.
u. Figure 3.21 Deduction of Duhamel integral under arbitrary load.
v. Figure 3.22 δ function load diagram.
w. Figure 3.23 Viscous damping energy dissipation hysteretic curve.
x. Figure 3.24 Working block diagram of electromagnetic worktable.
y. Figure 3.25 Typical hammer and impact pulse.
z. Figure 3.26 Working principle of acceleration sensor.
aa. Figure 3.27 Working principle of displacement sensor.
ab. Figure 3.28 Single degree of freedom active vibration isolation system.
ac. Figure 3.29 Single degree of freedom passive vibration isolation system.
 ad. Figure 3.30 Relationship of transmissibility with different frequency ratio.
4. Chapter 4
a. Figure 4.1 Multiple degree of freedom system with one point.
b. Figure 4.2 Multi-degree of freedom system.
c. Figure 4.3 Establishment of the equation of motion by elasticity method.
d. Figure 4.4 Dynamic balance of two degrees of freedom system.
e. Figure 4.5 The physical meaning of the stiffness factor.
f. Figure 4.6 Five degrees of freedom of system.
g. Figure 4.7 The reciprocity theorem of work.
h. Figure 4.8 Two-layer shear frame structure and vibration mode.
i. Figure 4.9 Three-particle simply support beam structure.
j. Figure 4.10 Unequal height single story plant.
k. Figure 4.11 Simplified calculation of simply supported beam.
l. Figure 4.12 Simplified calculation of the beam with concentrated mass.
m. Figure 4.13 Cantilever beam structure.
n. Figure 4.14 Simple support beam structure.
o. Figure 4.15 The constant average acceleration method.
p. Figure 4.16 Schematic diagram of linear acceleration.
q. Figure 4.17 Comparison of three basic assumptions for acceleration.
r. Figure 4.18 The basic assumptions of the acceleration method of the Wilson-θ…
s. Figure 4.19 External load excitation diagram.
5. Chapter 5
a. Figure 5.1 (a) Transverse vibration of a beam; the transverse displacement is v(x,t)…
b. Figure 5.2 Free body diagram of the beam in presence of an axial preload N.
c. Figure 5.3 Bar subjected to dynamic axial deformations: (a) bar properties and coordinates;…
d. Figure 5.4 Simply supported beam. (a) boundary conditions. (b) Mode shapes.
e. Figure 5.5 Free-free beam. The solutions of the equation cosμ = (coshµ)–1 …
f. Figure 5.6 Free-free beam. (a) Boundary conditions (b) Mode shapes (rigid body modes and flexible…
g. Figure 5.7 Simple beam-vibration analysis: (a) basic properties of simple beam; (b) first…
6. Chapter 6
a. Figure 6.1 Classification of stochastic vibration problems.
b. Figure 6.2 Sample function of stochastic process.
c. Figure 6.3 Unilateral and bilateral spectral density of stationary process.
d. Figure 6.4 Spectral density and sample function of narrowband process.
e. Figure 6.5 Spectral density and sample function of broadband process.
f. Figure 6.6 Filtered stationary white noise model.
g. Figure 6.7 Filtered white noise model and improved filtered white noise model.
h. Figure 6.8 Power spectrum of acceleration of strong earthquake filtered by multiple layer.
i. Figure 6.9 Time envelope function of quasi-stationary process.
j. Figure 6.10 Calculation diagram of single-layer structure.
k. Figure 6.11 Multi-layer shear structure under the action of seism.
l. Figure 6.12 Large-span structure under the action of seism.
7. Chapter 7
a. Figure 7.1 Acceleration records of El-Centro ground motion.
b. Figure 7.2 The process of calculating the accelerationg response spectrum.
c. Figure 7.3 El-Centro seismic wave acceleration response spectrum.
d. Figure 7.4 Earthquake impact coefficient α curve.
e. Figure 7.5 Bottom shearing force method illustration.
f. Figure 7.6 Two-story reinforced concrete frame structure.
g. Figure 7.7 Interlayer shearing force illustration.
h. Figure 7.8 Viscoelastic dampers illustration.
i. Figure 7.9 Rubber isolation bearing illustration.
j. Figure 7.10 Rubber isolation bearing illustration model.
k. Figure 7.11 Resilience-displacement curve of rubber isolation bearing.
l. Figure 7.12 Several calculation models of isolated structures.
m. Figure 7.13 Three modes of magneto-rheological damper.
n. Figure 7.14 Magneto-rheological damper commonly used in building structure shock absorption.
o. Figure 7.15 MDOF system.
p. Figure 7.16 Two different spring models.
q. Figure 7.17 Hysteretic curve of reinforced concrete beams.
r. Figure 7.18 Two commonly used resilience models.
s. Figure 7.19 The mass steel wire system which has geometric nonlinearity.
t. Figure 7.20 P – Δ effect of single particle system.
 8. Chapter 8
a. Figure 8.1 Uses of offshore pipelines and risers.
b. Figure 8.2 Regular long-crested waves.
c. Figure 8.3 Random long-crested waves.
d. Figure 8.4 Wave spectrum.
e. Figure 8.5 2048s zero mean wave amplitude time realization.
f. Figure 8.6 Frequency domain and time domain representation of long-crested waves.
g. Figure 8.7 Example Hm0 – Tp contour lines.
h. Figure 8.8 Flow field around pipe.
i. Figure 8.9 Fluid drag and inertia forces acting on a pipe section.
j. Figure 8.10 CL in shear and shear-free flow for 103< Re < 30 x 104.
k. Figure 8.11 Sketch of the riser system.
l. Figure 8.12 Element defintion for a finite element idealization of a vertical riser.
m. Figure 8.13 Vibrations of a cylinder submitted to vortex shedding.
n. Figure 8.14 Typical pipeline spans.
o. Figure 8.15 Vertical riser eigenmodes.
p. Figure 8.16 Flow regimes for smooth circular cylinders.
q. Figure 8.17 Relation between the Strouhal Number and Reynolds Number.
r. Figure 8.18 A schematic model of a 2-D freely oscillating rigid cylinder.
s. Figure 8.19 A schematic model of a spring mounted rigid cylinder undergoing two-dimensional…
t. Figure 8.20 Layout of the experiments (left); the vacuum tank (on the top of the structure)…
u. Figure 8.21 The modal shape up to transverse forth mode.
v. Figure 8.22 Vibration response of the riser. The cross-flow instantaneous deflected shapes of…

List of Tables
1. Chapter 7
a. Table 7. 1 Application of viscoelastic dampers in wind-resistant earthquake engineering.
b. Table 7.2 Experimental study on the structure with viscoelastic dampers.
2. Chapter 8
a. Table 8.1 Parameters of model data in Chaplin’s test.
b. Table 8.2 The modal frequencies from modal analysis.
 Scrivener Publishing
100 Cummings Center, Suite 541J
Beverly, MA 01915-6106

Publishers at Scrivener
Martin Scrivener ([email protected])
Phillip Carmical ([email protected])
Structural Dynamics

Yong Bai
Zhao-Dong Xu
This edition first published 2019 by John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07 030, USA and Scrivener Publishing LLC, 100 Cummings Center, Suite 541J,
Beverly, MA 01915, USA
© 2019 Scrivener Publishing LLC
For more information about Scrivener publications please visit www.scrivenerpublishing.com.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical,
photocopying, recording, or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at
http://www.wiley.com/go/permissions.
Wiley Global Headquarters
111 River Street, Hoboken, NJ 07 030, USA
For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com.
Limit of Liability/Disclaimer of Warranty
While the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or
completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a
particular purpose. No warranty may be created or extended by sales representatives, written sales materials, or promotional statements for this work. The fact that an
organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors
endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the
publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a
specialist where appropriate. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special,
incidental, consequential, or other damages. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work
was written and when it is read.
Library of Congress Cataloging-in-Publication Data
ISBN 97 8-1-119-60560-7
Preface
Dynamic problems of structures are ubiquitous in research. Therefore, it is very important for students majoring in civil engineering, mechanical
engineering, aircraft engineering and ocean engineering to systematically grasp the basic concepts, calculation principles and calculation methods of
structural dynamics. This book focuses on the basic theories and concepts, as well as the application and background of theories and concepts in
engineering. Since the basic principles and methods of dynamics are applied to other various engineering fields, this book can also be used as a
reference for undergraduate and graduate students in other majors.
The main contents include basic theory of dynamics, establishment of equation of motion, single degree of freedom systems, multi-degree of freedom
systems, distributed-parameter systems, stochastic structural vibrations, research projects of structural dynamics, and structural dynamics of marine
pipeline and riser.
This book was co-authored by Professor Yong Bai of Southern University of Science and Technology and Professor Zhao-Dong Xu of Southeast
University. The authors would like to appreciate Dr. Yong Bai’s and Dr. Zhao-Dong Xu’s graduate students and postdoctoral fellows who provided the
initial technical writing. The students in Southern University of Science and Technology are Ms. Xinyu Sun (Chapters 1 & 3), Mr. Jiannan Zhao
(Chapters 2 & 4), Mr. Zhao Wang (Chapters 5, 6, & 7), and Mr. Wei Qin (Chapter 8). The students in Southeast University are Mr. Yanwei Xu (Chapter
1 and proofread all), Mr. Hao Hu (Chapter 3), Mr. Yun Yang (Chapter 6), Mr. Shi Chen (Chapter 7), and Mr. Qiangqiang Li (proofread all). The
students in Xi’an University of Architecture and Technology are Mr. Zefeng He (Chapter 3), Mr. Zhenhua He (Chapter 4), and Ms. Tian Zhang
(proofread all). Thanks to all persons involved in reviewing the book.
About the Authors
Professor Bai received a doctorate from Hiroshima University in Japan and engaged in postdoctoral work in the field of ocean engineering in Technical
University of Denmark, Norwegian University of Science and Technology and University of California at Berkeley. He has published over 100 research
papers, 9 English academic treatises and 8 Chinese books on Ocean Engineering. Bai served as a professor at University of Stavanger, Harbin
Engineering University, Zhejiang University and Southern University of Science and Technology. He guided more than 50 graduate students and 30
doctoral students.
Professor Bai has a wealth of engineering experience and management skills. He worked in Det Norske Veritas, American Bureau of Shipping, JP
KENNY Company in Norway, Shell E & P Company and MCS in the United States. He has presided over dozens of large projects in the field of ship
structures, submarine pipelines and risers, design analysis and risk assessment of offshore platforms. Bai put forward the design concept of buckling
and ultimate load carrying capacity of deepwater submarine pipelines. He improved the design methods, analytical tools and design standards of
marine pipelines and reached the international leading level. He significantly improved the design methodology and criteria for subsea pipelines and
risers such as ultimate strength design, use of risk and reliability methods.
He contributed to subsea technology by publishing many papers and a recognized book entitled Subsea Engineering Handbook and promoted limit-
state design and use of risk and reliability by teaching at universities and publishing a book entitled Marine Structural Design.
Professor Zhao-Dong Xu is the professor at the Civil Engineering School of Southeast University, serving as doctoral tutor. His major research fields are
Anti-earthquake of Structures, Structural Control and Health Monitoring, Smart Material and Structures. Professor Xu got his Ph.D. in China, followed
by a series of teaching and research positions at Xi’an Jiaotong University, Ibaraki University, North Carolina State University and University of
Illinois at Urbana-Champaign. He is the Vice President of RC & PC Key Laboratory of Education Ministry. He has also been Changjiang Scholar
Distinguished Professor and the National Science Fund for Distinguished Young Scholars in China.
Professor Xu engaged in teaching and research on structural dynamics for more than 20 years. He has published more than 200 papers on the subject
of structure dynamics research, numerical analysis and application of civil engineering, etc. He has been honored with many awards—the 43rd Geneva
International Patents Exhibition Gold Award, the Second Award of National Award for Technological Invention in China, the Top Award of Chinese
Building Materials Technology Invention, etc. He has completed many significant research projects in the areas of structural vibration control and
structural health monitoring, and many research outcomes have been utilized in major real applications.
Chapter 1
Introduction
1.1 Overview of Structural Dynamics
Have you ever thought about the technology used in the Shenzhou spacecraft that we are all so proud of? Have you ever thought about what kind of
marvelous power it takes to make planes which weigh tons fly while we enjoy them? What’s the reason for the collapse of Tacoma Narrows Bridge
when it suffered 19m/s wind? Why are buildings with seismic resistance and isolation technology considered better in terms of seismic safety? All those
subjects exist in nature, and are aspects of the subject of advanced dynamics as well.
The theoretical study of dynamics began in the seventeenth century, and the publication of Analytical Mechanics by Joseph Louis Lagrange (1736–
1813) laid the foundation for the dynamic analysis in the linear system. With the development of science and technology, a variety of dynamic dives are
applied to different engineering structures, which allows the theory of structural dynamics to move forward constantly. Up to now, we can already
accomplish the dynamical analysis for huge complex structures with thousands of freedom degrees.
With regard to the design or analysis for a structure, static problems are always major areas which should be of primary concern. However, a structure
often comes to failure when critically subjected to dynamic loading. Structural dynamic analysis thus frequently plays as the control function in a
structure’s design, which may be far more critical than static load for the damages of a structure. Examples include seismic-induced structure collapse,
wind-induced failure of bridges or other long-span flexible structures, deformation of pile and destabilization of foundation under impact loads. Thus,
it’s indispensable to conduct the dynamic analysis for engineering structures’ study, design, and security evaluation. Despite the fact that numerous
pseudo-static calculation methods are adopted in some specifications for structural design and structural dynamic analysis for simplicity, such as the
response spectrum method in seismic design code or the equivalent statics wind stress which is used to substitutes the actual wind stress in wind-
resistance design, their theoretical basis is still from structure dynamics. Hence, it’s still essential to conduct the dynamic analysis in these solution
procedures, such as solving the structure’s natural period and modes in multi-degree-of-freedom system, all of which are necessary parameters
involved in response spectrum method.
Structural dynamics is a theoretical and technical subject to study the dynamic characteristics of structural systems (mainly referring to the period,
frequency, mode, and damping characteristics) and determine the dynamic responses of structures under dynamic loads (including internal force,
strain, displacement, speed, acceleration etc.). The fundamental purpose of this discipline is to provide a solid theoretical basis for improving the safety
and reliability of engineering structural systems in the dynamic environment.

1.2 Dynamic Loads

According to whether a load is time-varying or not, loads are divided into dynamic load and static load. For static load, its magnitude, direction, and
action point of static loads don’t change or change with time slightly, such as dead weight of structure, snow load, ash load, etc. On the other hand,
dynamic loads change with time. In addition, dynamic loads will also bring structure inertial forces, which cannot be neglected and must be taken into
consideration. Typical dynamic loads include simple harmonic oscillation caused by working machinery, wind loads, seismic loads, etc.

1.2.1 Simple Harmonic Loads

Simple harmonic loads are the loads varying with time harmonically and periodically, which can be represented by harmonic function, such as
P(t)=P0sinθt and Ρ(t)=P0cosθt. Analyses of structural response under the simple harmonic loads are of great importance, not only because these
dynamic loads actually exist in engineering structures (centrifugal load caused by cam axial rotation), but also any nonharmonic periodic loads can be
represented as a sum of a series of simple harmonic components. Thus, in principle, structural dynamic response caused by any periodic loading can be
translated into a superposition of responses created by a series of simple harmonic components. Furthermore, the responses of a structure under
simple harmonic loads can reflect its dynamic characteristics; therefore simple harmonic loads play a vast role in the structural dynamic analysis.

Figure 1.1 Classification of dynamic loads.

1.2.2 Nonharmonic Periodic Loads

Nonharmonic periodic loads are periodic functions of time, which vary with time periodically. They are different from simple harmonic functions.
Examples include the hydrodynamic pressure of calm waves on dams and thrust generated by a propeller of a ship.

1.2.3 Impulsive Load

Magnitude of impulsive load can increase or decrease rapidly in short duration, for example, impact load produced by explosion or blast.
1.2.4 Irregular Dynamic Load
Irregular dynamic loads are difficult to be expressed by analytic expression because of the complexity and arbitrariness of its magnitude, direction, and
position, such as earthquake action or wind load acting on the structure.
Four classifications of loads are shown in Figure 1.2.

Figure 1.2 Types of dynamic load.

1.3 Characteristics of a Dynamic Problem

A structural dynamic problem is intended to solve for the response of the structure under dynamic loads, which differs from its static loading
counterpart in the following two important aspects.
The first is the time varying nature of the dynamic problem. Because the dynamic loads vary with time, the analyst must calculate a succession of
solutions corresponding to all times in the response history when computing the dynamic responses of structure. Thus, a dynamic analysis is clearly
more complex and time-consuming than a static analysis.
Secondly, the inertia force must be considered in the dynamic problem. Compared with the static problems, the inertial force brought in by acceleration
due to rapid variation of displacement in the structural dynamic reaction will seriously influence the structural dynamic response, and the direction of
inertia force oppose the direction of acceleration.
If a simple beam is subjected to a static load, F, as shown in Figure 1.3, forces acting on the simple beam are only external force, F, and support
reactions. However, if F is a dynamic load, displacement of the beam will change rapidly. Therefore, in addition to external force F and support
reactions, there is also an inertial force distributed along the beam’s central axis acting on the simple beam. Magnitude of the inertial force depends on
motion of the beam, which is significantly influenced by the inertial forces themselves. Occurrence of inertial force makes analysis of structure
responses more complex. Especially when the loading rate gets faster, the additional responses induced by inertial force may be far bigger than the
corresponding responses caused by static force.

Figure 1.3 The difference between static and dynamic.

Dynamic calculations must be conducted in all time domain, while inertia force induced impact must be taken into consideration. Occurrence of inertial
force makes analysis more complex, but understanding and effective treatment can significantly simplify the complexity of dynamic analysis.
Structural dynamics differs from statics due to the inclusion of inertial force, in other words, taking the influence of inertia force into consideration.
Whether an engineering structure is analyzed as a dynamic problem depends on whether the load provokes a large acceleration response. If the
acceleration response is slight, the inertial forces merely represent a smaller portion of the total load equilibrated by the elastic forces of the structure,
there will be no significant distinction between effects of dynamic and static, and these kinds of problems can be solved by static structural analysis
procedures.

1.3.1 Methods of Discretization

The inertia force is the essential cause of creating structure dynamic response, which is resulted from structure mass. Thus, describing the position and
motion of the structural mass for a structure becomes the key factor in structural dynamic analysis procedure, which results in a different definition for
structural degree of freedom (DOF) in structural dynamics and statics. DOF in dynamics is defined as the number of requisite independence
parameters for ascertaining the spatial mass locations of system in dynamical analysis. These independence parameters are also called generalized
coordinate, which can be displacement, deflection, or other generalized quantities. Generalized coordinate and dynamic DOF will be presented in
Chapter 2.
Selecting a correct dynamic freedom is of great importance to describe inertial force of system accurately. All structural systems have a distributed
mass, and all of them have an infinite number of degrees of freedom. However, it is extremely difficult to conduct calculations on an engineering
structure with infinite DOFs, except some simple structures can be dealt with like a construction with infinite DOFs. In structural dynamics analysis,
structures are usually simplified as a finite DOFs system to avoid the mathematical complexity in the procedure of calculation, which is called the
procedure of discretization for a structure. The common discretization methods include lumped mass procedure, generalized displacements procedure,
and finite element procedure. By all accounts, discretization method is the way of transforming an infinite DOFs problem into a finite DOFs one.

1.3.2 Lumped Mass Procedure

Lumped mass procedure is the most commonly used in structural dynamic analysis, which centralize the quality of the actual structure in some
geometric points according to some certain rules, while the structural bars are massless. In this way, an infinite DOFs system will be translated into a
finite DOFs one.
Figure 1.4 shows two examples of converting an infinite DOFs system into a finite degree of freedom system by the lumped mass procedure for the
structures with continuous distribution mass. Figure 1.4a shows the procedure of translating a simple beam into a finite degree of freedom with three
masses by lumping its continuous distribution mass at three discrete points on the beam, that is, replacing the continuous distribution mass with
lumped masses m1 , m2, and m3. If taking the transverse movement in the plane of beam into consideration, the lumped mass beam has three
transverse degrees of freedom. Figure 1.4b shows a plane frame structure with three stories, if we centralize half mass of columns, walls, floor, beam of
every contiguous stories at the center of their floor. The three stories frame structure will be simplified as a finite DOFs system with three lumped
mass points.
Figure 1.4 Schematics of lumped mass procedure.

1.3.3 Generalized Coordinate Procedure

Generalized coordinates are the independent parameters which can determine the geometric position of the system. It can be adopted in the beam
system with the distribution mass to improve calculation accuracy. Series method is normally used to solve differential equations in the field of
mathematics, which can also be used to solve problems in structure dynamic analyses. As shown in Figure 1.5, the deformation curve of a sample beam
with distribution mass can be expressed as a sum of the Fourier series.

Figure 1.5 Discretization of a simple beam by generalized coordinate procedure.

(1.1)

Where L is the length of beam; sin iπx/L is shape functions, it is a series of given functions that satisfy the boundary condition; bi = bi(t) represent
generalized coordinates, it is a set of parameters to be determined, and it is a function of time in dynamics.
Because the shape function is a predefined and ascertained function, the deformation of the beam will be determined by multiple generalized
coordinates, this makes the simple beam have infinite degrees of freedom in the theoretical dynamic analysis. The same approaches are taken as in
mathematics, only the first few items of the series will be brought into the actual analysis. Its first N items are taken in the following example.
(1.2)

Thus, a simple beam is simplified as an N degrees of freedom system. A more generalized expression of structure displacement may be written as:
(1.3)

Where qi(t) is amplitude of shape function, that is, generalized coordinate; φ i(x) is shape function and it is a continuous function that satisfies the
boundary condition, take the simple beam shown in Figure 1.5 for instance, the boundary condition that shape function must satisfy is φ n(0) = φ n(L) =
0.
If the shape function represents displacement, the generalized coordinate will have the dimension of displacement. The generalized coordinates can
represent the amplitude of the shape function, but it is not the actual physical displacement, because it cannot convey the real displacement unless all
the N items are added.

1.3.4 Finite Element Method

As with the finite element method (FEM) used in the static problem, structural dynamics problems can also be solved by the FEM which combines
certain features of both the lumped mass and the generalized coordinate procedures. The procedure of FEM applied in dynamics analysis uses the
displacements of structure nodal points to represent the displacement state of each point in the structure. Firstly, the structure would be divided into a
series of units, and these units are connected with nodal points, then the displacements of nodal points are independent coordinates which determine
the displacements of all mass point in structure.
In discretization procedure of FEM, the assumption deflections of elements are added to obtain the deflection curve of the entire beam at a certain
time. These deflection curves used to describe the shape of each units are described as displacement function or interpolation function, whose
expression contains several parameters. These displacement functions should be smoothly continuous inside each element and satisfy the support and
deformation continuity conditions at two ends of each element. According to these conditions, the parameters in the displacement function can be
expressed by the nodal displacement. Thus, the entire structure system is transformed into a finite DOF system with unknow nodal displacement.
Take a continuous beam, for instance; it can be divided into N elements (beam segments), the intersections of interconnected elements are called nodal
points, and the parameters of nodal displacement (displacement u and rotation angle θ) are taken as generalized coordinates. This finite element model
has six generalized coordinates (displacement parameters): u1 , θ1 , u2, θ2, и3, θ3, the corresponding shape function is φ 1, φ 2, φ 3, φ 4, φ 5, φ 6.
For a cantilever beam discretization model with N elements, there are 2N generalized coordinates in total, and the displacement of the beam can be
described by 2N generalized coordinates and their shape functions, whose expression is shown below.
(1.4)

In this way, an infinite DOF beam can be transformed into a system with 2N finite DOF.
The finite element procedure combines the characteristics of both the lumped mass and the generalized coordinate procedure:
1. Similar to the generalized coordinate procedure, the finite element procedure also uses the concept of shape function. Different from the
interpolation (definition of shape function) in entire system (structure) in the generalized coordinate procedure, the finite element procedure
adopts the piecewise interpolation method (definition of piecewise shape function), so the expression (shape) of the shape function can be relatively
simpler.
2. Compared with the lumped mass procedure, the generalized coordinates of the FEM have real physical means too, the same as the lumped mass
procedure, which has the advantages of being direct and intuitive.
Among the above three procedures, the lumped mass procedure is relatively simpler and more practical, the generalized coordinate procedure needs to
select suitable shape function to satisfy the displacement boundary condition, which is suitable for simple structures merely. The FEM combines the
characteristics of the lumped mass procedure and generalized coordinate procedure, which is suitable for various complex structures and widely used
to solve the dynamic problems of engineering structure indeed.

1.4 Application of Structural Dynamics

1.4.1 Application of Structural Dynamics in Civil Engineering
When a building structure is subjected to earthquake, wind, vibration and other loads, the magnitude, direction and action point of such loads not only
change with time, but also change greatly, which makes the load response of the structure very different from the static load. Therefore, this kind of
load is defined as dynamic load. The essential difference between structural statics and dynamics is that the static load acts on the building structure to
produce static internal force, deformation and other responses; the dynamic load acts on the building structure, the vibration makes it produce
accelerated motion, the accelerated motion gradually attenuates due to damping effect. The dynamic load not only makes the structure deform; the
inertia force caused by acceleration also makes the structure respond to internal force, deformation, motion, displacement, acceleration, and so on.
Building structure is a space system that can be connected by several components. Building structures can be divided into concrete structures,
masonry structures, steel structures, wood structures and composite structures according to the building materials used.
Combined with functional characteristics, appearance design, planning and layout requirements, building structures are usually irregular in plane
shape. When a plane irregular building structure is under stress, some irregular parts interact with each other, which makes the whole structure
torsion. At the junction of irregular parts, stress concentration and deformation concentration are easy to occur. Since the wind load is less than the
earthquake action, it is necessary to analyze the regularity lateral resistance of building structure plane according to the earthquake action, earthquake
action + wind load action, dead load + live load + accidental load, indoor physical simulation test (jack test, shaking table test).

1.4.2 Application of Structural Dynamics in Ocean Engineering

With the increasing demand for offshore oil and mineral resources, ocean engineering has developed rapidly in recent years. A large number of topics
related to ocean engineering science need to be studied and solved. Marine structural dynamics mainly studies the interaction between waves and
marine structures, the interaction between structures and seabed foundations. A reasonable and practical mathematical model, theory and method for
structural analysis are proposed to solve the problems of structural vibration response, strength calculation and failure mechanism. A reasonable
structural form is proposed for engineering design and construction. It provides the necessary theoretical basis and design parameters. Marine
structural dynamics is an interdisciplinary subject which integrates structural dynamics, soil dynamics, geology, oceanography, mathematical statistics,
computational methods, experimental techniques and other disciplines. It is a research direction with a wide range of disciplines and development
prospects.
The marine environment is very harsh, so human understanding of the ocean is very limited. Most of the offshore structures are submerged in sea
water. The generation of structural load and the motion of the structure are closely related to the surrounding fluid medium.
The dynamic loads of ocean structures can be divided into two categories: natural loads and artificial loads. According to the causes, natural loads can
be divided into fluid loads, wind loads, earthquake loads and floating ice loads. Man-made loads refer to the dynamic loads generated in various
production activities (including ship and machine impacts, drilling operations, machinery, crane operations, etc.). These loads are mainly fluid loads,
which are to be discussed in this section. Earthquake and floating ice loads should be considered in some areas, which will be described separately
below.
Fluid loads are caused by waves, currents, eddy detachment, overshoot vibration, breaking and beating of waves, among which wave action is the main
one. Waves are caused by wind blowing on the water surface and will continue to exist after wind stops. The shape of the wave surface is close to the
ideal sinusoidal curve in deep water, and its amplitude is small. It tends to be in the same cycle in shallow water. The asymmetry of the wave surface
increases with the increase of the steepness.
There are two main factors indicating the wave characteristics, the period and the wave height. Degree propagates unidirectionally forward, but water
points in steady waves move regularly at lower speeds. In a wave cycle, a water point completes a near-closed orbit with a vertical diameter equal to
that height. In deep water, the trajectory of a water surface particle is close to a circle with a diameter varying with depth. When the depth exceeds one
wavelength, the water point becomes almost static; in shallow water, the orbit of the surface particle is approximately elliptical, its long axis is parallel
to the seabed, and its short axis decreases with the depth until the seabed has only horizontal motion. Waveform is caused by half of kinetic energy and
comes from water point velocity. The energy density distribution with frequency is called spectrum, which will be used for spectral analysis of
structures.
Wave parameters are determined by wind speed, wind time and wind zone. Actual ocean waves are superimposed by waves of different periods and
wave heights, which are caused by the changing size, range and direction. Therefore, ocean wind waves generally have broad spectrum characteristics
and are suitable for spectral analysis. When the waves move to the generating area or when the wind stops. Surge propagates very far and keeps a
certain shape over a long distance, and the wave with longer period has more components and narrow spectrum characteristics.
Tides and winds produce currents at estuaries. Currents play a significant role below the surface of waves (about a wavelength area). Directional
currents cause static loads and pulsating currents (turbulence) cause dynamic loads. Turbulence frequencies are low and have little effect on the
dynamic analysis of buildings. Both of them can be derived from the sea according to the formulas of resistance and inertia force.
The vortices are separated alternately from one side of the component in a steady flow, resulting in a vibration force perpendicular to the flow
direction-lift and upward pulsating resistance. The lift frequency is equal to the frequency of a pair of vortices detachment. The main frequency of the
pulsating resistance is twice that of the lift frequency. The vortex detachment frequency, dimensionless lift coefficient and fluctuating drag coefficient
depend on the section shape, relative flow direction and Reynolds number of the component.
Over-ride vibration is another form of vibration that occurs in a structure in an ocean current. It is a self-excited vibration in which the force induced
by the motion of the body is in its direction of motion. Thus, when the work of the fluid is done on the body, unstable vibration occurs when the input of
the work is greater than the energy dissipated by the damping of the structure. In the case of directional jets, if the axis of the structural member is
parallel to the sea surface, the component will alternately sink or expose to the water surface when the wave passes. The alternating impact loads can
also cause the structure to be excited. Under the conditions of deep water and storm, the wave often reaches the ultimate steepness and breaks up,
thus affecting the analysis of wave forces. Air entrainment will further complicate the problem. Direct breakage of waves on buildings may greatly
exceed the wave forces normally calculated. These problems are still limited to qualitative simulation tests, and their mechanisms need to be further
clarified.
Earthquake and floating ice are special loads encountered by marine structures, which can be ignored under normal circumstances, but structures
themselves in seismic or floating ice zones will become an important factor in the design load of marine structures. Load also has stochastic
characteristics; it is due to the ground motion acceleration caused by the whole structure mass inertia force caused by dynamic response.
Fatigue is an important problem in the analysis of marine structures. The reason is that the cyclic loads on marine structures are much more frequent
than those on land. Under long-term cyclic loads, the cracks and their propagating members can also cause serious damage under low stress conditions.
This is in accordance with the above maximum control. The first step is to calculate the stress response frequency, which is the combination of the
natural frequency and wave frequency of the structure. It is often much higher than the wave frequency. In addition, it should be considered that the
general local impact force can cause the movement of the component with its natural frequency. New attention is paid to the combinations of different
loads. Some loads cause little stress amplitude, but the frequency is higher.
There are also construction and installation dynamics problems for offshore structures. These problems are different from those under normal
production conditions and require special consideration, including the dynamic response of floating stability, towing, throwing and sinking components
in the course of transportation, the driving analysis of pipe piles, the local stress of caisson installation on non-uniform foundation, etc. In addition, there
are a series of problems, such as foundation stability, soil liquefaction, material dynamic performance, dynamic stress concentration, dynamic plasticity,
fracture, temperature difference, tsunami, landslide, uncertainty analysis and vibration tolerance standards.

1.4.3 Application of Structural Dynamics in Aircraft Technology

Aircraft structures are often subjected to a variety of dynamic loads in use, resulting in structural dynamic strength failure. With the improvement of
aircraft performance, it is more and more important to analyze the types and intensity of these vibration excitations and impacts. The degree of
damage is not only related to the motion state and environment of the aircraft, but also closely related to the dynamic characteristics of the aircraft
structure itself. Therefore, the dynamic strength of aircraft structure has become a very important issue in the process of aircraft design and
decommissioning.
In recent years, with the development of aircraft structural strength analysis and test technology, structural dynamics has made great breakthroughs
in the application field of aircraft design, and its technical level has been continuously improved and improved. Some remarkable achievements and
progress have been made in many aspects, such as aircraft complex structure dynamics analysis, dynamic test simulation and analysis technology,
large structure vibration test design technology, formulation and application of mechanical environment conditions of airborne equipment, composite
structure dynamics research, aircraft noise prediction and acoustic fatigue life analysis.
However, up to now, structural dynamics is still a developing subject, so there are inevitably many problems to be further studied and discussed in the
design of structural dynamics. For example, due to the complexity of vibration sources, the dynamic loads cannot be accurately described; because of
the difficulty of simulating boundary conditions, the dynamic response analysis of large complex structures is not accurate enough; the damping
characteristics of complex structures are still difficult to determine. Structural vibration induced fatigue failure mechanism, vibration induced fatigue
life prediction method, dynamic strength criterion and criterion, etc. have become the key technical problems of future aircraft design and
development.
In the field of structural dynamic strength technology, some innovative engineering applications are brewing, mainly including:
(1) Dynamics design technology of structures under dynamic loads, which is penetrating into modern design technology.
(2) Structural adaptive vibration control or active control technology, which is developing and crossing between structural dynamics technology
and modern control technology.
(3) Structural vibration induced fatigue and dynamic fracture technology formed by the combination of structural dynamics technology and
traditional fatigue technology.
The development of these new technologies will greatly enrich and expand the theory of structural dynamics, and the application of these theories and
new technologies to guide and solve the key problems of structural dynamics in aircraft design is an important task and goal of future aircraft structural
dynamic strength work.

Exercises
1.1 What is the purpose of structural dynamics analysis? What are the characteristics of structural dynamics?
1.2 Express the types and characteristics of dynamic loads.
1.3 What are the procedures of structural discretization? What are their characteristics?

References
1. Zhao-Dong, X. and Le-Wei, M., Dynamics of Structures, Science Press, Beijing, 2007.
2. Clough, R.W. and Penzien, J., Dynamics of structures, in: Dynamics of Structures, 2005.
3. Chopra, AK., Dynamics of Structures, Prentice Hall, 2000.
4. Wilson, J.F., Dynamics of Offshore Structures, 1984.
5. Forment, D. and Welaratna, S., Structural Dynamics Modification—An Extension to Modal Analysis. Sae Paper, 1980.
6. Herrada, F.J., García-Martínez, J., Fraile, A., Hermanns, L.K.H., Montáns., F.J., A method for performing efficient parametric dynamic analyses in
large finite element models undergoing structural modifications. Eng. Struct., 131, 2017.
7. Kawabata, N. and Fukunaga, H., Identification of structural dynamic response based on the design of modal sensor. Proc. 1992 Ann. Meet.
JSME/MMD, 2003, 0, 2003.
8. Harte, M., Basu, B., Nielsen, S.R.K., Dynamic analysis of wind turbines including soil-structure interaction. Eng. Struct., 45, 2012.
9. Pasha, H.G., Kohli, K., Allemang, R.J. et al., Structural dynamics model calibration and validation of a rectangular steel plate structure, in: Model
Validation and Uncertainty Quantification, vol. 3, Springer International Publishing, 2015.
10. Ozcelik, O., Attar, P.J., Altan, M.C. et al., Experimental and numerical characterization of the structural dynamics of flapping beams. J. Sound Vib.,
332, 21, 5393–5416, 2013.
11. Herbert, M.R. and Kientzy, D.W., Applications of Structural Dynamics Modification, 1980.
Chapter 2
Establishment of the Structural Equation of Motion
The purpose of structural dynamics analysis is to calculate the dynamic response of the structure under dynamic load, that is, to solve the history of
displacement, velocity, acceleration, strain, etc., of the structure over time. In most cases, applying an approximate analysis method with a finite
number of degrees of freedom is accurate enough. In this way, the problem becomes to find the time-history curve of the selected component. Before
solving the time-history curve, the equation of motion of the dynamical system under dynamic load must be established. This chapter will briefly
introduce some of the basic concepts of structural dynamics and the methods for establishing the structural equations of motion.

2.1 General
The degree of freedom is often talked about in structural dynamics, and it is necessary to be familiar with the concept of particles before describing
degrees of freedom. Particles are ideal models for simplifying objects. The model is considered as objects which have only mass and no size.

2.1.1 Dynamic Freedom

The number of independent geometric parameters required to describe the position of the system at any moment of mass during motion is called the
number of degrees of dynamic freedom of the structure. The number of structural degrees of freedom is not fixed, changes as the structural calculation
hypothesis changes for a structure. As shown in Figure 2.1, the single mass is depicted in Figure 2.1(a), and the mass which has two degrees of freedom
can move in x-axis and y-axis; cantilever beam ignoring axial effect is shown in Figure 2.1(b), where the right mass can move in x-axis, and the left one
can move in y-axis; a rigid beam is shown in Figure 2.1(c), if the stiffness is assumed to be infinite, whose three masses has single degree of freedom,
namely angle of rotation θ; the sketch of four-layered frame is shown in Figure 2.1(d), each mass can move in horizontal direction and the structure has
four degrees of freedom.

Figure 2.1 Definition of degree of freedom.

When analyzing a dynamic system, the first step is to determine the degree of power freedom and establish the differential equation of motion. Before
describing how to establish the differential equation of motion of a dynamic system, it is necessary to understand the basic components of the dynamic
system.

2.1.2 Basics of Dynamic System

A dynamic system is a simple representation of physical systems and is modeled by mass, damping and stiffness. For any system subjected to a
dynamic load, the main physical characteristics are the mass of the system, elastic recovery characteristics, energy dissipation characteristics or
damping, and the external disturbance or load of the system.
Inertia Force
Mass is a fundamental property of matter and is present in all physical systems. This is simply the weight of the structure divided by the gravity
acceleration. Mass contributes an inertia force (equal to mass times acceleration) in the dynamic equation of motion, which can be expressed as,
(2.1)

where, F I(t) represents the inertia force; m represents the mass; ü(t) is the acceleration.

Elastic Restoring Force

Stiffness makes the structure more rigid, lessens the dynamic effects and makes it more dependent on static forces and displacements. Usually,
structural systems are made stiffer by increasing the cross-sectional dimension, making the structures shorter or using stiffer materials. Stiffness is the
resistance it provides to deformations, mass is the matter it contains and damping represents its ability to decrease its own motion with time.
Assuming that the relationship between the force and displacement is linear, the restoring force of the spring is also referred to as the elastic restoring
force, which is equal to the product of the spring stiffness and displacement,
(2.2)

where, F S(t) represents the restoring force; k represents the stiffness of the spring; u(t) is the displacement.

Damping Force
Damping, in physics, is restraining of vibratory motion, such as mechanical oscillations, noise, and alternating electric currents, by dissipation of energy.
Unless a child keeps pumping a swing, its motion dies down because of damping. Shock absorbers in automobiles and carpet pads are examples of
damping devices. Whereas the mass and the stiffness are well-known properties and measured easily, damping is usually determined from
experimental results or values assumed from experience. There are several sources of damping in a dynamic system. Viscous damping is the most
used damping system and provides a force directly proportional to the structural velocity. This is a fair representation of structural damping in many
cases and for the purpose of analysis, and it is convenient to assume viscous damping (also known as linear viscous damping). For the single degree of
freedom system, the viscous damping can be written as,
(2.3)

where, F D(t) represents the damping force; c represents the damping coefficient; is the velocity of mass.
Viscous damping is caused by such energy losses as occur in liquid lubrication between moving parts or in a fluid forced through a small opening by a
piston, as in automobile shock absorbers. The viscous-damping force is directly proportional to the relative velocity between the two ends of the
damping device. Viscous damping is usually an intrinsic property of the material and originates from internal resistance to motion between different
layers within the material itself.
The motion of a vibrating body is also checked by its friction with the gas or liquid through which it moves. The damping force of the fluid in this case is
directly proportional to a quantity slightly less than the square of the body’s velocity and, hence, is referred to as velocity-squared damping. Besides
these external kinds of damping, there is energy loss within the moving structure itself that is called hysteresis damping or, sometimes, structural
damping. In hysteresis damping, some of the energy involved in the repetitive internal deformation and restoration to original shape is dissipated in
the form of random vibrations of the crystal lattice in solids and random kinetic energy of the molecules in a fluid.
There are other types of damping. Resonant electric circuits, in which an alternating current is surging back and forth, as in a radio or television
receiver, are damped by electric resistance. The signal to which the receiver is tuned supplies energy synchronously to maintain resonance. In
radiation damping, vibrating energy of moving charges, such as electrons, is converted to electromagnetic energy and is emitted in the form of radio
waves or infrared or visible light. In magnetic damping, energy of motion is converted to heat by way of electric eddy currents induced in either a coil
or an aluminum plate (attached to the oscillating object) that passes between the poles of a magnet.
Damping is the dissipation of energy from a vibrating structure. In this context, the term dissipate is used to mean the transformation of energy into
the other form of energy and, therefore, a removal of energy from the vibrating system. The type of energy into which the mechanical energy is
transformed is dependent on the system and the physical mechanism that cause the dissipation. For most vibrating system, a significant part of the
energy is converted into heat.

2.2 Formulation of the Equations of Motion

As mentioned earlier, the primary objective of a deterministic structural dynamic analysis is the evaluation of the displacement time histories of a given
structure subjected to a given time varying loading. In most cases, an approximate analysis involving only a limited number of degrees of freedom will
provide sufficient accuracy; thus, the problem can be reduced to the determination of the time histories of these selected displacement components.
The mathematical expressions defining the dynamic displacements are called the equations of motion of the structure, and the solution of these
equations of motion provides the required displacement time histories.
The formulation of the equations of motion of a dynamic system is possibly the most important, and sometimes the most difficult, phase of the entire
analysis procedure. In this text, three different methods will be employed for the formulation of these equations, each having advantages in the study
of special classes of problems. The fundamental concepts associated with each of these methods are described in the following paragraphs.

2.2.1 Direct Equilibration Using D’Alembert’s Principle

D’Alembert’s principle, an alternative form of Newton’s second law of motion, was stated by the seventeenth-century French mathematician Jean le
Rond d’Alembert. In effect, the principle transforms a problem in dynamics to a problem in statics. According to the second law, it states that the force
F acting on a body is equal to the product of the mass m and acceleration ü of the body,
(2.4)

Equation (2.4), indicating that force is equal to the product of mass and acceleration, may also be written in the form
(2.5)

In other words, the body is in equilibrium under the action of the real force F and the fictitious force –mü. The fictitious force is also called inertial force
and reversed effective force.
Example 2.1 The dynamic system is shown in Figure 2.2, where the mass m is applied with the force P(t). Try to derive an equation of motion for
the system.
Figure 2.2 Single body with SDOF dynamic system.
Solution: The body only moves along a horizontal direction. Therefore, the system is the single degree of freedom system. Assume the displacement
of the body is u(t), the spring restoring force is F s(t)=ku(t), and the damping force is . The equilibrium equation under the external force P(t)
can be written as,

Further,

Example 2.2 The dynamic system is shown in Figure 2.3, which is the dual body dynamic system. Try to derive an equation of motion for the
system.

Figure 2.3 Dual body with 2DOF dynamic system.

Solution: The above system includes two particles with horizontal movement. These two particles can be analyzed separately as shown in Figure 2.3.
For the mass m1 , the real forces include the left spring force –k 1 u1 , the middle spring force k 2(u2-u1 ), and an external force P1 (t). For the mass m2, the
real forces include the middle spring force k 2(u2-u1 ), the right spring force –k 3u2, and an external force P2(t).
Further, the equilibrium equations for two particles can be written as,

The matrix form can also be given as,

where,
M is the mass matrix
K is the stiffness matrix
P is the vector of the external force
u is the vector of the body displacement
D’Alembert’s principle is a very convenient device in problems of structural dynamics because it permits the equations of motion to be expressed as
dynamic equilibrium. The force F may be considered to include many types of force acting on the mass: elastic constraints which oppose displacements,
viscous forces which resist velocities, and independently defined external loads. Thus, if an inertial force which resists acceleration is introduced, the
equation of motion is merely an expression of equilibration of all forces acting on the mass. In many simple problems, the most direct and convenient
way of formulating the equations of motion is by means of such direct equilibrations.

2.2.2 Principle of Virtual Displacements

However, if the structural system is reasonably complex involving a number of interconnected mass points or bodies of finite size, the direct
equilibration of all the forces acting in the system may be difficult. Frequently, the various forces involved may readily be expressed in terms of the
displacement degrees of freedom, but their equilibrium relationships may be obscure. In this case, the principle of virtual displacements can be used to
formulate the equations of motion as a substitute for the direct equilibrium relationships. The principle states that if a system of mass particles which is
in equilibrium under the action of a set of externally applied forces is subjected to any virtual displacement, i.e., a displacement pattern compatible with
the system’s constraints, the total work done by the set of forces will be zero.
(2.6)

where
i is an integer used to indicate (via subscript) a variable corresponding to a particular particle in the system
F i is the total applied force (excluding constraint forces) on the i-th particle
mi is the mass of the i-th particle
üi is the acceleration of the i-th particle
miüi together as product represents the time derivative of the momentum of the i-th particle
δui is the virtual displacement of the i-th particle, consistent with the constraints
With this principle, it is clear that the vanishing of the work done during a virtual displacement is equivalent to a statement of equilibrium. Thus, the
response equations of a dynamic system can be established by first identifying all the forces acting on the masses of the system, including inertial
forces. Then, the equations of motion are obtained by separately introducing a virtual displacement pattern corresponding to each degree of freedom
and equating the work done to zero.
Example 2.3 The inverted pendulum system is shown in Figure 2.4. Try to establish the equation of motion for this system using virtual
displacement principle.

Figure 2.4 Inverted Pendulum system with SDOF.

Solution: In the tangential direction of the mass, we have the following forces,
Spring force
Inertia force
External force
The virtual displacement can be defined as shown in Figure 2.5,
Figure 2.5 Inverted Pendulum system with 1DOF.
According to the virtual displacement principle, the principle of virtual work can be derived as,

Substituting the virtual displacement equation into Equation 2.6 and cancelling out δφ 1 the following equation of motion is obtained,

The advantage of the principle of virtual displacement is that the virtual work contributions are scalar quantities and can be added algebraically,
whereas the forces acting on the structure are vectorial and can only be superposed vectorially.

2.2.3 Hamilton’s Principle

Hamilton’s principle is one of the most fundamental and important principles of mechanics and mathematical physics. For the non-conservative system
of n particles, based on d’Alembert’s principle which states that,
(2.7)

Note that in this principle the knowledge of force, whether it is conservative or non-conservative, and also the requirement of holonomic or non-
holonomic constraints, does not arise. We write the principle in the form,
(2.8)

Now consider
(2.9)

Since we have , therefore, we write

(2.10)

where T is the kinetic energy of the system

(2.11)

Substituting this in Equation 2.10, we get

(2.12)

Further
(2.13)

Integrating the above equation with respect to time interval between t 0 to t 1

(2.14)
Since, there is no variation in co-ordinates along any paths at the end points. i.e., . Hence from above equation we have
(2.15)

This is known as Hamilton’s principle for non-conservative systems. If the system is conservative, then the forces are derivable from potential.
In this case, it is helpful to separate the force vector F(t) into its conservative and non-conservative components as represented by
(2.16)

A potential energy function, V(t), is then defined such that the conservative force vector Fc(t), by definition, must satisfy the component relations
(2.17)

Hence equation 2.15 becomes

(2.18)

Upon considering a summation of equations of this type for all mass particles, it becomes apparent that equation 2.18 is also valid for any complicated
system, linear or nonlinear, provided quantities T(t), V(t), and Wnc(t) represent the summation of such quantities for the entire system.
Equation (2.18), which is generally known as Hamilton’s variational statement of dynamics, shows that the sum of the time variations of the difference
in kinetic and potential energies and the work done by the non-conservative forces over any time interval t 1 to t 2 equals zero. The application of this
principle leads directly to the equations of motion for any given system.
Example 2.4 Use Hamilton’s principle to find the equations of motion of a particle of unit moving as shown below.

Figure 2.6 Single body with 1DOF dynamic system.

Solution: The kinetic and potential energy of the particle is given by

The work done by the nonconservative force can be given by

The Hamilton’s principle states that

Further,

where

Since , and δu is independent and arbitrary, then we have

Finally,

In summary, the specific steps for establishing the equation of motion using Hamilton’s principle are:
1) Explicitly study objects, analyze constraints, determine system’s degree of freedom and choose appropriate coordinates
2) Calculate the kinetic and potential energy of the system.
3) Calculate the sum of virtual work done by non-conservative forces
4) Hamiltonian principle equation of the generation to obtain the differential equation of motion.
The advantage of the Hamilton principle is that inertial forces and elastic forces are not obviously used, and they are replaced by variations on kinetic
energy and potential energy, respectively. Therefore, for both of these terms, only scalar processing, i.e., energy, is involved. In virtual displacement
theory, the virtual work itself is scalar, but the force and displacement used to compute virtual work are all vectors.
The above variational procedure differs from the virtual work procedure used previously in that the external load as well as the inertial and elastic
forces are not explicitly involved; the variations of the kinetic and potential energy terms, respectively, are utilized instead. It has the advantage of
dealing only with purely scalar energy quantities, whereas the forces and displacements used to represent corresponding effects in the virtual work
procedure are all vectorial in character even though the work terms themselves are scalar.

2.2.4 Lagrange’s Equations

The equations of motion for an MDOF system can be derived directly from Hamilton’s equation by simply expressing the total kinetic energy T, the
total potential energy V, and the total virtual work Wnc in terms of a set of generalized coordinates, q1 , q2, …, qN.
For most mechanical or structural systems, the kinetic energy can be expressed in terms of the generalized coordinates and their first-time
derivatives, and the potential energy can be expressed in terms of the generalized coordinates alone. In addition, the virtual work which is performed
by the nonconservative forces as they act through the virtual displacements caused by an arbitrary set of variations in the generalized coordinates can
be expressed as a linear function of those variations. In mathematical terms the above three statements are expressed in the form
(2.19)

(2.20)

(2.21)

where the coefficients Q1 , Q2, … QN are the generalized forcing functions corresponding to the coordinates q1 , q2, …, qN, respectively.
Integrating the above equations with respect to t between t 0 to t 1 we get
(2.22)

Integrating the velocity-dependent terms in equation 2.22 by parts leads to

(2.23)

The first term on the right-hand side of equation 2.23 is equal to zero for each coordinate since δqi(t 0) =δqi(t 1 ) = 0 is the basic condition imposed upon
the variations.
Substituting equation 2.23 into equation 2.22 gives, after rearranging terms,
(2.24)

Since all variations δqi(i=1; 2; …, N) are arbitrary, equation 2.24 can be satisfied in general only when the term in brackets vanishes, i.e.,
(2.25)

Equations (2.25) are the well-known Lagrange’s equations of motion, which have found widespread application in various fields of science and
engineering.
It can be seen that the steps for establishing the equation of motion using the Lagrange’s equations are:
1) Identify the system and figure out what objects are included in the system under consideration. Then determine the degree of freedom and
choose a proper set of generalized coordinates.
2) Find the kinetic and potential energy of the system
3) Find non-conservative forces in the system
4) Substitute each of the above quantities into the Lagrange equations, the equations of motion are obtained by simplification.
The above describes four basic methods for establishing the equation of motion for a dynamic system. The direct equilibration method is a simple and
intuitive way to establish the equation of motion, which has been widely used. More importantly, the direct method establishes the concept of dynamic
balance. The method of establishing equilibrium equations in the structural static analysis is directly extended to dynamic problems. If the structure
has distributed mass and elasticity, it may be difficult to directly use the dynamic balance method to establish the equation of motion of the system. At
this time, it may be more convenient to use the virtual displacement principle. It partially avoids the vector operation, the scalar operation can be used
to establish the system’s equation of motion after obtaining the system virtual work. The Hamilton’s principle is another scalar method for establishing
equations of motion. Based on energy method, if the work (mainly the damping force) of non-conservative forces is not considered, it is a complete
scalar operation. The Lagrange’s equation used to establish the equations of motion is more common. It is the same as the Hamilton’s principle. For
damping force, it is obtained through experimental tests in practice, rather than complete mathematical and mechanical analysis of structural materials
and components. The mechanical damping of continuum medium is mainly caused by the medium itself, and the structural dynamics damping is
different. The source for structural damping is more complex, it cannot be simply derived, but the experimental coefficient is given by actual
measurement or experience.

2.3 Theory of Total Potential Energy Invariant Value of Elastic System Dynamics
2.3.1 The Main Idea of the Principle of Virtual Work
In order to clarify the principle of the total potential energy invariance of the elastic system dynamics, the idea and physical concept of the principle of
virtual work and the principle of static potential energy constant value are first introduced.
The principle of virtual work is derived from the sum of the forces of the balance force on the infinitesimal virtual displacement of the system equal to
zero or derived from the principle of conservation of energy. The virtual displacement is an arbitrary small displacement that people imagine to satisfy
the system deformation coordination condition (constraint condition), and has nothing to do with the actual force of the system, so it is called virtual
displacement. Let a particle maintain equilibrium under any force system. If there is any displacement for any reason, the total work done by this force
is equal to zero. This is the virtual work theory of the particle. Virtual work, that is, one of the force system and the displacement is dummy, and each
of them is independent, which is the basic concept of virtual work theory. Multiply the three equilibrium equations of elastic mechanics by the virtual
displacement. After mathematical demonstration, the following virtual work equations are derived.
(2.26)

The left side of the equation (2.26) is δW, the sum of the external force φ and the virtual force X of the physical force. The right side is the virtual strain
energy δUi of the system. Therefore, the equation (2.26) is abbreviated as
(2.27)

The negative value of force work W is its potential energy Ue, so, SW = -SUe it is substituted (2.27), which gives a more concise expression of the
principle of virtual work.
(2.28)

Equation (2.28) is called the system total potential energy (also known as potential energy) standing value principle, where U = Ui + Ue is the total
potential energy of the system and Ue is the potential energy of the external force (potential energy). When changing the total potential energy of the
system, it is necessary to maintain the essence of the virtual work principle of (2.28), the external force and the stress remain unchanged, only the
displacement and strain variation, let δ in equation (2.28) always maintain the effect of δ on displacement and strain variation in equation (2.26). It is
not possible to express the virtual work principle (2.26) as the total potential energy standing value principle (2.28), which will result in the meaning of
the variable semicolon δ deviates from the original meaning of the virtual displacement and the virtual strain in the equation (2.26), but δ is also
thought as a mathematical variational symbol for total potential energy U.
According to the above discussion, two points can be summarized.
Virtual displacement is an arbitrarily small deformation coordination displacement in an imaginary equilibrium system, is a possible trend,
independent of external forces and stresses in the system; it is not the actual displacement, and does not destroy the balance of the force system.
Therefore, the external force and the internal force do not change during the virtual displacement process;
Move the virtual displacement in equation (2.26) and δ before the virtual strain symbol to the integral number to obtain the principle of total
potential energy standing value of the elastic system. This movement of δ in equation (2.28) does not cause changes in external force and stress of
the system; so for the first-order variation of the total potential energy U of the system, only the displacement u and the strain ε vary, and the
external force and the stress are not strained, so as to ensure that the equation (2.28) returns to the equation (2.26). This just reflects the
requirements of the principle of virtual work.

