You are on page 1of 1023

ENGINEERING
SYSTEM
OYNAMICS

CONTROL ENGINEERING
.4 Series of Reference Booksand Textbooks
Editor
NEIL MUNRO,PH.D., D.Sc.
Professor
AppliedControl Engineering
University of ManchesterInstitute of Scienceand Technology
Manchester, United Kingdom

1. NonlinearControl of Electric Machinery,Darren M. Dawson,Jun Hu, and
Timothy C. Burg
Intelligence in Control Engineering,RobertE. King
2. Computational
3. Quantitative FeedbackTheory: Fundamentalsand Applications, Constantine H. Houpisand StevenJ. Rasmussen
4. Self-LearningControl of Finite MarkovChains,A. S. Poznyak,K. Najim,
and E. G6mez-Ramirez
5. RobustControl and Filtering for Time-DelaySystems,MagdiS. Mahmoud
6. Classical FeedbackControl: With MATLAB,
Boris J. Lurie and Paul J.
Enright
and Applications:
7. OptimalControl of Singularly PerturbedLinear Systems
High-AccuracyTechniques,Zoran Gajid and Myo-TaegLim
8. Engineering SystemDynamics: A Unified Graph-CenteredApproach,
Forbes T. Brown
Additional Volumesin Preparation
AdvancedProcessIdentification
Najim

and Control, Ensolkonen and Kaddour

Modem
Control Engineering, P. N. Paraskevopoulos
Sliding ModeControl in Engineering,Wilfrid Perruquetti and JeanPierre
Barbot

ENGINEERING
SYSTEM
OYNAMICS
A Unified Graph-CenteredApproach

Forbes T. Brown
Lehigh University
Bethlehem, Pennsylvania

MARCELDEKKER, INC.
DECKER

NEWYORK¯ BASEL

ISBN: 0-8247-0616-1
This bookis printed on acid-free paper.
Headquarters
MarcelDekker, Inc.
270 Madison Avenue, NewYork, NY10016
tel: 212-696-9000;fax: 212-685-4540
Eastern HemisphereDistribution
Marcel Dekker AG
Hutgasse 4, Postfach 812, CH-4001Basel, Switzerland
tel: 41-61-261-8482;fax: 41-61-261-8896
World Wide Web
http://www.dekker.com
The publisher offers discounts on this book whenordered in bulk quantities. For more
information, write to Special Sales/Professional Marketing at the headquarters address
above.
Copyright © 2001 by Marcel Dekker,Inc. All Rights Reserved.
Neither this book nor any part maybe reproduced or transmitted in any form or by any
means,electronic or mechanical, including photocopying,microfilming, and recording, or
by any information storage and retrieval system, without permission in writing from the
publisher.
Currentprinting (last digit):
10987654321
PRINTED IN THE UNITED STATES OF AMERICA

This book is dedicated to
Henry M. Paynter
Marjorie H. Brown
who in their distinct ways
madeit possible.

Series

Introduction

Manytextbooks have been written on control engineering, describing new
techniques for controlling systems, or new and better ways of mathematically formulating existing methods to solve the ever-increasing complex
problems faced by practicing engineers. However, few of these books fully
address the applications aspects of control engineering. It is the intention of
this newseries to redress this situation.
The series will stress applications issues, and not just the mathematics
of control engineering. It will provide texts that present not only both new
and well-established techniques, but also detailed examplesof the application of these methods to the solution of real-world problems. The authors
will be drawn from both the academic world and the relevant applications
sectors.
There are already manyexciting examples of the application of control
techniques in the established fields of electrical, mechanical(including aerospace), and chemical engineering. Wehave only to look around in today’s
highly automated society to see the use of advanced robotics techniques in
the manufacturing industries; the use of automated control and navigation
systems in air and surface transport systems; the increasing use of intelhgent control systems in the manyartifacts available to the domestic consumer market; and the reliable supply of water, gas, and electrical power to
the domestic consumer and to industry. However, there are currently many
challenging problems that could benefit from wider exposure to the applicability of control methodologies, and the systematic systems-oriented basis
inherent in the application of control techniques.
This series presents books that draw on expertise from both the academic world and the applications domains, and will be useful not only as
academically recommendedcourse texts but also as handbooks for practitioners in many apphcations domains. Engineering System Dynamics is another outstanding entry to Dekker’s Control Engineering series.
Nell Munro

Preface
Chapters 1 - 6 of this textbook/reference constitute a basic course in system
dynamicsprimarily for juniors in mechanicalengineering and allied fields. It
can be read at more than one level, and includes problemsof varying sophistication, so it also workswell for graduate students whohave not been exposedto
its unifying methodology.By avoiding someof the topics relating to nonlinear
systems, material from Chapter 7 or 8 or AppendixB maybe added to emphasize comprehensivelinear methods, automatic control, or classical vibrations.
A senior-level course emphasizinglinear methodscan be centered on Chapters
7, 8 and 11, and a senior-level course emphasizing modelingcan be contered
on Chapters 9, 10 and 12, where nonlinear systems predominate. These latter chapters work even better with graduate students. Details are given in the
section entitled To The Instructor that follows thi~ Preface. Aninstructor’s
manualgives further suggestions, as well as solutions to the problems.
Modelingis the inherently pivotal and difficult element in the process of
analyzing real physical systems. Therefore, the text is organized around the
behavior of physical systems rather than around linear mathematics. It starts
with graphical descriptions of the steady-state characteristics of engineeringdevices and the concepts of stable and unstable equilibrium. Nonlinearity is not
avoided in modelingor simulation, since it impedes only hand analysis. The
modelingof mechanical,electrical, and fluid systems is treated simultaneously,
in order to develop a sense of analogy. Emphasisis placed on real engineering componentsand hybrid systems. This is made possible through the use of
the bond graph language. The traditional modelinglanguages of linear mathematics, linear block diagrams, and medium-specificdiagrams such as circuit
diagrams, free-body diagrams and fluid-power diagrams also are used, when
appropriate. Linear methodsare developedfully despite their subservient role.
The powerof bond graphs is widely appreciated, but they have been taught
more to graduate students than undergraduates, missing the majority of potential users. The author set out explicitly to develop a textbook appropriate for
undergraduates. Throughthe crucible of several years of trial and error, including manypreliminary editions, he gradually discovered the special impediments
encountered by undergraduates, and howto address these di~culties. Students
learn the bond graph language easily enough, but the modeling concepts that
the language so concisely express are new to most undergraduates, and need
to be developed gradually and carefully. These concepts exist independent of
language, and are essential to an ability to modela range of physical systems
without over-particularization. Textbooksthat do not employbond graphs tend
to address modelingin a relatively superficial or ad hoc manner. Viewedthis
way, the use of bond graphs is the simplest wayto achieve competencein the
modelingof systems based on power, energy and forces.
Most topics are introduced through case studies. Exampleproblemshelp get
the student started, and the "guided problem" is designed to encourage vital
active learning. A problemis posed, and its relevance to the student’s learning
vii

viii

PREFACE

noted. A list of suggested steps or hints are then given. A solution is given a
fewpages later.
MATLAB
is employedextensively. Nonlinear models are treated with standard numerical integrators. Linear modelsare treated with standard operatorbased routines. It is important that students understand what these routines
mean, but it is no longer critical that they becomehighly proficient in hand
analysis and calculations.
The advent of the information age has impacted the cognitive processes of
young people in both positive and negative ways. Certain social and communicative skills have been enhanced.On the other hand, critical thinking likely
has declined, including its logical, visual and kinesthetic components.A decrease in the time children are given for self-directed creative play contributes
to this phenomenon.Faced with a problem, students today are apt to ask where
they can find a solution, not howthey can solve it themselves. Theyoften view
the sample problem, rather than fundamental concepts, as the repository of
knowledge. The change is partly justified by moderncommunicationnetworks,
but it hinders the learning of basic modelingand analysis skills.
The remedyincludes better integration of the engineering curriculum. Industry had taken a back seat to governmentin shaping engineeering departments, particularly because of Cold Warfunding. An increasing specialization
resulted, including teaching assignments and the developmentof a new traditional style that could be called "prescription learning." A. rebellion emerged
in the 1980s, in the form of an insistence on the primacy of "design." More
recently, most remainingcourses at major universities in the relatively specialized area of mechanical vibrations have been replaced by more general courses
in system dynamics,often with an introduction of control. It is ironic that the
resulting plethora of textbooks in systemdynamics,while nominallyintegrative,
itself tends to employthe prescription learning that is antithetical to integrative
learning. This text is intended to be moreintegrative, both in its subject and
in its cognitivestyle.
This booknever wouldhave been written had I not had the privilege of contact with HenryPaynter, the creator of bondgraphs, largely from the late 1950s
and through 1970. It also benefited from contributions by myson, Gordon,Professor KennethSawyers, Professor RamChandranat Kettering University, especially Professor Timothy Cameronwho is presently at Virginia Commonwealth
University, and most especially myLehigh colleague and co-teacher Professor
No DukePerreira.
Forbes T. Brown
¯ .. well-connected representations let you turn ideas aroundin your mind,
envision things ~rommanyperspectives ¯ .. that is what we meanby thinking.
Marvin Minsky
¯ .. there is somethingessential in humanunderstandingthat is not possible
to simulate by any computational means.
Roger Penrose

To the Instructor

The structure of the first part of the book favors a basic course in modeling
and simulation, with more emphasison nonlinear systems than is conventional.
By reducing this emphasis, you can fashion an alternative course that emphasizes comprehensivelinear methods, an introduction to automatic control, or
linear classical vibrations. Suggeste
d sequencesof sections are given in the diagrambelow; the detailed comments
that follow identify certain further possible
variations.
Suggestedsequencesof chapters and sections for first
undergraduatecourses, four different emphases
3.6-3.7’
4’

4.1-4.3

5

5.1(skip5.1.4),5.2.1-5.2.3,5.3,5.4.1-5.4.6

6.1-6.2.5
6.2.6-6.2.8~
6.3

6.1-6.2.5
optional )6.2.6-6.2.8
topics
6.3

optional
8 ~ topics
modelingand
linear
automatic
simulation
methods control

classical
vibrations

Modelingultimately is a nondeterministic process, and consequently this
text is not written in the minimalistic style of elementary textbooks in engi:
neering science. Rather, someengineering color or practical asides are.inserted
ix

x

TO THE INSTRUCTOR

from time-to-time. The objective is two-fold: to conveyinteresting practical
information, and to developthe students’ critical ability to distinguish core essentials from peripheral details. The students need to understand at least in
general terms what is expected of them, however, so you maywish to discuss
this issue with them at an appropriate time. The learning objectives should
depend on the maturity and level of the students, and you have considerable
latitude in the sophistication of the problemsyou choose to assign as homework
or discuss in class.
The first six chapters represent about three credit hours of workfor a typical
undergraduate. The remaining six chapters represent about five semester hours
more, assumingfew topics are skipped. The later chapters have less redundancy
than the earlier chapters. Graduate students typically not only can skip some
of the more routine mathematical sections, but.can proceed at a faster pace,
covering the entire unfamiliar material in about five semester hours. There are
manyoptional topics that can be deleted to form coherent courses with fewer
credits, depending on the objectives and the mathematical preparation of the
students. The author believes it is important to have all the topics under one
cover, however,to provide flexibility in course design and continuity between
successive courses, to allow the workingengineer to extend his/her capabilities,
and to demonstrate that the basic approach extends to sophisticated modeling
and analysis, unlike someother methodologies.
Chapter 1 introduces beginning students to system modelingand analysis,
discusses contrasting learning styles and addresses the treatment of dimensions
and units. Advancedstudents could skip someor all of this chapter.
Chapters 2-3 represent a first pass at modelingand analysis, with about 1
credit hour of content for the beginning undergraduate. Chapters 4-6 continue
with a second pass, significantly extending the domainof the modelsby allowing greater structural complexity for the same energy types. Chapter 2 starts
by developing the seminal idea of steady-state source-load synthesis, which is
strangely missing from manyengineering curriculums. This focuses attention
on the constitutive characteristic, which is represented graphically more than
analytically in order to develop the student’s basic grasp of its meaning.The
chapter proceeds to develop the special cases of static couplers known(in bondgraph language) as transformers and gyrators, and applies these, particularly,
to maximizethe power transmitted from a source to its load. Although there
is little dynamicsin the development, the student begins to address dynamics
qualitatively, including the determinationof stability for certain cases..
The first three sections of Chapter 3 completethe introduction of the fundamental modelingelements: the compliance, the inertance and the junction.
Examplesare limited to models with a single junction, because this element
requires the student to identify variables and the constraints between them,
whichlies at the heart of the modelingprocess and most challenges the student.
The .chapter proceeds to deduceand solve the linear differential equations for
these models. Sections 3.6 and 3.7 introduce nonlinear elements and the use of
numerical simulation particularly for nonlinear models. These sections maybe
skipped if other matters are assigned a higher priority.

TO THE INSTRUCTOR

xi

The heart of the basic modelinginstruction lies in Sections 4.1 and 4.2. Here,
as elsewhere, you can adjust the level of sophistication through your selection
of classroom examples and homeworkproblems. Someof Section 4.3 on model
equivalences and all of Section 4.4 on equilibrium can be skipped if little emphasis is intended on nonlinear models. Sections markedwith an asterisk in the
table of contents and the text in this and other chapters are optional, even if
nonlinear modelsare a priority.
Modelsare converted to differential equations in Sections 5.1 and 5.2. The
treatment of under-causal modelsthrough the use of virtual compliancesor inertances is unique to this book, and seems to be readily applied by students.
Youmayskip Sections 5.2.4-5.2.6 if difficult nonlinear modelshave lowpriority.
The introduction of operator methodsin Section 5.3 assumes no prior knowledge. The discussion of linearization in Sections 5.4.1-5.4.6 probably should be
included in any course, but the subsequent subsections are optional.
The classical solution of linear differential equations in Section 6.1 assumes
no formal training. The use of the phasor methodfor frequency response in
Section 6.2 is similar to that given in almost any course in automatic control, including the use of Bode plots. If you emphasize the use of MATLAB
rather than hand calculations, and place a low priority on the inverse problem
of systemidentification through frequencyresponse, you could skip subsections
6.2.6-6.2.8, saving considerable time.
There is considerable latitude in whether, or when, to introduce Laplace
transforms formally. As ordered in the text, the formal study of Laplace transforms is deferred until after their mathematicalbases in Fourier transforms and
in convolution are established. The earlier use of the operator (Heaviside) notation allows transfer function expansion to be given in Section 6.3, so that
problemswith zero initial conditions can be treated, including full use of the
Laplace transform tables. Chapter 8 on automatic control is accessible on this
basis. Should you nevertheless wish to introduce Laplace transforms early and
explicitly, particularly so that initial-value problemscan be treated, you may
insert Section 7.2 (less subsections 7.1.1 and 7.1.2) right after Section 6.1; the
details of the partial fraction expansioncan be introduced as needed.
A full presentation of linear methodsshould include Section 6.4 on convolution, but the topic is otherwise optional. The Fourier story in Section.7.1 is
important should you wish to focus on vibration, but otherwise is also optional.
Section 7.3 is a rather classic exposition of matrix methodsin the study of linear models. Section 7.4 is a modernexposition on the use of the loop rule to
expedite the determination of transfer functions for linear modelsrepresented
by bond graphs. The material also is part of the basis of somemodernbondgraph software. As indicated by the asterisks, however,Sections 7.3 and 7.4 are
optional for most courses of study.
At least one-half of a first coursedevotedexclusivelyto control, not including
modeling, is accommodated
by the text. This includes material in Chapters 5
and 6 as well as Chapter8. A briefer treatment can end after the first or second
of the three sections in Chapter8, and still represent a balancedcoverageof the
subject.

xii

TO THE INSTRUCTOR

Chapters 9 and 10 return to the subject of modeling, greatly extending the
domainof applicability. These chapters can be studied independently of Chapters 7 and 8. Chapter 9 focuses on modelswith variable couplers, such as modulated transformers and gyrators, activated bonds, nonconservativecouplers and
irreversible couplers. The irreversible coupler introduces thermal systems with
heat conduction, and connections to other energy media. Section 10.1 elaborates on the meaning and methodology of lumping. The balance of Chapter
10 focuses on coupled complianceand inertance fields, treating a variety of energy transducers and providing a unique insight into the distinction between
holonomic and nonholonomicconstraints.
Chapter 11 introduces the seminal idea of distributed-parameter modelsand
analysis, with restriction to one-dimensionalmodelsand linear media. Sections
7.1 and 7.2 are appropriate prerequisites. Emphasisis placed on wave-like behavior and on boundary-value problems. Someof the boundary conditions are
nonlinear. Models are categorized according to the numberof distinct powers propagated, and organized in such a fashion as to minimizecomputational
effort.
Chapter 12 addresses the modeling and analysis of thermodynamicsystems,
particularly with a flowing compressiblepure substance. The material in Chapters 7, 8 and 11 is not prerequisite. It is natural to adapt bQndgraphs to
thermodynamicsystems, since the graphs deal inherently with energy flux and
can be extended to deal with the distinctions described by entropy. Muchof
the material has not been published elsewhere, and maybe of special interest
to those whoare already expert at traditional bond-graph methods. Practical
codes are developed for the simulation of unsteady or nonequilibrium models,
including multiphase flows, without any need for iteration in most cases. Accurate analytical modelsare used for the properties of selected gasses and liquids.
Codesfor evaluating the properties are given in AppendixD and, together with
the simulation codes given in the chapter, can be downloadedfrpm the Lehigh
University website. Chemicalkinetics is introduced in the last section of the
chapter, employinga further extension of bond graphs.
Appendix A is a primer on MATLAB,
which together with the "help" windowshould suffice for the tasks given in the book. Youcan assign this material
early in the first course, or as needed.
AppendixB places special emphasison the particular advantages of the classical vibrations approach, whichis based on second-orderdifferential equations
rather than the first-order or state-space e}tuations used in the rest of the text.
The second-order approachis particularly useful in the treatment of high-order
models without damping, or in’ which dampingis small enough to treat in an
approximatemanner.~ A course that emphsizes mechanical vibrations also ought
to makespecial use of manyof the vibrations examplesand problems scattered
throughout Chapters 2-6. The material in Chapter 11 on distributed-parameter
modelsalso is highly relevant to mechanicalvibrations.
Forbes T. Brown

Contents

Table of Bond Graph Elements ..............................

inside front cover

Series Introduction by Neil Munro...........................................

v

Preface ....................................................................

vii

To the Instructor

ix

..........................................................

Chapter 1 INTRODUCTION
...........................................
Example; Modeling and Engineering Science; Modeling Languages;
Modeling for Control; A Word to the Wise About Learning; Treatment of Dimensions; Treatment of Units; References
Chapter

2 SOURCE-LOADSYNTHESIS ............................

2.1 System Reticulation ..............................................
2.1.1 Case Study: Induction Motor as a Source; 2.1.2 Case Study:
Water Sprinkler System as a Load; 2.1.3 The Source-Load Synthesis; Case Study; 2.1.4 Summary
2.2 Generalized Forces and Velocities ...............................
2.2.1 Efforts and Flows; 2.2.2 Electric Conductors; 2.2.3 Longitudinal Mechanical Motion; 2.2.4 Incompressible Fluid Flow; 2.2.5
Rotational Motion; 2.2.6 Lateral Mechanical Motion; 2.2.7 Microbonds; 2.2.8 Analogies; 2.2.9 Summary
2.3 Generalized Sources~ Sinks and Resistances. : .................
2.3.1 Independent-Effort and Independent-Flow Sources and Sinks;
2.3.2 General Sources and Sinks; 2.3.3 Linear Resistances; 2.3.4
Nonlinear Resistances; 2.3.5 Source-Load Synthesis; 2.3.6 Power
Considerations; 2.3.7 Summary
2.4 Ideal Machines: Transformers and Gyrators ..................
2.4.1 Ideal Machines; 2.4.2 Transformers; 2.4.3 Gyrators; 2.4.4 Mechanical Devices Modeled as Transformers; 2.4.5 Electric Transformers; 2.4.6 Transducers Modeled as Transformers; 2.4.7 Mechanical Devices Modeled as Gyrators; 2.4.8 Transducers Modeled
as Gyrators; 2.4.9 Summary
xiii

1

17
18

30

40

57

xiv

CONTENTS

2.5 Systems with Transformers and Gyrators ......................
2.5.1 CascadedTransformers; 2.5.2 CascadedGyrators; 2.5.3 Case
Study of a TransformerConnectinga Source to a Load; 2.5.4 Second Case Study of a Transformer Connecting a Source to a Load;
2.5.5 Case Study of a Gyrator Connectinga Sourceto a Load; 2.5.6
Transmission Matrices*; 2.5.7 Summary

72

Chapter 3 SIMPLEDYNAMIC
MODELS
...........................
3.1 Compliance Energy Storage .....................................
3.1.1 Linear Springs and Energy; 3.1.2 The Generalized Linear
Compliance;3.1.3 Electric Circuit Compliance;3.1.4 Linear Fluid
Compliance Due to Gravity; 3.1.5 Fluid Compliance Due to Compressibility; 3.1.6 Summary
3.2 Inertance Energy Storage .......................................
3.2.1 Mass, Momentum
and Kinetic Energy; 3.2.2 The Generalized
Linear Inertance; 3.2.3 Common
Engineering Elements Modeledby
Constant Inertances; 3.2.4 Tetrahedron of State; 3.2.5 Summary
3.3 Junctions .........................................................
3.3.1 Junction Types; 3.3.2 Mechanical Constraints Modeled by
1-Junctions; 3.3.3 Electrical Circuit Constraints Modeledby 1Junctions; 3.3.4 Fluid Circuit Constraints Modeledby 1-junctions;
3.3.5 MechanicalConstraints Modeledby 0-Junctions; 3.3.6 Electric and Fluid Circuit Constraints Modeledby 0-Junctions; 3.3.7
Simple IRC Models; 3.3.8 Summary
3.4 Causality and Differential Equations ..........................
3.4.1 Causalities of Effort and Flow Sources; 3.4.2 Junctions With
Elements Having UncoupledBehavior; 3.4.3 Junctions with Elements Having CoupledBehavior; 3.4.4 Writing Diffferential Equations; 3.4.5 Summary
3.5 Solutions of Linear Differential Equations .....................
3.5.1 The First-Order Differential Equation; 3.5.2 Responsesto
MoreComplexExcitations Using Superposition; 3.5.3 The Second=
Order Differential Equations; 3.5.4 Solution of the Second-Order
Differential Equations; 3.5.5 Summary
3.6 Nonlinear Compliances and Inertances ........................
3.6.1 Nonlinear Compliances; 3.6.2 Nonlinear Fluid Compliance
Dueto Gravity; 3.6.3 Nonlinear Compressibility Compliance;3.6.4
Junctions with Multiple BondedCompliances; 3.6.5 Nonlinear Inertances; 3.6.6 Kinetic and Potential Energies; 3.6.7 Summary
3.7 Numerical Simulation ...........................................
3.7.1 State-Variable Differential Equations; 3.7.2 Simulation with
MATLAB;
3.7.3 Integration Algorithms; 3.7.4 Second-Order
Runge-Kutta; 3.7.5 Fourth-Order Runge-Kutta; 3.7.6 Summary

91
92

*- optional topic (see ToTheInstructor)

102

110

134

145

162

174

CONTENTS
Chapter 4 INTERMEDIATE
MODELING
.........................
4.1 Simple Circuits ..................................................
4.1.1 SimpleElectric Circuits; 4.1.2 Fluid Circuits; 4.1.3 Mechanical Circuits; 4.1.4 Summary
4.2 System Models with Ideal Machines ...........................

xv
191
191

205

4.2.1 Electric Circuits; 4.2.2 Fluid/MechanicalCircuits with Positive DisplacementMachines;4.2.3 Losses in Positive Displacement
Machines*; 4.2.4 Losses with DCMotor/Generators*; 4.2.5 Case
Study with Source and Load*; 4.2.6 Twoand Three-Dimensional
Geometric Constraints; 4.2.7 Case Study: Pulley System; 4.2.8
Modeling Guidelines; 4.2.9 Tutorial Case Study; 4.2.10 Common
Misconceptions; 4.2.11 Summary
4.3 ModelEquivalences .............................................
242
4.3.1 Thevenin and Norton Equivalent Sources and Loads; 4.3.2
Passivity with Respect to a Point on a Characteristic*; 4.3.3 Truncation of Transformers and Gyrators Bondedto R, C, or I Elements; 4.3.4 Reduction of Two-Pair Meshes; 4.3.5 Transmission
Matrix Reduction of Steady-State Models; 4.3.6 Summary
4.4 Equilibrium .......
...............................................
258
4.4.1 Reductionof Steady-State Modelswith a Single Source; Case
Study; 4.4.2 Alternative Approaches to Reducing Steady-State
Models; 4.4.3 Removalof Elements for Equilibrium; 4.4.4 Case
Study with a Steady-Velocity Equilibrium; 4.4.5 Case Study with
Stable and Unstable Static Equilibria; 4.4.6 Case Study with LimitCycleBehavior; 4.4.7 NecessaryConditionfor Instability or LimitCycle Oscillation*; 4.4.8 Summary
Chapter

5 MATHEMATICAL
FORMULATION
..................

5.1 Causality and Differential Equations ..........................
5.1.1 ApplyingCausal Strokes; 5.1.2 Differential Equations for
Causal Models; 5.1.3 Case Study: A Linear Circuit; 5.1.4 Case
Study: Nonlinear Stick-Slip; 5.1.5 Case Study with Transformers
and Gyrators; 5.1.6 ModelsReducible to Causal Form; Order of a
Model; 5.1.7 Summary
5.2 Over-Causal and Under-Causal Models .......................
5.2.1 Treatment of Over-Causal Models; Case Study; 5.2.2 Equations for Under-CausalModels; 5.2.3 Algebraic Reduction Method;
Case Study; 5.2.4 Differentiation Method; Case Study*; 5.2.5
Method of Non:Zero Virtual Energy-Storages; Case Study Continued*; 5.2.6 CommercialSoftware for DAE’s;5.2.7 Case Study
With Meshes*; 5.2.8 Summary

281
281

309

xvi

CONTENTS

5.3 Linear Modelsand Simulation ..................................
5.3.1 Superposition and Linearity; 5.3.2 Linearity and Differential EqUations;5.3.3 Operator Notation; 5.3.4 Transformationfrom
State-Space to Scalar Form; 5.3.5 Transformation FromScalar to
State-Space Form*; 5.3.6 Transformations Using MATLAB;
5.3.7
Simulation of Linear Models Using MATLAB*;
5.3.8 Summary
5.4 Linearization .....................................................
5.4.1 Case Study with Linearization of a Resistance; 5.4.2 Linearization of a Function of OneVariable; 5.4.3 Essential Nonlinearities; 5.4.4 Linearization of a Functionof TwoVariables; 5.4.5 Linearization of a First-Order Differential Equation; 5.4.6 Linearization of State-Variable Differential Equations;5.4.7 Case Study with
Three Different Types of Equilibria; 5.4.8 Case Study: Stick-Slip
System with Inertia*; 5.4.9 Summary
Chapter 6 ANALYSIS OF LINEARMODELS,PART 1 .........

341

358

389

6.1 Direct Solutions ’of Linear Differential Equations ............
6.1.1 The Homogeneous
Solution; 6.1.2 Singularity SystemInputs;
6.1.3 Exponential and S~nusoidal Inputs; 6.1.4 Power-LawInputs;
6.1.5 The Methodof UndeterminedCoefficients; 6.1.6 Application
of Initial Conditions; 6.1.7 Solutions to ImpulseInputs; 6.1.8 Differentiation and Integration Properties; 6.1.9 Step and ImpulseResponses Using MATLAB;
6.1.10 Summary
6.2 Sinusoidal FrequencyResponse.................................
6.2.1 The Phasor Method;6.2.2 Bodeplots; 6.2.3 ModelsWithout
Damping;6.2.4 Models Comprising a Single Pole or Zero; 6.2.5
ModelsComprisinga Pair of ComplexPoles or Zeros; 6.2.6 Factorization of Higher-OrderModels;6.2.7 ,Bode Plots for Highei-Order
Models; 6.2.8 The Pure Delay Operator*; 6.2.9 Summary.
6.3 Transfer Function Expansion...................................
6.3.1 Impulse Responses; 6.3.2 Step and Other Transient Responses; 6.3.3 Determinationof Partial Fraction Expansions;6.3.4
Summary
6.4 Convolution*.....................................................
6.4.1 Decomposing
Signals into a Sumof Steps; 6.4.2 Discrete Convolution; 6.4.3 Discrete Convolution by MATLAB;
6.4.4 Convolution Integrals; 6.4.5 Summary

389

Chapter 7 ANALYSIS OF LINEAR MODELS,PART 2 .........
7.1 FourierAnalysis ..................................................
7.1.1 Fourier Series; 7.1.2 Responseof a LinearSystemto a Periodic
Excitation; 7.1.3 Fourier Transform;7.1.4 Digital Spectral Analysis*; 7.1.5 Fourier Analysis Using MATLAB*;
7.1.6 Summary

485
485

408

459

470

xvii

CONTENTS
7.2 The Laplace Transform ..........

’ ...............................

507

7.2.1 Development from the Fourier Transform*; 7.2.2 Development
from the Convolution Integral*; 7.2.3 Definition and Inverse; 7.2.4
The Derivative Relations; 7.2.5 Singularity Functions and Discontinuities; 7.2.6 Other Key Relations; 7.2.7 Finding Laplace Transforms of Output Variables; 7.2.8 Partial Fraction Expansions; 7.2.9
Initial and Final Value Theorems; 7.2.10 Summary
7.3 Matrix Representation

of Dynamic Behavior*

......

: .........

526

7.3.1 The Matrix Exponential; 7.3.2 Response to a Linearly Varying Excitation; 7.3.3 Eigenvalues, Eigenvectors and Modes; 7.3.4
Case Study: Three Fluid Tanks; 7.3.5 Case Study with Complex
Roots; 7.3.6 Modified Method for Complex Eigenvalues*; 7.3.7
Case Study: Vehicle Dynamics; 7.3.8 Application of MATLAB;
7.3.9 Response to Exponential and Frequency Excitations; 7.3.10
Representation in the s-Plane; 7.3.11 Summary
7.4 The Loop Rule* .................................................

561

7.4.1 Signal Flow Graphs; 7.4.2 The Loop Rule for Signal Flow
Graphs; 7.4.3 Converting Bond Graphs to Signal Flow Graphs;
7.4.4 Direct Application of the Loop Rule to Bond Graphs Without
Meshes; 7.4.5 Bond Graphs with Meshes; 7.4.6 Determination of
State Differential Equations; 7.4.7 Summary
Chapter

8 INTRODUCTION

TO AUTOMATIC

CONTROL ....

583

8.1 Open and Closed-Loop Control ................................
8.1.1 Example Plant; 8.1.2 0pen-Loop and Optimal Control; 8.113
Feedback Control; 8.1.4 Response to Disturbances; 8.1.5 Root Locus Basics; 8.1.6 Use of MATLAB;
8.1.7 Criteria for Stability; 8.1.8
Summary

583

8.2 Dynamic Compensation .........................................

605

8.2.1 Proportional-Plus-Integral
Control; 8.2.2 Proportional-PlusDerivative Control; 8.2.3 Proportional-Plus-Integral-Plus-Derivative
Control; 8.2.4 Phase Lead Controllers; 8.2.5 Phase Lag Controllers;
8.2.6 Phase Lead-Lag Controllers; 8.2.7 Digital Control Systems;
8.2.8 Summary
8.3 Frequency Response Methods ..................................
8.3.1 Polar or Nyquist Frequency Response Plots; 8.3.2 The
Nyquist Stability Criterion; 8.3.3 Measures of Relative Stability;
8.3.4 Nichols Charts; 8.3.5 Dynamic Compensation Using Nichols
Charts; 8.3.6 Approximate Correction for Digital Sampling; 8.3.7
Summary

629

xviii

CONTENTS

Chapter.

9 MODELSWITH STATIC COUPLERS ................

657

9.1 Modulated Transformers ........................................
9.1.1 Remotely Modulated Transformers; 9.1.2 Locally Modulated
Transformers; 9.1.3 Increase in the Order of a Model Due to Modulation; 9.1.4 Dependent Inertance with a Locally Modulated Transformer; 9.1.5 Inertance Dependent on Local Displacement; Case
Study; 9.1.6 Summary
9.2 Activated Bonds .................................................

657

9.2.1 Definition

682

and Application; 9.2.2 Causality; 9.2.3 Summary

9.3 Nonconservative Couplers ......................................
9.3.1 Causal Relations; 9.3.2 Equilibrium; 9.3.3 Dimensional Analysis; 9.3.4 DynamicSimulation; 9.3.5 Linear Couplers; 9.3.6 Summary
9.4 Irreversible
Couplers and Thermal Systems ..................

697

718

9.4.1 Effort and Flow Variables; 9.4.2 Heat Conduction; 9.4.3 The
Irreversibility
Coupler; 9.4.4 Application to Diction; 9.4.5 Use of
0- and 1-Junctions; 9.4.6 Thermal Compliance; 9.4.7 Pseudo Bond
Graphs for Heat Conduction; 9.4.8 Summary
Chapter

10 ENERGYSTORAGEFIELDS .........................

735

10.1 Field Lumping .......
i ..........................................
10.1.1 Scalar Fields; 10.1.2 Rigid-Body Vector Fields; 10.1.3 The
Role of Approximations; 10.1.4 Nodic Fields; 10.1.5 Planar Vector
Nodic Fields; 10.1.6 Estimating Upper and Lower Bounds for Field
Transmittances*; 10.1.7 Three-Dimensional Vector Nodic Fields;
10.1.8 End Corrections for the Inertance of Channels; 10.1.9 Multiport Vector Nodic Fields; 10.1.10 Summary

735

10.2 Discrete Compliance Fields ...................................
10.2.1 Generic Relationships; 10.2.2 Examples with Geometrically
Varied Capacitance; 10.2.3 Thermodynamic Coupling Between Mechanical and Thermal Energies; 10.2.4 Summary
10.3 Displacement Modulated Inertances .........................
10.3.1 Inertances Dependent on Holonomic Constraints; 10.3.2 Inertances Dependent on Nonholonomic Constraints;
10.3.3 Summary
10.4 Linear Multiport Fields ........................................

752

10.4.1 Mutual Inertances and Resistances: the Real Electrical
Transformer; 10.4.2 The Rigid Inertive Floating Link; 10.4.3 MultiCoil Transformer with Perfect Flux Linkage; 10.4.4 The Piezoelectric Transducer; 10.4.5 The Thermoelastic Rod; 10.4.6 The Piezomagnetic (Magnetostrictive) Transducer; 10.4.7 Generalized Linear
Media; 10.4.8 Reticulation of General Multiport Fields; 10.4.9 Further Equivalences; 10.4.10 Summary

760

771

CONTENTS

~x

Chapter
11 INTRODUCTION
TO DISTRIBUTED-PARAMETER
MODELS
...............................................................
11.1 Wave Models with Simple Boundary Conditions

.................

795
796

11.1.1 Comparison of Lumped and Distributed Models; 11.1.2 The
Pure Bilateral-Wave-Delay
Model; 11.1.3 Analysis of the Pure
Bilateral-Wave-Delay Model; 11.1.4 Fixed and F~ee Boundary Conditions; 11.1.5 Fourier Analysis with Fixed or F~ee Boundary Conditions; 11.1.6 The Hodograph Plane; 11.6.7 Summary
817

11.2 One-Dimensional Models ......................................
11.2.1 General Formulation; 11.2.2 One-Power Models; 11.2.3 Symmetric One-Power Models; 11.2.4 Multiple-Power Models; 11.2.5
Summary

825

11.3 WavePropagation ..............................................
11.3.1 Simplest Model: Pure Transport; 11.3.2 First Modification:
Thermal Leakage to a Constant-Temperature Environment; 11.3.3
Second Modification; Thermal Compliance in the Walls; 11.3.4 Dispersion and Absorption; 11.3.5 Group Velocity*; 11.3.6 Summary
11.4 One-Power Symmetric Models ................................

837

11.4.1 WaveBehavior; 11.4.2 Energy Velocity in Conservative Media*; 11.4.3 Boundary-Value Problems; Transmission Matrices;
11.4.4 Exponentially Tapered Systems*; 11.4.5 Summary
11.5 Multiple-Power Models .........................................

858

11.5.1 Symmetric and Anti-Symmetric Variables and Models;
11.5.2 Case Study of a Degenerate System: A Counter-flow
Heat Exchanger; 11.5.3 Case Study of a Symmetric Model: The
Bernoulli-Euler
Beam; 11.5.4 Summary
Chapter
12.1

12 THERMODYNAMIC
SYSTEMS .......................
The Convection

Bond and Compressible

867

Flow .............

867

12.1.1 Flow Through a Port; 12.1.2 The Convection Bond; 12.1.3
The RS Element for Fluid Flow; 12.1.4 Summary
12.2 Heat Interaction
and Junctions ...............................
12.2.1 The Reversible Heat Interaction Element; 12.2.2 The General Heat Interaction
Element; 12.2.3 The HRS Macro Element;
12.2.4 The 0-Junction for Convection Bonds; 12.2.5 Merging
Streams: the 0S Junction; 12.2.6 The 1-junction and IS-Junction
with Convection; 12.2.7 Summary

876

12.3 Case Study with Quasi-Steady

891

Flow .........................

12.3.1 Bond Graph Model; 12.3.2 Irreversibilities;
tation of Results

12.3.3 Compu-

...1 Application of Simple ThermalCompliance........ 979 919 943 968 APPENDIX D THERMODYNAMIC DATA AND COMPUTER CODE .................9..4...5 Programs and Data for Water Index ... ..... 12.... Data Files..... 12.2... 12. 12..6......1 Models With Two Degrees of Freedom ....6......2 Energy of Multiple Species.. 12... Scalar Calculations...5 Application to Water.3 Data for Refrigerant R-134A..9 Pseudo Bond Graphs for Compressible Thermofluid Systems. 983 D....... .5......1 Energy of a Pure Substance.4 Data for Refrigerant R-22.....5. 997 ..............I.Forced HarmonicMotion B.... 12...6 Case Study: a Refrigeration Cycle.7 Treatment of a Quasi-Steady-State Fluid Machine. B...4........1 Normal ModeVibration....... ComplexNumbers.. 12. D..xx CONTENTS 12...2 HelmholtzAnalytical Formfor State Properties... 12.....5....4..5 Reaction Rates... 12... Variables..2 Programs and Data for Refriferant R-12..4...8 Considerations of Reversing Flows..2.6........2......4.....2. B....3 Stoichoimetric Coefficients and Reaction Forces.....4........6 Case Study: A Piston-Cylinder Compressor.... 12.6. 12.12..1 Programs and Data for Air and Components. B.4 ThermodynamicCompliance ...6....4.....Arrays and Matrices.. 12.5.....4 ChemicalEquilibrium.....6.... 12.1. 12.....1 Modal Motions.2 The Initial Value Problem.8 Fluid Inertance with Compressible Fluid Flow.. 12.3 Application to Gases. Evaluating and Plotting Functions.5 Evaluation of ThermodynamicProperties .7 Application to the Liquid Region....2..... B... 12.... B...4 Modal Damping... 12.... Summary 12.....7 Modelsof Reactions With Mass Flows..6 Systems with Chemical Reaction* . 12... 12.... 12.2 Causality........4 The CS Macro Element.....5....5..2.. Control Flow Commands..4.. 12.......5 Example Using MATLAB..........6. 12.4....2.. B... D............6 Mode Reduction 965 965 APPENDIX C LAPLACETRANSFORM PAIRS .6 Models of Reactions Without Mass Flows..1 The Most Commonly Available Analytical Form for State Properties.......8 Summary 897 APPENDIX A INTRODUCTIONTO MATLAB.2 Higher-Order Models ..4 Application to Refrigerants.......... 12... D..... B. Function Files.5 Computations for the Ideal Gas.6...3 Forced Response... 12.. CommunicationBetween Files 953 APPENDIXB CLASSICALVIBRATIONS... Script Files...3 General ThermodynamicCompliance...5... 12....4. B..10 Summary 12... Fitting Curves to Data. 12.5...... D... B...5....

ENGINEERING SYSTEM DYNAMICS .

The design process you are engaged in is summarized.Chapter 1 Introduction This Chapter is directed primarily at undergraduates taking a first course in system dynamics and related areas. surely the discrepancies between the desired and the actual behavior require further iterative passes around the loop. You might have a small team under your direction. namely the original specifications. and likely use a computer to help determine the behavior.1. you interpret this need in terms of specifications. as a good designer. you model the designs abstractly. so that you can compare them with the desired result. At this point it is unclear which of the solutions are feasible. in Fig. If instead you follow the analytical option. At some point you decide either that . and what values of their various parameters would be necessary or desirable. You are directed in turn by a managementthat has a limited understanding of the technical possibilities. including quantitative performance objectives. or it may be an entire production machine. Your responsibility is to shepherd a new product idea from the conception stage to the production stage. that is a list of the specific functions the product should exhibit. If you follow the experimental option shown in the diagram. so you also play a role in choosing the product ideas to pursue and the marketing strategy. Howdo you do your job? You discover a need. Your job title is Project Engineeer. you conceive possible solutions. you display the results of the experiment or the analysis in an intelligible manner. 1. preferably several of them. and . machinists and production specialists. albeit in a very cryptic fashion. or one is presented to you. Part of the success of the company rests on your shoulders. It addresses three questions: Whyare you taking this course? What style of study works best? Howcan you reliably treat dimensions and units? Advanced students may proceed directly to Chapter 2. you construct and test physical models of the proposed solutions. It may be a component such as a pump or an automotive strut. or you might work alone. but in any case you have access to technicians. As designer. it may be a machine tool. if any. the quintessential assignment for a mechanical engineer or an engineer in an allied field such as aeronautics or agriculture. Next.the design has succeeded or failed. Then. In either case. you must then decide what to do next.

It is the modeling that is inherently difficult. even if it is a revolutionary advance in its field. that is. details of the system becomeof interest. however. You will spend more time constructing and executing analytical or computer models of your designs than you will conceiving them. Further. An unnecessarily complex model. This text deals largely with systems that can be characterized by how they . that makes engineering an intensely human activity. It involves intent. for it is not a "deterministic process. it is the "existential pleasure. as a final design is approached. produces unnessesarily complex results. Considerable attention also is needed on the analytical methods available to reveal this behavior. even if it is the abandonment of a poor idea that has been intriguing your boss. the construction of conceptual abstractions with well-defined meanings that behave in an appropriately approximate way to the real thing. the intended decision likely is simple: whether or not to pursue the particular concept further." the spice. time costs money. It is an art based on science. Whenyou first go around the design loop.2 CHAPTER 1._ computation1I modeling["Figure 1. Later. and how to predict performance with this ~nodel far quicker and cheaper than you could get a working physical model built and tested. perhaps requiring a rather fancy model. Together with the genius required to conceive new products.1: The engineering design process you go on to another design concept or another project. Howcan you carry out the analysis process rapidly as well as effectively? The answer is that you learn how to construct an abstract model of nearly any system you may design. and your competitors are not standing still. A crude model would be appropriate. even if constructed readily. A successful conclusion is rewarding to you and your company. you should learn how to make the abstract model appropriately simple. physical insight and experience. and can be expensive to execute computationally. Emphasis is placed on the process of modeling. need ~1 problem -1 lbrmulation ] I specification ~ ~ INTRODUCTION conceptualization solution ] experimental option t~s~ data reduction collection.I modeling analytical option analysis and I~ I conceptual L. whether they exist already or are in the conceptual stage. however. To be successful it must not take too long to accomplish. You should keep in mind the decision toward which the model is created. L I physical l. The purpose of this text is to empoweryou to analyze real engineering systems. recognition and judgement. Efficient design implies quick analysis.

justifying the title system dynamics. The driver then can release the energy stored in the accumulatorto achieve rapid acceleration. The vehicle has mass. The fluid drives the vehicle at a variable speed v through the use of a hydraulic motor with a volumetric displacement (volume of fluid passed per revolution of the shaft) that is varied by the driver. such as the informational considerations of computersystems. Fluid is stored in a hydraulic accumulator(a tank containing a compressedgas and hydraulic oil separated by a bladder or piston) at a high pressure. To reduce this time. are not addressed. It cannot accelerate the vehicle from rest to 27 m/s (about 60 mph)in under 18 seconds. etc. unlike conventionalbrakes. P. economicor political systems.1 Example As an exampleof the process. A wide variety of mechanical. v ~ The motor has a fixed field and a 12 volt armature voltage.1. e. particularly in Chapter 2. fluid. either in large or small quantities. thermodynamicand hybrid systems are included. Youquickly discover that a motor with a maximum powerof 30kWcan drive the vehicle fast enough. he can also makethe displacementnegative.which has inertia and resistance. Youwouldlike to have answers to such questions as: Howlarge (and heavy) a hydraulic accumulatoris necessary to accelerate the vehicle to 27 m/s in less than 6. without requiring a larger motor? Whatpressures should be used?. Someof the systems are in a time-invariant state.What maximum volumetric displacements (sizes) should the pumpand motor have to accomplish this? Howshould the volumetric displacement of the pumpbe automatically controlled? Whatwouldthe energy efficiency of various schedules of speed be? Ultimately. imaginethat you are a project engineer developing the basic concept of a vehicle poweredby batteries and a DCelectric motor. Systems that do not operate because of energy balances. however. This feature. electrical. 3 EXAMPLE employenergy or power. Its shaft rotates at a variable angular velocity $. shaft ~ accumulat°r/""x flow. called regenerative braking. Qp flow. Most of them are time-varying. is this concept worth pursuing? As a first step in answeringthese questions.1. Further. traffic systems. driving a hydraulic pump. recovers the energy of motionrather than dissipating it. Qm[ motor 1 velocity. frictional drag and an additional hill-climbing load. even with a perfectly variable transmission. you makea simplified modelof . braking the vehicle by having the hydraulic motor pumpfluid back into the accumulator.5 seconds. you consider the electrical-hydraulic drive represented diagramatically as: e RI t fixed field "i~¢. 1. at a steady 45 m/s (about 100 mph).

INTRODUCTION the system in the form of a bond graph: R. Ip. Someof the componentscan be described by parameters (constants): Ri. seconds A speed of 27 m/s is indeed achieved in 6. but recovers afterward. The values of TM and Tp are varied to control the flow of energy to or from the accumulator. The displacement is then changedto a certain larger value. Thus a subsequent burst of acceleration also wouldbe robust. Rp. if you like. This control also affects . The modelleads directly to a computersimulation for the dynamicperformanceof the systemfor whateversituation is of interest. you can simulate hundredsof different situations. The pressure and energy in the accumulatorfalls during the acceleration. Since each run takes only a couple seconds.4 CHAPTER 1. adjusting the various parametersuntil you are satisfied with the result. m/s 2O 10 0 0 50 100 150 time. G. R. by setting the volumetric displacement of the motor at a particular value. RL A bondgraph is a minimalist representation of the implied model. resulting in the samefinal steady speed. Its structure also directs in routine fashion the assemblyof the corresponding differential equations. Others require more complexconstitutive relations: C and RL. due partly to the automatic control of the pumpdisplacement that can be designed partly through analysis of the bond graph.5 seconds. and IL. Here are typical results after someof these changes have been made: 5O [ 40~ 30 f motor speed ~/10. rad/s /SP~P-’~’~o=7MPa speedv.

1.5 seconds is set to the value which is found to produce the maximumpossible steady speed. Most of the problems you will address in the course will be considerably shorter. fluid mechanics.1. The pressure drops to its minimum. MODELING AND ENGINEERING 5 SCIENCE the speed of the electric motor. As a student. etc. which of course also involve factors not discussed here. thermodynamics. at 507o efficiency.2 Modeling and Engineering Science Real engineering systems. The control successfully prevents inefficient speeds (below tad/s) and achieves the maximumpossible 947o efficiency at the steady speed. which is the charging pressure of the accumulator (the pressure when it is virtually empty of hydraulic oil). and address some comparable problem as a project after completing Chapters four and five. the displacement of the motor after 6. These textbook systems are in fact highly abstracted mode~s. As project engineer. however. seconds The motor speed becomes the 250 rad/s for which it produces the necessary maximumpower. do not closely resemble the pristine systems presented in introductory textbooks in mechanics. weight and reliability. . You hopefully absorbed some of the modeling idea. you are now in a position to render some judgments. rad/s t~~~pressur e P/IOPo Po =7 MPa oo 150 time. which is the desired 45 m/s: 5O speed v. usuMly largely pictoral. that convey a unique meaning. The motor is 507o efficient when its power is maximized (at 250 rad/s) and 100%efficient when its power approaches zero (at 500 rad/s). such as cost. In another simulation. but what you really practiced was the application of physical principles to translate the given model as expressed in one language into another language. m~ 40 30 20 speed ~/10. you should be able to address some simpler problems in this manner after studying Chapters two and three. like the example above.2. presented in the form of some language.