2.3.2 Derivation of the Principle of Total Potential Energy Invariant

After introducing D’Alembert’s principle and considering the damping force, the elastic system dynamics problem is transformed into the dynamic
equilibrium problem. The general form of the equilibrium equation is
(2.29)

In the formula
fs——System elastic force vector array
——System inertial force vector array

——System viscous damping force vector array

F——System Coulomb Friction Vector Array

P(t)——System interference force vector array
Q——System gravity vector array
At a given instant t, the two sides of the equation (2.29) are multiplied by the virtual displacement δu of the system, and the elastic force δuTfs is taken
as the variational δUi of the virtual strain energy Ui of the system, and we can obtain
(2.30)

Equation (2.30) is the general expression of the principle of virtual system dynamics. It can be reduced to the following simple form
(2.31)

In the equation
Ui——System elastic strain energy
——System inertial force to do negative work

——System viscous damping force to do negative work

——System Coulomb friction work negative value

VP = -uT P(t)——System interference force does negative work
Vg = -uT Q——System gravity potential energy
Because the virtual displacement is applied to the balance system, the time is instantaneously fixed, and all the forces acting on the system are
unchanged. Their work is only related to the starting position and the ending position of the displacement, and is independent of the moving path of the
force acting point. The definition, so they can all be seen as powerful. Thus, Vm, Vc, VF, Vp, Vg are the potential energy of the system inertial force,
damping force, Coulomb friction, interference force and gravity. Therefore,
(2.32)
is the total potential energy of the elastic system dynamics, and it is also the general calculation formula of the system.
Imitating the equation (2.28), we call the equation (2.31) as the principle of the total potential energy of the elastic system dynamics. The right
subscript of the variable semicolon in the equation also emphasizes the total potential energy Πd of the elastic system dynamics.
In the case of variation, it is necessary to maintain the essence of the virtual work principle of (2.31), that is, only for elastic strain and displacement
variation, inertial force, damping force, interference force, gravity and other force components are invariant, that is, in calculation, Vm, Vc, VF, Vp, Vg,
all strains and displacements are considered variables, and the corresponding forces are specified. The physical meaning of equation (2.31) is: after
introducing D’Alembert’s principle, at a given instant t, the first-order variation of the total potential energy Πd of the linear system dynamics must be
equal to zero, and Πd takes a constant value. The general formula hold value is based on the functional variational principle. Here, the total potential
energy Πd is constant according to the principle of virtual work. Obviously there is no reason to calculate the first-order variation of X by the
variational method.
The strain and displacement components in Πd should only be changed according to the physical concept of the virtual work principle.
In summary, the basic steps of establishing the equation of motion by the principle of the total potential energy of the elastic system are as follows:
Firstly, the potential energy of each force acting on the system is derived, and the total potential energy of the motion system is obtained Πd; then, the
vibration differential equation of the system is derived from the condition of δε Πd = 0. The greatest advantage of the principle of the total potential
energy of the elastic system dynamics is that it is not necessary to distinguish between forces and non-potentials in the process of system dynamics
modeling, and consider whether the external potential field is stationary and whether the constraints are stable.

2.4 Influence of Gravitational Forces

Consider now the system shown in Figure 2.7. In this case, the system of forces acting in the direction of the displacement degree of freedom is that set
shown. Using Equation (2.4), the equilibrium of these forces is given by
(2.33)

where W is the weight of the body taken as kust = W, then we have,

(2.34)

Figure 2.7 Influence of gravity on SDOF equilibrium.

Comparison of equations (2.4) and (2.33) demonstrates that the equation of motion expressed with reference to the static equilibrium position of the
dynamic system is not affected by gravity forces. For this reason, displacements in all future discussions will be referenced from the static equilibrium
position and will be denoted u(t) (i.e., without the overbar); the displacements which are determined will represent dynamic response. Therefore, total
deflections, stresses, etc., are obtained by adding the corresponding static quantities to the results of the dynamic analysis.

2.5 Influence of Support Excitation

Dynamic stresses and deflections can be induced in a structure not only by a time-varying applied load, but also by motions of its support points.
Important examples of such excitation are the motions of a building foundation caused by an earthquake or the base support of a piece of equipment
due to vibrations of the building in which it is housed.
The horizontal girder in this frame is assumed to be rigid and includes all the moving mass of the structure. The vertical columns are assumed to be
weightless and inextensible in the vertical (axial) direction, and the resistance to girder displacement provided by each column is represented by its
spring constant k/2. The mass thus has a single degree of freedom, u(t), which is associated with column flexure; the damper c provides a velocity
proportional resistance to the motion in this coordinate.
The equilibrium of forces for this system can be written as
(2.35)

in which the damping, elastic and inertial forces can be expressed as,
(2.36)

where u0(t) represents the total displacement of the mass from the fixed reference axis. Substituting for the inertial, damping, and elastic forces in
Equation (2.36) yields
(2.37)
Before this equation can be solved, all forces must be expressed in terms of a single variable, which can be accomplished by noting that the total motion
of the mass can be expressed as the sum of the ground motion and that due to column distortion, i.e.,
Expressing the inertial force in terms of the two acceleration components obtained by double differentiation of equation (2.37) and substituting the
result into equation (2.36) yields
(2.38)

or, since the ground acceleration represents the specified dynamic input to the structure, the same equation of motion can more conveniently be
written
(2.39)

In this equation, Peff(t) denotes the effective support excitation loading; in other words, the structural deformations caused by ground acceleration
üg(t) are exactly the same as those which would be produced by an external load Peff(t) equal to –müg(t). The negative sign in this effective load
definition indicates that the effective force opposes the sense of the ground acceleration. In practice this has little significance inasmuch as the engineer
is usually only interested in the maximum absolute value of u(t); in this case, the minus sign can be removed from the effective loading term.
An alternative form of the equation of motion can be obtained by using equation (2.37) and expressing equation (2.36) in terms of u0(t) and its
derivatives, rather than in terms of u(t) and its derivatives, giving
(2.40)

In this formulation, the effective loading shown on the right side of the equation depends on the velocity and displacement of the earthquake motion,
and the response obtained by solving the equation is total displacement of the mass from a fixed reference rather than displacement relative to the
moving base. Solutions are seldom obtained in this manner, however, because the earthquake motion generally is measured in terms of accelerations
and the seismic record would have to be integrated once and twice to evaluate the effective loading contributions due to the velocity and displacement
of the ground.

Exercises
2.1 Try to answer what are the basic elements for structural dynamic analysis and find out the mechanism and mode of damping.
2.2 What is D’Alembert’s principle? Explain how the principle is employed in vibration problems.
2.3 Write the equation of motion for the inverted pendulum using virtual displacement method.

Figure 2.8

References
1. Clough, R.W. and Penzien, J., Dynamics of Structures, Computers and Structures, 2005.
2. Wilson, J.F., Dynamics of Offshore Structures, Second Edition, John Wiley & Sons, Inc, 2003.
3. Chopra Anil, K., Dynamics of Structures, Prentice Hall, 2000.
4. Timoshenko, S., Vibration Problems in Engineering, Van Nostrand, 1975.
5. Paz, M., Structural Dynamics-Theory and Computation, Van Nootrand Company. Inc, 1985.
6. Hibbeler, A.C., Structural Analysis, Prentice Hall, 2011.
7. Lanczos, C., The Variational Principles of Mechanics, Dover Publications Inc, 1986.
8. Worden, K. and Manson, G., Random vibrations of a multi-degree-of-freedom non-linear system using the volterra series. J. Sound Vib., 1999.
9. Hsueh, W.J., Vibration transmissibility of a unidirectional multi-degree-of-freedom system with multiple dynamic absorbers. J. Sound Vib., 2000.
10. De Wachter, S. and Tzavalis, E., Detection of Structural Breaks in Linear Dynamic Panel Data Models, Computational Statistics and Data
Analysis, 2012.
11. Cluni, F., Gioffrè, M., Gusella, V., Dynamic response of tall buildings to wind loads by reduced order equivalent shear-beam models. J. Wind Eng.
Ind. Aerodyn., 2013.
12. Zhao-Dong, X. and Le-Wei, M., Dynamics of Structures, Science Press, Beijing, 2007.
Chapter 3
Single Degree of Freedom Systems
The concept of dynamic degrees of freedom has been introduced in Chapter 2, in which the system with one dynamic degree is called Single Degree of
Freedom System (SDOF). The SDOF system is the simplest dynamic system. It is of great significance to study the dynamic characteristics and
vibration problems of Single Degree of Freedom Systems for the following aspects: (1) SDOF system has the basic characteristics of a general dynamic
system, including all the physical quantities and basic concepts involved in the dynamic analysis of structure,–which is the basis of learning structural
dynamics. (2) Many practical dynamic problems can be analyzed and calculated directly as a SDOF system, such as one-storied factory and water
tower, etc. Figure 3.1 presents several mechanical models commonly used in the structural dynamic analysis as a SDOF system.

Figure 3.1 Commonly used single degree of freedom systems.

The following picture shows how to simplify complex structure into a SDOF.
This chapter deals with the SDOF systems’ dynamic characteristics as well as its responses under free vibrations at the beginning, then the solving
process of dynamic response under the action of harmonic, periodic, impulsive, and random loading will be introduced. Finally, this chapter will end
with the discussion of damping, vibration energy, vibration measurement, and the principle of vibration isolation.

3.1 Response of Free Vibrations

Chapter 2 has discussed how to establish the motion differential equation of dynamic systems. As is shown in Figure 3.2 (as example 2.1), the equation
of motion for this SDOF system is found to be
of motion for this SDOF system is found to be
(3.1)

Figure 3.2 To simplify a complex structure.

Structural free vibration is a vibrational process in which the structure leaves the equilibrium position when disturbed and no longer affected by any
external force, i.e., P(t)=0, and the equation of motion will be
(3.2)

When the dynamical system has the damping term, which means the system has an energy dissipation mechanism, the reaction of the dynamic system
will be influenced directly. The SDOF system without damping is called the undamped SDOF system, while the one that has damping is termed the
damped SDOF system. Free vibration of the undamped SDOF system will be introduced first.

3.1.1 Undamped Free Vibrations

For the undamped SDOF system, c = 0, according to the Eq. (3.2), the motion equation of the structure can be expressed as
(3.3)

Assuming the solution is in the following form

(3.4)

Where s is undetermined constant and A is constant coefficient.

Substituting Eq. (3.4) into Eq. (3.3) leads to
(3.5)

It is evident that the two eigenvalues of Eq. (3.5) are s1 = iω n and s2 = –iω n. Adopting the notation of , which is a constant parameter related to
the properties of the structure.
Utilizing the relationship between exponential function and trigonometric function of e ix = cos x + i sin x and e –ix = cos x – i sin x, solutions of Eq. (3.4)
may be expressed by sine function and cosine function in the following form
(3.6)

In this formula, A and B are two new undetermined constants which will be confirmed by the initial conditions of the system of A = u (0). If the initial
conditions are defined at t = 0, u (t) = u (0), the solution will be concluded, which is given according to Eq. (3.6)
(3.7)

So that leads to
(3.8)

Introducing Eq. (3.8) into Eq. (3.6) leads to the solution of the free vibration of undamped system:
(3.9)

Obviously, the undamped free vibration of the system is a simple harmonic motion. The structural dynamic characteristics can usually be represented
by three physical quantities of ω n, T n, and fn in its free vibration analysis.
ω n is the natural vibration circular frequency, which can be expressed as:
(3.10)

T n is the time required to complete a cycle of structural motion, which is called the natural vibration period of the structure, which can be expressed as:
(3.11)
fn is called as structural natural vibration frequency, namely the number of cycles of vibration per unit time, whose unit is Hz (times per second). It is
commonly employed as a measurement of the speed of structural vibration which is defined , i.e.,

(3.12)

It can be seen that the natural vibration circular frequency ω n is only related to the stiffness, k, and quality, m, while the natural vibration frequency,
fn, and the natural vibration period, T n, are only related to ω n. Thus, ω n, fn, and T n are all the structural inherent characteristics, which are only
related to the structure itself. They are all important physical quantities that can reflect structural characteristics.
Eq. (3.9) is plotted in Figure 3.3. It demonstrates the curve of the undamped free vibration response changing with time. The curve’s value is equal to
the initial displacement, u(0), and the slope of the curve goes to the initial velocity, , at the initial moment (t = 0). The curve changes along the
slope’s direction, and it will reach its maximum value u0 after a period of time, namely
(3.13)

Where u0 is the amplitude of free vibration system. In line with the trigonometric knowledge of higher mathematics, the response of undamped free
vibration of the Eq. (3.9) can be expressed in the following form:
(3.14)

Figure 3.3 Single degree of freedom system.

Phase angle φ is given by the following formula
(3.15)

The phase angle indicates that the motion of the dynamic system lags behind the angular distance of the cosine, as shown in Figure 3.4.

Figure 3.4 The response of undamped free vibration system.

What can be seen from Figure 3.4 is that the displacement response of the undamped SODF system is a simple harmonic motion centered on a
balanced position, which is connected with displacement u0, initial velocity , and natural vibration period T n.

3.1.2 Damped Free Vibrations

If damping exists in the dynamic system, its free vibration equation will be . Commanding u(t) = Ae st and substituting it into the
equation leads to
(3.16)

In the formula above, is the natural vibration circular frequency of undamped system which has been defined previously. This expression can
describe three different types of motions depending on the signs of the radical above, which can be positive, negative, or zero. When the stiffness and
mass of the structural system is determined, the value of the radical entirely depends on the damping coefficient, c. The value of the radical may be
greater than zero when c is larger, which will be opposite when c is much smaller. These two conditions correspond to two completely different kinds of
motion state, respectively. When the radical is positive, which results in two real solutions s1 and s2, the system will not undergo reciprocating
oscillation. But if the radical is less than zero, these two solutions will be of different pluralities, which will give the system a motion of reciprocating
vibration. There is also a possibility in which the radical equals 0, which is the demarcation line for these two motion states described above; the
damping in this state is called Critical Damping.

1. Critical Damping
To order the radical in Eq. (3.16) is zero leads to the value of critically damped c cr
(3.17)

So the critical damping c cr is also a constant completely determined by the stiffness and mass of the structure. The two characteristic roots of the Eq.
(3.16) will be:
(3.18)

Therefore, the solution of the equation will be:

(3.19)

In the formula, C1 and C2 are the undetermined coefficients which are confirmed by the initial conditions. Supposing the initial displacement and the
initial velocity are u (0) and respectively, then , so the final form of the critical damping system will be
(3.20)

What can be found out is that the critical damping system will not oscillate near the static equilibrium position. The displacement amplitude of the
damped system oscillates with decreasing exponentially up to zero over time, where the system will not oscillate anymore and return to the static
equilibrium position. The so-called critical damping is the minimum value of damping in the case of not oscillating in the free vibration responses.
s1 and s2 will be different negative real roots when the damping of structural system is greater than the critical damping, in which the displacement
amplitude of motion equation will also decay exponentially. In this situation, only if the damping of the system is less than the critical damping will the
system vibrate freely.

2. Damping ratio
The damping ratio, ζ, is often used in theoretical calculation and engineering applications to indicate the size of structural damping, and ζ is the ratio of
damping coefficient c to critical damping c cr, i.e.
(3.21)

(1) When c = c cr, ζ = 1, the damping is namely the so-called critical damping mentioned above
(2) When c < c cr, ζ < 1, the damping is underdamped, and the structural system is known as an underdamped system
(3) When c > c cr, ζ > 1, the damping is overdamped, and the structural system is known as an overdamped system.
The damping ratio ζ is generally determined by experiment, which will be described in the later chapter. The damping coefficient c = 2m ω nζ of
dynamic analytic calculations can be given by the damping ratio. Similarly, the damping ratio is a characteristic parameter to measure energy
dissipation capability of dynamic system. According to statistics, the damping ratio ζ is 0.02 for steel structure, and 0.05 for reinforced concrete
structures, while the damping ratio ζ of the energy dissipation system or the dynamic system added damper (shock absorber) described in Chapter 7
will be 0.10 ~ 0.20 generally.

3. Underdamped systems
Underdamped system indicates that the structural damping is less than the critical damping, i.e. c < c cr, ζ < 1, substituting c = 2mω nζ into Eq. (3.16)
leads to two characteristic roots:
(3.22)

will be a complex number which may leads the two characteristic roots are both complex numbers. Adopting the same analysis method used in
undamped free vibration, the solution of underdamped free vibration system satisfying initial conditions can be obtained.
(3.23)

Where ω D is the natural circular frequency of vibration of the system where .

By comparing the responses of free vibration between the undamped system (Eq. (3.9)) and the damped system, the characteristic of reciprocating
near the equilibrium position by circular frequency, ω n, and the constant amplitude of motion can be understood when free vibration occurs in the
undamped system. However, the free vibration of the underdamped system has an invariable circular frequency ω D while oscillating near the center
position, but its natural vibration frequency will gradually become smaller because of the existence of damping. Meanwhile, Eq. (3.23) indicates that
free vibration of the damping system is a vibration process of amplitude attenuation, in which the displacement amplitude decays in the regular of
exponential function e –ζωnt over time. It is also shown that the damping force constantly consumes energy and weakens the vibration until it
disappears. When ζ = 0, Eq. (3.23) will degenerate into the solution in the form of undamped free vibration as the Eq. (3.9).
The damping ratio ζ is relatively smaller in the actual engineering structure, and the maximum damping ratio is no more than 0.20. at this time,
. Therefore, the influence of damping on the circular frequency will generally be ignored in the dynamic calculation for a
practical engineering structure, namely ω D = ω n.
In order to clearly show the effect of the damping ratio ζ when taking different values on the free vibration motion, the free vibration curve is shown in
Figure 3.5.
Figure 3.5 Influence of different damping on free vibration motion.
4. Overdamped systems
Overdamped system means damping of the structure is greater than the critical damping, namely c > c cr, ζ > 1. In this case, these two characteristic
roots will be two negative real numbers
(3.24)

Where is the natural frequency of damped vibration of overdamped system when considering damping which gives:
(3.25)

The constants C1 and C2, in the formula are determined by the initial conditions. Substituting them in, Eq. (3.25) gives:
(3.26)

In this case, the system movement will be a periodic motion, which is attenuated by the exponential law, due to the damping being too large, as shown
in Figure 3.5.
Example 3.1
Figure 3.6 shows a calculation diagram of a single-story building. Assuming the beam stiffness is infinite, EI = ∞, the weight of roof system, beam, and
part of columns can be concentrated in the beam, which is m = 1×104 kg n order to determine the dynamic characteristics of the gantry in the situation
vibrating horizontally, following vibration experiments were carried out: Applying a horizontal force, P = 98kN makes the gantry have a displacement
of u0 = 0.5cm then the structure will have a free vibration. If the period T ʹ = 1.5s and the lateral displacement of the beam is measured to be u1 =
0.4cm after one cycle. Calculate the damping coefficient of the gantry and the amplitude after 5 cycles of vibration.

Figure 3.6 Influence of different damping on free vibration motion.

Solution: according to the subject
So

The damping ratio

The damping coefficient

3.1.3 Damping and Its Measurement

1. Damping form
The source and brief introduction of damping commonly used have been described in Chapter 2. Damping generally comes from the internal friction
when material deforms, the friction in the connection of structure and the external media around the structure. According to the energy dissipation
mechanism, three main different damping theories were proposed.

(1) Viscous damping

Viscous damping considers the magnitude of the damping force to be proportional to the velocity, while its direction is opposite to the speed which can
be expressed as:
(3.27)

Where the subscript D in F D(t) represents damping; F D(t) is the damping force; c is the damping coefficient; is the velocity of the particle. The
motion equation which established by the theory of viscous damping is easy to solve, so this theory is widely used in current dynamic analyses.

(2) Complex damping

Complex damping is also called structural damping, hysteresis damping, or material damping. This damping theory states that the damping force is
proportional to displacement in the simple harmonic vibration, and its phase is the same with speed, which means the phase of damping force is ahead
of displacement by 90° or the time ahead of the quarter cycle. The damping force can be expressed in the following complex number form
(3.28)

In the formula, η is the complex damping coefficient, k is the stiffness coefficient. Complex damping can better reflect the energy dissipation
mechanism of the friction in the material, which has wide applications.

(3) Friction damping

Friction damping holds that the damping force is a constant, which is equal to the friction, and its direction is opposite to the velocity. The friction
damping, also known as dry friction damping, can be expressed as:
(3.29)

Where μ is the coefficient of friction, N is the positive pressure between the friction of the contact surface. Friction damping is generally applied to the
dynamic system where the frictional resistance is dominant, while coulomb damping is more suitable in the engineering structures installing friction
dampers or the friction-type isolation structures.
Furthermore, there are other damping models such as fluid damping, exponential damping, etc., and interested readers can refer to the relevant books.

2. Motion attenuation and measurements of damping ratio

What can be seen from Figure 3.5 is the structural free vibration response is greatly affected by damping (damping ratio), attenuation of the free
vibration response of dynamic systems with larger damping ratio is faster than that of dynamic systems with smaller damping ratio. According to this
principle, we can obtain the damping ratio ζ of the structure through the structural free vibration test, as shown in Figure 3.7.
Figure 3.7 Free vibration decay curve.
If there are two amplitude values of un and un+1 which are separated by one cycle in the damped free vibration system, the simplest way to measure
the damping ratio ζ is to calculate the ratio of the two values. The Eq. (3.23) can be expressed as:
(3.30)

(3.31)

There is ω DΔt = 2π. so

(3.32)

It can be seen that the ratio of adjacent vibration peak value is only related to the damping ratio and has nothing to do with the value of n. Taking the
logarithm on both sides of the Eq. (3.32), we will get the logarithmic decay rate:
(3.33)

For a small damping system, the value of ζ is very small, so approximately (1 – ζ 2) ≈ 1. The approximate formula of the damping ratio of small
damping system obtained from Eq. (3.33) will be:
(3.34)

Using the above formula, the damping ratio ζ can be calculated from the two measured adjacent amplitude values. However, the free vibration will
decay slowly when the damping of the structure is minute, then using a vibration peak ratio of several cycles apart to calculate the damping ratio of the
structure will get higher accuracy at this time. The ratio of these two amplitudes which separated by m cycles is
(3.35)

Thus, the logarithmic attenuation rate can be obtained

(3.36)

(3.37)

In the experiment, the damping ratio is sometimes calculated by the number of times J50%, when the vibration peak decays to 50%, and at this time,
,

(3.38)

When the damping ratio of the system is calculated by the method mentioned above, there will only be information of the two peak points on the
system vibration decay curve being used. There is another way to obtain the exponential decay law of the vibration amplitude by fitting the peak point
decay curve, then the damping ratio which has an important effect on vibration attenuation can be calculated. Interested readers can refer to the
relevant books on this subject. The following sections of this chapter will also introduce other methods of measuring damping ratios.
Example 3.2
Structural parameters are the same as example 3.1. If using a jack to make m have a lateral displacement of 20mm, and then suddenly releasing it, the
system will vibrate freely. The lateral displacement was measured to be 10 mm after 4 cycles of vibration. Please determine: (1) the damping ratio and
damping coefficient of the structure (2) the amplitude after 10 cycles of vibration.
Solution: As can be seen from example 3.1

So

the natural vibration period the circular frequency

(1) Damping ratio and damping coefficient

The damping ratio is known from Eq. (3.39)

(2) As can be seen from Eq. (3.39)

Therefore, the amplitude after 10 cycles of vibration is 3.531 mm.

3. Free vibration test

The damping ratio ζ of an actual structure cannot be determined by the theoretical analysis method since this measurement could produce a large
error. It must be determined by the free vibration test. The free vibration test provides a way to determine the structural damping ratio. According to
the above derivation, when the structural damping is small (damping ratio <20%). Through Eq. (3.34), the damping ratio can be expressed as:
(3.39)

Because it generally uses the acceleration sensor (vibration sensor), measurement accuracy in certain measured spots is improved. The measurement
signal recorded by the sensor is the acceleration time can be proven by Eq. (3.30) and the damping ratio can be calculated directly
from the acceleration amplitude:
(3.40)

In the free vibration decay curve, the natural vibration period T n of the structure can be calculated by the interval of adjacent amplitudes.

3.2 Response to Harmonic Loading

3.2.1 Harmonic Vibration of an Undamped System
Assuming the structural system is subjected to a harmonically varying load P0 with a circular frequency of ω, and ignoring the damping of the system,
the motion differential equation of the SDOF system becomes:
(3.41)

Then the solution includes two parts; homogeneous and particular solutions:
(3.42)

uc(t) is the homogeneous solution of the equation which is equivalent to free vibration solution, according to Eq. (3.6), uc(t) can be expressed as
(3.43)

The particular solution of the equation is a special reaction caused by the dynamic load, which is dependent on the form of dynamic loading. In this case
of harmonic loading, it is reasonable to assume that the corresponding motion is harmonic and in the same phase with the loading. Thus, the particular
solution of the equation is:
(3.44)

Substituting Eq. (3.44) into Eq. (3.41) leads to:

(3.45)

In which is defined as the frequency ratio of the applied loading frequency to the natural free-vibration frequency. The complete solution of the
motion equation can be obtained by making use of Eq. (3.42), which is:
(3.46)

In the equation above, the coefficients A and B are determined by the initial conditions. Taking into account the initial displacement and velocity are
u(0) and respectively, it can be obtained by the Eq. (3.46):
(3.47)

Finally, the solution satisfying the initial conditions is obtained:

(3.48)

The third term in Eq. (3.48) is caused directly by the dynamic load, and its vibration frequency is the same as the external load frequency ω, which is
called the steady-state reaction, also known as the forced vibration reaction. The first and the second items of Eq. (3.48) are equivalent to free
vibration, and its vibration frequency is equal to the natural vibration frequency ω n of the system, which is known as the transient response. In
practice, the free vibration term will decay to zero soon due to the inevitable existence of the damping in real systems. This means there will only be
the steady-state reaction left finally, which is directly caused by the external load. With the increase of time, the free vibration term reduces gradually
until disappearing, and the steady-state solution of the system will be:
(3.49)

Where , is called the equivalent static displacement or static displacement, which may be interpreted as the static deformation due to the
amplitude P0 of the force, and u0 is the amplitude of the steady-state reaction. The displacement amplification factor is defined as the ratio of the
maximum value of the dynamic displacement to the static displacement, which is:
(3.50)

Figure 3.8, which shows Eq. (3.50) for Rd plotted as a function of the frequency ratio, permits several observations:
When ω = 0, Rd = 1, the dynamic problem will be transformed into a static problem.

Figure 3.8 The curve of the dynamic amplification factor Rd of the undamped system with the change of frequency.
When ω = ω n, Rd → ∞, the dynamic response tends to infinity, known as resonance.
When , the dynamic response of the system is going to be less than the static reaction.

When ω → ω n, the particular solution of the motion equation is:

(3.51)

And the complete solution satisfying zero initial condition is:

(3.52)

Figure 3.9 shows the process of how the dynamic response of the system increases with time under initial conditions of zero. When the system
resonates, the vibratory deformation increases gradually, instead of becoming unbounded immediately.
Figure 3.9 The dynamic response of undamped system increases with time during resonance.
Example 3.3
Figure 3.10 shows a non-compact beam with a motor with a weight of W=20kN in the midspan. The centrifugal disturbance produced by the
eccentricity of motor is , If the revolutions per minute of the machine n = 500 r / min and the stiffness of the beam EI = 1.008 × 104
2
kN · m . Find the maximum displacement and bending moment of the beam under the undamped case.

Figure 3.10 A non-compact beam with a motor.

Solution:
(1) Maximum static displacement of beam under the action of machine gravity

(2) The natural vibration frequency of the beam is

(3) The frequency of the disturbance of the machine is

(4) The dynamic coefficient of the system is

Maximum displacement and bending moment of beam cross section in the midspan will be
3.2.2 Harmonic Vibration of Damping System
Considering the effect of damping, the motion equation of the SDOF system will be
(3.53)

The two sides of the equation are divided by m, and let c = 2mω nζ. The following form of motion equation can be obtained
(3.54)

The solution of the Eq. (3.42) includes two parts: homogeneous solution and special solution.
Assuming ζ < 1, the homogeneous solution uc corresponds to the free vibration response of the damping system, i.e., Eq. (3.23)
(3.55)

Where is the natural vibration frequency of the damping system. The special solution up of the motion differential equation can be set to
the following form:
(3.56)

Substituting Eq. (3.56) into the equation of motion (3.53) leads to:
(3.57)

By the arbitrariness of time, the following two simultaneous equations on the coefficients of C and D can be obtained:
(3.58)

The solution of equations is obtained:

(3.59)

Where is the static displacement of the structure. Then the complete solution of the motion equation is:

(3.60)

It can be seen that the steady-state reaction is a forced vibration reaction, and the maximum displacement can be calculated from Eq. (3.56) and Eq.
(3.59):
(3.61)

At this point, the structural dynamic amplification coefficient is:

(3.62)

When the damping ratio ζ = 0, the above equation degenerates into the form of undamped systems. The result will be the same as Eq. (3.50)
The first term in Eq. (3.60) represents the transient response of the system, and the constants of A and B can be calculated from any given initial
condition. However, as time increases, the damping term will quickly disappear. The second term is the steady-state reaction only caused by the
external load, which has the same frequency and different phase to the external load, and it is also known as the forced vibration reaction. It should be
noted that in special cases, the term of transient response may be much larger than the term of steady-state reaction at the initial stage of the reaction,
thus it becomes the controlling quantity of the structural maximum response. For this case, the term of transient reaction cannot be neglected in the
structural dynamic response analysis or in the structural design. The term of transient response is vibrating by the natural vibration frequency of the
structure, so it demonstrates the dynamic characteristics of the structure, while the term of steady-state reaction is vibrating by the external load
exciting frequency, and it reflects the nature of the input load.
When the load excitation frequency is close to the natural vibration frequency of the system, i.e., ω ≈ ω n, the damping system resonates. At this time,
the dynamic amplification coefficient is the largest, and its dynamic response can be drawn by Eq. (3.60) as shown in Figure 3.11.
Figure 3.11 Resonance reaction time-history curve of damping system.
Compared with the resonance response time-history curve of the undamped system, it is found that the rate of dynamic response growth is less than
that of no damping due to the existence of damping force.

3.2.3 Dynamic Amplification Coefficient

In the dynamic analysis, the dynamic amplification coefficient is an important concept. The dynamic amplification coefficient is the ratio of the
maximum value of dynamic system steady-state reaction to the static displacement caused by static load equivalent to the magnitude of load. For a
damping system the calculation of dynamic amplification coefficient is adopted Eq. (3.62)
(3.63)

When ζ = 0, the dynamic amplification coefficient Eq. (3.50) of the undamped system can be obtained as:
(3.64)

It can be seen that for a damping system, the dynamic amplification coefficient Rd is a function of the damping ratio ζ and the frequency ratio . If
a different damping ratio is given, the dynamic amplification coefficient is a curve that varies with the frequency ratio.
As shown in Figure 3.12, the graph is called the amplitude-frequency characteristic curve. The steady state reaction of the damping system Eq. (3.56)
can be written as:
(3.65)

Figure 3.12 Amplitude-frequency characteristic curve of damping system.

The phase angle φ can be obtained by mathematical deduction:
(3.66)

Similarly, the phase angle φ is also a function of the damping ratio ζ and the frequency ratio . The phase angle versus frequency ratio at different
damping ratios can be plotted, as shown in Figure 3.13, called the phase-frequency characteristic curve. The curve that phase angle φ varies with
frequency β at different damping ratios ζ can be graphed, as shown in Figure 3.13, called the phase-frequency characteristic curve.
Figure 3.13 Phase-frequency characteristic curve of damping system.
By analyzing the amplitude-frequency characteristic curve of Figure 3.12 and calculation Eq. (3.62) of the dynamic amplification coefficient, it can be
found that
(1) When , Rd ≤ 1, the amplification reaction will not be turned out in the system.

(2) When , the peak value of dynamic amplification coefficient . As the damping ratio increases, the peak of

the dynamic amplification coefficient tends to decrease. The maximum value corresponding to the frequency is , indicating that the
maximum dynamic response of the damping system will not occur at ω = ωn, but take place at .

(3) When ω / ωn = 1, , it is always considered that the maximum reaction (resonance) occurs approximately at ω = ω n in practical
engineering.
(4) When , Rd ≤ 1, any ζ is tenable. In some dynamic system designs, if the system is designed rationally, the frequency ratio
can be made within this range, so that the dynamic response is effectively reduced.
Example 3.4
The basic dynamic system shown in Figure 3.14 has the following characteristics: m = 10kg, k = 104 N / m. If the system is subjected to a resonant
harmonic load (ω = ω n) and the initial state is stationary, please set the value of the dynamic amplification coefficient Rd(t) after five cycles (ωt = 10π),
(1) no damping system; (2) ζ = 0.02; (3) ζ = 0.05

Figure 3.14 Simple basic dynamic system.

Solution: According to the given information:

(1) c = 0, namely no damping

(2) ζ = 0.02

(3) ζ = 0.05
3.2.4 Resonance Reaction
It has been said that when , the peak value of dynamic amplification coefficient is the maximum, for the occurring of resonance. This is
obtained by solving the first derivative of the dynamic amplification coefficient Eq. (3.62), which is equal to zero.
When , it can be assumed , and the value of Rd is maximum, i.e., . This suggests that Rd is a large value in the case of

small damping with ω → ω n, which increases the amplitude dramatically, so the interval of is called the resonance region.

When the structure resonates, ω = ω n, from Eq. (3.59) can be obtained:

(3.67)

Substituting Eq. (3.67) into Eq. (3.60) and then command it to satisfy zero initial conditions can lead to:
(3.68)

Finally, the resonance reaction satisfying the zero initial condition is obtained:
(3.69)

When the damping is small, ω D = ω n, the sine term in Eq. (3.67) has little effect, and at this point, the Eq. (3.69) becomes:
(3.70)

When ζ = 0, the resonance reaction with zero initial condition can be obtained by the Eq. (3.70)
(3.71)

3.2.5 Solution of Damping Ratio

In the actual structure, the damping and energy dissipation mechanism has not been fully understood, so the damping coefficient of many structures
must be obtained from the experimental method directly. There are usually three methods to calculate the structural damping ratio by experiments:
logarithmic decay rate method, bandwidth method, and resonance amplification method.

(1) Logarithmic Decay Rate Method

This method has been introduced in the first section of this chapter in the free vibration of SDOF systems, namely using the logarithmic decay rate
method to calculate the viscous damping ratio of the SDOF systems. The damping ratio of the system can be obtained by the logarithm value δ of the
free vibration amplitude ratio (usually take the amplitude which has the interval of m cycles)
(3.72)

Where . The reason why the damping ratio ζ can be obtained by the free vibration method is that the speed of the free vibration decay is
controlled by ζ, or the attenuation law of free vibration, which can clearly reflect the influence of the damping ratio ζ. The main advantage of this free
vibration method is that it requires the least amount of instruments and equipment, while any simple and convenient method can be used to get the
vibration and the required measurement is only the relative displacement amplitude.

(2) Bandwidth Method

Bandwidth method, also known as half-power point method, is based on the characteristics of dynamic amplification coefficient Rd. From the
amplitude-frequency characteristic curve of the dynamic amplification coefficient Rd, it can be seen that the Rd curve shape is completely controlled by
the damping ratio, and when the damping ratio is large, Rd is fat (wide); when ζ is small, Rd is thin (narrow). The damping ratio of the system can be
determined by the width of the half power point. As the amplitude-frequency characteristic curve shown in Figure 3.15.
Figure 3.15 Calculating damping ratio on amplitude-frequency characteristic curve.
ω a and ω b are the two frequency points corresponding to the points of Rd whose amplitude is equal to times the maximum amplitude, which is
called the half power point. When the damping is relatively small, the relationship between the half power point and the damping ratio ζ is as follows
(3.73)

Or
(3.74)

The damping ratio in the above formula is calculated by the circular frequency, and may also use the engineering frequency to calculate the damping
ratio, i.e.,
(3.75)

Where fn is the engineering natural vibration frequency of the corresponding undamped dynamic system.

Notice and the amplitude is equal to .

The corresponding frequency satisfies the following equation

(3.76)

Taking the reciprocal of Eq. (3.76) on both sides and squaring it results in:
(3.77)

Eq. (3.77) is a quadratic equation about (ω / ω n)2, which can have two roots
(3.78)

Eq. (3.78) corresponds to the larger root ω b when it takes a positive sign, otherwise corresponds to the smaller root ω a. The damping of the general
engineering structure is relatively small, and the square term of the ζ in Eq. (3.78) can be neglected, so:
(3.79)

The two roots corresponding to the half power point are:

(3.80)

The relation between the half power point frequency, ω b and ω a, and the damping ratio ζ is obtained by Eq. (3.80)
(3.81)
Thus, Eq. (3.73) is obtained. If substituting the relationship from Eq. (3.80) into Eq. (3.73), Eq. (3.74) which used to calculate ω n is also
obtained.
Bandwidth method (half power point method) uses forced vibration tests, which can be used for not only SDOF system but also MDOF system. While
Making the resonant frequency sparse is required for the MDOF system. The multiple natural vibration frequencies should be far apart, and it must
not be affected by adjacent frequencies when determining the half power point corresponding to a natural vibration frequency. By using this method,
the required static response in the resonant amplification method can be avoided. However, it is necessary to accurately graph the half power range
and the reaction curve at resonance.

(3) Resonance Amplification Method

Similarly, starting from the dynamic amplification coefficient, the amplification coefficient of arbitrary frequency is the ratio of the response amplitude
of the frequency to the zero-frequency (static) response amplitude. From the previous analysis it has been shown that the damping ratio is closely
related to the resonance amplification coefficient, and . When resonance (ω/ω n = 1) occurs,

, obviously, once the dynamic amplification coefficient Rd curve is obtained, the damping ratio can be determined

(3.82)

Since it is not easy to determine u0(ω n) from the dynamic amplification coefficient curve, the value u0m is generally used instead, u0m = max(u0), then
(3.83)

When the damping ratio is relatively small (e.g., ζ<20%), the error caused by this substitution is very small. Determining the system damping ratio
with the resonance amplification method is simple; however, the dynamic amplification coefficient curve measured in practical engineering is generally
given by the u0 – ω. To use Eq. (3.82) or Eq. (3.83) to calculate the damping ratio of the system, it is also necessary to obtain the static displacement
value at zero frequency. For large structures, the actual measurement of static load displacement ust has some difficulties both of loading equipment
and recording (picking up) equipment, namely the equipment of realizing dynamic loading and measuring dynamic signals cannot work at zero
frequency. Therefore, the bandwidth method (half power point method) is used more in engineering to obtain damping ratio ζ from the dynamic tests.

3.3 Periodic Load Response

According to higher mathematical knowledge, any periodic function can be developed in the form of Fourier series. Assuming P(t) is an arbitrary
periodic load, P(t) = P(t + T p), and T p is the period of the external load P(t), which can be developed in the following Fourier series:
(3.84)

Where , , , ,

For each of the reactions in the series, the response is only under the simple harmonic load. Then using the principle of superposition (only for the
linear system), the total response is the sum of the response of each load term. The series includes the constant load (the average load value
represented by a0) and a series of harmonic loads with frequency ω j and amplitude aj and bj. Considering that the transient term will disappear with
time, if the cycle time is long enough, the transient response that occurred at the beginning has disappeared. Only the calculation of steady-state
reaction is needed, namely the particular solution.
Let u0a be the response of the system under the action of constant load, a0, ujs be the response of the system under the action of sinusoidal harmonic
load, bj sin ω jt, and ujc be the response of the system under the action of cosine harmonic load, aj cosω jt. In fact, the harmonic load is one of the
simplest and most representative periodic loads, and any periodic load can be decomposed into algebraic sum of simple harmonic loads.
The steady-state response under sinusoidal loading is Eq. (3.65), and the upper section has been described in detail. If the load is a cosine load, setting
external load as P(t) = P0 cos ωt, the equation of motion is:
(3.85)

Let the steady-state solution be:

(3.86)

Substituting the steady-state solution into the equation of motion leads to:
(3.87)

u(t) can be written as:

(3.88)

Where ,

It can be seen that the form of dynamic response of the system under the action of cosine load is similar to that of sinusoidal load. Combining Eq. (3.65)
and Eq. (3.88) can be easily obtained:
(3.89)
(3.90)

(3.91)

Where , ω n is the circular frequency of auto oscillation of the system. So the total steady-state response of the system under arbitrary
cyclic loading is:
(3.92)

If ζ = 0 is substituted into the Eq. (3.92), the total periodic response of the structural undamped system can be obtained:
(3.93)

Using the corresponding exponential form given by Euler’s equation instead of a trigonometric function, the Fourier series expression can be rewritten
into exponential form, and then the solution of the complex form is obtained. The periodic load can be expressed as:
(3.94)

(3.95)

Where Pj = P(iω j) is the amplitude of complex harmonic loads.

When only considering the steady-state response of the system, and introducing the simple harmonic load e iωt of unit complex into the equation of
motion leads to:
(3.96)

Let H(iω) be the complex amplitude value of the steady-state solution under the simple harmonic load eiωt of unit complex, and the steady-state
solution of this equation has the following form:
(3.97)

Substituting Eq. (3.97) into Eq. (3.98) leads to:

(3.98)

H(iω) is called the frequency response function, also called the complex frequency response function or transfer function. The frequency response
function and the unit impulse response function, which will be described later, constituting a Fourier transform pair. For the dynamic problem, these
two functions play a very important role in the frequency domain analysis and time domain analysis. The total steady-state response of a dynamical
system to any periodic load can be written as:
(3.99)

Example 3.5
Solve the dynamic response of the system under periodic load in Figure 3.16, m = 10 kg, k = 104 N / m, ζ = 0.05, P(t) = 600 + 200 sin 6πt – 50 cos
3πt (unit: N).
Figure 3.16 Simple dynamical system.
Solution: According to the meaning of the title

Order

Thus

Under the action of F 0(t)

Under the action of F 1 (t), ω 1 = 6π, P10 = 200N then

It can be obtained from Eq. (3.65)

The displacement at different times under the action of F 1 (t) can be calculated by the formula above.
Under the action of F 2(t), ω 2 = 6π, P20 = 200N, then:

It can be obtained from Eq. (3.88) that:

The displacement at different times under the action of F 2(t) can be calculated by the formula above. Thus, the displacement of the system under the
action of load P(t) at any moment is: u(t) = u0(t) + u1 (t) + u2(t).
The final displacement time-history diagram is shown in Figure 3.17.

Figure 3.17 Displacement time-history diagram.

3.4 Impulsive Loading Response

The impulsive loading consists of a single primary pulse, and its duration is usually very short. The impulsive loading is very important for designing
some components of the structure, such as explosion proofing of the building structure, crane brake, and the impact of start-up on plant, impulsive
loading of vehicle on bridge, and so on. Under the impulsive loading, the maximum reaction of the structure will be reached in a very short time;, while
the damping force has not had enough time to absorb much energy from the structure before that, so compared with the structure under harmonic
loading or cyclic loading, damping is less important in controlling the maximum response of the structure. Therefore, this section will only discuss the
undamped response of the system under impulsive loading.
The response of the SDOF system to the load is mainly related to the type of impulsive loading, the duration of the impulsive loading and the ratio of
the natural period of the structure. There are many kinds of impulsive loads, such as sine-wave impulsive load, rectangular impulsive load, triangular
impulsive load, etc.
The analysis of the response of the structure to the impulsive loading is usually carried out in two stages. The first stage is the forced vibration of the
load. The second stage is the free vibration without load, and the initial condition is the displacement u(t) and velocity at the end of the impact load
of the first stage. In this paper, the dynamic responses of structures subjected to these three kinds of impact loads are introduced respectively.

3.4.1 Sine-Wave Impulse

Consider the sine-wave impulse shown in Figure 3.18, and divide the reaction into two stages:
Stage I: In this stage, the structure is subjected to a simple harmonic load and moves from rest. The undamped reaction containing transients and
steady states is given by the following Eq:
When 0 << t << t 1 ,
(3.100)

Figure 3.18 Sine-wave impulse.

Stage II: The free vibration occurring at this stage is related to the displacement u(t) and the velocity at the final moment of the stage I, which can
be represented by the following expressions:
When ,
(3.101)

The magnitude of the dynamic response produced by the impulsive loading depends on the ratio of the duration of the load to the period of the
structural vibration.
Generally speaking, the maximum response caused by impulsive loading is very important in structural analysis, so the time of peak reaction can be
obtained by finding the derivative.
(3.102)

It can be obtained that

(3.103)

The maximum response occurs in the time of impulsive loading, and the load frequency tends to the free vibration frequency, i.e. ω → ω n. The time at
which the maximum reaction takes place can be calculated by Eq. (3.102), substituting n = 1 into it and take the negative sign in Eq. (3.103):
(3.104)

The maximum reaction amplitude can be obtained by substituting Eq. (3.104) into Eq. (3.100). This result is correct only when is assumed,
namely only when the case of is correct.

When , the maximum reaction occurs in the free vibration stage (stage II), and the initial displacement and initial velocity at this stage can be
obtained by substituting ωt 1 = π into Eq. (3.100)
(3.105)

(3.106)
The amplitude of the free vibration motion can be obtained as:
(3.107)

Therefore, the amplification coefficient of this case is:

(3.108)

3.4.2 Rectangular Impulse

The response of structures under impulsive loading can be divided into the two stages of loading stage and free vibration stage to analyze. The two
stages under the action of the rectangular impulse is shown in Figure 3.19.

Figure 3.19 Rectangular impulse.

Stage I: In this stage, the sudden applied constant load is called step load, and the special solution of step load is the static displacement caused by it.
(3.109)

From this result, it is found that the free vibration constants in the general solution can be determined by the initial conditions which satisfy the
stationary conditions, and thus the general solution can be easily obtained. When 0 << t << t 1 ,
(3.110)

Stage II: In this stage, the free vibration is also given by Eq. (3.101). When ,
(3.111)

For this kind of rectangular impulse, obviously, if , the maximum reaction will always appear in stage I, and the dynamic amplification coefficient,
Rd, equals 2. For the short duration load, the maximum reaction will appear during the free vibration of stage II, and the amplitude of the vibration
response is:
(3.112)

Where ,

(3.113)
Thus, when , the dynamic amplification coefficient is a sine function that varies with the load pulse length ratio .

3.4.3 Triangular Impulse

Triangular impulsive loading is a kind of loading that decreases with time. As shown in Figure 3.20, the blast impulsive loading is often simplified as this
sub-form.

Figure 3.20 Triangular impulse.

Stage I: in this stage, the load is , and the special solution under this load can be obtained as:

(3.114)

If the initial condition is assumed to be zero, the free vibration constant in the general solution can be calculated:
(3.115)

Stage II: At the end of the calculation stage I (t = t 1 ), resulting in:

(3.116)

(3.117)

Then substituting the above equation into Eq. (3.101) can give the free vibration response of the stage II.
It can be seen from the above analysis that in the undamped single degree of freedom structure, the maximum response caused by a given impulsive
loading type only depends on the ratio of the duration of impulse to the inherent period of the structure, namely . Through the analysis, we can get
two general conclusions about the structural response under impulsive loading:
(1) For long duration loads, such as , the dynamic magnification coefficient mainly depends on the growth rate of the load as it reaches its
maximum value. The dynamic amplification coefficient produced by the step load with sufficient duration is 2, which is 1 when the load increases
slowly.
(2) For the loads with short duration, such as , the maximum displacement value umax mainly depends on the magnitude of the action
impulse . The form of the impulsive loading has little effect on it.

Based on the second conclusion, the approximate solution under impulsive loading can be derived. As the time of impulsive loading is very short,
according to the law of impulse:
(3.118)

(3.119)

Considering the solution of the dynamic response Eq. (3.101) . Because the displacement term u(t 1 ) can be neglected
compared with the velocity term , and the velocity , the following approximate relations can be obtained:
(3.120)
 Example 3.6

Assuming an SDOF system is subjected to the triangular impulsive loading, as shown in Figure 3.20, , solve for the maximum dynamic
displacement and the amplification coefficient.
Solution: The time of impulsive loading is very short, so the influence of damping is not considered.
Let the initial condition be zero and substituting P(t) into Eq. (3.115) leads to:

Integrating the formula above, and noticing that mω 2 = k leads to:

In order to obtain the maximum displacement, we can make the first derivative of u(t) be zero, and then the time t m to maximum displacement can be
obtained:

By using trigonometric function, the formula above can be rewritten:

Thus available:

The maximum dynamic displacement is:

After finishing:

Amplification coefficient:

It must be noted that the above two formulas are effective only when t m < t 1 , namely

Solving this inequality can obtain:

When , the maximum displacement response will occur in the free vibration state at t > t 1 . Substituting t = t 1 into the solution of dynamic
response and its derivative of time t, the initial displacement and the initial velocity of the stage t > t 1 can be obtained:

By substituting ζ = 0and ω ʹ = ω into the free vibration equation and changing the time variable t to t – t 1 , the displacement expression at this time is
obtained:

And its amplitude is:

The amplification coefficient is:

It can be seen from the calculation formula of the amplification coefficient that the amplification coefficient Rd is only related to the value of ωt 1 , in turn,
only related to the value of .

3.5 Response of Arbitrary Load

In practical engineering, many dynamic loads are neither simple harmonic loads nor periodic loads, but loads varying arbitrarily with time. This section
will study the response when the system is subjected to general arbitrary load excitation. There are two methods to calculate the dynamic response of
the system under arbitrary loads. One is carried out in time-domain and regards the arbitrary load as a series of short action time of pulsed
superposition. The unit impulse response is obtained firstly, and then the superposition principle is used to superimpose a series of impulse responses
(or integral form) to get the dynamic response of the system. The other method is carried out in frequency-domain, and the solution is first carried out
in the frequency domain by the frequency response function and Fourier integral, then the frequency domain solution is transformed into time-domain
solution by Fourier inverse transform.

3.5.1 Duhamel Integral (Time-Domain Analysis)

1. Unit Impulse Response Function
As mentioned above, the action of arbitrary loading with time varying arbitrarily can be regarded as a continuous action of infinite impulse P(t) dt, as
shown in Figure 3.21. Since the equation obtained by the micro-amplitude vibration is a linear differential equation, so the superposition principle of the
force action can be applied; specifically, the reaction induced by any load is the sum of the reactions caused by all the impulses.

Figure 3.21 Deduction of Duhamel integral under arbitrary load.

The unit impulse is the impulsive loading with short acting time and P(t)dt = 1. Equivalent to the function of δ in mathematics, and the δ function
shown in Figure 3.22 has the following properties:
(3.121)

(3.122)

Figure 3.22 δ function load diagram.

Solving the response of the system under the action of the unit pulse is to solve the free vibration problem after the unit pulse. The action of the unit
pulse is equivalent to giving an initial condition, substituting the initial conditions of u(τ) = 0 and after impulsive action at time τ into the
general solutions Eq. (3.9) and Eq. (3.23) of free vibration of an SDOF system, which leads to the unit impulse response function of the dynamic
system.
The undamped system:
(3.123)

The damping system:

(3.124)
The unit pulse response function uses the t in h(t – τ) to represent the time of the dynamic response of the structural system, and τ is the moment of
the unit pulse.

2. Duhamel Integral
First, calculate the dynamic response of any pulse Ρ(τ)dτ if the external load Ρ(τ) acting on the structural system has been discretized into a series of
pulses. The undamped system is discussed in advance. Since the impulse of the pulse is equal to Ρ(τ)dτ, the response of the structure under the action
of the pulse can be obtained directly by the unit pulse response function:
(3.125)

The total reaction can be obtained by adding all the differential reactions produced in the load time-history, that is, by integrating Eq. (3.125) as follows
(3.126)

The Eq. (3.126) is called the Duhamel integral of an undamped system. In this equation, it must be assumed that the structure is in static state at the
beginning of the load. For other specific initial conditions: when u(τ) and are nonzero initial conditions, the complete solution of the problem is:
(3.127)

Similarly, for the damping system, the Duhamel integral equation for its dynamic response is:
(3.128)

3. Numerical solution of the Duhamel integral

(1) The Undamped System
Write Eq. (3.126) as:
(3.129)

(3.130)

Therefore, the numerical solution of Duhamel integral can be obtained after the two integrals in Eq. (3-130) are solved by numerical integration. In this
step, the Simpson method is primarily used. In this method, the integral interval is first divided into several (even) equal length segments, and the
integrand curve between each two adjacent subparagraph is approximately replaced by parabola, then determine the area between each parabola and
x-axis, and the approximate value of the above integral can be obtained.
The integral area formula between the two paragraph interval (x, x + 2Δx) is
(3.131)

With the above Eq. (3.131), the first equation of the Eq. (3.130) can be expressed as follows:
(3.132)

Where:
(3.133)

In the formula, represents the total value determined before the moment t – 2Δτ.

Similarly, the summation formula of the second formula in Eq. (3.130) is obtained
(3.134)

Where:
(3.135)

Substituting Eq. (3.132) and Eq. (3.133) into Eq. (3.135) leads to:
(3.136)
Thus, the Duhamel numerical integral operation can be carried out within (0, t) interval with two steps each time.

(2) The Damping System

When damping is taken into account, it can also be expressed as a form similar to that of Eq. (3.129):
(3.137)

(3.138)

It is found that the attenuation factor is included in the numerical integration of damping system. Using the Simpson method, the first equation in Eq.
(3.138) is
(3.139)

Where:
(3.140)

Likewise, the second equation in Eq. (3.138) is:

(3.141)

The accuracy of the solution obtained from the numerical integration is clearly related to the length of the time interval Δτ. Generally, satisfactory
results can be obtained when taking .

3.5.2 Fourier Transform (Frequency-Domain Analysis)

Time-domain analysis can be used to compute the response of any linear SDOF system subjected to any input; however sometimes it is more
convenient to be analyzed in frequency-domain. The basic idea of frequency-domain analysis is based on Fourier transform, which is similar to the
time-domain analysis of periodic load response described in the preceding section. Fourier transform is defined as:
Time-domain conversion to frequency-domain (positive transformation)
(3.142)

Frequency-domain conversion to time-domain (inverse transformation)

(3.143)

Where U(ω) is called the Fourier spectrum and u(t) is displacement. According to the Fourier transform properties, the velocity and acceleration of the
Fourier transform are:
(3.144)

(3.145)

Both sides of the motion equation of SDOF system are simultaneously transformed by positive Fourier transformation.

(3.146)

Where U(ω) and Ρ(ω) are the Fourier spectra of u(t) and P(t), respectively. It can be seen that the problem is changed from the time-domain (the
independent variable is t) to the frequency-domain (the independent variable is ω) by the Fourier transform, and it can be obtained from the motion
equation in frequency-domain Eq. (3.146) that:
equation in frequency-domain Eq. (3.146) that:
(3.147)

(3.148)

The meaning of the frequency response function Eq. (3.148) is the same as Eq.(3-96). After the analysis is completed in the frequency-domain, the
dynamic response of the system can be obtained by using the inverse Fourier transform in Eq. (3.144), transforming the frequency-domain solution
into the time-domain solution.
(3.149)

Integral calculation is often very troublesome by the Fourier transform with load function. In the actual calculation, the Fourier transform of discrete
data is used extensively.
The discrete Fourier transform firstly assumes that the load is a periodic load which period is T p. Then the function of continuous change over time is
separated into a series of N discrete data points with equal step length Δt, i.e. t m = mΔt.
(3.150)

(3.151)

If N = 2m, using the cyclical characteristics of the simple harmonic function e ± ix can obtain the fast Fourier transform (FFT). The application of FFT
is one of the most widely used methods in Fourier transform and it can greatly speed up the analysis and reduce the workload.
Special attention should be paid to the application of discrete Fourier transform method to analyze the dynamic response of the system under any
general load:
(1) The discrete Fourier transform cycles the aperiodic time function.
(2) In order to minimize the error in the nonperiodic load analysis, increasing the duration of T P by adding enough zero points to P(t) to ensure
that the displacement of the system can be reduced to zero within the calculated period [0, T p].
(3) Spectrum upper limit frequency (also known as Nyquist frequency) , .

(4) The resolution of the spectrum is , i.e. .

(5) The lower limit of the spectrum .

There are many methods for Fourier transform and inverse Fourier transform of continuous function or discrete data, and there are also many ready-
made programs. which is very convenient to carry out these transformations in Matlab. Interested readers can refer to the book Matlab Language in
Building Seismic Engineering Application (Zhao-Dong Xu, Guo Yingqing, ed., Science Press).

3.6 Energy in Vibration

Energy is an important physical quantity in vibration analysis, and it is also an important method in vibration analysis. The energy relation and energy
dissipation in vibration process are introduced below.

3.6.1 Energy in Free Vibration

For free vibration, the energy input to an SDOF system has two sources, the initial displacement u(0) and the initial velocity u(0); the former causes
the spring to generate strain energy, while the latter causes the particle to produce kinetic energy. Use E0 to represent the total energy of the initial
time of the system, then:
(3.152)

At any instant of time the total energy in a free vibration system is made up of two parts, kinetic energy EK of the particle and the strain energy ES
produced by the spring deformation.
(3.153)

(3.154)

First, let’s discuss the energy in an undamped system. By substituting the solution of the free vibration of the undamped system Eq. (3.9)
to the expressions of kinetic energy and potential energy (3.153) and (3.154), then:

(3.155)

(3.156)

The total energy is

(3.157)
By substituting (3.155) and (3.156) into (3.157), and using the relation formula , the total energy at any moment of the system can be obtained:

(3.158)

It is proved theoretically that the total energy conservation in the free vibration of an undamped system does not change with time, and it is equal to
the energy input at the initial time. Therefore, the amplitude of vibration is constant. For a damped system, damping always consumes mechanical
energy in the process of free vibration, which results in the decrease of the amplitude of displacement and velocity, and the system must keep the
energy input from an outside source to maintain the vibration. For a damped system, the energy dissipated by viscous damping at 0 to the t 1 moment
can be calculated by the integral of the work done by the damping force.
(3.159)

By Eq. (3.159), it can be seen that different from springs, damping always consumes energy during the vibration of the system, with t 1 → ∞, the total
energy in the system will be dissipated completely by damping, when t 1 → ∞, ED = E0.

3.6.2 Energy Dissipation of Viscous Damped System

The energy dissipation of an SDOF system under a harmonic loading P(t) = P0 sin ωt is analyzed below. WP is the work of the external force in a
vibration cycle; ED is the energy dissipation caused by damping in a vibration loop, which is the work of the damping force; WS is the work of the elastic
force in a cycle; WI is the work of the inertia force in a vibration cycle. Under the action of simple harmonic load P(t), the displacement of a single
degree of freedom system is u(t) = u0 sin(ωt – φ).
(1) Work, WP, of the external force:
(3.160)

Combining Eq. (3.64) , the Eq. (3.160) can lead to:

(3.161)

(2) Energy dissipation ED caused by damping

(3.162)

Using relational expression in the deduction of formula (3.162), it can be seen from (3.162) that the dissipation caused by
viscous damping is proportional to the square of the amplitude, and the dissipation is proportional to the damping ratio, ζ, and the frequency of the
external load, ω. The higher the frequency of vibration is, the more energy will be consumed by damping in a cycle, so high frequency vibration is
damped more easily than low frequency vibration.
(3) Work WS of elastic force:
(3.163)

(4) Work WI of inertia force:

(3.164)

It can be seen that in a cycle of harmonic vibration, the work done by elastic force and inertia force is equal to zero, and the energy dissipated by
damping is equal to the work done by external force.

3.6.3 Equivalent Viscous Damping

We have discussed the classification and calculation method of damping, and we know that damping directly results in the dissipation of energy in
vibration systems. Energy methods are often applied to simplify complex damping into equivalent viscous damping in the analysis of vibration systems.
Viscous damping is an idealized damping and the equation of motion obtained through viscous damping theory is a linear differential equation, which is
simple to solve. Therefore, this theory has been widely used.
The equivalent principle of equivalent viscous damping is that the energy dissipated by viscous damping in a cycle is equal to the energy dissipated by
non-viscous damping force that needs to be simplified in the same cycle. In order to calculate the energy dissipation easily, the equivalent viscous
damping is usually solved under the harmonic load excitation.

1. Hysteretic Curve of Energy Dissipation

The hysteretic curve of energy dissipation refers to the relation curve between damping force and displacement, which is F D – u curve. Substitute the
reaction u(t) = u0 sin(ωt – φ) of the system vibration under the harmonic load into the viscous damping force formula, we can obtain:
(3.165)

From the formula and u(t) = u0 sin(ωt – φ) we can obtain:

(3.166)

As shown in Figure 3.23 (a), it can be seen that the energy dissipation hysteretic curve of viscous damping is an ellipse. The work done by a damping
force in a circle equals the area enclosed by the hysteresis curve, which can be derived directly by the meaning of the work. For the hysteretic curve of
the damping force shown in Figure 3.23, that can be proved directly, the area of the hysteretic curve is:
(3.167)

Figure 3.23 Viscous damping energy dissipation hysteretic curve.

Taking the elastic restoring force F S into account besides the damping force F D, the corresponding hysteretic curve is shown in Figure 3.23 (b), it can
be seen that the ellipse has a stiffness characteristic, that is the diagonal line of the ellipse, and this is the elastic stiffness. It has the characteristics of
both elastic stiffness and viscous damping. It is called viscoelastic property. In the actual structural analysis, the energy dissipation curve is more in
accordance with the energy dissipation characteristics of the actual structure.

2. Equivalent Viscous Damping

The principle of the equivalent viscous damping ratio is based on the principle of equal energy dissipation, i.e., the energy dissipated by equivalent
viscous damping is equal to the energy dissipated by the actual damping force in a vibration cycle.
(3.168)
The equivalent damping ratio ζ eq can be determined by , because k, u0, ω, and ω n are known, we can obtain:
(3.169)

In practical measurement, the structural response is more obvious in resonance. Thus resonance measurements are generally used to represent it.
When the structure resonates, . Sometimes the maximum elastic strain energy can be represented by , equivalent viscous damping
ratio becomes:
(3.170)

Sometimes, the dissipation factor η is used to measure the energy dissipation capacity, which is widely used in viscoelastic damping theory. For single
degree of freedom systems:
(3.171)

The equivalent viscous damping can be easily solved by the equivalent viscous damping ratio:
(3.172)

It can be seen that the energy dissipation of viscous damping is related to the excitation frequency.

3.6.4 Complex Damping

The definition and brief introduction of complex damping have been given in the first section of this chapter. Although the form of the structural
equation of motion derived from the viscous damping model is concise, the test results show that the damping of many structures has nothing to do
with the excitation frequency in a large frequency range. Because of this, the theory of complex damping has been developed, which can be defined as a
damping force that is in phase with the velocity and is proportional to the displacement. As defined in Section 1 of this chapter, the complex damping
force is defined as F D = iηku(t). η is the complex damping coefficient, which is the dissipation factor in the preceding section, k is the stiffness factor. It
can be seen from the expression of the damping force that F D is ahead of the displacement in phase, and its magnitude is proportional to the
displacement, which is a characteristic of complex damping.
The equation of motion for a SDOF damped system under harmonic excitation can be expressed in complex form
(3.173)

In this formula, k(1 + iη) is called complex stiffness, that is:

(3.174)

Eq. (3.173) can be further written as:

(3.175)

Steady-state solution of equation (3.175) can be set up as:

(3.176)

The acceleration is:

(3.177)

Substitute Eq. (3.176) and (3.177) into the equations of motion Eq. (3.175), we can obtain:
(3.178)

Then lead to:

(3.179)

Therefore, the steady-state response is:

(3.180)

u(t) is a complex function. It can be written in the product of its modulus and unit complex number, that is:
(3.181)

Which leads to:

(3.182)
(3.183)

The time domain solutions can pick real or imaginary parts of either Eq. (3.180) or (3.181). (The real part is equivalent to the result of the action of P(t)
= P0 cos ωt, and the imaginary part corresponds to the result of the action of P=P0 sin ωt.) By comparing the above solutions which are under the
harmonic loads of complex damping with those of viscous damping, it can be found that the solution of complex damping theory is exactly the same as
the viscous damping theory when η = 2ζ(ω / ω n).
It can be proved that when complex damping is used, the damping energy consumed in each vibration circle is:
(3.184)

The complex damping theory accords with the experimental results. It is widely used in frequency domain analysis, which is harmonic response
analysis. Viscous damping and complex damping are different in energy dissipation. The energy consumption of complex damping is close to the actual,
while the viscous damping underestimates the energy dissipation capacity of the system when the excitation frequency is low, which is overestimated
when the excitation frequency is higher.

3.6.5 Friction Damping

Friction damping is briefly described in the first section of this chapter too, and its damping force can be expressed as . The direction of
F D is always opposite to the velocity of motion, and the work done for every 1/4 cycle is F Du0, so the energy consumed by Coulomb damping in one
cycle is:
(3.185)

If we equal Coulomb damping to viscous damping, by substituting (3.185) to (3.169), the equivalent damping ratio can be obtained:
(3.186)

3.7 Structural Vibration Test

A structural vibration test is done to analyze the vibration of a dynamic system by means of experiment. By combining the vibration test and
theoretical analysis, the simplified calculation model and calculation method extracted from complex structure system can be verified and refined.
At present, common vibration tests include forced vibration test (including sinusoidal loading test), shaking table test, pseudo dynamic test, impact test
(knock test or impact test), free vibration test, and on-site dynamic model test. For any dynamic test, the most basic equipment can be divided into
excitation equipment, vibration measuring instrument, and data acquisition and analysis system.

3.7.1 Introduction to Vibration Test

1. Sinusoidal Loading Test
A sinusoidal loading test is mainly to study the energy dissipation characteristics and failure mechanisms of either structures or components, and the
common loading devices include electromagnetic exciter and hydraulic actuator. The loading methods include displacement control, force control and
displacement and force mixed control loading. In the sinusoidal loading test, the loading frequency can be varied to investigate the influence of the
vibration system when it suffered different excitation frequency.

2. Shaking Table Test

A shaking table test is to reproduce the phenomenon of structural damage and mechanical properties when it undergoes a random excitation (such as
seismic wave). It can simulate the whole process of vibration, and the mechanical properties of the test structure at all stages, whose loading method is
to input seismic wave records or artificial seismic waves directly.

3. Pseudo Dynamic Test

A pseudo dynamic test is carried out by a computer, this method needs a hypothesis model of the restoring force characteristic; the displacement signal
which depends on the dynamic analysis result of the preset acceleration record is treated as input data. The purpose of the test is to control the
displacement of the structure.
Impact test and free vibration test are on-site dynamic tests (such as the ambient excitation test), and they are often used to test dynamic
characteristics of structures (such as natural frequency, vibration mode, and damping ratio), readers who are interested can refer to the relevant
structural test books.

3.7.2 Exciting Equipment

The vibration excitation equipment can be divided into vibration exciter, vibration table, and impact hammer according to different vibration modes.

1. Vibration Exciter
A vibration exciter is a kind of equipment that makes the measured structure undergo forced vibrations by applying the exciting force through a
connecting rod. It can not only excite the models, but can also excite the actual structures involved in the test. According to its working principle, it is
usually divided into mechanical, electromagnetic, and hydraulic categories.
The mechanical vibration exciter is usually designed depending on the eccentric mass, cam mechanism, and crank connecting rod theories. The
eccentric exciter generates horizontal (vertical) exciting force by synchronous reverse rotation of the symmetrical flywheel with the same eccentric
mass. The cam exciter is driven by eccentric rotation of the flywheel to generate an exciting force. The crank and connecting rod exciter rely on the
crank link mechanism to realize the reciprocating vibration of the exciting connecting rod to generate the exciting force. The common characteristic of
the mechanical vibration exciter is that the excitation force is large, ranging from hundred to thousand N; but their vibration frequency is relatively low
which is usually below 50Hz, and vibration form is harmonic; therefore, mechanical vibration is usually used on vibration tests of large size models.
In an electromagnetic exciter, the moving coil and the exciting rod are connected, and are supported on the electromagnet by the spring. When
alternating current flows through the moving coil under the magnetic force in the magnetic field, it will move back and forth in the magnetic field and
generate the alternating exciting force on the direction perpendicular to the magnetic field line.
Hydraulic vibration exciter is a device that uses hydraulic oil as exciting force, such as hydraulic jack, pulse jack, and hydraulic transmission system
controlled by servo valve. Owing to utilization of hydraulic transmission, the exciting force can be very large, up to several thousand KN. However,
when the frequency becomes high the liquid will generate self-excitation, and the waveform will be disturbed. The low frequency performance of the
hydraulic exciter is very good, and the minimum can reach 1/1000 Hz. Hydraulic exciters are always controlled by a computer, and it can achieve
many kinds of exciting forms, such as harmonic, arbitrary, and random waves. It is widely used in vibration tests of structural models.
2. Shaking Table
A shaking table is a table that is excited by exciters and some connecting rods. Therefore, the vibration table can be divided into three categories:
mechanical, hydraulic, and electromagnetic, according to the types of exciters. Shaking tables can be supported by springs (or connecting rods),
hydraulic devices, air cushion, etc. Vibration conditions of a shaking table are usually required to be the same. Thus, when the structure model is
excited, the stress state of each point on the contact surface would be exactly the same, that is equivalent to the structure being situated on a rigid
foundation. Therefore, the table needs to be rigid, such as a rigid shaking table. The biggest difference between shaking table and vibration exciter is
that the vibration exciter is to vibrate a certain point on the structure model (connection point of exciter and model), the motion state of the points will
be controlled by the exciter, whereas the shaking table can synchronously vibrate all the points on the base surface contacted with the model structure.
Figure 3.24 shows a working block diagram of general electromagnetic vibration tables.

Figure 3.24 Working block diagram of electromagnetic worktable.

With the development of electronic technology, especially computer technology, the shaking table has developed from one dimension, i.e., horizontal or
vertical, to two and three dimensions, which provides better test conditions for the shaking table.

3. Impact Hammer
Figure 3.25 shows a typical schematic of an impact hammer, which has a force sensor and an impact hammer fitted to the sensor. When the structure is
thumped and vibrated, the energy is passed to the structure and makes the structure vibrate in a very short period of time.
Figure 3.25 Typical hammer and impact pulse.
The impact test with hammer is simple, fast, and convenient. However, the force signal has a high crest factor, which makes it unsuitable for nonlinear
systems. The limited incentive bandwidth control is also one of its drawbacks.

3.7.3 Vibration Measuring Instrument

Vibration measuring instruments are also known as sensors. Due to the different types and parameters of vibration measurements, a wide range of
sensors are being developed. According to the working principle, the sensor can be divided into mechanical, electrical, optical, and photoelectric type
and other types. According to the measured vibration parameters they can be divided into strain sensor, acceleration sensor, displacement sensor,
speed sensor, etc.

1. Strain Sensor
Resistive strain sensors are one of the most commonly used strain sensors that converts a change in strain into a change in resistance to measure
response of the structure or the strain. It can measure both the static strain and the dynamic strain. By measuring the strain value, it can convert them
into displacement, velocity, acceleration, or energy. In addition to resistance strain sensors, there are some high-tech strain sensors, such as fiber-optic
strain sensors, wireless strain sensors, and biological strain sensors.

2. Acceleration Sensor
There are two commonly used acceleration sensors, piezoelectric acceleration sensors and servo acceleration sensors. Their basic component is a mass-
spring-damped oscillator system. The working principle of the piezoelectric acceleration sensor is that when it vibrates together with the structure,
under the action of acceleration the mass block in the sensor will generate inertia force to compress the piezoelectric chip. Piezoelectric chip will
generate changing voltage due to piezoelectric effect, and the change of acceleration can be obtained by measuring the change of voltage. The servo
acceleration sensor transfers the measured acceleration into the driving current. The vibration acceleration response can be obtained by measuring
current value. It can be used in low frequency range with high precision and sensitivity, and it works as a precise balance using principle of balance, so
it is also called force balance acceleration sensor. The acceleration sensor is shown in Figure 3.26.
Figure 3.26 Working principle of acceleration sensor.
Under the action of the base motion, the motion equation of the instrument is:
(3.187)

The actual measured acceleration is an arbitrary variation with time. It contains a series of harmonic components in certain frequency band, take the
measurement of a single harmonic component for instance. The acceleration time history of the instrument base is a simple harmonic wave:
(3.188)

In the formula, üg0 is the amplitude of ground motion acceleration. Substituting Eq. (3.188) into the equation of motion (3.187), the relative
displacement of the particle u(t) can be obtained:
(3.189)

For simplicity, only the amplitude of u(t) is discussed below.

(3.190)

From the Eq. (3.189) we can see the relation between the recorded value u0 and the measured acceleration is determined by the dynamic
magnification factor Rd, while Rd is a function of ω. Results of experimental analysis show that, if ζ = 0.7, frequency band satisfies , Rd ≈ 1
and it is constant.
That is, in the range of the above frequency, the amplitude of the instrument recorded u0 is linearly related to the acceleration amplitude to be
measured (That is true for different harmonic motion). In this situation, the response of instrument u can be used to measure the acceleration which is
to be measured. The method of improving ω n is used to ensure each frequency of different harmonic acceleration components satisfies . Due

to , and the mass m is usually unchangeable, procedure of increasing the spring stiffness k is taken to improve ω n. Therefore, the spring
stiffness of the acceleration sensor is relatively large, and the instrument is relatively rigid.

3. Displacement Sensor
The commonly used displacement sensors include mechanical displacement sensors, inductive displacement sensors, and inertial displacement sensors.
Mechanical displacement sensors are also known as the indicator; its principle is simple and won’t be expressed here. The inductance displacement
sensor can convert the vibration displacement into the change of inductance. It is composed of an iron core and a differential coil; when the iron core
vibrates with the structure, the iron core will change the position relative to the differential coil, and it will change the inductance of differential coil
simultaneously. The vibration displacement of the structure can be measured by the variation of inductance. The working principle of the inertial
displacement sensor is a spring-mass-damping SDOF system, as shown in Figure 3.27. It is often used to measure the displacement of the base.
Figure 3.27 Working principle of displacement sensor.
Any harmonic displacement is calculated as:
(3.191)

In the formula, ug0 is the amplitude of ground motion displacement. By substituting into the equation of motion (3.187), relative
displacement of the instrument is obtained:
(3.192)

The amplitude of the relative displacement response is:

(3.193)

Similar to the principle of an acceleration sensor, we expect that in the measured vibration frequency range, the ratio between u0 and ug0 is

close to constant for different frequency components. It is found that when the damping ratio ζ = 0.5 and frequency range , is close to

constant, therefore, an inertial displacement sensor can be used to measure the displacement of the frequency range ω > ω n. To ensure that the
frequency of the measured displacement satisfies ω > ω n, it can be realized by reducing the natural vibration frequency ω n of the instrument. In
practice, it can be realized by reducing the spring stiffness k or increasing the mass m. Therefore, inertial displacement sensors are generally more
flexible. From the above analysis we can see that both acceleration sensors and inertial displacement sensors work in a certain frequency band.

4. Speed Sensor
The common speed sensor is inertial speed sensor, the working principle is that when the sensor is vibrating with the structure or the top rod of the
sensor, while moving coil is connected with the structure. Because of the vibration of the structure, the moving coil of the sensor moves in the magnetic
field, the induced magnetic field is generated by cutting the magnetic lines now. The magnitude of the induced electromotive force is directly
proportional to the moving speed of the moving coil; therefore, the vibration velocity of the structure can be determined by measuring the induction
electromotive force.

3.7.4 Data Acquisition and Analysis System

The traditional method of vibration measurement is to change the continuously changing vibration into continuous voltage signals for display and
recording. The common instruments are pen recorders, ray oscilloscopes, tape recorders, and so on. The continuous changing signals are called analog
signal. The disadvantage of analog signal is the low precision of its display or recording, poor anti-interference ability, and it is inconvenient for further
analysis and processing.
Dynamic data acquisition is used to convert analog signals into digital signals. Digital signals are easy to store, transmit, and analyze, so the signal
conversion device is needed in the acquisition system. More information about electronics and cybernetics is presented here; readers who are
interested in the data collection and analysis system can refer to the relevant books.

3.8 Vibration Isolation Principle

The building or machine’s vibration which is beyond its allowable range affects its normal operation and service life. It is also a form of vibration
pollution, affecting the normal work of the surrounding equipment as well as human health. Isolating vibration effectively is very important. Vibration
isolation in engineering is divided into two cases:

(1) Active vibration isolation:

Prevent vibration output. For example, the operating device can produce oscillation force, and these forces may produce harmful vibration in the
support structure.

(2) Passive isolation:

Prevent vibration input. For example, isolation design in structural seismic problems, vibration isolation of precision instruments and equipment
installed on vibrating structures or foundations.

3.8.1 Active Vibration Isolation

The vibration source is a machine or a device, and the purpose of active vibration isolation is to reduce the force transmitted to the supporting
structure or foundation. In fact, it is the separation of forces, and its mechanical model is shown in Figure 3.28.
Figure 3.28 Single degree of freedom active vibration isolation system.
It is assumed that the unbalanced force produced by the machine is P0 sin ωt. ω is the rotation rate of the machine (angular speed); m is machine mass
(set as rigid mass); k and c are the total stiffness and damping of the isolator; PT is the force that is transmitted from the isolator to the foundation. The
machine is mounted on a SDOF system, as shown in Figure 3.28. The force transmitted through the isolator to the foundation is:
(3.194)

This is a simple harmonic vibration problem, and its solution is u(t) = ustRd sin(ωt – φ). At the same time, the velocity relative to the basement is
, and by substituting them to (3.194), the force on the ground can be obtained:
(3.195)

Then the amplitude of the action force PT is

(3.196)

substitute , c = 2mω nζ into formula (3.196), then:

(3.197)

We call the ratio of the maximum of the force act on the foundation PTmax and the magnitude of the force on the system P0 force transmissibility,
which we can denote by T p:
(3.198)

T P is the amount that reflects the positive vibration reduction effect when , T P < 1. That means improving the frequency ratio of vibration
isolation system (ω / ω n) can achieve vibration isolation. Therefore, in order to achieve the purpose of vibration isolation, the method of reducing the
frequency of the natural vibration circle ω n can be adopted, that is, the vibration isolation effect can be improved by reducing the stiffness of vibration
isolation element or increasing the mass of the instrument. Of course, reducing the stiffness of the isolator or increasing the quality of the instrument
will lead to an increase in the static displacement of the instrument or equipment. The actual vibration reduction design should be selected between the
smallest possible stiffness and the acceptable static displacement.

3.8.2 Passive Vibration Isolation

The vibration source is the ground motion or the movement of the supporting structure. The purpose of passive vibration isolation is to reduce
transmission of the vibration amplitude from the foundation or supporting structure to the superstructure or equipment. If the studied vibration
amplitude is the displacement amplitude, what measures the vibration isolation effect is the displacement transmissibility. If the studied vibration
amplitude is the acceleration magnitude, what measures the vibration isolation effect is the acceleration transmissibility. If the studied vibration
amplitude is the velocity amplitude, the velocity transmissibility measures the vibration isolation effect. It is assumed that the mass of the isolated
vibration m is born by the spring-damped system, which is placed on the base plate. Assume that the plate is subjected to simple harmonic vertical
motion, the vibration displacement of the base (ground) is ug(t) = ug0 sin ωt, as shown in Figure 3.29.

Figure 3.29 Single degree of freedom passive vibration isolation system.

The absolute displacement of mass m is:
(3.199)

In the formula, u(t) is the relative displacement. The requirement for vibration isolation of the base is ut < ug, that is, the vibration of the equipment or
structure is less than the vibration of the base. The relative displacement of mass u(t) can be obtained:
(3.200)

The total displacement of a particle ut(t) is:

(3.201)

At this point, the displacement transmissibility T n is:

(3.202)

It can be seen from (3.198) and (3.202) that the transmissibility of displacement is exactly the same as that of the force, indicating that the two
vibration isolation problems are the same; therefore, their vibration isolation design methods are also basically the same. The curves plotted by the
ratio of transmissibility to frequency at different damping ratios are shown in Figure 3.30.

Figure 3.30 Relationship of transmissibility with different frequency ratio.

The following phenomenon can be found from Figure 3.30:
1. Regardless of the damping ratio ζ, only when the frequency ratio , can transmissibility be less than 1 and vibration isolation effect will be
achieved. The greater the frequency ratio is, the better the effect will be. Because of , reducing the spring constant vibration isolator or
appropriately increasing the mass should be chosen. When the frequency ratio β > 5, the transmissibility decreases slowly, further reducing the
stiffness or increasing the mass cannot significantly improve the effect of vibration isolation, so in practice β=2.5~5 is often adopted.
2. When , transmissibility is larger than 1. That is to say, the isolator amplifies the dynamic response instead, especially when β ≈ 1 (Resonance
region) and the damping is relatively small. The transmissibility reaches a great peak, and the isolator has the completely opposite effect.
Attention must be paid to this and it must be avoided in vibration isolation design.
When calculating the isolator parameters, the transmission rate according to the design requirements can first be determined. Then the frequency
ratio β and damping ratio ζ is determined. Finally, the stiffness k of the vibration isolation spring can be calculated.
The isolation problem of building structures is similar to the single point system discussed above. For example, both of them attempt to improve the
isolation efficiency by reducing the natural frequencies of the system. The difference is that the building structure is a multi-degree of freedom system,
and the study of its isolation efficiency is more complex. Moreover, the ground motion is a wide-band vibration process, and there exists a frequency
component with the same frequency as the natural frequency of the structure. Therefore, we can’t achieve the seismic isolation by avoiding the ground
motion frequency. The structural isolation project covers a wide range, and it is one of the hot topics in the field of disaster prevention and mitigation
engineering as well as structural engineering.
Example 3.8
A reciprocating machine weighs 2 × 104 N, It is known that when the machine operates at a speed of 40Hz, it can generate a vertical harmonic force of
500N amplitude. In order to limit the vibration of the machine to the building, the four corners of its rectangular bottom are supported by a spring. To
reduce all the harmonic forces that the machine passes through the building to 80N, what spring stiffness is required?
Solution: In this case, the force transmissibility is

From formula (3.197), because of damping ratio ζ=0, then

Thus, the total spring stiffness is obtained:

Therefore, the stiffness of each spring in the four supporting springs is:

It is also noted that the static displacement of the support springs due to the weight of the machine is:

Example 3.9
A machine is mounted on an elastic support, The measured natural frequency is fn = 12.5 HZ, damping ratio is ζ = 0.15. The mass that participate in
vibrations is 880kg. It’s known that when the rotation speed of machine is N = 2400r / min, The magnitude of the unbalanced force is 1470 N.
Determine the amplitude, force transmissibility, and the magnitude of force that passed to the foundation.
Solution: From rotation speed N and natural frequency fn, we can obtain frequency ratio

The stiffness of the elastic support is

Thus, the amplitude of the vibration of the machine is

Force transfer rate is

Thus, the amplitude of the force transmitted to the foundation is

3.9 Structural Vibration Induced Fatigue

With the rapid development of modern industry, many machines and equipments are developing towards high speed, high temperature and high
pressure. The factors that cause the failure of engineering structures are also increasing, in which vibration and fatigue are two important factors.
Vibration hazards are manifested in the following aspects: failure of instruments and equipment, ergonomics and health loss of personnel, vibration
intensity of structures, vibration reliability of complex equipment systems, noise and pollution caused by vibration, etc. It is easy to be overlooked in
many circumstances, where there are many accidents.
In the past half-century, many scholars have carried out a series of research studies on the structural vibration induced fatigue problem. Many
achievements have been made, but a large number of results are still distributed in scattered literature, which makes it difficult to form a system for
the structural vibration induced fatigue research. Therefore, understanding the definition and characteristics of structural vibration induced fatigue is
helpful for further study in structural vibration induced fatigue.

3.9.1 Definition of Vibration Induced Fatigue

With the increasing demand for the accuracy of structural fatigue design, fatigue research has gradually developed from material mechanics to elastic-
plastic mechanics and fracture mechanics.
Up to now, there is no consensus on the understanding of vibration induced fatigue in academia and engineering circles, but great importance has been
attached to this problem in practice. Vibration induced fatigue involves the dynamic characteristics of structures, so it is very difficult to define it
accurately. Although many scholars have described the definition of vibration induced fatigue, these definitions do not reveal the dynamic nature of
vibration induced fatigue. Although the term “vibration induced fatigue” often appears in books and other existing literature, even in some standards
and specifications, it is not well explained.

3.9.2 Characteristics of Vibration Induced Fatigue

It is important to know the type of the problem before studying the specific problem. According to the different types of excitation, vibration induced
fatigue can be divided into tension-compression vibration induced fatigue, torsional vibration induced fatigue and bending vibration induced fatigue. On
the contrary, it is called non-resonant fatigue. According to the ratio of excitation frequency to structure fundamental frequency, vibration induced
fatigue is divided into high frequency vibration induced fatigue and low frequency vibration induced fatigue.

1. Resonant fatigue and non-resonant fatigue

Resonance is a synthetical equilibrium phenomenon between external force and structural inertia force, elastic force and damping force. It is
characterized by modal inertia force and damping force occurring in the structure. Damping force is an important factor to determine the size of
resonance response, so resonance fatigue depends on the distribution characteristics of strain modes which play a major role. Resonance types include
global resonance, component resonance and local resonance. Resonance fatigue is more related to component resonance or local resonance. Some
vibration excitations often cause local mode and excitation vibration coupling, and the damage site is usually in the local resonance where strain is large
and defective or stress concentration. Fatigue extensively exists in large-scale structures excited by impact, transient or random vibration. Non-
resonant fatigue is that the excitation frequency of the structure is far from the resonant frequency of the structure. It often exists in the case of single-
frequency vibration excitation or structural stiffness is large but the excitation frequency is low.
The amplitude of structural dynamic response in resonance is much larger than that in non-resonance under the same magnitude of excitation. The
amplitude of resonance fatigue response mainly depends on the magnitude of excitation and damping. A large number of medium-magnitude
excitations induce resonance fatigue failure. Controls of stiffness and damping, mainly contributing to fatigue failure, are small and larger orders of
excitation.

2. High frequency vibration induced fatigue and low frequency vibration induced fatigue
The low frequency vibration induced fatigue mainly studies the fatigue problem that the excitation frequency is less than half of the fundamental
frequency of the structure. Its characteristic is that the damping and inertia of the system are very small, the elasticity of the system plays a major
role, and the system is similar to a pure elastic structure. Therefore, the low frequency vibration induced fatigue failure is mainly controlled by the
elastic strain. The characteristics of fatigue are that the damping and elasticity of the system are very small, and the inertia of the system plays a major
role. The system is similar to a pure inertia structure for quality control.
In engineering, some large-scale structures (such as aircraft) may work in a wide-band excitation environment. Aircraft structures may always remain
resonant with the fatigue growth, which results in frequency-band excitation fatigue. Considering the influence of vibration nonlinearity, the vibration
response of the structure may also have some nonlinear characteristics, such as self-excited vibration. Dynamic, parametric vibration, multi-frequency
response, jumping and synchronization phenomena are prone to fatigue problems that cannot be explained by linear vibration theory.

Exercises
3.1 In the undamped and damped free vibration, what factors are associated with the natural frequencies and dynamic responses?
3.2 As the dynamic system shown in the picture, using jack to make m produce lateral displacement of 30 mm, then the jack is suddenly released, the
system begins to vibrate freely. The lateral vibration was measured 10 mm after 8 periods. Determine: (1) Damping ratio and damping coefficient of
the structure (2) Amplitudes after 20 weeks of vibration (3) If the lateral displacement is 15mm after 8 periods of vibration, while the damping and
mass remain constant, the stiffness must be set at what value?

3.3 Single degree of freedom system, known m = 100kg, k = 6000 N / m. Free vibration begins when initial condition u(0) = 20mm, .
Determine: (1) The natural frequency and period of the system (2) The displacement and velocity when t = 1.0s for two cases of c = 0, c = 200N.s / m.
3.4 The basic power system shown here has the following characteristics: m = 20kg, k = 3 × 104 N / m. If the system is subjected to a resonance
harmonic load(ω = ω n) and the initial state is static, specify the value of the dynamic magnification factor Rd(t) after five cycles(ωt = 10π), (1)
Undamped system (2) ζ = 0.02; (3) ζ = 0.05.

3.5 What are the methods for calculating damping ratio?

3.6 What are the types of impact loads, and what are the basic ideas and procedures for their dynamic responses?
3.7 Use MATLAB language to draw up a procedure that solves structural dynamic response (displacement response and acceleration response) with
Duhamel integral. Single degree of freedom, mass m = 2 × 104 kg, stiffness k = 3 × 106 N / m, is subjected to 0.2g, El, Centro seismic waves.
3.8 What are the basic steps in the frequency domain analysis of structural dynamic responses?
3.9 What are the structural dynamic tests? What are the common excitation and sensing devices?
3.10 A reciprocating machine weighs 2 × 104 N; it is known that when the machine operates at a speed of 40Hz, a vertical harmonic force of 1000N
amplitude is generated. In order to limit the vibration of the machine to the building, the four corners of its rectangular bottom are supported by a
spring. What is the spring stiffness required to limit all the harmonic forces that pass from the machine to the building to the 100N?

References
1. Clough, R.W. and Penzien, J., Dynamics of structures//Dynamics of Structures, 2005.
2. Chopra Anil, K., Dynamics of Structures, Prentice Hall, 2000.
3. Wilson, J.F., Dynamics of Offshore Structures, 1984.
4. De Wachter, S., Tzavalis, E., Detection of structural breaks in linear dynamic panel data models. Comput. Stat. Data Anal., 56, 11, 2012.
5. Cluni, F., Gioffrè, M., Gusella, V., Dynamic response of tall buildings to wind loads by reduced order equivalent shear-beam models. J. Wind Eng. Ind.
Aerod., 123, 2013.
6. Mackie, R.I., Dynamic analysis of structures on multicore computers – Achieving efficiency through object oriented design. Adv. Eng. Softw., 66,
2013.
7. Birman, V., Elisakoff, I., Singer, J., On the effect of axial compression on the bounds of simple harmonic motion, 1982.
8. Zavodney, L.D. and Nayfeh, A.H., The response of a single-degree-of-freedom system with quadratic and cubic non-linearities to a fundamental
parametric resonance. J. Sound Vib., 120, 1, 63–93, 1988.
9. Zhao-Dong, X., Le-Wei, M., Dynamics of Structures, Science Press, Beijing, 2007.
Chapter 4
Multi-Degree of Freedom System
The single degree of freedom system has previously been described in detail. Although most of the structures in engineering practice can’t be reduced
to a single degree of freedom system to calculate, simplifying the structures as a single degree of freedom system will bring about greater error. We can
simplify the water tower and single-story building into a single degree of freedom system, but it cannot be done for multi-story buildings.
It should be noted that a single degree of freedom system and a single mass point system are two different concepts. The reason we say water towers
and single-story buildings are not only single-point system but also a single degree of freedom system is that the mass of the two buildings is mainly
concentrated at the top, so they can be simplified into a single-point system by using the centralized mass method. On the other hand, for a structure, if
we only consider the certain horizontal direction (such as x-direction) vibration and disregard the vertical deformation or y-direction vibration, it can
also be called a single degree of freedom system. It should be noted that not all of the single-point systems are single degree of freedom systems. As
shown in Figure 4.1, the single point m contains two degrees of freedom of horizontal and vertical, and therefore should fall under the multiple degree of
freedom system.

Figure 4.1 Multiple degree of freedom system with one point.

In short, for the mass decentralized structure system, in order to reflect its dynamic performance more accurately, the structure system can be
simplified as a multi-particle system. For example, for a multi-story frame with rigid floors, the mass can be concentrated at the floors of each story.
For multi-span, unequal height, single-story plants, the mass can be concentrated at the roofs of each span. For other structures, such as chimneys, it
can also be divided into several sections according to the calculation requirement, and then every section can be converted into a particle to analyze
(see Figure 4.2).

Figure 4.2 Multi-degree of freedom system.

When a multi-point system is sufferedone-way vibration merely, the number of mass points equals to that of degree of freedom. It should be pointed
out that for the frame structure, the main purpose of emphasizing the rigid floor or shear frame assumptions is to ensure that the simplified particle
only has horizontal displacement without rotation degrees of freedom.

4.1 Two Degrees of Freedom System

Although the two degrees of freedom system is the simplest special case of the multi-degree of freedom system, the analytical method applied in the
two degrees of freedom system is universal in the multi-degree of freedom system. Therefore, the two degrees of freedom system is introduced firstly.

4.1.1 Establishment of Motion Equation of Undamped Free Vibrations

Similar to the single degree of freedom system, the dynamic balance method and the virtual displacement principle described in Chapter 2 are also
applicable to the multi-degree of freedom system. Here, only the dynamic balance method is used as an example to illustrate its application in multi-
degree of freedom systems. For that, an inertial force can be assumed to be exerted on each particle (n all), which transforms the multi-degree of
freedom system into n single degree of freedom systems, and treats the dynamic problem as a static equilibrium problem to solve. The concrete
application can be divided into the flexibility method and stiffness method. Here the flexibility method will first be briefly described, then the way to
establish motion equation by the stiffness method will be introduced in detail.
For the two degrees of freedom system shown in Figure 4.3(a) and (b), the displacement u1 (t) and u2(t) of the masses m1 and m2 can be regarded as
the static displacement generated by combined action of the inertial force F I1 = –m1 ü1 (t) and F I2 = –m2ü2(t). Apply overlay principle results in:
(4.1)
Figure 4.3 Establishment of the equation of motion by elasticity method.
In which Δ11 , Δ12, Δ21 , and Δ22 are the flexibility coefficients of the structure, the physical meaning of Δij is the displacement of point i generated by
the unit force acted at point j (shown in Figure 4.3(c), (d)), according to virtual principle, Δ12 =Δ21 . The above formula contains the flexibility
coefficient, so the above-mentioned way to establish the displacement equation is also called the flexibility method. It can be seen that the flexibility
method is used to establish the motion equations by forming the displacement equations.
The way to establish the motion equation of the two degrees of freedom system by the dynamic balance method of stiffness method will also be
introduced. Figure 4.4 is a shear-type frame structure simplified as a two-degree-free system, when subjected to transient load, it begins to perform
free vibration, taken the particle 1 as a separator part, the effect of the inertial force F I1 and elastic recovery force F S1 subjected on it are respectively:
(4.2)

(4.3)

Figure 4.4 Dynamic balance of two degrees of freedom system.

Similarly, for point 2:
(4.4)

(4.5)

It should be noted that the negative signs at the right end of F I1 , F I2, and F S2, which indicate that the force is opposite to the corresponding
displacement direction.
According to the dynamic balance equation F I+F S=0, the motion equations about m1 and m2 are:
(4.6)

After arranged, there are:

(4.7)
Written in the matrix form, there are:
(4.8)

In which, , the physical meaning of k ij is the counter-force of point j generated by the unit displacement

of point i, as shown in Figure 4.5. Since k ij represents the lateral stiffness of the system, the method of establishing the motion equation described
above is also called the stiffness method.

Figure 4.5 The physical meaning of the stiffness factor.

4.1.2 Natural Frequency and Vibration Mode Shape

The solutions of the differential equations group (4.8) are:
(4.9)

In which, ω n is the self-oscillation frequency, ϕ is the initial phase angle, and X1 and X2 are the displacement amplitudes of the particles 1 and 2,
respectively.
Substituting equation (4.9) into equation (4.7):
(4.10)

The above equation is a homogeneous equations group about X1 and X2. Although X1 =X2=0 is the solution of the equations group, it only means that
the system is in a static state, therefore it is not the solution of free vibration. According to the Cramer rule in linear algebra, if the homogeneous
system has a nonzero solution, its coefficient determinant must be zero.
(4.11)

Expanding it:
(4.12)

Solving it:
(4.13)

These two roots are positive roots; this is because if the two roots are negative real or complex number, then the ω n1 , ω n2 should be imaginary or
complex number, and the u1 and u2 obtained by substituting them into the equation (4.9) will be a hyperbolic function containing the time parameter,
which is obviously contradictory to that the system is stable in the equilibrium position (t = 0) under the effect of restoring force. Now the two positive
roots of the ω n can be obtained, knowing the frequency ω n of the system through Eq. (4.13) is related only to the parameters (mass and stiffness) of
the system itself, but not to other conditions. The smaller one is called the first self-resonant circular frequency or basic natural frequency
(fundamental frequency), the larger one is called the second self-resonant circular frequency. What should be noted is that ω n in this book indicates the
natural frequency of the system, ω ni indicates the i-th order natural frequency of the system, ω indicates the frequency of the system’s excitation.
According to the Cramer rule, the solution of Equation (4.10) is not independent because the coefficient determinant of Equation (4.10) is equal to zero.
Assuming that the corresponding solutions of the ω ni are Xi1 and Xi2 (the first subscript indicates the i-th order frequency and the second subscript
represents the corresponding particle). Although the solution of the system (the displacement amplitude Xi1 , Xi2) is not independent, the ratio is
constant.
Corresponding to ω ni
(4.14)

The particle displacements obtained by equation (4.9) are:

Corresponding to ω n1
(4.15)

Corresponding to ω n1
(4.16)

Substituting equation (4.15) and (4.16) into equation (4.14), respectively

(4.17)

It can be seen that the displacement ratio of the two particles in the vibration process is not only time-independent but also constant. In other words,
at any time during the structural vibration process, the displacement ratio of the two particles is always constant. This type of vibration is usually
called the main vibration mode shape, vibration mode for short, which is only related to the parameters of the system itself and is irrelevant to initial
condition. The vibration mode shape corresponding to ω ni is called the i-th vibration mode shape. In addition, since the main mode depends only on the
relative value between the mass displacements, generally, the displacement value of one of the particles (usually the first or last) is usually set to 1 for
the sake of simplicity. This process is called normalization of vibration mode shape.
In addition, for the first mode, it can be proved that the displacement ratio is always greater than zero, which means that the vibrations of m1 and m2
possess same phase and same direction. For the second mode, the ratio is always less than zero, which means that the vibration of m1 and m2 possess
reverse phase and opposite direction, so there is a point on the curve that vibration does not occur; this point is called the node.

4.1.3 General Solutions of the Equations of Motion

According to the theory of homogeneous linear equations solution space, the general solution of the equations group (4.7) should be a linear combination
of the fundamental solution systems (4.15) and (4.16):
(4.18)

In which, α 1 and α 2 are the combination coefficient, which can be contained in the displacement amplitude X, so the above formula can be simplified as
(4.19)

From Equation (4.17), it is clear that the ratio of displacement amplitude X is constant, so the above equation contains only four arbitrary constants,
X11 , X21 (this is because X21 , X22 can be obtained by X11 , X21 according to the ratio of displacement amplitude), ϕ1 , and ϕ2. They can be determined by
the initial condition and of the vibration. It can be seen that the vibrations of each particle represented by Eq. (4-19) are the
synthesis of two simple harmonic vibrations. It is not a simple harmonic vibration certainly. Only when the ratio of the initial displacement of each
particle is equal to that of the main mode and the initial speed is equal to zero, can the system follow the main vibration mode. Under normal initial
conditions, the system does not implement simple harmonic vibrations, and the ratio of displacement between particles is no longer constant and its
value will change over time.

4.2 Free Vibrations of Undamped System

4.2.1 Establishment of Motion Equation
We can generalize the form of the two degrees of freedom equations of motion to a multi-degree of freedom system. In order to facilitate expression
and computation, being similar to equation (4.8), the motion equations of the undamped multi-degree of freedom system are expressed in the form of
the following matrix:
(4.20)

In which,
Now a five degrees of freedom system (shown in Figure 4.6) is served as an example, the physical meaning of the coefficient k ij contained in the
stiffness matrix [K] is shown as follows:
(4.21)

Figure 4.6 Five degrees of freedom of system.

Being similar to equation (4.9), the solution of the equations group (4.20), the displacement of each particle can be expressed as follows:
(4.22)

In which:

Substituting (4.22) into equation (4.20) and noting that it is arbitrary results in:
(4.23)

The condition that the above equation possesses the nonzero solution is:
(4.24)

The natural frequency of the system can generally be obtained by solving the generalized eigenvalues of equation (4.23). According to the linear
algebra, we can see that when the mass matrix [M] is positive definite (the order principal minor determinant of [M] are all greater than zero), the
stiffness matrix [K] is also positive or semi-definite (the all order principal minor determinant of [K] are greater than or equal to zero), then all the
roots of the equation (4.23) of are positive real or zero. Those roots are called the eigenvalue of the above equations group. ω n is also called the
natural frequency of the system. Generally, the natural frequencies are arranged from small to large, the smallest of them is called the first frequency
or fundamental frequency of the system, and the corresponding period T 1 = 2π / ω n1 is called the first cycle or the basic period, since the first
frequency is the minimum frequency of the system, it means that the basic period should be the longest period. There are n natural frequencies for n
degrees of freedom. Those natural frequencies are only related to the parameters of the system itself, independent of other parameters such as
external loads, boundary conditions, and so on.

4.2.2 Vibration Shape and Its Orthogonality

Since the coefficient determinant (4.24) of the equations group (4.23) is equal to zero, the solution {X} of the equations group (4.23) is not independent.
Therefore, it is necessary to substitute the obtained natural frequency into equation (4.23) and let Xij= 1 (generally take j = 1 or j = n), so that the
equations (4.23) can be transformed into a non-homogeneous algebra group, the solution of the above equations, we can get the i-type main vibration
shape corresponding i-order frequency ω ni by solving the above equations.
(4.25)

According to the linear algebra, we can see that the above-mentioned i-th main mode is the eigenvector corresponding to one of the eigenvalues of
the equations group (4.23).
The so-called orthogonality of vibration shape refers to the following relationship existing between any two different mode vectors {X}i and {X}j
corresponding to the two different frequencies ω ni and ω nj in a multi-degree of freedom system or an infinite degree of freedom system.
(4.26)

(4.27)

Now it is proved as follows.

Proof Method 1:
According to equation (4.23):
(4.28)

(4.29)

And the transpose vectors of the main vibration shape of the j-th order are simultaneously applied to both sides:
(4.30)

Noting that:
(4.31)

According to the transitive rule of the linear algebra, the two ends of the equation (4.30) are transposed at the same time:
(4.32)

Similarly, corresponding to the frequency ω nj, both ends of the formula (4.28) are previously multiplied by {X}j first and multiplied by second,
which can obtain a resembling equation (4.30):
(4.33)

Subtracting equation (4.33) from equation (4.32) results in:

(4.34)

Due to ω ni ≠ ω nj:
(4.35)

Similarly, the orthogonality of the stiffness matrix [K] can be proved. Certainly, there are many ways to prove the above theorem, such as the
equivalence law of work of Betti’s law, as shown in Proof Method 2.
Proof Method 2:
According to the equivalence law of work, the work done by one set of loads on the deflections due to a second set of loads is equal to the work of the
second set of loads acting on the deflections due to the first.
Now the above-mentioned law is illustrated by the orthogonality between the first and second vibration shape of the three degrees of freedom, in
which the mode is used as the displacement and the inertia force, , as the force. The situation corresponding to the first mode is assumed as the
first state (Figure 4.7a); the situation corresponding to second mode is assumed as the second state (Figure 4.7b). Then the work done on the second
state displacement by the first state force is:
(4.36)
Figure 4.7 The reciprocity theorem of work.
And the work done on the first state displacement by the second state force is:
(4.37)

According to Betti’s law of work reciprocity, W12 = W21 , results in:

(4.38)

Since equation (4.37) can also be rewritten as follows:

So:
(4.39)

So equation (4.38) can be written as:

(4.40)

Because of :

Prove finished
Proof Method 3:
When the multi-free motion equation is established by the flexibility method:
(4.41)

In which, the flexibility coefficient matrix of system [δ] = [K]–1 .

Expanding the above formula in the form of equations group, corresponding to the frequency ω nj and amplitude of Xji:
(4.42)

After deforming the above equation, the expression of the amplitude X corresponding to ω nj and ω nk can be expressed as:
(4.43)
(4.44)

Multiplying the first formula, second formula … n-th formula of the equation (4.43) by ,… respectively, and then adding them
together, similarly, multiplying the first formula, second formula and n-th formula of the equation (4.44) by ,… respectively,
and then also adding them together, obviously, the right of the two above equations is exactly identical, so the left should also be equal, resulting in:
(4.45)

Because of ω nj ≠ ω nk, it can be obtained that:

(4.46)

The above equation is equation (4.35), proof finished.

4.2.3 Generalized Mass and Generalized Stiffness

For the equations (4.26) and (4.27), the precondition of the mode orthogonality is i ≠ j, if i = j, according to the property of determinant operation,
and will be unequal to zero. Mk and Kk are called generalized mass and generalized stiffness respectively corresponding to
the k-th mode, defined as follow:
(4.47)

(4.48)

Both sides of the equation (4.29) are multiplies by , it becomes:

(4.49)

According to the definition of generalized mass and generalized stiffness, the above equation is , therefore:
(4.50)

It can be seen that the corresponding frequency can be obtained directly by generalized mass and generalized stiffness.
For the n degrees of freedom system, the n×n order square matrix can be formed by ranging the obtained n modes {X}i from left to right, this square
matrix is called the vibration matrix. That is:
(4.51)

With this substituted into Eq. (447), it becomes:

(4.52)
The above equation is called the generalized mass matrix. Similarly, the generalized stiffness matrix can be obtained by using the modal matrix. That
is:
(4.53)

Example 4.1 As shown in Figure 4.8, a two-layer shear-type frame structure, each layer of mass is known, which is m1 = 60t, m2 = 50t, the first
layer’s lateral stiffness is k 1 = 5×104 kN/m, the second floor of the lateral stiffness is k 2 = 3×104 kN/m. Solve the problem of the system’s natural
frequency and vibration of the structure, and verify the orthogonality of its main mode.

Figure 4.8 Two-layer shear frame structure and vibration mode.

Solution: According to (4.8), the inter-layer stiffness coefficients of the frame layers can be obtained respectively:

From equation (4-12), the frequency equation is given as follows

Expanding the above equation, there are

Solving the above equation, there are ω n1 = 17.54rad/s, ω n2 = 40.32rad/s.
Now, the corresponding periods are respectively:

According to equation (4-17):

For the first vibration shape:

For the second vibration shape:

The above modes are shown in Figures 4.8(b) and (c), respectively.
Next, the orthogonality of the main mode will be checked, for the mass matrix and stiffness matrix, according to equation (4.26) and equation (4.27), it
can be obtained that:

The generalized mass and generalized stiffness are given by (4.47) and (4.48) as follows:

4.3 Practical Calculation Method of Dynamic Characteristics

The dynamic characteristics of the structure consist of the physical quantities such as the natural frequency, period, and mode of the structure. In the
dynamic analysis of the structure, determining the dynamic characteristics of the system should be done first. If the dynamic characteristics are
calculated only by analytic method, the calculated workload will be very heavy when the degree of freedom of the system exceeds three. For example,
from the formula (4.13), even for a two degrees of freedom system, the analytical formula of its natural frequency has been very complicated. In
addition, it is even unable to be solved using the exact method for some complex structure sometimes. So, in order to effectively deal with such a wide
range of computing problems, we need to resort to the simplicity and accuracy of the approximate calculation method, which is the reason why
dynamic characteristics of the practical calculation method is proposed. Here, some conventionally used calculation methods will be introduced.

4.3.1 Dunkerley Formula

It is important to estimate the fundamental frequency of the system quickly in the dynamic analysis. Therefore, the Dunkerley formula for estimating
the first frequency is introduced.
Taking the three degrees of freedom system as an example, the system characteristic equation established by the softness method is:
(4.54)

In which, δij is the flexibility coefficient of system, the expanded formulary of which is a cubic algebraic equation about .

(4.55)

Assuming that the three roots of the equation are , an equation can be written as:

(4.56)

Expanding it:
(4.57)

Comparing equation (4.57) with equation (4.55) results in:

(4.58)

Since the high frequency part of the engineering practice is much higher than that of the fundamental frequency, the approximate formula for the first
frequency of the system can be obtained by ignoring the high order frequency at the left end of the above equation.

(4.59)
It is the basic frequency calculation formula given by Dunkerley, which is the general form of Dunkerley formula for n degrees of freedom system.
(4.60)

Because of , the above equation can also be written as:

(4.61)

Dynamic analysis often requires a quick estimate of the fundamental frequency of the system when the quality or stiffness parameters of the system
changes. The Dunkerley formula can give a convenient calculation for this. If the fundamental frequency of the original multi-degree of freedom system
is ω n1 and the increment of each mass is Δmi, then according to the Dunkerley formula, the base frequency ω ʹn1 of the system after the mass increased
is
(4.62)

According to the formula (4.59), it is obvious that the Dunkerley formula is accepted by subtracting the high frequency term at the left end, so the
obtained fundamental frequency is lower than the actual value. In addition, the result from the Dunkerley formula is more precise when the frequency
of the higher-order mode of the strucyure differs greatly from its fundamental frequency.. Contrarily, the deviation of Dunkerley formula is greater
when frequencies of structure are close. So, for the dense spectrum of large-span bridges, large-span grids, and other structures, the calculation
obtained by the Dunkerley equation is less accurate.
Example 4.2 For the simply supported beam shown in Figure 4.9, if , find the fundamental frequency of the system by the
Dunkerley formula.

Figure 4.9 Three-particle simply support beam structure.

Solution: According to the figure multiplication, it can be obtained that the flexibility coefficient of system , substituting
them into the formula (4-59):

Solving it, results in , the more accurate result is with the deviation is -4.15%.

Example 4.3 As shown in Figure 4.10, it is known that the masses of the unequal high single-story plant are m1 = 59200N, m2 = 50000N. The
softness coefficients are δ 11 = 2.13×10–4 m/kN,δ22 = 5.65×10–4 m/kN. In addition, the base frequency of the plant has been obtained, ωn1 =
5.16(rad/s). If the mass m1 of the plant has increased by 20000N, estimate the base frequency of the plant after the mass changed.

Figure 4.10 Unequal height single story plant.

Solution: According to equation (4-62),
Solving it, ω ʹn1 ≠ 4.90(rad/s). The exact solution after the parameter changed is 4.95s-1 , and the estimated error is -1%.

4.3.2 Rayleigh Energy Method

According to the law of conservation of energy, if the energy loss of the system in the vibration process is neglected, for example, ignoring the damping
effect, the sum of the potential energy and the kinetic energy of the system will remain a constant at any time. As in the static equilibrium position, the
potential energy of the system is zero, so the system obtained the maximum kinetic energy at this time, recorded as Umaxo. Similarly, at the moment
that the system reaches the maximum displacement, because the speed is zero, the kinetic energy is zero, so all the energy switches into potential
energy, recorded as Wmaxo. From the energy conservation theorem we know:
(4.63)

(4.64)

In which, Y(x) is vibration mode shape, ω n is frequency. The above equation is the partial derivative of time t, and the vibration velocity of the system
is obtained. That is:
(4.65)

According to it, the kinetic energy of the system is obtained. That is:
(4.66)

When cos(ω nt – φ) = 1, the system obtained the maximum kinetic energy.

(4.67)

Figure 4.11 Simplified calculation of simply supported beam.

If only the bending deformation is consedered, the strain energy of the system is:
(4.68)

Similarly:
(4.69)

Substituting equation (4.67) and (4.69) into equation (4.63). It can obtained that:
(4.70)

In addition to the distribution mass m(x), if the system also possesses n concentrated masses (Figure 4.12). In this case, the kinetic energy of the
system should not only include the kinetic energy of the distribution mass m(x) as shown in equation (4.66), but also the kinetic energy of concentrated
mass, which is:
(4.71)
Figure 4.12 Simplified calculation of the beam with concentrated mass.
So, the maximum kinetic energy of system is:
(4.72)

And the equation (4.70) can be rewritten as:

(4.73)

It can be noted that the natural frequency of the system can be easily acquired simply when the vibration form Y(x) of the system is known. If the
displacement shape function Y(x) that has been assumed is exactly the same as the first principal mode, the approximation of the first frequency of the
structure can be obtained. If it is close to the second mode, the approximation of the second order frequency can be obtained. But the problem is that
the displacement function shape Y(x) is not known in advance. However, for the first frequency, according to some practical calculations, it can
illustrate that as long as the assumed mode curve satisfies the boundary condition of the system and roughly conforms to the fundamental vibration
form, the error of the natural frequency achieved by substituted which into the above equation will not be great. If the assumed mode function is
accurate, the natural frequency obtained is also accurate. In general, it is difficult to accurately assume a high-order mode function, so Rayleigh
(Rayleigh) is only used to calculate the basic frequency. Although the use of the Rayleigh method for high-order frequency effect is not good, for the
symmetrical structure, it is easy to give its symmetrical mode and anti-symmetrical mode, the corresponding two natural frequencies can be obtained
easily.
In order to solve the first frequency, Rayleigh suggests that the static displacement generated by the self-weight load on the system m(x)g can be
treated as the first mode Y(x) (note that if considering horizontal vibration, gravity should act in the horizontal direction.) So the work that it has done
can be expressed as
(4.74)

Thus, for the structural systems with both distributed and concentrated mass, equation (4.73) can be rewritten as:
(4.75)

Of course, it is also possible to take the elastic expression of the structure under the action of a static load q(x) as the approximate expression of the
mode curve, so that the above equation can be rewritten as:
(4.76)

Example 4.4 Solve for the first frequency of the simply supported beam by energy method (EI is constant, the distribution mass is ).
Solution:
(1) Suppose that the first-order mode curve Y(x) of the simply supported beam is a parabola.

when x=0, Y(0)=0; when x=l, Y(l)=0, It can be seen that this parabolic satisfies the boundary condition:

According to equation (4.73):

So, there is

(2) The deflection curve under uniformly distributed load q is taken as the mode shape curve Y(x):

Substituting it into equation (4.73) results in:

So:

(3) Assume that the shape function is sinusoidal

Substituting it into equation (4.73) can similarly obtain

It can be seen that since the sinusoidal curve is the exact solution of the first principal mode, the frequency obtained from it is the
exact solution of the first frequency. In addition, the first two selected curves meet the boundary conditions almoatly, so the deviation of the results are
minimal, and they are larger than the exact value. This is because the assumed curve is not the real vibration curve, which is equivalent to adding some
extra constraints to the structural system. Then the stiffness of the system is increased, so the resulting frequency will be larger.