Nevertheless. Structural science. But the gap between real engineering systems and approximate abstract models must be bridged. Not all engineers today are competent to do this with muchgenerality. The language chosen to represent these elements should be general enough to encompass virtually all systems which centrally involve the management of pow.3 Modeling Languages A limited number of generic building blocks will be identified from which you will construct models. Most non-structural engineering in the nineteenth century evaluated design concepts largely through experiment. became particularly well developed. The Boeing company.6 CHAPTER 1. for example. Linear equations were emphasized. The analysis of dynamic systems lagged. Experiment need be employed only sparingly. The language also should be chosen for the ease with which its objects can be reduced to mathematical form. Such a language allows you to take full advantage of the analogies that exist between seemingly different engineering systems. 1. Perhaps the . because only they could be solved routinely. routinely comparing the structure of the model or its consequences . now designs and builds its most complex aircraft without a single prototype being constructed and tested (although key component~are tested separately). The new expression was intended to reveal certain features or behaviors of the model which at first were hidden. This ap-. the goad of increasingly complex military hardware in World War II spawned a formal systern science. but those who can are in a different league from those who can’t.most serious hazard in modeling is the threat of making an outright mistake. The engineer then can adjust its parameters until the result is satisfactory. saving time and money and sometimes reducing danger. based on differential equation models. because of the difficulty in solving the describing differential equations. sometimes just so that their behavior could be predicted. Today an engineer can construct a computer model of nearly any proposed system. for example. The advent of the computer removed the computational roadblock. largely either to understand phenomenawhich defy analysis (often using relatively simple apparatus) or verify a final design. The science of thermodynamics. albeit often with considerable tedium. was born of the need to understand steam engines better. Everyone is an occasional victim. Alternatively. or until it appears that some different approach to the problem should be sought. and predict its performance far quicker and cheaper than a prototype could be built and tested. proach helped you develop a working understanding of physical principles that must be acquired before you can advance to considering real systems. the expert keeps a commonsense vigil. Slowly it became apparent that the analytical option can substitute. so critical to safety. components and systems were designed explicitly to act nearly linearly.er and energy. INTRODUCTION which usually was mathematical. and narrow enough to impose the conservation of energy as the default option.

4 Modeling for Control A model or its direct consequence is itself built into a control system. Considerable use is made of sketches. fortunately. but these languages tend to obscure mistakes. Several languages apply to specific types of systems. possible. Bond graphs are used by a significant fraction of mechanical and electrical engineers. A more generic language is sought." since everything dissolves in it. diagrams and plots. and minimize mistakes. MODELING FOR CONTROL 7 with expectations. Similar languages apply to classes of mechanical systems and to hydraulic systems. a principal modeling language is desired that automatically conserves energy. becoming invisible. and allow routine conversion to equations or equivalent computer code. it has not been taught widely in undergraduate core courses. however. Mathematics and derivative computercode are the final language for most models. and is not to be dabbled in lightly. The circuit diagram is prototypical. This conversion is made as automatic as. unless it is directed explicitly to do otherwise. Where practicable. It lacks adequate generality to represent all systems. Mathematics jokingly has been called "the universal solvent. Since the subject is systems that control power and energy. by the biennial International Conference on Bond Graph Modeling (ICBGM). compressible fluid flow and chemical reaction. It is helpful also if the language identifies the critical geometrical and analagous constraints of such systems. which today means most systems. The language has been extended to model systems with heat fluxes. The modeling language or languages you use to represent models also greatly affects your propensity to make and discover mistakes. and even by some life scientists. This book employs a graph-based approach to modeling. in the form of the controller with its control algorithms or alternatively its hardware configuration. Bond graphs were devised in their basic form in 1959 by Henry M. The models for most dynamic systems are converted to differential equations. you should knowthat the control engineer typically succeeds or fails according to the workability of the model or models that he devises to represent . By making its mastery a manageable and natural task. Regardless of whether you study the introduction to control in Chapter 8. the system behavior implicit in a graphically expressed model is also determined graphically. it is desired that the language be both graphical and precise in its meaning. Such a language should be optimally efficient.1. Finally. Paynter at MIT. the electric circuit diagram often is applied by analogy to represent non-electrical phenomena.4. since 1991. Graphical languages comunicate better to your senses than mathematics. 1. Actually. It has been promulgated by thousands of professional papers and. however. They fail to address interdisciplinary systems. Although the advantages of a knowledge of the bond graph language are well known.such as heat transfer or acoustics. this book aims to increase its popularity for the benefit of engineering. which are then solved. and is awkwardeven when it is usable. Such a language exists.

The controller must be tailored to the system it controls. covers a thousand times more situations than possibly could be addressed by sample problems. Typically. recall of Roger Penrose. He or she views these detailed and often voluminous patterns.” the picturing in your mind’s eye of what really is going on. on the other hand. The output of the controller then is sent to the plant for action. Without them. or “plant. despite your earlier experience. The numbered examples. This insight. Its cost depends on the complexity of the model. the most practicable design results from a compromise between accuracy and complexity. develops problem-solving prescriptions by examining the solutions t o sample problems. They represent only a first step. however. suggested in part (a) of Fig.5 A Word to the Wise About Learning Your primary challenge as a student is to learn how t o approach problems that you. 1. guided problems and assigned problems. The difference between these signals is an “error” which is used as the input to the controller part of the brain. have never seen before. A person engaged in the problem-specific or training mode of learning. Oxford University Press. as the information needed for success. which often is a measure of the desired performance of some plant variable such as a speed. Shadows of the Mind. The brain usually measures this “output” variable. and should be viewed as integral parts of the text.2. . Excessively detailed focus on solved problems would distract from your developing the essential “physical thinking. The celebrated mathematician Roger Penrose’ claims that human understanding and even consciousness itself are non-computational. They logically may be skipped. and compares it t o the command signal.” with a brain part. rather than the relative simple underlying concepts. comprise the essential skills. The breadth of the problems in this text may help convince you of the futility of reliance on superficial procedures. together with a firm grasp of the grand conservation principles. This book complements its direct presentation of topics with four different types of problems: case studies. INTRODUCTION 8 the behavior of the system to be controlled. Its design depends on a model of the system created by the engineer. and maybe everyone else also. 1994. an angle or a temperature. Its performance hinges on the accuracy of the model.C H A P T E R 1. are sample problems that merely demonstrate application of the basic concepts and procedures. Control systems combine a muscle part. that is. Case studies typically introduce and illustrate new ideas. 1. numbered and highlighted examples. The brain part is directed to act by some command signal. This feedback scheme also potentially helps correct for unwanted effects due to external disturbances on the system. The detailed minimalistic procedures that are the focus of many elementary courses address only specially conditioned and presented problems that could be solved by a clerk with a fancy computer program. cannot be simulated by a Turing machine (digital computer). Real-world engineering.

but are suggestive illustrations that help you develop your integrated network. suggested in part (b) of the figure. . Engineering becomes heady stuff. Is your pleasure growing with each successive course? To learn the true meaning of the core concepts and procedures you must actively solve relevant problems without reliance on highly specific patterns. solved problems. on the other hand. If some element is forgotten. rather. The variety of situations addressable through the learning mode. The problem-general learning mode. guided problems and assigned problems are not definitive patterns t o be imitated. grows exponentially with the number of the core concepts and procedures understood. which is briefly suggested at the outset.1.2: Contrasting modes of learning all existing engineering formulas is useless. You may then compare your solution with one given at the end of the section. There is no substitute for direct observation and contemplation of basic concepts. the structure becomes an increasingly redundant network of ideas. and the flaw itself often can be repaired without recourse t o outside information. Case studies. Enriching physical demonstrations also can help greatly. You must. recognize what basic concepts apply t o your real-life problems. stores the information needed for success in an integrated network of basic concepts and procedures. In this information age there are many ways you can find the details. and proceed with a list of “suggested steps” that is less than a unique prescription for a solution but is intended t o relate the solution t o the core concepts and procedures. It is inefficient to infer reality solely from its shadows. Fbrther. The “guided problems” in this text start with a problem statement.5. a detour around the flaw usually can be found. A W O R D T O T H E WISE ABOUT LEARNING sample problem sample problem sample problem sample problem 9 (b) problem-general or education mode (a) problem-specific or training mode Figure 1. A small set of concepts casts myriad shadows in printed solutions t o relevant problems. Each guided problem exists for a purpose. It may seem that the amount of specific information you need t o recall actually decreases with experience.

1. These equations are solved for unknowns. to substitute the dimensions of each of the terms.3. rotates concentrically in a sleeve or cylinder of length L and diameter d + 2e. if you have not done so already. INTRODUCTION Figure 1.3: Journal bearing for case study (Some guided problems are indicated as advanced.1) Mo . This space is filled with oil of absolute viscosity #. as functions of the knowns. You know the following dimensions: .10 CHAPTER 1. or dependent variables. usually letters of the Romanor Greek alphabets. which are written on the left sides of the equations. The shaft has diameter d. Consider that you are asked to find the torque required to rotate a lightly-loaded shaft in a well-lubricated journal bearing at the angular velocity u~. These dimensions. For only the engineers. however. M0: (1. within the several assumptions upon which it is based? You know that most potential mistakes violate dimensional consistency. A case study illustrates this advantage. before substituting numbers to get a specific value for the torque. which are grouped on the right sides of the equations. so that e << d is the radial clearance between the shaft and the sleeve. 1.) Perhaps you deduce the following relation to approximate the torque or moment. but is this one correct. The experienced engineer always keeps the dimensional equations in mind. in which numbers are represented by symbols. for they give him or her a tremendous advantage over the mathematician in the use of equations.6 Treatment of Dimensions Engineers and mathematicians write algebraic equation~. which you may address in a first course in fluid mechanics. Substitution of the numbers for the symbols then produce a second type of equation which gives numerical answers. or optional. or independent variables. you will usually benefit from the attempt. so you proceed.) You defeat their primary purpose by looking at the given solution before attempting to create your own. comprise the terms of a third type of equation. Even when you fail to produce a solution. as pictured in Fig.. (This is a classical problem. which themselves can be represented by words or symbols. do the symbols carry added meaning: physical dimensions.~r#d2 Lu~ 2e No simple model can be exact.

it is necessary to work with a minimal consistent or primary set thereof. which again is proportional to d. Finally. since they can be expressed in terms of the consistent set. The proper torque therefore is Mo. wheneverforce F appears in Table 1. perhaps. which follows from Newton’s second law: F = rna. The namesof someof these are listed in the first columnof Table 1.1. (The other columnsare discussed later. you reason as follows: the shear stress on the surface of the shaft is proportional to the velocity of its surface. d. TREATMENT OF DIMENSIONS 11 ~r has no dimensions L. and symbolstherefore are listed in the second column. Other dimensionsare called derived. Youthen return to your derivation. the term d2 in the numeratorof equation (1. whichalso is proportional to d.7r pd3 Lw (1. Therefore. The shear force also is proportional to the wetted surface area of the shaft. however. This scenario is typical of a goodengineer. which you knowis the correct dimensionsfor torque.1. 3 xx time You know.4) 4e Youare finally ready to substitute the numericalvalues of the parameters.3) gives force x length. whichin turn is proportional to the diameterd. Somethingis wrong! The equation for Mo~nust be changedto increase its dimensionby one length.6. this set of dimensions substitues M. In order to deal with dimensionseffectively.2) where the square brackets around Mdesignate "dimension of.x length 1 [M0] . equation (1. length (L). Youtherefore re-visit your derivation.3) [M0] = forcelength x time length 3 = force. the torque equals the shear force times the radius." Cancellations give (1. you substitute to get force x time length ~. electric charge(Q) and temperature(0). Specifically.~ length length time’ (1.1.1) should be d3. M. Withthis correction.) A second primary set of dimensions replaces the role of force with mass. but have gotten into the habit of catching them by checking dimensions. electric circuit and thermal systems is force (F). Or. The most commonand recommended primary set of dimensions in use by engineers today for treating mechanical. which is d/2. that the torque or momentM0has the dimensions of force x length. not force. and find that you forgot to multiply the shear force by the radius. e have dimensionof length a~ has dimensionsof l/time (giving units such as rad/second) # has dimensionsof forcex time/length squared Therefore. This so-called "absolute" consistent set of dimensionssometimesis featured in ele- . Goodengineers are humanand make mistakes.L/t"-.time (t).

ft second.1 Dimensions mass (m) angle velocity angular velocity acceleration angular accel. m second. or energy as having the dimensions ~-. K coulombs.se/ft= (radian) slugs ft/s 1/s (rad/s) ~ ft/s 21/t ft. dynamics texts tend to use the recommendedforce-based American Standard system. A generation of engineers have been confused.lb/s "~ lb/ft ft.lb/R mentary books on dynamics.L/Q N. It abandons much of the benefit derived from using dimensions. and in practice is rarely used. lb foot. with both mass and force.L F. For example. Fluid mechanics texts tend to use slugs.s2/m = kg (radian) m/s 1/s (rad/s) ~ 1/s Nm= J (joules) N-m/s= W (watts) N/m’-> -.7 Treatment of Units Most implementation of dimensions is carried out through the use of a specific system of units. it is relatively awkward for most engineering systems.andmass-based. This attempt prevents critical cancellations from being made. R and units: F. s degrees Kelvin. I Am.t2/L Lit 1/t ~ lit F. American Standard or British Gravitational units. Std. Rankine. N meter. Three systems are most common. particularly when interdisciplinary systems are involved. engineers do not often think of pressure as having the dimensions M/t2L.L/t ")F/L F. C Q INTRODUCTION and Units name I symbol I SI units primary set of consistent dimensions and units: force length time temperature electric charge derived dimension 1.L/O Q/t F.lb ft. energy power pressure entropy electric current electric potential F L t 0 newton. which if not converted represent a mass-based American Standard system. Old English Engineering units are by tradition mixed force. The first two can be treated as either force-based or mass-based.Pa (pascals) N. and are addressed here: SI units.12 CHAPTER Table 1.= J/K C/s= A (amperes) N-m/C= V (volts) lb. 1. thermodynamics texts tend to use pounds mass as part of a mixed old English Engineering system. Nevertheless. and the old English Engineering units. units pound force. and as a result are not listed in Table 1 but are treated separately below. These texts often fail to adequately delineate force-based and mass-based use of the SI system. also. s deg. M.m/K. . L2 /t Beginners often try to use a mixed system of diinensions.

gc m W -.174 ~ 1.1.174 lbm-ft 2 lb. the numerical values will be properly represented.0000. Values and units ~re given in the Table 1.81 ~5 kg. the role of mass should be replaced by force through the use of Newton’s second law.s of Mass to Force units of m W gc g 2lb.09 -~. lb in. force units are.0000 sluglet. 386.5) where gc is a constant of proportionality that may or may not have the numerical value 1. however.s ft . which are dimensionless though not unitless). which is a good general policy in analysis. except in the cases of the more careful thermodynamics texts in which g~ appears already. the units of m~s will be converted automatically to equivalent units using force. As advocated above. containing both units of mass and force.= --. unlike the fundamental dimensions. lb-s" 32. Use of the method above automatically replaces its pounds mass (lbm) with equivalent untis using pounds force (lbf. or just lb). in general.7. which can complicate their treatment.174 ~ gc m 1. The old English Engineering system of units by tradition is a mixed system.m 1.s ft 2 lb.6) Whenever you face a mass m in an equation.174.2. so use of the recommended method produces a symbol for every number (other than true conversion factors. TREATMENT 13 OF UNITS The confusion regards the treatment of mass.s slug ft lb ft 32. where 7 is the weight density).s "sluglet" in. m F = --gc a.g. Force units do not necessarily equal mass units times acceleration units.1. you may substitute either m/gc or Wig (and p/g¢ or 7/g for p.in.0000 2 N. second.0000.0000 slug.2 Units and Values for Conversion Equivalent units of values and units of m L F kg m N 9. The system also uses British Thermal Units (Btu) for energy. Table 1.ft 2lb. The result is virtual conversion to the force-based American Standard system. gc g or (1. regardless of the system of units being employed. The result is twofold: first. as desired and as listed in Table 1. lbm ft lb ft 32. not 1. (1. The standard weight W is therefore related to the mass m by m W = -. Note that the value of g¢ in the old English Engineering system is 32. Rather. merely proportional to the units of mass times acceleration. In general.s in 2lb.

14 CHAPTER I.1 Evaluate the classical formula for the natural frequency.s _ 1s ~z.ft 2~.9 rad/s . This unit of mass has no official or recognized name. or one-twelfth of a slug. INTRODUCTION rather than ft-lb.1 = 2 rad/s symbol equation : units equation : number equation : ~ [Wn] =~/lb ~-’lbm 1 lbm.s"/in. as well as all electrical circuits. The SI system of units has d particular advantage for electromechanical systems: a volt-coulomb equals a Newton-meter (equals a Joule). Table 1. (b) m = 2 slugs.~ . This book employsSI units for all such systems. for the following particular values of the mass m that is attached to ground by alinear. EXAMPLE 1. This namelessness underscores the preference of engineers for force rather than mass units. Manymore physical phenomena directly involve force than involve mass. w.m ~ kg N-s" 1 s .0 lb/in.2 = 13. k = 8 lb/ft. (d) W= 2 lb.32. it equals the weight Win pounds force divided by g in inches per second squared. The units of mass then are lb. Therefore.ft lb.s 1 ¯ ~ .lb to Btu is a dimensionless conversion factor (with value 778. (c) m = 2 lbm. such as in "psi" for pressure. Engineers probably use inches more often than feet. k = 12 lb/ft. k = 1. but this poses no real problem since the ratio of ft. k = 8 N/m. The rather whimsical name "sluglet" is given in the table for this unit. Solution: (a) symbol equation : w~ = ~/-~-~ units equation : number equation : (b) symbol equation : units equation : number equation : (c) [w~] = a~n = ~/N 1 kg. spring with rate k: (a) ra = 2 kg.’ slug lb.16).~ ¯ 1 = 2 rad/s Wn = [w~] = w~ = ~/lb 1 slug.1 gives electrical units only in $I. = v/~-/m. = 12.

evaluate the velocity and note the units.C. Write the numbersequation for the Btu example. 2 P. D. It aiso showsthe role of true conversion factors. Write the units equation for the SI example. The conversion factor between Btu and foot-pounds should be included. and (a) Ah = 10 kJ/kg and (b)Ah = 5 Btu/lbm.2. Suggested Steps: 1. This ~ is a circular frequency.NewYork. 2SystemDynamics: a unified approach. Are the dimensions and units of the answer correct? 5. 4. In all the c~es the result h~ the units s-~. L. Guided Problem 1.12 = la. Write the numbersequation for the SI example. .1 This first guided problem illustrates the role of gc even whenthe mass does not appear explicitly in a formula being evaluated. Karnopp. Write the units equation for the Btu example. References Textbooks that can serve as references for this text have been authored by D. Note that the Ah given is the enthalpy per unit mass. 15 TREATMENT OF UNITS (d) symbol equation units equation 1 ft ~-in = 1~ [~n] =~/lb ~n’l-~’sZ’ 1i number equation : a~ = 1.) 3. To’convert from radians per secondto cycles per second(Hz). ~ ¯ a2. Wiley. 2. or the specific enthalpy. evaluate the velocity and note the units.7. Margolis and R. except they likely woulddirectly substitute the value and units a86 in/s ~ for 9. whichis the radian. Gawthrop and L. There are 778.9 rad/s The las~ case aboveillustrates the most common style of Americanengineers. Rosenberg.0. The k3 should be expressed in terms of the primary consistent units. replace this implied mass by m/gc to modify the formula. Unless you already are quite competent with treating dimensionsand units.16 ft. you are strongly advised to at least attempt the problem before looking at the solution on the next page.. division by 2n is necessary. Find the velocity v of a perfect gas in the throat of a nozzle.C.1. write the units for this factor in your equation. given the formula v = x/~-~--~.lb per Btu. Are the dimensionscorrect? Is there a need for a conversionfactor? (If so. a dimensionless angle is implied.2nded. Therefore.1990.

6 slugs/ft Solution to Guided Problem 1. Ap = 10 lb/in. 5 ContinuousSystemModeling.Amsterdam. Boca Raton.CRCPress.°F and a density of 1.5 A hydraulic fluid suffers a pressure drop of 2000 psi in flowing through a valve. Breedveld.s 2 " Btu . assuming a specific heat of 0. North-Holland. guest editor P. Find the maximum velocity of the mass assuming a spring rate of 20 lbs/in:. Iv] = Vl--~-~m.1 N m 2.5.com. INTRODUCTION Smith. 1991. Thoma. .C. v = ~/2. PROBLEMS 1. 1990.Springer-Verlag. Breedveld and G.Springer Verlag. 1999. 6Modelingand Simulation of Engineering Systems ThroughBondgraphs. Mukherjee and R. Find the temperature rise 3. an initial spring compression of 5 inches from its free length and a mass weighing 10 lb.32. Cellier.4 Find the velocity of the flow of water (weight density 62.1991.E. Prentice-Hall.16 = 500 ft/s a MetaModelling:Bondgraphs and dynamicsystems.m Ns2 ’k--~= ~3. DauphinTanguy.for Engineers.U. s For conference proceedings and more see www. v = x/2.4 F.1992.1. 1000 = 141.174 ¯ 778.s 5.lb ft 4. Neglecting any effect of compressibility or heat transfer. 1.v 328 n5/6.C. 1. lb.4 lb/ft 3) in the throat of a nozzle. NewYork. the dissipated energy simply raises the temperature of the fluid.10.ft ft.16 CHAPTER 1. 1996 4Simulationby Bondgraphs:introduction to a graphical method. Berlin. Karmaka~ Notable collections and reviews of papers are given in a special issue of the Journal of the Franklin Institute 7 and a book edited by P. of th~ FranklinInstitute.2 The energy of a compressed spring accelerates a mass. 3 J.4 ~/Btu lbm. 5 and A. NewYork. Neglect the mass of the spring. and the formula v = ~. given the pressure drop from the rest upstream state. 1.BondGraph.3 A DCmotor operates at 70% efficiency with a power supply of 24 volts and 2 amps. Find the velocity at which it can raise a 10 kg object.5 Btu/lbm. [vl--v /kN-m ~gg " kg. 1. 7j. 2. S BondGraphs.1 State the standard weight of a mass of I lbm.

This is usually not the case. The engineer also wants to know how to configure a particular system to achieve the best possible behavior. In somewhat more complex instances the source and the load may be interconnected by a third component. generally change over time. into components with explicit boundaries. 17 . voltage and current. deals with components and systems that can be characterized in terms of variables that remain constant over time. or in equilibrium. a small disturbance on the system produces a radical change in the state of a system. Thus you are urged to pay serious attention. even if your primary interest is dynamics. however. Most of the mistakes that modelers of dynamic systems make can be traced to the geometric and quasi-geometric constraints of the system. Sometimes. such as force. Variables. These components or systems are said to reside in steady state. which apply to the equilibrium states as well. however. further. velocity. Each component is characterized in isolation. in which the variables change over time. such an equilibrium is said to be unstable. The challenge of modeling a system in equilibrium might seem vastly simpler than the modeling of a dynamic system. conceptually. or have one or more components in between to form a chain. justifying the name. and these characterizations are combinedto reveal the behavior of the assembled system. Sometimes more than one equilibrium exists for a given synthesis of components. This chapter considers sources and loads which either are joined directly. This chapter. In the simplest case only two components are defined: a source and a load.Chapter 2 Source-Load Synthesis An engineer usually starts to model an engineering system by subdividing it.

The Latin stem reti means (fish) net.1 Case Study: Induction Motor as a Source Consider an induction motor as pictured in Fig. In the study of statics and dynamics one draws a free body diagram of the system in question. 2. This is a complicated device. 2. therefore. In either case. Reticulation is not as straightforward as sometimes it may seem in practice.1: Partitioning 2. It should be no surprise.1 of the universe into a system and its surroundings System Reticulation In the study of thermodynamics. Conceptual partitions of space or bodies of matter also are basic to the study of mechanics.2. not to understand the details of its construction or operation. representing a region in space or control volume into which and/or from Which matter can flow.1. the universe is divided customarily into a system and its surroundings.18 CHAPTER 2. the boundaries of the system. to reticulate ineans to make into or like a net. SOURCE-LOAD SYNTHESIS surroundings~ syste~ ~ ) contro~ Figure 2. The system is characterized in terms . might be fixed or moveable. as you did in the study of statics and thermodynamics.nting a fixed quantity of matter or a control mass. The entire machine and its source of power is defined as the system. as pictured in Fig. This image helps you focus not only on the sub-systems. The system might be closed. or perhaps five or twenty different sub-systems. It is easy to overlook the simplifying assumptions that are implicit in this act. The process of endowing a system with structure often is called reticulation. that in the general modeling of physical systems one starts by carefully defining a system. with resultant trouble. The g0al is merely to characterize its behavior as it applies to users. letting only the shaft penetrate the system boundary or control surface. 2. In fluid mechanics one chooses a control volume.1. but also on the meshpoints or bonds of the net. by identifying a single system and its surroundings. represe. often called the control surface. On the other hand it might be open. The discussion here starts simply. giving a total of four different possible combinations.

~ (¢ is the rotational angle of the shaft). If the the work done is F58. (2. The shaft.1) To establish this result. in the shaft. is said to be the energy port or just the port of the system. usually measured in radians per second. or more precisely its intersection with the control surface. assuming ~ and M. The description "one-port model" distinguishes this representation from potentially more complex models having two or more defined ports. The first of two variables that most engineers would identify is angular velocity.1. a normal force F acting on pictured in Fig.3: Shaft and crank of appropriate variables that can be observed at the point of interconnection between it and the environment. as crank is displaced by an infinitesimal distance equals r 5¢.3.orque or moment.2. where 5~ is the corresponding angular .by the motor and propagated through the port (shaft) to the environment. Mine. I energy port (intersection motor system I of shaft and control I surface) Figure 2. oi: ’ ~ ~ ¢ M.~ are both defined as positive in the same sense a~ seen by an observer stationed near the motor: T~ = -~¢m~. The product of the two variables. SYSTEM RETICULATION control surface F ~ I r---"’--~ ..’ I I L 19 surroundings ~ ~ I shaft oC. is the power.2: Representatioh of an induction motor as a one-port system Figure 2. generated . But (is consider that the torque and the power arise from an imaginary attached crank of radius arm r. namely the shaft.. The second variable is the t. 2. P. M~.

5 150 f. or 7 ~ = Fr de/dr = Fr~. SOURCE-LOAD SYNTHESIS 200 2. rad/s 200 I 00 Figure 2. These two plots foll(~w the convention that the horizontal axis (abscissas) represents the independent variable on the right side of the equation. the environment is adjusted until J¥/m reaches some desired value.20 CHAPTER 2.1). This perspective could be called existential. that is of cause and effect. ~ you will employ a causal perspective of a characteristic plot. and then Mmis measured. Imagine an experiment in which the environment is adjusted until ~ reaches some desired value. so the work becomes Fr 5¢.0 100 rad/s 0.~).5 0 5O 0 50 I00 150 ~. for specific computational purposes. Mm. however. Care must be taken because there are two values of ~ for some values of Mm.2) can be plotted. This experiment is then repeated for many other values of ~ until a complete characteristic Mm = Mm(~) " (2. with the axes merely reversed. This convention actually is quite arbitrary.0 2. however. The power is the work per unit time. (L/~ter.4. the two plots represent the same characteristic. (2. Thus. at which point ~ is measured. and the vertical axis (ordinate) represents the dependent variable on the left side the equation. Alternatively.4: Torque-speed characteristic 1. 2. the two plots are completely equivalent. the product Fr is the applied moment.3) The corresponding characteristic for the induction motor is shown in part (b) of the figure. The general idea of a characteristic overlooks any considerations of independence and dependence. as opposed to causal. A typical characteristic for an induction motor is shown in part (a) of Fig. establishing equation (2. Finally.5 Nm 1.) .0 of the induction motor system rotation of the crank and shaft.o 1. The consequence of repeating this experiment for many different values of Mmcan be written ~ = ~(M.

The outward-directed half-arrow defines the direction of power flow when P = Mm~is positive. collectively. The new system can be represented summarily by a second word bond graph: Mp LOAD ~ SYSTEM Like the motor. this model of the system has only one port. with the conjugate variables ~ and M.~. for short..5: Representation of the load as a one-port system Either plot completely describes the motor as long as the motion is steady (no acceleration or inertial moments). the machine does not run backwards. etc. now can be considered as the system. which is a shaft with an angular velocity ~ and moment Mp. and you are not worried about failure of the power or the motor. It is represented by the word bond graph Mm INDUCTION MOTOR ~ The port is penetrated by a power bond or. The role of the system and the environment thus are inverted. Note one difference: the power . the motor becomes the new environment.1. bond. indicated by the horizontal line. 2.1. This model of the motor is described as having only one port.2 Case Study: Water Sprinkler System as a Load Consider now that the motor drives a pump which pumps water from a basement tank through a pipe to some upper floor of building where it passes through sprinkler heads and extinguishes a fire. you do not care about heat and noise.2. SYSTEM RETICULATION 21 control surface h~ sprinkler __~ surroundings load system -~Q shaft pump [ Me ~ ]_water supply ~ Figure 2. This new equipment.

and therefore only one possible steady operating speed. there are no intersections at all. This description alone does not tell you how much water flow is associated with any particular combination ~. .0 1. if their respective characteristics are plotted on commoncoordinates. for the sprinklers being on the tenth.e. This is the only difference.22 CHAPTER 2.5 I 50 I I 100 150 ~. so that M. in fact. ilnplying that the state variables are unchanging over time.0 SYNTHESIS ~ twentieth floor tenth floor 0." Therefore. the resulting equilibrium must be represented by their intersection. Such an equilibrium is called a steady state. Repeating this experiment for the whole range ~ > 0 of interest produces a load characteristic.m 1.5 N. between a "source" and a "load.~equals .5 2. For the tenth-floor case. 2. the load torque exceeds the motor torque for all speeds ~ > 0. not out. respectively. at which the torque Mp is measured.3 The Source-Load Synthesis. Moreover.1. Three such characteristics are plotted in Fig. rad/s Figure 2. Case Finally. 2. The motor simply will not run. 2. just recognize that the faster the shaft turns. the particular motor is joined to the particular the following word bond graph: ]l~l INDUCTION LOAD MOTOR SYSTEM .6: Torque-speed characteristics I 200 for the load system product ~Mpis positive when power flows into the load. as suggested by These two sub-systems now share a commonspeed (~) and a common moment. as in Fig.6. SOURCE-LOAD thirtieth floor 2. the greater the volume flow rate of water will be. Mp. it doesn’t even recognize the existence of water. there is only one intersection of the characteristics. For the thirtieth floor case.7." It is now possible to run an experiment in which the environment of this load (i. It might even burn itself out trying. These characteristics also could be found by inverted experiments in which Mpis set and ~ is measured. For now. flow and moment. as with the motor.hip. the motor) is adjusted to set any desired speed. twentieth or thirtieth floors. ~ Study load.

7: Synthesis of the source (motor) and load characteristics M (h~ (a) . This is because someequilibria are unstable. change~ i Figure 2. 2.1.is not a sufficient condition to insure actual operation. Onesuch pair of source and load states near the equilibrium state ~2 is indicated in the figure as points (a).8. A . rad/s 200 Figure 2. The electromagnetically induced torque of the motor can be different from the fluid-induced torque of the pump. A straight lin6 drawnbetween the two points therefore must be vertical. Since the shaft is assumedto be rigid." thirtieth floor ~ twentiethfloor tenth floor 1.K>K . Mp.0 0.2. For the motor-pump-sprinkler system this means a departure from the equilibrium angular velocity. Both of these intersections represent equilibrium states. that is fail to persist. which meansthat the source speed equals the load speed and the source torque equals the load torque. M. however.5 load characteristics. There are two intersections of the source and load characteristics at the speeds labeled ~ and ~2 in Fig.5 0 I 0 50 I I II 100 150 ~. the source (motor) and the load have the same angular velocity. The source and the load do not have the same torque.Mp>Mm (d)’ I. Equilibrium. Prediction of whetheror not a particular equilibrium or steady state is stable requires consideration of a nearby non-equilibrium or unsteady state. 23 SYSTEMRETICULATION 2. so that the torques of the source and the load are different. (~.8: Stable and unstable equilibria for the twentieth floor sprinkler The case of the twentieth floor sprinkler is moreinteresting.

whether the shaft accelerates or decelerates. SOURCE-LOAD SYNTHESIS difference between the two torques causes the shaft. you can see that the motor torque. the net torque and the deceleration vanishes. in the event that some auxiliary means were available to start the shaft rotating faster than ~. A shaft penetrating this surface defines an energy or power port of the system. and therefore the system is unacceptable. It is in equilibrium because there is no force to make it fall over. tiny errors in the initial state and tiny disturbances are impossible to avoid. producing a net negative torque and a decrease in the angular velocity ¢. however. As with case (c). As the departure from equilibrium grows. Acceleration would drive the speed further above the equilibrium.2. as indicated by the arrow. despite the stability of the speed ~2. At the same time it indicates that the equilibrium speed ¢1 is unstable. the steady-state behaviors of each such system. which produce a deceleration. Again. for any and all environments. the acceleration vanishes as the speed aproaches ~)2. In practice. you could simply call them system A and system B. an acceleration will drive the speed away from it. This means that the motor will not start. Mm< Mp.lp. This check can be done by considering the non-equilibrium states (b). ~. but to make sure you ought to check what happens if the starting speed is somewhat below ¢2. . and the pencil soon falls over. with its attached rotors in the motor and the pump. to accelerate or decelerate. This conclusion is reinforced by consideration of states (c). the source (motor) torque exceeds the load torque. the torque slowing it downalso de. produces a small force in a direction awayfrom the vertical equilibrium. consider a perfectly symmetrical sharpened pencil placed perfectly vertically on its tip in a perfectly still room. as indicated by the associated arrow. On the other hand. Its environment. or be a one-port system. As the speed decreases. Real pencils in real rooms never stand on sharpened tips for more than a brief moment. (It could be made practicable. or the slightest puff of air. If the system is not influenced by any other external variables.) As a more familiar example of an unstable equilibrium. The equilibrium appears to be stable.24 CHAPTER 2. ¢2. ~im. Whenthe speed reaches the equilibrium. Here.4 Summary A system is identified by its control surface. it is said to have a single port. consider the rest case (d). For states (a). similarly. however. deceleration would cause the speed to approach the equilibrium. could be described as another one-port system. the destabilizing force grows also. is less than the pump torque. The behavior of this power port can be described by the conjugate variables torque and angular velocity.1. The slightest deviation from the vertical orientation. indicating instability.he equilibium speed ¢2. Further. You need to know. One such means is suggested in Guided Problem 2. so the deceleration decreases. therefore. 2. Which is the system and which is the environment is quite arbitrary. This observation confirms the stability of t. no matter how close states (b) are to this equilibrium. so the speed of the common shaft accelerates. Finally. the product of which is the power passing into or out from it.creases.

This determination requires information not contained in the characteristics themselves. The thrust and drag are forces that are analogous (similar) to torque.1 This first guided problem involves a basic source-load synthesis.2. . before examining the solution at the end of the section. (a) Plot the steady-state position. An equilibrium can be stable or unstable. The throttle then is suddenly advanced to 100%.9. as shown. Similar considerations apply to systems with interconnecting rods. Such a relation is called the system characteristic. and an arrow pointing to higher speeds. If the source torque exceeds the load torque. The pattern of such arrows for different speeds shows which equilibria are stable and which are not. The purpose of presenting this case study far transcends systems with shafts. speed of the boat as a function of the throttle (b) Assume that the throttle has been at 30% long enough for the boat to reach equilibrium speed. 2. as least roughly. For the case of the motor-pump-sprinkler this information is two-fold: the shaft is rigid. and the linear speed is analogous to angular velocity. The thrust produced by the propeller depends on the throttle position and the speed of the boat. as you shall see. but the torques can be different. The drag force on the hull also depends on the speed of the boat. depending on whether an imposed departure from equilibrium spontaneously grows or shrinks. Find the instantaneous acceleration. Two one-port systems connected at a commonport experience an equilibrium state or states where they have commonvalues of both of the conjugate variables. This situation can be represented by a vertical line segment connecting the two characteristics at the particular speed. An internal combustion (IC) engine drives a propeller that powers a boat which weighs effectively 4000 lbs.1. as determinable by experiment and as representable by an equation or a plot. and connect the resulting points with a smooth curve. so both source and load have a commonangular velocity. determine the intersections of the thrust curves with the drag curves. Guided Problem 2. fluid pipes and electrical conductors. You are strongly advised to carry it out. the combined system accelerates. these equilibria are represented by their intersections. Also. as plotted in Fig. cross-plot the speeds of these points against the throttle positions. determine the acceleration when the speed of the boat has increased half way to its final equilibrium. Suggested Steps: 1. To answer part (a). SYSTEM RETICULATION 25 can be represented by the relation it imposes between its conjugate variables. If the two characteristics are superimposed on a commonplot.

twentieth and thirtieth floors. The proposed motor. characterized in Fig. (c) Discuss the stability and any peculiar operation of these equilibria. 800 ¯ % throttle 600 60-- SOURCE-LOAD 100 90.10. .~ ~ ~">< SYNTHESIS drag ~ ~ force. (b) Find the equilibrium speeds for the sprinklers on the tenth. and repeat steps 2 and 3. Guided Problem 2. One solution is to replace the motor. respectively. using the proposed motor.1 2. the one with a series capacitance is switched out of the circuit by a centrifugal switch at 75%of synchronous speed. To answer the final question. 3. find the average of the initial and final speeds. Get the net force which accelerates the craft by subtracting the drag of the hull at the equilibrium speed for 30% throttle from the thrust at the same speed corresponding to 100%throttle. Its value depends on your effort. and compare to the original motor.2 This problem illustrates stable and unstable equilibria as well as basic sourceload synthesis.26 CHAPTER 2. which is of the standard shaded-pole design commonfor small induction motors. Use F = ma to determine the acceleration. (a) Estimate the maximumpossible powers below and above the switching speed for the proposed motor.. F 400 lbf 200 0 0 10 20 30 40 speed. It has two windings. Ignore potential problexns with overheating. 4. ~. ff/s Figure 2. Clearly the water-sprinkler system you have examined is unsatisfactory for the twentieth floor and above. 2. is of the capacitor-start induction type.~thrust 80 ~_~.9: Thrust and drag for boat of Guided Problem 2.

27 SYSTEM RETICULATION mainand auxiliary winding ~~’--~. represented by the solid line in Fig.switching 2.~..1 and Fig. All you really need to find the equilibrium points is the motorcharacteristic.rad/s Figure 2. 4. Superimposethe load characteristics on the motorcharacteristics. Notethat the vertical segment of the load charateristic merely connects its two end points.2 (d) As project engineer for the system. . and the three load characteristics given in Fig. 2. 2.9 is traveling at a steady 20 ft/s the throttle is abruptly re-set at 80%.1. use your common sense to make a suggestion beyond the formal scope of the analysis as presented.10: The motor of Guided Problem 2. 3. youwill see soIne graphical meansthat will help you zoomin on the maximaquickly. 2. PROBLEMS 2.8.~/~ I I I I I I 50 150 200 100~.7. Consider whether the nominalsolutions seem practical. and find the intersections. Later.0 /’//~ 0 -~1 0 -’~ main windi~ oNy~ .2. The maximumpowers can be estimated by trial-and-error guesses of points on the characteristics. (b) Find the instantaneous acceleration of the boat. Suggested Steps: 1. switching (rapid?) may Occur. recommend a course of action. Carry out a stability analysis as in Fig. operation cannot occur at an intermediate point. 2. (a) Find the thrust and drag forces before the velocity has time to change.1 Whenthe boat of GuidedProblem2. 2. Instead. Power is the product M6.10.

4 can deliver.3 Consider the motor of Guided Problem 2.delivered when the torque is maximized. SOURCE-LOAD SYNTHESIS 2. over the (b) Repeat (a) for the speed ranges 100 < ~ < 130 rad/s and 130 < 150 rad/s. (c) Estimate the time required for the system to accelerate the switching speed of 150 rad/s. average net torque available for acceleration speed range 0 < ¢ < 100 rad/s. 2. Also.m (a) Estimate the.2 and Fig. % 1 O0 .2 Estimate the ma~ximumpower 7)maz the induction motor characterized in Fig.28 CHAPTER 2. Determine the stabilities of the three equilibrium points. The load has an effective inertia of 0. floor sprinkler system. 2. 2.015 kg.4 An induction motor and load have the characteristics shown below. M~ ~::~ ~ m -- load otor SOLUTIONS TO GUIDED PROBLEMS ~ 30 ft/s20 10 0 0 60 0 throttle. and the associated torque Mand angular velocity ~. find the ratio of this power to the power 7)ma~ M. from rest 2.10 driving the twentieth’~. but have some inertia. The shah and the load are rigidly connected.

4. ~. the difference between the maximumthrust and the drag is 770 . Note that the points must lie on the descendingportions of the respective characteristics.0 original motor ~. rad/s 200 Something must be done to eliminate the rapid switching for the 20th-floor case.m 1. and have the powers.0 " 10 Points a and d are seen to be stable equilibria. which gives an acceleration of 280 × 32. rad/s 200 2~ 3. e d 50 100 150 ~. which would burn out the switch quickly..0 M. Fg 600 x 32. as discussed in Section 2. At this speed.2. 50 0 100 ~ 150 ~.1. They can be found by trialand-error. Subtracting 270 lb for the drag force at the same speed gives a net force of 600 lbs. 2.-main winding only ~. so the motor switches back and forth rapidly between these twostates. . in watts.. only states b and c exist. indicated by the numbers.490 = 280 lbs. 29 SYSTEM RETICULATION The 100% thrust curve must be extrapolated to estimate the thrust at 13 ft/s: about 870 lb.4. Better matching between one of these motors and the existing pumpnevertheless presents an alternative class of solution.25 ft/s 4000 Guided Problem 2. The simplest solution employs a more powerful motor.0 M. The intersection of the 20thfloor characteristic with the dotted switching line wouldbe a stable equilibrium if the dotted line represented a real characteristic.83ft/s 13 + 35 The average of the initial and final speeds is 2 . 4000 mg 3. In reality. ..2 main and auxiliary winding The three maximum powers are at the points indicated by dots.2 a .__2 ~. N. ! 327 2. and likely would not be any more expensive than fancier solutions ~vhich retain one of the existing motors. Me. N-m 1.24 ft/s. _ 2.

but you should overlook this possibility at this time. pressure and temperatureare efforts. Whichis the effort and which is the flow are determined by the second part of the definition. Thus.30 2. for example. Powerflows along a bond in one direction or the other. This idea of factoring power into a product of two conjugatevariables also applies to mechanicaltranslation. and the symbolchosen to represent the powerconjugate flow variable is written below the horizontal bondor to the right of the vertical bond: 1The traditional bond graph symbol for flow is f.4) Equation(2. Theyalso reduce the variety of functionally different componentsin a modelof an interdisciplinary system. SOURCE-LOAD SYNTHESIS Generalized Forces and Velocities The power that is transferred through the bond between the two sub-systems in the case study of Section 2. 2. Flows are normally defined as directed. q. electrical power." and the flow or generalized velocity. (There are infrequent instances in which someadvancedbond graphers mayconsciously invert the roles of effort and flow. labeled as "e. . Mand ~ qualify as effort and flow. mechanicalshear. since P = Me.1. fluid power.2.) Scalars such as voltage. the symbolchosento represent an effort variable is written abovea horizontal bondor to the left of a vertical bond. while efforts are normallyeither true scalars or are treated as scalars.1 Efi‘orts and Flows The powerfactor variables for a generic (or general abstract) bond are called its effort or generalized force. Conventionally. equals the product of the torque Mon the shaft and the angular velocity ¢ of the shaft. This book substitutes 0 to emphasize its relation to the displacement.4) represents the first of the two. and to reflect the practice in analytical dynamics. labeled as "0". parts of the definition of efforts and flows.~ Thus the power becomes ~ (2. behavefunctionally in the same wayas the correspondingvariables in an analogous mechanical"circuit. Directed quantities such as electric current and volume fluid flow are flows. also. for example." These analogies enable you to transfer your knowledgeof one type of physical or engineering systemto another. Observation of the scalar or vector nature of the resulting conjugate variables establishes analogies(special similarities) betweenvariable types across these different domains. heat transfer and other energy domains. analogiescan be drawnalso between the various types of componentsthat the bonds interconnect.As a result.2 CHAPTER 2. The variables and componentsin an electric circuit.

2. while the current 0 = i is equal in magnitude and opposite in direction. (2. represented by either the symbol 0 or the symbol i. . 31 GENERALIZED FORCES AND VELOCITIES (a) (b) Figure 2.2 Electric Conductors Each power flow in an electric e. 2. is shown in Fig. Current.11. the half-arrows should be placed on the flow side of the bond.One detail: according to a standard set by a committee during the llth IMACS World Congress (1985).2. and current. If e is a true scalar. 2.5) Voltage is a symmetric or scalar variable with respect to the conductor. The corresponding bond graph symbol also is shown. on the other hand. i: circuit is represented by the product of voltage. 0. one grounded. Consider the examples of Fig. signals that u is the effort (generalized force) while v is the flow (generalized velocity). A pair of electric conductors. 2. Note that the voltage e is the difference in electric potentials of the two conductors. including the power convention half-arrow. It is identified therefore as a generalized velocity or flow. and v below and to the right.2.11: Convention distinguishing generalized force (effort) ized velocity (flow) from general- e et i: Figure 2. In case (a) the positioning of u above and to the left of the bond. it is identified therefore as a generalized force or effort. it flows one way or the other.12: Power propagated by a pair of electric conductors The half-arrows on the bonds indicate the direction that power flows when T’ > 0. In case however. is anti-symmetric with respect to the conductor. represented in this book also by the symbol e (the sy~nbol v also could be used). this is also the direction of positive flow. In practice one usually does not bother to draw the grounded conductor in a circuit diagram. only the location of the half-arrow resolves an ambiguity to show the same association.12. but nevertheless tacitly assumes its existence.

5. The cable has a tension.6) . on the other hand. or effort. refers to both the action on the systemand the reaction on its environment.7) Note that if the cable or the rod is in tension (Ft > 0 or F~ < 0) and k > 0. 2. If the velocity of the cable or rod. ~ "~dt= q. is defined as positive in the rightwarddirection. Forceis defined classically as an action on a system describable by a vector.Ft. 2. as mechanicaldisplacement.13.32 CHAPTER 2. [ (2.In this symmetric view the tension or the compressionacts between two systems like a scalar. equals the net electric charge or electric displacement.2. The tension F~ or compressionFc.ql.3 Longitudinal Mechanical Motion Consider an inextensible cable or a push-rod that connects two sub-systems. Ft or F~ mustbe the generalized force. A . The variable 2 is a real velocity aligned with the bond. Therefore. q. SOURCE-LOAD SYNTHESIS control surface subsystem A I subsystem B velocity~ push-rod: F veloci SUBSYSTEM A SUBSYSTEM B Figure 2. Ft. with dimensionsof force. which happily is traditionally represented by the samesymbol. the powerconveyedfrom the left to the right becomes 17)= -Ftgc or 7)= Fc~:. The generalized velocity or flow is directed similarly. as illustrated in Fig.~ (2. that is being propagatedto the left. The push-rod can sustain either tension or compression. the poweris negative.13: Subsystemsinterconnected by a cable or push-rod The time integral of the electric current.the choice of variable shownis compression. The "powerconvention" (7) > 0 for powerpropagatedto the right) is indicated by the half-arrow on word bond graph.

In our immediate example. 33 GENERALIZED FORCES AND VELOCITIES control observer A tension axis of bond observer B Figure 2. Assuming P is uniform over the section. If their perceptions are identical. If the variable is i.14. so does the other. and velocity ~ = v. whichcan be important particularly whenP is small.4 Incompressible Fluid Flow The flow of a fluid through a pipe or tube can be represented at any cross-section or control surface as the sumof the flows through a bundle of infinitesiInal stream tubes. -~ section gives the total power integration (2.9) omits the transport of both internal kinetic energies. so does the other. Each stream tube acts much like a push-rod with compression force dFc = P dA. each with cross-sectional. they . one sees a motion directed away from himself.14: The perspectives of two observers at opposite sides of a bond "generalized force" or effort is a generalized concept of such a variable. if one observer senses tension.2. Anti-symmetry implies a flow or generalized velocity.8) and (2. A powerful way to determine whether a variable has the directedness of a flow or generalized velocity or the symmetry of an effort or generalized force is to consider two observers facing each other head-to-head and toe-to-toe on either side of the control surface or either end of the bond. Nevertheless. as suggested in Fig. Thus its energy flow rate or power is d7) = ~ dFc = Pv dA. as suggested in Fig. 2.2. 2. voltage is another example.15. They both perceive the variable in question at the control surface through which the bond penetrates. however. while at the same time the other sees motion directed toward himself.8) of this power over the 2Theapproximationof equations (2. where P is the static pressure. if instead he senses compression. and the variable is an effort or generalized force.2. area dA. 2. the variable can be said to be symmetric with respect to the observers.

10) Whichof the two power factor variables.5 Rotational Motion The power transmitted along a rotating shaft was seen in Section 2. as shown in Fig. this implies symmetry and identifies P as an effort or generalized force. Note that the time integral of Q (representable by the generic symbol q) is the net volume of fluid which has penetrated the control surface. Conversely. and angular velocity. Q=.2. Sometimes.16. (The analysis of compressible flow in Chapter 12 employs mass flow rate instead.15: Channel with elemental stream tubes of area dA 7 ) = PQ. the other sees counterclockwise rotation. _~. momentor voltage. angular velocity and electric current. ~: T’ =/~. (Check this out by rotating your pencil apply with considerable accuracy to most hydraulic "fluid power" systems.1 to be the product of the moment. 2. l~ke simple velocity. as a result. . . f vdA.I or ~. and thus is a flow or generalized velocity. ~’"" ~~ SOURCE-LOAD SYNTHESIS stream tube micro-bond: ~’ control surface integrated: P vdA P Q Figure 2. for then Q is uniform along the tube or pipe and across any throttling restrictions. (2.34 CHAPTER 2. More complete representations for compressible flow systems are presented in Chapter 12.) The pressure P is the same whether the observer is facing upstream or downstream. the symbol V is preferred for this volume or generalized displacement. t" . 2. where it is shown that it is more accurate to substitute the stagnation pressure for the static pressure. If one sees clockwise rotation. M. (2. and the symbol ~’z is preferred fbr the volume flow rate. It is representable by the generic symbol e. S~treamlines .9) The symlJol Q stands for volume flow rate. is the effort and which is the flow? To answer this question. This representation is particularly useful when the density of the fluid can be assumed to be constant. like simple force. representable by the generic symbol 0. Q has a direction. imagine the two observers facing each other at the two ends of the shaft.

2. also. ~ is a flow variable. Your right hand then also exerts a clockwise moment..Thus. . If one observer sees the motion as rightward. either clockwise or counterclockwise. by placing a clockwise momenton one end of your pencil with your left hand.(Checkthis out. whichrepresents a top view. on the other hand. As suggested in Fig. and balancing that momentwith your right hand. to their respective right sides or respective left sides. vs: (2. the poweris the product of the shear force. 35 GENERALIZEDFORCESAND VELOCITIES control observer A as seen by both observers shaft Mas seen by observer A Mas seen by observer B observer B Figure 2.6 Lateral Mechanical Motion Transverse or lateral motion of a membertransmits power through a control surface if a corresponding shear force is present.as seen from its wrist.Both will feel the same direction.16: Observersviewing rotating shaft I or I I controlsurface Figure 2.) Thus.11) Observers facing each other at the two ends of the bondsee the shear force in the samesense.17: Power propagated by lateral motion in one direction and observing it from both ends. 2. Fs.let the two observers sense the torque or moment.2.2.2. Now. momentMis an effort or generalized force.17. and the lateral velocity. as seen by your left wrist. and vice-versa.) Thus. the other observer sees it as leftward. the force is the effort and the velocity is the flow.

15). 2. dA. and the velocity as the flow. you are permitted to choose whichis the effort and which is the flow. where Q = f v dA. the powerof a microbondis given directly by equation (2. model the situation. For the incompressible fluid flow above (Fig. the roles of the two variables becomereversed: the velocity is seen as the s~ne by both observers. . 2. This equation can be integrated to form the macrobond only if either P or v is uniformover the channel. To the extent that neither P nor v is uniform. while the shear force has opposite sense. But do you want the distinction betweeneffort and flow to hinge on whetherthe motionis lined up with the axes of the observers? In exceptional cases like this. (Most~nodelerschoosethe force as the effort. If P is uniformor is assumedto be uniform (the more commonassumption). observer A control SOURCE-LOAD SIWTHESIS vector examples: symmetric:(a) and (b) asymmetric:(c) and (d) axis of bond observer B Figure 2. Thus. however.18. the macrobonddoes not properly. the equation integrates to give the product PQ.36 CHAPTER2.its existence implies one or more simplifying assumptions. on the other hand: the arrow points in the direction that the powerflows whenthe product of the effort and flow variables happensto be positive.) The convention of the powerhalf-arrow is inviolate.2. however. in contrast to microbonds which describe the power propagated through a microport of infinitesimal cross:sectional area. whenneither variable is a true scalar.18: Transversevectors as seen by two observers at opposites sides of a bond Should the shear motion be vertical rather than horizontal. Howthe observers see transverse vectors is pictured in Fig. as noted above.7 Microbonds The bonds addressed so far can be called macrobonds. A bond or macrobondis thus the sum or integral of an infinite numberof microbonds. P is the effort and v dA is the flow.8). 2.

or 7) = 7vs[27rrdr]. the macrobond cannot model the situation precisely. Sometimest-he error is not small. If 2 is not uniform.19 part (a). 37 f flywheel -~flywheel ¯ I control surface control surface Figure 2. vs. If ~ is uniform throughout the cross section. To the extent that ~ is not uniform. 2. Mdoesn’t exist at a point. the effort on the microbond is the compressive force -a dA.and the shear velocity vs. the rotation is a lateral motion. GENERALIZED FORCES AND VELOCITIES ~ .13) shows that ) =M~ is correct onl y if you assume that ¢ is uniform.. as shown in Fig.2. The center and the periphery of this flywheel are muchmore likely to experience different . Only in extremely rare cases would the assumption of uniform (~ not be entirely appropriate.12) where r is the radius variable and a is the radius of the shaft. of inexactness. Therefore. over the area. The point is.19: The effect of power factoring control surface or port location on accuracy. Neither of these quantities is commonto all area elements dA. 3I = vr[27rr dr].2. that whenever you employ a macroscopic variable (such as Mor ~) you are inevitably introducing some measure of approximation. The boundary between the system and the environment is assumed to lie right in the middle of the flywheel. As an example.12) and (2.ations (2. vs = r~. At the microport level. namely the force. (2.13) A comparison of equ. due to the choice of Microbonds for longitudinal motion share the flow variable a? with the microbond. the very concept of the macroport is in error. (2. with the power density being the product of the shear stress ~. The moment M is r times the moment arm integrated over the area. the effort of the macrobondbecomes the integral of the stress. The rotating shaft also can be considered to be the sum or integration of an infinite set of microports. The power propagated is the integral of the shear stress times the shear velocity. and thus can be taken forward of the integral. where -or is the normal compressive stress. consider a "shaft" that is a thin but large-diameter flywheel. This happens when the cross-section of the shaft rotates as a rigid uuit. however.

the periphery might not follow in lock-step. you would be well advised to place the system boundary to the left or to the right of the entire flywheel. e or flow. for any subject that can be reduced to rules can becomevery dull. Table 2.) Cutting a motor through its middle. Should this phenomenon possibly exist. even though it would not be incorrect. Fluid. in fact. as you identify commontypes of elements that represent various classes of interactions between the efforts and flows or their time integrals or time derivatives. There are no inviolate rules for doing this. consequently. due to their commondefinitions with respect to power and directivity. incompressible: approximate t micro-bond Mechanical. if the core is oscillated rotationally at a high frequency. The various effort variables are said to be analogous to one another. and its practitioner is subject to being replaced by software. then.1. longitudinal: tapproxi~nate micro-bond Mechanical. modeling ultimately becomes an art. 2. likely would cause you grief.19 part (b). For most mechanicaland incompressible fluid cases they are small. a "natural frequency" might well exist at which resonance occurs. as shownin Fig. 2. . The implications will gradually become clearer.38 CHAPTER 2. It is a good idea. for example.8 Analogies The effort and flow symbols for the various media that have been examined are summarized in Table 2. to partition a system from its environment at a location where the interaction can be described simply (in the present case by the variables q~ and M). Or. for example. (You should not be discouraged by this fact. the core might continue to rotate through an additional small angle. If the periphery were suddenly braked. as are the various flow variables.2. transverse: rotation translation (shear) micro-bond Electric conductor: P Fc M Fs ~-dA e v dA v = vs vs i t The powersassociated with the transport of kinetic and potential energies are added in the Chapter 12. angles of rotation than would the center and the periphery of a small shaft.1 Effort SOURCE-LOAD SYNTHESIS and Flow Analogies generalized force generalized velocity or effort.

each representing an infinitesimal cross-sectional area. All use of (macro) bonds. GENERALIZED FORCES AND VELOCITIES 39 The direct analogy between mechanical and electric circuit variables associates mechanical force and torque with electric voltage. Represent the interaction with a bond. which should be defined. The practical question is not the presence of error. e0. The direction of positive power flow is indicated on a bond by a half-arrow. and indeed the very use of the word "modeling. 2. thereby producing some energetic effect." For basic physical reasonsincludingits extensibility to non-circuit situations. and linear or rotary mechanical velocity with electric current. however. but rather its extent and significance. The example of compressible fluid flow is treated in Chapter 12.the author strongly prefers the direct analogy. cable or shaft.2.2. This summingor integration inevitably introduces some degree of approximation. This simple factoring applies to longitudinal.5 A cable attached to the top of the machine shown on the left at the top of the next page is pulled upward. and the "flow" or "generalized velocity" 0 normally is anti-symmetrlc with respect to two observers facing each other from opposite ends of the bond. to electric conductors and to the flow of an incompressible fluid. A few physical interactions are not best ~epresented by simple bonds or product conjugate variables. 3Electrical engineers sometimesdrawthe "Firestone analogy"betweenforce and current on one hand and velocity and voltage on the other. The "effort" or "generalized force" e normally is symmetric.2.9 Summary The power flowing through a simple port or simple bond customarily is factored into a product of two variables." implies approximation. annotated with a power half-arrow and effort and flow variables. Only pure mathematics is exact." Similar conceptual and computational advantages in the transference of knowledge accrue from the other analogies. lateral and rotational motions of a rod. establishing important analogies l?etween variables used in these different domains. . This gives a greater apparent similarity betweencircuit diagramsand somecorresponding mechanical"circuits. 3 This analogy can help the student well grounded in mechanical systems transfer his knowledge to electric circuits. PROBLEMS 2. A standard port or macro-port and its bond can be considered to be the sum or integration of a bundle of micro-ports and their micro-bonds. Similarly. it can help the student well grounded in electric circuits to extend his understanding into mechanical "circuits.

where Q is a quantity of thermal energy called heat. ~.2 such static relationships were generalized. Relate this displacement variable to a commonvariable used in thermodynamics. Determine the flow variable.3 Generalized Sources.4O CHAPTER 2. ." Sources normally emanate power. Sinks and Resistances The case study of Section 2. (This question presupposes that you have been introduced to thermodynamics. M(or Mmor Mp). The source and resistance elements in each case might be characterized in one of several special ways. 0. q. Represent the interaction by an energy bond.7 Heat conduction is a special kind of power that can be treated as the product of an effort and a flow. The effort e and flow ~ in this graph could represent any of the energy types discussed in the last section. The graph could represent an electrical system. and express its integral as the displacement variable. or a mechanical system with either longitudinal.6 A lever attached to the machine shown above right is rotated upward. SOURCE-LOAD SYNTHESIS [cable ~ machine I !machine[ 2. 2.) 2. In Section 2. annotated by symbols for effort and flow which you should define and by a power convention half-arrow. like the other power types above. Further consideration of heat transfer and more general thermodynamics is postponed until Chapters 9 and 12. lateral or rotational motion. a system with incompressible fluid flow. which will now be discussed. The effort variable is absolute temperature. so that the form of this system becomes in which S stands for "source" and R stands for "resistance. in terms of these variables. The power itself is the rate of heat transfer. and the shaft angular velocity.1 was represented INDUCTION MOTOR M ~9 by the word bond graph LOAD SYSTEM Both the induction motor and the load system were modeled by particular static relations between the commonshaft moment. whereas resistances normally absorb or dissipate power. which you can label as Q. which you can label as T.

As this reversal or roles suggests.1 Independent-Effort Sources and Sinks and Independent-Flow An independent-effort source. are defined to haveefforts whichare independentof their flows. 2. as discussed below. The elements provide whatever effort that the systemrequires in response to the independent . a voltage source as from an ideal battery.21. The independent-flow source and the independent-flow sink.3. Physical examples approximated by the Se element include a force source as from a weight in a gravity field. as long as the motion never reverses direction.3. The elements provide or absorb any flow that the attached system mayrequire in response to the imposed effort. it is just nice to have different wordsto express the different functions normallyserved by the element. This meanseither that e is a constant. that is e0 > 0.20 by a horizontal line in the e-0 plane. These examplesare suggested in Fig..20: Characteristics of effort and flow sources and sinks 2..2. the flow maybe either constant or an independent function of time. more simply called the flow source and the flow sink.-~e 41 S s° --g2or . a torque source. are similar to the effort source and the effort sink except that the flow is independentof the effort rather than the effort being independentof the flow. and the sink with an inward-directed powerarrow: effort source: effort sink : Se ~ ~ Se They are really the same element. Coulombor "dry" friction also can be modeledas a force source. usually called simply an effort sink.~. Both elements are designated by the symbol Se. This behavior is represented in the plot of Fig. SINKS AND RESISTANCES Se--. and a independent-effort sink. and a pressure source or sink from a large body of water at some elevation.-~- or eo . or in general a function of time. qo Figure 2. 2. e = e(t). The only difference betweenthem is that the source is designated with an outwarddirected powerarrow. GENERALIZED SOURCES. since there is no absolute requirement that the powerflows in the direction of the powerarrow. usually called simply an effort source.

21: Exampleeffort sources and sinks . SOURCE-LOADS-YNTHESIS mechanicaltranslational: mechanicalrotational: fluid (gravity): electrical (ideal battery): Figure 2.42 CHAPTER2.

Notice again that a current sink of 5 amps is the same as a current source of -5 amps.20 as a vertical line.3. You will see in Chapter 4 that any general source can be represented well by an effort source or a flow source in combination with a simple resistance. 2. as discussed below. Thus." which in this book is given the same symbol. Thus. Any other constant velocity source. if at all.1 can be represented as a general source. It is helpful to distinguish different classes of these elements. In the most general case. Ultimately. is represented by a straight line drawn through the origin. as noted above. R. but does not recognize the torque limits.3 Linear Resistances The generalized resistance. Thus. 2. or vice-versa. constant current source or constant fluid flow source also can be represented by the element. 2. the relationship itself could be an explicit function of time. only as a convenient shorthand. as long as the load torque does not exceed some limit.2. determined by the fl’equency of the driving ACvoltage (usually 50 or 60 Hz). the induction motor of Section 2. Flow sources and sinks are designated by the symbol S~: flow source: S. in favor of the equivalent term resistance and its symbol R. in a plot of effort vs. you are permitted to avoid effort and flow sinks in favor of effort and flow sources.t flow sink: ~ Sf The generalized characteristic is plotted in Fig. The algebraic . a symbol Ri in a bond graph indicates that a resistance relation applies between the associated effort ei and flow ~i. although most modelers prefer to use both types.22. GENERALIZED SOURCES. A synchronous motor turns with a constant speed. The general sink element (~S) is not employed per se in this book. A source SI element models this behavior. The simplest class is the linear resistance which. as a pressure source of 5 psi is the same as a pressure sink of -5 psi.3. SINKS AND RESISTANCES 43 flow. Other examples include a model of a battery for which the voltage depends on the current. The slope of this characteristic is defined as the modulus of the resistance.3.2 General Sources and Sinks A general source (S ~) can represent any prescribed static (or algebraic) relationship between its effort and its flow. a model of a velocity-dependent force applied to a mechanical system and a model of a pressure source that depends on the fluid flow. flow as illustrated in part (a) of Fig. the general source is employed. indicated as e can have any functional dependencybetween its effort and its flow. or for short its "resistance. 2.

44 CHAPTER 2. ei (a) linear SOURCE-LOAD SYNTHESIS // -R (b) nonlinear qi (c) power as area (d) linear biased (e) multiple valued Figure 2. 2. This or any other algebraic expression is written separately from the bond graph. They are called dashpots. More generally. where R~ is a constant. the resistance can be approximated as linear. In your fluid mechanics course you will derive the relation that the pressure drop P across the tube is related to the volume flow rate Q through it by P = RQ. algebra is never included within a bond graph. flow through a porous plug and "creeping" flow through an orifice. The two meanings of the resistance are distinct. Examples include flow through a thin slit such as a leakage path in a machine. and e = Ri. but are very co~nmonly assumed in analysis. The electrical resistor may be the most familiar approximation to a linear resistance. This and other linear resistances are pictured in Fig. L is its length.8p_~_L (2. 2. A commonlinear fluid resistance results from the assumption of viscous laminar fully-developed steady flow through a circular tube. that is when the Reynold’s number is low. details this relationship. R =. Mechanical devices that can be approximated by a linear resistance are hard to make.14) where a is the inner radius of the tube. The effort is the voltage drop e across the resistor. The ratio .22: Types of resistance characteristics expression ei = Ri~)i. and # is the absolute viscosity of the assumed incompressible fluid. however. and are given the special symbol shown in part (c) of Fig.23.23. the flow is the current i which passes through it. and its modulus as a constant. whenever the viscous forces overwhelm the inertial forces in an incompressible flow through a passage.

SINKS AND RESISTANCES 45 (a) electrical resistance ---~i R te---’~R e~--~~ (b) fluid resistance P "~ e =Ri -~~~ Q ~’R L P=RQ (c) translational dashpot linear fluid resistance x F=tUc.23: Devices approximated by linear resistances . R=b X (d) rotational dashpot classical symbol thin film of fluid physical example bond graph relationship Figure 2.3.2. R=b M=R~. GENERALIZED SOURCES.

25.14). (2. 2. for example. The shear flow of a viscous liquid in a very thin annulus between a rotating cylinder and a cylindrical shell. As shown in Fig. or worse. 2. The general resistance is the same as the general sink. as shown in the example. which can be done by employing a piston and cylinder. the small amount of fluid trapped in the annulus is apt to overheat. however. The shock absorbers in your car are essentially of this construction. and complete dissipation of the kinetic energy is assumed to occur downstream.15) Here. If Bernoulli’s equation is used to model the flow between an upstream state with zero velocity and the point in the throat where it achieves maximumvelocity. the classical model for such a resistance comprises a constant or coulomb .611. and likely is determined experimentally. that is a rotational linear resistance. in this case. Cd = 0. 2. A rotational dashpot is designated by a similar symbol. A mechanical brake is another kind of nonlinear resistance. but is not proportional to it. it is desirable. can produce a torque proportional to angular velocity.3. Replacing the linear fluid resistance in the translational viscous damper in Fig. The characteristic given by equation (2. p is the fluid density and c4 is a flow coefficient which represents the fact that the flow separates from the walls of the orifice to form a narrower vena contracta or effective orifice area CdAo. for small bumps you want the shock absorber to act gently. in most other cases it is larger but less than 1. Then. As a practical matter.23 with an orifice flow produces a nonlinear mechanical damper. SOURCE-LOAD SYNTHESIS of the force to the velocity traditionally is designated by the symbol b.24.4 Nonlinear Resistances Any component with a single effort or generalized force that is a function of an associated flow or generalized velocity. and is used in its stead. One way to construct a linear translational dashpot is to convert the relative velocity of the two translating membersto the flow of ~: liquid.0. Cd is adjusted to accomodate the error. (For a sharp-edged orifice. 2.) The equation often is used even when the assumptions above are recogniZed as inaccurate. causing the fluid to oxidize and turn black. this liquid is forced through a passage that behaves as a linear fluid resistance according to equation (2. as shown in part (d) of the figure. causing its viscosity to drop. Translational and rotational dashpots often are described as "viscous dampers" because it is hard to approximate linearity mechanically without using a viscous fluid.46 CHAPTER 2. Flow through an orifice at anything higher than creeping Reynold’s numbers is an example. so the modulus of the bond graph element is R = b.24. like the motor oil in your car. as suggested in Fig.15) is plotted in Fig. can be described as a nonlinear resistance. 2. whereas for large bumps you prefer that it act more vigorously so as not to bottom out (reach its mechanical limit). Not only is the nonlinear behavior easier to obtain. the result is P=-~pv ~=~p ~ = ~ .

GENERALIZED SOURCES. the resistance torque balances whatever torque is applied. and it does not pass through the origin. This assumes application of a fixed force to the brake calipers. 2. P = eO= R(~)O". A linear biased characteristic (2.22 is nonlinear two ways: it is curved.3. R = R(c)) or R = R(e). such as the one that makes a violin sing or a dry door hinge squeel. for either the linear or the nonlinear case. 2.25: A mechanical brake and its ideal resistance characteristic friction torque for forward motion.17) is shown in part (d) of Fig. this model gives nothing but a constant torque. Whenthere is no motion. as shown. Algebraically.16) These equations also apply to the linear case. or is biased. and an equal but opposite torque for reverse motion.~ ~ plug: "--~~ valve or porous AND RESISTANCES P or F large hole with small orifice shock absorber: F 0o Figure 2. and can be modeled better as an effort sink. as indicated in part (c) of the figure. In Chapter . This chordal resistance therefore is not constant. associated with the kinetic coefficent of friction.22. The power dissipated in the resistance.2. of the rectangle defined by the origin and the operating point. is defined as the slope of the chord from the origin to the state point of interest. R. The modulus of the resistance. If the disk never rotates backwards. The resistance characteristic shown in part (b) of Fig. as long as it does not exceed the "breaknway"torque which is associated with the static coefficient of friction. however. but rather is a function of t) or e: e = R~). (2.24: Nonlinear fluid resistance and application isk ~o in a mechanical damper MI e Figure 2. equals the area. SINKS nozzle: --~--- 47 ----- p . You also will see more exotic friction models in Chapter 4. except that R becomes constant.

3.Anexampleof a multiple-valued characteristic is given in GuidedProblem2. biased characteristics are represented as combinationsof an effort or flow source and an unbiased linear or nonlinear (curved) characteristic. in3/s400 I . psi 60 ~ ~ 40 ~ ~750 rpm IMPELLERPUMP~ LOAD 20 0 load I 0 I 100 I I 200 I ~ I I I 300 Q. andits stability. by a shaft. but a lack of bias does not assure passivity. multiplezvaluedcharacteristics are shownin part (e). are globally passive. 2. typical of the flows through sharp restrictions as discussed above and as employedin the water-sprinkler nozzles of Section 2.1 The output pressure of an impeller pumpreduces the net flow below that of the geometrically slip-free flow. Finally. Determinethe equilibrium state for the combinationof these elements.48 CHAPTER2.1. The particular load characteristic plotted in the figure also is nonlinear.1. SOURCE-LOADSYNTHESIS 4. since they reside only in the first and third quadrants where the powerproduct e4 is positive everywhere. as plotted below. Note that any bias destroys passivity.5 Source-Load Synthesis The interconnectionof a rotary source to a rotary load. 8O P. The result is a nonlinear source characteristic. unlike the biased curves (b) and (d). EXAMPLE 2. The same type of synthesis applies to any simple interconnection in which the poweris represented by the product of a pair of conjugate generalized forces and velocities.3. nevertheless. The characteristics illustrated. was treated in Section 2.3 by finding the intersection of the two torque-speedcharacteristics.

3. The slope of a chord drawn from the origin to any point on a locus of constant power equals the magnitude of the slope of the tangent to the locus drawn at the same point. 2. GENERALIZED SOURCES. All points on the locus must have the same area. This fact.2.6 Power Considerations The power th’at is transmitted from a source to a load often is of interest. Note also that the chord and the tangent form isosceles triangles with the ~ axis. ibrcing the flow back toward equilibrium. the pressure of the pumpwould exceed that of the load. since {f operation is attempted at a larger flow. as suggested by the two other rectangles shown. illustrated in part (b) of the figure. 2.~ plot. If operation is attemped at a smaller flow. Loci of constant power can be constructed on an e . P=e~=4 3 e 3 e 2 0 0 1 2 0 3 (a) rectangles 4 2 0 0 i 2 0 3 (b) chords and tangents Figure 2. 4 . One such locus is shown in part (a) of Fig.26. The curve is a rectilinear hyperbola. SINKS 49 AND RESISTANCES Solution: The equilibrium state is given by the intersection of the characterisitics. the pressure of the load would exceed the pressure of the pump.3. again forcing the flow toward equilibrium.26: Loci of constant power as rectilinear hyperbolas. The power equals the cross-hatched area of the rectangle of height e and width ~. This equilibrium is stable. aids the sketching of constant power loci.

The slope of the load characteristic. of 2.5 0 0 50 100 150 ~.0 0. SOURCE-LOAD SYNTHESIS EXAMPLE 2.5 1. All other points on the characteristic have less power. which is also the resistance of the load. Thesestraight lines are drawndashed. but the chord from the origin to this point has the same magnitudeof slope as the tangent to the load characteristic. which is . 2. Not only does this point maximizethe area of the rectangle betweenit and the origin. shownby a dashedrectilinear hyperbola.2 Locate the point of maximum power on the torque-speed characteristic the induction motoras plotted below.50 CHAPTER 2. rad/s 200 The source-load synthesis of a linear electrical source and linear electrical load is given in Fig. is tangent to the cha[acteristic at the operating point for that power. and the slope of the tangent to the characteristic at that point. The locus representing the maximum power. including the point of maximulnmoment. is similarly knownas the load impedance.W 50 100 200 300 400 1.A simpler way to estimate the location of the maximum power equates the magnitUdeof the slope of the chord from the origin to the point.0 0. 2.27.5 0 0 50 100 150 ~. The magnitudeof the slope of the source characteristic is knownas the source impedance. rad/s 200 Solution: A family of loci for various constant powers is superimposed on the plot below.5 loci for constant powers. Maximum poweris transferred from source to load at the midpoint of the source characteristic.

Linear fluid resistances depend of the viscosity of the fluid. and are designated by the bond-graph symbol S.3. Resistances occur whenever electrical. but possibly dependent explicitly on time. its value of R is constant. The slope of a chord drawn from the origin to a point on the characteristic curve of a resistance is called the chordal resistance or just the resistance.2. An unbiased resistance is called passive if its characteristic remains in the first and third quadrants of the coordinates. Resistance characteristics that pass through the origin of the effort-flow coordinates are called unbiased. for example. 51 GENERALIZED SOURCES. and ¯ constant-effort sinks if it is drawn inward. General sinks normally are called resistances.3. General sources have other relationships between the effort and flow." and often depend on the viscosity of a fluid. It will be shown in Chapter 4 that both general sources and resistances can be represented by a combination of an effort or flow source or sink and an unbiased resistance. They are called constant-effort sources if their power arrow is drawn outward. mechanical or fluid energy is dissipated into heat. Nonlinear fluid resistances depend rather on the density of the fluid. They are called constant-flow sources if their power arrow is drawn outward.27: Synthesis of linear source and load to maximize power transfer the load impedance itself. but possibly dependent explicitly on time. Linear mechanical resistances or dashpots often are called "viscous dampers. regardless of the operating point. and is designated by the same symbol. Mechanical friction. 2. Thus. and are designated by the bond graph symbol R. but maybe approximated as~ an effort source or sink if its motion is always in one direction. Such "impedance matching" is standard between audio amplifiers and audio speakers. is an unbiased nonlinear resistance. as in a brake.7 Summary One-port elements with constant effort independent of flow. One-port elements with constant flow independent of effort. R. are designated by the bond-graph symbol Sy.~rce impedance i Figure 2. and constant-flow sinks if it is drawn inward. SINKS AND RESISTANCES ~ ~ d impedance e ~. The simplest resistance is linear and unbiased. are designated by the bond-graph symbol Se. maximumpower is transmitted when the load impedance equals the source impedance. since then the power 7) =e~ is always positive into the element. . others are called biased.

.. This peak is due the wavegenerated by the motion. f 6o 800 force. Unlike the displacement-type hull considered before.28. with the generated wavecrest being amidships. Find and plot the steady-state relationship betweenthe throttle position and the speed of the boat.~hrustfor indicated percentthrottle" 80 ~ 50 ~/dral ~ 2OO ~ 0 I I 0 10 20 30 40 speed.the bow is considerablyelevated over the stern. SOURCE-LOAD SYNTHESIS . Clearly indicate any hysteretic behavior. F lbf 400 ~ . plotted in Fig. Draw a word bond graph for the system.. . The IC engine and propeller drive from Guided Problem 2..1 and again in Example 2. ft/s Figure 2.3 50 I 60 for the boat in Guided Problem The source-load synthesis illustrated in Section 2." riding high and nearly level..~O. At higher speeds the boat "planes.1 can be generalized for any of the powertypes considered..1 is placed in a small boat with a planing-type hull.. Suggested Steps: 1. a ggneralized velocity and a generalized force. 2.. a load. so.. in effect the boat is continually trying to climb a hill of water..k. and reveals an important phenomenoncalled hysteresis. the drag characteristic of this hull.3 This problemoffers you needed experience in source-load synthesis. or history-dependent behavior including sudden jumpsin state. has a pronouncedlocal maximum (at about 24 feet per second). naming a source.52 CHAPTER 2. 90.28: Thrust and drag characteristics 2. The synthesis maximizesthe powertransmitted if the chordal resistance of the load equals the slope of the source characteristic at the equilibrium point: Guided Problem 2. At the speed of its maximum.

An n-channel enhancement MOSFETis driven by a 12 v battery and has a load resistance of 1.3.------~ ~/ 53 e~=5. It requires determination of a source characteristic. GENERALIZED SOURCES.5 kfL Given the characteristics plotted in Fig..5V io.4 Were it not already done._. 2.ov ¯ "~’~IDo 0. units and scales are compatible. Your instructor may advise you on its relevance to your situation. In this range the actual speed depends on the history of the throttle position as well as its current state.. Identify any jumps in behavior with arrows on the cross-plot._.2. generate more points by interpolating more source characteristics.. Note explicitly the range of throttle positions for which there are multiple speeds. you would superimpose plots of the source characteristics and the load characteristic.0 V 3. Guided Problem 2. making sure the dimensions. If and where it is not clear how to connect the points on your crossplot with a continuous curve.5V 3.29: Characteristics of a MOSFET for Guided Problem 2. plot the drain current iD and voltage eD as a function of the gate voltage . SINKS 10 ~5._. ~f~ ~-~ ~ 14 16 V @)e~ Figure 2. mA 6 4 AND RESISTANCES / [ ] ~. 1. Cross-plot all intersections of the source and load characteristics onto coordinates with speed on the ordinate and throttle position on the abcissas.4 This type of problem is familiar to electrical engineers.29.0 V 4.5 V / 4.5k 12 2 VT 4 6 8 10 12 eD.

shown below. Consider the battery and resistor as the power "source. eG. (These seemingly contradictory assumptions could be reasonable if low-friction polymer seals are used between the piston and the cylinder. Suggested SOURCE-LOAD SYNTHESIS Steps: 1. to create a standard type of shock absorber.. (b) Sketch-plot the force vs.23. Cross-plot the intersections of the source line and the given load characteristics onto plots of iD vs.) You will need to define relevant physical parameters (constants) in order to carry out the following: (a) Find the relationships between the force and the pressure difference across the piston. and show the power-convention half-arrow. the velocity.10 Answer the above problem when the cylindrical hole is replaced by a short orifice of area Ao. and express the resistance relative to the velocity as a function of fixed parameters to the extent possible. and between the velocity and the flow. 2. rotates concentrically journal with radial clearance e < < a and length L. Indicate what type of bondgraph element most simply represents this frictional behavior. in a fluid-filled .9 As noted in the text and pictured in Fig.11 A shaft of radius a. its intersections with the axes are readily found." and the given characteristics as the "load. Is this approximate resistance a constant? 2.54 CHAPTER 2. Label the effort and flow of its bond with symbols that indicate their meanings. Fluid leakage around the piston and friction between the piston and the cylinder walls may be neglected. 2. PROBLEMS 2. assuming that the shaft never reverses direction. Hint: It is a straight line with negative slope. 2. eG’ and eD vs. a viscous fluid damper can be made by placing a cylindrical hole through the piston of a fixed double-rodend cylinder." Superimposethe source characteristic on the given plot.8 A shaft drives a load that resists with a constant torque regardless of speed. 2.

GENERALIZED SOURCES.13 The impeller pump of Example 2.12 Write the chordal resistance of the flow through an orifice.14 Estimate the maxitnum power that the impeller pump of Example 2.3.1 (p. 2. (b) Discuss any major pract’ical deviation from the assumed behavior.2. as a function of the base current. in the form R -. 2. SINKS 55 AND RESISTANCES (a) Find the steady-state resistance relative to the angular velocity of the shaft. as modeled by 2. equation (2. ib. 48) drives a load for which P 20 + 0. transistor ec ~ load 8 R-. Give the corresponding pressure and flow rate.0. 2.10 Q psi. . 48) is driven by a hydraulic source described by P = 80 . 2.. but you should disregard everything in the figure except for the engine characteristic. where Q has units of in3/s. Estimate the equilibrium pressure and flow.1 (p. where Q has units of in 3/s. V for the load which comprises (b) Plot the load (emitter) current.) 2.17 An npn transistor in the commonemitter configuration has the characteristics plotted below for various values of the base current. 75) can produce.39 (p.15). ic.1500 4 50’ ~[ -- ~" ~ ’ ~" ic’mA " --e°=12V- ~ ~ 10 ~’~~ 00 (a) Superimposeon the plot the characteristic the voltage source eo and the resistor R. and approximate this relation algebraically. in place of the impeller pump. also. Estimate the equilibriura pressure and flow.5. in place of the plotted load.1 (p.15 The hydraulic load plotted in Example2. (This figure appears in Section 2.R(Q). 4 8 12 e.16 Estimate the maximumpower that the internal combustion engine characterized in Fig.. 48) could deliver if the load could be changed arbitrarily.2Q psi. 2.

200[- . 2.-"------~~.f 40 f~/s 3O 2O l0 ¯ jump~ ~~ I~/I PROPELLER F DRIVE A~ ~ hysteresis band 20 40 60 80 throttle. The force in the tow line for one skier is shown plotted below as a function of speed. 10 20 30 40 0 50 speed...28 pulls one or more water skiers... % 100 . SOLUTIONS TO GUIDED PROBLEMS Guided Problem 2.. -~ .56 CHAPTER 2.. (c) In a fancy maneuver. Estimate the maximumnumber of skiers that can be towed successfully after the transfer. .3 velocity. skiers are transferred to the boat from another boat. ..~ ..18 The boat of Guided Problem 2. note that a planing phenomenonexists which is similar to that of the boat hull. ~ 0 ~//. and estimate their maximumspeed.3 and Fig. but more pronounced. ft/s (a) Carefully estimate the maximum steady speed for towing a single skier.. . SOURCE-LOAD SYNTHESIS 2.. at full speed. (b) Estimate the maximumnumber of identical skiers that can be accelerated from rest simultaneously. and estimate their maximumspeed...

e: mA 8 ¢o 0 2. V Ideal Machines: Transformers Gyrators I 6 and An ideal machine transmits work at one of its two "ports" to work at the other "port. This simplification is done despite the fact that leakage or slippage in the presence of friction or analogous generalized forces inevitably has at least a small effect. Entropy also is not generated. V 16 12 I 2. 2 4 6 10 12 14 eo. or more properly the conversion of energy to heat. electric motors and piston pumps are examples of engineering components that may be approximated as ideal machines. In modeling physical or engineering systems one often neglects the dissipation of energy. an electric transformer might be assumed to have perfect coupling and zero resistance. Levers.2. an electric motor might be assumed to be 100%efficient in converting electrical energy to . a gear train might be assumed to be frictionless. gears.4 l0 i~. energy is not stored. generated or dissipated.4 IDEAL MACHINES: TRANSFORMERS AND GYRATORS 57 Guided Problem 2. For example." Although the effort and the flow usually are changed.4 8 ~ "1 I 4 I e~.

58 CHAPTER 2. Constant modulus or not. or. . As a result. 2. Thus the input power equals the output power. As well as being able to run forwards or backwards. including DC.in general may be a function of the displacement q~ or some other displacement. neither power can represent the flow of heat. the power could be negative. often it is represented separately. (The fatniliar electrical transformer is an electrical approximation of the transformer. that is to consume or produce zero net power at all times. SOURCE-LOAD SYNTHESIS mechanical form. One is called a transformer. Other authorg usually . a real electric motor can be represented as a frictionless or ideal motor plus an external resistance to represent electrical losses and another external resistance to represent mechanical friction. called the transformer. the ideal machine is thermodynamically reversible. called the modulus of the transformer.. may be constant. With the power convention of this graph. Until Chapter 9. 2.] (2.19) allows for many different types of ideal machines.) An ideal machine can be represented by the word bond graph IDEAL MA CHINE It is defined to conserve energy. but only if it is idealized so as to work perfectly at all frequencies. e101 = e_~02. and a fluid pump might similarly be approximated as converting mechanical energy to fluid energy completely.19) The ideal machine can be run in either direction.20) Thus T is defined as the ratio o/the generalized velocity or flow on the bond with the outward power arrow to the generalized velocity or flow on the bond with the inward power convention arrow. For example. attention is restricted to tranformers with constant moduli.2 Transformers Equation (2. and therefore flow from right to left. There are two particular generic (or generalized) types of ideal machine that command special attention. Even if energy dissipation is not neglected. which means that it generates no entropy.1 Ideal Machines All of these devices can be described as "machines. The most commontype." in a generalized sense. unlike a real electrical transformer. and models of them that neglect energy loss are called ideal machines.the symbol T within a bond graph designates this type of relationship.4. and the other a gyrator. Although a positive power flows from left to right according to the power convention arrows chosen arbitrarily above. This ratio. satisfies the additional relation =r0.4. (2.

other authors more commonly employ GYor. Such drives can be based either on non-sliding friction (shear forces) or positive action (normal forces). with MTF(for modulated transformer): ~.4 Mechanical Devices e_._~_T ¯ ql T ~. and can damp vibrations.4. Substitution of equation (2.4 IDEAL MACHINES: TRANSFORMERS 59 AND GYRATORS elaborate with the designation TF. ] (2. which by Thus. is defined as the ratio of the effort on one of the bonds -.2 Modeled although consideration as of non- Transformers Most mechanical drives can be represented as transformers with constant moduli. 2.4.) ~bu must not confuse the bond-graph symbol T with the modulus T.either one -. 2. or.1 ql el.19) gives a simple but profound result: [G1 = T~2 .23) el = G02.to the flow on the other bond.19) gives (2. only the modulus can participate in an equation.2. the modulus of the gyrator. constant moduli is deferred until Chapter 9. the modulus G need not be a constant. MGY: el G e2 e. if the modulus is not a constant. . as suggested in Fig.21) In words. The gyrator in a bond graph is designated by the symbol G.31. 2. as long as frictional losses are neglected.1 Like T. if the modulus is not a constant. the transmitted forces are limited by the frictional properties of the materials. such as T = 5._L__~ MG Y GY e. (Most other authors do not-designate ¯ commonsymbol for the modulus of transformers.2 q’2 el F~ ~o q~ MTF ~ ~ e2 O~ Subscripts can be added to the symbol T to distinguish different transformers that might appear in the same model. Friction drives are inexpensive. Friction drives include belt drives.3 Gyrators The second major possibility definition satisfies for the ideal machine is the gyrator. G. 2. as suggested in Fig. Substitution of equation (2. and rolling drives.30.20) into the conservation-of-energy equation (2. the ratio of the 9e~evalized forces of an ideal transformer equals the inverse of the ratio of the respective ~eneralized velocities.22) into (2. On the other hand.

SOURCE-LOAD SYNTHESIS Figure 2.30: Examples of pulley drives cylindrical conical Figure 2.60 CHAPTER 2.31: Examples of roiling contact drives .

Nevertheless. The nominal kinematics of most toothed drives can be analyzed as though the teeth or sprockets weren’t there.4 IDEAL MACHINES: TRANSFORMERS AND GYRATORS 61 (CourtesySeitz Corporation) (a) timing belt (b) chain and sprocket . and a pitch diameter. A cylindrical pitch surface possesses a pitch radius. The . for example by using gear teeth with involute shape. are the respective angular velocities.d me~nbers 1 and 2 at the point of contact is v = r~l = ro. The virtual surface of contact for a rotating member is known as the pitch surface. Examples are shown in Fig. 2..32: Examples of toothed drives Teeth added to a belt gives a timing-belt drive. Bostb-nGearWorks) (C) spur gear (Courte~. Similarly..GleasonWorks) (d) bevel gear Figure 2. where ~ and ¢.2. wheel or gear. slight manufacturing errors or deflections due to heavy loads occasionally produce significant vibrations not included in a constant-modulus transformer model. All of them can be designed to operate very s~noothly.32. the commonvelocity of the two couple. These positive-action drives can carry heavy loads and high power efficiently. For a pulley. and a chain of links and pins substituted for the belt gives a chain-and-sprocket drive. teeth added to the wheels of a rolling drive produce a gear drive. r.q~_9.

such as microphones. sometimes refer to idealized behavior that is equivalent to the (ideal) transformer. These convert relatively little power into electrical form.4. after the idea of transducing transformers and gyrators is introduced. therefore is the inverse of the ratio of their pitch radii or diameters. Whenunqualified. and only if resistance losses are neglected. Electric circuit symbols for both the real electric transformer and the (ideal) transformer are given in Fig. The non-ideal behavior of an electrical transformer is modeled in Chapter 10 by combining the (ideal) transformer with other elements.5 Electrical Transformers An electrical transformer acts like an ideal transformer only for frequencies of ex~citation that are neither too small nor too large. ..62 CHAPTER 2. pumps..4. Circuit diagrams. 2. strain gages and instruments that measure temperature. devices with internal moving parts that behave like fluid transformers can be synthesized from two component devices that act like transducing transfomers or gyrators. Sometransducers are transformational.. T = ~/~ = rl/r.. 2. however. compressors and turbines are power transducers that convert substantial power from one domain to another.6 Transducers Modeled as Transformers A two-port device that converts (or transduces) energy directly from one energy domain to another. without employing a thermal engine. the term transformer refers to the ideal element asdiscussed above.33: Standard electric circuit symbols for transformers and gyrators ratio of the angular velocities of the two pulleys or two gears etc. Some people reserve the word "transducer" for instrument transducers. An example is given in Section 2. actuators. since their function is to provide information. generators.5. Nevertheless. acceleration. Electrical and fluid motors. and therfore does not include the real electrical transformer. etc. 2. pressure. tachometers. (ideal) transformer SOURCE-LOAD SYNTHESIS electrical transformer (ideal) gyrator Figure 2. is called a transducer. Purely fluid devices (fluid-in and fluid-out) that act like transformers without any moving mechanical parts are almost nonexistent.e. i.33. while others are gyrational.

for in more difficult situations. In the example above. You can choose. however. The term pump usually refers to machines that permit continuous rotation. regardless of the orientation of the power convention half-arrow. ~quals ~h~ velocity shaft.2. or do both as a check. the input power on one side equals the output power on the other F~ = PQ. In any case. Students who have t~ken a course in statics often tend to recognize this result more readily. the transformer model is proper. which has been verified. the relation Fc = AP is derived from a velocity constraint combined with the conservation of energy.4 IDEAL MACHINES: TRANSFORMERS AND GYRATORS 63 EXAMPLE 2.3 Show that the positive-displacement inechanical-to-fluid power transducer in the ibrm of a piston-and-cylinder or rain.They. The power in a trausformer can flow in either direction. articulate the revolute joints of hydraulic robots. in fact. its bond graph must be Fc T~ The transformer modulus T is defined as the ratio Q]~. \\\\\\ Solution: The volume flow rate of ~h~ ~uid. It is important to get in the habit of identifying such geometric constraints directly. A simple force balance on the piston would ~lso give this result. The piston-cylinder device can be operated either as a pump or as an actuator (its most commonrole)~ Rotary limited-angle actuators. use of the principle of the conservation of energy such as in this example precludes the need to perform both a force balance and a geometric constraint analysis. Further. particularly when dynamics is involved. as shown below. Q. ~. the force balances can be relatively awkward. side. and therefore equals the area A. also are co~nmon. than they recognize the velocity constraint. can be modeled as a transformer if leakage and friction are neglected.34. times the area of the piston. such as drawn in Fig. A: of i~ Q= ~A. 2. for example. If the device is modeled as a transformer. . The transformer modulus also must equal the ratio F~/P. Find the modulus of the transformer in terms of physical constant(s). Substitution gives F~ = (Q/~)P = AP. Therefore.

(2.34: Rotary actuator such as the gear pumpof Fig. 2. and Mis the torque on the shaft.) It also ignores the unsteadiness of the flow that exists in most pumps. so powerout equals the powerin. or piston pumpof part (c). Thus.24) A pumpapproximatedas an ideal machinealso has no friction. Q = D~ (2. (The effects of modest leakage and friction on nominal positive displacement machines is considered in Chapter 4. then M~ = Q AP. Its definition assumes no leakage flow between the inlet and outlet chambers. the vane pumpof part (b).P2 ~ ~u iOnaryabutment M and vanes Figure 2. D.64 CHAPTER 2. if Apis the pressure rise from the fluid inlet to the fluid outlet. knownas the radian displacement of the machine. With these idealizations.35 part (a). PI SOURCE-LOAD SYNTHESIS AP = PI .25) . This type of machineis characterized geometrically by the volumeof fluid which passes through it per radian of shaft rotation.

..35: Common types of positive-displacement ~ Spherical washer pumpsand motors ...0u.2.0...4 IDEAL MACHINES: TRANSFORMERS 65 AND GYRATORS internal ~eal formed here Unbalancedforces causedby pressureat the outlet pump or (a) gear CourtesySauer-Danfoss Inc. motor ~ igh pressure LOwpressure inlet .... ~ internal seal formed here Rotations Shaf (b) vane pump or motor ........~tlutlet Casing.. Vanes Above armbelow:Reprintedwi¢h permissionand cou~esyof E~onCorporalion Cylinder block Spring force transmitting pin (3 spaced evenly) (c) axial pis[on spring pump or motor Shoeretainer plate (retracts pistons) P ~’::::::~::::~:~ ~ Ports are ....t ~ -- .n~_~..~ rotated 900/ flow Valve~ plate / Cylinder hlnnk ~ \ Piston \ Sho~ Figure 2..

) Positive-displacement hydraulic machines are widely used in industry and transportation because they handle vastly larger forces and powers than their . This implies that the fluid power is flowing into the machine. so that both AP and M become positive and T = lID. SOURCE-LOAD SYNTHESIS EXAMPLE 2. it is called a hydraulic motor. and find the relation between the momenton the shaft and the pressure rise. A device operating this way is not usually called a pump. Solution: Substituting equation (2. since packets of fluid with fixed volume are conveyed mechanically from one port to the other. In the vane pump they are the fluid trapped between succesive vanes. which rotate with the drive shaft like the barrel of a six-shooter. for it is neutral on the question of which way the power flows. shown cross-hatched. Pivotable piston shoes rotate with the pistons. Thus. Whenthe volume of a cylinder is expanding. 2. which equals the ratio M/AP. one might prefer to reverse the power arrows when use as a motor is intended.35. (These inlet and output ports or lines actually are centered at angles rotated 90° about the machine axis from those pictured. In the piston pump the packets are the fluid volumes that enter and subsequently are expelled from each cylinder.24) defines the transformer 5I ~ T AP The modulus of the transformer is the ratio Q/~. so that T = D. The transformer representation nicely represents this fact. it is connected through a kidney-shaped slot in a fixed val~ving plate to the inlet port. barring any practical problems such as rubber seals being over-stressed and failing. the van~s slide in and out of slots to prevent leakage. In practice.25) gives the sired relation between the momenton the shaft and the pressure rise: M = D AP. This result together with equation (2. What happens to the purap if AP is negative. and vice-versa. whereas the mechanical power is flowing out. rather.4 Show that the idealized positive-displacement pump is a transformer. it is connected through another kidneyshaped slot to the outlet port. including the machines shown in Fig.66 ~ CHAPTER 2. Whenthe volume is contracting. The pump/motors being considered are called positive displacement machines. and are also given a sinusoidal axial motion by sliding on a stationary angled cam pla~e.24) into equation (2. The axial design pictured has several cylinders bored around a common barrel. the volumetric displacement per radian. Determine the modulus of this transformer. however. corresponding to a pressure drop instead of a rise? The equations above and the bond graph still apply. so a negative value of the torque Mresults. In the gear machine the packets are the fluid trapped between adjacent teeth. a positive displacement pump run in reverse becomes a motor.