4.3.3 Ritz Method

It can be seen from the previous example that the accuracy of the first frequency of the system obtained by the energy method depends on the
accuracy of the assumed mode. Furthermore, only the upper limit of the fundamental frequency of the vibration can be acquired. In order to find the
approximation of the higher order frequency and to make the lowest frequency close to the exact solution, the Rayleigh method was developed. The
Ritz method is based on the Hamilton variation principle, which transforms the variation problem into the extreme value problem of multiple variable
functions.
Hamilton’s principle is a basic variation principle in analytical mechanics. It provides the criterion for judging the real and actual movements from all
movements that are possible and satisfying the constraint situation. For the actual movement, the kinetic energy U, potential energy W and virtual
work δW of the elastic system must meet:
(4.77)

For free vibration problem of the periodic (integral upper and lower limits take a cycle, that is 0 and 2π / ω), no damping structure of the system free
vibration problem, the virtual work of above formula is zero. The Hamilton principle can be expressed as: in all possible motion state, the exact solution
makes
(4.78)

Substituting equation (4.66) and equation (4.68) into the above formula, because the time integral range is taken as a period, results in:
(4.79)

Let Y(x) be the assumed mode curve and decompose it by the mode:
(4.80)

In which, y(x) are n independent displacement functions which satisfy the displacement boundary condition, a is undetermined parameter.
Substituting the above equation into equation (4.79), the equation becomes:
(4.81)

if , then

(4.82)
Using the stationary condition: , then:

(4.83)

Writing it into matrix form:

(4.84)

According to the law of Kramer, if the equations group has nonzero solutions (because the parameter ai are not all zero), the coefficient determinant
should be equal to zero, thus
(4.85)

Since the above equation is an n-th algebraic equation about the , an approximation of the first n natural frequencies can be obtained.
Example 4.5 Solve out the natural frequency of the equal section cantilever beam (shown in Figure 4.13) by using the Ritz method.

Figure 4.13 Cantilever beam structure.

Solution: Assume that the approximate mode function is:

By equation (4-81), the constant k ij and mi can be obtained. As shown in the following:

Therefore, the frequency equation is:

Namely,

the roots of the equation are

In order to improve the computational accuracy of ω n2, the following four functions are available.

In which, the first two terms y1 (x), y2(x) are in compliance with equation (a), and y3(x), y4 (x) are respectively expressed:

The first two order frequencies of the structure can be obtained as:

It can be revealed that it would be better to take more items in progression of hypothetical modes if a more accurate value needs to be obtained.

4.3.4 Matrix Iteration Method

The matrix iteration method is also called the Stodola method or power method, this method calculates the frequency and mode of the structure by the
gradual approximation method, which can be used to solve out the first few modes and frequencies of the structure.
As mentioned earlier, the free vibration equation (4.20) of a multi-degree of freedom system can be expressed as:
(4.86)
Multiplying the above equation by gives:

(4.87)

Let , the above equation becomes:

(4.88)

Now assuming that {X}0 is the first approximation of the first mode, and performing regularization (which suppose that the amplitude of one particle is
1, usually the amplitude of the first or the n-thparticle), substitutes it into the left side of the above equation and makes:
(4.89)

If {X}0 is the real solution of the first mode, it must be:

(4.90)

If the above equation is invalid, again let:

(4.91)

This iteration process is repeated until the two adjacent iterative results are similar.
It should be pointed out that the physical meaning expressed by the column vector of the mode is the relative displacement between the particles. {X}
is not the absolute value but the relative value, because of this, the mode {X} should be performed normalize before each iteration so that it is
convenient to compare between the two modes before and after the iteration. It is also more efficient to obtain the true value.
Example 4.6 A system with three degrees of freedom, its mass matrix [M] and stiffness matrix [K] is respectively:

Solve out the first-order frequency and mode of the structure by matrix iteration method.
Solution:
Because of

There is

Assuming {X}0 = {1 1 1}T , substituting it into equation (4.89), there are

After performing regularization of shape, there is:

here, now because of , the second iteration is needed to be performed.

At the second iteration, first let {X}0 = {0.717 0.953 1.000}T , and then substituting the equation (4.89) into it:

After the normalization, , repeating the above process, after the third iteration, there is ,
the result is very approximate , so the approximate solution of the first-order model is:
So from equation (4.87), there is:

Noting that the above equation is actually three independent equations, it can be solved according to any type of solution, such as by type 3:

so ω n1 , = 8.89rad/s
Then it will prove that the frequency and mode obtained by the above-mentioned iterative method are the first frequency fundamental frequency and
the corresponding mode.
From the equation (4.89) we can see that after two times iterations, the mode vector, (the subscript indicates the number of iterations), is:

Similarly, after k times iterations the equation becomes:

Since the assumed {X}0 can be expressed as a linear combination of the true mode vector {Xi} of the system:

In which, ai is a constant, {Xi} is the i-th mode of the system.

So, according to [β]{Xi} = α i{Xi} (i represents the i-th mode, ), resulting in:

Similarly:

Dividing both ends of the above equation by becomes:

Because of 0 < ω 1 < ω 2 < … < ω n, thus α 1 > α 2 > … >α n > 0, which means that when k is large enough, (α i / α 1 )k ≈ 0, therefore:

It can be seen that after k times iterations, the difference between the [β]k{X}° and the exact value {X1 } of the first mode is only a constant term ,
and the normalization done after each iteration can eliminate the influence of this constant term. So, after finishing many repeated iterations:

The mode iteration method can not only solve the fundamental frequency and first order mode of the system, but also can be used to determine the
high frequency and the corresponding vibration mode. However, in order to do this, it is necessary to deal with the assumed iteration vector
appropriately. Which means that in order to decide the second mode, the effect of the first mode must be eliminated in the assumed iteration vector.
When the third mode is determined, the effects of the first and second modes should eliminate the effects of the first and second modes, and so on.
Based on this law, before performing the iteration, let a1 = 0, the final result of the iteration will converge to the second mode; Let a1 = a2 = 0, which
will converge to the third mode, and so on. Therefore, the concrete way to obtain the system i-order frequency or mode by iteration method is as
follows:
Premultipling the linear combination of the mode vector {Xi} by {Xi}T [M], and using the orthogonality of the mode results in:

So:

In order to eliminate the preceding r-order mode component in the assumed mode, the initial iteration vector is taken as , substituting

aj into it becomes:

In which, the is the clearing type matrix.

In the actual iteration process, in order to avoid that the vibration may contain the former r-order mode components, it is required that which should
be premultiplied by the matrix has been cleared each time. In other words, when the first frequency of the system is obtained by the matrix iteration
method, the process can be carried out according to equations (4.88) to (4.91). For solving the higher order frequencies, replacing [β] in the formula
(4.88), then it can be obtained by performing the matrix iteration.
Example 4.7 Solve out of 2nd shape and frequency of example 4.6
Solution:
Because that {X1 },[Μ],[K],[β],α 1 have been known:

Assuming that the initial approximation of the second-order mode is {X2}0 = {1.00 0.00 – 1.00}T , executing the iteration by the matrix iteration
method of Example 4.6, in which [β] takes [β]1 .
So the resultant is {X2} ≈ {0.98 – 0.49 –1.00}T, ω 2 ≈ 27.20rad/s.

4.3.5 Subspace Iteration Method

The subspace iteration method is also called the parallel iteration method, which is essentially a repeated use of the Ritz method and the matrix
iteration method for a set of test vectors. Matrix iterations can only find one eigenvalue and eigenvector of matrix by one iteration, but subspace
iteration can find the first few largest eigenvalues and eigenvectors of matrix by one iteration. The basic principle is that the original n linearly
independent eigenvectors (modes) {X}1 , {X}2, …,{X}n of the matrix [X] compose an n-dimensional vector space. The p (p <n) linearly independent
vectors form n’s p-dimensional subspaces {Y}1 , {Y}2, …, {Y}p, Then, the Ritz method and the matrix iteration method are used repeatedly, so that the
low-order mode components of the p test vectors are continuously amplified and eventually approach to the subspace formed by the low-order
eigenvector {X}1 , {X}2, …, {X}p.
It should be pointed out that if only the p-dimensional subspace vector is iterated without orthogonalization, {Y}1 , {Y}2, …, {Y}p will eventually
converge to the first-order mode. Therefore, the orthogonalization process of the mass matrix [M] must be done in the iterative process, so that {Y}i
approaches to {X}i.
The specific process of subspace iteration is that if the first set of p eigenvalues of the [X] and the corresponding eigenvalues are required (p<n), let p
linearly independent vectors {X}1 , {X}2, …, {X}p constitute an n×p initial matrix and the initial matrix [Φ]0 is:
(4.92)

Premultiplying the above equation by matrix [β] = [K]–1 [M] results in:
(4.93)

Performing orthogonalization for [ψ]1 , so that the columns of the vector tend to different levels of vibration after iteration, respectively, rather than
trend to the first mode, for this reason let
(4.94)

In which, [Z] is undetermined coefficient matrix that can be expressed as [Z]=[{Z}1 , {Z}2, …,{Z}p], by using the generalized mass and generalized
stiffness matrix
(4.95)
The original problem is reduced to the p × p order eigenvalue problem:
(4.96)

Because of p«n, it can obtain the first approximation of the p-order natural frequency and the corresponding fixed-coefficient vector,
therefore it can obtain the first approximation of the p-step mode of the system:

Similarly, obtaining the [ψ]2 by premultiplying [Φ]1 by [β], and suing the second approximation of the undetermined coefficients
vector obtained by Ritz method, after repeating the iterative process, the calculated results will converge to the pre-p-mode frequency of the system.
Example 4.8 It is known that m1 = m2 = m3 = 1, all of the lateral stiffness of the floor is k = 1. Solve out the second-order mode by subspace
iteration.
Solution:

so, there is

The initial approximation of the first two modes is

Using

After performing normalization:

Solving generalized mass modes and generalized stiffness matrices:

If there is a non-zero solution for equation , there should be:

Solving it gives:

For ω n1 , there is Z 2 = –0.0002Z 1

For ω n2, there is Z 1 = –0.351Z 2
So:

Thus, the first approximation of the first and second order mode is

Then, after performing normalization:

[ψ]2 can be obtained by premultiplying [Φ]1 by [D],

After performing normalization:

The second approximation of generalized mass mode and generalized stiffness matrix is:

It is attributed to the Ritz eigenvalue problem again, if there is a nonzero solution for equation , then:

If there is a non-zero solution, then:

Solving it gives:

For ω n1 , there is Z 2 = –0.0606Z 1

For ω n2, there is Z 1 = –0.0002Z 2
So:

Thus, the second approximation of the first and second order mode is

Performing normalization results in:

The difference between the solution , obtained by the second-round iteration and the real solution is very small.

4.4 Mode Superposition Method for Damped System

For a multi-degree of freedom system, if the viscous damping is considered, the differential equations of the forced vibrations are expressed in the
form of a matrix:
(4.97)

There are two general methods to solve this motion equation. One is the direct integration method, which performs numerical integration directly on
the above differential equation according to the time course, that is numerical solution, and the commonly used numerical solutions are the central
difference method, Newmark method and Wilson-θ law. The other is the mode superposition method. In order to illustrate the superposition of
vibration shape, the following concepts will be explained first.

4.4.1 Coordinate Coupling and Regular Coordinates

From the equations (4.8) about two points given above, we can see that since the unknowns (coordinates) u1 and u2 of the equations group are coupled,
means that we must perform connecting solve to obtain the solution of u1 and u2. The equation is called coordinate coupling equation. The coordinate
coupling can be divided into stiffness (static) coupling, and inertia (acceleration, power, or mass) coupling. Therefore, the less the number of differential
equations is coupled, the simpler the solution of the equation is. For this reason, the concept of generalized coordinates (independent coordinates) can
be presented to decouple the differential equations.
For a system with n degrees of freedom, if the vibration shape of the system has been known, the linear transformation between the current
coordinate {q} and the original physical coordinate {u} can be introduced by a new set of coordinate {q} = {q1 ,q2, …,qj,, …,qn}T, that is
(4.98)

Written in the form of equations group:

(4.99)

In which, ui is the displacement of point i, that is the solution of the differential equations, Xij is the displacement amplitude of i caused by the vibration
shape of j. qi is the generalized coordinate corresponding to the shape of i, also known as the regular coordinate or the mode coordinate.
It should be noted here that since ui is a function of time and mode, Xij is a coefficient. qi should also be a function of time t. It can be seen that the
physical meaning of the linear transformation of (4.99) is very clear. It regards the physical coordinate ui as the linear superposition of each mode, and
qi is the participation factor or weighting factor of each mode.
Because of that, the equations group (4.99) can also be written as:
(4.100)

Premultiplying the both sides of above equation by gives:

(4.101)

Due to the orthogonality of the vibration of the quality matrix, at the right side of the above formula, only is unequal to zero, the other
items are all equal to zero, therefore:
(4.102)

Thus:
(4.103)

The above equation is the relationship between the generalized coordinates qi(t) and the actual displacement {u(t)}.

4.4.2 Damping Assumptions

For viscous damping, it is assumed that the magnitude of the damping force is proportional to the velocity of the particle vibration. The direction of the
damping force is opposite to the velocity. In the multi-degree of freedom system, each particle is vibrating, and the damping force acting on the particle
i is affected not only by the vibration velocity of this point, but also by the vibration velocity of the point j. Therefore, the damping of the multi-
degree of freedom system gives the matrix [C] as:
(4.104)

In which, c ij is the damping force of point i generated by the j-point unit velocity.
Through the above analysis, it can be indicated that the mode of the system is orthogonal to the mass and stiffness, but the damping matrix does not
satisfy the orthogonal condition. Therefore, at normal circumstances, it can only obtain a set of mutual coupling differential equations about the mode
coordinate. In order to facilitate the decoupling of the equations group, the modes should be orthogonal to the damping matrix. The damping
assumption of the two degrees of freedom systems will be presented next.
(1) Rayleigh Damp
Assume that the damping matrix [C] is a linear combination of the mass matrix and the stiffness matrix, thus, the system is a proportional damping or
Rayleigh damping, that is:
(4.105)

In which, α, β are proportional constants. Meanwhile, the modes of the system are orthogonal to the damping matrix, therefore:
(4.106)

in which, is the generalized damping coefficient of the j-th order mode. The damping ratio of the j-th mode can be obtained from the below equation.
(4.107)

In addition, if the first and second order frequencies of the system and the corresponding damping ratio are known, the proportional constant α, β can
be obtained by using the following equations.
(4.108)

Solving it:
(4.109)

After both α, β have been determined, substituting them back into equation (4-107) will obtain the natural frequency and damping ratio of the higher
order mode according to the each-order natural frequency.
(2) Cauchy Damp
Cauchy Damp is also known as the extended Rayleigh damp. By comparing the above Rayleigh damping, it can be known that the proportional constant
α, β are determined by the first and second order frequencies and the damping ratio. The higher order damping ratio is determined by equation
(4.107). which means that Rayleigh damp may only satisfy equaling to a given damping ratio at two natural frequency points, whereas higher order
damping is generally not consistent with the actual measured results. In addition, the higher the order of chosen vibration modes is, the greater the
error is, which is an inadequacy of Rayleigh damp hypothesis. Therefore, if it is required to satisfy equaling a given damping ratio at more frequency
points, it is necessary to construct a linear combination with more items. The damping matrix constructed by the following form can realize this aim,
which suppose the damping matrix is the matrix that is proportional to Cauchy matrix series, thus, the damping matrix can be taken as follows:
(4.110)

In which, α 0,α 1 , …,α n–1 are n undetermined constants. By using the generalized orthogonality, it can prove that the Cauchy damping matrix satisfies
the orthogonality, thus, the following formula is established when i ≠ j:
(4.111)

When i = j, the general diagonal elements in the generalized damping matrix [C] is:
(4.112)

It can be obtained by the above equation that:

(4.113)

Substituting the known n-order natural frequency and the measured n-order damping ratio into the above equation, n linear equations and n
undetermined constants can be obtained by simultaneous solution of those equations. This demonstrates that the assumed damping ratio and the
actual measurement results are completely consistent.
Example 4.9 The damping ratio of the first three order modes ratio of a three degrees of freedom system have been known, ζ 1 = ζ 2 = ζ 3 = 0.05,
and the mass matrix of the structure and the stiffness matrix are as follows. Solve out the damping matrix.

Solution:
The first three order frequencies of the structure are obtained by using the equation: ,

(1) Assuming to be Rayleigh damping

According to (4.109):

So, substituting it into the equation (4.105):

Using equation (4.107), it can obtain the third order mode ratio

It can be seen that there is a large difference (27.20%) between the third vibration damping ratio (basic damping 0.0636) and the actual measured
value (0.05) obtained by the damping ratios of the first and the second vibration mode.
(2) Assuming to be Cauchy damping
Applying the equation (4.113), it can obtain equations group as following

Solution α 0 = 0.847, α 1 = 0.00284, α 2 = –4.96×10–7 and then, according to equation (4.110):

Substituting the known conditions into it can obtain:

4.4.3 Mode Superposition Method

Using the Rayleigh damping assumption, the equation (4.97) can switch into:
(4.114)

Substitution equation (4.98) in above equation gives:

(4.115)

Per-mutiplying the above equation by , and using the principle of orthogonal modes results in:

In which, Mj is the generalized mass corresponding to the j-th mode.

For the remaining few items, similar results can be obtained. So, per-multiplying the both sides of the formula (4-115) by gives:
(4.116)

In which, the generalized load .

This is an independent motion equation for n mutual independence single degree of freedom systems. Each equation can be solved by the vibration
theory of a single degree of freedom system. It means that the coupling equations group of (4-97) can be decoupled by the linear transformation of
equation (4-98). Divide the two sides of the formula (4-116) by Mj, resulting in:
(4.117)

Let ,

so, there is:

(4.118)

It can be noted that equation (4.97) are actually n (j = 1, .., n) independent equations. If the load P(t) is a simple harmonic load or other load that can be
interpreted as an analytical expression, then each equation can be solved by the method of the single degree of freedom system. If the load P(t) has no
analytical expression or it is discrete, then the equation can be solved through the Duhamel integral.
If the initial displacement and the initial velocity of the equation (4.97) are both 0, that is , then it can derive out ,
using the Duhamel integral, then the solution of the equation (4.118) can be obtained.
(4.119)

In which,
If the initial condition does not equal to zero, i.e., the initial velocity and the initial displacement are not zero, the vibration caused by the initial
condition should be superimposed on equation (4-119), thus:
(4.120)

For the initial condition q(0) and , it can be obtained from equation (4-103) that:
(4.121)

After obtaining the response qi(t) of each vibration coordinate, the displacement response of the multi-degree of freedom system can be obtained by
substituting it into equation (4.98), and furthermore, the other dynamic reaction and internal force can be obtained.
The above-mentioned solution method is expressed by linear superposition of the modes with the solutions of motion equation, called mode
superposition method. This method switches the problem of coupled differential equations group into a problem to solve n mutually different
differential equations, thus, the calculation is greatly simplified. It should be pointed out that often only the first few lower modes play a role for a finite
degree of freedom system under external load. Therefore, when using the superposition method to solve the dynamic response of the structure, it
usually does not need to include all the vibration mode of the system, but only to retain a number of the lowest order modes according to the accuracy
requirements of the calculation.
Example 4.10 As shown in Figure 4.14, it has been known that the two self-frequencies of the structure are ω n1 and ω n2, respectively. The first
main mode and the second main mode of the structure are {X}1 = {1,1}T and {X}2 = {1,–1}T, respectively. Solve for the displacement and bending
moment of the structure under the sudden load P1 (t).
Figure 4.14 Simple support beam structure.
Solution:
(1) Using formula (4-98) to establish a regular coordinate transformation:
(a)

(2) Seeking generalized quality

According to equation (4-95):

(3) Finding the generalized load

(4) Finding the regular coordinates

According to equation (4.103):

(5) Finding particle displacement

According to the coordinate transformation formula (a):
(6) Finding bending moment
Supposing that denotes the sum of the load and the inertia force of the particle i at any time t

Thus, the bending moment of the cross section of the particles 1 and 2 is:

4.5 Numerical Analysis of Damping System

Through the depiction of the above chapters, we can see that whether it is a single degree of freedom or a multi-degree of freedom system, when the
external load P(t) is analytical function, using the Duhamel integral method of time-domain analysis method or the Fourier transform methods of
frequency-domain analysis method, etc., the analytical expressions of the system dynamic response can generally be obtained. However, these two
methods are based on the superposition principle, which requires the structural system to be linearly elastic, and when the load does not possess
analytic expression (such as seismic load) or the analytical expression is too complex, the superposition principle will not suitable. This means that
Duhamel integral, Fourier transform and mode superposition method lose effectiveness at this time, and only numerical analysis method can be used.
At present, the most common and most effective numerical analysis method is the gradual integration method in the time domain, called the direct
integration method and the time history analysis method. This is the method of determining the dynamic response of the structure by directly solving
the vibration equation. This can further be subdivided into: segmentation method, central difference method, average constant acceleration method,
linear acceleration method, Newmark-β method, Wilson-θ method, etc. Regarding whether it is needed to simultaneously solve the coupling system, it
can be divided into two categories: implicit methods and explicit methods. In implicit methods, such as the Newmark-β method, the Wilson-θ method,
etc., an expression that gives a new value at a given step contains one or more values associated with this step. It is necessary to assume the required
test value, and then through iteration to find the initial value of the next step. Explicit methods, such as the central difference method, is defined as
follows: the calculation of new reaction value in each step depends only on the amount already obtained in the previous step. The stepwise integration
method can be used either as a single degree of freedom system or as a multi-degree of freedom system, either for linear elastic structures or for non-
linear systems or discrete external loads.
The criteria for evaluating the advantages and disadvantages of the progressive integration method mainly include three aspects: (1) convergence,
which mean that the integral step Δt → 0, whether the numerical solution is convergent to the exact solution; (2) stability, which means that the total
length of the calculation increases, whether the numerical solution is far from the exact value, (3) the accuracy of the calculation, when interrupting the
relationship between the truncation error and the time step Δt, if the error EN+1 = O(Δt N+1 ), then it can say that the method has N-order accuracy.
The following single-degree of freedom system will serve as an example to introduce several of these methods briefly.

4.5.1 Central Difference Method

At time t, the vibration equation for a single particle is
(4.122)

If the isochronous time step (Δt = Δt i = t i+1 – t i) is used, the forward differences and backward differences of the velocity at time t i are:
(4.123)

Thus, the central difference is:

(4.124)

Similarly, the center difference of the acceleration is approximately:

(4.125)

Substituting equation (4.124) and equation (4.125) into equation (4.122):

(4.126)

If it is known that the displacement ui, ui–1 of the time t i, then the displacement иi+1 of time t i+1 can be solved by the following equation:
(4.127)

And the velocity and acceleration at time t i+1 can be obtained from equations (4.124) and (4.125). The above formula is the central difference method
for the single degree of freedom system. For a multi-degree of freedom system, it just needs to change the parameters m, c, k, and P of the above
equation into a matrix or vector form.
For the single-point vibration equation with the initial condition , the specific iteration steps of the central difference method are as
follows:
1. Calculating the basic data and the initial conditions
 2. According to motions at time t and the previous time, calculating the displacement ui+1 at time t i+1 by using equation (4.127).
3. Repeating the above calculation steps to obtain the dynamic response of the system all the time.
The above formula has second order accuracy, and the stability condition is ∆t ≤ T 1 / π, and T 1 is the minimum natural period of the system. Although
the central difference method requires conditional stabilization, but because of its explicit integration and the advantages of high computational
efficiency, it is widely used in many cases.

4.5.2 Average Constant Acceleration Method

Considering a single-degree of freedom system, it is assumed that the acceleration of the particle during the time ∆t is constant, which is equal to the
average of the acceleration ü(t i) at time t i and the acceleration ü(t i + ∆t) of the time t i + ∆t = t i+1 of particle (shown as Figure 4.15(a)), thus:
(4.128)

Figure 4.15 The constant average acceleration method.

So the particle in Δt presents a linear change (shown as Figure 4.13(b)), and the velocity at time t i + Δt = t i+1 is:
(4.129)

The displacement presents parabolic change during the time Δt (shown as Figure 4.13(c)), and the acceleration at the end of the period is:
(4.130)
For the single particle motion equation in Equation (4.97), it becomes:
(4.131)

Subtract Equation (4.122) from Equation (4.131), we can obtain the vibration equation expressed in incremental form.
(4.132)

In which,
(4.133)

Rewriting velocity into incremental form according to equation (4.129) gives:

(4.134)

Similarly, the incremental form of displacement is , substituting equation (4.134) into it gives:

(4.135)

According to equation (4.132):

(4.136)

According to equation (4.133), the acceleration increment can be obtained as:

(4.137)

And substituting it back into equation (4.134) results in:

(4.138)

Substituting the equation (4.137) and equation (4.138) into equation (4.132), there is
(4.139)

In which
(4.140)

(4.141)

Equation (4.140) and equation (4.141) are called equivalent stiffness and equivalent incremental load, respectively.
When the iteration is performed with the average constant acceleration method, Δu (t i) is obtained by formula (4.139) under the condition that the
initial displacement u (t i=0) and the initial velocity are known. Then is obtained by using equations (4.137), (4.138), and (4.133). The
displacement and velocity at time t i+1 can be obtained next by using equation (4.133). In order to avoid the accumulation of error, the acceleration
should be calculated by (4.136) directly.
For a multi-degree of freedom system, just rewrite the corresponding physical quantities in the above formulas into a matrix or vector form, which will
not be described here.
Finally, it should be noted that the average constant acceleration method is one of the acceleration methods, which is stable for structures with various
natural periods and for taking various time steps.

4.5.3 Linear Acceleration Method

The linear acceleration method assumes a linear change for acceleration within Δt time (as shown in Figure 4.16(a)), i.e.:
(4.142)
Figure 4.16 Schematic diagram of linear acceleration.
Performing one integral for the above equation results in:
(4.143)

And integrating once again gives:

(4.144)

Thus, the velocity and acceleration change according respectively to the quadratic parabola (Figure 4.16 (b)) and the cubic curve (Figure 4.16 (c)).
In equation (4.143) and equation (4.144), let t = t i+Δt:
(4.145)

(4.146)

For convenient calculation, expressing the by Δu(t i) gives:

(4.147)

(4.148)

Now let t = t i, and substituting the above two equations in equation (4.132) gives:
(4.149)
After arranging the above formula, the same equation as equation (4.139) can be obtained.
(4.150)

In which:
(4.151)

(4.152)

After solving out the displacement increment Δu(t i) by equation (4.150), substituting it into the equation (4.147) and equation (4.148), the velocity
increment and the acceleration increment can be attained. Then, the dynamic response of the t i + Δt = t i+1 time of the end of the period i can be
achieved by using equation (4.133). By repeating the iterative process, it is possible to find the total vibration response over the entire time range.
It should be noted here that the linear acceleration method is a conditional stable numerical integration method, the stability condition is time step
. When the time step is too large, it may not converge.

4.5.4 Newmark-β Method

From the description of the above two sections, we can see that the basic assumption of the average constant velocity method is that the acceleration
ü(τ)(t i ≤ τ ≤ t i+1 ) of the particle takes the average value between the start and the end of the period in the period Δt = [t i, t i+1 ], while the linear
acceleration method assumed that the ü(τ) changes according to the linear law in Figure 4.17. The Newmark-β method is characterized by assuming
that the acceleration is a constant that is between ü(t i) and ü(t i+1 ), and can be expressed as:
(4.153)

Figure 4.17 Comparison of three basic assumptions for acceleration.

The coefficient γ can be understood as the weight of the initial value of the acceleration ü(t i) and the final value of the period ü(t i+1 ) for the ü(τ), so
there are 0 ≤ γ ≤ 1. At the same time, in order to improve the iterative accuracy, by using other parameters 0 ≤ β ≤ 1, the above equation will be
expressed as follows:
(4.154)

According to equation (4.153), the velocity at t i+1 time can be obtained.

(4.155)

Similarly, according to equation (4.155), the displacement at time t i+1 can be obtained.
(4.156)

The formulas of the speed and acceleration at time t i+1 can be obtained from the above two formulas.
(4.157)

Substituting the above equations into the fundamental vibration equation (4.97) with multiple degrees of freedom, and rewriting it gives:
(4.158)
In which
(4.159)

(4.160)

The iterative process of the Newmark-β method is as follows:

1. Confirm of basic parameters and the initial condition ,
2. Choose the time step Δt and the parameters β,γ, generally take γ = 0.5, 0 ≤ β ≤0.25
3. Form the effective stiffness matrix by equation (4-159), calculate the equivalent load of the t i+1 time by the equation (4-160).
4. Solve out the displacement {u(t i+1 )} at time t i+1 by equation (4-158).
5. Calculate the corresponding velocity and acceleration by equation (4-157)
By repeating steps (3) to (5), all the dynamic responses of the system can be obtained.

4.5.5 Wilson-θ Method

The Wilson-θ method is a numerical integration method based on the linear acceleration method. As illustrated in Figure 4.18, this method assumes
that the acceleration changes linearly within the time period [t, t+θΔt]. First, the linear acceleration method is used to calculate the acceleration,
velocity, and displacement of the system at t i+θΔt, and then the motion response of the system at time t i+1 =t i+Δt is obtained by interpolated method.

Figure 4.18 The basic assumptions of the acceleration method of the Wilson-θ method.
Assuming that the acceleration changes linearly in an extended period of time τ = θΔt(θ≥1) (Figure 4.18). According to the derivation method of the
linear acceleration, the incremental quasi-static equilibrium equation in the Wilson-θ method can be obtained. The equivalent stiffness and equivalent
incremental load are as follows:
(4.161)

In which,
(4.162)

(4.163)

Similarly, the formula (4.147) can be written as

(4.164)

After solving out the Δuτ(t i) from equation (4.161), it is substituted into equation (4.164) to solve out Δüτ(t i), thus, the acceleration increment of
Δuτ(t i) corresponding to the normal step is:
(4.165)

Substituting Δü (t i) into the equation (4.145) and equation (4.146) respectively, so that and Δu(t i) can be obtained. By repeating the above
calculation step, the time course curve of the structural reaction can be obtained.
When θ = 1, the Wilson-θ method degenerates to a linear acceleration method, and when θ = 1.37, the method is unconditionally stable.
Example 4.11 It is known that the mass m = 200kg, stiffness k = 7200N/m, and damping ratio of a single degree of freedom system is 0.05 and
the external load subjected to time changes is shown in Figure 4.19. Determine the dynamic response of the particle using the linear acceleration
method.
Figure 4.19 External load excitation diagram.
Solution: The natural vibration frequency is , the natural period is T = 2π / 6=1.047s, the damped coefficient is c =
2mω nζ = 2 × 200 × 6×0.05 = 120, let the time step equal to 0.1s.
Equivalent stiffness:

The equivalent incremental load is:

when the initial conditions of the system is: t = t 0 = 0,

So:

4.6 Stability and Accuracy Analysis of Stepwise Integration Method

The stability of the solution of the stepwise integration method is divided into conditional stability and unconditional stability.
If any initial conditions are given, for any selected step length ratio (T is the period corresponding to any mode of the system), when the number of
time steps tends to infinity, if the solution given by the stepwise integration method is bounded, the solution of the algorithm is said to be
unconditionally stable. If the solution given by the algorithm is bounded only when it is less than a certain value, the solution of the algorithm is said to
be conditionally stable. The following takes the Wilson-θ method as an example to illustrate the stability analysis and accuracy analysis method of the
algorithm solution. The main significance of this analysis is to provide a way to select the integral parameters.
The equation of motion for n degrees of freedom system is
(4.166)

When C = a0M + a1 K, by regular coordinate transformation , we can get

(4.167)

Where the i-th equation is

(4.168)

The calculation result of the stepwise integration of each equation represented by the equation (4.168) by the same time step is equivalent to the
calculation of the stepwise integration of the equation (4.166) by the same time step. In addition, the n equations represented by the equation (4.168)
are independent of each other and constructed identically. Therefore, one of the equations can be represented (for the sake of simplicity, the subscript i
is no longer marked) for analysis as follows:
(4.169)

The following equation is obtained by the Wilson-θ method.

(4.170)

(4.171)

At t + θΔt, equation (4.169) should be satisfied, so there is

(4.172)

Formula (4.169) and (4.170) are substituted into equation (4.171) to solve . Then use the Wilson method to solve , and , then we can
obtain
(4.173)

In the formula, the stepwise integral operator A and the load operator L are respectively
(4.174)

(4.175)

In the formula, .

Equation (4.173) is the recursive relationship between the response at time t and the response at time t + ∆t. For example, the response at time t +
2∆t is obtained from the response at time t + ∆t, and t in equation (4.173) is equal to t + ∆t, so there is

By analogy, we can get

(4.176)

Equation (4.176) is the basic relation between the stability and accuracy of the analytical algorithm solution.
Since the load varies with the problem, when analyzing the stability and accuracy of the algorithm solution, only the free vibration of any initial
condition is analyzed, that is, . At this time, the formula (4.176) becomes
(4.177)
4.6.1 Stability Analysis of Algorithm Solutions
From the above definition of unconditional stability and the formula (4.177), when n → ∞, if An is bounded, the algorithm is unconditionally stable.
Known by linear algebra, for an arbitrary matrix A, there is a full rank matrix P such that:
(4.178)

(4.179)

It can be seen from equation (4.179) that when n → ∞ is bounded by An, Jn must be bounded, that is, the absolute value of all the eigenvalues of An is
required:
(4.180)

Equation (4.180) is the criterion for the stable solution of the algorithm.

4.6.2 Accuracy Analysis of Algorithm Solutions

In order to gain an essential understanding of the accuracy of the method and the Newmark-β method, Wilson used these two algorithms to calculate
the solution of the undamped free vibration described by equation (4.169), such as:
(4.181)

Adopt two initial conditions:

1. , . At this time, , the exact solution of (4.181) is .
2. , . At this time, , the exact solution of (4.181) is .
The free vibration described by equation (4.181) is not attenuated and the period is constant. In this way, the accuracy of the algorithm solution can be
discriminated by the stepwise integration method to obtain the change of the vibration.

Exercises
4.1 A three-layer frame as shown in Figure 4.1, the mass of each layer is m1 = 2560t, m2 = 2540t, m1 = 560t, the stiffness between layers is k 1 = 5.45
× 105kN/m, k 2 = 9.00 × 105 kN/m, k 3 = 8.20 × 105 kN/m. Find the natural frequency and mode of the system and verify the orthogonality of the
main mode.
4.2 A homogeneous rigid rod as shown in Figure 4.2, the total mass of the rod is m, the two ends are supported on springs with the stiffnesses of k 1 , k 2
respectively. Assuming that the rod has only two degrees of freedom, one for the vertical degree of freedom u1 and the other for the rotation degree of
freedom u2, if all the mass of the rod can be assumed to focus on point O, establish the system of free vibration equations of motion.

4.3 As shown in Figure 4.3, the cantilever frame consists of two rods with the same bending stiffness, EI. A particle with mass of m is connected to the
frame at its free end. The displacements of the mass point generated under dynamic load are ux and uy , respectively. Find the free vibration equation
of the system.
4.4 As shown in Figure 4.4, the two known mass points are m1 = 100kg, m2 = 160kg, the bending stiffness of the two rods is EI = 5.5 × 103kN·m2 and
the mass point m1 exerts a simple vertical harmonic load, in which P = 3kN, θ = 240r/min, regardless of the damping, calculate the amplitude of each
particle by the mode decomposition method without any damping.

4.5 Please prove that it is reasonable to obtain the natural frequency ω n1 , ω n2 by the formula (4-13), namely .
4.6 Please use the energy method to calculate the fundamental frequency of the frame structure of Figure 4.1.

References
1. Clough, R.W. and Penzien, J., Dynamics of Structures, Computers and Structures, 2005.
2. Wilson, J.F., Dynamics of Offshore Structures, Second Edition, John Wiley & Sons, Inc.
3. Chopra Anil, K., Dynamics of Structures, Prentice Hall, 2000.
4. Timoshenko, S., Vibration Problems in Engineering, Van Nostrand, 1975.
5. Paz, M., Structural Dynamics-Theory and Computation, Van Nootrand Company.Inc, 1985.
6. Hibbeler, A.C., Structural Analysis, Prentice Hall, 2011.
7. Lanczos, C., The Variational Principles of Mechanics, Dover Publications Inc, 1986.
8. Worden, K. and Manson, G., Random Vibrations of a Multi-Degree-Of-Freedom Non-Linear System Using the Volterra Series. J Sound Vib., 1999.
9. Hsueh, W.J., Vibration Transmissibility of a Unidirectional Multi-Degree-Of-Freedom System with Multiple Dynamic Absorbers. J Sound Vib.,
2000.
10. De Wachter, S and Tzavalis, E, Detection of Structural Breaks in Linear Dynamic Panel Data Models. Comput. Stat. Data Anal., 2012.
11. Cluni, F., Gioffrè, M., Gusella, V., Dynamic Response of Tall Buildings to Wind Loads by Reduced Order Equivalent Shear-beam Models. J Wind
Eng. Ind. Aerodyn., 2013.
12. Zhao-Dong, Xu and Le-Wei, Ma., Dynamics of Structures, Science Press, Beijing, 2007.
Chapter 5
Distributed-Parameter System
5.1 Overview
This chapter consists of a brief introduction to the continuous systems described by partial differential equations. It begins with the plane vibration of a
beam without shear, the vibration of a string, and the axial vibration of a bar. The end of this chapter considers the bending vibration of rectangular
and circular plates and the response of rotationally periodic structures to rotating forces.
The formal mathematical procedure for considering the behavior of an infinite number of connected points is by means of differential equations in
which the position coordinates are taken as independent variables. Inasmuch as time is also an independent variable in a dynamic-response problem,
the formulation of the equations of motion in this way leads to partial differential equations. Different classes of continuous systems can be identified in
accordance with the number of independent variables required to describe the distribution of their physical properties. For example, the
wavepropagation formulas used in seismology and geophysics are derived from the equations of motion expressed for general threedimensional solids.
Similarly, in studying the dynamic behavior of thinplate or thinshell structures, special equations of motion must be derived for these twodimensional
systems. In the present discussion, however, attention will be limited to onedimensional structures, that is, beamtype and rodtype systems which may
have variable mass, damping, and stiffness properties along their elastic axes. The partial differential equations of these systems involve only two
independent variables: time and distance along the elastic axis of each component member.

5.2 Establish Differential Equations for Motion

5.2.1 Euler-Bernoulli Beam
Consider the transverse deformation of a beam vibrating in a plane as shown in Figure 5.1. The beam has a symmetric cross section of area A and
moment of inertia I; the mass density is ρ and the Young modulus is E. It is subjected to a distributed load p(x, t); the transverse displacement is v(x,
t). The free body diagram of a small slice dx of the beam is shown in Figure 5.1, with all the forces acting on the two cross sections. According to classical
notations, M and S represent respectively the bending moment and the shear force acting on the left side, while M+dM and S+dS are those acting on
the right side. The inertia load –ρAϋdx is applied to the center of mass of the small element, in the direction opposing the motion.
The vertical equilibrium reads:
(5.1)

Or
(5.2)

Figure 5.1 (a) Transverse vibration of a beam; the transverse displacement is v(x,t). (b) Free body diagram of a slice dx of the beam. (c) Euler-
Bernoulli kinematic assumption.
The rotational equilibrium (moment about the point P on the left side of Figure 5.1, positive clockwise; the rotational inertia is neglected).
(5.3)

and, if one neglects the second-order terms dS.dx and (dx)2,

(5.4)

These are the equilibrium equations. The kinematic assumptions of the Euler-Bernoulli theory of beam are that (i) the plane cross section of the beam
remains plane after deformation, and normal to the neutral axis (no shear strain); (ii) there is no stretching of the neutral axis, and (iii) the loading acts
in the plane (x, y) which is a plane of symmetry, so that the beam remains in the same plane after deformation.
The vertical displacement, common to all points in the same cross section is v(x). Since the cross section remains plane and orthogonal to the neutral
axis, the axial displacements vary linearly with the distance to the neutral axis:
(5.5)

Every fiber in the beam is assumed to be in a uniaxial state of stress and strain:
(5.6)

Taking into account that a positive bending moment induces negative stresses above the neutral axis, the bending moment reads
(5.7)

Combining Equation (5.2), (5.4) and (5.7),

(5.8)

It is of the second order in time and of the fourth order in space.

5.2.2 Beam with Axial Pressures

If the beam is subjected to an axial preload N, the axial force interacts with the lateral displacement to produce an additional term in the moment
equilibrium equation, as shown in Figure 5.2. The moment associated with the axial loading is Ndv=Nv ʹdx (N positive in traction). It follows that the
moment equilibrium equation (moment about the point P) reads
(5.9)

Figure 5.2 Free body diagram of the beam in presence of an axial preload N.
and, if one neglects the second-order terms,
(5.10)

(5.11)

Using Equation (5.11) instead of Equation (5.4) and pursuing the same development, one gets the partial differential equation governing the vibration
of a prestressed beam:
(5.12)

5.2.3 Beam Flexure with Viscous Damping

The previously established differential equation does not consider the effects of damping. We will consider two forms of damping: (1) an external
damping force (2) internal resistance opposing the strain velocity.
The first of damping:
(5.13)

The section moment expression is changed to the form

(5.14)

Making these changes we will get the differential equations of motion with damping.
(5.15)

5.2.4 Beam Axial Deformations without Damping

While this bending mechanism is the most common type of behavior encountered in the dynamic analysis of one-dimensional members, some
important cases involve only axial displacements, e.g., a pile subjected to hammer blows during the driving process. The equations of motion governing
such behavior can be derived by a procedure similar to that used in developing the equations of motion for flexure.

Figure 5.3 Bar subjected to dynamic axial deformations: (a) bar properties and coordinates; (b) forces acting on differential element.
Start vibrating after applying a time-varying load at both ends of the beam. Axial stiffness EA and mass per unit length m vary along its length. The
axial displacement of the beam is u(x,t).
(5.16)

Axial strain

(5.17)

Substitute the above formula

(5.18)

5.3 Free Vibration of a Beam

The free vibration of a (Euler Bernoulli) beam without axial load is governed by Equ. (with p = 0). If the beam is uniform,
(5.19)

A non-trivial harmonic solution v(x,t)=V(x)ejωt exists if V(x) and ω satisfy the eigenvalue problem
(5.20)

Introducing the non-dimensional quantities η = V (x)/L and ξ = x/L, using

(5.21)
we get
(5.22)

Or
(5.23)

after introducing the non-dimensional frequency

(5.24)

The characteristic equation of this ordinary differential equation is

(5.25)

and the general solution has the form:

(5.26)

where C1 , · · ·,C4 are constants which depend on the four boundary conditions, two at each end. These can be kinematic η=0 (blocked displacement),
ηʹ=0 (blocked rotation), or natural, ηʺ =0 (no bending moment) or ηʹʺ =0 (no shear force).

5.3.1 Decoupling the Boundary Conditions

Solving for the constants C1 , · · ·,C4 involves the solution of a system of four equations with four unknowns. However, the solution of the problem may
be considerably simplified if, instead of using the exponential functions of Equation (5.19), one uses linear combinations of them defined as follows
(5.27)

Note that all these functions vanish at the origin ξ = 0 except c 1 and their derivatives are related by
(5.28)

The general solution may be written alternatively

(5.29)

so that the derivatives with respect to ξ are respectively

(5.30)

Note that the relationship between the constants A,B,C,D and the value of η and its derivatives at ξ = 0 is totally decoupled
(5.31)

which allows a considerable simplification of the problem, as we can see in the following examples.

5.3.2 Simply Supported Beam

Consider the simply supported beam of Figure 5.4(a); it is pinned at both ends, so that the displacement v=0 and the bending moment M=0 (or
equivalently vʺ =0) vanish at both ends. Thus, we have η(0)=0 and ηʺ (0)=0; it follows from the foregoing equation that B=0 and D=0. Expressing the
boundary conditions at ξ=1, η(μ)=0 and ηʺ (μ)=0 one gets the following system of equations
(5.32)
Figure 5.4 Simply supported beam. (a) boundary conditions. (b) Mode shapes.
In order to have a solution, the determinant must vanish:
(5.33)

(5.34)

The solution μ=0 would lead to η=0. The non trivial solutions are μ=kπ for which s1 (µ)=s2(µ)=sinh μ, thus leading to
(5.35)

or A=–C. Substituting this (and B=D=0) in (5.22), one finds

(5.36)

The natural frequencies are obtained by substituting μ=kπ in Equation (5.17).

Finally, the natural frequencies and the mode shapes are
(5.37)

As for all distributed systems, there is an infinite set of modes; the first four mode shapes are displayed in Figure 5.4(b). The natural frequencies vary
with the square of the order k of the mode and the inverse of the square of the length L.

5.3.3 Free-Free Beam

Consider a uniform beam in free-free support; the boundary conditions are S=M=0 at both ends, ξ=0 and ξ=1. Thus, η″(0)=0 and η′″(0)=0 and,
according to Equation (4.31), D = C = 0. Expressing the boundary conditions at ξ=1, η″(1)=0 and η′″(1)=0, one gets
(5.38)

The eigenvalues cancel the determinant

(5.39)

Or
(5.40)
This equation is illustrated in Figure 5.5. The eigenvalues correspond to the crossing between the two curves. Once the value of μ is obtained, the
natural frequency may be computed from Equation (5.17).

Figure 5.5 Free-free beam. The solutions of the equation cosμ = (coshµ)–1 are the eigenvalues of the problem.
Note that the two curves are tangent to each other at the origin, leading to a double root at μ=0. Substituting in Equation (5.16), one gets ηIV=0,
meaning that
(5.41)

and, taking into account the boundary conditions η″(1)=η′″(1)=0, a=b=0, and the solution is
(5.42)

It consists of a family of two rigid body modes (there are two free parameters c and d). On the other hand, the function (coshμ)–1 decreases extremely
fast, so that the solution converges quite quickly towards its asymptotic value
(5.43)

For small values of k, a better estimate can be obtained by an asymptotic method whereby a first-order correction εk is added to the asymptotic value:
(5.44)

Introducing into the characteristic equation (5.32) and expanding,

(5.45)

Taking into account that , , and using the approximations coshεk≃l, sinhεk ≃εk and sinεk≃εk, this becomes

(5.46)

and finally, keeping only the first-order term in εk in the left-hand side,
(5.47)

The first-order correction is alternately positive and negative, and decreases rapidly with k. For a given eigenvalue μ the mode shape is obtained from
anyone of the Equation (5.31); from the first one we get:
(5.48)
leading to the mode shape
The first few modes are represented in Figure 5.6.
Example 5.1 Considering the uniform simple beam shown in Figure 5.7, its for kown boundary conditions are

Figure 5.6 Free-free beam. (a) Boundary conditions (b) Mode shapes (rigid body modes and flexible modes).
Figure 5.7 Simple beam-vibration analysis: (a) basic properties of simple beam; (b) first three vibration modes.
The second partial derivative with respect to x, the shape functions can be written as

from which one obtains (A1 +A3)=0 and (-A1 +A3)=0; hence, A1 =A3=0. Similarly, the rest shape functions can be written in the form

after setting A1 and A3 equal to zero. Adding these two equations, after cancelling out a2, gives

Thus A4 =0 since sinhaL≠0. Only A2 remains a nonzero constant, therefore

Excluding the trivial solution A2=0, boundary condition η(L)=0 can be satisfied only when

5.4 Orthogonality Relationships

We now establish the orthogonality relationships between the mode shapes. Consider two modes with distinct frequencies; mode i satisfies
(5.49)

Multiplying by vj and integrating over the length of the beam, one gets
(5.50)
We then integrate by parts twice, in order to distribute evenly the spatial derivatives over vi and vj:
(5.51)

The two expressions in brackets vanish, because of the boundary conditions [at each end of the beam, either EIv″= 0 or v′=0, and similarly (EIv″)′=0
or v=0]; substituting into the previous equation, one gets
(5.52)

Similarly, permuting i and j, one gets

(5.53)

and, subtracting these two equations,

(5.54)

Thus, if ω i≠ω j, one must have

(5.55)

This is the first orthogonality relationship: the mode shapes corresponding to distinct frequencies are orthogonal with respect to the mass distribution
(ρA is the mass per unit length). Returning to the previous equation, it follows also that
(5.56)

The mode shapes are also orthogonal with respect to the stiffness distribution (EI is the bending stiffness).
(5.57)

(5.58)

where δij is the Kronecker delta (δij=1 if i=j, δ ij=0 if i≠j), μi is the modal mass (also called generalized mass) of mode i. From the foregoing equations,
(5.59)

5.5 Modal Decomposition

We now return to the general equation (5.8) and perform a modal decomposition according to
(5.60)

where vi(x) are the mode shapes and the modal amplitudes zi(t) form an infinite set of modal coordinates. Substituting Equation (5.60) into Equation
(5.8), one finds
(5.61)

Upon multiplying by vk and integrating over the beam length,

(5.62)

After integrating by parts the first term, as in the previous section, it is rewritten
(5.63)

and, upon using the orthogonality conditions (5.46) and (5.47), one finds a set of decoupled equations:
(5.64)

This equation states that every mode behaves like a single-degree-of-freedom oscillator of mass μk and frequency ω k; the generalized force is once
again the work of the external distributed force p on the mode v k. A point force F(t) applied at x0 is represented by p(x, t)=F(t)δ(x – x0) where δ(x) is
the Dirac function such that
(5.65)

so that the work is F(t).vk(x0). The modal damping can be added and the displacement at a point x1 along the beam is
(5.66)

Using (5.55) and the Frequency Response Function (FRF) between a point force excitation at x0 and the displacement at x1 reads
(5.67)

with Dk (w) being the dynamic amplification factor of mode k, given by as,
(5.68)

References
1. Clough, R.W. and Penzien, J., Dynamics of Structures, McGraw-Hill, 1995.
2. Clough, R.W. and Penzien, J., Dynamics of structures, in: Dynamics of Structures, 2005.
3. Chopra Anil, K., Dynamics of Structures, Prentice Hall, 2000.
4. Wilson, J.F., Dynamics of Offshore Structures, 1984.
5. Forment, D. and Welaratna, S., Structural dynamics modification—An extension to modal analysis. Sae Paper, 1980.
6. Zhao-Dong, X. and Le-Wei, M., Dynamics of Structures, Science Press, Beijing, 2007.
Chapter 6
Stochastic Structural Vibrations
6.1 Overview
Analysis methods of structural response to seismic excitation can be divided into two categories: deterministic analysis and stochastic analysis.
Deterministic analysis refers to the analysis of structural response when subjected to deterministic seismic excitation, which means the process of
movement can be described as a deterministic function of time t. For example, the equation of motion of an undamped free vibration involving a linear
SDOF system can be expressed as:
(6.1)

In which represents the inherent frequency of vibration of the system. The solution of Eqs. (6.1) is:
(6.2)

In which u(0) and represents the initial displacement and initial velocity of the system respectively. Clearly, the displacement, u(t) of SDOF at any
time t can be determined by Eq. (6.2). The velocity and acceleration with time t, and ü(t), also can be determined. Deterministic analysis, having a
history of over 200 years, is the content of traditional vibration theory, in which many problems have been studied systematically and thoroughly; its
research results have been widely used in engineering practices.
Stochastic analysis is to consider the vibration problem from another angle and developed on the basis of the stochastic process as a branch of
structural dynamics. It discussed the stochastic structural response to stochastic excitation, of which the statistical properties are known. That is to
say, the research objects are not one certain wave but the statistical properties of the set of waves. For example, wind pressure and earthquakes that
act somewhere in the structure, movement of some location and undulation of the sea are all unpredictable. Researchers cannot get the same
observation results even when the observations proceed in the same time period and observation conditions. That is to say, the results of every
observation can be expressed by a function of time t, which can be uniquely determined, while the results of independent repeated observations are
varied. Although the results seem disorderly, the set of them shows statistical regularity. So stochastic vibration can be described not by deterministic
function, but by statistical properties. The deterministic relation between excitation and response of the system is researched in the deterministic
vibration problem, whereas the statistical probability relation is studied through stochastic vibration problems.
The statistical characteristics of the stochastic process include probability density function of vibration, average amplitude, variance, self-related
function, power spectral density, maximum of given probability, etc. The character of stochastic excitation, stochastic structural response and relations
between them can be described by these statistical characteristics and their relationships. Mass, stiffness, strength, and damping, as structural
dynamic characteristics, are also discrete random values essentially. But given that the randomness of these values is not as much as that of external
excitation, these are usually considered as deterministic values to simplify the analysis.
Stochasticity exists widely in the natural world. When the degree of randomness is low, the discrete degree of the results of experiments is not high.
Consequently, problems can be solved on the basis of the average value of the results and the fluctuations around the average value are ignored. From
this point of view, deterministic vibration is a special case of stochastic vibration. When the degree of randomness is high or relatively high
requirements are set toward the description of the vibration characteristics of the system or the mechanism of system work is not clear in the vibration
issue. It is not appropriate to use deterministic vibration theory and we should adopt random vibration theory for analysis and processing.
The theory of stochastic vibration is widely used in mathematics, mechanics, engineering, and many other fields. The fundamental theory of structural
stochastic vibration should consist of three parts, including the mathematical model of structural stochastic excitation, analysis of the response of
stochastic vibration of linear or nonlinear system, and analysis of structural damage and reliability. Normal models of structural stochastic excitation—
stationary and non-stationary model, include pulsed wind speed and wind pressure model, ocean wave and wave force model and earthquake ground
motion model. In theory, among excitation, system, and response, knowing two of them can solve the rest. Vibration theory studies how to identify the
response when excitation and characteristics of the system are known. If there is randomness in excitation and the system, there is also randomness in
response of the system. That is to say the vibration of the system is stochastic. Stochastic vibrations can be divided into different categories according
to the characteristics of the system and stochastic process. Considering stationary or non-stationary excitation and linear or nonlinear system,
stochastic vibrations can be divided into stationary stochastic vibration of the linear system, non-stationary stochastic vibration of the linear system,
stationary stochastic vibration of the nonlinear system, and non-stationary stochastic vibration of the nonlinear system, as is shown in Figure 6.1.

Figure 6.1 Classification of stochastic vibration problems.

The linear system is an approximate model. Lots of structural systems can be processed into the linear system. However, when better accuracy is
required or problems are nonlinear essentially, we have to use nonlinear models in simulation. Nonlinearity of the system brings enormous difficulties
to analysize the stochastic vibration. Lots of efficient methods in analysis of linear systems do not work in analysis of nonlinear systems because the
superposition principle is not compatible with nonlinear systems. Besides, the response of a nonlinear system to Gauss stochastic excitation is not a
Gauss process any longer. As a result, only studying the secondary moment of response cannot help to acquire all the statistical characteristics though
it works in a linear system. It’s necessary to study higher-order moment of response or seek probability density function of response directly. Since the
1960s, higher emphasis has been laid on the study of nonlinear stochastic vibration and several effective methods have been developed, such as FPK
equation method, moments method, stochastic average method, statistical average method, and perturbation method, etc. For nonlinear systems and
excitation systems, the only method to locate the accurate solution is the Markov diffusion process method. This method seeks conditional probability
density function of response of nonlinear systems to excitation of Gauss white noise, which can be concluded to solve corresponding FPK equation in
mathematics. However, accurate solutions can only be found in some special first-order nonlinear systems. To locate an accurate solution of stationary
probability density function, quite strict constraints must be imposed on constitutive relations of the system and characteristic of excitation, which is
hardly realistic in nature. So, it’s urgent to develop various approximate methods, which are more practical to solve the stochastic vibration of
nonlinear systems.
Methods to solve non-stationary stochastic vibration of a linear system can be divided into time domain method, frequency domain method, or a
combination of these two methods. The time domain method is firstly used by Caughey et al. to solve the response problem of an invariable system
under sudden stationary stochastic excitation. In fact, it’s also an effective method to solve the response problem of an invariable system under non-
stationary stochastic excitation. Frequency domain method can be divided into two categories: double frequency spectrum method and evolved
frequency spectrum method. The double frequency spectrum method is proposed earlier but solving double frequency spectrum needs lots of raw data
of non-stationary process, which will cost too much. The double frequency spectrum method does not have distinct physical meaning itself.
Consequently, it has not been actually used. Evolved frequency spectrum method is applicable to problems of evolved stochastic process, in which
statistical characteristics are influenced by “slow-change” time and “stationary process” modulated by frequency function. Though the evolved
frequency spectrum method is practical, it’s much more complicated than the frequency domain method of stationary response. Furthermore, some
methods combining these two methods have been proposed and are still under study.

6.2 Stochastic Process

6.2.1 Concept of Stochastic Process
Stochastic process extends from the concept of stochastic variable. There are lots of examples of stochastic process in nature. For example, longtime
observation of one point on the wing and recording the results automatically are necessary when the vibration of the wing of a plane in flight is studied.
As a result, a deterministic function of acceleration-time, ü1 (t) can be acquired. This deterministic function of acceleration-time cannot be known
without experiments. However, the deterministic function ü2(t) acquired from another experiment conducted independently on the same condition will
be different from ü1 (t). In fact, because of the randomness of structural vibration, the same function of acceleration-time can never be acquired from
two experiments although the conditions remain unchanged. These functions are integrated as a family of functions, as is shown in Figure 6.2.
Consequently, the observation of structural vibration can be considered as a stochastic experiment, which is conducted during a period of time and
leads to a result of a function of time t.

Figure 6.2 Sample function of stochastic process.