2. since their behavior is more complex.25. a gyroscope with a horizontal shaft supported only at one end will not fall downif it is allowed to precess.4 IDEAL MACHINES: TRANSFORMERS F2 2 AND GYRATORS x~ 67 x axes for xI andx.g. motors and solenoids) for a given size and weight. produces a proportional angular velocity ~. Such coupling can be enormous if the angular momentumis large. a gyroscope indeed exhibits gyrational coupling between two axes of rotation that are both normal to each other and normal to the principal axis of rotation. As Fig. are normal Figure 2. Dynamic machines typically are used in relatively low pressure applications. however. and are not represented as two-port elements until Chapter 9. it suffices to represent a dynamic machine as a one-port fluid device. presuming a fixed speed or some other fixed condition for the rotor." This association is appropriate. which uses compressed air or other gases in similar linear and rotary machines. and vice versa. most mechanical systems do not combine . As a result. In manycases. equals the angular momentumof the rotation about the principal axis. varying from tiny impeller pumps and fans to enormous hydraulic turbines. Dynamic hydraulic machines do not have fixed displacements. As an example.36 attempts to indicate. The remainder of fluid power encompases pneumatics. 2. p 71).2/L on the other axis.36: Gyroscope with idealized model electromechanical counterparts (e. that is rotate slowly about a vertical axis through the support point (see Problem 2. a torque FIL on one of these axes. an impeller pump was modeled in the last section as a one-port general source. which are beyond our present interest. where L is the length of the shaft. The strength of the gyrational coupling. any mechanical system with a significant angular ~nomentum has a potential for gyrational coupling..2.7 Mechanical Devices Modeled as Gyrators The word "gyrator" may conjure up in your mind the word "gyroscope. Their use comprises the major part of the area of engineering known as fluid power. for when stripped of its complexities. that is the magnitude of the tnodulus G. Nevertheless.4. Thus.

is a motor with translational rather than rotational motion.8 Transducers Modeled as Gyrators The force on an electrically charged particle moving in a magnetic field is proportional to the charge. e = G~. M = Gi.. (2. such as that examined in Guided Problem 2. Equation (2.37: Idealized DC motor and gyrator M model large angular momentumwith a rotation of its axis. The force on an electrical conductor therefore is proportional to the electrical current and the field.26b) where G is a constant dependent on the geometry of the motor. 2. If. On the other hand. There are many similar electromechanical devices based on magnetic fields. Thus. further. much like the hydraulic motor and the hydraulic pump. 2. In a simple DC electric motor. as in a permanent magnet motor or a motor with a field circuit driven by a fixed voltage or current. 62). although most are complicated by varying energy storage and so are addressed later (in Chapter 10). an instrument in which the angular velocity of a shaft produces a proportional voltage which can be displayed. A large tachmeter qualifies as an electric generator. . such as solenoids..It also can be modeled as a gyrator.33 (p. Thus.G Figure 2. by a voltmeter.37. 2. the magnetic field is constant..26a) (2. justifying the neglect of gyrational coupling. however. the torque is simply proportional to the armature current. the symbol is used mostly to represent idealized electric circuit analogies for mechanical systems. as suggested in Fig.6. The electric circuit symbol for a gyrator is included in Fig. with an appropriate conversion factor. No simple passive electrical element for low-frequency circuit use acts like a gyrator. A translating coil device.02b) follows from the assumed conservation of energy. the armature wi)e is wound and the commutation so arranged that the force produces a torque on a shaft.68 CHAPTER 2. the electric motor and the electric generator are the same machine simply operated backwards. This torque therefore is proportional both to the current in the armature circuit and to the magnetic field. to the field and to the particle velocity. It directly suggests the idea of the tachometer or speedometer.4. SOURCE-LOAD SYNTHESIS field circuit e.

and examine their bilateral behavior. such as the sources and loads already introduced. 2. not chosen agrees with the modulus of step . The depth of the working chamber may be taken as w.5 This problem and the next should enhance your understanding of the nature and application of transformers and gyrators by having you evaluate the moduli of particular devices modeled as ideal machines. Evaluate the transformer modulus for the rotary actuator shown in Fig. and the input flow. Guided Problem 2. Suggested Steps: 1. and other parameters may be defined as appropriate. 2. Simpler representations are given in Chapter 9. Many actual machines can be approximately represented by these forms. Examples include capacitor microphones and piezoelectric transducers.9 Summary Ideal two-port machines are by definition conservative. and evaluate T as their ratio in terms of (fixed) parameters. beginning in the following section. hydraulic rams and hydraulic pump/motors. T.34. and the depth of the vanes and chamber. The effort on each of the two ports of a gyrator equals the product of a commonmodulus G and the flow of the opposite port. 2.4 IDEAL MACHINES: TRANSFORMERS AND GYRATORS 69 the class of electromechanical devices based on electric fields rather than magnetic fields can be shown to give transformational coupling. and rnore complicated models of these and other devices often have transformers or gyrators imbedded within them.4. The two simplest forms are the transformer and the gyrator. reversible and without stored energy. The input effort equals the product of the same T and the output effort. Engineering approximations to the gyrator include gyroscopes. Transducing gyrators also are useful in modeling dynamic fluid machines. 3. The output flow of a transformer equals the product of a modulus. Define with symbols the radius of the shaft and the radius of the vanes/ inner housing.2. Engineering approximations to the transformer include pulley and gear pairs. Choose whether to relate ¢~ to Q (easier) or 7~/ to Ap. detailed consideration is again deferred to Chapter 10 because of complications due to energy storage. Their modeling in this manner goes beyond the scope of this text. You will examine how these elements can be combined with one another and with other elements. Check to see that the relation 2. DC motors and other moving coil devices. electrical transformers.

PROBLEMS 2. connect the two bonds with the associated standard bond-graph symbol. roller . 3.70 CHAPTER 2. and express this modulus in terms of physical constants that also should be defined. B is the magnetic field strength and i is the electric current. as in a loudspeaker. 60).38 employs X permanent magnet. 2. define the transformer modulus. 2. Identify the type of ideal machine and its modulus from the given equation.38: The translating Guided Problem SOURCE-LOAD SYNTHESIS coil device of Guided Problem 2. 2.20 Answer the question of the preceding problem for the cylindrical drive shownon the left side of Fig.6 2.30 (p. Identify the power conjugate variables on the mechanical and electrical bonds which are related by this device. Note that it follows from the conservation of energy. Give its modulus and the relation between the voltage e and the velocity Suggested Steps: 1. neglecting losses. The axial force on the coil is F = 27rrNBi where 2~rr is the circumference of the coil.19 Identify a set of variables for a transformer model of the pulley drive systems of Fig. Find the relation between e and 2 from this bond graph. The example illustrated in Fig. Model this device as a form of an ideal transducer. 2.6 A translating coil device comprises a coil wrapped on a tube and positioned in a radial magnetic field. N is the number of turns. 2.31 (p. Figure 2. 60).

2. Neglect the angle of tilt. 2.34 (p. 2. 64).25 The shaft of a rapidly spinning rotor (or gyroscope) is horizontal and supported at a distance L from its center of mass. Determine this modulus. where I = mr~ is the mass momentof inertia and r~ is the radius of gyration.m.connecting the vertical axis of the turn with the axis of tilt equals the angular momentumof the two wheels. The rotor has mass m and spins at w rad/s.02 N. The axis of the shaft is observed to precess about a vertical axis at a steady rate ~. 2.35 part (c) (p. has a radius of 1.22 You are given a hydraulic rotary actuator such as pictured in Fig.0 feet. You Inay define the radius of piston as r~.21 Answerthe question of the preceding two problems for the bevel or conical roller drive system shownon the right side of Fig. the number of pistons as n a~d the angle of the cam plate from the normal to the axis as 8.1 feet and has a radius of gyration of 1. 2.4 IDEAL MACHINES: TRANSFORMERS AND GYRATORS 71 2. The axis remains horizontal. but are not given its specifications. 60). 2. Assumethe behavior is consistent with that of an ideal machine.2. If the machine is used instead as a motor. 2.23 Find the transformer ~nodulus for a steady-flow model of the piston pump/ motor shown in Fig. the distance between the machine centerline and a piston centerline as vb. Propose and describe an experiment that could determine the modulus of the transformer that models the actuator.26 A bicyclist negotiates a turn with a 20-foot radius at 24 feet per second. determine the rate of precession. . and note whether it is a transformer or a gyrator.24 A tachometer produces 1 volt for each 10 revolutions per second. 65). (a) The gyrator modulus. Knowing that this device can be represented by a gyrator with modulus Iw. Each of the wheels weighs five pounds. find the current needed to produce a torque of 0.31 (p. 2. (b) Determine the moment produced by the rotation of the wheels that partly counteracts the centrifugally-induced tilting moment.

depth of the vanes and chamber: w. .. which can be seen to be correct. The result is M = 2rF where r = (r~ + r.. e = 2rrrNB k. Howtransformers and gyrators act when cascaded head-to-tail. or r~. Radius of shaft: r~. G -= 2~rrNB.6 1.)/2 and F = ApAr. Therefore. .72 CHAPTER 2.. The volumeof one chamberis ~r(r~ -r~)(¢/2~r)w. radius of vanes or inner housing: r~.5 1.r’2s) w(r~.1 and 2. 3.r~ = Ar is muchsmaller than either r. Electrical generalized force and velocity: e and i. The flow Q enters two of the four chambers (and leaves the other two).3 describe how to predict the equilibria of sources attached to loads.r~)(r.----2-’-~- T e~ 2 The definition of a transformer requires 03 = T2g12 = ~T~o~. 2. Therefore. ~ G ~ From the given equation.5. so Q = (1/T)~ with lIT = w(r~ . respectively. Section 2. The relation M = (1/T)AP can be checked most easily in the limit when r. Guided Problem 2.. 2.4 introduces the transformer and the gyrator. the rate of change of the voluine of one chamberis w(r~ -r~)@2. and how they can be placed between a source and load to improve the behavior. + r~). (2. SOURCE-LOAD SYNTHESIS SOLUTIONS TO GUIDED PROBLEMS Guided Problem 2. 2. respectively.1 Cascaded Transformers A pair of transformers can be cascaded: e2 e 1 ~ T1 .27a) el = T~e: = T~e3.5 Systems Gyrators with Transformers and Sections 2.27b) which says simply that the cascade reduces to a single effective transformer with modulus equal to the product of the component moduli: e~ T O~ ~ q3 T = TIT~ . (2. e F 2. is now investigated. Mechanicalforce and velocity: F and ~?.

28a) el = Te~+l. 2.2.5 Showthat gear trains. (2. and that they act as if there were only a single gear pair...5. or single transformer. can be represented as a cascade of transformers if friction is neglected. (2. the result is a single effective transformer with modulus equal to the product of M1 the component moduli: 0n+I = T(~I. with a transformer modulus or gear ratio equal to the product of all the individual transformer moduli. (2. The inverse torque ratio has the same value only if friction is neglected. as noted in Section 2.28c) EXAMPLE 2. such as those pictured in below. if n transformers are cascaded.2 Cascaded Gyrators Cascading two gyrators produces a transformer with a modulus that equals the ratio of the two gyrator moduli.4. that is all the component gear ratios.. SYSTEMS WITH TRANSFORMERS AND GYRATORS 73 Thus. This result can be deduced step-by-step as follows: e2 ~1 q2 ~3 . Solution: The transformer modulus for each pair of meshing gears equals the ratio of their numbersof teeth or of their diameters. the model is invalid otherwise.4. . T.28b) T=T1T.. The overall train acts like a single gear pair..5.

showthat if the twopairs of electrical leads are interconnected instead (below right). gives a gyrator. EXAMPLE 2. II :2 12 T=~G~ G2 This device has an important advantage over the usual electrical transformer: it works at DC. (You should take a few momentsto demonstrate this fact for yourself.2.6 Showthat if the two shafts of ideal DCelectric motorswith constant fields are connectedtogether (belowleft).74 CHAPTER 2. the cascading of any even number of gyrators produces a transformer.~ G.29) showsthat connecting together the shafts of two such motors produces a transformer: el i~ M ~G~ ~ ~G2 e~ . The key ideas perhaps are best learned by a familiar example.) As a result. 2.-7-~ a2 ~ ~ M1 -~ r~ G~ r= Energylosses are addressedin Section 4. while the cascading of any odd numberof gyrators produces a gyrator.3 Case Study to a Load of a Transformer Connecting a Source Transformersroutinely interconnect sources and loads. Equation (2. Solution: Anideal DCelectric motorwith constant magnetic field has been represented as a gyrator.---~--~:---~. In addition. The force-speed characteristics of the load are plotted on the right .4. For the second case. Consider an automobileIC engine with a fixed throttle position whichdrives a vehicle through a fixed transmission along a level road. the voltages and currents of the twopairs of electrical terminalsare related like those of an ideal electrical transformer. the result is a mechanical transmission knownas a Ward-Leonardsystem: M1 e ~ ¢. Non-transducingtransformers often are chosenexplicitly to improve the matching of the source to the load.39.5. the two shafts have torques and speeds related like those of a transformer.-~--’~ ~" e2 e~ T--. The engine can be modeledas a source. Transducingtransformers do this intrinsincally. 2. SOURCE-LOAD SYNTHESIS Cascading a gyrator with a transformer. however. the torque-speedcharacteristics are plotted on the left side of Fig.

you need either to transformthe load curve into the torque-speedcoordinates of the engine plot. Both of these transformations are shownin the figure..--.--z . so these componentscan be modeled. VEHICLE X M S~M F T F~. The two cascaded transformers can be telescoped into one.--.collectively. 0 "1 i I I 0 100 200 300 400 speedof drive shaft. or the distance moved per radian of rotation of the shaft..’N \ vehicle drag. -" . /.39: Automobilesource and load characteristics with synthesis side of the figure. This drive train converts engine torque and speed to vehicle thrust force and speed. drive shaft and gear differential.--" . F ~/~ \ equilibriumS{ / equilibriumfl~ lbs.. The modulusof this combinedtransformer equals the ratio of the velocity of the vehicle to the angular velocity of the engine shaft.------. /\ ~/ / \ lOO . To determine the equilibirum speeds of the vehicle and the engine.5. labeled as T.--. 100 M ~ ~ ¢ ft-lbs 50 R. Assumethe vehicle is knownto movetwo feet for every rotation of the shaft. Losses in this system are neglected here. R The drive axles in turn rotate the wheels. . &..2. (This force is the sumof the air drag and the rolling drag.~ ° 200 ~.... or to transform the engine characteristics into the force-velocity characteristics of the vehicle...-~-.. rad/s 0 50 100 150 vehiclespeed. This action is modeledby the transformer Tw.) The drive train between the engine and drive axle comprises a transmission.k. whichis the radius of the wheels. T = 2/2rr = 1/~r ft/rad. shown as Td below: F M ENGINE~ DRIVETRAIN ~ WHEELS -~-=7--.. therefore.. ft/s Figure 2. 75 SYSTEMS WITH TRANSFORMERS AND GYRATORS 300 // 150 engine torque.. as a transformer.The modulusof this transformer represents the ratio of the linear motionof the vehicle to the rotary motionof the wheels.

Better yet. you can simply use trial-and-error. ft/s Figure 2. Presume you wish to find the drive train ratio Td which maximizes the steady-state speed of the vehicle for a given load characteristic and a given radius of the wheels. The transformations are carried out by choosi. as they should: & = 123 if/s. the transformer and the load are effectively combined into an equivalent load. You can reason as follows: maximum load speed means maximumload power. In the first case. Note specifically that the rectangular hyperbola representing maximumpower is tangent to the . SOURCE-LOAD SYNTHESIS 0 equiliy~f/ J load for /level road 50 100 150 vehicle speed. Since the transmission is assumed to be lossless. In the second case.3. F = 226 lbs. labeled as R’. 2. you can use one of the methods described in Section 2. / equilib~F2~ "\locus.6. labeled as S’. this means maximumengine power. rad/s / 300 M 100 0 load for 10%grad~ 0. you start by finding the point on the engine characteristic which gives the greatest power. Thus.of 341 lbs ft-lbs 5O / 0 ~ q P (but opposite sign) 100 200 300 400 speed of drive shaft. ~b. The equilibrium state is given by the intersections of the characteristics.76 CHAPTER 2. which are implemented on the left-hand side of Fig. M= 72 ft-lb. \\\locus -. \maximum F 400 lbs. ~. The power is the product of the torque and the speed. The relations ¢ = (1/T)Sc and M = (first case) or :b = Tq~ and F (1/T)M (s econd case) ar e us ed.ng individual points along the characteristic to be transformed.40 by dashed lines. The transmission ratio of a real car changes when you shift gears to address a changed loading condition. ~ = 386 tad/s. the engine and the transformer are effectively combined into an equivalent source. ~ 500 vehicle drag. The two approaches give the same answer.1 3500 lbs 150 engine torque.40: Calculation of the transformer modulus for maximumpower transmission using dashed lines.

without using a larger motor.2. you’ve madea mistake if both calculations don’t give you the sameanswer.1 feet. The maximum power is approximately T) = M.242 ft/rad or T = lt.. whereits intersection with the load characteristic gives the corresponding maximumspeed and force: approximately ~ = 132. as the gears in the transmission of the case-study vehicle aboveillustrate. the drag should be reduced by about 0.4 ft/s and F1 = 256 lbs.303) (2. The weight times the sine of the angle gives an additional force of 348 lb of drag.lm/F~ = 111.900 ft.4 Second Source Case Study to a Load of a Transformer Connecting a The matching of a source and a load often is improvedthrough the purposeful introduction of a coupling that acts like a transformer. there follows Td = 0. as shownin the figure.~ = 73.lb x304rad/s = 33.. 2. The newsystem. This is added to the air drag and wheel drag to give a newload characteristic and a new solution4: T = :~2/~...) Finally. can be represented by the bond graph S M_~_~_~T A/d R Theobjective is acceptableoperation for the twentieth and thirtieth-floor sprinklers.242 ft/rad. where Twis the radius of the wheel. shownin Fig. consider inserting an ideal pulley or gear drive betweenthe induction motorand the pumpof the water sprinkler system discussed earlier in Section 2. Assumethat the vehicle and its payload weighs 3500lbs.. For example. As a secondcase study.5 ft. for whichthe direct drive failed.305) 4The solution as presented overlooks a small correction for the case of the 10%slope that results frown the fact that the normal load on the pavement is reduced by about one-half percent below the weight.5. 2.1. 77 SYSTEMS WITH TRANSFORMERS AND GYRATORS characteristic at the maximum powerpoint. if this radius is 1..lb/s. (Clearly. Since about 40 pounds of the load force is rolling friction which would be approximately proportional to this force.. . the required transmission ratio Td is the ratio of T/T. 1 -Aid = ~Mra. The locus for maximum power can be transformed over into the right-hand plot. The angular velocity and the torque of the ideal pulley drive are q~ = T~. (This drive train ratio in turn is factored into the ratio for the transmissionand the ratio for the gear differential) Achieving maximum speed on an incline requires a lower transmission ratio (down-shifting). The transformer ratio which permits this to occur is therefore T = k~/~l = Mm/FI= 0.5/a04 = 0.396.2 lbs.4355 ft/rad.5.5/46~ = 0. and is climbing a long 10%grade. and that the slope of the chord from the origin to this point has the sameslope as the tangent to the characteristic at this point.n~m= 111.41. (2.

N’mF t o 2. rad/s Figure 2.io. SOURCE-LOAD SYNTHESIS Figure 2.42: Characteristics the pulley drive of the source combining the induction motor and where T is the ratio of the diameter of the motor pulley to the diameter of the pump pulley.8200 150 250 ~e. S’~Md R /An acceptable alternative combines the pulley system and the pump system into an equivalent one-port load.41: Addition of pulley drive to the induction motor system ~ locus for maximum power S’ ~ 3.~ f’\ . This is muchlike the choice you have in shifting gears of an automobile with a ~nanual transmission.0I 0 \~. The motor and the pulley system can be considered as a single equivalent source. Characteristics for four different values are plotted in Fig. all for the same motor speed and torque. Each point on the basic characteristic is mappedonto . whereas a small value of T (as with "first gear") gives low speed at high torque. The other characteristics are transformations of this basic characteristic..78 CHAPTER 2. 2. A large value of T (as ~vith "fourth gear") gives a high speed at low torque. the ideal pulley acts like a direct drive.0 M~.) The torque-speed characteristic of this new source depends on the value of T.42. 50 ~ locus for ~ maximum 100 one-half power . The characteristic for T = 1 is identical to that of the motor itself.

which in turn equals the largest possible source power. SYSTEMS WITH TRANSFORMERS _x locus t 3..The power. Rapid sketching is aided by the fact.0~ Md’ | N. The transformed point lies on a locus of constant power that passes through the original point. and for the tenth floor it is about T = 1. i.2.¢. The maximumpower that can be delivered by the source is indicated in Fig. that the rnagnitude of the negative slope at each point on the hyperbola equals the positive slope of the chord from the origin to that point.) show practical operating points ~\ for systems permitting reliable start-up jthirtieth floor ~ I xx Ma 1..43: Matching of motor/pulley system to pump/sprinkler load a correspond!ng point on the transformed characteristic according to equation (2. the intersection is close to T = 1.e. 2.~. The respective lower-speed intersections have the same speed and torque ratios as both the upper-speed intersections and the maximumpower points. using only knowledge of the values of T. This locus is a hyperbola with vertical and horizontal asymptotes.associated with the point is unchanged by the transformation.34).43 this locus is seen to intersect the load characteristic for the thirtieth floor whenT = 0.1 (interpolating). The largest possible load flow corresponds to the largest possible load power.42 by the locus of maximumpower. Are these solutions acceptable? The candidate design solution with T = 1 for the twentieth floor already . This locus intersects each source characteristic twice. 2. The various speed ratios and torque ratios equal the transformer moduli or their reciprocals.81 For the twentieth floor.5. The locus for maximumpower just graces each characteristic. All of these points of tangency represent the same maximumpower the motor can deliver.m / / 79 AND GYRATORS for maximumpower ~~//circles (. MdCa= M. Twosuch loci are drawn in Fig. the same rectilinear hyperbola discussed in the prior example. find and validate the corresponding point or points on another characteristic. The other locus drawn in the figure represents one-half the maximumpower.42 as dashed lines. as noted above. It is strongly suggested that you choose some arbitrary point or points on one of the characteristics and.3 0 0 50 100 150 200 250 rad/s Figure 2. 2. In Fig.

.. SOURCE-LOAD SYNTHESIS has been eliminated.80 CHAPTER 2. starting is too problematical to justify the choice. and the pressure drop across the restriction is plotted below right as a function of the flow through it. Thus. rad/s ~ ~.8 and 0. the latter gives virtually maximumpossible flow. both T -.5 for the thirtieth floor suffer the same problem. Then.5 in3/s. the starting torques for the motor and pump are about equal. moment.M~imumpowerfrom the motor occursat the mid-point of thestraight-line characteristic.2.6andshown specifically in Fig.Thisis the solution whentheloadpressure is zero. A locuswith thispoweris superimposed on the pressure dropplotbelowleft. The solutions T = 0..5 is seen to be acceptable for the twentieth floor. For T = 1.1. in3/s Assuming thepressure of theloadis zero. For the tenth floor. feature highlighted in Guided Problem 2. EXAMPLE 2.~d theassociated fl0w.0.3. unless you were to replace the motor.~ restriction- 0 1 2 3 4 5 6 flow.revealing an intersection ~t a pressure of 1000psiandflowof 3. The acceptable solutions are marked by heavy dots in the figure.. including friction which is being neglected.7 A DCmotor drives a frictionless and leak-free positive displacement pump which forces a liquid through a flow restriction to a load: fluid line restriction The motor has the torque-speed characteristics plotted below left.so thatthe pressure seenby the pumpequalsthepressure dropacrossthefluidrestriction. the solution for T = 0.8 and 1. lb/in2~ 1°°° f /’.repeatthe solution whentheloadpressure is knownto be 1000psi.ib. thismeansmaximizing the power from the motor. ~mr 1 8 4 pressure I’ ~ ’ ’ ’ ’ ’ ~l~ ’ drop. for example using the "capacitor-start".findthedisplacement of thepump that(approximately) maxilnizes theflowto theload. Allowing about a 10%margin to cover errors inherent in the modeling. producing thepower500x 7 = 3500in. Sincethe pumpis 100%efficient. as discussed in Section 2. Solution:Maximizing the flow to the load impliesm~imizing the power transfer.Thusthemotorshouldbe run at 500rad/switha torqueof 7 in.2.Ib/s.27. . ~ 0 500 I000 speed. since unfortunately the pump won’t start. you must settle for a lower output flow and power in all these cases.0 are acceptable.

.~restriction~ 0123456 flow.~.2. You wish to find the equilibrium torque and speed. The system can be modeled by the bond graph S e~G ~ M ----U The given information M= ai establishes the gyrator modulusas a. \ / solution pressure. This sumis plotted belowfor the secondcase of a 1000psi load pressure. Youcould combinethe source and the gyrator into an equivalent source..005 in3/rad. and the mechanical load has the torque-speed characterisitic shownon the right side. ~’t~t~l’~’~ ’4 lb/in21000 lO0~O~psi~. lb/in 2 ~ ~ /" ~ ~ .~ drop.007 in3/rad. which can be deduced from the ratio 3. t ~ . neglecting losses. 2.in3/s The pumpacts as a transformer. as shownabove right.which drives a load. The pressure generally seen by the load is the sumof the pressure drop across the restriction and the load pressure.44. so .. SYSTEMS WITH TRANSFORMERS AND GYRATORS 81 ~ocus of maximumpower pressure ~. .0. . The modulusof the transformer that produces this result is the volumetric displacement 2.5 Case Study of a Gyrator Load Connecting a Source to a Consider a DCmotor with constant field and momentM = ai. so you also knowthat e = a~.005 in3/rad or 7/1400 = 0.~ restriction~ 0 1 234 56 flow. a = 0.5/500 = 0.007 in3/rad or the ratio 7/1000 = 0. Its volumetric displacement is the modulusof the transformer..5/500 -. 2.5 in3/s.5. The electric power source for the motor has the voltage-current characteristic shownon the left side of Fig. . .ina/s The intersection with the same power locus as before is nowat 1400 psi and 2.15 N m/amp.\.5..

6 Transmission Matrices* The use of matrix notation expedites the telescoping of a chain of linear twoport elements into a single equivalent two-port element. and vice-versa..~kS’ dfor G =0. 2.5..~M 2O k. with its effort first and its flow second. . The mappingis shownin the figure by the dashedline labeled S~. rad/s 160 Figure 2. so the first approach is easier.15 N. Note that the left end of the source characteristic becomesthe right end of its transformation. SOURCE-LOAD SYNTHESIS S ~ G ~R 0. Further use of the methodis given in Chapter 4 and beyond.m/am N. volts " 10 0 0 ~ 1 0. you could combinethe gyrator and the load into an equivalent load.. This idea is introduced here for the special cases involving only transformers and gyrators. The state of a bondis treated as a two-elementvector. The equations describing a transformer of modulusT and a gyrator of modulus G become ..44: Source and load characteristics for DCmotor system that the source characteristic is mappedinto the load plot: S’ M R ¢ Alternatively.4 s [ \x~\ S "~quilibrium 2 3 i.m . so that the load characteristic is mappedinto the source plot: S e ~ i R’ The source characteristic is a straight line in the present case..82 CHAPTER 2. Younowsimply identify the intersection of the two characteristics as the equilibrium point. ~ps 0 4 40 80 120 ~.

7 Summary Transformerscascaded head-to-tail can be represented by a single transformer with modulusequal to the product of the individual moduli...’--’-7 - e2 T q2 __ 01 ez q3 gives l/TIT2 (t3 01 = T1T2 03 Denotingthe transmission matrix of the ith element in the cascade of n transformers as Ti. The source then can be combinedwith the transformer or gyrator to give an equivalent source.1. -el ~T1 . T:TIT9 """ rn. Graphical representations of the characteristics enables you to visualize the solutions of a . Modelsoften interconnect a source to a load by a transformer or a gyrator. which can be directly synthesized with the source.5. SYSTEMS WITH TRANSFORMERS AND GYRATORS 83 The cascading of two transformers. which can be directly synthesized with the load. is equivalent to a transformer with a modulusequal to the ratio of the individual moduli.33) whichagrees with equation (2. This meansthat an even numberof gyrators in cascade gives an equivalent transformer. there results 01 =T~T~ . while an odd numberof gyrators in cascade gives an equivalent gyrator.29). Alternatively. qe 01 q3 gives ° oll G2 l/G2 which agrees with equation (2. =[~ 0 ]lIT[en+l] [en] .28). the transformer or gyrator can be combinedwith the load to give an equivalent load.5. Jell [O~+~J’ 0.~__~ 2 ~. 2.5.T. The cascade of two gyrators. A pair of gyrators in series. on the other hand.. (2. A gyrator connected to a transformer gives a gyrator. as in Section 2.2.

Guided Problem 2. . The transformer modulus sought is the ratio ~/Q of these two points. Suggested Steps: 1. transformer and load with a bond graph.help you understand howto choose the modulusof a coupler so as to maximizethe performanceof an electrical.7 Solving this problem and the next should . Express the powerdelivered in terms of the bond Current and the parameters. Drawa bond graph of the system. The characteristics of the power supply and the load are plotted in Fig. Combinethe transfor~ner and the load and draw a new bond graph. Matrix relations can be used to represent the characteristics of linear twoport elements such as transformers and gyrators. Find a relation to characterize the source. 4. 2. Neglectfriction. Find the current that maximizesthis power. This approach is helpful but not necessary. powerpoint on the source characteristic. 5. 5. Represent the source. 2.8 A hydraulic power supply drives a hydraulic ram (piston/cylinder) against mechanical load. Checkthat the transformer modulusis also the ratio P/F for these two points.45. mechanicalor fluid system. Suggested Steps: 1. Evaluate the modulusof the new load. Find the maximum 3. Find the correspondingpoint on the load characteristic. such as determination of the moduli which give maximum powertransfer from the source to the load.84 CHAPTER 2. Choosethe modulusof an electric transformer which connects an amplifier with a 24-ohmresistive output impedanceto a speaker with a 6-ohmresistive impedanceso as to maximizethe powertransfer. Relate this ratio to the area of the piston. Guided Problem 2. 3. Find the transformer modulusthat gives this maximizedpower. Find the area of the piston which maximizesthe power delivered to the load. 2. 4. SOURCE-LOAD SYNTHESIS variety of problems.

5 I I 1.1 2. in. in3/s I 0 I I 0. out i 201 28 15i numbersof gear teeth ]] ! I 12 i ! I . Find the moduli of the componenttransformers and the overall equivalent single transformer.0 1. .000 1000 20.8 PROBLEMS 2. and. lbs 50.0 Figure 2.i:.000 F.28 Find the relationships betweenthe boundaryvariables for the following cascade.000 2000 40.000 0 I 0 I 0 20 30 40 Q. 85 SYSTEMS WITH TRANSFORMERS AND GYRATORS [ nower I tube "40013 ¯ P.5 .27 Modelthe gear train below as a cascade of transformers. Canthis systembe represented by an equivalent single transformer or gyrator. what wouldits modulusbe? el~T ~2 q2 G ~3 q3 2. if so. psi 30013 80.29 Carry out the previous problem using the methodof transmission matrices.5./s I I 2.2.45: The system of Guided Problem 2.4 I q~in 2.

4. (b) find the engine speed. 2. 75).5. 79) that corresponds to the bonding of the actual motor to an equivalent load: S M.33 An ideal hydraulic power supply (comprising a motor. pulley drive and pump load given in Section 2.. (d) find the acceleration of the vehicle. lbs 40O0 2500 pressure. pump and relief valve) has the pressure-flow characteristic as plotted below left.30 The vehicle considered in Section 2. 2. (a) find the drag force. Determine which of the candidate solutions T = 0. SOURCE-LOAD SYNTHESIS 2. 6000 force. Neglect friction. Using only the data given by the solid lines in Fig.32 The analysis of the system with the motor.5. moment and power. pulley drive and pump load analyzed in Section 2.2 (p.3 are acceptable for the tenth.5./s . Find the maximum possible speed of the piston.86 CHAPTER 2.31 The capacitor-start motor described in Guided Problem 2. 2. and the area of the piston that results in this speed. in. 2. 1.3 (which has T = 1/~r ft/rad and mg = 3500 lbs) is traveling at 100 ft/s on a level road.5.39 (p. 26) substituted for the basic induction motor used in the system with the motor. (c) find the thrust force (from the engine power and vehicle speed). It drives piston-cylinder with the characteristic shown below right. psi 2000 0 0 30 in3/s flow 0 0 6 12 18 velocity.0 and 1. 0. ~ R’ 2. twentieth and thirtieth floor sprinklers.4 employs an equivalent source bonded to the actual load: St Md Draw an annotated sketch of the replacement for Fig.43 (p.8.

5. 2. The powersupply is regulated to give constant voltage up to a limiting current. i i 24~~--~~~ i. also plotted below. 12 torque. as plotted below. rad/s 300 (a) Drawa bond graph modelfor the system. SYSTEMS WITH TRANSFORMERS AND GYRATORS 87 2. in. so poweris conserved. 1 0 400 800 1200 1600 speed. (c) Determine the corresponding speed and torque of the fan. 3. The load has the torque-speed characteristic. Neglect losses in the motor. and the pulley ratio that achieves them.The pulley ratio is not specified..-lb (to prevent overheating). and find this power.2. . amps o 0 100 200 ~. The motor has a constant magnetic field.0 in. rpm (a) Drawa bond graph model of the system. but the motorshould not be run with a torque higher than its continuousrating of 4. i . (b) Determinethe speed of the motor that maximizesits Mlowablepower. neglecting energy storage elements.01 e. Neglectbelt losses. The torque-speed characteristics of the motorand the fan are plotted below. (b) Determinethe maxirnumspeed that the load can achieve.35 An electrical power supply drives a DCmotor that in turn drives a rotational load. .34 A motor drives a fan through a belt drive.

2.5. . 54) using the reticulated F ~T P~R ~ Q 2. and determine its mechanical resistance. Rm.45 (p.39 Repeat Problein 2. The load torque on the shaft of the motor is M = a~ where a = 100 in. as shown below. and find the value of D that maximizes the speed of the load. S e = RLT’~ir T i 2. 2. R~. and is otherwise ideal.9 (p. Draw a bond graph model. S e~. (a) Determine the maximumpower that can be transmitted to the load. (b) Determine the value of a that accomplishes this transmission. R~ = 24 Q .10 (p.38 The armature circuit of an ideal DC machine with constant magnetic field is completed with an electrical resistance Re. 2. In the present case. The machine is characterized by the relation M= ai.. SOURCE-LOAD SYNTHESIS (c) Evaluate the modulus that describes the basic property of the motor.~i.37 Consider the system studied in Section 2. 2. an ideal DC motor with a constant magnetic field and a particular mechanical load with characteristics given in Fig. 82). 85) drives a rotary hydraulic motor with volumetric displacement D per radian.44 (p. SOLUTIONS TO Guided Problem 2. 2.7 1. in place of the piston load shown in the figure.36 A pumpwith the characteristics given in Fig. 2. 54) using the reticulated above.5 which comprises an electrical source. Show that the system acts like like a linear rotary mechanical dashpot. the modulus a is not specified. GUIDED RLTi rt~L Ti 2R~ = RLT model model of the problem PROBLEMS e =eo--R.88 CHAPTER 2.40 Repeat Problem 2.lb s.

~)RL Guided Problem 1. also.0.in.000 locus 2000 40. ’ 1:5 ~ 210 Y¢. ~ = 0 = eo .1 in-’.in-’/s of maximum~ 00 0:5 l:0 -r-. 79 = ei = (eo .. *--~= 0.-R F ~’’~imum power’~ max 80.050 in2. 1/(pistonarea) = P/F.~i.R.69 4.2.5. 4000 7 P. 5. S--~ 2.2R. = (eo/2R. e = RLli = RLT2i.000 00 -.¢~ = 1.0013 1000 20.8 T ---~--. lbs 60.3.2. i = ~R-’~’~’ e = eo/2. psi 3000 = 2.~i)i d79 eo 4. 5.000 F. Solving for T and using the above. 10 20 30 40 Q. SYSTEMS WITH TRANSFORMERS AND GYRATORS 3. Therefore. The piston area is l/T= 20./s 89 .

Chapter 3 Simple Dynamic Models The systems considered thus far respond instantaneously with a new state whenever their environment or boundary conditions change. The conversion of the bond graph models to differential equations in presented in Section 3.7. The behavior of the models. finally. The variables which define its state would approach their new equilibrium either monatonically (non-oscillatorily) or oscillatorily. a source. Section 3. Nonlinear compliances and inertances. Section 3.6. 91 . are introduced in Section 3. called inertances.1 starts the subject by introducing a major class of energy storage mechanisms called compliances. This chapter makes a first pass through the major ideas in the modeling and analysis of dynamic systems. a compliance and an inertance joined together by a junction. The dynamics can be thought of as a sloshing of energy between different energy storage mechanisms. for later release.2 continues by introducing the other major category of energy storage mechanisms. To accomplish this. This leads. and/or a gradual dissipation of energy in resistances of the type you have examined already.5. Junctions are virtually the only elements that can interconnect three or more bonds. Dynamic physical systems contain mechanisms that store energy temporarily. as represented by the solutions to the differential equations. Whena real physical system is subjected to a sudden change in boundary conditions.3. and therefore become the very heart of the structure of a bond graph. They are introduced in Section 3. Such systems. assuming stability. it might in fact respond noticeably slowly. to numerical simulation in Section 3. however. at most. are called dynamic. a resistance. Dynamic models are represented mathematically by differential equations. previously omitted. or the models that represent them.4. attention is restricted to simple models comprising. as contrasted to static or steady-state models that are represented mathematically by algebraic equations. is the subject of Section 3.

xl~’). Compliancesare designated in bond graphs by the capital letter C. sinks and resistances consideredin Chapter2.1 Compliance Energy Storage Elementsthat store energy by virtue of a generalized displacement are called compliances. x is defined so as to equal zero whenthe spring force is zero.1. Physical approximations include mechanical compliances (e. (3. springs). 3. (3. required to compress or stretch the spring from x = Xl to x = x~: Wl-~2 = F~ df F dx = kx dx = -~k(x2 . amongothers. so as to computethe input work.2.1.2) As suggested in Fig. as resistances do.g. capacitors). 3. the spring is characterized by a relationship between its generalizedforce and its generalized displacement. As illustrated in Fig.1 Linear Springs and Energy The mechanical spring is a familiar example of a one-port element that is totally different from the sources. fluid compliances(e.g. they store energy. If the process is subsequently reversed and the spring is returned to its original position. this workis the area under the spring characteristic between the two values of x. the spring constant or the spring stiffness. Note that in equation (3. 3. gravity tanks) and electrical compliances(e.~. the additional workis exactly the negative of the original work: W~-+l= ~k(xl = .g. This can be shownexplicitly by integrating the powerflowing into a spring over time. Springs do not dissipate energy. SIMPLE DYNAMIC MODELS F Figure 3.1: Linear spring and spring characteristic 3. WI~.rather than its velocity: F ~ SPRING F = kx.1) The coefficient k is knownvariously as the spring rate.92 CHAPTER 3.1).

¢: M = k¢. (3.6) are replaced by e and q.7) The same result applies when the M and ¢ of the rotary spring characterized by equation (3.1. Wl-~2 = ~22 - ~1. the energy it stores is _-1 ~=11F_~ 2 kx" 2k (3. 93 COMPLIANCE ENERGY STORAGE XI -~’2 X Figure 3. The constitutive relation is e = ~q. in general. just as the symbol R is used for both the resistance and its modulus. a constant. which is the time integral of the angular velocity. The result is summarizedgraphically in Fig. and its. Further. ." so C = l/k. ¢.3. for both translational and rotational mechanical springs. and replacing the displacement x by the generalized displacement q. (3.1. Note that the modulus of a compliance is defined in a manner similar to the definition of the modulus of a resistance. M.6) 3. displacment is an angle. 3.1) by the generalized force e. "Compliance" means the inverse of "stiffness.2: Work as area under characteristic It is customary to say that energy is conserved by the spring.5) A rotary spring acts like a compression or tension spring.3. except its effort is a moment. In bond graph form this is written and is called a compliance element or just a compliance. the symbol C is used for both the compliance and its modulus. (3.2 The Generalized Linear Compliance A "generalized linear spring" results from replacing the force F of the compression or tension spring in equation (3.4) so that. respectively.

. A capacitance usually is considered to be constant.1. The coincidence that "capacitance" and "compliance" both begin with the letter "c" contributed to the selection of that letter to designate the generic ratio of generalized displacementto generalized effort.4): "1?=-~ ~ q =-~ Ce " . Its applied voltage e and charge or electrical displacement q are modeledas q = Ce. (3.4 Linear Fluid Compliance Due to Gravity The mechanicalspring and the electrical capacitor have a generalized force which is proportional to a generalized displacement.7)~ revealing that.3 Electric Circuit Compliance An electric capacitor is indicated by its traditional symbolin Fig. (3. and that the capacitance C is analogous to the compliance C. The traditional symbol is capital V. the modelis linear. 3.4. 3. the capacitor is modeledas an electric circuit compliance. whichin this text is given as script Y (to distinguish it from volume). Think for a momentwhat this could be.94 CHAPTER 3.9) This equation agrees precisely with equation (3. A fluid complianceshould have the same thing: a pressure proportional to a volumetric displacement. 3. SIMPLE DYNAMIC MODELS e q Figure 3.8) This is knownas a potential energy.1.3: The generalized linear compliance ea C (a) circuit (b) bond graph i ~-C Figure 3.4: Capacitor as a compliance The energy stored in the linear generalized compliancefollows from substitution of the generalized variables into equation (3.

is the ratio of tile volume to the pressure: [c=. The equation of state for a perfect gas is P = pRO.pg"A (3. A fluid port enters at or near the bottom of the tank.e q-VP pgV/AV I .dP (3. where 0 is absolute temperature. the pressure and density of a perfect gas with constant specific heats are related by p/pk =constant. which can be designated either as ~’z or Q.3. 3.5: Liquid gravity tank as a compliance The most commonanswer is a fluid tank of constant area independent of the depth of liquid in it. Assuming small changes in density only.5.000 psi.10) The pressure in the tank at the level of the port equals the weight density of the liquid times the depth: P = pgz.12) Note that the time derivative of the volume of liquid in the tank is the volume flow rate into the tank. as shown in Fig. The depth of the liquid equals the fluid displacement V. such as found in considerations of air acoustics and nearly all problems involving the compressiblity of liquids. is treated as a constant.13) reveals that fl = P0 for small isothermal changes. Differentiation of this equation and comparison with equation (3. it is conventional to assume dP = ~--. for water with no gas bubbles this gives about one percent of 300. 3. 95 COMPLIANCE ENERGY STORAGE surface area A P~C Figure 3. C. for example. which is the volume of liquid in the tank above the port divided by the cross-sectional area of the tank: z = VIA.13) P in which the bulk modulus. (3. In the absense of heat transfer. Upon . that a one percent change in density is associated with a pressure change of virtually one percent of fl. (3.11) The compliance of the tank. P0 is the mean absolute pressure.5 Fluid Compliance Due to Compressibility A second type of fluid compliance is associated with the compressibility of the fluid. This equation says. fl.1. or 3000 psi.1.

de dP = -~-" (3. in the absence of heat transfer. A one percent change in volume of air at atmospheric conditions therefore. this gives/3 = kPo.4 * 14. would be associated with a pressure change of about . where k is the ratio of specific heats. w "’x applied cm+~)moment Solution: the block triangular side. EXAMPLE 3. The fluid compliance of this system is C = q dq _ Vo dp/p Vo I e . a area shown shaded below emerges above the waterline on the left an equal triangular area is submerged on the right side: .01 * 1.96 CHAPTER 3.6.206 psi. 3. the effective compliance would be larger. Find the compliance associated with small rotations induced by an applied moment about an axis normal to the page.1 A uniform rectangular block has half the mass density as that of the water in which it floats. i. For a clockwise rotation through a small angle ¢. SIMPLE DYNAMIC MODELS v (< <vd Figure 3. and "l ]~ ~ The center of mass is on the waterline.7 = 0. Note that this compliance is approximated as a constant. A rigid chamber of volume V0 filled with a slightly compressible fluid is shown in Fig. so it remains fixed when is rotated.14) Should the volume of the chamber expand with increasing pressure.6: Fluid-filled rigid chamber as a compliance differentiation.

3. The corresponding compliance. This symbol doubles as the modulus of the compliance. 3. The generalized displacement is the time integral of the generalized velocity or flow.6 Summary The reversible storage of potential energy can be represented by one-port elements for which the effort is a function of the generalized displacement.6. which stands for compliance. as shown in Fig. the electrical capacitor with its voltage and charge. This is the momentper unit length of the block in the direction normal to the page. times the weight density of the water. Guided Problem 3. F Figure 3. These forces act through the centroids of the triangles to give a pure moment or couple.pgw 12 3. The couple equals the force times the separation between the centroids.7: Guided Problem 3. Find the compliance of the system with respect to the displacement x and force F.q/e. pg. is the ratio of the angular deflection to the applied moment. by definition. the open fluid tank with its pressure and liquid volume. or pgw3 . Examples modeled by compliances include the mechanical spring with its force or momentand linear or angular displacement. or C--. COMPLIANCE ENERGY STORAGE 97 The buoyancy forces on the two sides change by the area of this triangle.7.1 .1. The rotation of a lever is resisted by a spring with rate k. and a container confining a compressible fluid.1.3l~I¢ . For the linear or constant compliance this modulus is the inverse slope of the characteristic. Such elements are designated by the bond graph symbol C. (w¢/2)(w/4) = w2¢/8.1 It is essential that you grasp the idea that the definition and value of a compliance depends on the displacement variables to which it is referenced. This problem is basic in this regard. Nonlinear compliances are discussed in Section 3. while the second problem is more sophisticated. that is C -.

The internal radius is r. 3. Suggested SIMPLE DYNAMIC MODELS Steps: 1. which you could call xs. although presently it is more important that you learn the method above. and find the normal axial stress. and wall thickness is t << r and the length is L >> r. Draw a cross-sectional pressure forces across the diameter and the resisting peripheral forces ta on the shell. of steps 1 and 2 to get the energy of the spring as a of step 3 to the known form. Place the 1. ]2 = (1/2C)x ~-. Draw a side view. Neglect the effects of the ends of the cylinder. and solve 5. Then. as a function of the parameters a and b and the displacement variable x. Use the results of steps 1 and 2 and the elastic parameters of the material to find the peripheral and axial strains. (a) Estimate the compliance of the chamber. Find the ratio of the change in volume to the original volume. as a function of the parameters a and b and the effort variable F. 4. Equate the result for C. and solve for the peripheral normal stress. You might also wish to solve for C using a more familiar force-based analysis. .2 A long slender-walled cylinder expands under the influence of an internal pressure. having a bulk modulus Suggested to a fluid which Steps: end view of one-half of the cylinder. Write the energy stored in the spring as a function of xs. Guided Problem 3. compute the ratio C. Finally. 3. Solve the equation from step 1 for x. of course. which you could call Fs. Find the force on the spring. = x/F. Balance the forces. and no variables. The answer. creating an effective compliance relative to an incompressible fluid within. 2. Combinethe results function of x. should contain only parameters. 2.98 CHAPTER 3. (b) Estimate the effective overall compliance relative itself compressible. The wall material has Young’s modulus E and Poisson’s ratio #. Find the displacement of the spring. 4. using the results of step 3 and the assumption of small strains. solve this equation for F.

1. Define parameters as needed.2 Find the compliance of a shaft in torsion with respect to its overall angle of twist. Find the compliance.3. then there are two compliances bonded to a single junction.1 Derive the compliance of a solid body floating on the surface of a liquid with respect to small vertical displacements. which is the ratio of the change in volume to the applied pressure. 6.3 A pendulum comprises a point mass attached to the end of a massless rigid rod which is pivoted about its upper end. the correcting (b) Find the compliance by considering the gravity energy associated with 1 2 the deflection.) 3. 3. . Determine whether this is a 0-junction or a 1-junction. 7. and find the combined compliance. If both the shell and the fluid are elastic.4 Solve the preceding problem substituting a uniform slender rod of mass m and length L for the massless rod and point mass. PROBLEMS 3./~. assuming the shell is rigid and the fluid has a finite bulk modulus.. Interest is restricted to small angles of deflection from the vertical equilibrium. which also has a third bond to represent the flow that can enter or leave the chamber through a port. Define parameters as needed. (a) Find the gravity-induced compliance by considering momentfor a deflection from the vertical. Hint: Use the approximation cos 0 ~_ 1 . 3. COMPLL4NCE ENERG’Y STORAGE 99 5. Evaluate the compliance.~0 . (You may refer to your textbook on the mechanics of materials.

SOLUTIONS TO GUIDED PROBLEMS Guided 1.6 Find the compliance of the piston/cylinder shown below with respect to F.~_ xa~-b ° 1 2 1 2-4. (b) Find the compliance of the disk with respect assuming it has only small values. Assume x << L. 12 = =kx~ = =k 2 2 a ~ = lx2. 3._. ( )~ x’~ --2C . Problem 3.5 For the rolling disk of mass m and radius r shown below: (a) Define a variable to describe the displacement of the disk from its equilibrium position.7 Verify that the bulk modulus for acoustic (small) and adiabatic compression of a perfect gas is ~ = kPo. and presume (a) isothermal behavior (b) adiabatic behavior. x.theref°re. SIMPLE DYNAMIC MODELS 3.a.1 Xs -~. where k is the ratio of specific heats and Po is the mean absolute pressure. 3. neglect friction and leakage.100 CHAPTER 3. 21 C= (-~) . to the displacement. Find the potential energy of the disk in terms of this displacement.

# ~-~ 4. neglecting V 6.1. AV = (2.#a~) E--~ .5 Dc = -- p ~rr’2L ~ ~ P Q=~/~ C=Cc÷Ci C =~rr2L (2.#aa) = 1 ea = ~ (aa . Guided C -= F = ~ Problem . z#) 5. C~ = /~ Q = _~p o ~’¢" "’’~Q C~ end effects.5- #)~-~ ¯ ~rr3L . therefore ap = Pr/t 2~rrtaa---. which is the same as above. volume: V = ~r[r(1 + ep)]2L(1 + e~) ~rr2L(1 + 2ep+ ea ) 7¢r3Lp change in volume: AV ~_ ~rr2L(2ep + ~a) = (2.2 2tap = 2rP. COMPLIANCE 101 ENERGY STORAGE that F ---- x~. therefore aa = Pr/2t 2 nrt = ap/2 a~~ 3. Fs = From step 1. 3. peripheral strain: axial strain: ( ep = ~(ap .. a Finally. x ----a+b~xs.2~) . For the fluid alone.5 ~. For the chamber alone.~rr2 p.3.

~ = T~ .17) 1 m . however. as its name suggests.) 3.8.102 CHAPTER 3.p = m2.2 Inertance Energy Storage Inertance energy storage. as shown in part (b) of the figure. Like potential energy. and is given in this text by script T (to distinguish it from transformer moduli). Consider the mass m of Fig. designated by the symbol I.2. also knownas generalized kinetic energy.15) to give ~:- fFdt m _ p m (3. The same type of development that gives equation (3. (More complicated models are considered in Chapter 10. SIMPLE DYNAMIC MODELS P (a) mass (b) characteristic Figure 3. and a corresponding linear momentum.%.16) The kinetic energy is designated traditionally by a capital T. Most models of this storage are representable as one-port elements.~ This energy is represented by the area under the characteristic. is associated with motion.18) .)2 d2 1. 5. 3. is the second of the two categories of energy storage. which is associated with displacement. Momentum and Kinetic Energy Kinetic energy.1 Mass. unlike potential energy.5) for the work associated with a change of displacement of a compliance gives the work required to change the velocity or momentumfrom state 1 to state 2: W~. (3. kinetic energy is potentially recoverable.8: Accelerating mass and characteristic 3. which is accelerated by a force F(t) from rest in an inertial reference frame to a velocity.(3. Newton’s law gives d2 F -m’ dt which can be integrated (3. It equals the work done by the force: T= F/ dx / = m-~2 dt d~ / = = -~m~" = 2m ._~p.

2. The kinetic energy is 31¢ or ½P2/I. The corresponding"constitutive relation" for the generalized inertance is (3.2. The generalized momentumis defined by lp=/edt or e-= (3. This is precisely equation (3.21) P’I The relationship of equation (3. I. (3. 3.9 part (a).16) if 0 is replaced by 2 and the generalized inertance. the inertance is said to be linear.20) which is a linear relation betweenthe generalized momentum and the flow.3 Common Engineering stant Inertances Elements Modeled by Con- Arigid bodywhichrotates about a fixed axis. 3. . Youshould not write J in the bondgraph. is replaced by m. in which case you can write I = ~/ for the modulus.19) which is a linear relation betweenthe generalized displacementand the effort. I. 3. such as a flywheel.3. p is the angular momentum. resists angular acceleration because of its massmomentof inertia. (Some authors use the symbol J to represent the mass moment inertia.8 changing the labels ~ to c~ and mto I. The generalized kinetic energy of a constant inertance is which can be comparedwith equation (3. so that I = mis a constant.8 is proportional to the velocity.since it is not a defined bondgraph symbol.17). consistent with equation (3.2 103 INERTANCE ENERGYSTORAGE The Generalized Linear Inertance The generalized compliance has been described by e = q/c. however. The bond graph symbolfor inertance uses the letter I in the samewaythat the complianceuses the letter C: e=$ I The momentum of the mass in Fig.22). 3.2. as suggested in Fig.20) is represented in part (b) of Fig. 3.) In this application.

9. of rotation.9: Examples of constant inertances EXAMPLE 3.104 CHAPTER 3. as shown in part (b) of Fig. ~mRfor a uniform disk of radius R and (1/12)m(2R) 2 for a rod of length 2R. ~ (a) rotational inertia SIMPLE l=J DYNAMIC MODELS M I=L (b) electrical inductor (c) fluid tube P~ t’IIIQ IIII/1 I=pL/A " Q Figure 3. the modulusI equals . Solution: The angular momentum is and the kinetic energy is where r is the radial and the integrations the first equation to with O = ~.2 Deduce a formula for the value of I for a rigid mass rotating about a fixed axis. This is the mass moment of inertia. all about their respective normal centroidM axes. In this application. is I = f r2dm.20) or the second equation to equation(~. the results should agree. familiar from introductory dynamics courses. It also can be modeled in bond graph language by an inertance. 3. An inductor in an electric circuit impedes the rate of change of current. for example. the result distance of the element dm from the axis. Either comparing equation(a. mR2 for a slender hoop of radius R. Use a momentummethod and then an energy method. It gives. p is still fe dt. are carried out over the entire body.22).

2. L.3 Deduce the inertance of an incompressible fluid of density p in a uniform channel of area A and length L. Many engineers use the symbol I for inductance anyway.10: Inertance of two cascaded fluid channels . Q --~ ~A~ If- pL.10. and agrees with the right-end term of the previous equation. 3. T= 5mv"= (pAL) =5 Here.p-. The term on the right within the parentheses is the desired inertance.9. The standard variables e = P and 1. since it assumes that the velocity is uniformly equal to Q/A whereas the velocity is more apt to be nonuniform over the cross section. of two cascaded constant-area channels sum. m is the total mass of fluid in the tube and v is its commonvelocity. entrance and exit losses will need to be added. the inertance relation presented here will not fully represent the behavior when the dynamic pressure is significant relative to the static pressure. Use both a momentumapproach and a separate energy approach. The inertances Fig. 3. ~ = Q can be used EXAMPLE 3. A1 12 = pL2 -~2 Figure 3.3.dp/dt Q dQ/dt by the successive _ P1. Approaching this result from the perspective of kinetic energy. Therefore. Its use in a fluid circuit also can be represented in a bond graph by an inertance. Solution: The momentumapproach can be represented steps in the multiple equation below: I-. and in fact somewhat underestimates the inertance. as notec~ before. P1 . The resulting formula is approximate. The mass of a fluid in a tube impedes the rate of change of the volume flow rate. in particular.P~_ p__~_L A dv/dt A ’ The final step above directly uses Newton’slaw. as suggested in iUse of e = P neglects the dyr~atnic pressure. 105 INERTANCE ENERGY STORAGE 1 "" 1 "2 the inductance.P2 equals the net force divided by A times the mass pAL and the acceleration dv/dt. as shown in part (c) of Fig. and the generalized kinetic energy is _~Lwor ~p /L.

4 Tetrahedron of State* The four essential variables of state for physical systems are represented by the vertices of the tetrahedron of state. or =p (3. as suggested below: The overall inertance is the sumof these. Solution: A channel of continuously varying area can be considered as a cascade of infinitesimally long segments.4 Find an appproximate expression for the inertance of a fluid channel of varying area A(s).) --~ds. The edge betweenthe vertices for generalized momentum p and the effort or generalized force e -. edge betweenthe vertices for e and 0 suggests the relationship that characterizes a resistance. The.11. pd~O~ /( )% q q dt Figure 3. The result of Example3.ib also represents a simple time differentialtion/integration relationsh~ip. each of which has the inertance p ds/A. 3. wheres is the distance along the channel.C and I elements that are defined by relationships betweenthem.2. It is a conception of HenryM.2 is the special case for A = constant. SIMPLE DYNAMICMODELS EXAMPLE 3. picture in Fig. Paynter.106 CHAPTER3. 3.11: Tetrahedron of state . Manynew bond graph modelers like to . The edge of the tetrahedron between the vertices for the generalized displacement q and the flow or generalized velocity ~ represents a simple time differentiation/integration relationship.use this diagramto help them rememberthese variables and the R.

~ I. respectively. is to the generalized velocity (or flow) 0. is to the effort. as shown in Fig. A slender rod of length L andmass Mis welded to the pinion. The inertance relationship is represented in a bond graph by the symbol I.3 It is esssential that you grasp the idea that the definition and value of an inertance depends on the generalized velocity to which it is referred. rotational inertia.12. This problem offers a basic example.p. fluid inertia and electrical inductance are examples of a generalized inertance or generalized kinetic energy. The generalized potential and kinetic energies equal the areas under their respective characteristics.5 Summary A linear one-port inertance or generalized kinetic energy can be characterized by the proportionality between its flow or generalized velocity and its generalized momentum. q. between p and q. The proportionality constant is the modulus of the inertance.12: Guided problem 3. General formulas have been developed for the rotational and fluid cases.2. ~ =. where the generalized momentumis the time integral of the effort. Finally.3 . INERTANCE ENERGY STORAGE 107 ~R. Guided Problem 3. has no special significance. or a general source or sink~ S~or~S. the generalized momentum. (The hidden remaining edge. 7"= ~1102 = ½(1/I)p Mass. and also is called the inertance and is designated by the symbol I.2.3. the edge between the vertices for p and 0 suggests the relationship that characterizes an inertance. II/11///I//1111 Figure 3. 3. ½ (1/C)q=’~~Ce ~ 2and ~-. ~ C. The edge between the vertices for e and q suggests the relationship that characterizes a compliance. Table 3. Find the inertance of the unit relative to the horizontal velocity of its center. as shown. e. A disk-shaped pinion of radius r and mass m is engaged to a fixed straight rack.1 summarizes some of the more primitive physical elements commonlymodeled as compliances and inertances. 5:. which for the linear cases considered are.) 3. Thus. as the generalized displacement.

8 Define the symbols L. (f) gas accumulator.og I=pL/A Steps: 1. E. Write the kinetic energy of the unit due to the horizontal velocity of its center of mass as a funtion of 5. Solve for I. (d) dulum. 3. A. (e) compliance volume. 1 "2 4.108 CHAPTER 3. Find the angular velocity of the unit as a function of 5. (c) torsion rod. Table 3. Find the kinetic energy of the unit due to its angular velocity as a function of this angular velocity.1 as they pertain to the (a) cantilever beam. . (b) floating block. PROBLEMS 3. (g) gravity tank and inertance channel. p. /3 and l~b in Table 3.1 Primitive ~ ~ M Compliances compliance gravity strain C=l/k Physical SIMPLE DYNAMIC MODELS C=L3/3EI and Inertances inertance C= l/Apg M C= 1/k C=Vo/~6 Suggested C=2L/zraG P IQ C 1/mgr C=A/. and set equal to Fix . which should be the desired answer. and then use the result of step 2 to convert this to a function of ~. r. Add the energies of steps 1 and 3. I. G. 2.

3.b) Usethe result of step (a) to find the inertance of the disk relative ¢. as in Problem3.12 Solve the preceding problem substituting a uniform slender rod of mass mand length L for the massless rod and point mass(as in Problem3.13 For the rolling disk of Problem3.9 A disk of mass mand radius r rolls without slipping downan incline with angle ~ relative to the horizontal.4.5 (p.3. Represent the situation with a bond graph. using the volumeflow rate Q as the flow variable but neglecting entrance and exit losses. (a) Find the momentrequired to producean angular acceleration. 4.2. and find the acceleration of the center of the disk. The pressure at its entrance is a constant P = pgh and is zero at the other end. SOLUTION TO GUIDED PROBLEM Guided Problem 3.11 A pendulum comprises a point mass m attached to the lower end of massless rod of length L pivoted at its upper end.3. 99). 3. neglecting the gravity compliance property. using the velocity of the center of the disk as the flow. and thereby deduce the inertance with respect to angular velocity. . Find the rate of changeof the flow.energy of the disk. T~= ~(m+M)2 2. INERTANCE ENERGY STORAGE 109 3. as in GuidedProblem3. Determinethe inertance. %+~=~ ~ =~ ~m+ 1+~ M ~. where ¢ is the angle betweenvertical and a line drawnfrom the center of curvature to the center of the cylinder. 100): (a) Find the kinetic . 3. (b) Checkto makesure that 5I¢ "2 . first in terms of any velocity but finally in terms of ¢. A is filled with and its entrance surroundedby an inviscid liquid of density p. (. 3. 99).3 1 2. with your I frompart (a). ~ =Sir. 1. p. is the correct kinetic energy.3 (p. Suggestion: Use the concept of kinetic energy to find the inertance.10 A tube of length L and internal area. Represent the situation with a bondgraph.

Application to a variety of more complex systems is made in Chapter 4 and subsequent chapters. For the 1-junction type. 3. creating nor dissipating energy.1 Junction Types An arbitrary number of bonds can be joined by a junction. electrical and fluid constraints are then modeled by these junctions. The conventional bond-graph abstractions for junctions are given below first. showing that there are precisely two types. For the 0-junction (spoken zero-junction).3.. requiring el(~l + e2~2 ’: e3~3+ . The following rules are imposed to define two types of junctions: with Rule (i): For both types.23) . Therefore.en~n. Mechanical. q. The first rule implies that the input power equals the output power. and a single junction is used at the heart of models for very simple systems.110 CHAPTER 3. the efforts on all the bonds are equal.. r eo represented as (3. Rule (ii): For the 0-junction type. the temporary symbol J: identified here by e3 The power convention half-arrows in this example are chosen arbitrarily. which is accomplished through the use of simple junctions. Junctions The models presented thus far exhibit no branching. Most models of engineering systems require branching. the junction is ideal. 3. neither storing.3 SIMPLE DYNAMIC MODELS I= ~m+ 1+~ AI. the first two being inward and all others being outward. the flows on all bonds are equal.

~ have been removed from the respective bonds. is recognized as being commonto all the bonds. Analysis and Design of EngineeringSystems.e2 = e3 q. The 1-junction is called the common flow junction. which through equation (a.-. which may not be of interest.M.13. The commonflow or generalized velocity is indicated by the 9 (or a specialized symbol. if one prefers) written below and to the right of the 1.. ¯ e. and vice-versa. and was the most crucial step in his creation of bond graphs. written above and to the left of the junction symbol.~ en ~. with the bond vanishing.3.27) The labels 01. Mass..~ have been removed from the respective bonds. The 0-junction is the common-effort junction..Paynter... It is the dual of the 0-junction. 111 JUNCTIONS the second rule requires el ~ @2=. (3.9. not of necessity (they could be left on) but because the effort e.26) gives el -+.24) which when introduced into equation (3.e. respectively. It is often more useful to recognize its property described by equation (3. ¯ .[ the second rule requires 01 = 0u = 93 .23) gives (3. Two junctions of the same type connected by a bond can be collapsed into a single junction. For the 1-junction. again because they would be redundant.. The MITPress.25). a.. since the constraints imposed on the efforts in one case are the same as the constraints imposed on the flows in the other. represented as e3 q" e. The only difference between the expanded and contracted versions of the graphs shown is the presence or absence of the variable ea or 93. 2H. + en.e3 = . so it is also the flowsumming junction. The concept and notation for these junctions was introduced by H. 1961. Cambridge. as shown in Fig..2a) 0n = 9.. It also is called the effort-summing junction.M.25) The labels el . .3. (3. Paynter 2. (3.

28) F. The other two are the geometric constraints betweenthe displacements or velocities (sometimes called "continuity"). A mechanical examplefor translation is shownin Fig.14. as implementedin the bond graph by the 1-junction. = FA + FB. SIMPLE DYNAMIC MODELS e5 e2 ~4 l~2 q5 Figure 3. The conservation of powerrequires Ft~ -~ FAS: + FB~. All three push rods share the commonvelocity k. the third is redundant. It is extremelyadvantageousfor . Anytwo the three are sufficient. 3.13: Bond-graphjunction equivalences Figure 3." Generalized force balances represent the "conservation of momentum" which is one of three great principles used to analyze combinationsof mechanicalelements. and the conservation of energy. (3.3.112 CHAPTER 3. (3.14: Junction for mechanical translation 3.29) so that Equation (3.2 Mechanical Constraints Modeled by 1-Junctions Common-flow junctions abound in models of engineering and natural systems.29) can be found alternatively from fr ee-body analysis which is traditionally taught in courses entitled "statics" and "dynamics. A push-rod with compression force Ft drives two push-rods with compression forces FA and FB.

6 Model the parallel combination of three dashpots below by a bond graph. 113 JUNCTIONS you to be able to employ any one of the three consequent approaches. the use of the energy balance and the geo~netric constraints tends to be clearer and simpler than either co~nbination using the force balances. can be combinedinto a single resistance characteristic. EXAMPLE 3. When the modeling becomes complex or confusing. F EXAMPLE 3. It also is more generalizable to non-mechanical phenomena. show how the individual characteristics. Also. so only one element of each type remains.3. Elements of different types cannot be merged.3.5 Model the parallel combination of two masses. a spring and a dashpot shown below with a bond graph. as plotted by solid lines. force . The two inertances (masses) sum to give an equivalent inertance (mass). F Solution: The 1-junction of the bond graph model below represents the fact that all four elements share a commonvelocity. Constant inertances arrayed around a 1-junction always can be summedin this manner. Reduce the graph. if possible.

which are inversely proportional to the three compliances. It is quite correct. whereas the force F is the sum of the three individual forces. The combination rule therefore is 1 1 1 1 ¯ . b~ 1.114 CHAPTER 3. kl Solution: The three individual forces sum to give the total force. the three forces sum for each value of 5: to give the overall characteristic.i. that can represent the combination. as shownby the dashed line. These forces are proportional to the three springrate constants. It follows that the equivalent overall resistance R equals the sum Ra + Rb + Re. if the resistances are functions of 2 and the characteristics are curved lines. C. Determine the modulus of a single compliance element.7 Each of the three springs below has a constant spring rate.~--~R force R~ reduces to: Mathematically. as with the dashpot example above. Therefore. This equation is most easily understood if the resistances are constants and the characteristics are straight lines through the origin. SIMPLE DYNAMIC MODELS Solution: All three dashpots experience a commonvelocity (5:). nevertheless. F = Fa + Fb + Fc = Ra~ + RbiC + Rc~ = RS:. EXAMPLE 3.

The "pump"includes its fluid load.. 0 O._M. EXAMPLE 3. The presence of a 1-junction maynot be always obvious.3.1.. as shown. ’200 250 ’ 150 ¢.~.8 Modelthe fire-extinguishing system of Chapter 2 with an additional load driven from the sameshaft. M’I. load B and their combination._~. they have the same angular velocity. show how the individual characteristics of the pumpand the auxiliary load. "~a" need only be noted below and to the right of the "1" symbol.~r BELT Md M..3. also as shown. or it might be a separate pump. ~ auxiliary l°a d~a~ "~ This auxiliary load might be a siren. ~Dtotal ~- p~ = M~. . are p~ = Mp~. that is operated at the same time as the pump. first answerthis: do these three elementsshare one of their respective conju. This fact is recognized by a 1-junction: MOTORM. ’ 100 ’ . respectively.. for example. DRIVE T ~d"~AUXILIARY LOAD For emphasis. Also. ~ pulley drive . all three bonds are annotated with "~a.~. ~_~-_~. 5." This is redundant. PUMP ~. Cd. the 1-junction represents the geometric constraint in which three or more bonds share a common generalized velocity or flow. as the following examplesuggests. because they have a common shaft. The powers delivered to the load A.~_M. Mv~d+ Ma[~ = (Mp + Ma)~a= Md~d. combineto give an equivalent overall resistive load.gate variables? The answer is yes. rad/s Solution: To rePresent the combination of the drive and the two one-port devices. 115 JUNCTIONS In general. as plotted below.

-. The bond graph follows directly.15: Electrical exampleof a 1-junction: series circuit .Ma COMBINED MOTOR ..5 0o 50 100 150 . 3. SIMPLE DYNAMIC MODELS Thus. Z2 and Z3 are connected in series. and common-current implies series interconnection of components. This curve is the effective one-port characteristic of the equivalent combinedload. as shownin Fig.3 Electrical 1-Junctions Circuit Constraints Modeled by Common-flow implies common-currentfor electric circuits. It can be used as before to carry out a source-load analysis or optimization. The total voltage of the source equals the sum of the voltages (or voltage drops) of the three components.5 ~. rad/s Mm BELT -.3.----~-~ ~a ~" LOAD Cm DRIVE 3.~= Mp+ The summationof the last equation above mayimplementedgraphically by "vertical" addition of the two componentcharacteristics.~3 (same). as suggested below.Suppose that three components arbitrarily labelled Z1.the sum Figure 3..15. the total torque provided by the belt drive is M. 2.116 CHAPTER 3.M~=M~+Mo 1. 0.to give the characteristic drawnwith a dashedline.5 Nm 1. 200 250 Ca.

Sometimesthe effect of an element is small enoughto justify its removal from a model. the sununation of the voltages implies summingthe reciprocals of the capacitances. In this case.4 Fluid Circuit 1-Junctions Constraints Modeled by Fluid flow is analogousto electric current.2 + l/Ca. if the two remaining powerarrows are directed unailaterally. ~hydraulic ~ motor I t~l reservoir Solution: The first step is to recognizethat all four elements havethe same flow. Q. A series interconnection of components such as pumps. If the elements are capacitors.valves. their bondsare joined by a 1-junction: VALVE Pm PUMP ~P !Q’~-~ ! MOTOR RESERVOIR . describe how the model can be simplified if the pressure drop across the valve is negligible or the reservoir pressure is virtually atmospheric. If the elements are inductors.9 Model the hydraulic system pictured below. Example 3. such as the masses discussed earlier. the total inductance of the circuit is the sum I1 + I2 + Ia.MOTOR and RESERVOIR.3. This is like the springs bondedto a common 1-junction as discussed earlier. The same is analogous to any other type of inertances bondedto a common 1-junction. 3. 117 JUNCTIONS of the voltage drops around a loop equals zero. it is permissible to removethe junction altogether and connect the two dangling bonds. Also. Electrical resistances in series combinelike mechanicaldashpots in parallel.The order of these bonds is immaterial. hydraulic motors and reservoirs can be represented by a 1-junction with a bond radiating outwardto the appropriate 1-port element for each individual component. the total resistance of the circuit is the sumR1 + R2 -I.R3. This action can leave a 1-junction with only two bonds attached. Thus. Thus if the elements happen to be resistances.3.VALVE. or 1/C1 + 1/C.3. You mayrepresent the four elements in the bond graph by the words PUMP.

it therefore always has the same force at its two ends. If the valve is wide open. is the pressure drop across the valve. EXAMPLE 3. the reservoir pressure P. so if atmospheric pressure is employedas the reference (to give gage pressure). despite the nonlinearities. maybe considered to be zero gage. whichtherefore is the flow on the resistance bond. a dashpot can be represented by a three-ported 0-junction and a resistance. Finally. Thus. 3.:~2. the junction itself can be removedand the two bonds simply connected together.16: Mechanical exampleof a 0-junction. if only twobondsare left on the junction. this case.10 Showhow a cascade of dashpots can be modeled by a bond graph.. and the powerdissipated in the valve is virtually zero.5 Mechanical Constraints Modeled by 0-Junctions A dashpot is the idealization of a damperwith zero mass. two-endeddashpot .118 CHAPTER 3. X1 X~ Figure 3. Pv -~ 0. SIMPLE DYNAMIC MODELS Note that the pressure on the valve bond.3. the bond and the reservoir element maybe removedsimilarly. and howthat graph can be simplified by combiningthe characteristics of the individual dashpots.16. the reservoir bond carries no power. as shownin Fig. The pressure Pmis similarly the pressure drop across the motor: The sumof these two pressure drops plus the reservoir pressure equals the pumppressure. The force is a function of the velocity difference :b~ . P~. This means that the resistances of the components also sum. 3. In this case the valve and its bondsimply can be removed from the bond graph model. assuming the two bonds have powerconvention half-arrows in the same direction.. Similarly.

-----F0 R 1-22~ (c) l-X4 x. in the case of a rotary spring) is proportional to the difference between the displacements at the two ends. by a three-ported 0-jur/ction and a compliance. The deflection. Then. so like a dashpot it has the same force at each end. The same statement applies if the resistances are nonlinear. 119 JUNCTIONS Solution: The bonds of the successive elements can be joined as shown in (a) below. the dashed line representing the overall resistance characteristic is found by a horizontal summationof the individual characteristics: x Mathematically. as shown in Fig. the overall reciprocal ofthe resistance. An idealized spring has zero mass. since each dashpot has the same force. Should one end of the spring be motionless.. since their bonds all have the same effort (force). the three O-junctions can be coalesced as shown in (b).which causes the force (or torque. as long as the individual resistances are expressed as functions of the commoneffort e rather than the non-commonvelocities. It therefore can be represented similarly.3. Therefore. Fo. 3. ql . R~ ¢R2 (b) (a) ¯ ---fo---. 1/R~ called the overall conductance. because of the commonvelocities at the connections.-. Finally.q’_. the bonds of their respective resistances can be joined together with a O-junction. there is no power on the associ- . as in (c).17. I ~ X3 X4 R~ R~ R~ R~ The generalizied velocities of the individual resistances sum~o give the overall generalized velocity. =F + F +F F R Thus.3. equals the sum l/R1 + l/R2 + l/R3 of the individual conductances.

into a single compliance.6 Electric and Fluid by 0-Junctions Circuit Constraints Modeled A common-voltagecurrent-summingjunction is typified by a soldered joint or a parallel combination of elements.. Part (d) of Fig. C3 etc. satisf~ving the sameequation as the springs of Example3. Anexampleis shownin Fig. whichtherefore maybe erased. SIMPLE DYNAMIC MODELS M F’~’~"~x~ F M~ (b) rotary spring (a) translational spring ¯ ~eo ~eo ~:~>-~--~. and a relative displacementis proportional to a compliancefor a given force. It can be removed.1. 3.19. or interconnection of fluid lines at a point of common pressure. Solution: Since the forces are common and the relative displacements sum. as suggested in Fig. and therefore serves no purpose. 3.17: Double-endedsprings ated bond. the compliances sum: C=C~+C2+C3+.. The result is a single bond and a compliance.A0-junction for a fluid circuit is typified by the pipe tee.3. The 0-junction is left with only two bonds. .11 Determine the combination rule for combining a cascade of springs with individual compliancesC1.. identical to the case whichintroduced springs in Section 3. and its two bonds joined (assuming the power conventions of the two bonds are compatible). Capacitances sum similarly. C. Conductances sum in the same way as they do for mechanical 0-junctions. Finally. 17 showssuch a case. ~ C (c) bond graph C C (d) right side i~obil~ed Figure 3. EXAMPLE 3.18. you need to consider howinertances connected to a common0-junction can be combined.120 CHAPTER 3. 3. C2.10. which applies to any type of compliance.

eo ..18: Electrical exampleof a 0-junction: parallel connection P P Figure 3.---~eo---~.3. JUNCTIONS 121 t e Zl .3.19: Fluid examplesof a 0-junction: parallel connection . eo ~ Zl Z~ 23 Figure 3.

Special RCand IR models result from deletion of the I and C elements. Consideration of more complex models is deferred to Chapter 4. causing the effort on the 0-junction to be zero. All of these models include a source. Newmodelers often are uncertain which junction to use in a particular situation.8 Summary 0 and 1-junctions normally are employed as the only means of constructing a branched model. respectively. The "geometric constraints" of mechanical circuits are analogous to the "continuity" or "conservation of mass" constraints of fluid circuits. the overall inertance is zero also. 3.7 Simple IRC Models Mechanical.) Since flows sum. These examples are the major vehicle used in this chapter to introduce the modeling and analysis of dynamic systems.20. bonded to a common Solution: The effort is commonto all the inertances. a resistance. for example. Should any of the inertances be zero. resistances and energy storage elements. The rate of change for the flow of each inertance is inversely proportional to that inertance. compliances and reciprocal inertances (called susceptances) sum. and (as will be illustrated later) transformers and gyrators. and therefore play a major role in bond-graph modeling. SIMPLE DYNAMIC MODELS EXAMPLE 3. 3. junctions. Therefore. although the R. 3.3. which is a required step no matter what type of modeling language is used.122 CHAPTER 3. The junction structure of a model comprises its bonds.12 Determine the combination rule for multiple inertances 0-junction. a compliance and an inertance arrayed around a commonjunction. while the flow (and its rate of change) sums. electrical and fluid examples of simple IRC systems and bond graph models thereof are given in Fig. This dilema is reliably resolved only by careful identification of the key variables in the physical system. conductances. You are urged to cover up the bond graphs in the figure and practice modeling these systems yourself. The complete model merely adds one-port elements to his junction structure: sources. C and I elements could be the result of combining two or more elements of the same type. (The reader aware of the difference between the stagnation and the static pressures should recall that this difference is being neglected for the time being. The bonds on a 0-junction have a commoneffort and flows that sum. the combination rule is 1 1 1 1 = This means that the overall inertance is less than any individual inertance. The bonds on a 1-junction have a commonflow and efforts .3.

3.20: Simple IRC models .3. 123 JUNCTIONS F k~~ l=m R=b C=l/k e I=L R=R C=C I=pL/Ap ~A~_~ C= At /Pg ~" QI (a) O-junction type orous plug ~’Ap R C=l/k R=b C=C e R=R ~~ Q ~ Ap A~ I=pL/A~ C=A./pg R=8~/~ (b) 1-junction type Figure 3.

Suggested Steps: 1. . in terms of the moduli C~. two or more compliances bonded to a common0-junction produce an equivalent overall compliance larger than any of its components. whereas compliance is proportional to admittance. 3. On the other hand. Twoor more compliances bonded to a common1-junction. produce an equivalent compliance smaller than any of its components. The same statements apply to inertances. SIMPLE DYNAMIC MODELS that sum.4 Needed practice dealing with compliances as well as junctions is provided by this problem. electrical or fluid. Three constant compliances are combined two different ways: C1 (i) ~0~C2 C1 (~i) ~1~C2 l C3 (a) illustrate mechanical. Guided Problem 3. Trying instead to remember which type of junction applies to series connections and which applies to parallel connections usually leads to mistakes. The difference results from the fact that resistance and inertance are proportional to impedance.C. Extracting the behavior of these models comes next. electrical and fluid implementations. 2. and (b) find compliance of the equivalent integrated compliance.124 CHAPTER 3. the reciprocal of impedance. C2 and C3. and compare the result with the desired" form. You are expected at this point to be able to find appropriate bond graph models for si~nple systems such as those pictured in Fig. Part (b) should become routine to you. since the answer depends in too complicated a way on whether the system is mechanical.20. conversely. ~. Multiple resistances bonded to a common0-junction combine to give an overall resistance smaller than any of its components. Multiple resistances bonded to a common1-junction can be combined to give an overall resistance greater than any of its components. For part (a) focus on the type of variable (e or q) which sums and the type of variable which is commonto all three compliances. For part (b) sumthe variables of the appropriate type to get a total value.

Guided Problem 3. The mass moment of inertia of the drum is known to be 0.12 in lb s ’~. and label this on a 1-junction.22. 3. basic ~nodeling problems which you should be able to do at this point. 3. 2. Estimate the overall inertance relative to the total volume flow. A drum viscometer comprises a cylinder with a rotating drum submerged in the fluid being tested. The rate at which the drum coasts to a stop is observed. Model the system with a bond graph.3.4 x 10-7 ~. .21: Guided Problem 3. Suggested Steps: 1. 125 JUNCTIONS Figure 3.21. Find the inertances of the four separate channels relative to their individual flows. as pictured in Fig. m~dthe viscous drag on it produces the moment M.02 in and # = 1. The inertance resulting from step 2 shares a com~nonflow with the fourth channel.27rLr3# (b W where L = 6 in. Identify the key flow variable. Water pumpedfrom a river to cool the steam in a power plant passes through the circular channels shown in Fig. lb s/in Suggested Steps: 1. r -. Combine the three inertances which share a commonpressure.3.6 This is the first of three short. and evaluate the moduli of its elements. 3. Combinethe inertances to get the overall inertance.5 This problem requires the consolidation of distinct inertances into a single inertance.5 Guided Problem 3. w = 0.3 in.

6 d ~d ¯ density p Figure 3. ter following a non-equilibrium initial condition the system with a bond graph. and evaluate the preparation for an analysis. State or find the moduli of the elements of part 2.126 CHAPTER 3. Guided Problem 3. using the given information. Represent these as bond graph elements appended to the 1-junction. The motion of the wais of ultimate interest.23: U-tube for Guided Problem 3.7 2.22: Drum viscometer for Guided Problem 3. Model parameters of your model in viscosity.23. 3. as shown in Fig.7 A U-tube is filled with water. Neglect the effects of 3. SIMPLE DYNAIVIIC MODELS Figure 3. Identify the important physical mechanisms that perpetuate and those that dissipate energy. .

Guided Problem 3. Determine the effort or flow variable that is commonto the four elements. Small perturbations in the current source are inevitable. and the effort or flow variables that sum.8 A current source drives an essentially inductive load.)l IR~ -.24: Guided Problem 3.3. Express the moduli of the I and C elements in terms of the parameters given in the drawing.8 Suggested Steps: Compliance and inertance elements are present. 2.24. to moderate these pulsations. If so. Draw and label a junction accordingly. You are asked to model the system. 3. Ili/¢. You may have more than one energy storage element of the same type in your model. Bond the four elements to this junction. Suggested Steps: 1.-C source filter load Figure 3. and produce undesirable variations in the voltage. determine its modulus. in preparation for an analysis directed at determining appropriate values for the resistance and capacitance.11 ’. as shownin Fig. combine these elements into a single element of the same type. and draw the reduced bond graph. A shunt resistor and capacitor are added to the circuit. Identify the variable which is commonto them.3. _1_. . Bond the C and I elements to the junction. and deduce which kind of junction is proper. 127 JUNCTIONS r r -7 ’.

16 Drawa bond graph to represent the electric circuit below. as shownbelow.) . (Youmayuse references in your personal library.17 Anapplied force F is resisted by the bending of a slender cantilevered beamof length L. Modeleach of the components by an appropriate bond graph element and assemble a corresponding complete bond graph model.generator electrical load ~ mechanical load I I. Find the modulusof R’ assumingconstant moduli as given.14 An engine drives both a virtually ideal DCgenerator with constant magnetic field and a Inechanical load on the same shaft.128 CHAPTER 3. 3. ~ I engine ~.15 The bond graphs below can be reduced to a single resistance. R. and find its modulus.--0 R2 RI R4 R2 R4 ~R bothcases: R~ = 4 R2=3 R3= 2 R4=l 3. SIMPLE D}%TAMIC MODELS PROBLEMS 3. 3. thickness t and width w as well as a spring with rate k. Reducethe load seen by the voltage source to a single resistaace. (a) R 1 R 3 R3 ~0-~--~1 --1---~. F (a) Find the moduli of the two componentcompliances in terms of fixed parameters.

18 Twolinear springs with rates kl and k. C1 and C_~. and the right end. Investigate its effect on the behavior as suggested by the steps below. where 2 is the velocity of the large rigid mass. connected in (a) series 3. (b) Interpret the kinetic energy of the spring as an inertance relative the velocity ~. 3. by integration. (Ask yourself: which of these variables is commonto the two compliances. as shown below. 3. (a) Assumethat the mass of the spring. Define a position variable y such that the left end of the spring represents y = 0. the velocity of the segment is (y/L)2.~ predicted by your model.3. which can be assumed to be divided equally between its two ends. Similarly. of an ideal mass-spring system is shown in Section 3. is not great enough to significantly upset the uniformity of the strain from one end of the spring to the other under dynamic conditions. Find the compliance of the combinations. Find the w. 129 JUNCTIONS (b) Represent with a bond graph how the two compliances combine. labeling F and 2 appropriately. Each infinitesimal segment dy therefore has mass m~(dy/L).21 A spring with stiffness k connects a mass Mto ground.3. express the total inertance relative to ~. is relatively small.19 Find the capacitance of two capacitors.~ are connected in (a) series (b) parallel. (c) The natural frequency w~. The mass of the spring. which has the same compliance as if the spring had zero mass. but not insignificant. and which is the result of a summation?) (c) Find the combined compliance. ms. the kinetic energy of the entire spring as a function of ms and ~. may entail.qualitatively. (d) Comment.5 to equal V/~/IC. ms. y = L. which you may define as I. Then. on any limitations you suspect this analysis . Evaluate. Give a bond graph model for this damper~ and relate the moduli of its elements to physical parameters. 3.20 A physical damper has significant mass. (b) parallel. as a function of M and ms.

24 Define variables. 3. (b) Represent how the compliances combine with a bond graph.23 Define variables. slow) (ii) adiabatic (i. graphs.22 A vertical tube of diameter d and length L is capped with a closed chamber of volume ~/b. Estimate the effective compliance of the system with respect to the entering water.130 CHAPTER 3. and air is trapped above. as shown below. and that the compression of the air is (i) isothermal (i. labeling P and ~ appropriately. Water enters from the bottom. water (a) Evaluate separately the compliances due to gravity and the compression of the air.e. (Ask yourself: which of these variables is common to the two compliances. and model the mechanical circuits below with bond . fast). SIMPLE DYNAMIC MODELS 3. and which is the result of a summation?) (c) Find the overall compliance. 3. (a)~ and model th~ electric circuits below with bond graphs.e. Assume that the water never enters the upper chamber. the following steps are suggested.

The fluid can flow unimpeded through an auxiliary channel. Model this system in preparation for an analysis which will be used to choose the springs and the dashpot. Neglect fluid inertia. combine these elements and give the modulus of the new element in terms of k. and evaluate its parameters.0 m2 drains through a porous plug in its bottom with a fluid resistance of 98.0 cm." . Model the system in preparation for an analysis to determine its motion. A standard scheme. Model the system with a bond graph.4 lb/ft ~-’~ #=2x10~ lb’s/in2 ~w~.28a lb/in 3) is released from rest in a fluid-filled cylinder. Specify the moduli of your bond graph elements. If your model has separate elements to represent the separate springs. lporous ~ I plug tube ~ 3. ’ tank area: 10 in2 3pg=62. valid for y-motion only. neglecting the inertance of the fluid.fluid ~I 3. and assume the plunger remains centered. etc.3. It also drains through a tube of length 10 m and diameter 2. 3. and thereby prevent large oscillations from shaking the floor. 131 JUNCTIONS 3. .25 Model the tank and tube shown below left.28 Often it is impractical to reduce the vibrational forces of a machine to an acceptable level.7 kN s/m5. as shown above right. mass m k~ ~b k ~ support (conceptual) ~ floor .27 A water tank of area 2.but it is practical to isolate the machine from its foundation. Specify the values of the moduli of your bond graph elements. Neglect the effect of viscosity. creating noise.26 A steel plunger (Ps9 = 0.3. is shown below.

83 lb s~/ft 3.. = 6.5 1.69 lb s2/ft . Combinationof three small channels: I = ~5.S(1/C~)q~ I/C = 1/C~+ 1/C~. 5.94 x 100 5.49 lb s2/ft 7r(3)2 2.83 = 8. + 1/C3 Guided Problem 3.86 + 1.4 Problem (a) (i) common effort R PROBLEMS mechanical: electrical: _1. SIMPLE DYNAMIC MODELS 3. .49 = 1.29 Presume that the LRC filter shown below drives an instrument with a virtually infinite input impedance. A 7rd2 = ~ Each small channel: ls = 4 × 1. The relation betweenthe voltages ei and eo is of interest.~ C.132 CHAPTER 3.94 × 20 5. Large channel: IL = pL _ 4pL 4 x 1.e ~ C = CI 3+ C~_ + C (ii) e = ~’ei = . since the latter is intended to be a smoothed but not smothered version of the former. 5.5. (i) q ’~ qi = .86lb s’~/ft 7r(6). Model the system with a bond graph in preparation for au analysis. Combinationof all four channels: 1 = 6. _l_ _[_ fluid: (ii) common displacement mechanical: electrical: fluid: (b) Themodulineednot be constantsif chordaldefinitionsare used. L SOLUTIONS TO GUIDED Guided 3.

~. since equations (3. The key flow variable is the angular velocity of the drum. or R = 27rLr31~/w= (2zr×6×33×1.12).3.12) (p. I = pL/A = 4pL/Trd The two compliances share a common flow or displacement.4×10-7/0. all four associated bond graph elements share a common0-junction. Note there are two compliances. each compliance is C A/p9 = ~rd2/4p9. The commonvoltage at the bottom ends of the elements is taken as ground.generalized velocity either the commonvertical velocity of the surfaces of the water or the volumeflow of the water. the two stiffnesses sum. or zero. and the kinetic or inertance energy is a function of the velocity of the water.3 (p.8 1. Energydissipation takes place because of the viscous drag in the narrowannulus. so their efforts or pressures sum. The latter choice is madehere.6 1.2. R. the inertance is ~. so this gives a resistance. The associated generalized displacements are the common vertical displacements of the surfaces of the water and the volumedisplacement of the water.12 in lb s The resistance is the ratio of the torque Mdue to viscous drag to the angular velocity ~. This means that the reciprocal of the combined compliance equals the sum of the reciprocals of the individual compliances. 133 JUNCTIONS Guided Problem 3. Theshear force is a function of angular velocity. 105) gives the inertance with reference to these variables.s. 95) the compliances and Example3. Note that half the compliance meanstwice the stiffness..z c . There is kinetic energy ½I~2 associated with this velocity. M-~-~’~R The mass moment of inertia is the integral 1 = fr2dm 2. From equation (3. The bond graph that represents these observations is I~ 1. Guided Problem 3.3. Guided Problem 3. Therefore. All four elements share the same voltage difference. = 0.02) = 0. with symbolsQ = g" and V. where I is the mass momentof inertia of the drum.3. as shownin the bond graph.00713 in-lb. R s<--77--o o ----.one for each free surface. so that the cmnbined compliance is C’ = C/2 = 7rd2/8pg. respectively. One could choose as the.7 The potential or compliance energy is a function of the displacement of the water. Fr om Example 3.2.

4. however. and the following section deals with their solutions.com. (Netherlands) has a programcalled 20-sim. informa’~ioncan be found http://www. This rapidly evolving software is not discussed further in this book.or analytically. and solving the equations numerically -. which are directly plotted. and write one equation for each one-port element.~ forces the flow ~ into the system. Causality SIMPLE and Differential DYNAMIC MODELS Equations The dynamic behavior of a well-posed model with energy storage elements is found by expressing it as a differential equation or set of equations. The procedures are largely deterministic. to give a solvable set of differential equations. in the case of linear models.134 3. unless an effective over-arching plan is applied.be. This bilateral causality can be indicated with either of the notations Se e--~ ~ ~-q or Se e ~ q Similarly. analytical solution. It is possible to define all the efforts and flows in a bond graph with symbols.informationcan be found at http://www. 1974. unlike in modeling. They can be expedited by software.1 Causalities of Effort and Flow Sources The effort source Se imparts the effort e as an independent variable onto the system to which it is attached. two equations for each transformer and gyrator. and one equation for the variables which sum about each junction. which responds with the effort. N. whereas e is caused by 3See R. the flow source S.nl/20sim/clp.htm.lorsim. A middle course is chosen in this text between blind reliance on software and total reliance on mind and hand.el. in favor of the development of a more general and basic understanding of the analytic procedures and the use of more general-purpose and widely used software. Such a comprehensive plan best develops one first-order differential equation for each independent energy storage element. Controllab Products B. so that you can benefit from modern methods and understand their origin.4 CHAPTER 3.called simulation -.Y. Rosenberg. . Such programs are indeed available for bond graph models3. Users of such software (sometimes coupled with separate numerical simulators) do not have to write the differential equations themselves or even find their solutions. This usually large number of equations then can be combined.V. information is available at http://www. A Users Guide to ENPORT-4.CadsimEngineering(California) has a product called CAMP-G. however. This section concerns the determination of the differential equations. The associated flow ~ becomes dependent. Wiley.bondgraph. 3. whereas (~ is caused by the reaction to this by whatever system (unshown below) is attached to the right-hand end of the bond.rt. This state-space formulation is ideally suited for both numerical simulation techniques and. You can say that Sf is the cause of 9.utwente. e. requiring care but no special insight. This procedure is excessively complex in practice. Lorenz Simulation SA(Belgium)has a package called MS1. You can say that Se is the cause of e. Also. The plan also ought to be so automatic that it can be programmed into software. eliminating the variables of minor interest.

The flow ~. This causal pattern is recognized by placing causal strokes at the outer ends of the bonds for these three flows.4. respectively. These three elements are said to be given admittance causality.).25. to a C. on the other hand. ~3 and ~4. Consider now the specific situation shown in part (b) of the figure. it is the sum (~ = eo/R + C deo/dt +po/I. This means that their flows respond to an induced effort. assuming constant moduli R. C and I. in which the three bonds are attached. . as shown in part (a) of Fig. and are expressed as functions of that effort or its time integral. The behavior of the three ele~nents is said to be uncoupled. must be separately caused by whatever (unshown) elements are attached to the outer ends of these bonds. 3. and the flow on the I bond as f eo dt/I =po/I. ~3 and 04. which means that the flow on one of them is independent of the size or even the presence of the others. This causality can be indicated as S! or The short causal stroke drawn transverse nationally: The causal orientation entation. with signs determined by whatever power convention half-arrows are part of the model. The flow on the source bond completes the picture. the fiow on the C bond as C deo/dt.4. All bonds on the 0-junction have the same effort. an I and an R element.3. p0. The flow on the R bond then can be written as eo/R. CAUSALITY ’(a) AND DIFFERENTIAL 135 EQUATIONS e01~ (b) l. is the sum of 02. The flows ~. is specified or caused by the effort source. q2 eo~O4 q~ C-~ eo~eo /R po = f eo dt R Figure 3.2 Junctions ior Sy I~ to the bond is recognized inter- is completely independent of the power convention ori- with Elements Having Uncoupled Behav- Consider an effort source bonded to a 0-junction. e. by the definition of the junction. This effort. then.25: Effort and flow sources bonded to a 0-junction behavior to give uncoupled the reaction to this (~ by whatever system (unshown below) is attached to right-hand end of the bond. 3.

Are the behaviors of these elements coupled or uncoupled? Solution: A flow source is bonded to a 1-junction. all the bonds on any 0-junction must have their causal strokes at the remote end.the flow 00 imposedby the source. I and R elements with the respective efforts expressed as ROo. Then explain the annotations for the efforts and flows given on the bondgraph belowright. as in the O-junction example. This fact is underscored by the placements of the causal strokes on the bonds. I d(lo/dt and R(?o + qo/C + I d~o/dt.13 By analogy to the exampleabove. (1/C) f ~odt = qo/C.136 CHAPTER3. TheI element is said to have integral causality.4:3 Junctions with Elements Having Coupled Behavior The system represented in Fig. C and I elements must have its causal stroke placed adjacent to the junction. since there is only one effort. Three of the bonds are attached to C. and the I elementis said to have differential causality. since their efforts are functions of the flow or its integral or its derivative. Youwill find that these facts makecausal strokes very useful. Similarly. the other two causal strokes must be placed at the outer ends of their bonds. because each effort is independentof the presence of the other two elements. But which of the three possible patterns do you use? .26 differs from that in Fig.25 in that the source does not directly determine either the efforts or the flows on any bonds other than the source bonditself. except for precisely one whichhas its causal stroke at the remote end. Oneof the bondsfor the R. The C elementalso is said to haveintegral causality. All the bonds and the elements are forced to have. except for precisely one which has its causal stroke adjacent to the junction. explain whythe causal strokes in the bond graph below left are drawnwhere they are. EXAMPLE 3. all the bonds on any 1-junction must have their causal strokes at their junction ends. SIMPLE DYNAMICMODELS The C element also is said to have differential causality. Theseelements are said to have impedancecausality. In general. including giving appropriate namesto the causalities of the R. since its causal output is a function of the time integral of its causal input. 3. They are said to be uncoupled. 3. 3. since its causal output is a function of the time derivative of its causal input. in common. C and I elements.

i qi Ii Oi =_C’~-~-.26. Ci (3.3.i q~ ci ei = ei(q~) q~. given a choice. This forces impedance causality on the R element. CAUSALITY C~ 0 AND DIFFERENTIAL "---~’I EQUATIONS C’~---’~ 137 0 ~I R (b) integral causalites added R (a) model " R R (d) annotation completed (c) annotation of causal bonds Figure 3. you use integral causality. (3. The reason should become clear later.31b) Ii For the present model.4.26: Flow source bonded to a 0-junction to give coupled behavior The short answer to this question is that. 3.31a) 1 [ ei dr. but for now simply recognize that differential causalities are defined only for C and I elements (the energy storage elements).30a) ei = Ii . and under the following causal conditions: compliance : ---~k-~Ci O~ inertance: Integral causalities conditions: compliance: ~. integral causality is applied to both the C and I elements in part (b) of Fig. (3. depending in general on the directions of the power convention inertance : ei . dei (3.30b) apply only to these same elements under the inverse causal ~.1 J C~ gt~ dr. The next thing you do is annotate the effort and flow sides of the C and I bonds as follows.----~I~ qi 0i = O(Pi) Pi I~ .