Stochastic experiments can be described by the sample space made up of all the possible experiment results. Likewise, the vibration of the wing of a
plane in flight can be described by a family of functions of acceleration-time acquired from longtime observation. Now introduce the concept of
stochastic process using this example.
Provided that E is a stochastic experiment and S={e} is its sample space, for every e ∈ S, a corresponding function of time t can always be determined
according to a certain rule.
(6.3)

(T is the variation range of t). So, a family of functions of time t can be acquired for all the e ∈ S. The family of functions of time t is called stochastic
process and every function of this family is called a sample function of this stochastic process.
So, the vibration of the wing of a plane in flight is a stochastic process and every function of acceleration-time acquired from observation is a sample
function of this stochastic process.
Stochastic process can be characterized by the typical sample function X(e,t) of the family. For a determined ei ∈ S, or for a determined experiment
result, X(ei,t) is a determined sample function, which can be regarded as a physical implementation of stochastic process. Since stochastic process is a
collection of sample functions, we can also consider it from another angle. For a arbitrary time t i ∈ T, X(e,t i) is a stochastic variable defined on S, shown
in Figure 6.2. In engineering, X(e,t i) is sometimes called as a stochastic process at the time of t = t i. According to this layer of meaning, the stochastic
process can also be defined in another form: If X(e,t i) are stochastic variables for every certain t i ∈ T, it is called a stochastic process. In other words,
the stochastic process is a family of stochastic variables that depend on time.
The two definitions of the stochastic process are the same essentially, but different in description. The second description method is preferred in the
theoretical analysis while the first description method in the actual measurement. So, these two methods are complementary in both theoretical and
practical fields. For the sake of simplicity, e in equation (6.3) is omitted. The stochastic process is indicated by the notation X(t), and its sample function
is indicated by x 1 (t).

6.2.2 Probability Description of Stochastic Processes

The state of the stochastic process at any time is a stochastic variable. So, the probability description method of stochastic variables can be used to
describe the statistical characteristics of the stochastic process.
Provided X(t) is a stochastic process, for each certain t i ∈ T, X(t 1 ) is a stochastic variable, of which the distribution function is generally related to t 1 ,
denoted by:
(6.4)

This is called the one-dimension distribution function of stochastic processes. Provided there is a binary function f1 (x 1 ,t 1 ) establishing the equation
below:
(6.5)

The function f1 (x 1 ,t 1 ) is called one-dimension probability density of the stochastic process X(t).
In order to describe the connection between a stochastic process at any two moments t 1 and t 2, the distribution function of a two-dimension stochastic
variable is introduced, which generally depends on t 1 and t 2, denoted by:
(6.6)

This is called the two-dimension distribution function of the stochastic process X(t). Provided there is a function f2(x1 ,x2;t 1,t 2), the equation below can
be established:
(6.7)

Which is called a two-dimensional probability density of a stochastic process.

In general, when the time takes any n values t 1,t 2, …,t n, the distribution function of the n-dimension stochastic variable is denoted by:
(6.8)

Called the n-dimension distribution function of the stochastic process X(t). Provided there is a function fn(x1 ,x2, … xn,;t 1,t 2, …,t n), the equation below is
established
(6.9)

Which is called a n-dimension probability density of a stochastic process.

The n-dimension distribution function (or probability density) can describe approximately the statistical characteristics of the stochastic process.
Obviously, the larger n is, the more prefect characteristics of the stochastic process described by n-dimension distribution function. In general, the
family of distribution function {F 1 ,F 2,···} or the probability density {f1 ,f2,···} completely determines the overall statistical characteristics of the
stochastic process.
In the engineering vibration problem there are more than one stochastic process to be studied. So, it is necessary to consider statistical information of
two or more stochastic processes. Provided there are two stochastic processes X(t) and Y(t), and any two groups of real numbers, the distribution
function of the (n + m)-dimension stochastic variables (X(t 1 ),X(t 2),···,X(t n)), (Y(s1 ),Y(s2),···,Y(sm)) is denoted by:

Called the (n + m)-dimension joint distribution function of the stochastic process X(t) and Y(t). The corresponding (n + m)-dimension joint probability
density is denoted by:

If for any of the integers, n and m, and the arrays t 1 ,t 2,···,t n and s1 ,s2,···,sm, the joint distribution function satisfies the equation below
(6.10)

the stochastic processes X(t) and Y(t) are independent of each other.

6.2.3 The Numerical Characteristics of Stochastic Processes

The probability characteristics of stochastic processes can be described by the numerical characteristics. Although they cannot describe the stochastic
processes as well as the probability distribution, they can center on the main statistical properties of the process. In seismic engineering, only a few
problems can be addressed from the perspective of analysis and obtain the probability distribution function. In most practical problems, only solutions
of numerical characteristics can be obtained. Thus, in the practical problem, the numerical characteristics are widely used to describe the probability
characteristics of a stochastic process. The numerical characteristics of the stochastic process can be divided into the numerical characteristics of the
time domain and the frequency domain. Some basic numerical characteristics of the stochastic process are discussed below.

1. Time-Domain Numerical Characteristics

(1) Mean (mathematical expectation) and variance (or standard deviation) Provided X(t) is a stochastic process, when t 1 is fixed, X(t 1 ) is a stochastic
variable and its mean or mathematical expectation is generally related to t 1 , denoted by:
(6.11)
In the formula above, f1 (x 1 ,t 1 ) is the one-dimension probability density of X(t); μX(t) is the mean of X(t).
If the second-order origin moment of stochastic variable X(t) is denoted by ψ2x(t), namely
(6.12)

This is called the mean square value of stochastic process X(t). If second-order center moment of stochastic variable X(t) is denoted by σ2x(t), namely:
(6.13)

This is called the variance of stochastic process X(t). The square root of the variance, σX(t), is called the standard deviation of the stochastic process
X(t). The mean function μX(t) represents the center of the stochastic process at each moment, and the variance (standard deviation) describes the
degree of deviation of the stochastic process from the mean μX(t) at time.
(2) Correlation function
Mean and variance are important numerical characteristics that characterize the statistical properties of stochastic processes at various isolated
moments. In order to characterize the connection between at two different moments of stochastic processes, it is necessary to introduce new a
numerical characteristic named two-dimension probability density.
Provided X(t 1 ) and X(t 2) are the status of the stochastic process X(t) at any two moments t 1 and t 2, f2(x1 ,x2;t 1 ,t 2) is the corresponding two-dimension
probability density. The second-order origin mixed momentis called the auto-correlation function of the stochastic process X(t), which is denoted by:
(6.14)

In short it is called the correlation function.

Similarly, the second-order central mixing moment is called autocovariance function of the stochastic process X(t), which is denoted by:
(6.15)

This is called the covariance function, in short.

The relationships between the numerical characteristics of the stochastic process X(t) are shown below.
(6.16)

(6.17)

(6.18)

Among the above numerical characteristics, the most important ones are, variance, and autocorrelation function. From a theoretical point of view, the
study of mean, variance, and autocorrelation function alone cannot replace the study of the whole stochastic process. However, because they do
characterize the main statistical features of the stochastic process and are far more observable than the family of finite-dimension distribution function
and practical in calculation, these three numerical characteristics often can play an important role in solving practical problems.
The second-order origin mixing moments determined by the twodimensional joint probability density of X(t) and Y(t), called the cross-correlation
function of stochastic processes X(t) and Y(t):
(6.19)

The reciprocal variance function of two stochastic processes is defined as:

(6.20)

If for any t 1 and t 2, the reciprocal variance function of these two stochastic processes X(t) and Y(t), CXY(t 1 ,t 2) = 0, these two stochastic processes are
said to be irrelevant. At this time:
(6.21)

It can be said that if the two stochastic processes are independent of each other, then they must not be relevant; on the contrary, if the two stochastic
processes are irrelevant, they may not be independent of each other. If a stochastic process is composed of two stochastic processes, namely
Z(t)=X(t)+Y(t), equations below hold according to the definition above:
(6.22)

(6.23)

If these stochastic processes do not correlate, then the autocorrelation function is simply equal to the sum of their respective autocorrelation functions.
(6.24)

Specifically, provided t 1 = t 2 = t, equations below hold.

(6.25)

(3) Correlation function of derivative process

Sometimes dealing with derivative problems is necessary in analysis of engineering vibration, such as solving the speed by the displacement, solving
the acceleration by the speed and so on. In the stochastic vibration analysis, the problem is expressed as a derivative of the stationary stochastic
process. This part is related to the knowledge of the mean square calculus of the stochastic process, to which a comprehensive introduction won’t be
made here. The need for in-depth understanding of this content can be addressed by referring to the monographs of stochastic process theory.
Supposed that X(t) is a stationary process and is its derivative. Considering that the mean of the stationary process X(t) is a constant, the mean of
is:
(6.26)

The expectation and differential operation are both linear operators, so the order of operations can be exchanged.
The derivative of the autocorrelation function of the stationary process X(t) is:
(6.27)

Exchange the order of the derivative operation and the operation of expectation, then:
(6.28)

So:
(6.29)

Similarly,
(6.30)

That is, the second derivative of the autocorrelation function of the stationary process X(t) is equal to the correlation function of X(t) and . Then,
(6.32)

Thus, another expression of the second derivative of the autocorrelation function is obtained, that is, it is equal to the autocorrelation function of the
negative derivative process. In the same way, the higher order derivative of the stationary process is also easy to obtain.
Note that RX(τ) is an even function, which is symmetrical about the vertical axis and obtains the maximum value at the origin, so the derivative value
at the origin is zero:
(6.33)

Using equation (6.29) can obtain:

(6.34)

Which means that the stationary process X(t) is orthogonal to its derivative process .

(4) The properties of the correlation function.

The correlation function has the following properties:
1) The autocorrelation function of the deterministic function is equal to zero.
2) The deterministic function factor has the characteristic of “accompanying traversal” after the operation by autocorrelation function operator.
That is, if Y(t) = φ(t)X(t), in which φ(t) is a deterministic function,
(6.35)

3) The autocorrelation function of the stochastic process is equal to the sum of the cross-correlation functions of the components.
4) Suppose stochastic processes, X(t) and Y(t), are independent of each other. If Z(t) = X(t) + Y(t), then
(6.36)

If Z(t) = X(t)Y(t), then

(6.37)

5) The correlation function of the differential of the stochastic process X(t) can be obtained by solving the mixed partial derivative of the
autocorrelation function of X(t).
(6.38)

In the formula, X(n)(t) represents the n-order derivative of the process X(t). For a stationary stochastic process, the formula for solving the auto-
correlation function derivative can be obtained from the above equation.
(6.39)

In the formula, represents the n-order derivative of the process RX(τ).

6) If , then:

(6.40)

It should be noted that the operation rules of the cross-correlation function can refer to the operation rules of the autocorrelation function. The above
rules are applicable except for the property 4 in the calculation of the autocovariance.

2. Frequency-Domain Numerical Characteristics

(1) The Power Spectral Density of the Stationary Process
The correlation function describes the characteristics of the stochastic process in the time-domain. However, in order to fully grasp the characteristics
of the structural vibration process, it is necessary to study the frequency structure of the stochastic process. In other words, the study consists of how
many frequency components are included in the process and the magnitude or energy of each frequency component. This problem is accomplished by
Fourier transformations, a powerful mathematical tool.
Construct a truncation function of the stationary process X(t), –∞ < t < + ∞
(6.41)

The Fourier transformation of this function is as follows:

(6.42)

Use Parseval equation,

(6.43)

Divide both ends of the equation by 2T,

(6.44)

Obviously, the integrals in equations (6.43) and (6.44) are random. The limit of the mean of the left end of equation (6.44);
(6.45)

is defined as the average probability of a stationary process.

Exchange the order of integral and mean operation in formula (6.45), then:
(6.46)

The average probability of the stationary process is equal to the mean square value of the process.
Using equation (6.44) and equation (6.46), it becomes:
(6.47)

The integrand in the equation (6.47) is called the power spectrum density of the stationary process X (t), denoted as S X(ω),
(6.48)

Using the notation S X(ω), equation (6.47) can be abbreviated as:

(6.49)

When the mean of the stationary process is zero, the equation (6.49) becomes:
(6.50)

The power spectral density S X(ω) is also commonly referred to as the self-power spectral density, which is the most important numerical
characteristic to describe the statistical law from the frequency point of view. The physical meaning indicates the distribution of the mean power of X(t)
over frequency.
(2) Power Spectral Density of Non-Stationary Processes
For a general non-stationary process, the power spectral density function is defined by the Fourier transformation of the correlation function. For
example, for a general stochastic process X(t), its self-power spectral density function is defined as:
(6.51)

(6.52)
(3) Properties of the Power Spectral Density
The power spectral density has the following important properties.
1) Non-negative. It can be seen from equation (6.48) that S X(ω) is a non-negative function of frequency ω.
2) S X(ω) is a real even function. In fact, in equation (6.48), the component
(6.53)

is a real even function of ω, so the limit of its mean is also a real even function.
3) The power spectral density S X(ω) is the Fourier transformation of the correlation function RX(τ),
(6.54)

The correlation function RX(τ) is the inverse of the Fourier transformation of the power spectral density S X(ω)
(6.55)

They are collectively referred to as the Wiener-Khintchine formula. Owing to space reasons, this formula is not derived but other literatures can be
referred to.
The positive and negative values of the spectral density defined above are defined as “bilateral spectral density”. In general, the negative frequency is
not well explained, and in order to facilitate the actual measurement, the even-based function is used to convert the spectral density in the negative
frequency range to the positive frequency range, thus defining the “unilateral spectral density”:
(6.56)

The relationship between single-sided spectral density and bilateral spectral density is:
(6.57)

To show the relationship between unilateral spectral density and bilateral spectral density more clearly, the pattern is shown in Figure 6.3.

Figure 6.3 Unilateral and bilateral spectral density of stationary process.

(4) The Spectral Density of the Derivative Process
Suppose X(t) is a stationary process with a mean of zero, and its derivative process and both exist. According to the Wiener-Khintchine
formula:
(6.58)

(6.59)

Substituting equation (6.58) into equation (6.59) gives:

(6.60)

(6.61)

Comparing equation (6.58), (6.59), (6.60), and equation (6.61) yields the relation:
(6.62)
(6.63)

Similarly, comparing with equation (6.49), we can get the mean square value of the derivative process:
(6.64)

(6.65)

(5) Cross-Spectral Density and Its Properties

Suppose X(t) and Y(t) are both stationary stochastic processes.
(6.66)

Equation (6.66) is the cross-spectral density of the two stationary stochastic processes, X(t) and Y(t). X(t) and Y(t) are determined by equation (6.42).
For a number of stochastic processes Xi(t), (i = 1,2) (such as stationary processes), the cross-spectral density of any two processes can be obtained
using the method given above, in the form of a matrix,
(6.67)

Using the Fourier transformation, the corresponding cross correlation function matrix is as follows.
(6.68)

The diagonal elements in the upper matrix are the self-spectral density and the autocorrelation function.
From the equation (6.66), it can be seen that the cross-spectral density is no longer a positive, real, even function, but has the following properties:
1) . In other words, S XY(ω) and S YX(ω) are conjugate functions.
2) The requirement for that the cross-correlation function is absolutely integral.
(6.69)

(6.70)

3) The real part of S XY(ω) and S YX(ω) is an even function of ω, the imaginary part is an odd function of ω.
4) There is an inequality between the cross-spectral density and the spectral density.
(6.71)

5) Suppose the stochastic process Z(t) = X(t) + Y(t), its self-power spectral density is as follows.
(6.72)

In particular, when X(t) and Y(t) are independent of each other,

(6.73)

In the above equation, S (n)(ω) is the n-order derivative of S X(ω).

6) Coherence Function
The Coherence function (also called cohesive function) is useful in seismic engineering, for example, in the consideration of multi-point input, damage
identification, noise effects, and multi-dimensional ground motion effects. The coherence function is defined as
(6.74)

The physical meaning of the coherence function is: If X(t) and Y(t) are the input and output of a linear system respectively, and Z(t) are the effects of
noise or other input upon the output Y(t), the following equation can be obtained:
(6.75)

According to the above formula, when ρ2XY(ω) = 1, the noise spectrum is zero at this frequency ω, indicating that the output Y(t) is completely from
the input X(t), that is, the input and output are completely related. Otherwise, when ρ2XY(ω) = 0, S z(ω) = S Y(ω), indicating that at this frequency ω,
the noise spectrum output S Y(ω) is a complete noise spectrum. That is, the output is completely irrelevant to the input.

6.2.4 Stationary Stochastic Process

The stationary stochastic process is the most important type of stochastic process in practical applications. Its characteristic is that the statistical
properties of the process do not change with time. For any real number ε, The n-dimensional distribution function of the stationary process satisfies the
equation below.
(6.76)

If the probability density exists, the above stationary condition is equivalent to:
(6.77)

Suppose X(t) is a stationary process. Equation (6.77) is applied to its one-dimensional probability density, and ε = –t 1 . Then:
(6.78)

It can be seen that the one-dimensional probability density of the stationary process does not depend on time and it is denoted by f1 (x1 ). So, the mean
of X(t) should be constant, denoted by μX.
(6.79)

In the two-dimensional probability density of the stationary process, if ε = –t 1 ,

(6.80)

which shows that the two-dimensional probability density, which is denoted by f2(x1,x2;τ), only depends on the time interval τ = t 2 – t 1 , and is
irrelevant with the individual values of time t 1 and t 2. Thus, the autocorrelation function of X(t) is a function of a univariate τ,
(6.81)

The covariance function is:

(6.82)

In particular, if τ = 0, then:
(6.83)

It can be seen that the mean of the stationary process is constant and the autocorrelation function is a function of the univariate τ. It is very difficult,
and even impossible, to determine whether it is a stationary process according to equation (6.77). So, the stationary process is only considered in the
range of relevant theory in engineering. This stationary process is called a wide stationary process or a generalized stationary process, and it satisfies
that E[X(t)] is a constant and E[X2(t)] < +∞, E[X(t)X(t + τ)] = RX(τ). The stationary process defined by Eq. (6.77) is called a strictly stationary process
or a narrow stationary process.
In general, the spectral density of the stationary process is distributed over the entire frequency domain. But in engineering, we are more concerned
about which frequency the spectral components of high strength in the actual signal concentrate on. The frequency range of the main components of
the signal spectrum is often expressed in terms of bandwidth, and the spectral components out of the bandwidth are less intense and can be ignored in
practical applications. In this way, according to the “narrow” and “wide” bandwidth, the stationary process can be divided into narrowband stationary
process and broadband stationary process. The power spectral density of the narrowband process has a peak characteristic, mainly distributed in the
narrow frequency range near the center of gravity frequency ω 1 . The power spectral density of the broadband process is distributed over a wide
frequency range. The fluctuation of pressure on the surface of the aircraft caused by the turbulence and jet noise of the boundary layer, atmospheric
turbulence, and seismic ground acceleration all have the characteristics of broadband process. The single-sided spectral density and sample functions of
typical narrowband processes and broadband processes are shown in Figures 6.4 and 6.5, respectively.

Figure 6.4 Spectral density and sample function of narrowband process.

Figure 6.5 Spectral density and sample function of broadband process.
As one of the most important particular cases of the broadband process, white noise process (white noise) is widely used. When the spectral density
S X(ω) = S 0 of the stationary process X(t) satisfies S X(ω) = S0 or the unilateral spectral density satisfies G X (ω) = 2 S 0 (S 0 is a constant), the
stationary process X(t) is called stationary white noise. The spectral density of the white noise is evenly distributed over the entire frequency range
from –∞ to ∞, that is, the harmonic component of all frequencies is equally contained, which is very similar to the white light. The reason for the term
“noise” is that this process in the acoustic simulation is a hissing “Cha Cha” noise.
The process of dividing the stationary process into narrowband processes and broadband processes is of great practical significance in engineering.
Because the energy of the narrowband process is mainly concentrated in a very limited frequency range. So, if the signal from the excitation source
that causes the structural vibration is a narrowband process, it is possible to avoid a violent vibration response by making the natural frequency of the
structure far away from this range of frequency in structural design. The energy of the broadband process is distributed over a wide frequency range,
which means the type of structure it affects much more than the narrowband process does.

6.2.5 Several Important Stochastic Processes

1. Gaussian Stochastic Process
If the status of a stochastic process X(t) at any time t = t i is a Gaussian stochastic variable, or X(t) is a sequence of stochastic variable of any n states,
the joint probability density can be expressed as:
(6.84)

The stochastic process is called the Gaussian process, in which x1 ,x2,···xn is the stochastic variable value of X(t) collected at time t = t 1 ,t 2,···t n, and μX
(t 1 ),μX(t 2),···μX(t n) is the corresponding mean. K is the covariance matrix.
(6.85)

|K|jk is the minor after putting off the j-th row and k-th column and the diagonal elements is the variance of the process at all times.
The Gaussian process has many characteristics, which bring great convenience to stochastic vibration research. First, many natural phenomena can be
approximated by the Gaussian process, such as some excitation sources (seismic processes, waves, atmospheric turbulence, etc.) of random vibrations.
Secondly, the linear transformation of the Gaussian process is still a Gaussian process. Consequently, for linear systems of constant parameters when
the input is a Gaussian process, the output must also be a Gaussian process. Moreover, as long as the first moment and the second moment of the
Gaussian process is known, we can calculateits probability density function. So, when the input and output of a system are both Gaussian processes,
their statistical properties are fully determined by corresponding first and second moment.

2. Poisson Stochastic Process

To begin with, let us examine a sample function space. From a moment, the number of times when the positive value N reaches the given number is
related to the time t. It is a stochastic process, denoted by N(t). N(t) only takes positive values and is a countable process. This countable process exists
widely in engineering practice, such as the number of cars arriving at the toll of the expressway, the number of times a certain intensity of the waves
hit the coastal facilities, the number of calls in the daily call during a certain time interval, etc. They are all countable processes.
Suppose that N(t) is a specific countable process. P(n,t) is the probability of a stochastic variable N(t) = n at time t and P(n1 ,t 1 ;n2,t 2) is the joint
probability of N(t 1 ) = n1 at t = t 1 and N(t 2) = n2 at t = t 2. The countable process N(t) is called the Poisson process if the following conditions are met.
1) Independent Increment Assumption: If the number of stochastic arrivals during [0,t 1 ] is N(t 1 ) and the number of stochastic arrivals during
[0,t 2] is N(t 2), the number of stochastic arrivals during [t 2 – t 1 ] is [N(t 2) – N(t 1 )]. Similarly, the number of stochastic arrivals during [t 3 – t 2] is
[N(t 3) – N(t 2)]. If the events [N(t 3) – N(t 2)] and [N(t 2) – N(t 1 )] are independent of each other, that is to say, the number of stochastic arrivals at
a certain time in the future is independent of the number of stochastic arrivals at a certain time in the past, then N(t) satisfies the independent
increment assumption.
2) Suppose λ(t) dt is the probability of one arrival in a minute interval [t,t + dt] (a deterministic function), and the probability of two or more
arrivals in the minute interval [t,t + dt] is negligible. That is, the number of stochastic arrivals is 1 or 0, so the average of the number of stochastic
arrival is λ(t) dt.
3) The probability P(n,t) has the following form
(6.86)

This distribution is called a generalized Poisson distribution.

It is worth noting that the Poisson stochastic process is widely used in the simulation of seismic wave and structural reliability analysis.

3. Markov process
In the stochastic vibration, the fatigue analysis of vibration system, stochastic parameter system response analysis will use the Markov process. The
characteristic of this process is that the state of the process at the current moment t = t i is related only to the state at the moment t i–1 before t i, but
not to the previous states before t i–1 , which is referred to as no aftereffect. In the case of a condition of X(t i) = xi(i = 1,2,···,n – 1), the conditional
distribution of the state of the process X(t n) is only relevant to the state X(t n – 1 ) = xn–1 , and is not related to the state X(t n–2,) = xn–2, … X(t 1 ) = x1 ,
namely
(6.87)

Then the stochastic process X(t) is Markov process.

6.2.6 Stochastic Model of Seismic Ground Motion

As described in Chapter 1, seismic ground motion is a stochastic excitation, and its properties are mainly composed of amplitude characteristics,
spectral characteristics, and holding time characteristics. A large number of actual ground motion records show that the characteristics of the ground
motions which have occurred are different, and the future possible ground motion characteristics are unpredictable. This stochastic process of ground
motion generally establishes its stochastic process model based on the statistical analysis of the actual ground motion records that have occurred. This
section will introduce several stochastic models of ground motion in structural stochastic vibration analysis.
1. Stationary White Noise Model
The stationary white noise model was the earliest stochastic process model to simulate seismic ground motion, first proposed by Housner, an American
scholar, in 1947. This model assumes that the power spectral density function of the seismic ground acceleration has a constant value in the
frequency range of (–∞ < ω < + ∞).
(6.88)

Using the Wiener-Khintchine formula, the autocorrelation function of is

(6.89)

In which δ(τ) is Dirac δ function.

Stationary white noise is an extreme case of the broadband process, which equally contains the harmonic components of all frequencies. From equation
(6.50), the mean variance of the stationary white noise is:
(6.90)

Therefore, stationary white noise cannot be realized physically. However, due to the mathematic simplicity of white noise, it is often used as an
approximate or idealistic model of some physical phenomena. For the system of linear elastic structure, no large response is produced by the high
frequency excitation signal. Therefore, the broadband stochastic load acting on the structure is abstracted into a white noise process, which causes no
big errors but makes the analysis of structural stochastic vibration greatly simplified.

2. Filtered White Noise Model

Due to the characteristics of the source, the propagation path and site conditions and many other factors, site ground vibration is extremely complex.
Based on the analysis of the data of strong ground motion, it is found that the power spectral density function of the actual ground motion has a clear
peak value. The ideal white noise model cannot reflect the actual characteristics of ground motion. Therefore, the Japanese scholars Kiyoshi Kanai and
H. Tajimi proposed a stationary filtered white noise model named Kanai-Tajimi model, which can reflect the frequency characteristics of the ground
vibration better. The model is based on the following principle: the earthquake causes the movement of bedrock, and then spreads through the surface
soil structure, as shown in Figure 6.6. The model assumes that the seismic bedrock acceleration Ü(t) is stationary white noise, of which the mean is zero
and the spectral density is S 0, the surface soil layer is equivalent to a filter. The cover soil is considered as a linear SDOF system, equivalent to a filter.
At this time, the movement equation of the cover soil is:
(6.91)

Figure 6.6 Filtered stationary white noise model.

In which ü(t) is the acceleration of the ground relative to the bedrock; ω g and ζ g are the frequency and damping ratio of the cover soil, respectively.
Thus, the surface acceleration is:
(6.92)

The power spectral density function of the ground acceleration can be obtained.
(6.93)

In the filtered white noise model, S 0 is a spectral parameter that reflects the degree of earthquakes. ω g and ζ g are related to the characteristics of the
cover soil, taking different values according to the general degree of softness and hardness of the soil. The filtered white noise model takes into account
the influence of the surface soil on the spectral characteristics of the ground motion, and has a clear physical meaning. It is a model of seismic ground
motion widely used in seismic engineering. However, the model has the disadvantage that it does not properly exaggerate the energy of low-frequency
ground motion and may obtain unreasonable results in seismic response analysis of certain structures (especially low-frequency structures). And the
seismic ground displacement, speed, and derivative of the acceleration cannot be determined by this model. Therefore, the filtered white noise model is
more suitable for the analysis of stochastic seismic response of medium and high frequency structures.

3. Improved Filtered White Noise Model

In order to improve the irrationality of the filtered white noise model in the low- frequency range, Yuxian Hu put forward the improvement scheme,
and corrected the defects of the filtered white noise model. The power spectral density function of the ground acceleration in the improved filtered
white noise model is:
(6.94)

In which ω c is cutoff frequency in low frequency range.

Comparison of the power spectral density function about the filtered white noise model and the improved filtered white noise model is shown in Figure
6.7. It can be found that the improved filtered white noise model only modifies the low frequency part of the filtered white noise model without
substantially changing the middle and high frequency part. The improved filtered white noise model fitted with the data from the observation of strong
earthquakes is more suitable for the actual seismic ground motion.

Figure 6.7 Filtered white noise model and improved filtered white noise model.
4. Housner model
Housner proposed power spectral density function of ground motion acceleration which can be expressed as follows from the study of the rate response
spectrum of an undamped SDOF system excited by the white noise process excitation.
(6.95)

In the equation, S V is the velocity response spectrum of an undamped SDOF system excited by the white noise process excitation and t 0 is the
duration. Housner’s study opened up the transition from response spectrum to power spectrum.

5. Multi-Peak Power Spectrum Model

The power spectrums of the ground motion models above, other than the stationary white noise model, are all single peak spectrums. But for a multi-
layer foundation, the multi-degree of freedom system should be considered as the soil filter. So, the power spectral density of the ground motion is
filtered by multiple layers can be expressed as:
(6.96)

In which j is the ordinal of the layer of foundation and N is the number of layers of foundation. As can be seen from the above equation, the multi-peak
power spectrum is obtained by superimposing all peak power spectrums together. In the study of practical problems, the parameters ξgj,ω gj,S 0j can be
adjusted according to the number of peaks required and the magnitude of the amplitude to obtain from the required multi-peak power spectral density
function. Figure 6.8 shows the simulation results of a three-layer foundation.
Figure 6.8 Power spectrum of acceleration of strong earthquake filtered by multiple layer.
6. Stabilized Stochastic Process Model
Models introduced above are all stationary, but it can be seen from the record of the seismic ground motion that they are all non-stationary stochastic
processes, which can be divided into three stages: the initial stage, when the seismic ground motion increases quickly; then the stationary stage, when
seismic ground motion maintains its average strength; lastly the decay stage, when the seismic ground motion decreases. Thus, the researchers usually
use the product of the intensity function which varies from time and the stationary process to reflect this non-stationary change which is called the
stabilized stochastic process. Its general form is as follows:
(6.97)

In the above equation, X(t) is a stationary stochastic process and f(t) is deterministic intensity envelope function.
At present, domestic and foreign researchers have conducted some research in this field. Intensity envelope function commonly used can be
summarized into two categories: segment function type and continuous function type. The intensity envelope function of segment function type is:
(6.98)

In the formula, c is the attenuation coefficient. t 1 , t 2 and t 3 are the time of the end of the ascending segment, the time of the end of the stationary
section, and the total time of the earthquakes. All four parameters can be calculated from the empirical formula of historical seismic records according
to the risk analysis method. The illustration of the intensity envelope function is shown in Figure 6.9.

Figure 6.9 Time envelope function of quasi-stationary process.

Intensity envelope function of continuous function type has the following several forms.
First:
(6.99)

Second:
(6.100)

Third:
(6.101)
In the formula, α,β,a,b,c,t p are all parameters to be determined.

6.3 Stochastic Response of Linear SDOF System

It has been mentioned that in the linear vibration system, the resilience of the system, the damping force, and inertial force can be expressed as a
linear function of the displacement, velocity, and acceleration of the system motion. The most essential characteristic of the linear vibration system is
that the excitation and the reaction are also linearly related, and consequently the linear superposition principle can be applied to the reaction analysis
of the system. In other words, the reaction of the system under multiple excitation is equal to the sum of the reactions under each excitation. In
Chapter 3, we have already mentioned that the dynamic response of a linear SDOF system can be obtained by time-domain analysis (Duhamel
integral) and frequency-domain analysis (Fourier transformation). If the seismic excitation is considered as a stochastic process, the structural
stochastic response can also be obtained by time-domain analysis and frequency-domain analysis. The time-domain analysis of the structural
stochastic response is to solve an analytic formula of the correlation function of dynamic response of the system. The frequency-domain analysis of the
structural stochastic response is to solve the power spectral density function of the dynamic response of the system, and then calculate the mean
square response of the system by the integral operation.

6.3.1 Time-Domain Analysis Method

The time-domain analysis method of dynamic response analysis of the linear SDOF structural system has been described in Chapter 3 – Duhamel
Integral. The process imposing external load P(t) on the structure is considered to be composed of a series of continuous short-time pulses. The
response of the system under every short pulse is obtained first, and then the total response of the system can be obtained using the superposition
principle. Therefore, the calculation of the SDOF system under the action of the unit pulse load, namely the pulse response function, is a very critical
step.
Assuming that the linear SDOF structural system is subjected to a unit pulse load δ(t – τ) at moment τ, the dynamic response of the system is:
(6.102)

In the formula, ζ is the damping ratio and ω D is the natural frequency of the damping structure. , in which ω n is the natural frequency of
the undamped structure. Set τ = 0, then the response of the SDOF system under the action of the unit pulse is obtained, namely the unit pulse
response function, denoted by h(t – τ).
(6.103)

The structural dynamic response caused by the impulse Ρ(τ)dτ at the moment τ is P(τ)dτh(t – τ). The total response caused by the external load P(t)
during the period [0,t] is:
(6.104)

The equation above is called Duhamel integral. Equation (6.104) can be written as follows.
(6.105)

The analysis of the stochastic response of linear system is the process to analyze the stochastic motion of the system and probability characteristics of
the solution process. According to the conclusion of linear stochastic differential equation, in the case of linearity, the mean square solution and the
deterministic solution have the same expression when the solution exists. As for the probability characteristics of the solution process, the easiest ones
to find in the response analysis of stochastic vibration and the most practical ones are its first moment and second moment, that is, the mean and the
correlation function.

1. The Mean of the Response

The output u(t) of the linear time-invariant SDOF structural system is a stochastic process. Its mean value is:
(6.106)

According to equation (6.104),

(6.107)

From equation (6.103), we can obtain:

(6.108)

When t → ∞,
(6.109)

mu(t) tends to be constant, and if the mean of the external excitation mp(t) is zero, the mean of the response mu(t) will be zero.

2. The Autocorrelation Function of the Response

The autocorrelation function of the stationary response process u(t) is:
(6.110)

Substitute the structural response of the formula (6.104) into the formula (6.110):
(6.111)
Exchange the order of mathematical expectation and integral operations:
(6.112)

Notice that E[P(t – u)P(t + τ – v)] = Rp(τ + u – v)]. So, the autocorrelation function of the response process can be expressed as follows.
(6.113)

Double integral operation is necessary in calculating the autocorrelation function of the structural stationary response process using the formula (6.113)
and the autocorrelation function Rp(τ) is difficult to obtain in actual engineering stochastic excitation process. Therefore, it is very rare to use the
formula (6.113) to calculate the structural stochastic response directly. However, once the autocorrelation function of the structural stochastic response
process is obtained, it is very easy to calculate the mean square response of the structure. Set τ = 0, we can obtain the following equation.
(6.114)

6.3.2 Frequency-Domain Analysis Method

In Chapter 3 we introduced the Fourier transformation method, carrying out a series of superposition calculation in the frequency domain. Similarly,
equation (3-146) is the frequency response function.
(6.115)

H(iω) is usually an imaginary number and can also be written as the following form:
(6.116)

In the formula above, Rd = |H(iω)| reflects the magnification of the response of the linear structural system to the external load. Angle φ represents
the phase difference between the external load and the structural response of the harmonic components of the same frequency. Rd and φ not only
depend on the characteristics of the structural system but also relate to the frequency of the excitation ω. Rd and φ are easily obtained through the
following equations:
(6.117)

(6.118)

According to the definition of the frequency response function, it’s easy to see that linear structural excitation and response have the following relation:
(6.119)

In the formula, U(ω) and Ρ(ω) are the Fourier transformation of the excitation of system - u(t) and P(t), respectively. The excitation mentioned here
refers to not only the external load acted on the structural system but also the ground motion. The structural response can refer to the displacement,
velocity and acceleration, and it can also refer to the internal force, such as, bending moment, shear force, and so on. For the sake of convenience, the
excitation is called the input, denoted by x(t) and the response is called the output, denoted by y(t) in the following content.

1. The Self-Spectral Density of the Response Process

According to the Wiener-Khintchine formula, the self-power spectral density function of the stationary response process is:
(6.120)

Using the equation (6.112), we can obtain:

(6.121)

Exchange the order of the integral operations and then sort out:
(6.122)

So:
(6.123)

Suppose H*(iω) is the conjugate function of H(iω).

(6.124)

The self-spectral density function of the structural response can be expressed as:
(6.125)

It can be seen from equation (6.125) that the self-spectral density of the response and the excitation have a very simple relationship, and it is easier to
calculate |H(iω)| than the double integral operation of the equation (6.113). So, calculating the self-spectral density is more common in engineering
analysis. If necessary, it can be transformed into autocorrelation function through the Fourier transformation.
The structural mean square response can also be obtained by using the spectral density function:
(6.126)

Compared with equation (6.114), the formula above requires integration. It can be seen that solving the correlation function is difficult, but it is easy to
calculate the mean square value through it. The spectral density function of the response is easy to find, but it is difficult to solve the mean square
response. The two methods have their own advantages and disadvantages. Therefore, they should be selected according to the actual situation.

2. The Relation between Pulse Response Function and Frequency Response Function
The pulse response function h(t) and the frequency response function H(iω) describe the characteristics of the structural system in the time domain
and the frequency domain respectively and they have some connections. In fact, h(t) and H(iω) constitute a Fourier transformation pair:
(6.127)

6.3.3 Cross-Correlation Function and Cross-Spectral Density of Excitation and Response

According to the definition of the cross-correlation function of the stochastic process, the cross-correlation function of the excitation process and the
response process is:
(6.128)

Using equation (6.111), we can obtain:

(6.129)

Exchange the order of expectation operation and integral operations,

(6.130)

(6.131)

The cross-power spectral density function of the excitation process and the response process is:
(6.132)

Substitute equation (6.131) to the above equation to obtain:

(6.133)

Exchange the order of the integral operations:

(6.134)

Resulting in:
(6.135)

Example 6.1 Derive the spectral density of the filtered white noise model of the acceleration of seismic ground motion.
Solution: This model is based on the fact that earthquakes cause bedrock movement that passes through the surface soil to the structure, as shown in
Figure 6.6. The acceleration of the bedrock caused by the earthquake is assumed to be white noise whose mean is zero. The soil layer on the bedrock is
considered as a linear SDOF system. Therefore, the equation of motion of the surface soil can be written as follows.
(6.136)

In the above equation, ζ g and ω g are the damping ratio and natural frequency of the surface soil, respectively. Xg(t) is the displacement of ground
relative to the bedrock. Ü(t) is the absolute acceleration of the bedrock.
The absolute acceleration of the ground is:
(6.137)

The equation above is the model of the seismic ground acceleration used as a structural excitation. The spectral density of the absolute acceleration in
the stationary state is the spectral density of the stationary filtered white noise model that we require.
From the equation of motion (6.136), the absolute acceleration ag(t) can be expressed as:
(6.138)

Thus, the correlation function in the stationary state a (t) can be expressed as:
(6.139)

Noticing that , the equation above can be transformed into:

(6.140)

Substitute and into equation (6.140):

(6.141)

In the equation above, , , ,

Using the relationship between spectral density and correlation function, we can obtain the spectral density of ags(t) from equation (6.142):
(6.142)

Substitute and into the above equation and take a notice of the formula of . We can

obtain
(6.143)

which is the spectral density of the white noise model simulating the acceleration of seismic ground motion. Since the mean of bedrock acceleration is
zero, the mean of this model is also zero.
Example 6.2 Analyze the stochastic response of the SDOF system under the excitation of the earthquake ground motion, shown in Figure 6.10.
The ground acceleration of the earthquake is a stationary white noise model established in Example 6.1. The equation of the motion of the system can
be written as:
(6.144)

Figure 6.10 Calculation diagram of single-layer structure.

Transform it into the standard equation.
(6.145)

In the equation above, , .

Considering that the mean of the excitation ags(t) is zero, we can know from the equation that the mean
of the response X(t) is also zero. Substitute equation (6.143) into
and Then we can obtain the correlation function of non-stationary and
stationary displacement response.

6.3.4 Fatigue Predictions for Narrowband Systems

In order to predict fatigue life of a narrowband SDOF system, we assume that the material follows a prescribed S-N relationship (S = harmonic stress
amplitude, N = number of cycles to failure) and Miner’s linear-accumulative-damage criterion applies.
The accumulative damage (AD) can be expressed in the discrete form

In which n1 , n2, n3, … are numbers of harmonic stress cycles applied to the material at amplitudes S1 , S2, S3, …, respectively. Failure occurs when the
accumulative damage reaches unity, that is, AD = 1.
If the system is responding as a narrowband system, the accumulative damage can be expressed in the continuous form

Where n(S)dS represents the number of harmonic stress cycles with amplitudes between S and S+dS. If a stationary response process of duration T is
assumed, the total number of stress cycles will be equal to ωΤ/2π, in which case

Where p(S) is the probability density function for stress amplitude S. The accumulative damage can be given in the form

Setting the accumulative damage equal to unity and solving for T gives the expected time to failure

6.4 Stochastic Response of Linear MDOF System

The analysis of the stochastic vibration of the SDOF system is a single-input and single-output problem. For a MDOF system, it is necessary to analyze
the response of multiple degrees of freedom, and the structure is often excited in more than one degree of freedom. So, stochastic response analysis is
usually a multi-input and multi-output problem. This section will introduce two basic methods for solving the stationary stochastic response of a linear
time-invariant MDOF structural system, specifically the direct method and the mode superposition method. The direct method starts from the system
vibration equation to find the stationary stochastic response of the system. It is an accurate method, but the size of the problem cannot be too large. So,
it’s suitable for analyzing the stochastic response of a system whose number of degrees of freedom is relatively small. The mode superposition method
takes advantage of the orthogonal characteristics of different vibration modes. It reduces the number of degrees of freedom significantly using
vibration transformation, so as to calculate the system stochastic response. The mode superposition method can solve the problem of large-size
structural analysis, but it is an approximation method requiring a linear system with a classical damping in general.

6.4.1 Direct Method

1. Time-Domain Analysis Method

(1) Vibration Response of MDOF System

Suppose that the system has m inputs, xj(t), j = 1,2, ···,m and n outputs, yk(t), k = 1,2, ···,n. According to the principle of superposition, each of the
outputs of the linear system can be obtained by superimposing the responses caused by the individual inputs. Thus, for each output yk(t), there are m
pulse response functions, namely hkj(t), k = 1,2, ···,n, j = 1,2, ···,m following equation holds:
(6.146)

Corresponding to all n outputs, there exist n × m pulse response functions. Arrange these functions in a matrix as follows.
(6.147)

The matrix above is called the pulse response matrix of the system. If the n outputs of the system are expressed as a vector as follows:

And the inputs are:

Then the response of the system can also be written in the form of matrix, namely:
(6.148)

(2) The Mean Vector of the Response of System

The mean of the n stationary response process is:
(6.149)

Substitute equation (6.148) into equation (6.149):

(6.150)

Exchange the order of expectation operation and integral operation:

(6.151)

Considering that m inputs are all stationary process, we can obtain:

(6.152)
In the above equation, {mx} is the mean vector of the stationary excitation process of the system.
Because [H(iω)] is the Fourier transformation matrix of [h(t)],
(6.153)

(6.154)

Obviously, the mean response of the MDOF system can also be calculated using the way of statics, that is, applying the average of the m input
processes as statics to the system and then calculating the responses they cause in the k-th degree of freedom using the way of statics, after which we
can obtain the mean of the stationary process of the system in the k-th degree of freedom by superimposing the m static responses.

(3) Correlation Function Matrix of the Response of the System By arranging m autocorrelation functions and the cross-correlation functions of the
inputs process in a matrix form, we can obtain the correlation function matrix of the input:
(6.155)

Similarly, the n × n order matrix consisting of the autocorrelation functions and the cross-correlation functions of the output process is called the
correlation function matrix of the output.
(6.156)

From equation (6.148), we know that the transpose vector of {y(t)} is:
(6.157)

Substituting equation (6.148) and equation (6.157) into equation (6.156) gives:
(6.158)

Exchanging the order of expectation operation and integral operation results in:
(6.159)

So, the correlation function matrix of the response of the system is expressed as the following.
(6.160)

The main diagonal elements of the correlation function matrix of output describe the autocorrelation functions of the stationary response process of the
system. The other elements are the cross-correlation functions of the different response processes. If the main diagonal elements are picked up to form
a vector, we obtain the autocorrelation function vector of the response of the system.
(6.161)

The symbol ⊗ indicates that the corresponding elements in the two vectors are multiplied.
In the equation (6.161), set τ = 0 and then we obtain the value of the each mean square response:
(6.162)

(4) Cross-Correlation Function Matrix of Excitation and Response Arrange m cross-correlation functions of inputs and n cross-correlation functions
matrix of outputs in a m × n order matrix, denoted by:
(6.163)

Substituting equation (6.157) into equation (6.163) gives:

(6.164)

Exchange the order of expectation operation and integral operation:

(6.165)

So, the cross-correlation function matrix of excitation and response is:

(6.166)

Similarly, arrange n cross-correlation functions matrix of outputs and m cross-correlation functions of inputs in a n × m order matrix, denoted by
(6.167)

After similar derivation, we can obtain the cross-correlation function matrix of response and excitation.
(6.168)

Notice that [Rxy(τ)] and [Ryx(τ)] has the following relation:

(6.169)

2. Frequency-Domain Analysis Method

(1) Vibration Response of MDOF System

Fourier transformation of equation (6.147) is carried out to obtain the complex frequency response matrix of the system.
(6.170)

In the equation above,

(6.171)

The output and the input have a relationship in the frequency domain, as is shown below.
(6.172)

In this equation, {Y(ω)} and {X(ω)} are the Fourier transformation of the input vector {Y(t)} and the output vector {X(t)}, respectively.

(2) Spectral Density Function Matrix of the Response of the System Self-spectral density functions and cross-spectral density functions of m input
processes constitute an m × m order power spectral density function matrix of the input, denoted by
(6.173)

Self-spectral density functions and cross-spectral density functions of n output processes constitute a n × n order power spectral density function
matrix of the output, denoted by:
(6.174)

Substituting equation (6.171) into equation (6.174), we can obtain:

(6.175)

Exchanging the order of integral operations, we can obtain the following equation after arrangement.
(6.176)

Put another way:

(6.177)

In the equation above, .

The self-spectral density function vector of the response of the system is

(6.178)

It is a vector constituted by the main diagonal elements of the spectral density function matrix of the output. The mean square response of the system
in the k-th degree of freedom is:
(6.179)

In this equation, S yk(ω) is the k-th element of {S y(ω)}.

(3) Cross Spectral Density Function Matrix of the excitation and the response
Cross Spectral Density Functions of m inputs and n outputs constitute an m × n order matrix, denoted by:
(6.180)

Substitute equation (6.166) into equation (6.180),

(6.181)
Exchanging the order of integral operations, we can obtain the following equation after arrangement:
(6.182)

So, the cross spectral density function matrix of the excitation and the response is:
(6.183)

Similarly, cross spectral density functions of n outputs and m inputs constitute an n × m order matrix, denoted by:
(6.184)

Substitute equation (6.166) into equation (6.184),

(6.185)

The following equation can be obtained after exchanging the order of integral operations:
(6.186)

So, the cross spectral density function matrix of the excitation and the response is:
(6.187)

Notice that [S xy(ω)] and [S yx(ω)] have following relation:

(6.188)

6.4.2 Vibration Mode Superposition Method

Suppose that the degree of freedom of the linear MDOF system is N. The vibration equation of the system under the excitation of the input vector {x}
is:
(6.189)

In the above equation, [M], [C], and [K] are the mass matrix, the damping matrix, and the stiffness matrix, respectively, which are all N × N order
matrix. {y} is the output vector of the system, the dimension of which is N.
In general, the equations in equation (6.189) are coupled. In order to solve the above equations, set
(6.190)

In this equation, [Φ] is the vibration mode matrix constituted by the first n vibration modes (n << N here). {q} is the generalized coordinate vector.
Substitute equation (6.190) into equation (6.189) and multiply the result by [Φ]T in the left.
(6.191)

In this equation, , , , . As described in Chapter 4, and are both diagonal matrixes. If the
system has classic damping, then the equations in equation (6.191) can be decoupled, that is, is also the diagonal matrix. The equations in equation
(6.191) are not coupled to each other, and each equation corresponds to a vibration mode, which is equivalent to an SDOF structure. According to the
analysis method of the stochastic response of the linear SDOF system, the statistics of the response of each vibration mode coordinate qj(j = 1,2,···,n)
can be obtained. Finally, using the transformation relationship of equation (6.190), the statistical characteristics of the output {y} of structural system
are calculated.
The mean square value of the response of the system’s j-th degree of freedom is:
(6.192)

In this equation, φ jk is the k-th element of the j-th vibration mode.

Obviously, there is a correlation between the vibration coordinates qi(j = 1,2,···, n). It is generally believed that when the natural frequencies of the
system are close to each other, the correlation between the coordinates of corresponding vibration modes, qi and qj, is strong. By contrast, when two
natural frequencies of the system are far from each other, qi and qj tend to be independent of each other. Therefore, it is necessary to consider the
combination of vibration modes when calculating the mean square response of the system. At present, a variety of vibration combination methods have
been proposed; the following three methods are used more commonly.
1. SRSS Method
(6.193)

The method assumes that varies vibration modes are independent of each other.
 2. ABS Method
(6.194)

The method assumes the correlation coefficient between all modes is ρ = 1,–1.
3. CQC Method
(6.195)

In this equation, σji = φ jiσqi, ,

Example 6.3 Stochastic excitation of the foundation’s movement acted on the MDOF system. The multi-layer shear structure under the action of
seismic ground motion is shown in Figure 6.11. Solve the mean and the correlation function of the response.

Figure 6.11 Multi-layer shear structure under the action of seism.

Solution:
The equations of motion of the system can generally be written as:

In this equation, {l} is N-dimension vector whose all elements are 1. F(t) = ag(t) is the acceleration of the movement of the foundation. For the multi-
layer shear structure under the action of seismic ground motion, .

Writing in component form,

In this equation,

The stochastic response of the displacement of the system can be expressed as:

In this equation, . The mean of the response of the displacement can be obtained from the above equation.

The cross-correlation function is:

When F(t) is a one-dimension earthquake stochastic model in different form, the statistical moments can be obtained by using the above method.
Example 6.4 When the long-span structure is excited by the seismic ground motion, as shown in Figure 6.12, if the acceleration of the seismic
ground motion at the first support is ag(t), the excitation of the seismic ground motion at the second support may either lag or advance for a period of
time u2. Thus, the acceleration of the seismic ground motion at the second support can be expressed as ag(t + u2). The acceleration at the second
support is the same stochastic process as that at the first support, but only lag or advance for a period of time u2. When it lags, u2 is a negative amount;
when it advances, u2 is positive quantity, and so on. The acceleration of the seismic ground motion at the i-th support can be expressed as ag(t + ui).

Figure 6.12 Large-span structure under the action of seism.

Solution:
If the distance between the j-th and the j + 1-th support is Lj,j+1 and the average velocity of seismic wave propagation in this span is Vj,j+1 , the time
difference when earthquake wave arrives the i-th support is:

When i = 1, ui = 0.
Suppose the process of the acceleration of the seismic ground motion at the i-th support is denoted by F i(t), namely:

The cross-correlation function of the acceleration of the seismic ground motion at the i-th and the j-th support can be expressed as:

In this equation, Rag is the correlation-function of the acceleration of the seismic ground motion at the first support. It can be seen that the correlation
function or cross-correlation function of seismic excitation at each support can be expressed by the correlation function at the first support.
In general, if the MDOF system is subjected to the same stochastic excitation at a different time in a different position, its equation of motion can be
expressed as:

In this equation,
{F(t)} = {F 1 (t),F 2(t),···,F N(t)}T = {F(t+u1 ),F(t+u2),···,F(t+un)}T is the same stochastic excitation of different time and positions. If the time of the first
component F 1 (t) is taken as reference time, u1 = 0.
If the correlation function of F(t) is RF(t 1 ,t 2), the cross-correlation function of the stochastic excitation F i(t) and F j(t) can be expressed as:

From the equation:

we can obtain the stochastic response of the displacement of the system:

So, the mean is:

The correlation function is:

When F(t) is a modulated non-stationary process and the mean is zero, considering:

the components of the stochastic excitation {F(t)} can be expressed as:

The cross-correlation function is:

Substitute the above equation into:

After arrangement we can obtain:

In this equation:

When F(t) is an evenly modulated non-stationary process and the mean is zero, considering:

the components of the stochastic excitation {F(t)} can be expressed as:

The cross-correlation function is:

In this circumstance, the cross-correlation function of the response of the displacement can still be expressed by the following equation:

All that is necessary is to set A*(s+ul,ω) = f(s+ul) in the equation

When F(t) = F S(t) is a stationary process and the mean is zero, considering:

the cross-correlation function of the stochastic excitation F i(t) and F j(t) can be expressed as:

In this equation, S FS(ω) is the spectral density of the stationary process F S(t).
Substitute the above equation into:
We can obtain:

Jsil(ω,t) in this equation is defined as:

In the above equation, I k(ω,t) is defined as:

After quite a long time t, the response tends to be stationary. And in the stationary state, the above equation transforms into:

In the above equation, Vg(ω) is defined as:

When F S(t) is white noise and its mean and density is zero and S 0 respectively, the cross-correlation function of the component of the stochastic
excitation {F(t)}, F i(t) and F j(t), can be expressed as:

In this equation, τ = t 2 –t 1 ; τ ij = uj –ui. And cross spectral density of F i(t) and F i(t) is

In the case where the damping of each vibration mode is relatively small and the frequencies are not close to each other, the cross-correlation function
of the stationary response of the displacement Xi(t), Xj(t) can be expressed as:

In this equation, set i = j, τ = 0, then the mean square value of Xi(t), the stationary response of the displacement, can be obtained.
It can be seen from the discussion that the stochastic response of the linear MDOF system can be analyzed by direct method. When the number of the
degree of freedom of the system is large, the vibration mode superposition method can be used to approximate the solution. When using the vibration
mode superposition method, the system must have the classical damping, and the coupling effect between the different modes is neglected. In order to
overcome these limitations, Professor Lin Jiahao proposed a virtual excitation method for structural stochastic response analysis. This method can
automatically take into account the cross-correlation between the vibration modes participating in the vibration and the cross-correlation of the
excitation when dealing with large-scale complex structures subjected to stochastic excitation in multiple positions. In theory, an exact solution is
obtained. At the same time, this method is very convenient for dealing with the matrix of non-classical damping. In the finite element program which
has the function of simple harmonic vibration or time-history analysis, we can further develop the self-spectrum of various stationary or non-
stationary responses such as displacement, internal force, and strain. Taking advantage of the knowledge of the stationary response of the structure
subjected to simple harmonic excitation, it can analyze rather difficult stochastic structural vibration. It can not only solve the shortcomings of the
mode analysis method, but also make full use of the advantages of the modern digital computer, which helps greatly improve the computational
efficiency. As for the virtual excitation method, interested students can refer to the relevant literature.

6.5 Nonlinear Structural Stochastic Response Analysis

In general, the nonlinear problems in engineering include geometric nonlinearity and nonlinearity of material properties. Their forms are mainly
expressed in the damping term and the stiffness term or the coupling term of damping and stiffness. Here we only discuss the problem of nonlinearity
of material properties.
Strictly speaking, the vibration of the structural system is always in various degrees of nonlinearity. Only in the case of the micro-vibration in small
deformation, the majority of the nonlinear structural system don’t show obvious nonlinear characteristics so that these nonlinear structural systems
can be described as equivalent linear systems model. However, for large-scale vibration or vibration of the system with nonlinear elements, we must
describe the nonlinear model and discuss the corresponding method of nonlinear response analysis.
The nonlinear vibration response analysis is much more complicated and difficult than the linear case. The fundamental reason is that the superposition
principle is not suitable for nonlinear vibration. For this reason, the Duhamel integral method and the mode decomposition method, which suit the
analysis of the linear vibration, are not applicable for the analysis of the nonlinear vibrations. In addition, there is another problem in the nonlinear
stochastic vibration, which is that the response of the system is not necessarily normal type under the normal stochastic excitation because of the
nonlinearity. This makes it infeasible for us to obtain the probability distribution of the response directly from the second-order statistics of the
response.
The deterministic vibration of the nonlinear system is very different from that of the linear system. Three main differences are as follows:
1) The superposition principle isn’t applicable. When the excitation is f(t) = f1 (t)+f2(t), the response of the nonlinear system is:

Due to this characteristic, the method of Duhamel integral and mode decomposition method, which are based on the superposition principle, are
not applicable for the analysis of the response of the nonlinear vibration.
2) The free vibration frequency of the nonlinear system is related to the amplitude and is no longer an inherent value.
3) The forced vibration of the nonlinear system has a jumping phenomenon, which is an inherent characteristic of nonlinear forced vibration.
At present, although there are many research results on stochastic seismic response analysis of nonlinear structures, there are still many problems
that have not been solved. There are still many difficulties and differences in practical engineering design. One reason for this is the complexity of the
problem itself. Another reason is that there is a lack of the records of strong earthquakes, and that the existing records of strong earthquakes vary in
duration, making it difficult to obtain enough stochastic sample function for statistical analysis. Response analysis of the nonlinear vibration doesn’t
have a universally applicable general method, only some special methods can be used to reveal the movement state of the system as much as possible.
Only a few representative methods are introduced briefly here.