The R bondis forced to have impedancecausality. SIMPLE DYNAMICMODELS Ci ~Ci half-arrows: ~ or Thecircles drawnaround(li and/5i are a logically optional device that will prove to be helpful later. since all the flows are the sameand its specification should comeonly from one place. The next step is to annotate the efforts and flows for the remaining bonds. which substitues an effort source for the flow source in Fig. 3. are madein part (c) of Fig. combined with its effort. Is this causality compatible?Then. This is strictly a causally output variable. The causal stroke on the C bond. EXAMPLE 3. so as to achieve integral causality on both energy-storageelements. annotate all the efforts and flows except for the flow on the source bond. The only remaining unannotated effort or flow is the effort on the source bond. s~ C’~---~.138 CHAPTER3. The R-element responds. . 3. which maybe labeled with a large X to signify disinterest. These annotations. Since this reasoning allows a choice. Note also howthe sign changes compensatefor the reversal in the powerconvention half-arrow. as shownin part (d) of the figure. and then integral causality for the I bond. with its admittancecausality.1 R Solution: All but one of the four bondscan and must have its causal stroke placed adjacent to the 1-junction. signifying disinterest. and substitues a 1-junction for the 0junction. Here. you can choose integral causality for the C bond.26.14 Apply causal strokes to the bond graph below. Note that ibi is a wayof writing ei.26. and need not be labeled unless it is of interest. dictates that the effort on the R bondbe q/C. by giving the flow q/CR. it is simply labeled with a large cross.

26.4. which because there is only one effort on the junction equals the causal input. 3. Thus. The final result is as follows: se Rp/l~p/l R 3. the flow ~ = dq/dt is a causal output of the junction. CAUSALITY AND DIFFERENTIAL EQUATIONS 139 Se The causal strokes are shown above left. and this flow comes.3.32a) dt- (3. and equals the sumof the causal inputs S0 (with plus sign due to its power convention half-arrow). you observe that the causal input to the R element is its flow. the writing of the associated differential equations is a simple matter. Finally. q/C. The effort on the R element is the causal output of the element. (3. causally. and is written as R times the flow. Next. The left side of each respective equation comprises the term which is circled in the annotation of its bond. you place the designations for the effort and flo~v on the bonds that have integral causality. The nmnber of such elements or equations is called the order of the model.325) dp 1 ~q" . as shown above right.4. The right side comprises what this term is equal to. as dictated by the bond graph with its annotations. There is one differential equation for each C or I element with integral causality. to complete the annotation of all the efforts and flows in the graph. The rate of change of momentum [9 = dp/dt is another causal output of the junction.~P . and q/CR (with minus sign).~-~q.4 Writing Differential Equations Once the bonds in a bond graph have all their efforts and flows annotated according to the the rules given above. from the I element. p/I (with minus sign due to its power convention half-arrow). the equations become 0 -junction case : dq 1 1 d--~ = 40 . and equals p/I. For the 0-junction case of Fig.

using the method described and illustrated above. q/C (with minus sign) and Rp/I (with minus sign). they can be considered to be degenerate. The only other variables are the independent or causal input variables 00 and e0. (3.14.~p. 3. First-order models without coupling between the C . since they are the only dependent variables in the equations. The equations therefore are 1 -junction case: dp 1 R d--~. which could either be constants or specified functions of time. 3. and the pairs of equations are solvable.25 and Example3. and equals the suru of the causal inputs e0 (with plus sign). The fact of uncoupled behavior vastly simplifies these cases. The variables p and q are called the state variables. Should you be interested in some variable other than one of the state variables.15 W¥ite the state differential equations for the bond graph of Example 3.33) dp dt (3. The only other letters in the equations represent parameters. you use 0 = p/I. or constants: R. The annotated bond graph provides adequate information.= 00. Most problems of interest do not involve such uncoupled behavior. which because there is only one flow on the junction equals the causal input. the number of differential equations reduces to one. C and I. Each case has therefore two equations and two unknowns. q/C.14-3.26 or Example 3.15 be zero.140 CHAPTER 3. For the 0-junction case you get dq d-~. Their solutions are considered in the next section. if you are interested in the flow on the 1-junction. and the model is said to be of first-order. or missing.13? The answer is yes. should you want the effort on the zero junction.= e0 . you first solve for the state variables. you get one for the single instance of integral causality in each case. For example.34) and for the 1-junction case you get eo. The rate of change of flow 0 = dq/dt is another causal output of the junction. SIMPLE DYNAMIC MODELS EXAMPLE 3.]P" The differential equations above represent the mathematical ~nodels of their systems. Do you get differential equations for the uncoupled cases of Fig. you use e = q/C. Should one of the C or I elements in Fig. Solution: The effort D = dp/dt is a causal output of the junction.~q . dq 1 dt . and then express the variable of interest in terms of them.

Their signs are dictated by the power convention arrows. EXAMPLE 3. . Modelsoften are called by the elements they contain: RCmodels.3. The result is @ R d-7= eo.27: First-order models or I element and the R element are presented in Fig. The methodologyis the same. At that point (Chapter 5) the logical underpinning of the procedures introduced here should becomemore apparent. This description is not unique.16 Find the single differential equation for the IR modelgiven in Fig. and terms on the right side are dictated by the efforts annotated on the other bonds. Solution: The only variable that is encircled is .~--~q. the single differential equation dq 1 (3. or which junction type joins them. which according to the causal stokes are causal inputs to the junction. This is placed on the left side of the equation. 3. 3. and so are the results.27.35) d-~ = 00 .unless you specify whetheror not their elements are coupled. IR models and ICR models.5. gyrators and multiple junctions.~-p. The RCmodelrepresented in Fig. including the use of transformers. 3. CAUSALITY AND DIFFERENTIAL EQUATIONS after step I: eolX after step II: 141 ealX p/I (a) IR model after step I: XTOo after step II: X-~o (b) RCmodel Figure 3.27. except for the missing terms.27 gives. on inspection.4. whichrepresents the only variable that is encircled. however. Theconceptof causality will proveto be particularly crucial in the writing of differential equations whenmore complexmodelsare considered.

for each energy storage element with integral causality. Each causal output becomesa function of its state variable. specifically qi/Ci or pi/~i.142 3. and the flow in the other.expedites the writing of the state differential equations for a system. 3. The present applications involve modelswith a single junction. Youshould never annotate an effort or flow in violation of the causal strokes. and then (since it is allowed)apply integral causality to the energy-storageelement. each effort and flow thus can be indicated as a function of the state variables and/or the input variables. Its treatment for uncoupledmodels has been addressed. such an action frustrates the purpose of the causal method. that is the generalized velocity dqi/dt for each C element and the generalized force dpi/dt for each I element. and is called the order of the model. (Exceptions will be noted in Chapter 5.4. even if what you wouldwrite is functionally correct. Its treatment for other cases is deferred.28. Differential causality sometimesis forced. This refers only to computation. as indicated by a transverse markcalled ~ causal stroke. The numberof first-order differential equations that model an entire system therefore equals the numberof energy storage elements with integral causality. as shown in Fig. SIMPLE DYNAMICMODELS Summary Eachbond in a bond graph maybe assigned a bilateral causality. that is the determination of independent and dependent variables. Suggested Steps: 1. whichis a very special case. The use and meaning of the causal method is expanded upon in Chapter 5. to the causal input of its bond. The causal strokes on the other bonds in the system usually allows these state functions and any input variables to be propagated to the other bonds in the model.5 CHAPTER3. The signs are determinedby the powerconvention halfarrows. The causal inputs to these elementsequal the time derivatives of the state variables. and mayor may not be related to physical causes and consequences. .the C’s and I’s -.) It is usually desirable to annotate all the bonds before attempting to write the differential equations.9 Solving this problemand the next can be considered simple and essential first steps in your writing of differential equations from bond graph models. and write a correspondingfirst-order differential equation(s). Guided Problem 3. The effort on the bondis causal in one direction. The use of integral causality for the energy storage elements -. A simple RCmodel is excited by an indepedent effort source. Apply the mandatory causal stoke to the effort source bond. Define the state variables. Onedifferential equation of first order results from equating the derivative dpi/dt or dq~/dt.

3. PROBLEMS 3. Annotate the effort and flow on the remaining bond according to the mandates of tile causal strokes. 133). and flows on the source bond and the bond with 3. Identify the order of the model.31 Determine whether or not the behaviors of the R and C elements shown above right are coupled. Complete the annotation of any other bonds. 4. 4.10 Write differential equation(s) (pp. using integral causality if possible. Annotate the efforts integral causality. . if necessary.constant.30 Determine whether or not the behaviors of the R and I elements shown below left are coupled. Apply the causal strokes. 126-127.4. Write the differential equations by equating the circled terms to whatever the bond graph dictates.3.28: Guided Problem 3.7 Steps: 1. Guided Problem 3. 3. Write the differential equation by equating the circled term to the sum of terms dictated by the causal strokes. Determine their behaviors for e0 = constant.9 2. Annotate the energy storage bonds in the standard way. CAUSALITY AND DIFFERENTIAL EQUATIONS 143 Figure 3. Determine their behaviors for ~0 -. 2. Suggested for the U-tube modeled in Guided Problem 3.

3. 130).39 Write appropriate differential Problem 3. .26 3. The behaviors or the two energy storage elements are uncoupled.35 Write appropriate differential Problem 3. and describe mathematically the relations between the behaviors of the sources and the I and C elements.36 Write appropriate differential Problem 3.24 (p. 131). equation(s) for the plunger of Problem 3.25 (p. equation(s) 3. SIMPLE DYNAMIC MODELS 3.32 I and C elements are bonded to a commonjunction which is excited by an effort or flow source.34 Write appropriate differential lem 3.1 ~-----~C GUIDED equation(s) for the fluid tank and pipe for the vibration isolator equation(s) for the filter circuit of Problem PROBLEMS The model has one energy storage (the compliance). equation(s) for the mechanical circuits 3.27 (p. 132).9 Problem R~-------. the order of the modelis one. 131). 131). 131). SOLUTIONS TO Guided 3.40 Write appropriate differential 3.144 CHAPTER 3. 3. and therefore must be given this causality. ASa result.38 Write appropriate differential 3. Identify the possible combinations of junction and source types.33 Define state variable(s) and write the corresponding differential for the model represented by the bond graph below. equations equation(s) for the electric circuits of Prob- 3. equation(s) for the fluid system of Problem 3. this can be given integral causality.28 (p.29 (p. 130).23 (p.37 Write appropriate differential (p. 3.

-h-dq Guided Problem 3.13 considered in Problem 3. ] The constant r has the dimensions of time. 3.27. 1 p-~//I 3. C ’~ 1 --~-~I 2.1 The First-Order Differential Equation The first-order version of the general linear differential equation with constant coefficients.3.5. In this section. such as equations derived above for the RCand IR models given in Fig. Once found. dV 1 dPdt . the time is either (3. they can be used as components in the responses to more complex excitations. SOLUTIONS OF LINEAR dq 1 DIFFERENTIAL 145 EQUATIONS 1 4.5 (Makesure you have the minus sign. Even much higher-order differential equations are decomposed into first and second-order components. with one-half the compliance of the individual compliances.31.10 1. C ’-~y~ This bond graph assumes the two original compliances have been combined into one. and is known as the time of the model. 3. solutions to first and second-order differential equations are found in the absence of excitations and for stepwise excitations. and for many RC models such as that equation (3. For many IR models such as that given by Example 3.) Solutions of Linear Differential Equations The forms of the linear first and second-order differential equations found above are repeated over and over in the modeling of systems with constant parameters. This step is not necessary. can be written in the form ~-~-[ + x = u(t). -~ =-~eo.9.36) constant and that given by constant .~V1 3.5. natural frequencies and damping ratios.35) and that addressed in Guided Problem 3. These solutions are characterized by certain values called time constants.

30.has the dimensions of time can be seen by comparing the dimensions of the two terms on the left side of equation (3. (3.718. This is the only parameter that characterizes the models in the autonomous case of u(t) = Whenthe right side of equation (3. equals the homogeneoussolution plus any particular solution. 3.38) This decaying exponential is plotted in Fig.29: Bond graphs for basic RC and IR models T = RC or T = I/R. SIMPLE DYNAMIC MODELS Figure 3. During the same time interval the value of the function is reduced by the factor e = 2.36) above.37) The four classic cases cited are reproduced in Fig.36) is set equal to zero. These facts expedite the sketching of the curve. called a forced response. describing autonomous behavior. as well as underscoring its nature. 3. The solution to a non-homogeneous linear differential equation with constant coefficients. The fact that ~. As suggested in the plot. . Only the details of the excitation term on the right side of the equation vary. the tangent to the curve at any time t intersects the axis Xh = 0 at the time t + T. and its solution is known as the homogeneous solution.146 CHAPTER 3.29. xh(t): [ xh(t) = xoe-t/L ] (3. the equation is known as the homogeneous equation.

3.31.0 t/r 3. In the special case. Solution: The simplest particular solution is x = X f.6 0. that is when dx ~-~.2 0 0 0. and the corresponding value of x be x(T).0 Figure 3.5 2. . Substitution of this value into the solution above gives Xo = Ix(T)z~]e~/~. The only difference is that the final state is not zero.0 1.5 1. The completesolution is therefore x(t) = xoe-t/’~ + xy. because with this solution the derivative term vanishes.17 Find the solution to the differential equation whenthe right side is a constant.with T = 0 and x(0) = 0 this is often called the "step response" of the first-order model. The undetermined coefficient x0 depends on the initial condition.or x = Ix(T) z~]e-(’-v)/~ + x~t >T.0 2.4 0. let the initial time be called T.+ x = xf = constant.3. 147 SOLUTIONSOF LINEAR DIFFERENTIAL EQUATIONS |o0 x ~ 0"8~x p(-t/r) r = RC or r = I/R 0.5 4.5. whichis plotted in Fig.5 3. a final state xf is approachedexponentially from an initial state. Note that the behavior is essentially the sameas the homogeneous response.30: Homogeneous solution of first-order models EXAMPLE 3. To be general.

139) and the answer to Example 3.31: Responseof the first-order modelto a step excitation 3. (It does not extend to nonlinear. The two equations can be combinedinto a single second-order .) Since the excitation in the case of present interest has been decomposedinto a sum of steps. differential equations. justified or not. Theindividual responsesare given by dashedlines in part (b) of the figure.32. In other cases. Each of these is of the form of the solution to Example3. Thesimplicity and predictability of the behavior of linear modelsis the principal justification engineers havefor designingsystems that behavelinearly. The property of superposition is the basis of muchof the analysis of linear models. The time constant in this exampleis rather long comparedto the intervals between the steps. Their sumis given by the solid line. the individual steps could be madesmaller and closer together.. This is knownas the property of superposition.15 (p.148 CHAPTER 3. If the approximationstill is unacceptable.17 with Xi(T) = since they are independentof each other. 3.. X0 . the simplicity of linear analysis also tempts the assumptionof linearity.as you will see later on. The response to the sumof steps therefore is not muchdifferent from the response to the actual continuousexcitation. SIMPLE DYNAMIC MODELS -~ x(r)( I I I I O0 T T+ r t Figure 3. Examplesare given by equations (3. the response to this sumof steps equals the sumof the responses to the individual steps. The response consequently is considerably smoothed. 140).3 The Second-Order Differential Equations Second-ordermodelscomprise a set of two coupled first-order differential equations.5. 3.5.2 Responses perposition to More Complex Excitations Using Su- A continuous excitation u(t) maybe approximated as a sum of two or more steps.32) (p. Anillustration is given in part (a) of Fig.. The solution of a linear-differential equationfor an excitation whichequals the sumo] two or morecomponentexcitations equals the sumof the responses to the individual excitations...

q" + 1 dq 1 (3. SOLUTIONS OF LINEAR DIFFERENTIAL EQUATIONS 149 Au(t) ~ I ~ AU2 ~ /lu3 ~ (negau "v e) t (a) excitation and step approximation note: x(to)=Oassumed sum of step responses=x(t) toT~ T2 T3 T4 T~ t T6 (b) step responses and sum~ereof Figure 3. This second-orderequation is of the form d’~x dx a2-~ + al -~ + aox = bl ~t ) + bou(t).5.32). = x/-~o/a2. (3. -~ + x -ao dt bo u(t) + --ao -. dampingratio : ( = al/2 ax/-d~-~.40a) (3.f(t).42) . (3.39) Dividing this equation by a0 and makingthe following definitions natural frequency : w.26 and Example3.3. Taking the derivative of the first equation and substituting the first gives a second-order equation with the dependentvariable q: dt---T~ ~ ~-~ + ~--~q = -~-. (3. with one of the dependentvariables sacrificed in favor of the other. dqo ar~.32: Approximationof a response of a first-order modelusing steps and superposition differential equation.14 (p 137-139) are perhaps the most commonbond-graph models of second-order systems.41) The cases given in Fig. 3.40b) gives the conventional form 1 d2x 2( dx bl du(t) w-~n ~ + -~. The first-order equations for the 0-junction case are given by equations (3.

42).2 Note also that the right sides of the two differential are different. hand. both the first variable p: SIMPLE DYNAMIC MODELS taking the derivative of the second equation and substituting and the second gives a second-order equation with the dependent d2p 1 dp 1 dt ~ + ~-~ + -i-~P 1 = -~qo.43) and the two equations in the solution to Example3.2wn I .+-~-~ dt--~ R dq + -i--~q 1 = ~eo. the on the other : ~ . 1 EXAMPLE 3. Solution: When the four equations natural frequency is seen to be are compared to equation (3. taking the derivative of the second equation and substituting both the first and-the second gives a second-order equation with the dependent variable q: d2 q .150 CHAPTER 3. in both cases and all four equations. 1 R RV~_~ ~. (3. are different for the two cases: " O-junction 1-junction: The damping ratios.41).43) EXAMPLE 3.18 above. (3.21-R ’ I.18 Combinethe two first-order differential equations found in Example3.19 Find the natural frequencies and damping ratios corresponding to equations (3. equations for each case .2wnRC .14 for the 1-junction case into a single second-order differential equation in terms of (i) the dependent variable p and (ii) the dependent variable Solution: Taking the derivative of the first equation and substituting second gives a second-order equation with the dependent variable p: the d2p R dp 1 deo dt --~" + 7-d~ + -]-~P = ~-" Alternatively. Alternatively.

44) and its first and secondderivatives into the differential equation (3.33 for the commonly encountered special case in whichthe initial conditions are xh(0) = 1 and ~h(0) --~ 0. with the dampednatural frequency equaling precisely the natural frequency.5.41). whichapplies generally for second-order models. This special case requires Xhl : 1.3. Equation(3. since it applies broadly to the second-order modelsof the simple types considered in this chapter. Dampingoften is neglected to simplfy the analysis of vibrating systems. Homogeneous solutions are often called transient solutions as a result. Real passive systems normally have at least a pinch of damping.dd.) . w~often is called the undampednatural frequency.2.41). howeve2. The special case with no damping(~ = 0) producesa non-decaying sinusoidal oscillation. 3. In general. is lWd---. and by comparisonof its coefficients to those on the left side of equation (3. ~. you should be able to write firstorder differential equations directly from the annotated bond graph.since they apply only to special cases.41).45) This is such an important result that you are urged to verify it by substituting equation (3. SOLUTIONS OF LINEAR DIFFERENTIAL EQUATIONS 151 Youshould memorizethe left side of equation (3.althoughthe difference is very small for most commonmodest dampingratios ~ < 0. but not all second-order models.4 Solution of the Second-Order Equations Differential The solution to the homogeneous equation is Xh : (Xhl COS ~2dt ~. Note that WdiS smaller than wn. and then substituting the initial conditions. Xh2 -- (3.Xh2 sin03dr ) ~ <1 or 02 d > (3.V~--~: Wn. 3.5.46) -02d ~ ~ Youare urgedto verify this result also.44) is plotted in part (a) of Fig.44) in which the dampednatural frequency. that is a positive value of ~.] (3. determinethe natural frequency. To underscore the distinction. (Recall that a transient is someonewhodoesn’t stick around for long. regardless of whetherit is recognizedin a model. even if ~ is very small. You are not urged to me~norize the equations for the dampingratios or the excitation terms on the right sides of the differential equations. Damping causes the homogeneous solution to decay toward zero over ti~ne. It is found by substituting the solution into the differential equation. The solution for the special initial conditions becomessimply Xh = Xhl COSW~t.the dampingratio and the forcing function. It is also useful to rememberthe boxed equation for the natural frequency. combine theln to get a single second-orderequation.

0 0 2 6 10 12 (a) homogeneous response for x(O)=l.5 -1.33: Responseof the second-ordermodelto initial conditions and a step excitation .152 CHAPTER 3.5 0 -0.0 X/Xo 0.5 0 0 2 4 6 8 10 12 (b) responseto a step Figure 3.5 1.0 1. x(O)=O 2. SIMPLE DYNAMIC MODELS 4 8 1.0 0.

47) or (3.33. (3. Rather than using imaginary coefficients. which is known as critical damping. Perhaps the most commonconditions of interest specify that the values of x and its time derivative 2 are zero at the initial time.5. namely.Xhle -t/r1 -t/r -{.w. the sum of the responses of the model to the individual steps.48) is used for the homogeneous part. 3. The solution for this special case is xh = (Xhl +xh~. SOLUTIONS OF LINEAR DIFFERENTIAL EQUATIONS 153 For 0 < ~ < 1 the behavior is oscillatory.44). The values of these coefficients depend on the specific situation of interest. leaving only the particular or forced response.4s) The same initial conditions as above (xh(O) = 0. can be estilnated by approximating the excitation by a sum of steps. Twoconditions must be specified. Tile response of a linear second-order model to a stepwise excitation is. Note that any degree of damping causes the homogeneous or transient part of the solution to decay with time. the result for the case of ~ < 1 is x(t)=xl II--e-. for t > 0. (3. then. This idea is refined in Chapter 5.50) This response is plotted in part (b) of Fig.49) The particular solution forced response of the sec0nd-order equation (3. t = 0. Xp = xI.2Xh2e ~ > 1.47) For ~ > 1.41) for the excitation f(t) constant = x).t)e -~n* ({ = 1 or wd = 0). according to the property of superposition. regardless of whether equation (3. (3.~nt (COSWdt--~d sinwdt)] ~<1. Wd becomes an imaginary number. . As ~ is increased within this range the oscillations decay faster. (3. It has two unspecified coefficients. is the sameas th e p arti cular solut ion of the first-order model to the same disturbance. Thus the response to any excitation. Cases with { _> 1 also are plotted. is the sum of this constant and the homogeneoussolution.2V/~-. They vanish altogether for { >_ 1. After some algebra.nonly referred to as a step response.3. like that of the linear first-order model as illustrated in Section 3. kh(0) = 0) give V/~--1 ± ~ Xhl.5.4 above or of any order linear model. The general solution.2 -.I (3. This situation is com. one writes the solution in terms of real numbers as follows: Xh---.

form of equation (3. The final result is ¢(t) = ¢o coswnt.20 The floating. pitching block of Example 3. The motion is a sinusoidal oscillation at the natural frequency. or m -. which forces ¢: = 0. w. SIMPLE DYNAMIC MODELS EXAMPLE 3. ~ = 0 at t = 0. so therefore ~1 = ¢o. the total solution is the homogeneoussolution.hi1 + (h/w)~] Since there is no forcing and consequently a zero particular solution. the damping ratio ~ = 0 and the natural frequency ~n .½whp.~ .154 CHAPTER 3. . the inertance is I = hw(w"~ + h~)p/24. Neglect damping and the effect of any virtual inertia due to the water. 149) with the forcing function input f(t) = O.Also. you can take the derivative of the second equation and substitute the first: d2¢_ ldp dt "~ I dt 1 This is in the . The differential equations are dp 1 ~ = -~¢.41) (p. which from equations (3. de 1 To combine these equations and eliminate the variable p.~.44) and (3.m(w"~ ÷ h~)/12. The angle ~ is specified as ~0 at t = 0. bonded to a common1-junction with the angular velocity ~: pi/Ii Since the mass equals the mass of the displaced water. Solution: The model comprises a compliance with the value found in Example 1.45) (p. Characterize the behavior of the block after it is released from rest at an angle of ~b0. where m is its mass. and an inertance with the value above. 151)is ¢ = ¢1 cosw~t + ¢~ sinwnt.1 has a mass moment of inertia about its center of mass I -.

is X = Xh + Xp = (Xhl + Xh2Wnt)e -w"t + 2.5. The complete solution is therefore x(t) = -(1 + 3t)e -3t + 2.21 A system is described by the second-order differential equation d~ x 6 dx dt --~ + dt +9x=18’ in which the units of time are seconds. Since ~ = 1. 149): 1 d~" 2 dx 6dt-. and find x(t) assumingx(0) = 1.3.The first initial condition is x(0) = 1 = x~a + from whichXh~= --1. from which xhz = --1. SOLUTIONS OF LINEAR DIFFERENTIAL EQUATIONS 155 EXAMPLE3. The secondinitial condition is 2(0) = 0 = --Xh~Wn+ Xh~Wn= 3(-. Find the natural frequency and dampingratio of the system. so that 23 The particular solution is the const~t Xp = 18/9 = 2.-. Initial conditions nowcan be recognized.47) (p. the homogeneous solution is given by equation (3.+5-Yi+x =2. The total solution.1+ Xh~). .41) (p. Thecoefficient of the first term gives the natural frequency: 1 w~ = --~ = 3 rad/s The coefficient of the secondterm equals 2~/wn. 2(0) Solution: Dividing by 9 places the equation in the form of equation (3. 153). therefore.

Wd. C and R elements are coupled.form e Whenconstant I. Suggested Steps: 1. Integrate and .38). and a damping ratio. that is with terms of -~/r the .36) or through the use of equation (3. 3. Evaluate the time constant by comparing the differential tion (3.1 produces a special analytical form. v.5 CHAPTER 3.~the t. The response of any model described by a linear differential equation to virtually any excitation can be estimated by approximating the excitation by a sum of steps. equals the sum of the responses to the individual steps.4. Wdno longer exists as a real number. Evaluate the initial sketch-plot the result. Find and sketch-plot the speed and position as a function of time. equation to equa- 4. The homogeneous solution and the response to a step disturbance feature an oscillation at a damped natural frequency. ~. envelope e Whenthe effect of the R element becomes so large that ~ > 1. that is very close to w~ unless the damping is so large that oscillation barely can be discerned. wn. The oscillations decay within -. The solution is given by equation (3.5. The response. The borderline case of ~ -. Write the differential equation.37). the result is a linear secondorder model that. and of the speed. A large effect sometimes results from a large value of R.6 coasts to a halt from an initial velocity of 1. The time responses then are represented by the sum of two exponential terms with distinct time constants. 2. 5. The position of the drum is the time integral sketch-plot. The drum viscometer of Guided Problem 3. SIMPLE DYNAMIC MODELS Summary Whena constant resistance and either a constant compliance or a constant inertance are coupled.156 3. the result is a linear first-order model that can be characterized by a single time constant. in Section 3. according to the property of superposition.11 The problems below give practice in solving for the behavior of first-order and second-order dynamic models. Guided Problem 3. The homogeneous solution and the response to a step excitation approach equilibrium exponentially. Apply the causal strokes to the bond graph and annotate all the efforts and flows according to the procedure developed. and other times results from a small value.0 radian per second. can be characterized by a natural frequency. if the effect of the R element is not too large. velocity.

Only one undetermined parameter should remain at this point.01 Henry. 2. semble the solution.7 (p. Determine the natural frequency of the motion following a non-equilibrium initial condition. 126) has a diameter d = 0. Write the two coupled first-order differential equations.4.8 (pp. the fluid is water (pg = 62. Choose values for the resistance. R. Find the simplest particular solution possible (with no undeterinined coefficient). Youare interested in the load current. Compareyour differential equation to equation (3. Use the limitation of the maximumallowed voltage to find an equation that determines the remaining parameter. 149). Therefore. chosen in step 3 becomes the de- 5. Use the initial conditions to determine the undetermined coefficients. to choose an effective damping ratio. perhaps using plots given in the book for various dampingratios. 8. Examine the nature of the solution with this constraint in mind.5. Identify the values of the natural frequency and the damping ratio. equation. Apply the causal strokes to the bond graph and annotate all the efforts and flows according to the procedure developed in Section 3. 4. 10.25 inches and a length L = 10 inches. Combine the equations so the variable pendent variable. Note the pres- 7. Sumthis with the homogeneoussolution to get the total solution. Suggested Steps-" 1. As- 9. The voltage should not exceed 15 V. The limiting voltage is proportional to the rate of change of this current. Guided SOLUTIONS Problem OF LINEAR DIFFERENTIAL EQUATIONS 157 3. 133) undergoes a step change in time. choose as the variable for your combined differential equations the state variable proportional to current. Find the homogeneoussolution of the differential ence of two undetermined coefficients. C which satisfy the constraint and provide a good response.4 lb/ft3). and plot the current as a function of time. 127. and capacitance. PROBLEMS 3. .41) (p. The load inductance is 0. The limiting voltage effectively sets the maximumrate of change of the current. You should now be able to give values for both C and R. 6.3.41 The U-tube of Guided Problem 3.12 The current source of the electric circuit of Guided Problem 3. from 0 amps to 1 amp. 3.

26 (p.4(A)¯ Finally. defined as the natural logarithm of the amplitude of successive peaks. Find the subsequent depth as a function of time. 130) and Probcharge q0. 130) and 3. 144) is leased from rest 0.42 The capacitors of the electric lem 3.05 in3/s begin abruptly.50 The mass-spring-dashpot system below exhibits the response plotted at the top of the next page when the mass is struck by a hammer.23 (p. 3.34 (p.36 (p.46 Tile water tank of Problems 3. Find its subsequent velocity. SIMPLE DYNAMIC MODELS circuits of Problems 3. 3.24 (p.35 (p. Find and sketch-plot the position of the mass as a function of time. A.29 (p. 131) and 3. but the springs are initially unstressed. 144) have an initial inductors as a function of time.0 m.28 (p. Find the subsequent depth of the water in the tank.0 ohms.27 (p. 144) are given an initiM velocity 50. 3. invert the relationship to give a function 4 -.) (a) Estimate the natural frequency (w~) and the damping ratio (4) system. Find the position of the mass as a function of time for (i) system (a). 131) and 3.47 The vibration isolator of Problems 3. 3. but suffers no other excitation. C = 10 #f and R = 2. 144) is released rest.25 (p. 131) and 3. (ii) system (b). assuming L = 1.43 The masses of Problems 3. Assume k = 900 N/m.158 CHAPTER 3. Assume 3. 132) and 3. equals the constant -2~r4/x/~ . 144) has the Q0 = 0. Find the currents through the 3. from 0 to 1 volt.49 An underdamped second-order model oscillates with decaying amplitude in the absence of a continuing excitation. 144) experiences a step change in the voltage ei. 3. Next. 131) and 3. Hint: The damping ratio is so small that there is an indistin- . Find the output voltage eo.0 mh. m = 2 kg and b = 12 N s/m. 3.40 (p.48 The filter circuit of Problems 3.45 The plunger of Problems 3. Show ~hat the logarithmic decrement. relate A to the log of the amplitude ratio for n cycles. 3.42. (The blow effectively imparts a velocity to the mass instantaneously.02 m from its equilibrium position. 144) has an initial depth of 1.39 (p.44 The fluid tank of Problems 3. The tube is full of water that has a zero initial velocity.37 (p.38 (p.

8 -’1 0 9 4 6 8 ’10 ’~9 14 "16 18 90 time. Damping is to be neglected.2 ^ 1 0 -0. I motion (a) Find the change in the bouyancy force resulting from a vertical displacement to the disk from its equilibrium position. Interpret this result to give a compliance. (b) Estimate the spring constant.35 Hz. (d) Find the length of a cylinder of water of the same radius r that has mass equal to the virtual mass. It bobs vertically in a tub of water of density 1000 kg/m3 with an observed frequency of 1. inches °.3. SOLUTIONS OF LINEAR DIFFERENTIAL 159 EQUATIONS guishably small difference between the natural frequency and the dainped natural frequency.51 The cylinder shown below has a radius and height r = h = 10 cm and density equal to 750 kg/m3. (b) Solve for the effective inertance. (Note: In estimating the inertance of the . position. (c) By comparing the result of part (b) with the mass of the cylinder. using the observed natural frequency and the value of the compliance determined in part (a).5.’l[ II II // a 0. given that the mass is 4 kg. determine the "virtuM mass" of the water that effectively moves with the cylinder. seconds 3.4 -0.

100 ft 2 and length L = 3000 ft carries water from a large lake to a turbine.Oe-t/16.52 A long tunnel or penstock of area Av -. SIMPLE DYNAMIC MODELS fluid inside a circular tube with thin-walled ends surrounded by fluid.1. ~" = R 0.84 seconds 4. Such a correction is widely used in acoustics. seconds 4o How .160 CHAPTER 3. this length also should be added to each end. - . ~ . initially 1000 ft 3/s. o ~o :o ~o . 12e_.= -~--p dt GUIDED PROBLEMS (Makesure you have the minus sign. A large surge chamber of area As = 1200 ft -~ relieves the pressure surge that occurs. If the end of the tube is surrounded by a flange. Neglect all damping and the inertia of the water in the surge chamber. the end correction should be about 39%greater.s4 5. seconds 40 0 ~o :o ~0 t. and is especially important if the tube is short.8.t. surge ~///~chamber penstoc~ _- turbine ~ (a) Define variables and parameters and draw a bond graph model for the system. (c) Solve these equations analytically and sketch-plot the results. ceases abruptly. In an emergency shutdown the flow through the turbine. and when does it peak? SOLUTIONS TO Guided 3.00713 = 16.) 3. (b) Write differential equations of motion. These cases are discussed further in Section 10. high does the water in the surge tank rise.11 Problem -.) 3. ~) = ~)oe-t/~ = 1.

1 dp 1 4.g0) (p.~(e-¢~"t)(-iocoswat dt +ih~ sinwdt)]lt=o = in~wa + ~w~io {h : e-{w=t(ihl Therefore. Therefore.3. however. Multiplying this equation by IC ~d substituting p = Ii gives i~d2i I di ~+~+i=io Comp~ing~his with equagion (a. 9. ~d . so ihl = --io.19. To determine ih2 it is necess~y ~o take ~he derivative of i(t) at t = 0 and set this equal to zero: di It=o= 0 = [(e )(zowasmwdt+ih2wacoswat)-~w. Unforguna~ely. 1 dq 1. The plots in p~g (b) of Fig.a l~ge ~ is desired. 151): sinwat). ~he solution becomes .) 6.0.44) (p. SOLUTIONS OF LINEAR DIFFERENTIAL EQUATIONS Guided Problem 1.12 R q/C~q/RC ~o q/C~ p C dq 1 1 2. 8. ~_~sin~et whichagrees wi~h equation (a. ~he voltage which must no~ exceed 1~ V equals q/C = dp/dt = Idi/dt. i=i0 ih2 1-e -~ COSNdt + ih2 = --~Wn/Wd : --iO~/ coswet ~-- if2. -~ = io .5. 1~2) reveal ghag the l~ger ~ is chosen. laa). ~he curves appro~h their final value with an unnecess~ily long ~ail. Theinitial current is zero. Therate of changeof this current also is zero. Since the excitation is a step of amplitude i0 = 1.0 amps. When~ is ~eater than 1. a. q will be eliminated from the equations abovein favor of p.41) (p. 150. Thetotal solution is i = ih + io. d2p 2dt C dt RC C -~-~P = -~o RC dt ~P 5. Assuming~ < 1. ~herefore. 149) gives ~=~. ~ = 1 is a good solution.aa (p.yp dp 1 d-~ = -~q 3. 161 3. p. the p~ticul~ solution is i~ = io = 1. ghe equations above fN1 for ghis critical damping.-~-~q . ghe l~ger wn can be wi~houg ghe slope of ~he curve exceeding ~y p~ticul~ m~imum. wa = ~ ~w= 7. ~= 7~-~v~ (These results can be seen to agree wi~h those in Example3. 1 q 1 1. the homogeneoussolution is given by equation (3. The load current i equals p/I.

3. cantilevered at one end. Therefore the spring becomes progressively stiffer as it deflects. This member.34 comprises a uniform strip or beam of elastic material.4 0.w~t)e The m~imum voltage occurs whendZi/dt ~ is zero.owever. SIMPLE DYNAMIC MODELS -~t. Few nonlinear differential equations possess analytic solutions. that wraps around a fixed curved memberas it deflects.01 . The presence of a nonlinear R.h.1 Nonlinear Compliances The spring shown in Fig.015 pf. Numerical simulation.6 Nonlinear 0. On the other hand.01 = 4077 rad/s = 649 Hz.6.6.0 and Inertances The compliances and inertances considered thus far are constant.00 amps. and therefore give linear relations between effort and displacement or momentum and velocity.(4077)e 1x . This type of characteristic is represented by a curved line as in part . The concept of nonlinearity is now extended to compliances and inertances. C or I element in a dynamic model produces a nonlinear differential equation. ms Compliances 2. using MATLAB. 3. At this ti~ dt]m~ = Iw~e = 15V = e from which w~ = 15eli = 15 x 2. The differential equations considered thus far also are linear. or whenwrit = 1. i = io + (ihl + ih’2wnt)e The initial conditions render ih2 = ihl = --i0 = --1.0 0 . o.3. -.8 1. is employed in Section 3.1 -~ 6. CHAPTER 3. from which di ~. manyof the resistances considered in Chapter 2 are nonlinear.7 as a practical alternative.6 t. foreshortens the free length of the beam as x increases. and 1 C = w~l .718/.2 1.015 x 10 1. Without the curved memberit would have a linear characteristic for small displacements.o~t ¯ d-~ =~" r e d’2i dt ~ = io(w~ .162.

which is used in the linear case. the causal input 0 is integrated to give the state variable q. A generalized nonlinear compliancecan be described by a nonlinear characteristic. can be retained for nonlinear cases.51) whichalso is true of the linear special case.34: Exampleof a nonlinear spring ei qi (a) linear (b) nonlinear Figure 3. or V(x) = J g dx. and the causal output is expressed as the appropriate (nonlinear) function of The relation ei = qi/Ci. The term chordal compliance sometimesis used to underscore this definition. For the preferred integral causality. this approachcorrespondsto defining the complianceas the inverse slope of the chord drawnfrom the origin to the point of interest on the characteristic. (3.6. As can be seen from part (b) of Fig. if you so prefer. The energy stored in the spring equals the area under the curve.35. NONLINEAR COMPLL4NCES AND INERTANCES 163 F (a) system (b) characteristic Figure 3. whichcan be expressed algebraically either in a form appropriate for integral causality or in a form appropriate for differential causality.3. 3. It allows the writing of differential equations as thoughthe element were linear. Thus.35: Definitions of compliance (b) of the figure. .

qi = qi(ei). the pressure is proportional to the depth. Froma practical viewpoint. there is usually advantage to the approach using equation (3. For arbitrary nonlinear compliancewith generalized displacementq and effort e. The two approaches are analogousto the use of ei = Ri0i as compared to ei = ei (Oi) for a nonlinear resistance with the same (impedance)causality. The correspondingparticular relation for the tank is found by using the relations P = pgh.52).164 CHAPTER 3. The differential relation implied for the general complianceis qi(ei) Ci (ti = ~qi. . Se ~.54) The constant-effort source. as illustrated in Fig. (3. applications of differential causality for both linear and nonlinear compliancesand inertances are given in Chapter 4. if the surface area of the tank varies with the depth of the liquid. the relation is e = e(q).it is useful to consider the constant-effort source as a unique element. SIMPLE DYNAMIC MODELS P fQdt= V Figure 3.53). Nevertheless. The general relation is P = P(V).36. since it serves to store and dispense potential energy and can be represented by an effort-displacement characteristic. could be thought of as a very special type of nonlinear compliance.6. 3. and V = f2~=: A(z) dz.it can be treated directly. however. however. where z is a running variable for the depth.52). where h is the depth of the water. but the volumeis not.the co~nputational requirement is that Ci be a funtion of the state variable qi: [ Ci : Ci(qi). therefore.36: Nonlinear fluid compliancedue to gravity If C is not constant. Placing differential causality on a nonlinear complianceshould be avoided. if possible. ] (3. as given by equation (3. This happens. 3.2 Nonlinear Fluid Compliance Due to Gravity The gravity complianceof a tank of liquid is nonlinear if the pressure at the port near the bottomof the tank is not proportional to the volumeof the liquid stored abovethis level.53) ComputingCi involves about the same effort as the direct computation of ei using equation (3. As a practical matter.

~i. d--~=Qi’-Q(P)=Qi’~-A° =Qin-Ao 2 2g ~t~n This equation is solvable. and write a solvable differential equation in terms of V. Determineits fluid complianceas a function of the volumeV of liquid in the tank. . with Qi. Then.6. Solution: The compliancerelation can be found as follows: V = ~rr ~" dz = 7~z 2 tan 2 adz = 3 3 " This result is inverted to get the desired form P = P(V): P=pg ~z 3V ~/~ ¯ The expression P = V/C still applies. . treated as an input.I c P(V) ’_ P(v) _ UsingBernoulli’s equation for the nonlinear resistance that gives Q(P).3. and the orifice flow resistance relation Q(P). since the only unknowns on its right side are the given excitation Qin and the state variable V. although at this point finding C is an unnecessaryextra step: C(V)= V_ 1 P(V) 1/3 [7~tan2c~V2~ 3 The fully annotated bond graph below includes the effort P(V) above. 165 NONLINEAR COMPLL4NCES AND INERTANCES EXAMPLE 3. give a fully annotated bond-graphmodel.22 A conical-shapedtank ends in a small orifice of area A0.

The integral of this flow equals minus the changeof volumeof the gas. as shownin part (b). from whichthe characteristic in the desired form is P0 (3. is a rigid chamberinto whicha liquid such as water or oil enters or leaves.shownin Fig.55) k P . The system no longer conserves mechanicalenergy. whichis relatively incompressible. An accumulator can be modeled at least crudely over a wide variation in pressure by a nonlinear compliance. for which heat transfer is reversible.It is positive for flow directed inward. as also shown. The resulting pressure can be estimated from an appropriate equation of state for the gas. For the present this complication is neglected. and whichalso contains a fixed amountof a gas such as air or nitrogen (the latter to reduce the fire hazard with mineral oils).(1 . Accumulatorsserve the purposes of storing significant amountsof liquid and energy.3 CHAPTER 3.V/Vo) Here. Nonlinear Compressibility SIMPLE DYNAMIC MODELS Compliance A hydraulicaccumulator.166 3. The chordal definitions of complianceallow the combinationrules for linear compliances to be retained for nonlinear compliances. Isothermal conditions. the three efforts sum for a given commondisplacement. Po is the "charging pressure" that exists whenthe accumulator is empty of liquid but contains the full charge of gas. assumingdependencies . The gas maybe separated from the liquid by gravity or. are present only whenthe charging and discharging is so slow that significant temperature differences do not develop. The flow on the compliancebond refers to the flow of the liquid.38. to the extent that there is too little time for significant heat transfer to occur. 3. 3. In any case the pressures of the two fluids are virtually equal. Whenthe compliances are bonded to a common0-junction.6. Adiabatic conditions for an ideal gas give p/pk =constant.Isothermal conditions give the same equation except that k is replaced by 1. Whenthese compliances are bonded to a common1-junction. Actual charging and discharging rates usually lie somewherebetween these extremes. Compressingor expanding a gas by a large proportion changes its temperature significantly. the three displacementssum for a given common effort. the combinedcharacteristic is the horizontal sumof the three characteristics.4 Junctions With Multiple Bonded Compliances The characteristics of three arbitrary compliancesare given in part (a) of Fig. and Vois the total volumeof the chamber. and reducing pressure and flow surges in hydraulic systems. on the other hand.Heat transfer over a significant temperaturedifference is irreversible. more effectively. It is addressed in Chapter 10. Thegeneral shape of the characteristic is shownin part (b) of the figure. potentially causing heat transfer with the surroundings.6. and cannot be modeledaccurately as a pure compliance. Adiabaticconditions are present only. by a piston or a flexible barrier such as a diaphragmor a bladder. 3.37 part (a). the combinedcharacteristic is the vertical sumof the three characteristics.

37: Hydraulic accumulators .6..V x<~ valve that closes "l~:~]’whenV=Oto prevent neck ~]bladder ext~edfrom through neck~~ ~l ~] being Q neck by gas pressure piston type bladder type (a) majorconstructiontypes (b) compliancecharacteristic Figure 3. NONLINEAR COMPLIANCES AND INERTANCES valve for adjusting _~~s 167 ~1 valvefor adjusting charge charge gas ~e.Vo-V p~ston liquid standard symbol P.3.

5 Nonlinear Inertances Aninertance is defined as nonlinear if p is not proportional to 0.38: Graphical reduction of multiple compliancesbondedto a junction P relativistic mechanics inductancewith saturation Figure 3. and therefore inertances are linear. and for the 1-junction. Therefore. A happy fact: in classical mechanics. 3.) The electrical inductor has a limiting generalized momentum because the magnetic flux of real materials reaches a limit.The shapes of the characteristics for these cases are shownin Fig. 1/C = ~ 1/Ci(q). This is analogous to a nonlinear compliancein which q is not proportional to e.39: Nonlinear inertances are expressed as functions of the common variable.and (more importantly to engineers) in the electromagneticdomain.massesare not functions of their velocities. C = ~-~i C~(e). 3. (The axes have been reversed to makethem compatible with the next figure.168 CHAPTER3. for the 0-junction. Youwill find nonlinear inertances only in relativistic mechanics. q (a) individualcharacteristics (b) combinedcharacteristics Figure 3.39.including fluid mechanics. SIMPLE DYNAMIC MODELS for commonq C~ / add ¯ q’s for commone C~ e \ qi _.6. the phenomenon .

are suggested in Fig.56) causality they are dpi ~. (3. this is called potential energy. is the product of the effort and the flow. as you have seen.3. 169 NONLINEAR COMPLIANCES AND INERTANCES s called magnetic saturation. and is given the traditional symbol Y: In similar fashion. as you knowalready. roman(non-script) Vand T traditionally also are used for potential and kinetic energy.40. I (3. linear or nonlinear. the energy stored is its ti~ne integral: /~90dt=/edq-p.57) 3. f Od (3.61) The dual nature of the functions e = e(q) and c~ = O(p) and the energies ~. this is called generalized kinetic energy and is given the traditional symbol.59) The integral f e dq above is integrable. 3./-4: [T= f O(p)dp. (The weight of many electronic packages such ~s stereos often resides mostly in their transformers.6 Kinetic and Potential Energies The power flowing into a simple compliance or inertance. which contain considerable tron in order to minimize magnetic saturation. Both are state variables. .) 4Asnoted before.5S) Since this power is flowing into a reversible energy-storing one-port element. the integral f c)dp above also is integrable if 4 is a function of p. Pi = f ~ dt.6. if e is a function of q. and 7. Note that linearity is just a special case. or P = eo = ~0.P(~]i)d (3.) The preferred integral causality and the associated computation are similar to those for the compliance: (~)’~1~i O~(p) Eor differential (~i = (ti(Pi).6.respectively. (3. (The axes of the inertances are inverted from those given before in order to emphasize that momentump plays the same role for kinetic energy that displacement q plays for potential energy.

SIMPLE DYNAMICMODELS e=p (a) constitutive relations andenergy (b) definitions of C and (c) linear case Figure 3.170 CHAPTER3.40: Generalized compliancesand inertances .

Similarly to inertances. C = C(q). In engineering practice. the inductance (electrical inertance) of the coil in a solenoid depends on the displacement of the moving mechanical member. In nonlinear cases it is usually more convenient mathematically to use the particular relation in the form e = e(q) rather than e q/ C(q). Consideration of displacement-dependent inertances is deferred to Chapters 9 and 10. The definition of the chordal inertance. most nonlinear inertances occur because of magnetic saturation in electrical inductances. although either of these forms imply the favored integral causality. is recalled in part (b) of this figure.q e =. Constant inertances are linear. .3. these cases usually are most readily treated computationally using 0 = 4(p) rather than gl = p/I(p). that is the reciprocals of the compliance and the inertance. I = I(p). These cases are deferred to Chapters 9 and 10. C = C(q) and the compliance is said to be nonlinear. The slopes of the chords. Se. Sometimes a generalized kinetic energy or potential energy depends on more than one independent variable. The use of differential causality is treated in Chapter 4. a change in something else. conversely. the associated C is constant. inertances are linear. might also cause an inertance to change. not infrequently a displacement. the capacitance (electrical compliance) of a capacitance microphone varies with the mechanical displacement of its diaphragm. When proportionality does not exist. NONLINEAR COMPLIANCES AND INERTANCES 171 The definition of the chordal compliance. Potential energies dependent on two displacments are considered in Section 10. some linear compliances vary because of changes in a variable other than the two used to define the linearity. produce an equivalent compliance smaller than any of its components In classical mechanics. the fluid inertance of a channel that is empty toward the downstream end depends on the displacement of the fluid interface. and the compliance is said to be linear. Constant compliances are linear. for the compliance.62) Compliance and inertance moduli are assumed to be chordal unless otherwise described.2. It is not necessary for an inertance to be constant for its relation between 0 and p to be linear. is given in a similar fashion as the reciprocal of the slope of the chord: q. Analagous to the compliances.6. For example.7 Summary Whenan effort and its associated displacement are proportional to one another. including fluid mechanics. where they are treated as compliances with two ports or bonds. Wh en the generalized force is independent of the displacement. For example. Two or more compliances bonded to a commonI-junction. Two or more compliances bonded to a common0-junction produce an equivalent overall coinpliance larger than any of its components. are called generalized stiffness and generalized suseeptance.P lc -= ~ e ’ -: "q (3. however. respectively.6. you should substitute an effort source. In another example. however. 3.