6.5.1 Perturbation Method

In addition to the existence of a main vibration, there are small additional vibrations in a vibration system. These small additional vibrations are like a
micro-disturbance and so-called perturbation. The perturbation method is the method using perturbation to solve nonlinear vibration systems, which
is also called small parameter method in mathematics. The perturbation method in nonlinear stochastic vibration is basically the same as that used in
deterministic analysis.
Consider the equation of the motion of the following nonlinear SDOF system.
(6.196)

F(t) in the equation (6.196) is the stochastic excitation of the system and ε is the small parameter (0 < ε < 1). Suppose , the function
indicating the nonlinearity, is continuous, and independent variable y and velocity is the first-order partial derivative. According to perturbation
method, let the solution of the equation (6.196) be taken as the rising power polynomial of ε.
(6.197)

So
(6.198)

(6.199)

According to the Taylor series formula of the multivariate function, can be written as:
(6.200)

The last item to the right of the above equation is the remainder and 0 < θ < 1. Substitute equation (6.197) and equation (6.198) into equation (6.200)
and notice that . After arranging all the items by the order of the power, we can obtain.

(6.201)

Substitute equations (6.197), (6.198), (6.199), and (6.201) into equation (6.196), let the power of ε be the same and we can obtain:
(6.202)

Equation (6.202) indicates that each item of equation (6.197) satisfies a linear differential equation with stochastic input, which can be solved as a
stochastic linear equation. Obviously, after finding the first approximation y1 (t) and the second approximation y2(t) of y(t), the later calculation
becomes more and more complicated. But we normally only have to find y1 (t) and y2(t). This depends on the size of ε and accuracy requirements.
It can be said that the perturbation method is an effective way to solve the nonlinear problem with elastic nonlinearity (nonlinearity without history
characteristics). But for a strong nonlinear problem with historical characteristics, the perturbation method is no longer applicable.

6.5.2 Equivalent Linearization Method

The equivalent linearization method is also an approximation method applied to the deterministic nonlinear problem. The basic idea is that a nonlinear
equation can be approximated by an equivalent linear equation, where the equivalent parameter is determined by the condition where the mean
square of the difference between the two equations is the smallest. Since these parameters (frequency, stiffness, or damping) are related to the factors
to be determined, it is generally done by iterative method.

1. Nonlinear SDOF System

1. Nonlinear SDOF System
As for the stationary stochastic vibration of a nonlinear SDOF system, let the equation of the motion be:
(6.203)

In equation (6.203), is the nonlinear function of the displacement y(t) and the velocity , of which the initial condition is f(0, 0) = 0. F(t) is
a stationary Gaussian process and its expectation is E[F(t)] = 0.
The equivalent linearization method is to approximate the nonlinear equation of equation (6.203) by a linear equation (6.204). In this case, as long as
the mean square of the difference between the two formulas is the smallest.
(6.204)

Suppose:
(6.205)

Then:
(6.206)

(6.207)

The key of this method is to solve c and k, the equivalent parameters to be determined from equation (6.207).
The problem is that the stationary joint probability density function in equation (6.206) is unknown and it is necessary to make some approximate
assumptions about the distribution of the probability density function according to the nature of the problem. For example, assuming that the
nonlinearity of the original system is weak, the joint probability density of the linear system can be approximated. For a linear system, the input is a
stationary Gaussian process and the response is generally stationary Gaussian process. So, choose the following formula whose mathematical
expectation and correlation coefficient are both zero.
(6.208)

In this equation, y = y(t). c and k, the parameters given by equation (6.208) and are commonly the implicit functions of σY and . As a result, iterative
method is always used to determine the equivalent parameters c and k.

2. Nonlinear MDOF System

When a n-degree of freedom system is subjected to horizontal stationary stochastic excitation F(t), its nonlinear equation can be expressed as:
(6.209)

In equation (6.209), is the vector of nonlinear generalized restoring force and [M] is the mass matrix. {I} is the vector, in which the
element corresponding to the direction of horizontal displacement Y is 1 and the rest is zero.
According to the aforementioned equivalent linearization direction, the following linear equation can be used to replace the equation (6.209)
(6.210)

In equation (6.210), [C] and [K] are the damping matrix and the stiffness matrix respectively. Due to the conditions, the error in the stochastic vector
{ε} is extremely small,
(6.211)

(6.212)

Equivalent matrix [C] and [K] can be obtained. In the equation (6.212), Cij and Kij are the element in [C] and [K] respectively. It can be known from
the equation (6.212) that:
(6.213)

According to the equation (6.213), the equation (6.211) can be expressed as:
(6.214)
When {f(y(t), ẏ(t))} is known, [C] and [K] can be solved using equation (6.214). For the application of the equivalent linearization direction, interested
readers can refer to the relevant reference book.
The equivalent linearization method can be used to solve the nonlinear problem in which the restoring force has history characteristics. However, as
the degree of nonlinearity becomes stronger, the degree of approximation deteriorates, and the above method isn’t applicable when large plastic
deformation is considered.

6.6 State Space Method for Structural Stochastic Response Analysis

The modern control theory established the analysis method based on the state space method, which changed the complex frequency response function
analysis method based on the classical control theory. The state space method is a method for the analysis of a dynamic system. It can be applied
widely and provide more information of the system. It can solve many previously unsolvable or difficult problems. At the same time, it’s easier to
calculate by digital computer. So in the modern engineering control theory, this method has received great attention and has been widely applied.
Linear structure system is a linear mechanics system with constant coefficients. It can be analyzed by the powerful advantage of state space method.
The same goes for the solving of the statistical characteristics of stochastic response.

6.6.1 Basic Concept of State Space

The dynamic properties of the structural system are generally described by differential equation of motion, which establishes a bridge between the
input and output of the system, which is the most familiar description method. Differential equation is not the only way to describe the dynamic
characteristics of a system. In fact, the state of the system at any time can be described by state variables. The state here refers to a complete
description of the behavior of the system. If the initial conditions of the structural system at the moment t = t 0 are y(t 0) and ẏ(t 0), the output can
generally be written as:
(6.215)

In this equation, {z(t 0)} = {y(t 0), ẏ(t 0)}T is the state of the structural system at the moment t = t 0. x(t) is called the input (or excitation) of the system
and y(t) is called the output (or) of the system.
Thus, the output of the system is not only related to the input x(t) in the duration [t 0, t], but also to the initial state at the moment t = t 0. In general, if
the initial state {z(t 0)} of the structural system and the input x(t) after the time t 0 are known, the output y(t) can be uniquely determined by equation
(6.215). When t ≥ t 0, such n variables, z1 (t), z2(t),···zn(t), which is the least to uniquely determine the output y(t). z1 (t), z2(t),···zn(t) are known as the
state variables and the vector constituted by them is called the state vector, of which the state space, a n-dimension space, is composed. The state of
the system at any time can be described by a point in the state space. State variables describe the state of the system at some moments.
The state of the system in the moment t 0 contains all the historical information of the system before this moment. Knowing the state of the system at
t 0 and the input after t 0 is enough to solve the output of the system at any time. Since the choice of the time t 0 corresponding to the initial state is
arbitrary, the state of the system at time t includes all the information of the system at that time. So, the equation (6.215) can be written as:
(6.216)

The above equation indicates that the output y(t) of the system is determined by the state {z(t)} at the moment t and the input x(t).

6.6.2 SDOF System

The equation of motion of a linear SDOF system subjected to the excitation x(t) is:
(6.217)

In this equation m, c, and k are the mass, damping, and stiffness of the system, respectively.
Choose the state vector:
(6.218)

The equation (6.218) can be written as mż1 – cż2 + kz2 = x(t)

And then add the equation mż2 –mz1 = 0.
Write it into the form of matrix.
(6.219)

In the equation, , ,

The equation (6.219) is called the equation of state, which is not unique. The system is transformed into equation (6.219) to keep the symmetry of [M]
and [K]. The characteristic equation of the system is:
(6.220)

After expansion we obtain mp2 + cp + k = 0. Considering the case of small damping, namely ζ < 1, the characteristic equation (6.220) of the system has
a couple of complex eigenvalues.

In this equation, . The eigenvector equation corresponding to the eigenvalues of the system is:

(6.221)

The two eigenvectors of the system are obtained , , denoted by:

(6.222)
Apply complex mode transformation to the above equation,
(6.223)

Substitute equation (6.233) into equation (6.219) and multiply the result by [Φ]T on the left. After arrangement we obtain:
(6.224)

In this equation, , .

Noticing that equation (6.224) has been decoupled, its solution is:
(6.225)

In this equation [h(t)] is the matrix of pulse response function, which has the following form.
(6.226)

According to equation (6.225), we need to know the correlation function matrix of generalized excitation {F(t)} when compute the correlation function
of complex vibration response {q(t)}. To this end, we assume that the input x(t) of the linear structural system is a stationary process with its mean
equaling to zero, and that its autocorrelation function is RX(τ). Then, the generalized stochastic excitation of the system also has a mean of zero, and its
correlation function matrix is:
(6.227)

In this equation, . So, the complex modal response of the system is also a stationary process of zero mean, and its correlation
function matrix can be written as:
(6.228)

Substitute equation (6.225) into the above equation and exchange the order of the operation of expectation and the integral operation,
(6.229)

Considering [h*(t)] is a diagonal matrix, [h*(t)] = [h*(t)]T. Equation (6.229) is transformed into:
(6.230)

Substitute equation (6.227) into the above equation,

(6.231)

Using equation (6.223), we can obtain the correlation function matrix of the response of the state vector of the system,
(6.232)

Since the stationary process is orthogonal to its derivative process, [RZZ(τ)] should be a diagonal matrix.

6.6.3 MDOF System

Considering that the linear system subjected to the vector {x(t)} = {x1 (t), x(t),···, xm(t)}T constituted by m inputs, its vibration equation is
(6.233)

In this equation, [M], [C], and [K] are the mass matrix, damping matrix, and stiffness matrix of the system, respectively. Suppose that the order of the
response vector {y(t)} is N.

Introducing the state vector we can rewrite equation (6.223).

(6.234)

In this equation, , , .

Notice that and maintain symmetry. The characteristic equation of system (6.234) is as follows.
(6.235)

From equation (6.235), 2N eigenvalues of the system, pj(j = 1,2, ···, 2N), can be obtained. The eigenvectors equation corresponding to these eigenvalues
are:
(6.236)

Thus, 2N eigenvectors can be obtained, {φ}j (j = 1,2, ···, 2N). If it is denoted by:

It can be proved that the eigenvectors of the system have the following weighted orthogonality. That is to say, when j ≠ k,
(6.237)

When j = k, there exists the following relation:

(6.238)

Apply modal transformation,

(6.239)

Substitute equation (6.239) into equation (6.234) and multiply the result by [Φ]T on the left. We can obtain:
(6.240)

In this equation, , , , . Noticing that

equatin (6.240 has been decoupled, the quation can be written as:
(6.241)

In this equation, .

It can be seen that the complex vibration mode response equation (6.241) of the MDOF system is exactly the same as the complex modal response
equation (6.224) of the SDOF system in the form, except that the number of decoupled equations of the MDOF system has been expanded to 2N by
using complex vibration transformation. Therefore, the results of the stochastic response analysis of the aforementioned SDOF system can be
generalized for the MDOF system after similar derivation.

Exercises
6.1 Which types of stochastic models can be used to simulate seismic ground motions?
6.2 What is the train of thought in analyzing stochastic vibration of the linear freedom system in the time domain and frequency domain?
6.3 What are methods for analyzing the stochastic vibration of nonlinear structures? What is their basic thought?

References
1. Zhao-Dong, X and Le-Wei, M., Dynamics of Structures, Science Press, Beijing, 2007.
2. Clough, R.W. and Penzien, J., Dynamics of structures, in: Dynamics of Structures, 2005.
3. Chopra, A.K., Dynamics of Structures, Prentice Hall, 2000.
4. Wilson, J.F., Dynamics of Offshore Structures, 1984.
Chapter 7
Research Topics of Structural Dynamics
The majority of structural research in Civil Engineering, Mechanical Engineering, and Aeronautical Engineering is research about structural dynamics.
It refers to dynamic response, vibration control, modal analysis, damage identification, nonlinear analysis, and many other research projects. No matter
what kind the structure it is, those are the most prominent and realistic problems. This chapter will give an introduction to these popular research
topics.

7.1 Analysis of Structural Seismic Response

Earthquakes are among the main natural disasters that threaten human security. Now, the mainland of our country has entered the fifth seismic active
period. In recent years, several violent earthquakes have happened at home and abroad, such as the Hanshin Earthquake in Japan in 1995, the Indian
Ocean Earthquake and Tsunami Earthquake in 2004, and so on. This makes us learn more about the destructive force of earthquakes and accelerates
the developments of the seismic design theory and design technique at the same time.
People use seismographs to research and record earthquake intensity, which refers to earthquake magnitude. The seismograph can record the
velocity, acceleration, and displacement of the ground motion during each earthquake. The first seismograph in the world was invented by the Chinese
scientist Zhang Heng in AD 132. But in the field of structural seismic resistance, the first modern seismograph is the standard seismograph invented by
American Wood Andson in 1935. Figure 7.1 is the earthquake data recorded at the El-Centro Seismic Station during the 1940 US Imperial Valley
earthquake, which is nowadays the most widely used, famous El-Centro Wave. Since time records can’t be continuous, an important index is the time
interval Δt. Taking El-Centro wave as an example, the time interval of acceleration record is Δt = 0.02s.

Figure 7.1 Acceleration records of El-Centro ground motion.

The most important parameter between seismic recording and structural seismic resistance is the acceleration record. As can be seen from Section 2.4:
When the ground acceleration üg(t) is known, the equivalent external load of the structure Peff(t) = –müg(t) can be found by using Newton’s second
law. Of course, this external load does not directly act on the building structure, but the following analysis shows that the size of the equivalent external
load can reflect the structural seismic response.
China’s current Seismic Design Code for Buildings (GB50011-2010) provides the main seismic action calculation method, including vibration mode
decomposition method, bottom shear method and time-history analysis method. The theoretical basis of the first two methods is the seismic response
spectrum theory of the forced vibration of an SDOF system.
The differences between the structural seismic response and the vibration problems of the SDOF or MDOF system mentioned in Chapters 3 and 4 are:
Earthquake “load” is a series of discrete acceleration records, and can only be converted into a true external load in the forced vibration equation after
being multiplied by the mass of the corresponding structural system. Moreover, the external force of the dynamic equation in structural dynamics
refers to the force directly applied to the system or the particle, but the earthquake is only the irregular motion of the ground. So, the external force
must be transformed into the structural system through the “foundation or basic incentives” described in Section 2.4. Finally, it should be noted that,
due to the complexity and uncertainty of the earthquake, how to determine the intensity of earthquake action that a building can encounter, and what
parameters can be used to evaluate the impact of an earthquake to a building structure, are the problems that should be addressed in the section.

7.1.1 Brief Introduction to the Calculation Method

First of all, the development of calculation methods of seismic action will be briefly introduced. The calculation method of seismic action has been
gradually developed and so far, there are mainly four types of calculation theory of seismic action.

1. Static Theory
The scholar who first proposed the seismic load calculation equation is the Japanese seismologist Professor Dasen. He assumed that the horizontal
acceleration of the various parts of the structure was exactly the same as that of the earthquake itself. Therefore, if the weight of the structure is
denoted by Q, the absolute value of the maximum inertial force acting on the structure due to the seismic action is:
(7.1)

In which, τ 0 is the absolute value of the maximum horizontal acceleration of the ground and g is the gravity acceleration. Because the vibration of the
structure itself is not taken into consideration, it is called static theory.

2. Definite Function Theory

Because the static theory does not take the structural dynamic performance into account, there is a big flaw. Later, the Soviet Union’s Professor
K.C.Завриев criticized the theory proposed by Dasen Fangji and put forward a new dynamic theory. He used the form of cosine function of time to
describe the ground motion, which is: u(t) = a0 cos θt. On this basis, the equation (7.1) can be transformed into:
(7.2)

In the equation, β is a coefficient related to structural stiffness; Ke is a coefficient related to seismic intensity.

3. Response Spectrum Theory

Response spectrum theory is based on the response of the single-point elastic system under the actual earthquake action to analyze the structural
seismic response. In 1942, Biot of the United States firstly published the relationship between the acceleration response and the self-oscillation period
T of the single-point system, which started the research on response spectrum theory. Professor Hu Yuxian and Professor Zheng Zhemin first
introduced the response spectrum theory to China. Later, Professor Liu Huixian also conducted a more comprehensive summary. At present, this
theory is widely used in seismic design codes for building structures in many countries, including China.

4. Direct Dynamic Analysis Theory

Direct dynamic analysis theory is based on the direct integration of the structural dynamical equation to solve the seismic response along the change of
time and finally obtain the so-called time-history curve. Therefore, this method is also called time-history analysis method. The method is based on the
development and widespread use of computers. The actual seismic acceleration time-history record can be input to a computer to analyze the
structural seismic response.
China’s current Seismic Design Code for Buildings provides: For particularly irregular buildings, Class A building, buildings at Class I, Class II site in 8
degree area, buildings higher than 100m in 7 degree area, buildings higher than 80m at Class III, IV site in 8 degree area and buildings greater than
60m in 9 degree area, elastic time-history analysis method is necessary as supplementary calculation of the seismic bearing capacity and deformation
under normal earthquake.

7.1.2 Horizontal Seismic Action of SDOF Elastic System

When the base of the building structure moves horizontally, according to equations (2.28) and (2.29) the inertial force F I(t) acted on the particle in an
elastic SDOF system can be obtained:
(7.3)

The damping force on the right side of the above equation is a slight trace that can be omitted relative to the elastic restoring force ku(t), so:
(7.4)

So, under the action of the earthquake, the relative displacement u(t) of the particle at any moment will be proportional to the momentary inertial
force – at that moment. From (7.4) we can see: u(t) ≈ F I(t)/k, and u(t) is the relative displacement of the particle promoted by the
earthquake, and F I(t)/k is the relative displacement produced by the inertial force. Therefore, the inertial force can be understood as an equivalent
load that can reflect the effects of earthquakes. Using the maximum value to conduct the seismic check of the structure, we can make the seismic
design of the dynamic calculation problem be equivalent to the static calculation problem under static load.
The absolute acceleration a(t) of the particle can be determined by Eq. (7.4):
(7.5)

The expression of the seismic displacement response u(t) is substituted into the above equation:
(7.6)

If S a is equal to the maximum value of the absolute value of the above equation, we must calculate S a in order to obtain the largest seismic action
suffered by the structure during the duration of the earthquake. It can make seismic design easier because the acceleration of the ground motion üg(τ)
varies with time:
(7.7)

From the above equation, the absolute maximum acceleration S a of the particle depends on the acceleration of the ground motion üg(τ) during the
earthquake, the vibration period T of the structure and the damping ratio ζ of the structure. However, due to the fact that the acceleration of the
horizontal ground motion is very irregular, we can’t use a simple analytic expression to express it. Therefore, we generally use the numerical
integration method to calculate it.

7.1.3 Seismic Response Spectrum

According to (7.7), if the acceleration record of the ground motion for the given earthquake [Figure 7.1(a)] and the damping ratio of the system are
given, the curve describing the relationship between the maximum acceleration response S a about the particle and the system self-oscillation period T
can be acquired. This relationship curve S a – T is called the acceleration response spectrum. Of course, according to a similar approach, we can also find
displacement response spectrum or velocity response spectrum of the system. If the damping ratio is changed, the shape of the response spectrum will
also change.
Since the seismic acceleration response spectrum is the most widely used in the earthquake resistance of the building structure, we will only introduce
the calculation method of the acceleration response spectrum. Displacement and velocity response spectrum will not be introduced for simplicity. The
calculation process of the acceleration response spectrum is shown in Figure 7.2.

Figure 7.2 The process of calculating the accelerationg response spectrum.

Figure 7.3 shows the acceleration response lines plotted according to the 1940 El-Centro seismic waves. Its abscissa is the vibration period T of the
structure, of which the unit is seconds; the ordinate is the acceleration, of which the unit is the gravity acceleration g. As can be seen from Figure 7.3,
the acceleration response spectrum is a multi-peak curve. For a 2% damping ratio, the maximum acceleration response is greater than one
gravitational acceleration g. When the self-vibration period T of the structure is larger than 1s, the acceleration response of the particle is reduced. In
addition, as the damping ratio decreases, the spectral value of the acceleration response spectrum increases. When the vibration period of the structure
is small, the spectral value increases sharply with the increase of the period, but after the peak point the spectral value gradually decreases with the
increase of the period and gradually becomes gentle.

Figure 7.3 El-Centro seismic wave acceleration response spectrum.

For different seismic waves we can obtain different acceleration response spectrum curves, but this will not facilitate the comparison and statistics of
the dynamic responses between different seismic waves. To this end, we set:
(7.8)
In the equation, |üg(t)|max is the maximum of the absolute value of the corresponding acceleration record. The relationship curve between the
dynamic coefficient and the self-oscillation period T is called the β response spectrum or the standard response spectrum, the ordinate of the standard
response spectrum is defined as dimensionless quantity.
In addition, in 1959, American scholar Housner proposed the concept of average response spectrum in order to use the obtained strong earthquake
records as the basis for seismic design. The average response spectrum is a completely smooth curve obtained by simply averaging multiple reaction
spectrum curves or after a certain statistical averaging. The average response spectrum can be a dimensioned acceleration response spectrum, or it
can be a dimensionless standard response spectrum.
According to the response spectrum curve, for any SDOF elastic system, if we know the self-oscillation period T and the damping ratio ξ, the maximum
acceleration response S a of the system under this specific seismic record can be obtained from the curve. The product of S a and the mass of the
particle is the absolute maximum of the horizontal seismic action:
(7.9)

Based on the known seismic acceleration response spectrum, the standard response spectrum and the average response spectrum, in order to facilitate
the design calculation, China’s Seismic Design Code for Buildings (GB50011-2010) uses single point absolute maximum acceleration relative to the
gravity acceleration, g. The relationship between S a / g and the system’s self-oscillation period T is used as the design response spectrum, and S a / g is
represented by α that is called the earthquake impact coefficient. Therefore, the design response spectrum is also called α response spectrum. In this
way, the equation (7.9) can be written:
(7.10)

The seismic response coefficient α of the building structure should be determined by the seismic intensity, site type, design seismic grouping, structural
self-vibration period, and damping ratio according to Figure 7.4.

Figure 7.4 Earthquake impact coefficient α curve.

α response spectrum curve consists of four parts: In the range of T<0.1s, an upwardly inclined line is used, that is, a linear ascending segment is used;
in the range of 0.1s ≤ T ≤ T g, a horizontal line is used, that is, the maximum value of a is set to η2α max; In the range of T g < T ≤ 5T g, the curve
descending section shown in equation (7.11) is used; in the range of 5T g < T ≤ 6.0s, the straight descending section shown in equation (7.12) is used. It
should be noted, however, that the design response spectrum is beyond its scope in the range of T > 6.0s, and that in this situation, the seismic
response coefficient of the structure should be studied exclusively.
(7.11)

(7.12)

In the equation, amax is the maximum value of the earthquake impact and γ is the attenuation index of the curve descending section which is
determined by (7.13); η1 is the adjustment factor of the downward slope of the straight down section which is determined by (7.14) and when η1 < 0,
there is η1 = 0; η2 is the damping adjustment factor which is determined by (7.15) and when η1 < 0.55, there is η2 = 0.55; T is the structural self-
oscillation period in units of seconds and T g is the characteristic period, which corresponds to the period at the inflection point of the peak of the
response spectrum, it can be determined based on site type, seismic magnitude, and other parameters.
(7.13)

(7.14)

(7.15)

In the equation, ζ is the damping ratio of the structure, generally taking 0.05 for the reinforced concrete structure and 0.02 for the steel structure,
0.10 ~ 0.20 for the damping structure. γ, η1 , η2 are 0.9, 0.02, and 1.0, respectively.

7.1.4 Vibration Mode Decomposition Method

In order to simplify the calculation of the dynamic analysis of the building structure, the structures whose mass is concentrated can be regarded as a
single particle system when we analyze the structural seismic response. However, for structures whose mass is dispersed, in order to truly reflect its
dynamic performance, should be simplified as a multi-particle system when the structural seismic response is analyzed.
The inertial force acted on the particle in the elastic MDOF system is the seismic action of the particle. If torsional coupling is not considered, the
seismic action on particle i is:
(7.16)

If the damping frequency of the particles is assumed to be approximately equal to the undamped natural frequency, noting that the acceleration
response üi(t) of the particle i is numerically equal to the inertia force per unit mass (mi = 1), the following equation can be obtained according to
equation (3.126):
(7.17)

Then let γ j be the j-mode vibration mode participation coefficient, the expression is:
(7.18)

In the equation, Xji is the vibration coordinate of the i-th particle of the j-th mode, and mi is the mass of the i–th particle.
And because:
(7.19)

The above equation proves as follows:

(7.18) is rewritten as a matrix and a vector:

So there is:

With equation (7.19), the right side of equation (7.16) can be rewritten as:
(7.20)

(7.16) can be expressed as:

(7.21)

From equation (7.9) we can see that the maximum of the absolute value of F ji, the horizontal earthquake force acting on the i-th particle of the j-th
mode is:
(7.22)

According to the design response spectrum theory, let , G i = mig so there is:

(7.23)

In the equation, aj is the earthquake influence coefficient corresponding to the j-mode, and can be determined according to (7.11) and (7.12); γ j is the
mode participation coefficient of j-mode vibration and can be calculated according to equation (7.18); Xji is the horizontal relative displacement of the i-
th particle in the j-th mode, that is, the vibration mode displacement; G i is the representative value of the gravity load concentrated on the i–th
particle.
After F ji, the seismic action on the i-th particle in the j-th mode is obtained, the seismic effect S j (bending moment, shear force, axial force and
deformation) of the structure can be calculated by the general mechanics method. According to the vibration decomposition method, the seismic action
of the structure at any time is the sum of the seismic action of each mode and the seismic action F ji corresponding to each mode is the maximum. In
this way, the earthquake effect S i obtained by F ji is also the maximum. However, at any moment the seismic action of other modes (its corresponding
effect) does not necessarily reach the maximum value when the seismic action of a certain mode (its corresponding effect) reaches its maximum value.
The problem of using the maximum seismic effect of each mode when we sum the total seismic action effect of the structure, that is, how to combine
the modes to produce a reasonable seismic effect.
According to the analysis, if the ground motion is assumed to be a stationary stochastic process during the earthquake, the effect of the earthquake
caused by each stationary mode can be approximated by the method of SRSS:
(7.24)

In the equation, S is the horizontal seismic effect, and S j is the effect of the horizontal seismic action of the j-th mode including internal force and
deformation.
It is important to note that the seismic effect of each mode, which is calculated by the SRSS, is different from the seismic effect of each mode, which is
combined by the SRSS and then calculated. Because it is the high-order mode, the seismic actions can be either positive or negative, but after they are
squared they all become positive. Therefore, the seismic effect of the structure will be exaggerated by the latter method.
In general, the contribution of each mode to the total earthquake response will decrease rapidly as its frequency increases. So several modes with the
lowest frequency often control the structure of the largest earthquake response. In the actual calculation, we generally use the first three vibration
modes. But taking into account the various natural frequencies of the long-period structure are relatively close, so the Seismic Design Code for
Buildings stipulates: when the basic self-vibration period is greater than 1.5s or the house aspect ratio is greater than 5, it is appropriate to increase the
number of modes.

7.1.5 Bottom Shearing Force Method

When we solve the structural seismic response of an MDOF system by vibration mode decomposition method, we need to calculate the natural
frequency and vibration mode of the structure in various orders, which is a complex operation. In order to simplify the calculation, we can use the
bottom shearing force method when the height of the structure does not exceed 40m, the deformation is mainly shear deformation, and the mass and
stiffness are evenly distributed along the height or the structure is similar to a single particle system. The bottom shearing force method is an
important knowledge for civil engineering undergraduates and is the core of graduation design and seismic calculation. This method applied is to
calculate the total horizontal seismic action on the structure, that is, the shear acting on the bottom of the structure. Then we reallocate the total
horizontal seismic action to each particle according to certain rules.
From equation (7.23) we can see that the horizontal seismic action of the i-th particle of the j-th mode is F ji = α jγ jXjiG i. The bottom shearing force F Ej
of the structure in this mode should be the sum of F ji for each particle:
(7.25)

Transforming the above equation appropriately, it can be rewritten as:

(7.26)

From equation (7.24), the bottom shearing force F Ek of the structure should be:
(7.27)

Let:
(7.28)

(7.29)

Then equation (7.27) becomes:

(7.30)

In the equation, F Ek is the standard value of the total horizontal seismic action (bottom shearing force) of the structure; α 1 is the seismic influence
coefficient corresponding to the basic period T 1 ; G eq is the equivalent weight of the structural load, which can be calculated by according

to the Seismic Design Code for Buildings.

If the total horizontal seismic action F Ek of the structure is known, we can obtain the seismic action F i of each particle according to the following
method. For a structure whose deformation is mainly shear deformation and the mass and stiffness are evenly distributed along the height distribution
[Figure 7.5(a)], its seismic response is usually the first type of vibration-based and the shape is close to the inverted triangle [Figure 7.5(b)]. For this
reason, the first mode is assumed to be an inverted triangle as shown in Figure 7.5(c). The displacement amplitude of the particle i is X1i = ηΗi, η is the
proportional constant.
Figure 7.5 Bottom shearing force method illustration.
From equation (7.23), the horizontal seismic action of particle i under the basic mode (the first mode) is:
(7.31)

The bottom shearing force is:

(7.32)

By the above equation:

(7.33)

Substituting η into equation (7.31), the horizontal seismic action of particle i is:
(7.34)

The above equation is applicable for structures with a shorter basic period. When the basic period is relatively long, due to the influence of high
vibration mode, the top layer seismic shearing force is small and, as a consequence, should be adjusted. The method of adjustment is to consider part of
the total seismic action of the structure acting as a concentrated force that acts on the top of the structure and then assign the remaining part to the
various points by the inverted triangle. According to statistics of the analysis of the results, this additional concentrated horizontal seismic action ΔF n
can be expressed as:
(7.35)

In the equation, δn is the top additional seismic action coefficient which can be determined according to the Seismic Design Code for Buildings. So
equation (7.34) can be rewritten as:
(7.36)

Example 7.1 As shown in Figure 7.6, there is a two-story cast-in-place reinforced concrete frame structure. Fortification intensity is 8 degrees,
designed basic seismic acceleration is 0.20g, Class II site, designed seismic grouping is the first group (T g = 0.35s); Damping ratio is ξ = 0.05. Floor
representative gravity load is G 1 = G 2 = 2000kN; Interlayer lateral stiffness is k 1 = 2.8 × 104 kN/m, k 2 = 3.5 × 104 kN/m. Structural vibration period
is T 1 = 0.89s, T 1 = 0.89s.
Figure 7.6 Two-story reinforced concrete frame structure.
The vibration modes of the structure are respectively: {X12, X11 }T = {1.495,1.00}T, {X22, X21 }T = {–0.669,1.00}T. Use the vibration mode
decomposition method and the bottom shearing force method to calculate the interlayer seismic shearing force of each layer.
Solution:
(1) Use vibration mode decomposition method
From F 1i = α 1 γ 1 X1iG i, the following is available:

Similarly, from F 2i = α 2γ 2X2iG i, the following is available:

So the interlayer seismic shearing force is:

The interlayer seismic shearing force is shown in Figure 7.7.

Figure 7.7 Interlayer shearing force illustration.

(2) Use bottom shearing force method
It can be seen that the results of the two methods are similar. The reason is that the difference of the first two frequencies is quite large (T 1 = 0.89s, T 2
= 0.32s) and we only combine the first two modes.

7.2 Structural Vibration Control

7.2.1 Concept and Classification
Structural vibration control is a new type of anti-vibration method. It is different from the traditional way of relying on the structure’s own strength,
stiffness, and ductility to resist wind-induced (earthquakes). Structural vibration control is to install a variety of control devices in the structure to bear
the wind (earthquake) effect together with structure, so as to reduce the structural wind (earthquake) response and to ensure the safety of the
structure. There has been more than 30 years of research and application of structural vibration control in civil engineering. Structural vibration
control can be divided into passive control, active control, semi-active control, and hybrid control.

1. Passive Control

Passive control is control without external energy. Its control force is generated passively by the deformation of the control device when vibrating
along with structure together. Passive control devices are mainly viscoelastic dampers, viscous dampers, metal dampers, friction dampers, FM mass
dampers, FM fluid dampers, and base isolation devices. When the structure deformation occurs due to wind or earthquake, the passive control device
also produces a corresponding deformation to dissipate the vibration energy of the structure. Passive control devices are currently the most widely
used form of structural control. Professor Zhao-Dong Xu of Southeast University has applied the viscoelastic dampers to the seismic response control of
Xi’an Fujin Jiayuan, Xi’an Petroleum Hotel, Pakistan Vision High School and many other vibration damping (earthquake) projects.

2. Active Control

Active control is control with external energy. Its control force is applied actively by the control device according to a certain rule using external
energy. The active control system consists of three parts: sensor, operator, and actuator. The sensor transmits the measured ground motion, the
structural response, or both to the operator. The operator then processes the information and calculates the required control force according to the
given control algorithm. The required control force passes through the loop and becomes the control signal and then the control signal is transmitted to
the actuator. With external energy to generate control over the structure to reduce the dynamic response, active control has an active cable system,
active support system, active quality drive system, active wind shield, pulse generator, etc. The United States, Japan, and China have conducted much
research on active control. Since the late 1980s, China has begun to study active control. At present, the application of the active control in civil
engineering has reached more than 30 items, such as Japan’s Takenaka Experimental Building and Kankyu Chayamechi Building.

3. Semi-Active Control

Semi-active control is control with a small amount of external energy. Although its control force is generated by the movement of the control device
itself, in the control process, the control device can use the external energy to actively adjust its own parameters, so as to adjust the control. The
existing semi-active control techniques include: magnetic (electrical) rheological dampers, semi-active variable stiffness devices, semi-active variable
damping devices, semi-active frequency modulation dampers, semi-active frequency modulation liquid dampers, etc. Professor Zhao-Dong Xu of
Southeast University has applied the magnetorheological dampers to the wind response control of Hanzhong Xi’erhuan Bridge.

4. Hybrid Control

Hybrid control is the simultaneous application of passive control and active control over the structure, or the application of more than one passive
control device at the same time, so as to give full play to each control form and control device, overcome their weaknesses, and obtain better control
effect. At present, the hybrid control methods are as follows: hybrid control system combining base isolation and energy dissipation, hybrid control
system combining semi-active control and base isolation, and hybrid system combining multi-dimensional isolation and damping. Professor Zhao-Dong
Xu of Southeast University has successfully developed a multi-dimensional seismic isolation devices for large span reticulated structure, building
structure, high-speed train bridge or large size pipeline. The world’s first installation of a hybrid control system is located in Tokyo, at Japan’s water
technology research institute. At the same time, Professor Zhao-Dong Xu got many applications of hybrid control, such as Yan-An petrochemical
pipelines, Luo-Yang petrochemical pipelines.
Among these four kinds of control technology, active control is the best. However, the large size of building structures has caused the need for larger
external energy. The required control algorithm of the control device is more complex and there is a delay problem. Application of active control is less
than the other three kinds of control technology. Passive control has low costs, a good damping effect, and is easy to achieve. As a consequence, its
growth is the fastest and is used the most widely. In the latest Seismic Design Code for Buildings, it has been partially mentioned. The rubber base
isolation technology has especially matured and has been promoted and applied to a certain degree. But the universality of passive control for
earthquake excitation is poor. It means that the damping effect of the control structure which is designed according to a certain earthquake excitation
is not necessarily good under other earthquake excitations. Semi-active control is between active control and passive control. Its control accuracy is
higher and the cost is lower than the active control, it does not need a larger power source and has a broad application and development prospects.
Hybrid control combines the advantages of several control methods with good control effect and broader development prospect. This section will only
introduce the representative three devices which are applied and researched most widely: the damping of the viscoelastic dampers which represents
the energy dissipation technique, the rubber isolation which represents the base isolation technology, and the MR damper which represents the semi-
active control and active control techniques.

7.2.2 Vibration Reduction Technology of Viscoelastic Dampers

The viscoelastic dampers dissipate the vibrational energy of the structure by hysteretic energy dissipation of the viscoelastic material in order to
reduce the structural dynamic response. It was first used in the damping of military, aerospace, and mechanical engineering. Only in the nearly ten
years this technology was adopted in structural seismic response control.

1. Viscoelastic Damper Performance

Viscoelastic dampers are composed of viscoelastic materials and constrained steel plates. According to the deformation of the viscoelastic layer,
viscoelastic dampers can be divided into tension and compression dampers, and shear dampers. Figure 7.8(a) shows the tension and compression
dampers which are applied to the beam-column joints. The shear dampers can be divided into square-shaped dampers [Figure 7.8(b)], cylindrical
dampers [Figure 7.8(c)], and conventional dampers [Figure 7.8(d)].
Figure 7.8 Viscoelastic dampers illustration.
The performance of viscoelastic dampers is often expressed by energy storage stiffness Kd1 , loss factor η and energy consumption per lap Ed.
(7.37)

(7.38)

(7.39)

In the above equation, Av is the area of the viscoelastic layer, nv is the number of viscoelastic layers, and hv is the thickness of the viscoelastic layer. In
the sinusoidal excitation, the energy consumption per lap, Ed, is:
(7.40)

Because the main factors that affect the performance of viscoelastic materials are temperature, frequency, and strain amplitude, the main factors that
affect the performance of viscoelastic dampers are also temperature, frequency, and strain amplitude.
The impacts of the temperature and frequency are more significant than strain amplitude. If the parameters of the viscoelastic material G 1 , G 2 and η
are determined, the energy storage stiffness Kd1 and loss stiffness Kd2 of the viscoelastic dampers can be solved by equations (7.37) and (7.38).
In order to describe the stress-strain relationship of viscoelastic dampers accurately, many researchers proposed a variety of computational models,
including the Kelvin model, the Maxwell model, the standard linear solid model, the complex stiffness model, the four-parameter model, the finite
element model, and the equivalent standard solid model.

2. Analysis Method of Structure Installed with Viscoelastic Dampers

The dynamic characteristics and dynamic responses of the structures installed with viscoelastic dampers are mainly expressed as follows: Dynamic
characteristics: the stiffness and damping of the structure increase, the natural frequency increases and the period decreases. Dynamic responses:
Viscoelastic dampers provide a significant damping for the structure to make the structure appear elastic and the seismic response is greatly reduced.
Displacement, acceleration, interlayer displacement, and interlayer shear are significantly reduced. The seismic which is energy input to structure is
mostly absorbed by the hysteretic energy dissipation of the viscoelastic dampers. Structural hysteresis, which easily leads to damage of the structure,
only accounts for a small part. Therefore, the structure installed with a viscoelastic damper produces less plastic hinges and cracks under the same
earthquake. Tests showed that the damage that occurs is less than half of that which occurs in the original structure. Whether it is steel or reinforced
concrete structure, the installation of viscoelastic damper has shown the above characteristics.
The stiffness and damping are significantly increased after the structure is installed with the viscoelastic damper; the purpose of vibration reduction
can be achieved by using viscoelastic dampers to change the dynamic characteristics of the structure. The calculation of the dynamic characteristics of
the viscoelastic damping structure mainly solves the stiffness and damping of the structure after adding the viscoelastic damper. Currently, the
following methods are used for this calculation: Modal Strain Energy Method Incremental Stiffness and Incremental Damping Method
Improved Modal Strain Energy Method. After solving the structural stiffness and damping with these methods, the dynamic response of the structure
can be solved by the time-history analysis method, the vibration mode method, the transfer function method, the modern control method, and the
random vibration method described in the previous section. The structure can be further designed according to the dynamic response of the structure.
3. Engineering Application and Experimental Research

The application of viscoelastic dampers in mechanical engineering and aeronautical engineering has a long history. The early stages of application in civil
engineering concentrated on the structure of the wind damping engineering. In recent years it has begun to be used in structural seismic vibration
absorption engineering. Table 7.1 lists some engineering application profiles.
Table 7. 1 Application of viscoelastic dampers in wind-resistant earthquake engineering.
Building name Location Height Application Placement
World Trade Center Building New York 110 layers Reduce wind vibration Truss chord
Columbia Center Building Seattle 77 layers Reduce wind vibration Main oblique strut joints
TV tower United States 100 Meters Reduce wind vibration Oblique cable
Atomic reactivity building France Reduce wind or earthquakes
Dry bridge cableway Shanghai Reduce wind vibration Oblique cable
Changsha, Hunan Bridge Changsha Reduce wind vibration Oblique cable
Chongqing Yangtze River Bridge Chongqing Reduce wind vibration Oblique cable
Pittsburgh Steel Building United States 64 layers Reduce wind vibration
A frame structure Japan 11 layers Reduce earthquakes
A frame structure United States Reduce earthquakes Cross support
A steel frame United States 13 layers Reduce wind or earthquakes
ChenTaiRailway Station Taiwan Reduce wind vibration roof
Suqian City Traffic Building Jiangsu Province 13 layers Reduce earthquakes Braces
Huaiyin City Government Park Building Jiangsu Province 14 layers Reduce earthquakes Braces
Nanjing Traffic Building Jiangsu Province 26 layers Reduce earthquakes Braces
Suyu County Family Planning Commission Building Jiangsu Province 7 layers Reduce earthquakes Braces
Xi’an Petroleum Hotel Xi’an 7 layers Reduce earthquakes Braces
Xi’an Fujinjiayuan Xi’an 105 Meters Reduce earthquakes and wind vibration Braces
Seattle Columbia Center Building, 77 floors, 291 meters high, a steel-concrete building whose plane was a triangular. In order to reduce the effect of
wind vibration, 260 large viscoelastic dampers, each weighing 272 kilograms, were installed on the inclined braces where the moving parts and the
stress parts were larger. The project first analyzed the internal forces of the bars by computer to determine the most efficient placement of the
viscoelastic dampers. As a result, 1/6 of the diagonal braces were provided with a viscoelastic damper. Xi’an Petroleum Hotel is a 7-story reinforced
concrete frame structure, seismic fortification intensity of 8 degrees. The original design failed to meet the seismic requirements. If the traditional
transformation method that increases beam and column sections and reinforcement is used, it is difficult to meet the design requirements and will also
lead to higher costs. To solve this problem, Professor Zhao-Dong Xu of Southeast University used 56 viscoelastic dampers in the structure. The peak
value of structural displacement and acceleration under the 200 gal artificial seismic wave were reduced by 54% and 26%, respectively, which meets
the engineering requirements.
The wind tunnel test, the seismic simulation shaking table test, and the large number of structural analyses of the viscoelastic damper structure shows
that the installation of a viscoelastic damper in the structure can reduce the wind vibration response and seismic response by 40% to 80% and ensure
the safety of the main structure in strong winds and earthquakes. The comfort of the structure is controlled within the specified range when the
structure is under strong winds. The top layers of Seattle Columbia Center Building originally had obvious discomfort issues due to the impact of wind-
induced. After the installation of a viscoelastic damper, there is no longer an issue with feelings of discomfort. If the same effect was to be achieved by
increasing the stiffness, the existing column size would have needed to be doubled, costing roughly an additional $8 million. Obviously, the use of
increased stiffness is difficult to advocate due to costs. On the other hand, testing and installation of viscoelastic dampers only cost 700,000 US dollars
for this project. By adding viscoelastic dampers to solve the problem of high-precision transformation, Xi’an Petroleum Hotel. was able to save a total of
230,000 yuan. It can be seen that the use of viscoelastic dampers can effectively reduce the wind or vibration of the building while remaining
economical.
A large number of performance tests have been carried out on viscoelastic dampers at home and abroad in order to further research and development.
At the same time, in order to explore the dynamic characteristics of the viscoelastic damper structures and to verify the damping (shock) performance
of the viscoelastic dampers, researchers have conducted a shake table test on the structures using viscoelastic dampers. The main experimental
studies are shown in Table 7.2.
Table 7.2 Experimental study on the structure with viscoelastic dampers.
The experimenter Age Experimental model framework
Lee, etc. 1990 Five-story full-scale steel frame and a2/5scale model frame
Lin and Liang, etc. 1991 1/4three-story steel structure model framework
Bracciand Shen, etc. 1992 1/3three-story reinforced concrete model frame
Chang and Lai, etc. 1993 2/5five-story steel structure model frame
Chang and Shen, etc. 1995 2/5three-story steel structure model framework
Lai, etc. 1995 1/3three-story steel structure model frame and five-story full-scale steel frame
Chang, etc. 1996 2/5three-story steel structure model framework
Oujin Ping, etc. 1999 1/816-story steel structure model frame
Xu Zhao-Dong, etc. 2000 1/5three-story reinforced concrete model frame
The following conclusions can be drawn from a large number of experimental studies and engineering applications:
1) Viscoelastic dampers are simple in construction, easy to manufacture, and low in cost.
2) Viscoelastic dampers are reliable, with strong energy dissipation and no severe aging.
3) Viscoelastic dampers can effectively reduce the wind vibration and seismic response of buildings.
4) The use of viscoelastic dampers to reduce wind vibration and seismic response saves a certain amount of money and reduces the cost of
construction.
5) Viscoelastic dampers have a wide applicability. They can be used for structures requiring wind and vibration damping, as well as providing
structures with earthquake damping. They can be used in building structures, in tow mast structures, and bridge structures. They can be used for
new construction, seismic reinforcement, and post-earthquake repair.
6) Viscoelastic damping structures have a good overall efficiency and social significance. They can avoid secondary disasters and economic losses
caused by the earthquake and strong winds to ensure the safety of people’s lives and property.
Practical engineering applications show that viscoelastic dampers have many advantages and can solve practical problems and achieve their purpose,
while remaining economical. So, viscoelastic dampers have broad application and development prospects.

7.2.3 Rubber Base Isolation Technology

Base isolation means the provision of an isolation between the building foundation and the superstructure. The superstructure can be isolated from the
foundation which can limit the energy of the ground movement to the upper part of the building, reducing the seismic response in the upper part of the
building in order to ensure the safety of personnel and equipment in buildings. The core part of the base isolation is the isolation bearing. More
commonly used isolation bearings are rubber isolation bearing, sliding friction isolation bearing, friction roll isolation bearing, etc. Rubber isolation
bearings were developed and are the most widely used.

1. Performance of Rubber Isolation Bearing

Rubber isolation bearings are composed of multi-layer rubber and laminated steel plates alternately laminated by high temperature vulcanization, as
shown in Figure 7.9. When the bearing is pressed, the steel plate constrains the lateral deformation of the rubber sheet so that it has a very small
deformation and a large vertical stiffness. When the bearing is subjected to horizontal loads, the steel plate in the horizontal direction can’t constrain
rubber sheet, shear deformation. The shear deformation is the shear deformation of the rubber itself. It maintains the inherent flexibility of the
rubber, showing a small horizontal stiffness.

Figure 7.9 Rubber isolation bearing illustration.

In order to meet the damping requirements of an isolation structure system we can add lead to the rubber bearing or use high damping rubber
material. Common rubber isolation bearings are common rubber isolation bearing, lead rubber isolation bearing and high damping rubber isolation
bearing. The rubber isolation bearing must have sufficient vertical compression stiffness, horizontal stiffness, and a certain tensile stiffness.
Restoring force models of rubber isolation bearing include vertical models and horizontal models. While isolation bearings consume energy horizontally
the vertical must be able to effectively support the upper structure. So the vertical stiffness of the bearings are very large. We use a linear elastic
model between the vertical force and the deformation, that is, the vertical bearing is equivalent to a linear spring. Its corresponding equation is:
(7.41)

In the equation F V is the vertical load of the bearing, KV is the bearing vertical stiffness, xV is the bearing vertical deformation.
Because the rubber isolation bearing has good elastic performance but also has certain viscous damping characteristics, the horizontal restoring force
model of the rubber isolation bearing can be a combination of the elastic restoring force model and the viscous damping force model, as shown in Figure
7.10.

Figure 7.10 Rubber isolation bearing illustration model.

(7.42)

In the formula, F h is the elastic restoring force of the rubber isolation bearing; Kh is the horizontal rigidity of the rubber isolation bearing; x and ẋ are
the horizontal displacement and the velocity of the rubber isolation bearing respectively.
Due to the material properties of ordinary rubber isolation bearings, the energy consumption capacity is small. The restoring force and displacement
curve are approximated as straight lines, as shown in Figure 7.11(a). Due to the increase in damping ratio of the lead rubber isolation bearings or high
damping rubber bearings, they have a certain energy consumption.
Figure 7.11 Resilience-displacement curve of rubber isolation bearing.
Currently, the nonlinear analysis model that is the most commonly used of lead-core rubber isolation bearing is the bilinear analysis model. Readers
with an interest can refer to the relevant books.

2. Design and Analysis of Seismic Isolation Structure with Rubber Bearing

The seismic isolation analysis method includes the isolated structure calculation model and the corresponding motion equation. When we analyze the
dynamic response of the isolated structure, we must first establish a computational model that reflects the actual deformation properties of the isolated
structure and makes the calculation simple and feasible. The general calculation models available are single particle model, multi-particle model,
torsional vibration model, and spatial model, as shown in Figure 7.12.

Figure 7.12 Several calculation models of isolated structures.

(1) Single Particle Calculation Model
The stiffness of the superstructure of the general base isolation house is much greater than that of the base isolation layer. The upper structure is
designed according to elasticity. The input energy of the earthquake is mainly absorbed by the isolation layer. Since the vibration of the superstructure
has little effect on the vibration of the isolation layer, the upper structure can be regarded as a rigid body and the isolation layer can be simplified as an
equivalent spring and a damper. Then, according to that, we can calculate the dynamic response of the isolated structure system. The calculation
diagram is shown in Figure 7.12(a) below.
The motion equation of the single particle calculation model is:
(7.43)

In the equation m is the total mass of the structure, c and Kh are the damping coefficient and horizontal stiffness of isolation layer. And:
(7.44)

(2) Multi-Particle Calculation Model

When a detailed seismic response of the superstructure is required, a multi-particle model or a spatial analysis model can be used. These models can be
considered as the conventional structural analysis model which bottom is added the isolated layer simplified model. It can be analyzed by time-history
analysis method or response spectrum method to seismic response of isolated structure.
When the isolation structure is simplified as an MDOF interlayer shear model, the base isolation layer is the first layer and the interlaminar shear
stiffness is taken as the total horizontal stiffness and the mass of the layer is taken as the sum of the total mass of the isolation bearing and the weight
of the bottom half. Using this model to calculate the dynamic response, we can calculate the deformation and interlayer shear state of each stage of the
structure accurately. The calculation diagram is shown in Figure 7.12(b). The establishment of the motion equation of the isolated structure is similar to
the equation of motion of the multi-degree of freedom system in Chapter 4. It is not described here in detail.

(3) Torsional Vibration Calculation Model

If structural translation and twisting need to be considered at the same time, a torsional vibration calculation model can be used. The model assumes
that the floors are rigid in the plane and the motion of the plates consist of two horizontal lines of displacement and angular displacement. In this model,
the isolation layer is simplified as a damping spring unit. The calculation is shown in Figure 7.12(c).

(4) Spatial Computing Model

The spatial model assumes the structure to be rigid in the plane of each floor and takes two horizontal motion components, an in-plane rotation and a
vertical component, into account. The model can be used to calculate the structure more accurately. For the general structure, the stiffness of each
floor in the plate is very large and the impact between the vertical deformation of the floor of the structure is smaller. The isolated structure is
arranged by the isolation bearing to reduce the torsional response of the structure. In order to simplify the calculation, the layer shear model is
generally used and can analyze the dynamic response of the basic isolation house.

(5) Design Method

The Seismic Design Code for Buildings uses the two concepts of segment design and horizontal damping coefficient to link seismic and isolation design.
The resources and technology of seismic design can continue to be applied in isolation design or can be used by a slight change. Partition design method
of isolation structure divides the seismic isolation system into four parts: the superstructure, the isolation layer, the lower structure, and the
foundation. For details, the reader can refer to the relevant books and codes.
The engineering applications and experimental research of rubber isolation structure are very common and this technology is compared proverbially to
others. Thus, they are not listed here.

7.2.4 Vibration Reduction Technology of Magneto-Rheological Damper

The magneto-rheological damper is a new kind of vibration control device which is made by using fast rheological properties of magnetorheaters under
strong magnetic field. In the 1940s Rabinow first discovered that the magnetorheaters had rheological properties. Subsequently, the
magnetorheological technology has been rapidly developed and research surrounding the magneto-rheological damping device has shown encouraging
results, suggesting a bright future for people.
The magneto-rheological damper is installed in the support between columns of the building structure, in the beam and column joints, the truss under
the chord, and other deformation parts. Depending on the actual deformation of the structure, external seismic (wind-induced) excitation, and the
specific control law, the external magnetic field should be controlled to change the mechanical properties of the magnetorheological damper. This
changes the characteristic parameters of the structural system in an instant so that the structure achieves the best control effect. The use of a
magneto-rheological damper for intelligent vibration damping (vibration) control research in foreign countries began in the early 1990s and has
entered our country just in the last few years. The research is mainly divided into the study of the performance of magneto-rheological damper and the
study of the structure with magneto-rheological damper.

1. Performance of Magneto-rheological Damper

Magneto-rheological dampers are composed of magnetic fluid, cylinder, piston, and controllable magnetic field. According to its force mode, the
magneto-rheological damper can be divided into pressure-driven, shear, and extrusion types. Its working principle is shown in Figure 7.13. The
pressure-driven type is commonly used in servo control valves, damper, and shock absorbers. The shear type can be used for clutches, brakes, locking
devices, dampers, and other devices. Extrusion can be used to make magneto-rheological dampers with small displacement and large damping.
Application of semi-active control dampers of the structure commonly use the pressure-driven type. Figure 7.14 shows a commonly used magneto-
rheological damper.

Figure 7.13 Three modes of magneto-rheological damper.

Figure 7.14 Magneto-rheological damper commonly used in building structure shock absorption.
The damping force of the magneto-rheological damper is composed of two parts: the viscous damping force caused by the zero magnetic field viscosity
of the uncontrolled magneto-rheological fluid and the Coulomb damping force which can be controlled by the external magnetic field. Its performance is
affected by the displacement amplitude, magnetic field strength, excitation frequency, and other factors. The law of influence is:
1) At a fixed current, the peak force increases as the amplitude of the displacement increases.
2) The larger the magnetic field strength, the greater the viscosity of the magnetorheater, the greater the yield force, the greater the energy
consumption of the magneto-rheological damper. When the magnetic field strength reaches the saturation magnetic field strength, the yield force
and the energy consumption performance of the magneto-rheological damper are almost unchanged.
3) The excitation frequency has little effect on the force-displacement curve of the magneto-rheological damper. But the frequency affects the
nonlinear relationship between force and velocity. The higher the frequency, the greater the nonlinearity.
In order to truly describe the mechanical properties of the magnetorheological damper, the researchers proposed a number of models. These models
can be divided into two main types: nonparametric models and parametric models. There are mainly Bingham model, Bouc-Wen model, equivalent
viscous damping model, nonlinear double damping model, nonlinear hysteresis double damping model, Sigmoid model, and phenomenological model.

2. The Control Algorithm of Magneto-rheological Damper Structure

The control algorithm is a core problem in semi-active control and active control which directly affects the control effect. The control algorithm of the
structure with magneto-rheological damper is mainly composed of two-state optimal control, control strategy based on Lyapunov stability theory,
decentralized bar-bar control, limiting optimization control, tri-state control and fuzzy neural network control. The following only describes the two-
state optimal control, tri-state control, and fuzzy neural network control.

(1) Two-State Optimal Control

Using the instantaneous optimal control algorithm, the instantaneous quadratic objective function can be taken as:
(7.45)

In the equation, Q and R are the weight matrix of the state vector Y(t) and the control force vector U(t), respectively. They are used to adjust the
relative importance between the structural response and the control force. The optimal control theory can be used to obtain the optimal control force to
minimize the quadratic objective function.
(7.46)

In the equation Δt is the sampling period and B is the state matrix of the system. Since the magneto-rheological damper is to adjust the magnetic field
strength to adjust the damping force generated by the damper, it is impossible at any instant to achieve the optimal control expressed in equation (1-
14). But only by adjusting the magneto-rheological damper parameters so that it produces the damping force closing to the optimal control force. The
semi-active control strategy can be obtained by adjusting the parameters of the magneto-rheological damper:
(7.47)

In the equation, F min and F max are the damping forces generated by the magnetic damper when the magnetic field is zero and the magnetic field is the
largest. ui and are the displacement and velocity generated by the i-th damper. F i is the control force produced by the i-th damper.

(2) Tri-State Control

According to the actual failure characteristics of reinforced concrete structures in the earthquake process, Professor Zhao-Dong Xu of Southeast
University proposed a three-state control method in 2002. Reinforced concrete structures generally go through elastic stages, post-cracking stages,
and post-yield strengthening stages during earthquakes. When the structure is in the elastic stage, the structure does not need, or require only a small,
control force. When the structure is cracked, the structure requires a damper to provide a medium-sized control force to compensate for the structural
loss of stiffness due to cracking. When the structure continues to lose, or even enter the post-yield strengthening stage after cracking, the structure
requires the damper to provide greater control against external seismic forces. According to this theory, the tri-state control method can be written as
follows:
(7.48)

In the equation, k, k 1 and k 2 are the actual stiffness of the layer of the structure, the initial stiffness and cracking stiffness. α is the current regulation
factor, 0 < α < 1. The tri-state control method embodies the seismic requirements of the actual structure. The magneto-rheological damper provides
minimal control when the structure does not require control and requires only a small amount of control. The magneto-rheological damper provides
medium-sized control when the structure requires a certain degree of control due to cracking to compensate for the loss of stiffness. The magneto-
rheological damper provides maximum control when the structure is further destroyed and needs greater control. Therefore the tri-state control
method avoids the overshoot of the magneto-rheological damper parameters and damping force at the initial, final and small earthquakes. It does not
cause the dynamic response of the controlled structure to be partially amplified. The three-state control method is simple in concept, clear, and only
sets up three current files which is easy to realize in the actual control process.

(3) Fuzzy Neural Network Control

The fuzzy neural network control method was proposed by Professor Zhao-Dong Xu of Southeast University in 2002. In this control method, the
neural network technology is used to predict the dynamic response of the structure at the next moment to solve the delay problem of semi-active
control, and at the same time the fuzzy logic controller is used to implement the current selection of the magneto-rheological damper to solve the
instantaneous selection problem of current. Input the excitation acceleration and the state of the building structure (displacement, velocity, and
acceleration) predicted at each time to the fuzzy logic controller under the fuzzy rule and the membership function under the fuzzy logic controller and
get a control current. The control current command signal is passed to the magneto-rheological damper so that the magneto-rheological damper
generates control force and applies it to the structure.
The use of a magneto-rheological damper for the damping control of the building structure is semi-active control. It not only has the characteristics of
high control accuracy but also has the characteristics of low energy consumption. Therefore, more and more experts and scholars at home and abroad
have joined the research and have achieved results. However, many problems still remain that must be solved by the researchers.

7.3 Modal Analysis and Theory

The modal analysis transforms the physical coordinates in the motion equations of the linear dynamical system into modal coordinates and decouples
the equations into a set of independent equations describing modal and modal parameters in order to obtain the modal parameters of the system. It is
the basis of dynamic design, vibration isolation, vibration control, health monitoring, and fault diagnosis. The actual vibration of buildings, machines,
aerospace vehicles, ships, cars, and other complex structures are different and modal analysis provides an important way to study the actual structural
vibrations.

7.3.1 Modal Parameters

In the case of the N-degree-of-freedom linear stationary system shown in Figure 7.15, the differential equation of motion is:
(7.49)

Figure 7.15 MDOF system.

In the equation, the meanings of the symbols are the same as those in Chapter 4. The above equation is the equations of motion described by the
physical coordinates {ü}, , and {u} of the system. Each equation contains the physical coordinates of the system particles so it is a set of coupling
equations. When the degree of freedom of the system is very large the solution is very difficult. The fundamental task which the modal analysis needs
to solve is determining if the coupling equation can be transformed into an uncoupled and independent differential equations.
As described in Chapter 3, Equation (7.49) can be obtained by Fourier transform:
(7.50)

From the theory of vibration we can see that for a linear time-invariant system, any response of the system can be expressed as a linear combination
of modal responses. The response of l can be expressed as:
(7.51)

In the formula, φ lr is the modal coefficient of the l-th measure point and the r-th mode. The column vector composed of the modal coefficients of N
points is:
(7.52)

Equation (7.52) is called the r-th order modal vector. It reflects the vibration shape of the order mode and is also known as the vibration mode. The
matrix composed of the modal vectors is called the modal matrix or the vibration mode matrix:
(7.53)

qr(ω) in equation (7.51) is the r-th order modal coordinates. Its physical meaning can be understood as the contribution of each mode to the response
of the modal and its mathematical meaning can be understood as the weighting factor. The contribution, or coefficient, of the modes to the response are
not the same and are related to the frequency of the excitation and structure. Usually the lower-order mode has a larger weight coefficient than the
higher-order mode.
The response vector of the system can be obtained from (7.51) and (7.53):
(7.54)

In the equation, {Q} = [q1 (ω) q2(ω) ···qN(ω)]T. Chapter 4 has already discussed the orthogonality of the mass matrix, the stiffness matrix, and the
Rayleigh damping matrix which satisfies the decoupling condition:
(7.55)

(7.56)
(7.57)

Mr in equation (7.55) is the r-th order modal mass, Kr in equation (7.56) is the r-th order modal stiffness, and Cr in equation (7.57) is the r-th modal
damping. They are all numbers, not matrices. Using the modal coordinates instead of the physical coordinates in (7.49), multiplying by [Φ]T on the left
and taking into account the above orthogonal conditions, we can get:
(7.58)

In the equation, is modal load. Modal vector, modal mass, modal stiffness, and modal damping are collectively referred to as modal
parameters.

7.3.2 Real Modal Analysis

Modal can be divided into real and complex modes, depending on if the modal parameters are real or complex. From equation (7.58) the r-th modal
coordinates can be obtained:
(7.59)

In the case of a single point of motivation, if the excitation force is applied to the p point so the excitation force vector becomes:
(7.60)

Modal force Pr = φprPp(ω), equation (7.59) can be written as:

(7.61)

Substitute the above equation into equation (7.51):

(7.62)

Therefore, the frequency response function between the measurement point l and the excitation point p is:
(7.63)

The above equation indicates that the response of the point l is caused by the excitation force of the point p alone. Meaning, this is the complex
response generated at point l after the application of a unit of sinusoidal force at point p. It can be seen that the frequency response function is
independent of the magnitude of the excitation force.
In fact, the response and the frequency response function of the system are all the superposition of modal. That is the complete modal set. But in fact,
not all modal contributions to the response are the same. For low frequency response the effect of high-order mode is small. For the actual structure,
we are often interested in its first or tenth order mode and high-order mode is often abandoned. Although this will cause some errors, the frequency
response function matrix dimension will be greatly reduced so that the calculation workload is greatly reduced. Practice has proved that this is entirely
desirable and this is also a major advantage of modal analysis. This approach is called modal interception. Therefore, in the modal analysis we often use
a non-complete modal set.

7.3.3 Complex Modal Analysis

In the real modal analysis, it is considered that the vibration phase difference of each particle of the power system is 0° or 180° and the modal
coefficient φ is real. The actual structure, due to the existence of damping, lead to the different amplitude and the different phase, which indicates that
the phase difference is not necessarily 0° or 180°. So that the modal coefficient becomes a complex number. That is the complex mode.
As described in Chapter 6, the motion equations of the MDOF system in equation (7.49) can be written in the form of a state equation:
(7.64)

In the equation, the state vector is:

An MDOF system is used as an example to illustrate the complex mode analysis. The state equation is:
(7.65)

Assume that the solution of equation (7.65) is:

(7.66)

In the equation {Ψ} is the complex modal vector and substitute the above equation into equation (7.64):
(7.67)

Solving equation (7.67) yields 2N complex eigenvalues and eigenvectors which are written separately as: Zhao-Dong Xu

For the r-th order mode there is let , , equation (7.67) can be written as:

(7.68)

For the r-th order mode there is . According to the vibration theory, the complex modal also has the orthogonality of the matrix of
coefficients [A] and [B]. That is:
(7.69)

(7.70)

For all modes, the modal matrix is:

The characteristic matrix is:

λ r is the r-th complex modal frequency of the dynamical system and is the r-th complex modal vector of the dynamical system. The state vector is
transformed with complex modal coordinates:
(7.71)

In the above equation, {Q} and {Q}* are complex modal coordinates. The reaction of the dynamical system can be obtained by equation (7.71).
From the above solution we can see that complex mode analysis has the following characteristics:

1. Complex Conjugate Characteristics

In the complex modal analysis, the eigenvalues and eigenvectors of the system are complex numbers and conjugate pairs which have a total of 2N
complex eigenvalues λ 1 , λ 2,···λ Ν, λ 1 *, λ 2*,· · λ Ν*. There are 2N corresponding complex vectors:

2. Orthogonality

In the complex modal analysis, the 2N-dimensional main space formed by the complex eigenvector as the basis vector is orthogonal. The real mode is
orthogonal in the N-dimensional main space.

3. Decoupling

In the complex modal analysis, the equations of motion of the dynamical system can be decoupled in the 2N dimensional space of the complex
eigenvector. The motion equation of the dynamical system is decoupled in the N-dimensional space when the real mode is analyzed.

4. Movement Characteristics

In the complex modal analysis, the dynamical system presents some different movement characteristics with the real-mode analysis:
1) In the complex modal analysis, the phase of each particle of the power system is different so there is a phase difference. For real modal analysis,
the phase difference of each particle is 0° or 180°.
2) Since there is a time (phase) difference in the motion of each particle in the complex mode, they do not pass through the equilibrium position of
the vibration at the same time. For real modal analysis, the particles pass through the vibration balance position at the same time.
3) In the complex modal analysis, the dynamic system has no fixed vibration mode. That is to say, the particle is not necessarily fixed but for
periodic movement, which is very different from the real mode.
4) The free vibration of the complex mode is the attenuation vibration and the attenuation rates of the points are the same. The magnitude of the
attenuation rate is represented by ar in equation (7.69), which is the same as the real mode.

7.4 Structural Dynamic Damage Identification

Both traditional methods of detection (such as visual observation) and non-destructive testing techniques (such as ultrasound, acoustic emission,
radiation, etc.) are structural damage detection methods. It is difficult to predict the degradation of the overall performance of the forecast structure
and real-time damage diagnosis cannot be achieved. It is known that damage to the structure will inevitably lead to the changes of structural
performance parameters (stiffness, frequency, vibration). Damage identification, which is the overall detection method of the structural damage, allows
for the acquisition of the change in these parameters and can be used to locate and quantify the structural damage.
In structural damage detection, the following problems need to be solved:
Whether the structure is damaged. Judgment of structural damage position. The severity of structural damage. The effect of structural
damage on the structural performance, or the prediction of the residual life of the structure. The method of damage diagnosis mainly includes dynamic
identification method, static identification method, neural network method, statistical analysis method, and so on. This section focuses on commonly
used dynamic damage identification methods.
The dynamic damage identification method is to compare the observed dynamic parameters’ change with the reference parameters and select the
most likely change to determine the true state of the structure. Structural damage identification can be roughly divided into four steps: Select the
vibration observation signal. Extract the feature quantity associated with the broken state. Identify structural damage. Identify the location,
nature, and degree of the damage. At present, the commonly used dynamic damage identification methods are: frequency base damage identification
method, modal base damage identification method, compliance method, stiffness method, energy method, and sub-structure method. The following is a
brief introduction to the above methods and sub-structure methods will be described in the next section.

7.4.1 Frequency Base Damage Identification Method

The natural frequency of the structure is the easiest and most accurate measurement of the dynamic parameters. According to the method of damage
identification based on the natural frequency change, it is considered that the structural rigidity is reduced and the structural mass change is neglected
when the damage happens in the structure.
The natural frequency is processed into a dimensionless unit and the relative change ratio of the natural frequency before and after the occurrence of
the damage is considered as a dynamic fingerprint. As shown in the following equation:
(7.72)

(7.73)

In the equation, FFCi is the i-th natural frequency of the relative rate of change. f0i and fdi is the natural frequency of the structure in health and
damage state respectively. NFCRi (i = 1,2, …, m) is the natural frequency change rate after normalization.m is the number of natural frequencies
measured. Using natural frequency changes in damage identification the natural frequency is easy to obtain in the test. The test accuracy is relatively
high and the frequency is independent of the selected measuring point position and the method is simple. But the natural frequency is not very
appropriate to reflect the local micro-damage of the structure and is not very sensitive to the early damage of the structure. Damage at different
locations may cause the same frequency change and the location of the injury cannot be determined.

7.4.2 Modal Base Damage Identification Method

The modal base damage identification method is also known as the vibration foundation damage identification method. There are many methods to
identify structural damage by modal changes. Among these methods, the modal confidence criterion method, displacement modal method, modal
curvature method, and strain mode method are more commonly used.

1. Modal Curvature Method

The modal curvature method is the most commonly used method in modal base damage identification method. The difference of modal curvature
before and after injury is called the damage localization parameter. If the structure is broken, the stiffness at the damage will decrease and the modal
curvature difference will increase. The change of modal curvature increases with the increase of damage. Therefore the position of the damage can be
determined based on the change in the curvature of the mode.
By the geometric significance of curvature, the equation is:
(7.74)

In the equation, x is the coordinate along the length of the beam and v is the vertical displacement. If we consider the displacement of two points in the
x-direction of the structure, let h represent the horizontal distance between the two points, and the above equation can be further differentiated:
(7.75)

The modal curvature difference is the difference between the curvature when the structure is damaged and when no damage occurs. The equation is:
(7.76)

In the equation, v″d and v″0 are the curvatures when the structure is damaged and when no damage occurs, respectively. Since the displacement
variable is a variable related to the modal shape, from the equation (7.75) we can see that the local mutation of the displacement will result in a local
mutation in the curvature mode. Using this mechanical relationship, it is easy to determine the damage and damage degree of the structure by the
difference of the local modal numerical change from theoretical analysis or test.

2. Modal Confidence Criterion

The modal assurance standard (MAC) is defined as:

(7.77)

In the equation, Φ d and Φ 0 are the mode matrix of a lossy and intact structure respectively. The subscripts i and j represent the i-th and j-th modes.
The value of MAC(i, j) is 0~1. Approaching 1 means that the structure is healthy and approaching 0 indicates that the structural damage is serious.

3. Displacement Modal Method

The difference between the displacement modes before and after injury is taken as the damage localization parameter. When the damage occurs, the
modal difference in the degree of freedom affected will appear in the lesion area with a relatively large value. The position of the damage can be
identified using the modal difference. The displacement difference before and after the damage is:
(7.78)

In the equation, ΔΦ u is the displacement modal difference. Φ d and Φ 0 are the mode matrix of a lossy and intact structure respectively using
displacement data.

4. Strain Mode Method

Modes reflect the inherent characteristics of the structure, and strain and displacement are first derivative relationships. As the strain mode changes
significantly in the local damage position, the absolute value of the strain modal difference before and after the damage can be used as the identification
index of structural damage which is recorded as:
(7.79)
In the equation, ΔΦ ε is the strain modal difference. Φ d and Φ 0 are the mode matrix of a lossy and intact structure respectively using strain data.
Modal state is more sensitive to local stiffness changes. So, the modal base damage identification method has a high accuracy. However, these methods
require precise modal measurements, very intensive measurement points, in order to obtain accurate modal. Otherwise the error of the modal base
damage identification method will increase. In the test process the measured mode is often incomplete (not only the number of modal measurements is
less than the number of analytical models, but also the number of degree of freedom measured is less than the degree of freedom of the model) and the
high-order mode is especially difficult to detect. This makes the method of dynamic damage identification based on modal face the problem of
incomplete measurement and noise. Some scholars recommend to use incomplete measurement mode in structural damage identification. The
standard is the method based on the sensitivity and numerical values, which will not be introduced in detail; interested readers can refer to the
relevant literature.

7.4.3 Damage Identification Method Based on Stiffness Variation

When structural damage occurs, the stiffness of the structure will change and the modal parameters of the structure (eigenvalues, eigenvectors, etc.)
will also change. If the change of the mass matrix caused by structural damage is neglected the eigenvalue problem can be expressed as:
(7.80)

(7.81)

There is:
(7.82)

In the equation, [Kd] is the stiffness matrix of the damage structure. α i is the unit stiffness correction coefficient of the i-th unit when the structure is
damaged. [Ki] is the contribution matrix of the i-th unit element stiffness matrix to the global stiffness matrix. α i(j) is the element stiffness correction
coefficient of the i-th unit of the damaged structure corresponding to the j-th mode. λ dj and {φ j}d are the j-th eigenvalue and eigenvector after the
damage of the structure, which can be obtained by dynamic measurement.
Usually we also need to consider the error coefficient that the impact of modeling brings about and take into account the posterior unit stiffness damage
coefficient [η] = {η1 ,η2,,,ηη}T. Among them, ηi = βi, – α i. From the size of ηi the degree of damage to the unit can be determined. ηi is small and the
corresponding unit damage level is small. ηi is large and the corresponding unit damage level is large. The stiffness sensitivity of different units is
different for different structural vibration modes. When the structure produces minor damage (such as <5%) the stiffness method will be difficult to
use to get the correct damage identification conclusion.

7.4.4 Damage Identification Method Based on Flexibility Change

The flexibility matrix is a function of the mode and the reciprocal of the frequency. So, as long as the measurements of the previous few low-order
modal parameters and frequency, the flexibility matrix can be obtained with good accuracy. According to the difference in the matrix of the two
flexibility matrices before and after the damage, the largest element in each column of the difference matrix is obtained and the damage location can be
found. The relationship between the flexibility matrix F and the modal data is:
(7.83)

(7.84)

(7.85)

In the equation, ω i is the i-th modal frequency of the structure. φ i is the i-th modal mass normalized mode. ΔF is the flexibility difference matrix. F 0
and F d are flexible matrix of a lossy and intact structure respectively. δj represents the maximum absolute value of the j-th column element in ΔF.
The flexibility method ignores the higher modal order and is more accurate than the stiffness method. But in the case where the damage is not
symmetrical, the position of the injury is closer or the amount of damage is different, recognition accuracy is poor and can even incorrectly identify the
most seriously damaged locations.

7.4.5 Energy-Based Damage Identification Method

Damage identification based on the difference of energy (such as strain energy) before and after structural damage is called the energy-based damage
identification method. Within the energy-based damage identification method, the most commonly used is the modal strain energy method. The strain
energy of the i-th mode in the j-position is related to the modal curvature of the point. It can be represented as:
(7.86)

In the equation, and are the curvature of the first mode in the healthy and damaged states of the structure, respectively, according to the
definition in the curvature mode method. a and b are the boundary of the structural damage. When there are N modalities, take βij’s average.
The modal strain energy method utilizes both modal and finite element information. The basic principle is to use the difference in strain of the structure
before and after the early damage to identify damage. Professor Zhao-Dong Xu of Southeast University, has proposed an energy spectrum evolution
index based on the acceleration response. That is:
(7.87)
In the equation, Ε(ω) is the energy spectral density function of the acceleration response. is the Fourier transform of the acceleration response.
Based on this index the energy difference index and energy curvature difference index are established and the damage location and quantification of a
long span cable-stayed bridge are analyzed and studied. The results show that the proposed energy difference index and energy curvature difference
index have a good ability to locate damage, quantification and reduce noise pollution under single load excitation.

7.4.6 Prospects for Research on Dynamic Damage Identification

The method of dynamic damage identification in structural damage identification is mainly based on the change of frequency, vibration mode, modal
curvature, frequency response function, strain mode, modal flexibility matrix, transfer function, power spectrum, energy transfer ratio (ETR), modal
assurance guidelines (MAC, COMAC), and other dynamic parameters to characterize the change of structural traits. Power fingerprints can be divided
into three categories: Using a single dynamic characteristic, such as Hearn frequency ratio, Stubbs overall damage estimation method, power
spectrum method, vibration difference method, modal guarantee standard method (MAC), coordinate mode assurance standard method (COMAC),
strain Modal Method and Modal Curvature Method Derived from the First Order Second Derivative Concept of Oscillation (Displacement). Using
multiple dynamic characteristics (frequency and vibration), there are: flexibility difference matrix, stiffness difference matrix, load-deformation
method and energy-based energy damage fingerprints and energy differential fingerprints. Using other test responses, such as FRF waveforms
presented by Samman, etc.
In summary, some experts and scholars at home and abroad have done a lot of research on structural damage identification methods and achieved
many gratifying results. However, there are still many important issues in the field that require our active efforts to explore.
1) More reliable damage identification indexes need to be developed so that the index of different load incentives will not be mistaken or missed.
To achieve this purpose, the power parameters used must be sensitive and accurate to measure.
2) It is valuable to study the effect of the change of test parameters, the change of environmental parameters (such as temperature) and the
change of excitation parameters on structural damage identification.
3) Research on Damage Detection which does not rely on external excitation source. For precision machinery and aerospace structure,
piezoelectric sensors can be used as actuators. For the civil engineering structure, because of its large size is difficult to use, and the use of
environmental vibration on structural damage identification is worthy of further study.
4) Because of the small impact of small damage on the system parameters coupled with the impact of noise, many of the above methods will lose
recognition. However, for large-scale civil structures (such as large-span bridges, etc.), the accumulation of small damage will lead to structural
damage in a short time. To prevent unawareness that the whole structure is being destroyed, how to accurately identify small damage is an
important research topic in both practical and theoretical fields.
5) For damage identification of large complex structures, combining multidisciplinary techniques such as test technology, vibration theory, signal
processing, pattern recognition, artificial intelligence, control theory, and material structure to establish online real-time monitoring technology is a
development trend.

7.5 Nonlinear Problems of Dynamic Analysis

In the preceding chapters, for either SDOF or MDOF systems, the basic motion equations of the system can be expressed as follows:
(7.88)

For the SDOF system, the above equation is a second-order ordinary differential equation with constant coefficients. For an MDOF system, the above
equation becomes an equation or matrix. But it is still a constant coefficient. The reason for the emphasis on constant coefficient is that the physical
meaning can be understood in this way: Elastic force is proportional to the displacement and the damping force is proportional to the velocity and the
inertia force is proportional to the acceleration. The system is known as the linear system. The biggest feature of this system is that the superposition
principle can be used to solve the dynamic response of the system. For example, if there are a number of concentrated forces on a beam the dynamic
response of each force can be solved separately. Then superimpose them and the final dynamic response can be obtained. Likewise, if the applied load
is multiplied the dynamic response of the system will also increase exponentially.
However, in engineering practice, not all structural systems are able to meet the requirements of linear systems. For example, when the elastic force of
a spring does not scale with it there is an important nonlinearity. As shown in Figure 7.16, for a linear system the elastic force of the spring is
proportional to the displacement and the slope of the thin line in the figure is the stiffness coefficient of the spring. While the other two curves b,c in the
figure represent “hardening springs” and “softening springs” The physical meaning of the hardening spring is that the growth rate of the elastic force is
faster than the growth of the displacement, whereas the softening spring is the opposite.

Figure 7.16 Two different spring models.

For the “hardening spring” or “softening spring”, there is no linear relationship between the elastic force and the displacement. So, the elastic force
formula will no longer be established. That is, the basic vibration equations of the linear system (7.88) will no longer be established and this is the
background of the dynamic nonlinear problem.

7.5.1 Physical Nonlinearity Problems in Dynamic Analysis

The physical nonlinearity in structural dynamic analysis mainly refers to the nonlinear problem of structural restoring force. Resilience is the ability of
a structure or component to restore its original shape after external load removal. For example, according to (2-29), the motion equation of the
structure under earthquake action is:
(7.89)
Since the formula for the linear elasticity is no longer true, the above equation can only be rewritten as:
(7.90)

In the equation, f(u) is the restoring force and it is a function of the displacement u.
It can be seen that the key to solving physical nonlinear dynamic problem is to determine the restoring force model. Once the restoring force is
determined, it can be solved according to the linear acceleration method, Newmark-β method, and so on from Chapter 4.
Restoring force model refers to the restoring force curve model, generally including three components: the skeleton curve, hysteresis characteristics,
and stiffness degradation. The methods of determining the restoring force curve are the test fitting method, system identification method, and
theoretical calculation method. The restoring force characteristic curve can be expressed by the correspondence between the bending moment of the
component and the corner, bending moment and curvature, load and displacement, or stress and strain.
The characteristic curve of the load-displacement restoring force of the reinforced concrete beam is shown in Figure 7.17. The members form a series
of hysteresis loops under the repeated action of the load P. When P is small, the structure is flexible and the formula F s(t) = ku(t) can be established. As
the load increases, the stiffness of the beam begins to drop. That is, the rate of increase of restoring force is smaller than that of displacement and the
structure is nonlinear.

Figure 7.17 Hysteretic curve of reinforced concrete beams.

In the seismic response analysis, if the above curve is used, the calculation is too complicated and therefore needs to be modeled. At present, the
commonly used nonlinear restoring force model can be divided into curve model and polyline model. For reinforced concrete structures and
components, the most commonly used models are bilinear and degenerate trilinear model.
The bilinear model [Figure 7.18(a)] is the simplest restoring force model. The forwardly loaded skeleton curve uses two straight lines 0→1 and 1 →2
whose shape is determined by the yield strength Py, elastic stiffness k 0, and yield stiffness k 1 . The reverse loading skeleton curve is the same as the
forward direction. The loading and unloading stiffness remain constant and are equal to the elastic stiffness k 0.
Figure 7.18 Two commonly used resilience models.
The forward loading skeleton curve of the degenerate trilinear model [Figure 7.18(b)] consists of three straight lines 0→1, 1 →1, and 2→9 whose shape
is determined by the crack load Pcr, yield load Py, and stiffness of each stage. The reverse loading skeleton curve is the same as the forward direction
and the unloading stiffness of the model remains the same which is equal to the secant stiffness of the yield point (slope of 0→2 line). The loading
stiffness takes into account the degradation phenomenon, which makes the hysteresis line point to the maximum displacement point of the previous
cycle. The model can well reflect the characteristics of reinforced concrete elements with bending failure.
In addition, the commonly used restoring force models also include the Romberg-Osgood model under the plane force system, the Clough degenerate
bilinear model, and other models under the state of the space force. They will not be described one by one here.

7.5.2 Geometric Nonlinearity Problems in Dynamic Analysis

The geometric nonlinear problem is: If the deformation of the bar is large, you cannot analyze the balance of the force on the basis of the original
geometry, but should analyze on the basis of the deformed geometry. In this way, there is a nonlinear relationship between force and deformation
which is called geometric nonlinearity.
Here is an example to illustrate the impact of geometric nonlinearity on dynamic problems.
Example 7.2 As shown in Figure 7.19(a), a small mass m is located in the middle of a tensioned steel wire AB with a length of 2l. The mass is
subjected to an initial pulling force of the size F 0. When the mass moves laterally from its equilibrium position by a distance u and the wire produces an
elastic restoring force. In this way, the mass m begins to vibrate freely. Find its equation of motion.

Figure 7.19 The mass steel wire system which has geometric nonlinearity.
Solution: According to the D’Alembert principle, the mass vibrates only by the inertial force and the elastic restoring force as shown in Figure 7.19(b).
So, the motion equation of this system is:
(a)

In the equation, E, A, Δ represent the elastic modulus of the wire, cross-sectional area, and the change in length l due to displacement u, respectively. θ
represents the angle between the wire and the vertical line.
From the geometric relationship we can see:
(b)

Substituting equation (b) into equation (a), there is:

(c)

According to the series expansion (b), its approximate value can be obtained and there is Δ ≈ u2 / 2l, sinθ ≈ u / l. Substituting them into equation (a):
(d)

And let k = 2F 0 / l, there is:

(e)

It can be seen that both the exact equation of motion (c) and the simplified form (e) of the system are non-linear. If the initial tension F 0 is high and the
displacement u is small, then the cubic term of equation (d) can be neglected and this is the linear motion equation. Otherwise, this motion equation due
to the larger initial displacement u0 will be a geometric nonlinear motion equation.
In addition, another geometric nonlinear problem that often occurs in engineering structures is the P – Δ effect. This means that when the structure is
subjected to a horizontal force and the lateral displacement Δ occurs, the vertical force P causes the structure to produce additional moments and
additional lateral displacements so that the total moment and total lateral shift increase.
As shown in Figure 7.20, a mass of mass m produces a relative displacement ux, uy under horizontal loads Px and vertical loads Py. The rod length is l,
the stiffness and the damping coefficient in the two directions are k x, k yand c x, c y, respectively.

Figure 7.20 P – Δ effect of single particle system.

When the structure has a large horizontal displacement, the bending moment of the substrate under the action of vertical load and inertial force is:
(7.91)

If this additional moment is replaced by an equivalent horizontal force F eq, the following equation exists:
(7.92)

k eq is horizontal impact stiffness and k eq = (Py + müy)/l. At this point, the horizontal motion equation for the particle is:
(7.93)

Substituting equation (7.92) into the above equation gives:

(7.94)

It can be seen that P – Δ effect is equivalent to reducing the stiffness of the structure (k x – k eq), so the dynamic response of the structure will increase.
In addition, the equation (7.94) is linear in form, but in essence, it is a nonlinear equation. This is because the horizontal impact stiffness k eq also
includes the effects of ux.

7.6 Sub-Structure Method

Large civil structures and fine aeronautical structures are large in scale and complex in structure. The number of degrees of freedom and the number
of unknown parameters are enormous. This poses a great challenge to structural analysis and damage identification. It’s thought that large or complex
fine structures can be divided into sub-structures to reduce the size and difficulty of the problem, which is the origin idea of the sub-structure. The
substructure method is usually dividing the structure into finite sub-structures and through the analysis of a limited sub-structure, the analysis of the
entire structure is achieved. In the analysis process, a set of elements need to be condensed into a single matrix and this single matrix element is called
a supercell. The use of sub-structure is mainly to save time and allow solving large-scale problems on the basis of a limited number of computer
equipment resources. Its effectiveness is reflected in nonlinear analysis and analysis with a large number of repetitive geometries. In the nonlinear
analysis, the linear part of the model can be constructed as a substructure so that the unit matrix of this part does not have to be repeated in the
nonlinear iteration process. In models with repetitive geometries, you can generate supercells for duplicate parts and then copy them to different
locations, which can save a lot of machine time. If the disk space a problem requires is too large, the user can analyze the problem through sub-
structures. The sub-structure method is usually used for finite element analysis or damage identification of large structures and complex fine
structures. The following will briefly introduce the finite element analysis and damage identification of sub-structure method.

7.6.1 Finite Element Analysis of Sub-Structure Method

The finite element equation of structural analysis is:
(7.95)

The analysis of the sub-structure method is to decompose the equation (7.95), obtain the local solution and then combine the local solution into the
whole solution. The node displacement of the sub-structure can be divided into two groups: The first group is shared with other substructures or units
and has a displacement coordination relationship which belongs to the boundary node displacement. It has the subscript R. The second group has no
displacement coordination relationship with other substructures or units which is denoted by subscript I. Therefore equation (7.95) can be decomposed
into:
(7.96)

Equation (7.96) can be expanded as:

(7.97)

Find the internal node displacement:

(7.98)

The first term on the right side of the above equation is the local solution of the internal node. The remaining solutions on the internal nodes are due to
the local solution generated by the motion boundary nodes and are the second term on the right side of the equation. After the expansion:
(7.99)

That is:

The above equation is repeatedly assigned to different substructures and the stiffness matrix of the whole system is:
(7.100)

In the equation, nm is the number of units in the main structure. [KiM] is the stiffness matrix of the first principal structure. ns is the number of sub-
structures that connect the main structure. It can be seen that the displacement and stress of all sub-structures can be solved once the boundary node
or sub-structure node displacement is calculated. Since the order of the boundary stiffness matrix [KRR] is much smaller than the order of the matrix
of the sub-structure stiffness, the order of the structural stiffness matrix obtained by the respective sub-structures of the final set is greatly reduced.
So the sub-structure method can be applied in the computer to solve the power problem of large-scale structure.
At present, the existing finite element analysis software can conduct sub-structure analysis. The ANSYS sub-structure software consists of three
steps: the Generation, the Solution, and the Extension.
The Generation: Condense a series of ordinary finite element elements into a super-unit. Cohesion is achieved by defining a set of main degrees of
freedom. The degree of freedom is used to define the boundary between the supercell and the other elements in the model and to extract the dynamics
of the model. At the beginning of the model it is necessary to plan two parts of the model, that is, to determine how the super-unit and non-hyper-unit
parts are connected. To ensure that the connection is correct you should ensure that the contact part of the node number is consistent.
The Solution: the part of analysis by connecting super-unit and the whole model. The whole model can be a supercell, or as described above, a cell
connected by supercell and other non-supercell. The solution of the use part includes only the condensing solution for the supercell. Degree of freedom
calculations are limited to degrees of freedom and full solution for non-hypercomponents and trying to turn supercells into part of the model.
The Extension: From the results of the cohesion, calculate all degrees of freedom in the supercell. If there are multiple units in the part of the Solution,
a separate extension process is performed for each supercell.

7.6.2 Damage Identification by Sub-Structure Method

It has been mentioned that the finite element analysis of large-scale civil structure and complex fine aeronautical mechanical structures can be carried
out by sub-structure method for rapid calculation. Similarly, in order to quickly identify the damage of these structures, the sub-structure method can
also be used to reduce the difficulty of identification and calculation of the workload.
It is common to divide the structure into finite sub-structures and to identify the entire structure by recognizing the finite sub-structure. It can be
seen as a theory based on the modal correction and it is a sub-structure algorithm that corrects structural stiffness by structural modal data. It can be
represented by the equation:
(7.101)
(7.102)

In the equation, [Kd] and [Md] are the stiffness matrix and the mass matrix of the damage structure respectively. [K0] and [M0] are the stiffness
matrix and the mass matrix of the intact structure respectively. [Ki] and [Mi] are the stiffness matrix and mass matrix of the structures to be
modified. α i and βi are the degree of damage of the matrix stiffness matrix and the mass matrix. p and q are the number of damaged sub-structures.
Only part of the vibration mode vector can be measured and the cohesion method can be used to eliminate the measured vibration vector. That is:
(7.103)

And it is also:
(7.104)

In the equation, superscript 1 is the measured vector and superscript 2 is the unmeasured vector.
In the calculation we usually do not consider the damage to the mass matrix, substitute the correction matrix into the power equation, use the
measured vibration mode to modify the original structure, and use the least squares method to estimate the degree of damage to the structure.

7.7 Dynamics of Offshore Structures

To evaluate the fluid-induced forces acting on a structure or on any of its components, we need to know its surrounding hydrodynamic flow field. For
offshore structures, this flow field arises from time varying natural processes: winds, currents, and surface gravity waves. Although these processes
nearly always occur in various combinations, the scope of this chapter is limited to descriptions of surface gravity waves, which are usually wind-
generated. It has been observed that these time-dependent waves occur on two different scales. The shorter time scale, measured in minutes or
seconds, is useful for describing detailed features such as wind gusts and surface wave periods. This shorter time scale corresponds most closely to the
response time of fixed offshore structures. The longer time scale, measured in terms of hours, days, or even years, is useful for describing variations in
the wave intensity and in its statistics. The longer time scale is important too because structures may fail in low cycle fatigue fracture after months or
years of service, a topic discussed by Etube (2001). In this present chapter, we give a brief synopsis and critique of the classical wave theories used
most often by engineers for preliminary calculations of forces on offshore structures. Here, the focus is on the shorter time scale. For comprehensive
bibliographies of wave theories, see Sarpkaya and Isaacson (1981) and Young (1999).
Illustrated in Figure are the two fundamentally different descriptions of surface gravity waves: deterministic and probabilistic. Deterministic
descriptions, analytical or numerical, are used to characterize the short time scale features of waves. Deterministic analytic descriptions encompass
classical wave theories, which in turn are subdivided into linear and nonlinear types. Probabilistic descriptions are used to characterize the long-time
scale features of offshore waves. In both the deterministic and probabilistic descriptions, linear wave theory is important to engineers for two reasons:
it is simple to apply when estimating forces on offshore structures during the preliminary phases of design; and it affords a simple basis for estimating
the probability of failure of a given structure.

7.7.1 Descriptions of Offshore Waves

Before discussing the most important classical wave theories, we first introduce some fundamental definitions and concepts. The natural occurrence of
waves is usually complicated by the simultaneous superposition of waves of many different shapes and energies. Thus, we define a single wave or a
train of waves of pure form that always behaves in exactly the same way on passing any given point, and that propagates without change of shape
when the surrounding conditions remain the same. Further, we consider only plane or long-crested waves of permanent form, waves that are
independent of the coordinate normal to the wave propagation direction. Therefore, their crests or troughs can be represented by straight lines in a
horizontal plane. By permanent form, we mean that the field of motion, pressure distribution, and surface configuration are maintained as one follows
the wave at a velocity c, the designation of the phase velocity or celerity. (In a strict sense all real waves have finite length crests, a factor that is
ignored herein.) Consider a wave form that is simple harmonic so that at any time t the wave has a sinusoidal shape with reference to the still water
line (SWL).
The theory of linear waves is alternatively known as Airy’s theory, small amplitude theory, and first-order theory. Developed primarily by Airy (1845)
and by Laplace (1816), it is the most important of the classical theories because it is both easy to use and it forms the basis for the probabilistic spectral
description of waves. For a full development and discussion of the theory, the reader should consult the intellectually entertaining work by Kinsman
(1965). We now summarize the assumptions, the governing equations, and the solutions for the wave velocity and pressure profiles useful for
predicting wave-induced forces on offshore structures.

7.7.2 Introduction to Wave Spectra

A short time record representative of measured irregular surface wave forms offshore is shown in Figure 6.1. For such irregular forms, the wave
periods and wave heights take on new meanings as statistical ways of describing these waves are now explored. Procedures are now defined to portray
the important fluctuations and to condense these data into manageable forms.
Consider first the time interval at which the wave record is sampled at discrete points to give an adequate representation for η(t). For instance, the
record in Figure 6.1 can be marked off in equally spaced time intervals of Δt=1sec such that the sequence of ordinates, when connected by a smooth
curve, reproduces the important details of the record. One suspects that there is an optimum time interval which is neither too fine nor too coarse and
which best approximates the record. It is apparent that a 10-sec interval is too large, but also the choice of 100 points over the time interval 10
seconds is far too many. The Nyquist sampling theorem aids in the selection of an optimum interval Δt.
The interest in wave spectra arises from the need to select a representative design-loading history acting on the proposed offshore structure. The
design wave spectrum, usually furnished by a specialist in hindcasting or forecasting, is chosen to represent the worst possible combination of waves
which would lead to the highest loading for the structure. To complicate matters, there may be many different surface water wave elevation histories
characterized by essentially the same spectrum. In other words, a one-to-one correspondence between the surface water wave elevation record and
the spectrum does not generally exist. However, all records that yield identical spectra do have the same statistical properties, although details in the
records may vary widely. The suggested procedure, then, is to select a number of water wave elevation records corresponding to the design wave
spectrum and to analyze the proposed offshore structure under these alternative conditions.
It is important that the mathematical model used to represent the sea surface (the time history of wave height) should conform as closely as possible to
the actual waves. In practice, there are two ways to accomplish this. First, a purely deterministic method can be used in which the sea surface records
are fitted to a harmonic or Fourier series. This representation is unaffected by nonlinear wave effects. Second, a statistical method can be used in which
the sea surface records are fitted to a spectral function in the frequency domain. It is emphasized that the second method is based on the concept of
superposition and is adequate only for linear seas. Various efforts have been made to include nonlinear phenomena in spectral descriptions but these
methods are not widely utilized. For purposes of structural design, however, linear models usually provide adequate descriptions of the sea surface.

7.7.3 Frequency Domain Analysis

The derivation of response statistics that follow are based on a linear model in the form of equation, written in terms of the displacement coordinate v.
However, similar results can be obtained based on a linear model in the form of equation, written in terms of the rotational coordinate θ. After going
through the following derivation for the displacement model, the reader is encouraged to then derive the response statistics for the rotational model.
Given the linear structural model
(7.105)

for which the load spectral density S P1 (ω) is known through equation, the task now is to calculate the response spectral density S v(ω) and the variance
. Because these results are so important in applications, all of the assumptions and mathematical details needed for this derivation are included here.
Because two particular response functions for equation are required eventually, these are repeated for convenience. One is the harmonic response
function given by equation:
(7.106)

The other is the impulse response function given by equation

(7.107)

A preliminary step is to relate H(ω) to h(t) through the Fourier transform.

A linear, single degree of freedom flexible structure is subjected to a total wave force p1 (t), in line with the motion of the structure. Assume a
distribution of simple, linear waves for which the wave height η(t) is stationary, ergodic, and Gaussian, with a zero mean and with a spectral density
S η(ω). Starting from basic definitions, relate the spectral density of the wave load S p1 (ω)to η(ω) through a known transfer function G(ω)
First, rewrite the structural load-wave height relationship as
(7.108)

The required relationship is deduced as

(7.109)

Exercises
7.1 Use Matlab Programming to solve the Acceleration Response Spectrum of El-Centro Seismic Wave.
7.2 Calculate the seismic shearing force of each floor of the three-layer frame structure shown in Figure 7.2 by using the mode decomposition response
spectrum and the bottom shearing force. The following are known paraments:
m1 = 5.95 × 104 kg,m2 = 11.085 × 104 kg, m3 = 11.660 × 104 kg. The vibration period and vibration mode are presented in Figure 7.2. α max = 0.16, T g
= 0.4s.

7.3 What types can structural vibration control be divided into? Explain the ideas of various control methods.
7.4 What are the methods of dynamic damage identification? And what are the basic ideas of them?
7.5 In dynamic analysis, what is physical nonlinearity? And what is geometric nonlinearity?
7.6 What is the basic idea of substructure law? When is it often applied?

References
1. Camera, E., Giunta, G., Petrolo, M., Beam structures: classical and advanced theories, John Wiley & Sons, 2011.
2. Bhimaraddi, A., Free vibration analysis of doubly curved shallow shells on rectangular planform using three-dimensional elasticity theory. Int. J.
Solids Struct., 27, 7, 897.913, 1991.
3. Clough, R.W. and Penzien, J., Dynamics of Structures, McGraw-Hill, 1995.
4. Chopra, A.K., Dynamics of Structures, Prentice Hall, 2000.
5. Wilson, J.F., Dynamics of Offshore Structures, John Wiley & Sons, Inc, 2003.
6. De Wachter, S. and Tzavalis, E., Detection of structural breaks in linear dynamic panel data models. Comput. Stat. Data Anal., 56, 11, 2012.
7. Cluni, F., Gioffrè, M., Gusella, V., Dynamic response of tall buildings to wind loads by reduced order equivalent shear-beam models. J. Wind Eng. Ind.
Aerodyn., 123, 2013.
8. Mackie, R.I., Dynamic analysis of structures on multicore computers – Achieving efficiency through object oriented design. Adv. Eng. Software, 66,
2013.
9. Zhao-Dong, X. and Le-Wei, M., Dynamics of Structures, Science Press, Beijing, 2007.
Chapter 8
Structural Dynamics of Marine Pipeline and Riser
The offshore industry requires continued development of new technologies in order to produce oil in regions. Marine pipelines and risers for the
transportation of oil and gas have become a safe and reliable way to exploit the valuable resources below the world’s seas. The design of pipelines and
risers is a relatively new technology and continues to evolve in its quest to reduce costs and minimise the effect on the ocean environment.
The purpose of this chapter is to introduce the basic theory of structural dynamics of submarine pipelines and risers. Calculating environmental loads
and load combinations to understand the basic concepts of waves, currents and hydrodynamic forces is the first step in marine pipeline and riser
design. A systematic dynamic analysis method for structural analysis of marine pipelines and risers has been developed to predict the dynamic
response against wave and current loads. Vortex induced vibrations and the associated dynamic effects are also an important factor in both the
structural design and vibration control. Basics of vibration analysis will be covered in the relevant section of this chapter.

8.1 Overview
The first pipeline was built in the United States in 1859 to transport crude oil (Wolbert, 1952). Through the one-and-a-half century of pipeline
operating practice, the petroleum industry has proven that pipelines are by far the most economical means of large-scale overland transportation for
crude oil, natural gas, and their products, clearly superior to rail and truck transportation over competing routes, given large quantities to be moved on
a regular basis. Transporting petroleum fluids with pipelines is a continuous and reliable operation. Pipelines have demonstrated an ability to adapt to a
wide variety of environments including remote areas and hostile environments. Because of their superior flexibility to the alternatives, with very minor
exceptions, largely due to local peculiarities, most refineries are served by one or more pipelines.
Subsea pipelines are used to transfer oil from the production platforms to storage facilities or to the shore. One type of vertical offshore pipeline is the
marine riser, which is shown in Figure 8.1. A riser is a pipe that connects an offshore Floating Production Structure or a Drilling Rig to a sub-sea system
either for production purposes such as drilling, production, injection and export, or for drilling, completion and workover purposes. Due to the dynamic
environment and the movement of host facility, the dynamic effect has been significantly taken into consideration for riser design.

Figure 8.1 Uses of offshore pipelines and risers.

8.2 Environmental Conditions

8.2.1 General
Every offshore facility is subjected to several types of environmental loads, such as winds, currents and waves during its lifetime. Such loads occur both
on a short time scale of seconds and minutes, induced by periodic vortex shedding, wind gusts, and wave slamming, and over longer periods of hours,
days, and even years, where the loads are induced by steady wave, tides, and hurricanes.
In this book, the environmental conditions generally considered in the marine pipelines and risers include:
Waves
Currents
Environmental data includes: (1) wave heights, time periods and directions; waves generate hydrodynamic forces on sub-sea structures. (2) wind
speeds and direction, which drive sea currents. (3) currents and directions, which also cause hydrodynamic loads on pipelines and risers.

8.2.2 Linear Wave Theory

Linear wave theory was first published by Airy in 1845 and is thus named Airy wave theory. The linear wave theory assumes that the waves consist of
small amplitudes in comparison to the wavelengths. It is for this reason that it is often referred to as small amplitude wave theory, and first-order
theory. Other assumptions made in developing linear wave theory include:
Incompressible, homogeneous fluid;
 Neglect surface tension and Coriolis;
Uniform and constant pressure at the surface;
Ideal fluid without viscosity;
Horizontal, impermeable seabed;
Long-crested waves (2D approach);
The fluid motion is irrotational.
Figure 8.2 shows 2D regular long-crested waves defined by their wave amplitude and frequency, giving the following expressions for the wave
kinematics:
(8.1)

Figure 8.2 Regular long-crested waves.

Fluid velocity component in the x-direction,
(8.2)

Fluid velocity component in the z-direction,

(8.3)

Fluid acceleration component in the x-direction,

(8.4)

Fluid acceleration component in the z-direction,

(8.5)

Dynamic pressure,
(8.6)

Figure 8.3 shows 2D random long-crested waves defined by significant wave height, peak frequency, and etc. The formulation in this section is based
on the use of a wave spectrum, as shown in Figure 8.4, to define the characteristics of the sea-state.
Figure 8.3 Random long-crested waves.

Figure 8.4 Wave spectrum.

As an example, the Joint North Sea Wave Project (JONSWAP) proposed the JONSWAP spectrum based on measurements in the North Sea, which can
be defined as:
(8.7)

where:

f : frequency;
fp : spectral peak frequency;
g : acceleration of gravity;
ap : Phillips’ constant;
σ : spectral width parameter;
α : a constant that relates to the wind speed and fetch length. Typical values in the northern north sea are in the range of 0.0081 to 0.01.
γ : the peakedness parameter of JONSWAP spectrum, determined from,
(8.8)

Nowadays, it is more common to use Goda’s form, which specifies the spectrum in terms of Hs, Tp and γ,
Where:

From the wave spectrum we can find several properties. λ n denotes the nth (stress) spectral moment defined by:
(8.9)

H1/3 is the significant wave height and can be calculated by:

(8.10)

The average period measured at zero crossings is:

(8.11)

and the spectral band width parameter ε as:

(8.12)

By performing an inverse transformation, the wave amplitudes (ai) and frequencies (ω i) of each wave component is extracted from the wave
spectrum.
Extraction of amplitudes and frequencies from the wave spectrum is for each wave component done according to:
(8.13)

where:
(8.14)

Δf: the constant difference between successive frequencies.

(8.15)

An example of wave amplitude time realization by using JONSWAP spectrum analysis is shown in Figure 8.5.
Figure 8.5 2048s zero mean wave amplitude time realization.
If a random phase angle α i, uniformly distributed between 0 and 2π is assigned to each wave component, the wave kinematics are thus represented as
a sum of linear components as shown in Figure 8.6.

Figure 8.6 Frequency domain and time domain representation of long-crested waves.
If “N” is the number of wave components, typically a number power of 2, the sea state at a particular time and location can be represented by: Surface
elevation,
(8.16)
Velocity component in the x-direction,
(8.17)

Velocity component in the z-direction,

(8.18)

Acceleration component in the x-direction,

(8.19)

Acceleration component in the z-direction,

(8.20)

Dynamical pressure,
(8.21)

8.2.3 Nonlinear Wave Theory

One of the most fundamental nonlinear wave theories was developed by Stokes in 1847. It is based on a perturbation solution which is developed as a
power series. Stokes wave theory uses the Airy wave as initial exact solution in order to develop the power series. Additional harmonic waves are then
added to the basic harmonic of the linear wave theory using a perturbation parameter of ε=kA, a product of the wave number k and amplitude A.
(8.22)

with the initial linear solution:

(8.23)

In principle, if Stokes theory is carried to a sufficiently high order, it would be adequate for describing water waves in any depth of water. In practice,
this is only possible for waves in deep water. In shallow water the convective terms become relatively large, the series convergence is slow and erratic,
and a large number of terms is required to achieve a uniform degree of accuracy. Therefore, Stokes wave theory is of direct practical use for waves of
intermediate and deep water.

8.2.4 Current
The most common categories of currents are:
Tidal currents, which are associated with astronomical tides
Circulation currents, which are associated with oceanic-scale circulation patterns
Storm-generated currents
Loop and eddy currents

Figure 8.7 Example Hm0 – Tp contour lines.

The vector sum of these currents is the total current. The variation of current speed and direction with elevations are represented by a current profile.
The total current profile associated with an extreme storm sea-state should be specified for the design. In certain geographic areas, the current force
The total current profile associated with an extreme storm sea-state should be specified for the design. In certain geographic areas, the current force
can be one of the governing design loads. Consequently, selecting the appropriate current profile requires careful consideration.
For the situation where a steady current exists the effects of the bottom boundary layer may be accounted for. Seabed currents in design data are
normally given at a reference height above the seabed. The location of the pipeline in the velocity boundary layer lowers the effective velocity at the
pipeline height, and the mean current velocity over the pipe diameter may be applied in the analysis. According to DNV (2007), this has been used in
the sub-sea pipeline analysis assuming a logarithmic mean velocity profile.
(8.24)

where:
vc(zr) : current velocity at reference measurement height;
zr : reference measurement height (usually 3m.);
z0 : bottom roughness parameter;
D : total external diameter of pipe (including any coating).
The effect of the seabed roughness z0 is taken into account in the above equation. The rougher the seabed, the thicker the boundary layer and the
lower the average velocity over the pipe height.
The total water particle velocity is obtained by adding the velocities from waves and currents together:
(8.25)

8.3 Hydrodynamic Loads

8.3.1 Hydrodynamic Drag and Inertia Forces
A pipeline section exposed to a flow will experience hydrodynamic forces, due to the combined effect of increased flow velocity above the pipe and flow
separation from the pipe surface. Figure 8.8 shows the water particle velocity distribution around the pipe. This section will explain the different
components of the force vector and the expressions that are used to calculate these components.

Figure 8.8 Flow field around pipe.

Pipeline Exposed to Steady Fluid Flow
Fluid drag is associated with velocities due to steady currents superposed by any waves that may be present as shown in Figure 8.9. The drag is mainly
the result of the high pressure in front of the pipe and the low pressure region in the wake behind the pipe. The drag is influenced by the width of the
wake; the expression below gives the transverse drag force component per unit length of the pipeline:

Figure 8.9 Fluid drag and inertia forces acting on a pipe section.
(8.26)

where:
CD : transverse drag coefficient;
ν n : transverse water particle velocity;
ρ : density of seawater;
D : total external diameter of pipe.
Pipeline Exposed to Accelerated Fluid Flow
Waves produce cyclic loadings on the water particles in the water column. These cyclic loads accelerate and decelerate the water particles in both the
horizontal and vertical directions. A pipeline exposed to an accelerated fluid experiences a force proportional to the acceleration, this force is called the
inertia force. The following expression gives the transverse inertia force component per unit length of a pipeline:
Transverse inertia force,
(8.27)

where:
CM : (Ca+1), transverse inertia coefficient;
an : transverse water particle acceleration;
ρ : density of seawater;
D : total external diameter of pipe.
The Complete Morison’s Equation
The formula given above does not take into account that the pipe itself may have a velocity and acceleration. The inline force per unit length of a pipe is
determined using the complete Morison’s equation.
(8.28)

where:
ρ : seawater density;
D : outer diameter;
U : instantaneous (time dependent) flow velocity;
y : in line displacement of the pipe;
CD : drag coefficient;
CM : inertia coefficient; = (Ca +1) where Ca is the added mass coefficient;
δ/δt : differentiation with respect to time.
Drag and Inertia Coefficient Parameter Dependency
In general, the drag and inertia coefficient is given by:
(8.29)

(8.30)

Reynolds number indicates the present flow regime (i.e., laminar or turbulent) and is given as:
(8.31)

where:
U : flow velocity;
L : characteristic length (Diameter for pipelines),;
ν : fluid kinematic viscosity.
The Keulegan-Carpenter number gives information on how the flow separation around cylinders will be for ambient oscillatory planar flow (U=UM
sin((2π/T)t +ε)) and is given as:
(8.32)

where:
UM : flow velocity amplitude;
T : period;
D : diameter;
ε : phase angle;
t : time.
The current flow ratio may be applied to classify the flow regimes:
(8.33)

where:
Uc : typical current velocity normal to pipe;
Uw : significant wave velocity normal to pipe given for each sea state (HS, T p, θw).
Note that α = 0 corresponds to pure oscillatory flow due to waves and α = 1 corresponds to pure (steady) current flow.
The presence of a fixed boundary near the pipe (proximity effect) has a pronounced effect on the mass coefficient. The added mass will increase as the
pipe approaches a fixed boundary, as shown in the below equation,
(8.34)
where:
e /D : gap ratio.
The natural period of the pipe oscillation will increase as the added mass increases.
The roughness number (k/D) has a large influence on the flow separation and therefore also on the drag and mass coefficient. (k=Characteristic cross-
sectional dimension of the roughness on the body surface).
There is a connection between the VIV (Vortex-Induced Vibrations) and the drag force. A crude approximation can be given as:
(8.35)

where:
CD : drag coefficient with VIV;
CD0 : drag coefficient with no VIV;
AZ : cross–flow vibration amplitude.
This formula can be interpreted as saying that there is an apparent projected area D+2AZ due to the oscillating cylinder.

8.3.2 Hydrodynamic Lift Forces

Lift Force Using Constant Lift Coefficients
Lift force is produced in the same way as flow over an airplane. The presence of the seabed introduces an asymmetry between the flow over the top of
the pipe and the flow underneath. This causes slower flow or no flow underneath the pipeline with high pressure and higher velocities over the top with
low pressure, resulting in lift force.
The lift force per unit length of a pipeline can be calculated according to:
Vertical lift force,
(8.36)

where:
CL : lift coefficient for pipe on a surface;
ν n : transverse water particle velocity (perpendicular to the direction of the lift force);
ρ : density of seawater;
D : total external diameter of pipe.
Typically, drag, inertia and lift coefficients of hydrodynamic forces for a pipeline on seabed are empirically chosen as CD =0.7, CM =3.29, and CL =0.9.
Lift Force Using Variable Lift Coefficients
As can be imagined, the hydrodynamic lift coefficient (CL) will vary as a function of the gap that might exist between the pipeline and the seabed. The
lift coefficients according to Fredsøe and Sumer (1997) are given in Figure 8.10. It can be seen from the figure that a significant drop in the lift
coefficient is present even for very small ratios of e/D. This is true both for the shear and the shear-free flow.

Figure 8.10 CL in shear and shear-free flow for 103< Re < 30 x 104.

8.4 Structural Response Analysis

8.4.1 Global Deformation Due to Environmental Loads
A vertical marine riser can be regarded as a hollow ‘beam column’. The difference between a column subjected to lateral loading and a marine riser is
that the riser is subjected to both internal and external hydrostatic pressure as well as axial and lateral loadings.
If the verical riser is simply approximated to a beam column then the governing differential equation for lateral static deflection is,
(8.37)

where EI is the riser bending stiffness, Τ is axial tension in the riser pipe wall, at the riser botton T=T 0, w is the weight per unit length of riser and
contents and F is the lateral force per unit length. The co-ordinate system used is shown in Figure 8.11 with x measured from the bottom of the riser
and positive upwards while y denotes horizontal riser deflection from a vertical through the riser base.

Figure 8.11 Sketch of the riser system.

From the above equations, if the water depth, riser length, submerged weight are known, the touch down tension T 0 can be calculated iteratively.
The complexity of the governing equations for a vertical riser requires that the solutions be obtained through a numberical finite element computation
as described in Section 4.6. A typical beam element and riser idealization is given in Figure 8.12.