3.7. Make sure the two answers agree. assuming integral causality. Find the chordal compliance as a function of the angle (although this is not needed in practice). and take its derivative with respect to the angle to get the moment. find the chordal compliance. Problem SIMPLE DYNAMIC MODELS 3. The shape of the tank is described by y = ax2 and the width. between the gravity- 3. Get the relation between the moment and the angle a different way: express the gravity energy as a function of the angle. Perform a force-moment balance to get the relation induced moment and the angle. Note that the methods PROBLEMS 3.172 Guided CHAPTER 3. 4. w. and briefly discuss its desirability. Suggested Steps: 1. Place causal strokes on the bond graph. . 2. Find the compliance relation between the moment and the angle without restricting the size of the angle. Suggest how this can be accomplished with special shaping and mounting of coil springs. A simple pendulum comprises a slender uniform rod of mass m and length L pivoted about its upper end. and annotate the efforts and flows using the results above. 5. It is suggested that you start with a 1-junction labeled with the angular velocity. equations.54 Evaluate the nonlinear compliance relation for the two-dimensional paraboliC tank shown below. Draw a bond graph model for the system. Also. 3.13 This problem gives needed practice in determining a mechanical compliance relation from a force balance and also from the use of potential energy. Thes~ equations are solved in Section 3. 6. and the state differential equations that describe the oscillation.53 A well-advertised mattress gets abruptly stiffer above a certain deflection. Write the nonlinear state differential introduced thus far cannot solve these equations.

(a) Determine the compliance relation.3.22 (p. 3.22 (p.55 Solve the problem above for the tank being axisymmetric about the y axis.6. Write a set of state differential equations to model the system. .57 The conical tank of Example 3.5 (p.13 M -. drains to atmosphere through a vertical tube of area A and length L.13 (p. Viscous effects may be neglected. 109) is given a large initial displacement. so the compliance is not constant.56 The fluid tank of Example 3.mg-~sine ¢ C (~mgL~)sin ¢1 1 p’~//t dp ~= -(mgL/2) sin ¢ vd-~ = i I = mL’2/3 equations. 3. 1 Include the inertance effects of the tube.(mgL/2) sin ¢ L V = m9-~(1 c= ¢--= M mgL cos¢). but with no supply tube. (b) Write a complete set of first-order state differential SOLUTION TO GUIDED PROBLEM Guided Problem 3. and the resistance assumed to correspond to a single head loss in the tube (giving a pressure drop of ~_pv’). sin dV L M=a¢)-~’---. Repeat the analysis for this case. 165) is now formed in the shape of two-dimensional "V" rather than an axysymmetrical cone. 165). 100) and Problem 3. 3.58 The rolling disk of Problem 3. NONLINEAR COMPLIANCES AND INERTANCES 173 3.

t). This "state-space" notation is written as dx = f(x. slavish focusing on them is no longer warranted or necessary..5 and used throughout this book. indeed. such integrators can behave erratically.7. Nevertheless. some knowledge of their algorithms is worth the modest effort required. A maze of patch cords gave structure to a particular model. x. (3. the very design of those systems.1 State-Variable Differential Equations A model first should be reduced to a set of n independent first-order differential equations. Until the 1940s other differential equations could be tackled only through limited and laborious graphical or numerical procedures. The state vector. The number of state variables. including the MATLAB integrators described below and elsewhere.63) Boldface is used to indicate vectors. SIMPLE DYNAMIC MODELS Numerical Simulation Only linear differential equations and rather simple nonlinear differential equations allow analytical solution. is a vector of the state variables.174 3.. . The use of vector notation is helpful. The convenience. when written by hand. differential equations and the systems they represent. Even if you have little facility for solving differential equations analytically. Linear systems indeed possess some intrinsic virtue that commendsthem for engineering design. The first major tool developed to solve time-based differential equations automatically was the analog computer. accuracy and increasing speed of the digital computer finally pushed the analog computer into obsolescence. which was realized most fully when they were hybridized with digital computers. This impacted the modeling of dynamic engineering systems and. the typesetters’ symbol for boldface is recommended: a wavy underline. In the 1950s these "simulators" evolved away from their mechanical origins to become largely electronic circuits based on the operational amplifier in which vaxious voltages behaved in time as did the associated variables in the equations. you can utilize them effectively in the analysis of dynamic systems. You don’t even need to understand the numerical methods that the software employs. or Ix1 x2 . Analog computers were used into the 1970s because of their potentially high speed. In particular. is called the order of the system. often using grindingly slow mechanical calculators. and banks of adjustable resistors allowed parameters to be set. great emphasis was placed on linear or linearized models. On the other hand. flexibility. x~]T. and partly because the MATLAB simulator that is used below is based on it. n.7 CHAPTER 3. This form results naturally from the procedure described in Section 3. partly for its economy when a model is of high order. 3.

It is employed frequently is this book..dxn/dt are represented by the vector f...." again using the global declarations. x~. The derivatives dxl/dt. The MATLAB simulators presented below "integrate" any set of linear or nonlinear differential equations of the form given by equation (3. f(1). This is done special subprogramcalled a function M-file. x(2) . any parameters that are needed to evaluate the derivatives must be defined within the function file.3.2 Simulation With MATLAB MATLAB is a commercial software package that competes with and can be used in conjunction with C.7. It is especially easy to use. the only arguments that are communicated to this subprogram from the main program are the time t and the state vector x. C++ and Fortran. that is much more efficient computationally because it employs analytic solutions. not the name given on its first line. ¯. presented in Section 5. They may reside on your floppy disk.63). x~.. x~. .x) Unless you make special provision. x~. These optional elements of good programming practice are presented in Appendix A under the heading Communication Between Files.. With this basic scheme.63)) themselves. or better yet on the hard drive.. You also can write your main program as a special "script M-file.. To avoid the repetition. Unless you make special provision. The function M-file is called with its file name. are communicated back to the main program.. The first active line of this file is as follows: £unc~ion f = ~name~(~. which comprises the state variables x(1). The subprogram normally ends with differential equations in the form f(1) = <first function oft.. which has components f(1).. All M-files must be given the extension .... and be properly opened. x~. but you should make them the same. Before you proceed further it is desirable that you study Appendix A. This file may be executed thousands of times during a single simulation. f(2) .x(n).7. xn > f=f.m. and has features that commendit particularly for the analysis of dynamic systems.. f(n). The first step in carrying out a simulation is to code the coupled first-order state differential equations (equations (3... x. values can be communicatedback and forth between the main program and the function M-file through the use of a "global" declaration. Their directory ." dx:/dt. > Xn f (n) = <n_th function of t. The last statement above converts a row vector to a column vector.. rather than merely entering it into the commandwindow. f(2).3.. . f(n).. so the procedure can be quite inefficient.. only the elements of this vector. MATLAB also has a simulator for linear models. in order to be recognized.xn f (2) = <second function of t..)...7. 175 NUMERICAL SIMULATION 3..

Procedural details are given in Appendix A under the heading Script Files. The color code (which depends on the version of MATLAB being used) gives you the state vectors in the order they have been defined.176 CHAPTER 3. or to designate individual data points by various symbols. Details are found by clicking on " ODESET"within the help window. The command plot(t. These.l). SIMPLE DYNAMIC MODELS location should be entered into the path that MATLAB uses. Likely different variables have different magnitudes. Should you wish to plot x(1) and 100 times x(3).x0) The command"ode" stands for ordinary differential equation. The plot command will execute only if the number of rows in t equals the number of rows in x. at the expense of execution time. Securing a plot of the results is probably your final step. with a commonscale. as a function of time. Accuracy is increased. Further optional arguments can be added to control the accuracy of the simulation. The time is stored as a vector of the same length. use the "help" window. for example. in turn. by decreasing the size of the time steps. The third argument.x) will plot all the variables.x(:. Default values are set. Should you wish to abort a simulation that seems endless. Often it is wise to try a very small value for the final time. For example. The second argument on the right is a two-element vector comprising the initial time and the time the simulation should stop.3)*lO0) The state vector x is stored as a three-column matrix. Before the simulation can be executed. is the initial condition noted above.t. otherwise. enter plot(t. and some may be of little interest.x(:. until you are sure the behavior is reasonable. . the colon (:) is a wild ¯ card. the initial state vector Xl(0) = 1. Next. function file [0 10] . x~(0) 0. the initial conditions must be specified as a vector. are set automatically with respect to error indices that you can specify. x0. x3(0) = 2 can be entered x0 = [1 0 2]. Mistakes in the differential equations often result in instabilites in the solution which take almost forever to execute. For information on howto control the colors of individual plots.x]=ode23(’<name>’. type control-C. the "23" is a descriptor of the algorithm used (several alternative available integrator algorithims are described under "help: Function functions and ODEsolvers"). the commandto carry out a simulation using a particular for the differential equations can be issued directly: [t.

from 0 to 3 seconds. the function m-file becomes functionf = pendulum(t.x) L=I. with its natural frequency of w. g=32. Plot the angle as a function of time for a period of three seconds. or one cycle in 0. f(1) = -W*(L/2)*sin(x(2)). The differential equations are dp L ~ = -mg-~ sin ¢. As the angle approaches 180° . on the other hand.[O 3].13 exhibits interesting behavior because of the nonlinearity of the gravity-induced compliance.[0 pi/18]). The main program.2.23 The pendulum of Guided Problem 3. mg=1 lb and g = 32.9498 rad/s.. W=I. The frequency for an initial angle of 90° . >> plot(t. with initial conditions comprising zero velocity and an angle of ~r/18 = 10°. could comprise only two lines: >> [t.The frequency of the oscillation for the 20-degree swings can be seen to be almost the same as the frequency for the 10-degree swings. entered into the command window with its prompt >>. This shows that the assumption of linearity. the frequency approaches zero. enter the command >>hold on Someresultsareplottedat the top of the nextpage. Does this agree with your intuitive sense? . is approximately valid for angles as large as 20° . Should you wish to superimpose plots for simulations with different initial conditions. I=L^2*W/(3*g).904 seconds. = 1//x/~ = v/3 * 32. NUMERICAL SIMULATION 177 EXAMPLE 3.7.2/2 = 6.2 ft/s 2. is noticeably smaller. f=f. Notice the hesitation as the pendumlumpasses through its maximum angle. Simulate the system. dt de 1 d--[~=-~P" Solution: Letting the state vector be x = [p q]’ and the parameters be L = 1 ft..x] = ode23(’pendulum’. The final statement above converts a row vector into the necessary column vector. in degrees.3.2).x(:. f(2) = x(1)/I. starting from rest at various angles.180/pi) This produces a simulation and plot of the angle.

Oncex(t0 + h) is found. robust and easy to program. t) and x(to).7. reducing computer time.0 Integration Algorithms 2"Otime. to is reset to equal the previous to + h. but makes the methodself-starting.3 0 1. without the error exceedingset limits. you ought to avail yourself of current literature and software packages. There are manydifferent methodsavailable. the presentation belowis limited to the simple class of Runge-Kuttasingle-step algorithms. This choice is based on their simplicity and robustness. The fourth-order version with automatic setting of the time increment is a satisfactory work-horsefor most problemsthe engineer is likely to . Should you at sometime do a great deal of numerical simulation. The methodpresented below is a basic l~unge-Kutta integration scheme. can be varied at any stage in the solution is a special advantage. SIMPLE D~NAMIC MODELS ° 180 °120 °60 °0 -60° °-120 °-180 3. Runge-Kuttaalgorithms estimate the state x(~o + h) based solely on knowledge of f(x. h.encounter. The fourth-order Runge-Kuttamethodwill be approachedinductively. This allows the time step to be madeas large as possible.0 The numericalintegration of differential equations is a major subject in itself. The ease with which the step size. whereh is the incrementbetweensuccessive times at which the state is coraputed. plus their reasonable if not optimumefficiency.178 CHAPTER 3.seconds 3. start- . Throwingaway the known history in this mannerreduces the potential efficiency of the process. and the "single-step" process is repeated. Rather than survey these methods. even for the ordinary differential equations of concern here.

7. whichcan be extendedto higher orders.~ actual solution area is estimated ] ~tual / =f (area change in x ~ curve is the actual Io t to+h t underneath the change in x) (a) classical Eulerintegration actual k =f improvedestimate [ of the change in ~.~ ~l/estimated (trapezoidal ~ integration) I to to +h t (b) second-orderRunge-Kutta ~. 3. 3.41.64)). NUMERICALSIMULATION estimate of X(to+h) I ~. but can be estimated using a linear extrapolation of ± = f.66) is the sameextrapolation as in the lowest-order method(equation (3. (3.xo + hf(xo. the second-order symmetrical Runge-Kutta algorithm becomes h x(to + h) -~ x(to) + ~ [f(xo.64) wherexo --= x(to).179 3.67) where f+(to) is found from equations (3. 7.+ .~f. 3. . A moreformal methodof derivation. whichis x(to + h) . This is the classical Euler formula.41.64) is replaced by the average of f evaluated at to and f evaluated at to h.. to + in which x+(to) = xo + hf(xo.41: First and second-order Runge-Kuttaintegration ing with the lowest order. as shownin part (a) of Fig.. (3. as shownin part (b) of Fig. to). Hence.68) .(~o+~ Figure 3. employsa Taylor’s series expansionfor x(to + h): Of f(to) x(to + h) = x(to) + f(to)h + ~xx to + ~-~ -~. to) in equation (3. (3. to) + f+(to)]. the estimated state is direct extrapolation of the slope it(to).65) and (3. (3..65) f+(to) = f[x+(to).4 Second-Order Runge-Kutta Afar moreaccurate estimate of x(to + h) results if the term f(xo.66). the estimate of f at time to + h is (3. The latter is not knownin advance..

74a) z.75a) (3. (Of/OX)to is an n × n Jacobian matrix of derivatives.180 CHAPTER 3. and #2 are constants to be determined. through second order gives A2 = 1 .754) z4 = hf[x(t0) + z2.67).73) x(to + h) ~. and the cumulativeerror is proportional to the fourth powerof h. (3. permits the equation to become x(to+h) ~ x(to)+(Al+A2)f(to)h+A2 #1 Of f(to) "~X to + which can be compared directly with equation (3. A2. SIMPLE DYNAMIC MODELS Here. to + ~h].)’ to Equating the terms (3.x(to) + z2 1 e = ~(-zi + 4z2 .71b) where A1# 1 but otherwise is arbitrary.f(to)h + A2f[(x(to) + #. to). 1 #. 3.67).f(to)h.74b) in which 1 1 (3.~h]h. Twoequivalent forms of this are 1 x(to + h) -~ x(to) + ~(zl + 2z2 + + Z4). to + #. h. to + ~h]. (3.z4). .7. = #2 = 2(1.74) is proportional to the fifth powerof the time interval.75b) 1 (3. Z3=hf[x(to) + ½z2. #. but a highly stable and commonchoice is the symmetrical form AI = A2 .68) can be approximated by the general second-order Runge-Kutta formula x(to + h) ~ x(to) + A. (3. Equation(3.A.5 Fourth-Order Runge-Kutta The symmetrical fourth-order Runge-Kuttaformulae probably are in most commonuse.. Infinitely manychoices exist.68). A second Taylor’s series expansion.75c) z2 = hf[x(t0) + ~Zl. = #2 = 1 (3. = hf(xo. this time for the rightmost term of equation (3.65) to (3. The error e in equations (3. (3.2z~ .A. (3.72b) whichgives precisely the result of equations (3. to + h].72a) It.’ (3.69) in which A.71a) (3..

The results should be insensitive to large changes in the error gage. which itself might require iterative numerical methods with uncertain success. or would eat up excessive ruu time. Basic fourth-order Runge-Kutta is the most practical algorithm presented. 3. Although it is not the most efficient scheme available.7. although an excessively large error gage will produce excessive time steps and resulting error. Somefurther discussion of the use of these programs is given in Section 5. It is especially useful for nonlinear models. Sometimes it is even used merely to find the equilibrium of a complex system in order to avoid having to. as in a mechanical impact. and the calculation repeated. where analytic solution is typically extremely difficult or impossible. A good integrator such as the fourth-order Runge-Kutta often uses very small time steps near the beginning of a simulation. the time interval halved. The MATLAB integration algorithms are largely hidden from the user. There is no point in this latter case for throwing away the previous calculation. Guided Problem 3. The problem is more acute for systems demonstrating sudden changes such as occur when a variable is abruptly limited. where the superscript T means transpose.3. MATLAB also offers other integrators with significant particular virtues. much longer time intervals can be used and the process usually is more efficient. The magnitude of the vector e or its square eTe. The simulators ode23 and ode45 are Runge-Kutta algorithms of between second and third-order and between fourth and fifth-order respectively. which subsequently increase by a large factor. A fixed-step integrator would either be relatively inaccurate at first. and allows comparison of simulated and analytical solutions.14 This first simulation problem for you to do is about-as simple as a nonlinear system can get. The algorithm ode45 is a hybrid fourthto-fifth order Runge Kutta. they are amongst several MATLAB integrators that are simple to use but not fully documented.6 Summary Computersimulation is an alternative to analytical solution of the differential equations which describe the behavior of dynamic models.2 and Appendix A. MATLAB also offers other integrators with special virtues. and possesses reasonable efficiency for a very broad range of systems and conditions. which are described in the "help" window. if eTe is less than some small preset fraction of e~. and some of the features of a third-order. . is a very useful gage for the expected error. the time interval for the succeeding ti~ne step can be doubled. The algorithm ode23 has some of the features of a second-order Runge-Kutta. Similarly. solve difficult algebraic equations. Thus if eTe is greater than some preset maximum e~o. it is simple and robust. which is merely more accurate than necessary. NUMERICAL SIMULATION 181 Thus. for equivalent accuracy to the second-order formulae. the entire calculation for the time step can be thrown away.7.

SIMPLE DYNAMIC MODELS A water tank has an area of 6 ft 2 and an initial depth of 2 ft. Since you do not knowa priori how long this takes. The hub is vibrated vertically with the applied sinusoidal force. noting the nonlinear resistance. and to compare the result to . You may neglect the masses of the hub and the disk in your analysis. Write the function m-file. F = kx + cx k --. 8o Investigate errors that result from choosing a large error index. The resulting behavior is surprisingly exotic. c = 30.42.0 pounds. which can be done in as few as two lines of code. Solve the differential equation analytically. and the rim acts like a mass at the end of the spring. The rim weighs 3. choose a short time to start with. as shown.0 in 2. A vibration absorber comprises a flexible disk with a fused rigid hub and fused outer ring. 3. This is a rare example of a nonlinear equation that is simple enough to be solved: the method of separation of variables applies.182 CHAPTER 3. Comparethe result to your simulation.3001b/in.001. A drain orifice has an effective area of 1. You could change the index by factors of ten. The disk acts like a spring. as pictured in Fig. The force-deflection characteristic of the disk is nonlinear: 3. Model it assuming Bernoulli’s equation.an analytical solution.15 This problem gives you practice simulating a nonlinear second-order model. without having to d(~ muchcoding. Guided Problem 3. Plot your result. o Annotate the graph in the usual way. 3. Find a differential equation relating the volumeof water left in the tank during draining to time. using various error indices. or flow coefficient. The "effective" area is meant to include the effect of the vena contracta. The relation between the orifice flow and the tank pressure is nonlinear. You are asked to simulate the emptying of the tank. Note that the default index for ode23 is . and for ode45 is le-6 (see MATLAB help. and extend it with subsequent runs. 0001b/in . Suggested Steps: Model the system with a bond graph. What does your model say if the volume were to become negative? . Simulate the system from the given initial volume until it is empty. "ODESET").

Next. and 3.15 The linear coefficient k results from the bending stiffness of the disk. Suggested Steps: 1. Add the damping term. repeat the last case.059 inches.02 lb s/in. The nonlinear coefficient c results from radial or membranestresses set up in the disk because the rather rigid rim resists being pulled inward as the disk bends. starting from rest and proceeding for 0. manner.3. Finally. Add a damping term to the force-balance equation with a damping coefficient b = 0. Commenton your observations. repeat the simulation for an amplitude ten times higher and a frequency 15% higher~ for two seconds or longer. with rim position nqutral). . Write the differential equations. Model the system with a bond graph. Plot the position of the rim as a function of time in each case. Then.0001 inch at the calculated natural frequency. except for introduction of the initial values h = x -. It is necessary to introduce the rim position as a third state variable simply to have it available for plotting. you are asked first to simulate its behavior. increasing the 0. Note particularly that while the force on the disk spring and the rim is common. Apply causal strokes to the bond graph in the standard annotate the bonds according to the standard procedure.0.058 inch (compression.7.058 inches to 0.42: Disk-and-rim vibration absorber for Guided Problem 3.5 seconds. carry out a third simulation that is the same as the second. omitting the damping. the absolute deflection of the rim is not equal to the relative deflection of the disk spring. The effect of non-zero initial conditions does not decay to reveal the forced part of the behavior unless some damping is introduced. 2. for a small hub displacement equal to hosinw~t with the small amplitude ho = 0. 183 NUMERICAL SIMULATION top view section view ~induced vertical motion Figure 3. To explore a little of the behavior of the disk.

Increase the amplitude and the frequency as indicated. (a) Approximatethe characteristics given in Fig. Use integrator ode45with the default error indices. the cubic term with the coefficient c in the force characteristic can be neglected. 6. .m2 and a linear resistance of 0.39 by simple polynomial expressions.3 (pp. and carry out the second simulation for two.184 CHAPTER 3. It weighs 3500lbs. which is virtually unaffected by the damping:wn = 1/v~-~.Doesyour result seemto agree with this description? 7. SIMPLE DYNAMIC MODELS 4. The response of a stable linear model approachesthe samesteady-state oscillation regardless of the initial condition. The load comprises an inertia of 2.2 (pp. however.0001 inch. but specify the rather large initial deflections indicated. 26-27). Repeatstep 8 with the almost infinitesimally larger indicated initial conditions.59 Consider the vehicle of Section 2. Plot the result.0 kg. The results should be drastically different. Calculate the resulting resonant frequency. except for the effect of damping. 74-77) accelerating on level groundfroman initial velocity of 20 ft/s until equilibrium is reached. Runthe simulation with zero initial conditions for the specified 0.60 Carry out the same steps as in the previous problemfor the acceleration from rest of the motor of GuidedProblem2. For very small deflections. three or four seconds. and the displacement associated with the excitation velocity at 0. 3.5 seconds. Set the fre5. Codethe differential equations in a MATLAB quency at the frequency computedin step 4. 2. The amplitude of the oscillation Of a linear oscillator excited with a sinusoidal signal at its natural frequency growslinearly in time. (b) Write the differential equation for the speed: (c) Carry out a numerical simulation. Plot the resulting deflection as a function of time. Doesit in this case? 9. Can you explain this? PROBLEMS 3. Repeatthe secondsimulation. function file. The amplitude of motion for a sinusoidal signal at any frequency other than the natural or resonant frequency should ultimately reach an equilibrium value. Plot the result.5. Doesit? 8.m-s.005 N.

Y) A0=1/144. and (optional) compare the result to the analytical solution.62 The differential equation for the conical tank of Example 3. 3. Enter into the commandWindow: [t. can be transformed to the tbrm dV~/dt ’ = -(W) 1/6 by defining a nondimensional "volume" V~ = V/Vo. [0 100]. 131) and 3. and compare the result with the analytic solution (which is readily found despite the nonlinearity). NUMERICAL SIMULATION 185 3. 158). and an appropriate nondimensional "time" t ~. f=-A0*sqrt(2*V*g/h).where Ao = i ~ .64 Simulate the response of the fluid tank system of Problems 3.6.3 (p. PROBLEMS -~ c-~-----1 .VJ =ode23(’ tank’. C WC I Ao 2V~gy. g=32. 3. ~. 3. with zero input flow.8 (p. Q = Ao ~/-~. Simulate the emptying of the tank numerically. 12) plot(t.61 Carry out the same steps as in the previous two problems for the acceleration of the boat of Guided Problem 2.-~= (Note that the density p does not affect the result. function f = tank(t. The plot for 350 seconds (rather than the 100 seconds above) is as follows: .12 (p. 165). 52) from an initial velocity of ft/s.V) The result shows that the final time needs to be extended.~ ~ 2.25 (p. 144) for the initial conditions specified in Problem 3.~---~---~¢ ~=.14 Problem I.) 4. 157) to the step change in the excitation current.63 Simulate the response of the electric curcuit of Guided Problems 3. 5. where ~b is the initial fluid volume in the tank. Compare result with the analytic solution which was found because the model is linear.7.2. 127) and 3.44 (p.36 (p.22 (p. A=6.3. 3. Assume50%throttle and a total effective weight of 3000 lbs.R A dV -Ao 2V _Ao%/2~g c=~. SOLUTIONS TO GUIDED Guided 3.in ~.

= . SIMPLE DYNAMIC MODELS 12 ~ V. such as at the instant of a mechanical impact.Ao dt 2~lWo = -ao~C-~tl’o 2(v/V. The plot of this equation is indistinguishable from that above found using ode23. the behavior of high-order models is apt to be more sensitive to the eror index. the square root of the negative value is treated as zero.5].003284t) The tank empties at t = 1/. Normally. V=12*(1-0. seconds If the volumeis started negative. Note the array power (the dot before the ") plot(t.ft 8 0 0 100 200 300 t. .’2.0. or becomesslightly negative due to numerical errors. This result applies to first-order modelsonly. MATLAB treats the square root of a negative number as an imaginary number. on the other hand.2 2x12x e12(1 .1 32. The equation can be plotted with the MATLAB instructions t=[0:0.V) Increasing the error index by a factor of 1000produces a negligible difference. The most commoncause of a significant error is an abrupt changein a forcing function.186 CHAPTER 3.v’-~o) -Aow~t 2 A g = 12 1 .003284 seconds.5:304.003284*t).

NUMERICAL Guided 187 SIMULATION Problem 3.02. f(2) = ~(4)-f(3). --dr - d =-~(xosinwt)= xowcos~t dy 1 cl-~ = ~P dx ~ dt ~ ~0~ COS~t ~.2)) The resulting response: plot (with some notational refinements) indeed shows a growing . I = 3/386.47.~k "[ compressionX~__ q =x-y C kq+cq~ 0 ~---~- I s~ dp -~ = kq + cq 3 + damping force = kq + cq 3 + do dx ~c=--d~ dq _ & _ 1 ~p. 6. x0 = 0.5]. ~(4) = x0*om*cos(om*t).4).[0 0.x]=ode45(’disk’.15 Y*-.t.x(:. x]’.7. Then.3. f(3) = x(1)/I. Let the state vector be x = [p. om = 196. c = 30000.x(:. Enter into the command window: [t.47rad/s 5. q. = 1/x/7-C= x/gT-J7= x/K~7~ = x/a00x as6/a= 196. b = 0. f=funct±on d±sk(t. y.[0 0 0 0]). f(1) = k*x(2)+c*x(2)’3+b*f(2).0001.x) k = 300. plot(t.

188

CHAPTER

3.

SIMPLE

D}%rAMIC

MODELS

0.004
inches
0.002

0

-0.002

-0.0040

0. l

0.2

~
0.3
0.4
t, seconds

0.5

Changing xo to 0.001 in, increasing w by the factor 1.15 and increasing the
final time to 4 seconds gives the following result. The ratio of the arnplitudes
of the rim to the hub approaches a constant value of 4.0.
0.008
solid line: rim, y(t)

inches
0.004

0

1

2

3
t, seconds

4

Introducing the initial condition [0 0. 058 0 0. 058] gives the following result.
Although the amplitude starts large, it ultimately decays to the same =t=0.004
inches as when zero initial conditions are specified.
0.08
inches

rim motion for hub h(t)= 0.001 sin(1.15%t) and
initial compressiondisplacementof 0.058 inches

0.~

-0.04

-0.08

0

1

2

3
t, seconds

4

3.7.

189

NUMERICAL SIMULATION

Increasing the number 0.058 inches slightly, to 0.059 inches, causes a major
change in behavior: the rim amplitude does not decay, but approaches a value
about 17 times larger than before.
rim motion for hub h(t)= 0.001 sin(1.15w, t) and
initial compressiondisplacememof 0.059 inches
0.08
inches
0.04

-0.04

-0.08

0

1

2

t, seconds

3

4

There are in fact two grosssly different resulting riln amplitudes for the same
hub amplitude, depending on the initial conditions. Should the rim motion be
momentarily damped, for example, the large amplitude would vanish leaving
only the smaller motion, a phenomenon known as the jump effect. The short
explanation is that increasing amplitude means increasing stiffness,
due to the
nonlinear force term with the coefficient c. Increasing stiffness in turn means
increasing natural frequency.

Therefore, if the amplitude is large enough
at first,
the natural frequency will rise
to match the applied
frequency,
15%
larger than the small-amplitude natural frequency, and the disk resonates with a rimto-hub amplitude ratio of about 67. The behavior depends on the history of the state,
or demonstrates hysteresis,
as the resonance diagram opposite suggests.

resonant
amr;l~mde]

~[~
0

frequ_~
~ ’,jumps and

~eresis
1 w/w.

Chapter 4

Intermediate

Modeling

This chapter represents a second pass at the modeling of dynamic systems.
The physical systems considered require bond graph models with more than
the single junction assumed in the previous chapter. Models for the simple
circuits considered in Section 4.1 below employ junction structures comprising
any number of bonds and junctions. Transformers and gyrators are added in
Section 4.2, with special attention on hydraulic-mechanical, electromechanical
and mechanism-based systems.
Models often can be simplified or transformed to make them more amenable
to analysis, without losing meaning. The most commonmodel equivalences that
accomplishthis objective are presented in Section 4.3. This leads to a consideration of the equilibrium state of dynamic models in Section 4.4. Stability,
instability and limit cycle behavior are critical properties that emerge.
The translation of the bond graphs to differential equations, and the finding
of numerical or analytic solutions for imposed initial or boundary conditions,
are deferred until Chapters 5 and 6.

4.1

Simple Circuits

Circuits comprise components, each of which has two sides, interconnected at
simple junctions. Either the generalized forces or the generalized velocities are
commonat each junction. The structure of a curcuit is represented by its
pattern of interconnections, that is its topology. The geometric orientations of
the components, on the other hand, either are i~nmaterial or are constrained
in a simple universal way. Electric, fluid and mechanical circuits are addressed
below. Someof the systems considered in Section 4.2 are circuits, also.
191

192
4.1.1

CHAPTER 4.
Simple

Electric

INTERMEDIATE MODELING

Circuits

Simpleelectric circuits compriseinterconnectionsof resistors, capacitors, inductors, voltage and current sources. Three-port and higher ported elements such
as transistors are excluded, and transformers are deferred to the following section. The junction structures.of the bondgraphs are independent of the nature
of its .one-port elements. As a result, one-port elements sometimesare designated in this bookwith the generic symbolZ, in both the circuit diagrams and
the bond graphs.
The following routine procedure can be employedfor simple circuits without
transformers:
1. Represent each node by a zero-junction, since the voltage is commonto
all joined conductors. It is helpful to label the junction with a symbolfor
its voltage.
2. Represent each branch i by the combination

whereZ~ represents the ith element (source, resistor, capacitor or inductor). This says that the current entering the element on one side emerges
unchangedon the other side. It is helpful to label the junction with a
symbolfor this current.
Discard all bondsfor whicheither e = 0 (ground or reference voltages)
i -- 0, since they represent zero power. (An exception is the rare case in
which capacitors prevent all of the conductors from having any current,
and you still are interested in the distribution of voltages.)
Eliminate all junctions from which only two bonds emanate, and join the
bonds. Coalesce all directly bonded zero-junctions and directly bonded
one-junctionsinto single junctions of the respective types.
Steps 3 and 4 do not changethe meaningof the bondgraph, but serve to simplify
its appearance.
Recall that the electrical and bond-graphsymbols for resistance and for
capacitance are R and C, respectively, whereasthe electrical symbolfor inductance is commonlyL and the bond-graph symbol is I. Do not write L in the
bondgraph;it is undefined. Instead, write I -- L off to one side.

4.1.

193

SIMPLF~CIRCUITS
EXAMPLE 4.1
Applythe four steps aboveto get a bondgraph for the electric circuit

Solution: The bond graph after step 2:

After step 3:

and finally after step 4:

After somepractice it is possible to draw the final bond graph directly,
without separating the four steps.
EXAMPLE 4.2
Drawa bond graph for the Wheatstonebridge with arbitrary elements Z~.
The bond graph elments also may be labeled Z~ (as word bond graph elements). This circuit could comprise any combinationof types of elements,
althoughin usual practice the four legs are of the sametype.
e2

e’~~[~ ~]~6.

,~

194

CHAPTER 4.

INTER_MEDIATE

MODELING

Solution: The bond graph after step 2 is

eb
Z’\l./ \ 1./z2
e~o ~ 1~--~------1
Z4~ i~,~x

0

Zs/1/S~Zs

and after step 3 i~ is
Z~

O~

Z~

z

x

Z~

1i~0

¯

Z~

Finally after step 4 it is

z~

You are urged to try this problem using a different node as the reference
or ground; although the benzene-ring-like form of the resulting bond graph
will remain, the order of the elements will change. This change corresponds
to the different choices of variables, and does not imply different physical
behavior.

4.1.2 Fluid Circuits
Fluid circuits include gravity tanks, compression chambers and fluid restrictions
such as orifices or valyes interconnected by pipes or channels that may exhibit
significant resistance and/or inertance. A very large gravity tank might be
approximated as a constant-pressure source, called a reservoir or surap.

4.1.

195

SIMPLE CIRCUITS

Themodelingsteps for fluid circuits are analogousto those for electric circuits. Afluid restriction is modeledlike an electrical resistor, with an R element
bondedto a 1-junction, since the input and output flows are the same. A junction of three pipes (a pipe "tee") or moreis represented by a 0-junction, like
electric circuit junction, since the pressures are common
and the flows sum.
EXAMPLE 4.3
Modelthe hybrid mechanical/fluid sytem pictured below in the form of a
bond graph:

r~-~

.~

~

~

surge tank
area A
~~~--~toload,

¢

P,

i .--.-.~-

area)

Solution: The gravity tank is analogous to a shunt capacitor in an electrical circuit. (Series fluid compliances cannot be madewithout energy
transduction to mechanicalform, whichyou will see later.) The pressure in
the reservoir is considered in the modelbelowto be negligible comparedto
other pressures.
C
S ~ T ~ lrl---.--PO
O[

R

~ ln---2A,

- LOAD

T = displacementper radian
since S~ = O, its bondc~ be dropped

4.1.3

Mechanical

Circuits

Anassemblageof masses or inertances, springs and dashpots that moveeither
in simple translation or simple rotation is called a mechanicalcircuit. A translational exampleis given in Example4.4 below. Twoor ~nore componentsjoin
at a "mechanical node," where they have a common
linear or angular velocity.
This is different froman electrical node, at whichthe efforts (voltages) rather
than the flows or generalized velocities (currents) are common.The two sides
of a spring or dashpot have the sameforce but different velocities, whichalso
is unlike electrical capacitors or resistors that have common
flows rather than
efforts. Onthe other hand, the two sides of a masshave the samevelocities but
different forces, like the two sides of an inductor that have the samecurrents

196

CHAPTER 4.

INTERMEDIATE MODELING

1but different voltages.
Steps similar to those for electric circuits are as follows:
1. Representeach mechanicaljunction with a 1-junction; if inertia is associated with the generalized velocity of a junction, add its inertance:

I
It is usually desirable to label the linear or angular velocity with a customized
symbol.
2. Represent each spring and each dashpot, respectively, with a 0-junction
bonded to the C Or R element:
~0~

0

Again, it is usually desirable to label the common
force or torque with
particularized symbol.
3. Coalesce bonded junctions of the same type into a commonjunction.
4. Representeach input by S, S)~ or S~, as appropriate.
5. Delete any bondsfor whicheither the generalized force or velocity is zero.
6. Eliminate all junctions from which only two bonds emanate, and join the
bonds.
The exampleproduces a loop of bonds, called a mesh. This and similar fourbond meshes can be reduced to tree-like bond graphs, as is shownin Section
4.3.
1Somemodelers use the Firestone analogy between mechanical and electrical variables in
order to make the modeling diagrams of similar assembledges of electrical and mechanical
components look alike. In this analogy, velocity and voltage are knownas "potential" or
"across" variables, and force and current are knownas "flow" or "through" variables. Inertial
elements are treated as having zero velocity on one side. Fluid flow is treated like a current
or a force, and pressure is treated like a voltage or a velocity. This analogy turns a hydraulic
cylinder into a gyrator, and a DCmotor into a transformer. "Line graphs" (originally known
as "linear graphs") represent the models. These graphs are more complicated than bond
graphs, particularly when transformers or gyrators are employed. This author considers the
identification of force as a fiow and velocity as a potential to be unnatural, and prefers to define
variables by their nature at a single cut in space rather than with respect to a circuit element"
with its two sides. Furthermore, the effort-flow concept which emerges from the single-cut
idea extends better to thermodynamic systems and distributed-parameter models.

4.1.

197

SIMPLE CIRCUITS

EXAMPLE 4.4
Translate the system drawn below in the language of mechanical circuits
into the language of bond graphs:
R!

F~

m,

XI

F2

m.

~’
.C

X2

X3

Solution: After step 2:
Rl

x3 l x4

After step 3:

Steps 4, 5 and 6 do not apply to the model as pictured. If the right end
were fixed to ground, however, the bond with i4 would be deleted (step
5), leaving a two-port 0-junction that could be eliminated by bonding C2
directly to the 1-junction for i3 (step 6).
4.1.4

Summary

Simple electric, fluid and mechanical circuits comprise assemblies of simple components each having two ports with a commoneffort or a commonflow; The
components thus can be modeled with a bond graph having two boundary bonds,
a 0-junction or a 1-junction and an R, C or I element. Interconnections also
occur at 0 or 1-junctions. It is strongly recommendedthat you define symbols
for all commonvariables, and place these symbols on the system sketch first and

198

CHAPTER 4.

INTERMEDIATE
L

k!

R~

MODELING
R
2

~

(a) mechanical circuit
e(t~~

P~
(b) fluid circuit
h!

accumulator

tank

(c) electric circuit
reservoir i

Figure 4.1: Guided Problem 4.1
the corresponding bond graph junctions second. Bonds with either zero effort
or zero flow, and therefore zero power, may be erased from the graph.
Guided

Problem

4.1

The three circuit diagrams in Fig. 4.1 imply unique models. Translate the
language of these diagrams into the language of bond graphs. It is critical that
you attempt these simple cases before you examine the solutions.
Suggested

Steps:

1. Define symbols for whichever variable is commonfor the two ports of each
element, and place them on the diagrams.
2. Carry out the steps for electric circuits (which apply by analogy to fluid
circuits also) or mechanical circuits, as given in this section. Label each
junction with the proper symbols from step 1. Do not attempt more than
a crude .modeling of the accumulator.

Guided

Problem

4.2

Part (a) and step 1 of this problem are a simple example of the use of junctions.
The balance of the problem gives insight into some of the virtues and liabilities
of major conventional hydraulic circuits.
Hydraulic motors often are driven from simple fixed-displacement
pumps
run at constant speed so they produce virtually constant flow. See Fig. 4.2,
which employs~ standard fluid power symbols. The speed of the motor usually

4.1.

199

SIMPLE CIRCUITS

bleed-off:
pump

control valve

~

valve
~co:~’ntrol-’~~load

meter-in:
~

pump

relief valve

pump

relief valve

motor

motor

meter-out:

Figure 4.2: Hydraulic system of GuidedProblem4.2

lve

200

CHAPTER

4.

INTERMEDIATE

MODELING

is controlled by an adjustable valve in one of three ways. In bleed-off control,
excess pump flow is diverted through the valve. In meter-in control, the valve
admits only the desired load flow, while the excess, if any is diverted through
a relief valve. The relief valve is constructed so as to limit the pressure of
the pump, /~p. Meter-out control is similar, but the control valve is placed
downstream of the hydraulic motor. The small symbol at the bottom of each
vertical line represents the vented reservoir or tank, which is common.(This is
like the ground symbol in electrical circuits.)
(a) Draw bond graphs to model the systems, neglecting leakage and friction and assuming constant pump flow. The terms "RELIEF VALVE,"
"CONTROLVALVE" and "MOTOR/LOAD"may be used for the relevent
components.
(b) Find the overall efficiency of these systems when operation is at the
load-to-relief
pressure ratio (PL/PR) of 0.1 and 0.9, and the load-to-pump
flow ratio (QL/QP) is 0.2 and 0.8 (for a total of four combinations), neglecting leakage and friction.

Suggested

Steps:

Draw bond graphs for the systems, focusing on 0-junction
and 1-junction for series interconnections.

for pipe tees

Plot the pumpcharacteristics, as modified by the relief valve (if any),
a map of flow vs. pressure. Also, plot the four operating load points.
Find and plot on the map the corresponding
operates.

points at which the pump

Computethe et~iciencies ~ as the ratios of the motor power to the pump
power, power being the area of the rectangles defined by the operating
points and the origin.

PROBLEMS
4.1 Find the bond graph for the circuit of Example 4.1, moving the ground
from the lower end of the inductor to the uppper end. Is the structure of the
graph changed? Are the meanings of the effort and flow variables changed?
(Pay particular attention to signs.)

4.1.

SIMPLE CIRCUITS

201

4.2 Model the systems below with bond graphs. Identify variables and parameters where appropriate on copies of the diagrams and on the bond graphs.

(a)
porous plug

]’ tube

4.3 A hydraulic circuit includes a positive-displacement pump, rigid fluid lines,
a rotary hydraulic motor and a load with a rotational inertia. Give the structure
for a model of the system in bond-graph form. Describe in words what kinetic
energies are represented by the inertance(s) in your model.
4.4 Model the translational

mechanical system below with a bond graph.

F

4.5 Model the rotational

mechanical system below with a bond graph.

flywheel

4.6 Draw a bond graph for the circuit

flywheel

shown below.
L

coupling

202

CHAPTER 4.

INTERMEDIATE MODELING

4.7 Infections of the middleear frequently are stubborn and potentially serious,
particularly for small children. It is common
to insert a small tube surgically
through the eardrumto act as as a drain for the infectious material. (The tube
is rejected by the body several months later.) The direct effect of the tube
on the hearing, presumingthe tube and middle ear are free of liquid, is the
present concern. The mechanical loading of the eardrum by the tube appears
to be negligible; the question is the effect on the vibration of the eardrumby
the acoustic bypass hole through the tube.

Assumethat a pressure drop of 1 lb/ft 2 across the eardrum produces a
volumetric displacementof 0.5 x 10-6 ft 3. Also, assumethat the effect of the
small boneswhichtrans~nit the vibration is s!mplyto increase the effective mass
of the eardrum by 50%.Other parameters are listed below.

2( area = O. 10 in
eardrum~ thickness = 0.04 in.
3\ (,. density = 2.0 slugs/ft
~
~~~J4~~/

~bones

that

vibration
~’~er

to
ear

ddle
~’~-//////~~.~/eVaC~U__m~.~
\~,~

transmit
inner

ear

(neglect)
irnni3

~ area of hole i~.0 ~0~1 in
,~~ (diameter
~ 0.035
in.)
density of air = 2.4 x 103 3slug/ft

Describethe physical significance of each of the elementsin the simple bondgraph modelshownbelow. Relate the moduli of its elements to the parameters
of the physical system, and evaluate those parameters. The motion of the upper
1-junction is transmitted to the inner ear by an "activated" transformer which
you need not consider. (Problem 6.21 (p. 447) requests the response of
systemto sinusoidal excitations, both with and without the drain tube.)

4.1.

203

SIMPLE CIRCUITS

.1

~

~ T ~ to inner

ear

o ---.-c,,

o
1

SOLUTIONS

Guided

TO

Problem

GUIDED

PROBLEMS

4.1

(a) mechanical circuit
1.

~x2

[.~x~,

forcesin springs:Fk,, Fl~2
force in dashpot:Fb

~
2.

after step2:

l,

~
i

~
R

C~

12

C
2
(note: further reductionis
possible usingFig. 4.18)

I~

~

12

204

CHAPTER

4.

INTERMEDIATE

MODELING

(b) fluid circuit

11

CI

R~

(c) electric

I2

C,

Rs

R~

circuit

2. after step 2:1
L

~

~ R3

R,

I n°te: /=L I1; ~ 0 ~ li~"~

2.

Rf-~----

"1

s, e(.-~t) li

e,_

R,

li--~C

ls i-..~-- S

R,.
after step 3:

after step 4:
I

R~-.~----- ~

S~

li~

li~C
li~Sy

liaR3

i

R.,’
--R2+R
3

Se e(’~ t) 1,
2’ .~,--~-~-elo

--,~-~-R

R~ Sf~ li’---"~’C

4.2.

205

SYSTEM MODELS WITH IDEAL MACHINES

Guided

Problem

4.2

1. bleed-off:

Pe
a~

~ 0

PLQL
= -P¢ MOTOR/LOAD

CONTROLVALVE
meter-in and S
meter-out: ~

2,3.

Pe
Q~,

o
. 1^ Pt MOTOR~LOAD
L
RELIEF VALVE CONTROL
VALVE

_~. pump, bleed-off-~ .~ pump, meter-in
or meter-out
Q~’[~. pump characteristic
~

o

0.8o~-~,-~............ .:.,,;2
I t four combinations
I ’; .of~,o.L~, . 1 I’
00.1 PR

0.9 P~PR(relief valve setting)
P

Bleed-off: ~1 = ~12 = 0.8; ~/a = ~14 = 0.2. Meter-in or out: r/1 = 0.1 x 0.8 =
0.08; r~2 = 0.9 × 0.8 = 0.72; r/3 = 0.1 ×0.2 = 0.02; r/4 = 0.9 x0.2 = 0.18.
(Bleed-off control gives higher efficiency, but meter-in or meter-out control is
moredirect and less susceptible to errors in the pumpflow, etc.)

4.2

System Models With Ideal

Machines

The models of Section 4.1 comprise one-port elements (S, Se, Sf, R, C, I) interconnected by junction structures of bonds, 0-junctions and 1-junctions.
Constant-parameter
transformers and gyrators (T,G) are now added to the
junction structures to permit modeling a much broader set of systems. Circuitlike systems are considered first. Certain commonlyoccuring two- and-threedimensional geometric constraints are then recognized.

4.2.1

Electric

Circuits

An idealized electric transformer is shown in Fig. 4.3. The modulus of the
transformer, T, equals the turns ratio of the two coils. The presence of four wires
but only two different currents is represented by the use of two 1-junctions. The
four-port model reduces to the two-port model if the voltage differences across
the terminai pairs are adopted as the effort variables. This simpler model also
applies directly if the voltages on the lower terminals are considered as ground.
The bonds with zero voltage carry no power, and may be excised from the graph.
This model implies that the transformer behaves the same at all frequencies,
including DC. In fact, a real transformer fails completely at DC, and exhibits

206

CHAPTER 4.

INTERMEDL4TE MODELING

eI =
2 Te

i2 = Til
(a) circuit representation
1.-~-~- T -------r

1/2~

or

el
.-~-T
II

.~I1

- e2
.---~-12

(b) bond graph representation
Figure 4.3: Ideal electrical

transformer

deteriorated
performance for very rapid changes or high frequencies. These
limitations are explored in Section 10.4; for the present they are neglected.
Step 2 of the procedure for electric circuits given in the preceding section
is expanded to include placement of the four-ported combination of the transformers and 1-junctions.
EXAMPLE 4.5
Model the following circuit

with a bond graph:

Solution: The bond graph after

step 2 is

4.2.

207

SYSTEM MODELS WITH IDEAL MACHINES
The final graph with the grounds accounted for reduces to

4.2.2

Fluid/Mechanical Circuits with Positive
Displacement Machines

The piston-and-cylinder idealized in Example2.3 (p. 63) has a potentially serious practical problem:if the piston fits too tightly, friction is apt to be severe;
if it fits too loosely, leakage is boundto result. Further, high pressure expands
the cylinder walls, opening or wideningthe leakage path. Someleal~ge across
the piston usually is tolerated, becauseit is "internal," producingno puddles on
the floor. Fluid is ported to both sides of the piston. The symmetricdoublerod variation shownin Fig. 4.4, can be represented by a combination of two
1-junctions (to equate the inlet and outlet flows and the motions of the two
rod ends) and an ideal transformer, muchlike the electical transformer. The
modulusof this transformer is the annulus area of the cylinder. If the power
conventions arrows across T are reversed, the modulusbecomesthe reciprocal
of the annulus area.