Figure 8.12 Element defintion for a finite element idealization of a vertical riser.
The angular inclination of any element to the global axis (X, Y) is very vital to the estimation of the nodal parameters relative the global axis.
(8.38)
where for each element, we have

The angle is used to compute the transformation matrix between the fixed seabed and each element, the essence being to convert every force from the
local coordinate system (x, y) to the global system of coordinate (X, Y). The transformation matrix for each node therefore will be (show the beam
element)
(8.39)

and for the entire element we have,

(8.40)

Note that the transformation matrix is same for both nodes as the angle does not vary within the element and the zeros here represent 2x2 null
matrices.

8.4.2 Mass Matrices

As noted above, the finite beam element could have its mass lumped at the nodes or evenly distributed using an interpolation polynomial.
Using the interpolation polynomial for a two-dimensional system given in
(8.41)

the mass of each element is

(8.42)

(8.43)

The structural mass of each element is computed by expanding the above, taking into consideration the fact that the mass will always be unidirectional
at each node,
(8.44)

A and Aint = cross-sectional and internal area of the riser respectively

ms and mint,fluid =structural mass of the riser and mass of internal fluid respectively
ρs and ρfluid= density of steel and internal fluid respectively l= length of each element
The internal fluid mass matrix will be similar with the structural mass matrix; the only difference will be in the cross-sectional area and density
(8.45)

The matrix for the added mass is related to the external fluid and external area since the added mass effect is due to the interaction between the riser
and the external fluid giving the riser additional mass, the added mass matrix is given as
(8.46)

(8.47)
 mak, [M]k=added mass and total mass matrix for each element
ρext,fluid= density of external fluid
Aext = external area
This individual element mass matrix will first be transformed into the global reference frame and then assembled using matrix concatenation method
thus
(8.48)

Hence the global mass matrix on concatenation will be

(8.49)

8.4.3 Stiffness Matrices

The stiffness matrices are derived by noting that the variation of the total potential energy at a boundary is equal to the work of external force on the
virtual displacement. Mathematically we have
(8.50)

where the first term of the left side is the total body force in each element; the second term of the left side is total external force acting on the boundary
of the element; the right-side term is the change in the potential energy (virtual work) of the system.
For a linear system
(8.51)

This equation can be expressed in the form

(8.52)

where
{f} and {a} = force and displacement vectors and
[K] = the global stiffness matrix. In this form
(8.53)

where
[B] = First derivative of the shape function of each element
[D] = matrix of the modulus of elasticity
P = external load on the riser
β = the body force per unit volume
Substituting for δE, δu and τ in above equation as below,

Using the same interpolation function as in the mass matrix above we have
(8.54)
Substituting into equation 8.54 above, we get
(8.55)

To reflect the number of degree of freedom of each element, the elastic stiffness matrix is written as
(8.56)

Also, the geometric stiffness matrix which relates deformation in the lateral direction is given as
(8.57)

(8.58)

This is the stiffness matrix of each element in the local frame; it must be transformed to the global reference frame before it can be concatenated to the
global stiffness matrix. The procedure is the same as that for the mass matrix; the difference is that the stiffness matrix above is used instead of the
mass matrix.

8.4.4 Damping Matrices

The structural acceleration and velocity of the riser pipe as a result of wave and current forces are continually attenuated (damped) as the motion
progresses. Effectively, damping has to do with the dissipation of the energy of a vibrating system giving rise to decay in the amplitude of the vibration
circle. The effect of damping is factored into the dynamic analysis of a system using (damping) ratios which are a measure of the energy lost per cycle
to the total energy of the undammed system. In general, modeling the phenomenon of damping is quite involved but the current practice follows the
Raleigh method which assumes that damping is proportional to the mass and stiffness of the entire system. Mathematically, the damping matrix is
expressed as
(8.59)

The damping coefficients υ1 and υ2 are often expressed in terms of the damping ratio and the first and second mode natural frequency
(8.60)

(8.61)

The damping ratio ξ is often taken as 5% for the first two modes in analysis.
ξ1 and ξ2= damping ratios
ω n1 and ω n2 = first and second mode natural frequencies

8.4.5 Riser Deformation

Knowledge of the deformation will give information on the stress-strain condition of that location in the riser pipe. The deformation is gotten from the
solution of the dynamic equation
(8.62)

Where [M], [C], [K] are the mass, damping and stiffness matrices and q, , are the structural displacement, velocity and acceleration respectively.
While F(t) is the Morison’s force above, i.e., the global force vector.
(8.63)
The solution to this set of equation (in time domain) is an iterative process solved using numerical integration method e.g., fourth-order Runge Kutta
method, Wilson-theta algorithm, Newmark Beta method, etc., to get the structural acceleration, velocity and displacement.

8.5 Vortex Induced Vibrations

8.5.1 Introduction
Offshore structures exposed to current and waves may experience vortex induced vibrations (VIV). The general requirement for VIV occurrence is
that the vortex shedding frequency is close to an eigenfrequency of the structure, leading to a fully coupled fluid-structure interaction.
When vortices shed in the wake of a circular cylinder, the forcing frequency in the lift direction occurs at the frequency of vortex shedding. This is
expected since one vortex or group of vortices sheds to one side of the cylinder once per cycle and one vortex or group of vortices sheds to the other
side of the cylinder once per cycle. In the drag direction, however, all vortices are shed downstream, as shown in Figure 8.13. Therefore, the drag force
associated with vortex shedding occurs at twice the frequency of the lift force since both vortices are shed to the same side of the cylinder in the drag
direction. When the vortex shedding frequency is close to the natural frequency of the elastically mounted cylinder, the response motions resulting
from these fluctuating forces are in relatively large cross-flow motion (perpendicular to the direction of fluid flow). In-line vibrations (in the direction of
the fluid flow) are not as large as cross-flow motions and are typically less than 0.5 diameters. For this reason, many studies have simplified
experiments by only considering cross-flow vibrations of the structure.

Figure 8.13 Vibrations of a cylinder submitted to vortex shedding.

VIV is normally considered to be self-limiting type of response. This is easily understood if one realizes that the dynamic force that drives the
vibrations depends on the response amplitude. If the amplitude exceeds a certain level, the vortex shedding process will no longer transfer energy
from the fluid flow to the structure, but reverse the energy transfer process and hence lead to damping.
VIV may occur in typical slender structures such as lines, risers, free spanning pipelines and conductors. The pipeline spans are unsupported pipe
sections above the seabed. The pipelines tend to form spans rather than follow the topographical features of the seabed if the seabed is very rough.
Figure 8.14 illustrates typical pipeline spans formed on a seabed, in which (a) shows a single span and (b) shows multiple spans. The response will then
have a frequency close to an eigenfrequency for the structure, and in many cases be dominated by the associated mode shape.
Figure 8.14 Typical pipeline spans.
A long, slender, vertical riser has a large number of natural frequencies as shown in Figure 8.15. The riser will oscillate when the vortex shedding
frequency is close to any eigenfrequency. VIVs of these structures in the field can be very complex due to the high number of structural natural modes
and sheared ocean currents.

Figure 8.15 Vertical riser eigenmodes.

VIV will contribute to fatigue damage accumulation and have a critical influence on the safety and integrity of the offshore structures. Meanwhile the
oscillations will also lead to an increase of drag forces, which means that the static deformation of offshore structures will be influenced by VIV.

8.5.2 Analysis of Vortex-Induced Vibration

Vortex shedding from a smooth, circular cylinder in a steady flow is a function of Reynolds number. The Reynolds number is based on free stream
velocity V and cylinder diameter D,
(8.64)

where v is the kinematic viscosity of the fluid. The major Reynolds number regimes of vortex shedding from a stationary circular cylinder, summarized
in Figure 8.16.
Figure 8.16 Flow regimes for smooth circular cylinders.
At very low Reynolds numbers, below about Re=5, the fluid flow follows the cylinder contours. In the range 5≤Re≤45, the flow separates from the back
of the cylinder and a symmetric pair of vortices is formed in the near wake. As Reynolds number is further increased, the wake become unstable and
one of the vortices breaks away. A laminar periodic wake of staggered vortices of opposite sign is formed. The Reynolds number range
300<Re<1.5x105 is called subcritical. In this range, the laminar boundary layers separate at about 80 degrees aft of nose of the cylinder and the
vortex shedding is strong and periodic. In the transitional range, 1.5x105<Re<3.5x106, the cylinder boundary layer becomes turbulent, the separation
points move aft to 140 degrees, and the cylinder drag coefficient drops to 0.3. In the supercritical Reynolds number range, Re>3.5x106, regular vortex
shedding is reestablished with a turbulent cylinder boundary layer.
The most relevant Reynolds numbers for marine pipelines and risers subjected to current are in the sub-critical region from 103 and up to 2x105. This
is due to small diameters, usually in the order of 0.5 meters, and generally low current velocities. But for some high current cases the flow is in the
critical or supercritical regime.
For a stationary cylinder, it is observed that by varying the current velocity V, the vortex shedding frequency fs is proportional to U/D. Strouhal
number S t is introduced as a dimensionless constant between the predominant frequency of vortex shedding and the free stream velocity divided by
the cylinder diameter,
(8.65)

This number is by nature a stochastic value with a mean and a variation. Both the mean and the variation are functions of the Reynolds number and
the roughness of the cylinder. A good approximation for Strouhal number in the subcritical flow regime is 0.2. The Strouhal number as a function of the
Reynolds number for a circular cylinder is shown in Figure 8.17. The shedding of vortices behind the cylinder alternating from one side to the other
generates a time variable pressure over the cylinder. Integrated over the surface one may find time variable force components both in the drag and
the lift direction. The drag force will consist of a mean value and an oscillating component with peak frequency at twice the vortex shedding frequency.
The lift force will have zero mean and the peak oscillatory component at the vortex shedding frequency.
Figure 8.17 Relation between the Strouhal Number and Reynolds Number.

8.5.3 Harmonic Model

Because vortex shedding is a more or less sinusoidal process, it is reasonable to model the vortex shedding transverse force imposed on a circular
cylinder as harmonic in time at the shedding frequency.
(8.66)

where
ρ : fluid density (must be in mass units);
V : free stream velocity;
D : cylinder diameter;
CL : lift coefficient, dimensionless;
ω s : circular vortex shedding frequency;
F L : lift force (perpendicular to the mean flow) per unit length of cylinder;
This lift force is applied to a spring-mounted, damped, rigid cylinder, shown in Figure 8.18. The cylinder is restrained to move perpendicular to the
flow. The equation of the motion of the cylinder is
(8.67)

where
u : displacement of the cylinder in the vertical plane from its quilibrium position;
m : mass per unit length of the cylinder, including added mass;
ξ : structural damping factor;
k : spring constant, force/unit displacement;
ω s=(k/m)1/2=2πfn : circular natural frequency of cylinder;

Figure 8.18 A schematic model of a 2-D freely oscillating rigid cylinder.

Solutions to this linear equation are found by postulating a sinusoidal steady-state response with amplitude A, frequency ω s and phase φ,
(8.68)

and substituting this equation into. The result is (Thompson 1988)

(8.69)

where the phase angle is defined by

(8.70)

The phase angle shifts by 180 degrees as the cylinder passed through fs=fn.
The response is largest when the shedding frequency approximately equals the cylinder natural frequency, fs=fn, a condition called resonance. Further,
the resonance vibration amplitude is
(8.71)

The right-hand form of the above equation is obtained by incorporating the Strouhal Number St and the natural frequency, fn=(1/2π)(k/m)1/2. The
resonant amplitude decreases with increasing reduced damping δr, which is defined as the mass ratio times the structural damping factor,
(8.72)

where, m is the mass per unit length including added mass, ξ is the damping factor usually measured in still fluid, ρ is the density of the surrounding
fluid, and D is the cylinder external diameter.
Models have been developed for vortex-induced cylinders in the subcritical Reynolds number range amplitude dependence of lift (Basu and Vickery,
1983; Blevins and Burton, 1976). In the Blevins and Burton model, the experimental response data are fitted to a three-term polynormial using to
obtain CL as a function of A/D,
(8.73)

The coefficients in the curve fit are

(8.74)

As cylinder vibration at resonance with vortex shedding increases in amplitude from zero, the cylinder motion organizes the wake and spanwise
correlation increases. The vortex strength increases, and the lift coefficient with it. As the cylinder amplitude increases beyond approximately one-half
diameter, the cylinder begins to outrun the shedding vortices and the lift coefficient diminishes. This is reflected by c<0 in the polynomial fit.

8.5.4 Wake Oscillator Model

The nature of self-excited vortex shedding suggests that the fluid behavior might be modeled by a simple, nonlinear, self-excited oscillator. The basic
assumptions of the model are as follows:
1. Inviscid flow provides a good approximation for the flow field outside the near wake.
2. There exists a well-formed, two-dimensional vortex street with a well-defined shedding frequency.
3. The force exerted on the cylinder by the flow depends on the velocity and acceleration of the flow relative to the cylinder.
An elastically supported rigid circular cylinder with diameter D is free to move in transverse direction and subjected to a steady uniform flow with an
upstream velocity V as shown in Figure 8.19. The governing equation of the structure in 2-DOF can be expressed as
(8.75)

where ms, c s, and k represents the structural mass, damping and spring stiffness respectively; ma is the potential flow added mass; c f is referred to as
the fluid added damping; F V represents the fluctuating vortex force caused only by vorticity in the wake. The vortex force can be obtained by
introducing empirical parameters as discussed by Xu and Wei, and the vortex force is assumed and expressed as the following form,
(8.76)

where the vortex force F v is expressed by vortex circulation Г which is associated with the evolution of nascent vortex interacted with the separation
point and convection velocity, α is the non-dimensional parameter.
Figure 8.19 A schematic model of a spring mounted rigid cylinder undergoing two-dimensional VIV.
For the sake of simplicity, a dimensionless wake variable q which is assigned as Γ= UDq which is assumed to satisfy the van der Pol equation with a
forced term relying on the coupled action between structure and fluid. It can be expressed as
(8.77)

where ε and A are empirical parameters; ω st is vortex shedding angular frequency, which follows the Strouhal relation as ω st = 2πSt U / D, where St
denotes the Strouhal number.
This model can be easily extended to three-dimensional form, to further study the VIV characteristics of slender structures, such as marine riser,
mooring cables and subsea pipeline spans.
A vertical riser model can be idealized as an Euler-Bernoulli beam with top tensioned. A Cartesian coordinate system with its origin at the bottom of the
riser has been used, in which the z-axis is vertical, and the transverse displacement u is perpendicular to the incoming flow V(z). The governing
equation of the riser in a general dimensional form is described as,
(8.78)

where u(z) represents the transverse displacement along the pipe; m is the mass of the riser per unit length, c is the damping characteristic; EI is the
flexural stiffness and T(z) is the effective tension. The external fluid force F V(z,t) is the fluctuating vortex force. Herein, the three-dimensional effect
along the pipe is assumed to be ignorable, thus the fluid force acting on every cross section of riser can be expressed as,
(8.79)

where α is empirical parameter; V(z) is current velocity profile along the riser. The wake variable q represents the displacement of circulation, which
satisfies a forced Rayleigh equation. The 2-D wake oscillator can be extended to model the distribution fluctuating forces on the riser, expressed as,
(8.80)

Note that all direct coupling of the wake variables in the Z-direction is neglected here: a diffusion coupling was analyzed by Facchinetti (2002) that it
plays a negligible role when the cylinder motion is significant, as will be disregarded here.

Example 8.1 As shown in Figure 8.20, an experiment of a vertical tension riser exposed to a stepped current was carried out at Delft Hydraulics in
the Delta Flume by Chaplin et al. (2005). The model riser was 13.12 m long and had a diameter of 28 mm, giving a maximum Reynolds number of 28
000. The mass ratio, calculated as the mass of the model divided by the mass of fluid displaced, was 3 and the structural damping measured in free
decay tests in air was around 0.3%. The effect of a stepped current was achieved by mounting the model on the towing carriage with the upper 55% of
the model inside a chamber sealed at the top and with its other end open just below the water surface of the flume. By evacuating the air from the
chamber, the upper 55% remained in still water while the lower 45% was exposed to a current. The effective data is given in the following table. Use
the wake oscillator mode to calculate the dynamic response of the vertical riser.
Figure 8.20 Layout of the experiments (left); the vacuum tank (on the top of the structure) is mounted on the carriage (right).
Solution: According to the model analysis method in Section 5.2.2, the beam is divided into 130 elements. Table 8.2 shows the model analysis results
compared with experimental test under different top tensions. Wherein the added mass is considered, the added mass coefficient is taken as Ca=1.0.
Figure 8.21 shows the modal shape from 1 st mode to 4th mode. Due to the gravity effect, the tension along the pipe is changed, and the amplitude of the
modal on top is greater.

Figure 8.21 The modal shape up to transverse forth mode.

Table 8.1 Parameters of model data in Chaplin’s test.
Parameter Unit Value
Outer Diameter, D mm 28
Length, L m 13.12
L/D ratio, R - 470
Top Tension, T N 380-1900
Mass per unit, M kg/m 1.85
Submerged Mass, W N/m 12.1
Mass ratio, m* - 3
Structural damping ratio, ξ % 0.33
Towing Velocity, V m/s 0.1-1.0
Reynolds number, Re - 2500-25 000
Table 8.2 The modal frequencies from modal analysis.
Mode Top Tension = 813N Top Tension = 1185N Top Tension = 1546N
Num. Experimental Numerical Experimental Numerical Experimental result Numerical result
result result result result
Hz Hz Hz Hz Hz Hz
1st mode 0.633 0.657 0.782 0.807 0.899 0.930
2nd model 1.329 1.319 1.423 1.618 1.815 1.862
3rd model 1.954 1.990 2.394 2.436 2.741 2.802
4th model 2.625 2.675 3.203 3.266 3.614 3.751
5th model 3.312 3.377 4.024 4.110 4.561 4.712
6th model 4.028 4.102 4.875 4.972 5.525 5.690
Figure 8.22 shows the transverse displacement response of riser under carriage speed of 0.85 m/s and top tension of 939 N. The instantaneous
deflected shapes of experimental measurements and numerical predictions are shown in Figure 8.22(a) and (b), at intervals of 0.015s through ten
periods of the transverse oscillation. The color maps show the evolution in time of the displacements distribution along the axis of the riser predicted by
the present model and a standing wave dominates the whole riser predicted as seen in Figure 8.22(c).

Figure 8.22 Vibration response of the riser. The cross-flow instantaneous deflected shapes of (a) experimental measurements and (b) present model,
and (c) evolution of cross-flow displacement with time and space. Carriage speed=0.85m/s; top tension 939N.

Exercises
8.1 A simple, shallow water is defined when the ratio of water depth to wavelength satisfies d / λ < 0.05. In this case, show that the wave celerity is
given approximately by . If λ = 156m, what is the maximum water depth for which this relationship holds?
8.2 Calculate the vortex shedding frequency from a marine riser 0.1m in diameter in a current at 3m/sec. Please note that the Reynolds number must
be calculated first.
8.3 For vortex induced vibration harmonic model, starting with Equations (8.66)-(8.68), derive Equation (8.69) through (8.72).
8.4 Calculate the displacement of a riser as parameters given below. On the top, the riser has 20m offset. Sea linear current profile at the surface the
velocity is 0.2m/s and at the bottom is 0m/s.
 Riser length 457m Density of water 1025 kg/m3
Outer diameter of riser 0.533m Density of steel 7850 kg/m3
Inner diameter of risr 0.508m Density of internal fluid 300 kg/m3
Modulus of elasticity, E 210 GPa Water depth 470 m

References
1. Clough, R.W. and Penzien, J., Dynamics of Structures, McGraw-Hill, 1995.
2. Aguiar, L.l., Almeida, C.A., Paulino, G.H., Dynamic analysis of risers using a novel multilayered pipe beam element model. Elsevier Ltd., 44, 211-231,
2015.
3. Seo, M., Kim, Y., Park, D., Effect of internal sloshing on added resistance of ship. Elsevier Ltd., 29, 1, 13-26, 2017.
4. Lee, H. and Roh, M., Review of the multibody dynamics in the applications of ships and offshore structures. Elsevier Ltd., 167, 65-76, 2018.
5. Yang, H., Xiao, F., Xu, P., Parametric Instability Prediction in a Top-Tensioned Riser in Irregular Waves. Ocean Eng., 70, 39-50, 2013.
6. Faltinsen, O.M., Hydrodynamics of marine and offshore structures. Elsevier Ltd., 26, 6, 835-847, 2015.
7. Hirdaris, S.E., Bai, W., Dessi, D., Loads for use in the design of ships and offshore structures. Elsevier Ltd., 78, 131-174, 2014.
8. Molin, B., Hydrodynamic modeling of perforated structures. Elsevier Ltd., 33, 1, 1-11, 2010.
9. Chen, Jing-pu, Zhu, Numerical Simulations Of Wave-Induced Ship Motions In Time Domain By A Rankine Panel Method. J. Hydrodynam., 22, 5,
408-415, 2010.
10. Sarpkaya, T., Analytical study of separated flow about circular cylinders. Phys. Fluids, 12, 279-280, 1969.
11. Sarpkaya, T. and Schoaff, R.L., Inviscid model of two-dimensional vortex shedding by a circular cylinder. AIAA J., 17, 1193-1200, 1979.
12. Facchinetti, M.L., de Langre, E., Biolley, F., Coupling of structure and wake oscillators in vortex-induced vibrations. J. Fluids Struct., 19, 123-140,
2004.
13. Xu, B. and Wei, Q., Using vortex strength wake oscillator in modelling of vortex induced vibrations in two degrees of freedom. Eur. J. Mech. B Fluid,
48, 165-173, 2014.
14. Chaplin, J.R., Bearman, P.W., Cheng, Y. et al., Blind predictions of laboratory measurements of vortex-induced vibrations of a tension riser. J.
Fluids Struct., 21, 25-40, 2005.
Answers to Exercises
Chapter 1
1.1 What is the purpose of structural dynamics analysis? What are the characteristics of structural dynamics?
The purpose of structural dynamics analysis:
To study the dynamic characteristics of structural systems (mainly the period, frequency, mode shape and damping characteristics). To determine the
dynamic response of the structure under dynamic load (including internal force, strain, displacement, velocity, acceleration, etc.) The fundamental
purpose is to provide a solid theoretical basis for improving the safety and reliability of engineering structure system in dynamic environment.
Characteristics of structural dynamic problems:
First, the dynamic problem has the characteristics of changing with time. Because the dynamic load changes with time, it is necessary to solve a series
of solutions at all time points for calculating the dynamic response of the structure, which is more complicated and consumes more calculation time than
the static problem.
Secondly, the influence of inertia force must be considered in dynamic problems. Compared with the static problem, since the displacement of the
structure changes rapidly with time in the dynamic response, and the inevitable acceleration produces the inverse inertial force, the inertial force has
an important impact on the structural response.
1.2 Express the types and characteristics of dynamic loads.
Types of dynamic loads:
According to whether the load is pre-deterministic, the dynamic load can be divided into prescribed loads and random loads. According to the law of
load changing with time, dynamic deterministic loads can be divided into periodic loads and nonperiodic loads: for periodic loads, it can be divided into
simple harmonic periodic loads and non-harmonic periodic loads, and nonperiodic load can also be divided into impulsive loads and other regular loads;
random loads can be divided into wind loads, seismic force loads and other nondeterministic loads.

Characteristics of dynamic loads:

Prescribed loads: The law of load variation with time has been predetermined and is a completely known process.
Random loads: The law of load changing with time is uncertain in beforehand, and it is a stochastic process.
Periodic loads: Load changes harmonically periodically with time. Nonperiodic loads: Load changes periodically with time.
Impulsive loads: Load amplitude increases or decreases sharply in a very short time.
Arbitrary dynamic loads: The magnitude, direction and action point of the loads are complex and random, so it is difficult to express the loads by
analytic function.
1.3 What are the procedures of structural discretization? What are their characteristics?
The procedures of structural discretization include lumped mass method, generalized coordinate method and finite element method, etc.
First, the lumped mass method. Concentrating the mass of the actual structure on some geometric points according to certain rules, the members of
the structure except these points are massless. The infinite degree of freedom system is transformed into a finite degree of freedom system.
Second, the generalized coordinate method. Using the series expansion method, we use generalized coordinates to represent the size of shape functions.
Generalized coordinates themselves are not real physical quantities, only when n terms are superposed can they be real physical quantities.
Third, the finite element method. The displacement of each point on the structure is expressed by the displacement of structural nodes. In this
method, the characteristics of lumped mass method and generalized coordinate method are synthesized. Similar to the generalized coordinate method,
the concept of shape function is adopted, but the interpolation of segments is used, so the expression of shape function is relatively simple. Similar to
the lumped mass method, the generalized coordinates of finite element method also use real physical quantities and have direct and intuitive
advantages.
Chapter 2
2.1 The basic components in structural dynamic analysis are: mass elements, elastic elements, and damping elements.
Damping mechanism: The physical meaning of damping is the attenuation of force, or the energy dissipation of an object during motion. In layman’s
terms, it is to stop the object from continuing to move. When an object vibrates by an external force, a reaction force that attenuates the external force
is generated, which is called a damping force (or damping force).
Damping mode: Friction damping, radiation damping, etc.
2.2 D’Alembert principle: At any moment of the system’s motion, if the main force and the restraining reaction force of the actual action structure are
added, plus the imaginary inertial force, the system will be in an imaginary equilibrium state at this moment.

2.3

Solution: In the tangential direction of the mass, we have the following forces,
Spring force
Inertia force
External force
The virtual displacement can be defined as shown in Figure,

According to the virtual displacement principle, the principle of virtual work can be derived as

Substituting the virtual displacement equation into Equation xx and cancel out δφ1 the following equation of motion is obtained,

Chapter 3
3.1 In the undamped and damped free vibration, what factors are associated with the natural frequencies and dynamic responses?
In undamped free vibration, the natural frequencies are related to the stiffness K and the mass m of the structure, and the dynamic responses are
related to initial displacement, initial velocity, stiffness K and the mass m of the structure.
In damped free vibration, the natural frequencies are related to the stiffness K, mass m and damping ratio of the structure, and the dynamic responses
are related to initial displacement, initial velocity, stiffness K, mass m and damping ratio of the structure.
3.2 As the dynamic system shown in the picture, using jack to make m produce lateral displacement of 30 mm, then the jack is suddenly released the
system begins to vibrate freely. The lateral vibration was measured 10mm after 8 periods. Determine: (1) Damping ratio and damping coefficient of
the structure (2) Amplitudes after 20 weeks of vibration (3) If the lateral displacement is 15mm after 8 periods of vibration, and the damping and
mass remain constant, the stiffness must be set at what value?
1) damping ratio

2) At j = 20

3)

3.3 Single degree of freedom system, known m = 100kg, k = 6000N / m. Free vibration begins when initial condition u(0), 20mm, .
Determine: (1) The natural frequency and period of the system (2) The displacement and velocity when t = 1.0s for two cases of c = 0, c = 200N.s / m.
1)

2) At c = 0,

At t = 1.0s, u(1) = 0.02 cos 7.75 × 1 + 0.013 sin 7.75 × 1 = 0.015m

At c = 200N·s/m,
At t = 1.0s,

3.4 The basic power system shown here has the following characteristics: m = 20kg, k = 3 ×104 N / m. If the system is subjected to a resonance
harmonic load (ω = ω n) and the initial state is static, specify the value of the dynamic magnification factor Rd(t) after five cycles (ωt = 10π), (1)
Undamped system (2) ζ = 0.02; (3) ζ = 0.05.
1) Rd → ∞

2) At

3) At

3.5 What are the methods for calculating damping ratio?

(1) Logarithmic Decay Rate Method
Damping ratio is measured by the ratio of two amplitudes separated by one period. When the structural damping is very small and the free vibration
attenuates is slow, the structural damping ratio can be calculated by the ratio of amplitudes separated by several periods, and better accuracy can be
obtained.
(2) Bandwidth Method
From the amplitude-frequency characteristic curve of the dynamic amplification coefficient, it can be seen that the shape of the curve is completely
controlled by the damping ratio, so the damping ratio of the system can be determined by the width of the half power point. Forced vibration test is
used in this method. It can be used not only in single-degree of freedom system but also in multi-degree of freedom system. It requires sparse
resonance frequency for multi-degree of freedom system.
(3) Resonance Amplification Method
The amplification coefficient of any given frequency is the ratio of the response amplitude of the frequency to the zero frequency response amplitude.
When resonance occurs, the dynamic amplification coefficient can be obtained and then the damping ratio can be obtained.

3.6 What are the types of impact loads, and what are the basic ideas and procedures for their dynamic responses?
The types of impact loads include sine-wave impulse load, rectangular impulse load, and triangular impulse load.
The response to impact load is usually carried out in two stages: the first stage is forced vibration under load, the special solution under this load is
obtained, and then the full solution satisfying the initial conditions is obtained by considering the initial displacement and velocity; the second stage is
free vibration without load, and the initial conditions are displacement u(t) and velocity at the end of the first stage of impact load.
3.7 Use MATLAB language to draw up a procedure that solves structural dynamic response (displacement response and acceleration response) with
Duhamel integral.
Taking the time interval of 0.02s, using MATLAB programming analysis, the program is as follows:

fid=fopen(‘elcentr.txt’,’r’); dzhbo=fscanf(fid, ‘% f’,401);

% response analysis of structure-elastic time history analysis method

% structure parameter

k=3e+6; % initial stiffness

m=2e+4; % mass

% earthquake parameter

ct=1.4; % wilson cita method

dt=0.02;

xs=200/max(abs(dzhbo)); % adjust acceleration amplitude

ag=dzhbo*0.01*xs; % earthquake wave,400 epoches,dt=0.02s,unit-m?s^2

ndzh=400;

ag1=ag(1:ndzh);

ag2=ag(2:ndzh+1);

agtao=ct*(ag2-ag1);

% initial value

wyi1=0;

sdu1=0;

jsdu1=0;
wyimt=0;

sdumt=0;

jsdumt=0;

[x,d]=eig(k,m);

w=sqrt(d);

c0=2*m*w*0.05;

for i=1:ndzh

kxin=k+(3/(ct*dt))*c0+(6/(ct*ct*dt*dt))*m;

dpxin=-m*agtao(i)+m*(6/(ct*dt)*sdu1+3*jsdu1)+c0*(3*sdu1+ct*dt/2*jsdu1);

dxtao=kxin\dpxin;

dtjsdu=6*dxtao/(ct*(ct^2*dt^2))-6*sdu1/(ct*ct*dt)-(3/ct)*jsdu1;

jsdu=jsdu1+dtjsdu;

dtsdu=(dt/2)*(jsdu+jsdu1);

sdu=sdu1+dtsdu;

dtwyi=dt*sdu1+(1/3)*dt^2*jsdu1+(dt^2/6)*jsdu;

wyi=wyi1+dtwyi;

jsdu=-m\(m*ag2(i)+c0*sdu+k*wyi);% % justify acceleration

wyi1=wyi;

sdu1=sdu;

jsdu1=jsdu;

wyimt=[wyimt wyi*1000]; % unit mm

sdumt=[sdumt sdu]; % unit m/s

jsdumt=[jsdumt jsdu]; % unit m/s^2

end

t=0:dt:ndzh*dt;

subplot(2,2,1);

plot(t,wyimt(1,:),’r-’);

subplot(2,2,2);

plot(t,jsdumt(1,:), ‘r-’);

3.8 What are the basic steps in the frequency domain analysis of structural dynamic responses?
(1) Fourier transform of external load P (t) to P (ω);
(2) According to Fourier spectrum P (ω) and complex frequency response function H (iω), the frequency domain solution U (ω) of structural response
is obtained.
(3) Using inverse Fourier transform, the time domain solution U (t) is obtained from U (ω).
3.9 What are the structural dynamic tests? What are the common excitation and sensing devices?
(1) Common dynamic tests include forced vibration test (including sinusoidal loading test), shaking table test, pseudo-dynamic test, impact test
(percussion test and impact test), free vibration test and field dynamic simulation test.
(2) Excitation equipment: exciter, shaking table, impact hammer.
(3) Sensing equipment: strain sensor, acceleration sensor, displacement sensor, speed sensor.
3.10 A reciprocating machine weighs 2 × 104 N, it is known that when the machine operates at a speed of 40Hz, a vertical harmonic force of 1000N
amplitude is generated. In order to limit the vibration of the machine to the building, the four corners of its rectangular bottom are supported by a
spring. What is the spring stiffness required to limit all the harmonic forces that pass from the machine to the building to the 100N?
In this case, the force transmissibility is
From formula (3.197), because of damping ratio ζ=0,

then

Thus, the total spring stiffness is obtained:

Therefore, the stiffness of each spring in the four supporting springs is:

It is also noted that the static displacement of the support springs due to the weight of the machine is:

Chapter 4
4.1 The mass matrix and stiffness matrix of the structure:

The characteristic equation of the equation of motion:

Frequency equation:

Obtain:

Assume φ 3n = 1

Then the vibration mode equation is:

Substituting separately:

Obtain:

Verify orthogonality:
4.2 All the mass of the rod can be assumed to be concentrated at point O, so the following equation can be established according to the equilibrium
relationship:

So we can obtain:

4.3 The system has two degrees of freedom, so the flexibility method is adopted.:

hence: , ,

Dynamic equation :

where:

4.4 The system has three degrees of freedom.

(1) The natural frequency and vibration mode of the calculation system are easily obtained by the flexibility method:
Mass matrix:

Frequency equation:

We can obtain:

1. First mode: X11 = 1, :

obtain:

so:

(2) Establishment of regular coordinate transform:

(3) Generalized mass:

(4) Generalized load:

(5) Canonical coordinates:

Since the load is a harmonic load, the steady-state solution of the system can be obtained.:

Thus, it can be obtained.:

(6) Particle displacement:

The vertical amplitude of particle 1 is: A1 = 0.196mm

The horizontal amplitude of particle 1 is: A2 =–1.240mm
The horizontal amplitude of particle 2 is: A3 = –1.02mm
4.5 If the two are negative real numbers or complex numbers, then ω n1 , ω n2 should be imaginary numbers or complex numbers, then u1 , u2
which obtained in the substitution formula (4.9) will be hyperbolic functions containing time, which obviously contradicts the stability of the system
under restoring force at the equilibrium position (t = 0). From this, we can find two real roots of ω n, namely .
4.6
Chapter 5
5.1

Equation of motion

Boundary conditions of the uniform

Chapter 6
6.1 Stationary White Noise Model; Filtered White Noise Model; Improved Filtered White Noise Model; Housner model; Multi-Peak Power Spectrum
Model; Stabilized Stochastic Process Model.
6.2 The time-domain analysis of the structural stochastic response is to solve an analytic formula of the correlation function of dynamic response of the
system.
The frequency-domain analysis of the structural stochastic response is to solve the power spectral density function of the dynamic response of the
system, and then calculate the mean square response of the system by the integral operation.
6.3 Perturbation Method: In addition to the existence of a main vibration, there are small additional vibrations in a vibration system. These small
additional vibrations are like a micro-disturbance and so called perturbation. The perturbation method is the method using perturbation to solve
nonlinear vibration systems, which is also called small parameter method in mathematics. The perturbation method in nonlinear stochastic vibration is
basically the same as that used in deterministic analysis.
Equivalent Linearization Method: The basic idea is that a nonlinear equation can be approximated by an equivalent linear equation, where the
equivalent parameter is determined by the condition where the mean square of the difference between the two equations is the smallest. Since these
parameters (frequency, stiffness, or damping) are related to the factors to be determined, it is generally done by iterative method.
Chapter 7
7.2 Vibration Mode Decomposition Method
Bottom Shearing Force Method

7.3 What types can structural vibration control be divided into? Explain the ideas of various control methods.
The modal analysis transforms the physical coordinates in the motion equations of the linear dynamical system into modal coordinates and decouples
the equations into a set of independent equations describing modal and modal parameters in order to obtain the modal parameters of the system. It is
the basis of dynamic design, vibration isolation, vibration control, health monitoring, and fault diagnosis. The actual vibration of buildings, machines,
aerospace vehicles, ships, cars, and other complex structures are different and modal analysis provides an important way to study the actual structural
vibrations.
7.4 What are the methods of dynamic damage identification? And what are the basic ideas of them?
In structural damage detection, we mainly need to solve the following problems:
Whether the structure is damaged. Judgment of Structural Damage Position. The severity of structural damage. The effect of structural
damage on the structural performance, or the prediction of the residual life of the structure. The method of damage diagnosis mainly includes dynamic
identification method, static identification method, neural network method, statistical analysis method, and so on.
7.5 In dynamic analysis, what is physical nonlinearity? And what is geometric nonlinearity?
The physical nonlinearity in structural dynamic analysis mainly refers to the nonlinear problem of structural restoring force.
The geometric nonlinear problem is: If the deformation of the bar is large, you cannot analyze the balance of the force on the basis of the original
geometry, but should analyze on the basis of the deformed geometry. In this way, there is a nonlinear relationship between force and deformation
which is called geometric nonlinearity.
7.6 What is the basic idea of sub-structure law? When is it often applied?
Sub-structure method is usually dividing the structure into finite sub-structures and through the analysis of a limited sub-structure, the analysis of the
entire structure is achieved. In the analysis process, a set of elements need to be condensed into a single matrix and this single matrix element is called
a supercell. The use of sub-structure is mainly to save time and allows solving large-scale problems on the basis of a limited number of computer
equipment resources. Its effectiveness is reflected in nonlinear analysis and analysis with a large number of repetitive geometries. In the nonlinear
analysis, the linear part of the model can be constructed as a sub-structure so that the unit matrix of this part does not have to be repeated in the
nonlinear iteration process. In models with repetitive geometries, you can generate supercells for duplicate parts and then copy them to different
locations, which can save a lot of machine time. If the disk space a problem requires is too large, the user can analyze the problem through sub-
structures. The sub-structure method is usually used for finite element analysis or damage identification of large structures and complex fine
structures.
Index
Aircraft technology application, 14–15
Amplification coefficient, 87, 89
Arbitrary load response, 89–97
Duhamel integral (time-domain analysis), 89–95
Fourier transform (frequency-domain analysis), 95–97
Artificial loads, 12
Average constant acceleration method, 187–191
Axial pressures, 210–211
Bandwidth method, 70–74, 427
Beam, free vibration, 213–221
boundary conditions, decoupling, 214–215
free-free beam, 217–221
simply supported beam, 215–217
Beam axial deformations, 211–214
Beam flexure, viscous damping, 211
Beam with axial pressures, 210–211
Betti’s law, 139, 140
Bilateral spectral density, 244
Bottom shearing force method, 317–322, 443
Boundary conditions, 15, 137, 156, 214–217, 219, 220, 222, 441
Central difference method, 172, 186–187
Chaplin’s test, 412
Civil engineering application, 10–11
Coherence function, 247–248
Complex damping, 103–105
Complex modal analysis, 345–349
complex conjugate characteristics, 348–349
decoupling, 349
movement characteristics, 349
orthogonality, 349
Continuous systems, 207
Coordinate coupling equation, 173
Correlation function, 236–238
of derivative process, 238–239
properties of, 239–241
Covariance function, 236
Cramer rule, 133
Critical damping, 47–49
Cross-spectral density, 246–247
Current, 12, 13, 19, 20, 52, 108, 110, 173, 253, 306, 308, 338, 340, 341, 369, 375–377, 384–386, 389, 399–401, 403, 405, 411
D’Alembert’s principle, 21–23
Damage identification, 247, 305, 350–358, 365, 368–369
Damped free vibrations, 46–52
critical damping, 47–48
damping ratio, 48
overdamped system, 51–52
underdamped systems, 48–49
Damping, 13, 18–20, 32, 38
Damping and measurement, 52–57
damping form, 52–53
free vibration test, 56–57
motion attenuation and damping ratio, 53–55
Damping assumptions, 174–179
Cauchy damping, 176–179
Rayleigh damping, 175–176, 179
Damping coefficient, 48
Damping force, 19–20, 22, 32, 34–36, 49, 52, 53, 64, 80, 99, 101, 102, 104, 106, 122, 174, 175, 211, 260, 308, 333, 338–341, 358
Damping form, 52–53
complex damping, 52–53
friction damping, 53
viscous damping, 52
Damping matrices, 399–400
Damping ratio, 49
Damping ratio solution, 70–74
bandwidth method, 70–73
logarithmic decay rate method, 70
resonance amplification method, 73–74
Damping system numerical analysis, 185–198
average constant acceleration method, 187–191
central difference method, 186–187
linear acceleration method, 191–193
Newmark-β method, 193–195
Wilson-θ method, 195–198
Decoupling the boundary conditions, 214–215
Definite function theory, 307
Deformations, 19, 39, 211–212
Degree of freedom (DOF), 6, 7, 10, 17, 18
Differential equations for motion, 208–212
beam, free vibration, 213–221
beam axial deformations, 211–214
beam flexure, viscous damping, 211
beam with axial pressures, 210–211
Euler-Bernoulli beam, 208–210
modal decomposition, 223–225
orthogonality relationships, 221–223
Direct dynamic analysis theory, 308
Direct method, 32, 271–279, 290
Discrete Fourier transform, 96, 97
Discretization, 8, 9
methods, 6
Displacement amplitude, 47
Displacement time-history diagram, 78, 79
Displacement transmissibility, 117
Distributed-parameter system, 207–225
differential equations for motion, 208–212
overview, 207
Double integral operation, 263
Duhamel integral (time-domain analysis), 89–95, 181, 185, 261
of damping system, 92, 93–95
of undamped system, 91, 92–93
unit impulse response function, 89–91
Dunkerley formula, 147–150
Dynamic amplification coefficient, 65–68
Dynamic analysis nonlinear problems, 358–365
geometric nonlinearity problems, 362–365
physical nonlinearity problems, 359–361
Dynamic characteristics, calculation method, 146–172
Dunkerley formula, 147–150
matrix iteration method, 160–167
Rayleigh energy method, 150–156
Ritz method, 156–160
subspace iteration method, 167–172
Dynamic fracture technology, 15
Dynamic freedom, 6, 17–18
Dynamic loads, 2–4
classification of, 3
impulsive load, 3, 4
irregular dynamic load, 3, 4
nonharmonic periodic loads, 3, 4
simple harmonic loads, 2–4
Dynamic problem characteristics, 4–10
discretization methods, 6
finite element method (FEM), 9–10
generalized coordinate procedure, 7–9
lumped mass procedure, 6–7
Dynamic response problem, 207
Dynamics, theoretical study, 1
Dynamics design technology, 15
Dynamic system basics, 18–20
damping force, 19–20
elastic restoring force, 19
inertia force, 18–19
Earthquake, 228, 255, 256, 258, 259, 267, 269, 284, 292, 305–313, 315–317, 323, 325, 327–332, 334, 340, 341, 359
Earthquake loads, 12, 13
Elastic restoring force, 19, 101, 308, 333, 362
Elastic system dynamics, 32–37
Electromagnetic exciter, 108
Energy dissipation
hysteretic curve of, 101–102
of viscous damped system, 99–100
Energy in vibration, 97–106
complex damping, 103–105
energy dissipation, viscous damped
system, 99–100
equivalent viscous damping, 100–103
free vibration, 97–99
friction damping, 106
Environmental conditions, 376–386
currents, categories, 384–386
linear wave theory, 377–384
nonlinear wave theory, 384
Equation of motion, 75, 77, 104, 199
Equivalent incremental load, 198
Equivalent linearization method, 294–297
nonlinear MDOF system, 296–297
nonlinear SDOF system, 294–296
Equivalent viscous damping, 100–103
energy dissipation, hysteretic curve, 101–102
Euler-Bernoulli beam, 208–210, 410
free vibration of, 213–221
Fatigue predictions, 270–271
Finite element method (FEM), 9–10
Five degrees of freedom system, 135
Floating ice loads, 12, 13
Fluid loads, 12
Force transfer rate, 121
Free vibration, 41–54, 56–59, 62, 70, 80–85, 88, 90, 97–99, 106, 107, 128–132, 135–146, 156, 160, 201–203, 213–214, 227, 291, 349
Free vibrations response, 41–57
damped free vibrations, 46–52
damping and measurement, 52–57
undamped free vibrations, 43–46
Frequency-domain analysis method, 263–266
pulse response and frequency
response function, 266
self-spectral density, 264–266
Frequency domain method, 230
Frequency-domain numerical
characteristics, 241–248
coherence function, 247–248
cross-spectral density, 246–247
power spectral density, 241–244
spectral density, derivative process, 244–246
Frequency response function, 77
Friction damping, 106
Gaussian stochastic process, 251–252
Generalized coordinate procedure, 7–9
Generalized mass, 142–146, 168, 170–171, 180, 223
Generalized Poisson distribution, 253
Generalized stiffness, 142–146, 168, 170, 171
Gravitational forces, 37
Hamilton’s principle, 26–30, 156
Harmonic loading, response, 57–79
damping ratio solution, 70–74
damping system, harmonic
vibration, 62–65
dynamic amplification coefficient, 65–68
resonance reaction, 68–69
undamped system, harmonic vibration, 57–62
Harmonic model, 406–409
Harmonic vibration, 290
damping system, 62–65
undamped system, 57–62
Hydraulic vibration exciter, 108
Hydrodynamic drag and inertia forces, 386–390
accelerated fluid flow, 387–388
complete Morison’s equation, 388
drag and inertia coefficient
parameter dependency, 388–390
steady fluid flow, 386–387
Hydrodynamic lift forces, 390–391
constant lift coefficients, 390–391
variable lift coefficients, 391
Hydrodynamic loads, 386–391
hydrodynamic drag and inertia forces, 386–390
hydrodynamic lift forces, 390–391
Impulsive load, 3, 4
Impulsive loading response, 80–89
rectangular impulse, 82–84
sine-wave impulse, 80–82
triangular impulse, 84–89
Inertia forces, 5, 6, 386–390
Irregular dynamic load, 3, 4
Isolation technology, 1, 325, 332–336
Keulegan-Carpenter number, 389
Kinetic energy, 12, 20, 27, 30, 97, 98, 150–153, 156
Kramer, law, 158
Lagrange’s equations, 30–32
Lift frequency, 13
Linear acceleration method, 191–193, 196
Linear algebra, 137
Linear MDOF system stochastic response, 271–290
frequency-domain analysis method, 276–279
time-domain analysis method, 272–276
vibration mode superposition method, 280–290
Linear SDOF system stochastic
response, 260–271
cross-correlation function, 266–270
cross-spectral density, 266–270
frequency-domain analysis method, 263–266
narrowband systems, fatigue
predictions, 270–271
time-domain analysis method, 260–263
Linear wave theory, 369, 377–384
Logarithmic attenuation rate, 55
Loop and Eddy currents, 385
Lumped mass procedure, 6–7
Magneto-rheological damper
control algorithm of, 339–341
fuzzy neural network control, 341
performance of, 337–339
tri-state control, 340–341
two-state optimal control, 339–340
vibration reduction technology, 337–341
Man-made loads, see Artificial loads
Marine pipeline and riser, 375–415
environmental conditions, 376–386
hydrodynamic loads, 386–391
overview, 375–376
structural response analysis, 392–401
vortex induced vibrations (VIV), 401–415
Marine structural dynamics, 11
Mass-spring-damping system, 42
Matrix iteration method, 160–167
Maximum dynamic displacement, 87
Mechanical vibration exciter, 107
Modal analysis and theory, 341–349
complex modal analysis, 345–349
modal parameters, 342–344
real modal analysis, 344–345
Modal assurance standard (MAC), 352
Modal base damage identification method, 351–354
displacement modal method, 353
modal confidence criterion, 352–353
modal curvature method, 351–352
strain mode method, 353–354
Modal decomposition, 223–225
Mode superposition method, damped system, 172–184
coordinate coupling and regular coordinates, 173–174
damping assumptions, 174–179
Motion attenuation, 53
Multi-degree of freedom system, 127–202
damping system numerical analysis, 185–198
dynamic characteristics, calculation method, 146–172
mode superposition method, damped system, 172–184
stepwise integration method, 199–202
two degrees of freedom system, 128–135
undamped system, free vibrations, 135–146
Multi-point system, 127
Natural circular frequency of vibration, 49
Natural frequency, 14, 50, 107, 119, 120, 124, 131–134, 137, 145–147, 153, 158, 168, 176, 177, 217, 251, 261, 267, 314, 317, 327, 350, 351, 400, 402,
407, 408, 424, 435
Natural loads, 12
Natural vibration circular frequency, 44
Newmark-β method, 172, 193–195, 202
Nonharmonic periodic loads, 3, 4
Nonlinear problems, 291, 358–365
Nonlinear structural, 291–297
Nonlinear structural stochastic response analysis, 291–297
equivalent linearization method, 294–297
perturbation method, 292–294
Nonlinear wave theory, 384
Numerical characteristics, 234–248
Ocean engineering application, 11–14
Offshore structures dynamics, 369–372
frequency domain analysis, 371–372
offshore waves descriptions, 370
wave spectra, 370–371
Orthogonality relationships, 221–223
Overdamped system, 48–50
Over-ride vibration, 13
Parallel iteration method, see Subspace iteration method
Partial differential equations, 207
Periodic load, 76
Periodic load response, 74–79
Perturbation method, 229, 292–294, 442
Phase angle, 66
Phase-frequency characteristic curve,
damping system, 66
Physical nonlinearity, 359–361
Poisson stochastic process, 252, 253
Power spectral density, 241–244, 256
of non-stationary processes, 243
properties of, 243–244
of stationary process, 241–243
Principle of total potential energy invariant, 34–37
Principle of virtual displacements, 24–26
Principle of virtual work, 26, 32–34, 36, 423
Probability description, 232
Pseudo-static calculation methods, 1
Rayleigh energy method, 150–156
Rectangular impulse loading, 82–84
Resonance reaction, 68–69
Response, 2, 4, 41–97, 227
Response spectrum theory, 227, 228, 307–308
Reynolds number, 404, 405
Ritz method, 156–160, 169
Rubber base isolation technology, 332–336
design method, 336
multi-particle calculation model, 335–336
rubber isolation bearing performance, 332–334
seismic isolation structure, rubber bearing, 334–336
single particle calculation model, 334–335
spatial computing model, 336
torsional vibration calculation model, 336
Seismic excitation, 227, 260, 285
Seismic ground motion, stochastic process, 253–260
filtered white noise model, 254–256
Housner model, 257
improved filtered white noise model, 256–257
multi-peak power spectrum model, 257–258
stabilized stochastic process model, 258–260
stationary white noise model, 254
Seismic resistance, 1
Seismic response spectrum, 310–313
Simple harmonic loads, 2–4
Simply supported beam, 42, 149, 151, 154, 215–217
Sine-wave impulse loading, 80–82
Single degree of freedom systems (SDOF), 41–125, 187, 227
arbitrary load response, 89–97
energy in vibration, 97–106
free vibrations response, 41–57
harmonic loading, response, 57–74
impulsive loading response, 80–89
periodic load response, 74–79
structural vibration induced fatigue, 121–123
structural vibration test, 106–114
vibration isolation principle, 114–121
Softness method, 147
State space method, 297–303
basic concept, 298–299
MDOF system, 302–303
SDOF system, 299–302
Static load, 11
Static theory, 307
Stationary stochastic process, 248–251
Stationary white noise model, 254
Stepwise integration method, 199–202
algorithm solutions accuracy analysis, 202
algorithm solutions stability analysis, 202
conditional and unconditional stability, 199
Stiffness matrices, 170, 397–399
Stiffness method, 131
Stochastic, 13, 227–305, 316, 405, 420
Stochastic process, 230–260
concept of, 230–232
frequency-domain numerical characteristics, 241–248
Gaussian stochastic process, 251–252
of linear MDOF system, 271–290
of linear SDOF system response, 260–271
Markov process, 253
numerical characteristics of, 234–248
Poisson stochastic process, 252–253
probability description of, 232–234
seismic ground motion, 253–260
stationary stochastic process, 248–251
time-domain numerical characteristics, 235–241
Stochastic structural vibrations, 227–303
linear MDOF system stochastic response, 271–290
linear SDOF system stochastic response, 260–271
nonlinear structural stochastic
response analysis, 291–297
overview, 227–230
state space method, structural stochastic response analysis, 297–303
stochastic process, 230–260
Stodola method, see Matrix iteration method
Storm-generated currents, 385
Strouhal number, 405
Structural adaptive vibration control/active control technology, 15
Structural dynamic damage identification, 350–358
energy-based damage identification method, 356
flexibility change, 355
frequency base damage identification method, 350–351
modal base damage identification method, 351–354
research prospects, 357–358
stiffness variation, 354–355
Structural equation of motion, 17–39
direct equilibration, D’Alembert’s principle, 21–23
dynamic freedom, 17–18
dynamic system basics, 18–20
formulation of, 21–32
gravitational forces influence, 37
Hamilton’s principle, 26–30
Lagrange’s equations, 30–32
support excitation influence, 38–39
total potential energy invariant value theory, 32–37
virtual displacements principle, 23–26
Structural response, 2, 85, 103, 227, 228, 262, 264, 265, 323, 339, 392–401, 419, 430
Structural response analysis, 392–401
damping matrices, 399–400
global deformation, environmental loads, 392–394
mass matrices, 394–396
riser deformation, 400–401
stiffness matrices, 397–399
Structural seismic response analysis, 305–322
bottom shearing force method, 317–322
calculation method, brief introduction, 307–308
definite function theory, 307
direct dynamic analysis theory, 308
response spectrum theory, 307–308
SDOF elastic system, horizontal seismic action, 308–309
seismic response spectrum, 310–313
static theory, 307
vibration mode decomposition method, 314–317
Structural vibration control, 323–341
active control, 323–324
concept and classification, 323–325
hybrid control, 324–325
passive control, 323
rubber base isolation technology, 332–336
semi-active control, 324
vibration reduction technology, magneto-rheological damper, 337–341
vibration reduction technology, viscoelastic dampers, 325–332
Structural vibration induced fatigue, 15, 121–123
characteristics of, 122–123
definition of, 121–122
high and low frequency vibration, 123
resonant fatigue and non-resonant fatigue, 122–123
Structural vibration test, 106–114
data acquisition and analysis system, 114
pseudo dynamic test, 107
shaking table test, 107
sinusoidal loading test, 106
vibration excitation equipment, 107–109
vibration measuring instrument, 110–113
Subspace iteration method, 167–172
Sub-structure method, 350, 365–369, 444
damage identification by, 368–369
finite element analysis of, 365–367
Support excitation, 38–39
Three degrees of freedom system, 147
Time-domain analysis method, 260–263
autocorrelation function, 262–263
mean, 262
Time-domain method, 89–96, 185, 235, 241, 260, 272–273
Time-domain numerical characteristics, 235–241
correlation function, 236–241
mean and variance, 235
Time-varying applied load, 38
Total potential energy invariant value theory, 32–37
principle derivation, 34–37
virtual work principle, 32–34
Total spring stiffness, 119
Transitive rule, linear algebra, 138
Triangular impulse loading, 84–89
Turbulence frequencies, 13
Two degrees of freedom system, 128–135
equations of motion, solutions, 134–135
natural frequency and vibration mode shape, 131–134
undamped free vibrations, motion equation, 128–131
Undamped free vibrations, 43–46
Undamped system, free vibrations, 135–146
generalized mass and stiffness, 142–146
motion equation establishment, 135–137
vibration shape and orthogonality, 137–142
Vibration energy, 41, 97–106, 323
Vibration excitation equipment, 107–109
impact hammer, 108–109
shaking table, 108
vibration exciter, 107–108
Vibration isolation, 41, 114–121, 341
Vibration isolation principle, 114–121
active vibration isolation, 114–116
passive vibration isolation, 116–121
Vibration measurement, 41, 110, 114
Vibration measuring instrument, 110–113
acceleration sensors, 110–112
displacement sensors, 112–113
speed sensor, 113
strain sensors, 110
Vibration mode decomposition method, 314–317
Vibration mode shape, 131–134, 151
Vibration mode superposition method, 280–290
Vibration reduction technology, 325–332
analysis method, structure installed, 327
engineering application and experimental research, 328–332
magneto-rheological damper, 337–341
rubber base isolation technology, 332–336
viscoelastic damper performance, 325–327
viscoelastic dampers, 325–332
Virtual displacement, 32–34, 36
Virtual displacements principle, 23–26
Viscous damping, 19, 20, 172
Vortex induced vibrations (VIV), 375, 390, 401–415
analysis of, 404–406
harmonic model, 406–409
introduction, 401–403
wake oscillator model, 409–415
Wake oscillator model, 409–415
Wave parameters, 12
Wavepropagation formulas, 207
Wiener-Khintchine formula, 244, 254
Wilson-θ method, 172, 195–200
Wind loads, 12
WILEY END USER LICENSE AGREEMENT
Go to www.wiley.com/go/eula to access Wiley’s ebook EULA.