...........

P~|Q

T

T=A

P~tQ

Figure 4.4: Double-rod-endcylinder as an ideal machine

208

CHAPTER 4.

INTERMEDIATE MODELING

EXAMPLE 4.6
Three symmetricdouble-rod cylinders are placed in series, as shown. Draw
a bondgraphfor this circuit.
Pl

three cylindersin series

3

Solution: Since all three cylinders have the same flow rate Q, their fluid
bonds should emanate from a common-flow1-junction. Also, no mechanical
forces are placed on the left sides of any of the piston rods, eliminating the
need for any 1-junctions for the mechanicalbonds.

r,
PI

~

T 2 r~

2f

EXAMPLE 4.7
The single-rod cylinder is cheaper and more commonthan the double-rod
cylinder. The two ports of this cylinder sometimesare interconnected, as
shown, to give a regenerative circuit with a single external fluid port.
Modelthis system with a bond graph, neglecting friction and leakage.

2.as represented by the 0-juction. as shown in Fig. This result can be deduced directly. F ~ ~.4. as shown in part (c). Finally. which means that the single modulus T on the reduced bond graph. The flows Qa and Qc are different in magnitude (as well as direction). one having its modulusequal to the total area of the piston. while the other modulus equals minus the area of the annulus.2. The now familiar structure has two 1-junctions and an ideal transformer. because the areas of the two faces are different. the flows sum and the pressure is common. etc. Therefore.5 part (a).1 -----~ T¢ [’ F ~’P oo 10_~ 0’ - P T The regenerative circuit also can be viewed as an integrated unit with a single fluid port and a single mechanical port.5: Ports of idealized (b) ’/ (c) model of a hydraulic motor/pump . like the electric transformer and the doublerod-end cylinder. 4. the model reduces to a two-port ideal transformer. the two flows are joined at a pipe tee. which is the area of the shaft. If one of the pressures and one of the torques are zero. The flows Q~ and Qd sum to give Q. If the shaft does not penetrate through the machine. The sum of the volumes of the fluid and the rod within a fixed control volume drawn around the unit is a constant. Mechanical friction and fluid friction (pressure drops due to orifice flow. Two transformers represent these geometric relations. A rotary hydraulic motor/pump with a shaft that emerges at both ends can be modeled as a four-port machine.3 Losses in Positive Displacement Machines* The models of positive displacement machines presented thus far neglect the effects of friction and leakage. 4. the three-port version of part (b) of the figure applies. SYSTEM MODELS WITH IDEAL 209 MACHINES Solution: The two faces of the cylinder have the same velocity. and vice-versa. equals the sum Tc + T~.) can be modeled by bonding the resistances Re (a) Figure 4. every net cubic inch of fluid pushed into the cylinder pushes out one cubic inch of rod. drawn on the right side above. however. which is recognized in the bond graph below by a 1-junction.

labeled QD. although this is not included in the bond graph or the following analysis. Twotypes of leakage flows are represented in i)art (b) of the figure. Qea Qe2 represent external leakage. the 0-junctions indicate the literal diversion of fluid. R~2 and Ri can be included or excluded from the complete model. Sometimesa viscous resistance is also significant enough to be considered. INTERMEDIATE MODELING (b) leakage Figure 4. The brushes of a conventional machine impose a nearly constant frictional torque on the shaft. Re~. Note that the frictional resistances R~ and RQ can be removed if their moduli are sufficiently small. given as R in the model of Fig. 4. almost always is important. Any of the five resistances R~. The displacement flow of the machine.4 Losses With DC Motor/Generators* The resistance of the armature circuit of a DC motor/generator. The internal leakage Qi occurs as backflow across the piston or the vanes or the gear teeth. is the theoretical flow which would exist were there no leakage. whereas the leakage resistances R~. As with the other . as shown in Fig. 4. etc. depending on the magnitudes of the phenomena and how accurate you wish to be. The resistance to this leakage is Ri. RQ.6: Models for friction chanical machines and leakage in positive displacement fluid/me- and RQ to the respective terminal 1-junctions. Note the familiar four-port structure of the bond graph.2. 4. from the high pressure side of the machine to the low pressure side. the bearings contribute some additional torque. The leakage resistances are Rea and R~2.7. it can be appended readily as an R element. which either drips on the ground or floor or preferably is carried back to the reservoir by a special drain line at essentially zero (gage) pressure.210 CHAPTER 4. Re2 and Ri can be removed if their moduli are sufficiently large. The bond graph of part (b) of the figure therefore shows an effort-source element Se with its constant torque labeled as Mo.6 part (a).

6 0.5 G(~/e (c) typical torque-speed characteristic 0 0.4 0.7: Modelof electrical/mechanical machinewith losses 211 .0 ~o / Ge (d) maximum powersand efficiencies Figure 4.8 1.0 0.2 0.4.~ ~/]J R’-~----- 1 e-Re K --schematic e2[i:(M+M°)/~G e~ (a) 1 RMo Ge 0 1 0 M2l¢: (b) bond graph (e-Ri)/G 1.2. SYSTEM MODELSWITH IDEAL IVlACHINES M1 l~ M 0 (friction) el .

giving the nondimensionalized torque-speed relationship RM -l-Q---. R. respectively. . Mo and G. You should convert these to SI units to take advantage of the fact that one Newton-meter/second equals one Watt and one volt-amp. a~m~ v _ 1 . This information is sufficient for deducing the values of R. (4. the no-load condition. Specifications for permanent magnet DC motors ideally include values for e. the torque. This equation plots as a straight line in part (c) of the figure. speed and current are specified instead at two (or more) conditions..4~) (These results follow from setting the derivative of an expression for the efficiency equal to zero. perhaps including stall. is 1 (1 . Complete specifications include a coefficient of proportionality between the power being dissipated and temperature. = (4. (4.Ri)/G.212 CHAPTER 4. INTERMEDIATE MODELING machines. Q= RMo/Ge. defined as M~/ei.~.Q)/2. M0and G.Q)2 (1 + Q)-I ~max The maximumenergy efficiency lower power: ~ ~ ~ is higher.2) of this operating point. so that 2e P. e (4. Oftentimes. or a "rated" condition.3) ¯ and occurs at a higher speed and .4b) ~. The net applied voltage and torque are taken below as e and M. e. Forced air convection increases the range of allowable operation. for which RM/Ge = ~G/e = (1. As with other such straightline source characteristics. they should then give the same number.Q)2 ~-~’ The energy efficiency (4. These two relations can be combined to eliminate the current i. or ~ a "back EMFconstant" with units of volts/rpm. The bond graph reveals that i = (M + Mo)/G and ~ = (e .1) in which the parameter Q characterizes the quality of the machine.) The m~imumpowers and efficiencies are plotted in part (c) of the figure as functions of the quality index Most electric motors will overheat if run continuously at high torque and low speed. all at some voltage. maximumoutput power occurs at its mid-point. one or two of these ports may be removed if its effort is zero. Ge G~.na~ = (1 . The modulus G is typically described as a "torque constant" or "torque sensitivity" with the units oz-in/amp.

Sometimes this is easier. through a viscous coupler.0. by the intersection of the charactertistics R and S": Mm-= 0.67 ohms as plotted in part (c). A bond graph model is shown in part (b) of the figure. and is transcribed into part (e). 4. The result at this point is a source characteristic S’ bonded through a gyrator to a load characteristic R. The frictional torque M0 then is added to this characteristic. To find the equiSbrium. while the angular velocities (flows) sum. This characteristic is found by a horizontal addition of the characterisitics R2 and R3.4. armature resistance of R1 -.2. A different approach deals directly with velocities rather than the displacements.1. 4. One approach develops direct relations between the displacements of the inputs and the outputs of the mechanisms. I t i s f ound by subtracting the effort for R1 from the effort for S.e2 and e~. It is not the purpose of this book to present a course in kinematics. ~. the combination S~ ~ G ~ is reduced to S" ~. but certain key ideas are reviewed.8 is driven by an electrical source with characteristic plotted as S in part (c). The relations between velocities are found by differentiating the resulting equations with respect to time.and-three-dimensional geometric constraints is not shared by electrical or other circuit-like systems.2. the source S and the armature resistance R~ are combined in part (c) of the figure to give an equivalent source. particularly when there are multiple independent ve- .6 Two- and-Three-Dimensional Geometric Constraints The complexity of two. Next. one for its output power and one for the the difference or dissipated power. The torque (effort) is commonto the three bonds. The viscous coupler has three bonds. = R~i. 4. The viscous coupler has a slip velocity proportional to the torque.16 N m) and q~ = 67 rad/s (which gives ~3 = 60 rad/s). one for its input power. the bonds are joined by a 0-junction. Therefore. vertically because of the 1-junction. since ~1 = ~2 + ~3 and ~ = M/R2. since e3 -~ el . and friction torque M0= 0.M = 0.5 ¯ SYSTEM MODELS WITH IDEAL Case Study With Source MACHINES and 213 Load* The DCmotor of Fig.mplemented in part (e) of the figure.) -. as shown by the dashed arrows from part (c) to part (e). the load characteristic R3 and the viscous coupler characteristic R2 are combined to give the lower dashed line plotted in part (d) of the figure.2. It drives a mechanical load.24 N m (which gives .025 N s/rad as plotted in part (d).08 N m. The motor is characterized by torque constant of G = 0. Next. finally. and is one of the fascinations of mechanical engineering. The equilibrium operating state is given. as indicated and i.15 N m/amp. with characteristic plotted as R3 in part (d). with coefficient R. to give the higher dashed characteristic labelled R. This represents the overall load. The definition of the gyrator (e2 = G¢~ and M = Gi) is used to establish the end points (or any other points) of the S" characteristic.

~o~ 0 ~ ~" 40 80 120 ~.~o°upler.s" ~ 4 // ~..’~" 0 ~. R~. ¯ S R s~ ~’~=~_ I I ~I ~..-.¢~e..’~- INTERMEDL4TE MODELING t.. 160 0 ".. 2 -. raa/s (e) ~rther reduction S"andsynthesis Figure 4.4 .. raa/s (d) given R~. load M e3 G’~’. .Mm 1.m~ 0 4~ "’.2 (e’s shownfor volts I~ ~ particular i) and reduced S" s ---~-G R 0 Mm ~ 0." S I ~ s ~ equilibrium . M0~d reduced R ] J 0 \\1 k~ N.R 3 Mo ~ e2 R1 Se R2 (c) given characterisitics S and RI 20~. ~ J ~ ~ or. .. 40 80 12( ~..8: Exarnple of DCmotor with a particular source and load . ~ps ~x .. . ~~ (a) system (b) bond graph S ~e~1.214 CHAPTER4.

9: Kinematical constraints for rigid bodies locities or displacements.9: The vector velocities related by for arbitrary points A and B on a rigid body are VB = VA + ~) × tAB.) 4. unlike the corresponding points for pinned members. VA = VB. Two instantaneously contacting points A and B which belong to separate membersin rolling contact also satisfy equation (4. (4.e. 215 SYSTEM MODELS WITH IDEAL MACHINES (b) pinned joint between two bodies (a) two points on a body parallel to surfaces at contact (c) rolling contact between two bodies (d) sliding contact between two bodies Figure 4. (4. i.6).5) where tAB is a geometric vector from point A to point B. (The accelerations of these two points are different. Point A on one member and point B on another member have the same velocities if both points are located coextensively at a pinned or swivel joint between the members. however. as illustrated in Fig.4. The approach is based largely on the following key kinematical relations. Two instantaneously contacting points A and B which belong to separate members in sliding contact have zero relative velocity in the direction . 4.6) 3. 2.2. and ~ is the angular velocity vector for the body.

The velocities of points along a vertical diameter of the pulley are parallel.8b) and solving for 2c.85) The constraints represented by equations. including their vector orientations. (4. 4.8a) and (4. and the angular velocity. the constraints interrelating the various 1-junction velocities are sought and represented by proper junction structures. That is.10. Next. and any dashpots.9a) r~= 1~(~.2A)- (4. by kA = 2C -. the first equation is represented by the left-hand 0-junction. its construction is the critical step. (4. are modeled by attached R. Finally. (4.216 CHAPTER 4. .7) The development of a bond graph model normally starts with identification of the critical velocity variables.). (VA-. Specifically. and equation (4.Pulley System Consider the pulley and parallel cables shown in Fig.Vt~) ¯ n = 0 (4. 1 kc ~(2A + 2. then.9b) These equations are represented directly by the bond graph in part (e) of the figure.8) are incorporated into the bond graph shownin part (c) of the figure. separate 1-junctions may represent its orthogonal components. . Specifically. if n is a vector normal to the surfaces of contact. 4. This chapter focuses only on geometric constraints that can be modeled with junctions and constantmoduli transformers. B and C are assuraed to be horizontal. The velocities ~A and kB are related to the velocity of the center of the pulley. An alternative perspective results from adding equations (4.r~. Study this graph carefully.8a) 2~B = ~C "~ r~.. If the orientation of a velocity is not constrained by the geometry of the system. where ~A and ~B are chosen arbitrarily.2. FB and Fc and the moment Mare added in part (d) of the figure. and subtracting one from the other and solving for r~b: . Points A and B are commonto both the cable and the pulley. The applied forces FA.7 Case Study: .More complex mechanisms are addressed in Chapter 9.. equation (4. ~c.8a) is represented by the upper 0-junction. etc. a 1-junction is introduced to represent the magnitude and sense of each such velocity. The procedhre may be understood best by examples. and the corresponding label is attached to the junction.8b) is represented by the lower junction. springs and inertias. (4. They form a trapezoidal vector diagram as shown in part (b) of the figure. These should first be noted directly on a drawing of the physical system. any inputs or outputs are represented by bonds or sources. C or I elements. The velocities of points A. ~. INTERMEDIATE MODELING normal to the surfaces in contact. In the bond graph.

\0~ XB T=l/r (c) bondgraphwith kinematicalconstraints (d) bondgraph with forces added "~’.4.10: Pulley systein . Xc TI ~ T~ = I/2 Tz= I/2r (e) alternateto (c) -T2 Ohx ~0 2r--’~ \I~ (f) alternateto (d) Figure 4. 217 SYSTEM MODELS WITH IDEAL MACHINES I (a) physical configuration (b) velocity diagram 2.2./A =2r~ k~X. x / re 0-~-~---.

INTERMEDIATE MODELING and the second by the right-hand 0-junction. 4. the resulting erasures from the bond graphs of parts (e) and (f) lead simpler interpretation.. This constraint ~Tc~O TA -- rOB tAB ’ TB rDA : ’FAB --. Which ought to be preferred in practice depends on additional circumstances. Model the system with a bond graph. since the kinematical relationships and the boundary powers are proper. The only substantive difference is the addition of a force applied at point D. The key geometric constraint regarding point D is ~gD : TArA + TBJ:B. EXAMPLE 4.10 is adapted below represent the entire system except for the presence of a force at’ point D. Both bond graph models must represent correct relationships between the forces and the moments.~/r~. A:XA ]~c : FA r~ I (small deflections assumed) T~ =ros/r~s Tc= 1/2 rs=r. if the moment is zero. as long as the lever remains nearly vertical. The forces and moments are added in part (f) of the figure. 73=1/r~ Solution: The bond graph of part (f) of Fig. is implemented by a 0-junction in the bond graph: ~O~x 0r~o¢ T~ ¢ .218 CHAPTER 4.8 The floating lever below represents a minor modification of the pulley system. For example.

~= T~= r~ (~s " Xs X. ~ 1. and the centers of the planet gears also move with the spider."-~t-.i-. ~s.-~. one of the three input shafts is held motionless at any moment in time. In usual practice. All these members can rotate about their centers. ~p These four equations are represented in the bond graph by four transformers. reducing the system to a net two-port transmission. 2~. note: The 1-junctions are written and labeled first to aid in drawing iCp T. s "~ = ~(z~ 1+" ~s).4. The planet gears share a comnlon velocity with the sun gear at their point of mesh.2.rp gears "~~~. a ring gear and a spider which contains the bearings for the planet gears. Model the syste~n with a bond graph. 219 SYSTEM MODELS WITH IDEAL MACHINES EXAMPLE 4. The last equation also requires a 0-junction to represent its summation of velocities. A~..9 An epicyclic gear train comprises a sun gear. and a commonvelocity with the ring gear at their point of mesh. The kinematical relations are i. . by some external means..-~ ring gear ~_’~"1~’~ ~spider ~onAA Solution: The key velocities are defined in the diagram below left.-. the graph. They may be removed later since they have only two ports. 3 planet~ ~\ ~ . three kinematically redundant planet gears. Tp = ll& T~= 1/2 T.

220

4.2.8

CHAPTER 4.

INTERMEDIATE

MODELING

Modeling Guidelines

There is no universal modeling procedure. Nevertheless,
as a useful guideline for most cases.

the steps below serve

1. Identify any input and output variables, define symbols for these variables,
and place these symbols appropriately on a sketch of the system being
modeled. If any of these variables are efforts, also define and label the
corresponding flow variables. Use arrows to show directionality.
2. Define and label on the sketch all other variables that appear to be important in relating the input to the output variables.
3. Identify whatever generalized geometric constraints exist between the various flow variables you have defined. This can be done in the form of
equations, and often involves defining parameters such as the lengths of
levers or the areas of pistons. This step may utilize knowledge gained in
your engineeering science courses or in some of the examples given in this
book.
4. Start to form a bond graph by drawing any boundary bonds for the input
variables. Then draw a 1-junction without bonds for each other flow
variable identified in step 2, and annotate it with your symbol for that
variable. These ~nay be placed on the sheet, in isolation, wherever you
imagine will leave enough room for the missing bonds and elements.
5. Implement the constraint equations of step 3 by placing appropriate junctions, bonds and transformers. Identify the meaning or values of any
transformer moduli. At this point you may have a completed junction
structure, that is an interconnection between all the flows that you have
defined and are needed to relate the input variables to the output variables.
6. If one or more of the key variables are not yet interconnected by bonds,
the system may possess one or more gyrational couplings, that is proportionalities between efforts and flows. Identify these, and represent them
by gyrators. Note their moduli.
7. Should there still be a key dangling variable or two, examine each of the
steps above to identify what is missing.
8. Identify the one-port effort and flow sources, resistances, compliances and
inertances that you believe are important. Represent each by a bond graph
element Se, SI, R, C or I, using distinguishing subscripts if necessary, and
place these where they belong on the bond graph. This step and the next
may draw heavily from your knowledge of engineeering science. Whether
to include a possible element depends on the relative significance of its
power or energy compared with the power o~ energy of the other bonds
or elements.

4.2.

SYSTEM MODELS WITH IDEAL

MACHINES

221

9. Identify the constitutive relations for each 1-port element introduced in
step 8, that is, give the values of constant moduli and equations or plots
for nonlinear characteritstics.
Consideration of power or energy may help
here, also.
10. If the display of your bond graph is messy or awkward, re-draw it to suit
your taste. Eliminate any unnecessary 2-port junctions. You, may wish
also to simplify the graph in preparation for analysis. (Some simplifications are presented in Section 4.3.)
It is important to complete steps 1 - 3 before starting to draw the bond
graph. Most of the critical modeling decisions are made here, and have nothing
to do with bond graphs or whatever other modeling language you plan to use.
The essential virtue of the bond graph is that its junction structure imposes
the conservation of energy, saving you from having to find the relationships
between the various efforts independently. You can substitute such relationships
in the modeling process for the generalized geometric constraints between the
associated flows, but this approach often is harder, and is not recommended
as a general procedure. Examination of force relationships is most helpful as a
check; this becomes an optional step 11.
4.2.9

Tutorial

Case

Study

The steps above are applied to the hydraulic/mechanical system pictured in part
(a) of Fig. 4.11. Hydraulic fluid drives a single-rod cylinder which is connected
to a weight through a cable and a pulley. The flow emerging from the rod-end of
the cylinde~ drives a rotary hydraulic actuator which in turn drives a flywheel.
Friction and leakage are to be neglected in the model you are asked to create.
Step 1 asks for the key variables to be defined. The input volumetric flow
rate Q and pressure P are given already. The velocity of the piston and cable is
labeled in part (b) of the figure a.s ~. The angular velocity of the output shaft
the rotary actuator is labeled as ¢. These are the key input and output variables.
The flow rate between the cylinder and the rotary actuator also appears to be
a useful internal generalized velocity, so it also is labeled (Q~) as part of Step
Step 3 asks for the key geometric constraints. The velocity ~ is related to
the flow Q by Ap&= Q, where Ap is the area of the piston. The flow Q~ is in
turn related to & by Q~ = Aa&, where Aa is the area of the annulus (the area
Ap minus the area of the shaft.) This flow also is related to the angular velocity
~) by Q~ = Dq~, where D is the volmnetric displacement of the rotary actuator
per radian of rotation. This displacement can be related more specifically to
the diameters and width of the actuator chamber (Problem 2.22), if desired.
Only now that the key variables and constraints are identified ought one
start to draw a bond graph. The input boundary bond and 1-junctions
to
allow the key generalized velocities to be registered are shown in part (c)
the figure, following the instructions of Step 4. Their placement on the sheet
is arbitrary, but enough room for likely further development should be allowed.

222

CHAPTER 4.

INTERMEDL4TE MODELING

hydraulic ram

!(Q~.[J....~~]

.|

(~

pulley

cable

rotary motor
/
with attached ~ I"~--~--~.~ I
flywheel
(not shown) reserv°ir
(a) given information

(b) labeling of key velocityvariables
P
(c) start of bond
graph
1Qr

(d) geometric
constraints
represented

Figure 4.1i: Tutorial example

I ]

4.2.

SYSTEM MODELS WITH IDEAL MACHINES

223

Step 5 now directs that the constraints found under Step 3 be implemented,
as shown in part (d), in the form of three transformers. This structure implies
relations between the pressures, forces and torques that have not been directly
considered. In place of these relations, however, the conservation of energy has
been assumed implicitly. As a result, these relations must be correct. This is
an automatic benefit of using the bond graph.
Steps 6 and 7 do not apply to the present situation. Step 8 starts by asking
about effort sources. The mass at the right is pulled downwardby a constant
gravity force, which is represented in part (e) of the figure by an effort source.
Acceleration of the mass and the piston requires or imparts an "acceleration
force," which is represented by an inertance element bonded to the 1-junction
for }. The fiywhe.el itself represents a substantive inertia, which is added to
the 1-junction for ¢ by a second inertance. The constitutive relations for these
added 1-port elements, requested by Step 9, are Se -- mg, Ir~ = m (ignoring
the mass of the piston and piston rod, etc.) and I¢ -- J, where J is the mass
momentof inertia of the flywheel.
The bond graph is now complete, barring any refinements to represent friction, leakage, inertive or compliance phenomena. Step 10 suggests you consider
streamlining or simplifying the graph. One possibility,
shown in part (f)
the figure, telescopes the transformers T2 and T3 into a s!ngle transformer with
modulus T2T3, removes the unnecessary 1-junction for ¢, and straightens out
the layout. Subsequent analysis is considerably simplified by combining the two
inertances into a single inertance, as you will see later.
4.2.10

Common

Misconceptions

Should you experience difficulty in constructing bond-graph models, the following list of commmonmisconceptions and comments may help.
Misconception
element.

1: Each physical

component is represented

by a bond graph

Comment: Some physical components require several bond graph elements for
their representation. On the other hand, one bond-graph element sometimes
can represent more than one physical component. The bond graph is an
integration of a set of relationships rather than a set of components. Never
invent ad hoc bond graph elements such as three-ported T or G elements or
(before readingChapter 10, at least) two-ported R, I or C elements; they
have not been defined. And, never use ad hoc symbols in a bond graph, such
as L for an inductance or b for a dashpot.
Misconception 2: A "LOAD" always can be represented
by a resistance
element.
Comment: Elements indicated by a word in a word bond graph do not have
their type of behavior uniquely specified, unlike the standard bond graph
elements. A "LOAD,"for exa~nple, could include inertance or compliance
effects or a source.

224

CHAPTER 4.

INTERMEDIATE

MODELING

Misconception 3: One can usually draw a bond graph before defining
variables.

the key

Comment: Defining the key variables on a drawing of the physical system is
the key step in the modeling process, regardless of the choice of the modeling
language. The definitions implicitly represent manyof the assumptions being made. You also should determine the relations between these variables
before drawing the bond graph.
Misconception 4: Gravity implies potential energy, and therefore is always
represented by a compliance elelnent.
Uomment: Oftentimes an Se element, which also can store energy, is much
more appropriate, particularly to represent the (constant) weight of a solid
object. In general, first determine the relation between the generalized force
and the generalized displacement. Also, rememberthat the inertia of a mass
is treated separately from its weight.
Misconception 5: Each unique kinetic energy, T, has a unique inertance, I.
1
Comment:The unique kinetic energy equals ~.Iq-,
presuming linearity applies,
but there maybe several different possible definitions of 0. Thus, I is referred
to 0- The same is true of resistances.
Misconception 6: Each unique potential energy, ~2, has a unique compliance, C.
Comment: C depends on the q it is referred to, just as I depends on its 0Sometimes the choice is important; for example, the formula w,~ = 1/x/T~
requires that C and the I are based on the same q and its time derivative,
0, respectively.
Misconception
7: The bonds around a junction need not have either
a
commoneffort or a commonflow.

Comment: You know better than this when you think about it, but are you
always consistent? Following the 10 steps of the modeling guidelines will
minimize this type of accident.
4.2.11

Summary

Couplers such as electric transformers, fixed-field DCmotors and positive displacement fluid/mechanical transducers often are modeled with two efforts but
with a single flow at one or both of their ends. The corresponding bond graphs
have a 1-junction and two boundary bonds at these ends, joined by a transformer or a gyrator. Frictional losses can be represented by resistances bonded
to lojunctions, while slip or leakage losses can be represented by conductances
(or resistances) bonded to 0-junctions.
A hard mechanical constraint between two velocities within a machine can
be represented by a transformer. Such kinematical constraints often are between
the sums or differences of the velocities used as the variables of the model. Such
sums and differences can be represented by a structure of 0 and 1-junc~tions. If
the structure is based on knownvelocity constraints, the conservation of energy
implicit in the structure automatically ensures that the forces are properly related, apart from frictional losses, which can be added in the form of resistances,

4.2.

225

SYSTEM MODELS WITH IDEAL IVlACHINES

~

flywheel

pa~r [[ spring

Figure 4.12: Guided Problem 4.3
and kinetic energy storage, which can be added in the form of inertances. If
the structure is based on known force relationships in the absense of friction
and inertia, the proper velocity constraints also must result. Most students who
have studied elementary mechanics tend to favor force-based approaches at first,
but ultimately recognize that velocity-based approaches are more powerful. You
are encouraged to develop skill in dealing with constraints in terms of velocities,
and in practice to determine the relationships before attempting to draw a bond
graph.

Guided

Problem

4.3

This is a relatively simple problem with geometric constraints. It is important
that you attempt it before viewing the author’s solution.
For the system pictured in Fig. 4.12, define variables for a dynamic model,
draw the bond graph for this model and annotate the graph with the variables.
Suggested

Steps:

1. Identify distinct components and the variables used in the definition
their characteristics.

of

2. Represent the components by bond graph models. If you have not done
so already, identify which variable is commonin the key junction of the
model, and represent this by a 1-junction.
3. Assemble the elements of your model into a complete bond graph.

Guided

Problem

4.4

This straightforward problem involves finding a steady- state ~nodel for a circuit
with hydraulic actuators.
The hydraulic system shown in Fig. 4.13 drives two hydraulic cylinders and
one rotary actuator. One of the cylinders is connected regeneratively,
while
the other has one side connected directly to a reservoir at a negligible pressure.

226

CHAPTER 4.

INTERMEDIATE MODELING

cylinder 2
area of cylinder1:
area of cylinder 2:
Ac2
area of shaft 1:
area of shaft 2:
As2
diameter of vane hub: dh
diameterof vanetips: d,
width of vanes (normal
to plane of drawing):
Figure 4.13: The hydraulic system of Guided Problem4.4
Drawa bondgraph for the system, identify its elements in terms of the geometric
parameters listed in the figure, and relate the input pressure P and fiow Q to
the forces F1 and F2, momentM, and velocities 21, 22 and ~. Neglect friction
and leakage.
Suggested Steps:
Drawa bond graph for the system, representing each of the actuators
as a simple transformer. (The details of the regenerative circuit are not
addressed at this .point.) Pay special attention to the need for. any 0 or
1-junctions. Label the variables P, Q, F1, 21, F2, 22, Mand ¢.
Find the modulusof the transformer for cylinder 1, which by convention
equals the ratio of the output generalized velocity to the input generalized
velocity as indicated by the powerconvention half-arrows that you have
chosen. (Both half-arrows must be oriented the same way.)
Find the modulusof the transformer for cylinder 2, whichis regenerative.
This can be done either of two ways:
(a) The brute-intelligence way employsa bond graph for the subsystem, as in Example4.7 (pp. 208-209), with a 0-junction and
1-junction. Note the negative transformer modulus(which could be
madepositive by redirecting the two relevant power-conventionarrows).
(b) The insightful genius way notes that for every net unit volume
of oil that enters the cylinder, an equal volumeof steel shaft must
emerge.
4. Find the modulus of the transformer for the rotary actuator. This is

4.2.

227

SYSTEM MODELS WITH IDEAL MACHINES

piston area:

Ap

shaft area:

A
s,

cyhnde
piston and rod mass: rn
~ tubing

dashpot

~

~__~

spring

pulley mass:

M

pulleyradius:

r

springrate:

k

dashpotcoefficient: b

mlley

Figure 4.14: Guided Problem 4.5
directly related to the volumetric displacement, D, defined as the volume
of fluid displacedper radian of rotation.
The three loads are attached to the core of the modelby 0 or 1-junctions.
Write expressions for the efforts and flows on the load bondsconnected to
these junctions in terms of the load variables and the transformer moduli.
Perform the appropriate summationsfor these junctions in order to find
expressions for the source pressure and flow in terms of the load efforts
and velocities.

Guided Problem 4.5
This problemincludes a regenerative hydraulic cylinder as in the problemabove,
but emphasizesmechanical constraints ~vith rotation and dynamics.
Incompressiblefluid is ported to both sides of a piston, as shownin Fig. 4.14.
A belt with a spring at one end and a dashpot at the other end is wrappedwithout sliding arounda pulley comprisinga disk of uniform thickness. Parameters
are listed in the figure; phenomena
not represented by these parameters, other
than gravity, maybe neglected. Modelthe ~ystem, considering the pressure of
the fluid, P, to be an independentexcitation and the motionsof the solid parts
to be the responsesof interest.

228

CHAPTER 4.

INTERMEDIATE

MODELING

is connected in a regenerative circuit;

model it ac-

Suggested Steps:
1. The piston/cylinder
cordingly.

The velocity of the piston is an obvious state variable. Something about
the rotatiorl of the pulley also is needed; choose between its angular velocity and the velocity at its periphery relative to the.velocity of the piston.
Place corresponding 1-junctions on your paper labeled with symbols for
the chosen velocities.
Write an equation to represent the kinematical constraint between the velocities you have chosen, and represent it by a structure of bonds, junctions
and transformers. Take care with the power convention arrows.
At this point you have completed the junction structure of the model,
which is the difficult part of the modeling process. Complete your model
by adding R, I and C elements, as suggested by the drawing and the list
of parameters. Also, add something to represent gravity.

Guided

Problem

4.6

This is a typical problem regarding the interpretation of sales literature for a
constant-field
DCmotor.
A permanent-magnet DC motor is advertised as behaving as follows for a
rated armature voltage of 40 VDC:the no-load speed is 1153 rpm, for which
the armature current is 0.27 amps; at a "rated" speed of 900 rpm, the torque is
120 oz-in and the current is 3.00 amps. Find the values of G, R and Mo. Also,
estimate the stall torque, the maximumpower the motor can deliver and the
associated speed and efficiency. Finally, estimate the maximumefficiency the
motor can deliver and the associated speed and power. Assume that the model
given in Fig. 4.7 (p. 211) applies.
Suggested

Steps:

1. Extrapolate the current to the stall condition on a plot of current vs.
speed. Compute the armature resistance, R.
2. ComputeG as the ratio of the change in torque to the change in current.
3. Compute M0from the no-load current

and the value of G.

4. Computethe stall torque by extrapolating the characteristic as defined by
the two given conditions to zero speed, or use the plot in part (c) of the
figure or equation (4.1) (p. 212).

SYSTEM MODELS WITH IDEAL MACHINES

229

The speed for maximuinpowerlies at the mid-point of the characteristic.
The associated poweris the product of the speed and the torque. It can
be found from equation (4.2), also. The associated efficiency can be found
fromequation (4.3) or picked off fromthe plot in part (d) of Fig.
The maximum
efficiency and the associated speed and powercan be found
from equations (4.4). The efficiency and poweralso can be picked off from
the plots.
PROBLEMS
4.8 Modelthe hydraulic system below with a bond graph.
accumulator

DC motor
whe
~~

el
pump

~

,~,

hydra~lic/~
v.~e

motor

--~_

~o..~s~~

R

flY

I , coupling
reservoir

reservoir

4.9 Represent the electric circuit below with a bond graph.
C~

4.10 A motor drives a pumpto supply a hydraulic cylinder that lifts a 3000-1b
weight. Thetorque-speedcharacteristic of the motoris plotted on the next page.
The displacement of the pumpis 0.50 in3/rad, and the area of the cylinder is
5.0 in2. The flow passes through a valve with the characteristic also plotted.
weight

cy
~ lind
P~’

er
val’~ve
Q~

Pc

~

230

CHAPTER 4.

600
M,
in.lb
4OO

INTERMEDIATE MODELING

1200
pressure,
psi
800

200

0

0

400

40

80

120

~
160 200
~, rad/s

0

0

40

80
12
Q, in~/s

(a) Modelthe system with a bondgraph, neglecting inertia, friction and
leakage.
(b) Determinethe pressure Pc in the cylinder, and add it to the plotted
pressure Pp-Pcto get an effective load characteristic as seen by the pump.
(c) Transformthe given torque-speedmotorcharacteristic to plot an equivalent pressure-flow source characteristic at the outlet of the pumpon the
sameaxes as the plot of part (b).
(d) Determinethe equilibrium pressure Pp and flow Q, the speed at which
the weight rises and the torque and angular velocity of the motor.
4.11 A motor, gear pump, mechanical load and two hydraulic loads are interconnected as shownbelow.

pair

~

(a) Modelthe system with a bond graph, identifying the loads as resistances.
(b) Relate the moduliof any transformers in your bondgraph to physical
parameters of the system.
(c) Express the torque on the motor as a function of the shaft speed and
the various parameters.

4.2.

SYSTEM MODELS WITH IDEAL

MACHINES

231

4.12 Modi~’ the solution of Guided Problem 4.4 (pp. 225, 239) to include friction in the three actuators and internal leakage in the rotary actuator. Write the
corresponding equations, leaving the friction and leakage relations as unspecified
functions.
4.13 A double-rod-end cylinder exhibits internal leakage across its piston proportional to the pressure difference across it. Model the device with a bond
graph.
4.14 The three-port hydraulic system below can be used to increase or decrease
pressure or flow. (It is used as the heart of a device knownas a hydraulic "intensifier.")
Model the system for steady velocity with a bond graph, neglecting
leakage, inertia and friction. Start by labeling the flows for all three ports.
Relate all transformer moduli to the areas ALp, Asp and ASH.
/ / / / / / / //.

filled

with incompressible liquid

///////~
~
shaft
area:
ASH
large piston area: Ate ,,~ ~ small piston area: Asp
4.15 Water is pumpedfrom a well to an elevated tank without the use of electric
power. The level of water in a well is significantly above the level of a nearby
pond. So~ne of the well water (flow Qm)drives a hydraulic motor on the way to
the pond. The motor and pump have displacements of D,. and Dp, respectively.
Assmne steady-state
conditions; the heights zw and zt may be assumed to be
given. The water has density p.

Find a steady-state

bond-graph model. Identify

any parameters in terms of

232

CHAPTER 4.

INTERMEDIATE MODELING

the information given. Neglect friction and leakage. The seepage into the well
(flow Qs) maybe represented by a general source.
4.16 A motor drives a fan through a belt drive. The torque-speed characteristics of the motor and the fan are plotted below. The pulley ratio is not
specified, but the motorshould not be run for long with a torque higher than
its continuous-dutyrating of 4.0 in..lb (to prevent overheating).

12
torque,in-lb

4
0

I

0

I

400

I

I

I

I

I

I

II

1600
800 1200
speed, rpm

(a) Drawa bond graph model of the system, neglecting energy storage
elements.
(b) Determinethe speed of the motor that maximizesits allowable power.
(c) Determine the corresponding speed and torque of the fan, and the
pulley ratio that achieves them. Neglect belt losses.
(d) Re-drawthe bondgraph, including the effects of kinetic energy associated with the motorshaft and with the fan shaft.
(e) Estimate the torque available to accelerate the system, reflected to the
motorshaft, for speeds less than one-half the equilibrium.
~,
(f) The motor and fan have rotational inertias of 0.2 and 1.2 lb.in..s
respectively. Estimate (4-10%) howlong it takes the system to go from
rest to one-half its steady speed.
4.17 Derive equations (4.4) (p. 212) that describe the operating point
maximum
efficiency of a DCmotor with constant field.
4.18 You are asked to purchase a permanent-magnet DCmotor which will be
driven by a power supply that has a maximum
voltage of 24VDC.The maximum
load is to be 10 oz.in, at 2000rpm. A wide assortment of fair-quality motors
with Q =_ RMo/Ge= 0.04 are available.

4.2.

SYSTEM MODELS WITH IDEAL MACHINES

233

(a) Determinethe values of G, R and Moand the associated stall torque
and no-load speed that wouldrepresent the smallest motor possible (which
would therefore have to operate at maximum
power).
(b) Determinethe valnes of G, R and M0and the associated stall torque
and no-load speed that wouldrequire the. smallest powersupply possible
(which wouldtherefore have to operate at maximum
efficiency).
4.19 A commercial permanent-magnet DCmotor is advertised as having a
speed of 2200rpm at 24 VDC
and no load, a 4 oz.in, friction torque, a "torque
sensitivity" of 14.8 oz-in./amp, a "back EMF"of 11.0 V/Krpmand an armature
resistance of 1.9 ohms.
(a) Find the value of G in volt-seconds from the "back EMF,"and compare
to the value of G in N.m/ampfound from the "torque sensitivity." Do
they agree?
(b) Find the quality parameter Q =_RMo/Ge,assumingthe modelof Fig.
4.7 (p. 211).
(c) Use the information from parts (a) and (b) to computethe no-load
speed. Does this approximatelyagree with the advertised value?
(d) Estimate the maximum
power this motor can deliver, and the associated speedand efficiency.
(e) Estimate the maximum
efficiency this motor can provide, and the
associated speed and power.
4.20 A shunt-controlled DCmotor is shown below. The gyrator modulus for
the ideal machinepart of this systemis proportional to the magneticfield and
thus to the current if passing through the field circuit; you mayassumeG =
wherea is a constant.
i~ R~

(a) Assumingconstant supply voltage, ea, find the torque Mand sketchplot as a function of the shaft speed¢.
(b) Represent this system (with constant ca) by a bond graph with fixedparameter elements. (Hint: Twoversions are possible, one using a 1junction and the other using a 0-junction; you might try to find both. No
gyrator or transformer is needed.)

234

CHAPTER 4.

INTERMEDIATE

MODELING

4.21 A Ward-Leonard drive comprises two DCmotors with their electrical leads
interconnected. As pointed out in Example 2.6 (p. 74), the combination acts
like a rotary mechanical transformer. The transformer modulus can be varied
through adjustments in one of the circuits for the magnetic fields. Investigate
the effects of the total of the resistance in the armature curcuits, R, and the
brush frictions M0a and M02. The gyrator moduli can be written as G1 and G2,
and assumed to be known. The following steps are suggested:
(a) Draw a bond graph of the system, neglecting energy storage. Label
the input and output speeds and torques, the friction torques and the
electrical current.
(b) Use the diagram to deduce expressions for the two moments as functions of the speeds and the parameters. (The use of causal strokes can
provide helpful guidance in assembling the equations for the elements.)
(c) Plot the output moment as a function of the output speed, assuming
constant input speed. The use of nondimensional ratios is preferred.
(d) For the particular
"quality"
values ql = RMo~/G~ = 0.01
q2 -- RMo2/G1G~.~
= 0.01, find and plot the efficiency of the drive
function of the dimensionless speed r -- G~.~)2/GI~I. Also, find the
efficiency, and compare the power delivered at this efficiency to the
power that can be transmitted.

and
as a
peak
peak

4.22 The pulley system shown in Fig. 4.10 (p. 217) has negligible moment,
(implying negligible mass momentof inertia, since Mincludes inertial torque).
Give a reduced bond graph model. (Does it make sense?)
4.23 The pulley system shown in Fig. 4.10 (p. 217) has negligible force, Fc
(implying negligible mass, since Fc includes inertial force.) Give a reduced
bond graph model. (Does it make sense?)
4.24 A cylinder rolls without sliding. One cable is wrapped around the periphery; another is attached to a small central shaft.
cable
fm ~cable

(a) Model the system with a bond graph, using the appropriate geometric
constraint(s)
directly. Omit all inertance phenomena. Give the modulus
of any T or G element.

4.2.

SYSTEM MODELS I~TTH IDEAL MACHINES

235

(b) Find a. bond graph by a different route: specialize the result of Fig.
4.10 (p. 212) by noting that 2b = 0. Is the result compatible with that
part (a)?

(c) Introduce translational

and rotational

kinetic energy into your model.

(d) Reduce the model of part (c) so it contains si ngle in ertance, an d gi ve
its modulus. (Reference: Guided Problem 3.3, p. 107.)

4.25 Model the mechanical system below with a bond graph, and relate the
elements in the graph to the given parameters. Assumethat the link rotates no
more than a small angle.

4.26 Determine the speed ratio of the rotating shafts of the epicyclic gear train
of Example 4.9 (p. 219) when the ring gear is clamped. Also, find the result
when the spider is clamped, instead.
4.27 The lever pictured below can be assumed to be massless, and to rotate
only through small angles from that shown. The physical dimensions of the
mass m are small relative to the lengths rl and r2. Define variables, and model
the system with a bond graph. Identify the moduli of the model elements in
terms of the given parameters.

4.28 Repeat the above problem with the lever assumed to be slender and uniform with mass M. Let m = 0.

236

CHAPTER 4.

INTERMEDIATE MODELING

4.29 Oil passes through a valve to drive a hydraulic
cylinder whichin turn raises a cable that lifts a
heavy steel ingot, as pictured. Youare asked to
define appropriate variables, using symbolsplaced
directly on the drawing and explained by words.
Also, relate the moduliof all elements to physical
parameters of the systemwhich you should define.
Includethe effect of the principal inertia.

to reservoir

valve

ingot

4.30 A DCmotor with constant field, armature resistance R and brush friction
M0rotates a winchof diameter d that drives a vertical cable. The cable raises
one end of a uniform heavy beamof mass m and length L that is pivoted about
its other end, as shownbelow. Nevertheless, the beamremains not far from the
horizontal position. Define variables and modelthe system with a bondgraph,
including the principal inertial effect. Relate the moduliof bondgraph elements
to the given parameters.

DCmotor

~

)
cable

m

4.2.

237

SYSTEM MODELS WITH IDEAL MACHINES

4.31 A DCmotor with armature resistance rotates a pinion which drives a rack
supported by a linear ball bearing. The rack in turn drives a second gear that
is attached to both a flywheel and a spring. Note that the second gear can
translate as well as rotate. Define appropriate generalized variables and label
the sketch accordingly, state the geometricconstraints betweenthese variables,
draw a bond graph model, and relate the moduli of your model to the given
primitive physical parametersand any others whichyou define. Inertial effects
apart from the mass m maybe neglected. (Hint: Makesure you identify four
distinct mechanicalvelocities.
DCmotor and pinion
R

~/~

~wheel
k

//II//////I/I/~I/I//
4.32 A motor drives a cable with a winchof radius rw, as depicted below. The
cable is wrappedaround a massless pulley of radius rp, and is terminated in a
dashpot of modulusb. The pulley pulls a weight W1which is supported by two
rollers that each weigh

(a) Define key velocities and forces on a copy of the drawing. Note particularly that the centers of the rollers W2do not moveat the sameveloc!ty
as the slab W1.
(b) Drawa bond graph of the system, including the effects of both the
weightsand the inertias of the slab and the rollers.
(c) Evaluate the moduli of the elements in your bond graph in terms
the given parameters. Suggestion: The kinetic energy method may be
particularly helpful for the rollers.

238

CHAPTER4.

INTERMEDIATE MODELING

4.33 A proposed sound ~novie camera must have a steady flow of film past the
sound head despite the discontinuous loads and motion at 24 Hz (plus multiples) of the gate, with its claw drive, which acts on the sprocket holes. One
path for vibrations to reach the sound head from the claw drive is through the
pulley system. To minimizethis transmission, flexible belts are proposedon the
pulleys, as shown.
Treat the claw mechanismsimply as an independent unsteady force on the
top of the claw drive system. Find a bond graph model and identify each
componentwith a couple words (or more). Neglect variations in the magnetic
torque in the motor.

’top view

claw drive

I sounddrive
i

!

sound

drive~
opticalaxis
fixed pin
flywheel

flywheel

side view
pulleyswith"flexible belts

SOLUTIONS

TO

GUIDED

PROBLEMS

Guided Problem 4.3
1. Motorshaft: effort is torque, Mm; flOWis angularvelocity, ~m.
Flywheel:effort is torque, Mr;flowis angularvelocity,
Largergear and spring: effort is torque M~;flow is angularvelocity ~.

4.2.

SYSTEM

2.

MODELS

motor:

WITH

IDEAL

S e---L--~ 1 .---~-

239

MACHINES

G ---~-

Ro
gear pair:

viscous coupling:

~ 0

flywheel:

T
R,.

spring:

~ C

assembled bond graph:

Guided

Problem

4.4
I^.__._.,_T 2 F...

1.

",-Po

F~ l~
2.

notes: Q1 is the flow to actuator 1, and Q2 is the common flow
through actuator 3 to ~ctuator 2.
The power to the reservoir
shown
is neglected, because its pressure is
virtually zero

A"
2

M

T~-

3. (a)

~-.-~
~

To-~I.

F.__~

r° -’-’- x:
Q.,---7-° -E-~
52 _ 1
Ta ~ Qa
A’c2
Qe Q~ + Q~

Q~/~

(b) Q~ = A~; therefore,
4. Q2 =

D6; therefore,

Tb~

~2
Qb -

+ Q~/~z

1
Ac2
- As2

A~ - (A~

- A~z)

T
Ts - Q2 - D - w(r~ - r~)’

where r~ is the radius

to

the tip of the vanes, rs is the radius of the shaft and w is the width (not the
thickness) of the vanes. See Guided Problem 2.5 (p. 69) and its solution
further explanation.
P=TIFI = TzF: + T3M

T,F, + T,M

=~11 + ~ T,F, ~ =2,/T,
(choice)
TI

T.,,M
T3

= area of the shaft desirable simplemodel: l/As =l/r =m+M =l/k R=b MODELING .5 1.= areaof the piston A. Pep 0 alternativedetailedmodel: To = 1/Ap: To =A~-As ~ T2 A~. detailedmodel: A~-A~ T~ = A~ 4.240 CHAPTER Guided Problem At INTERMEDIATE 4.

4 N.The value is slightly different from that found using the given torque.Mo= 0.m M~t~.147x0.310. The difference could be due to round-off by the advertiser.084 RMo . rpm e 40 -= 3.u = Gist.00 G ~ 1~ S~ 9~ 1153 speed.22481 = 0.~u .27) x 16 x 39.27 0 R = R ¯~ 3.6 ~---stall current = 12.m/amp The citing of the 120 oz.3104 x 12. e-Ri 0. or errors in the assumedmodel. therefore. amps M~ --~ 1.37 x 0. SYSTEM MODELS WITH IDEAL Guided Problem 241 MACHINES 4.~ .86 ~maw = T 1153 27r X 6~ = 117 watts X ~ 3. G e-Ri= I (e-.2.3104 x 40 = ~(~ - ~)~ = ~.Mo= 3.084 N.3242 volt.71amps i. of torque is redundant.m.Ri =40-3.0213 Ge 0.86 N.71 .3104 × 0.27 0.in.4. The linear extrapolation gives the samevalue.0.Ri = G~.71 120 G = change of torque _ _ change of current (3 .27 = 0.147 ohms istau 12. WhenM = 0. Mo = Gi = 0. 3.0.145 x 0.s/rad These units are the same as N-m/amp. The value of G can be found without this datumas follows: e .

The same type of structure also can represent biased loads. The pressure of the water sprinkler system of Section 2. The resistances which result are constants. If they are not so directed. The creation of a biased sink from an unbiased resistance is shown in part (a) of Fig. is not zero at zero flow. Further key reductions are given below. C and I elements bonded to a single 1-junction or 0-junction. without any loss in accuracy despite someloss in detail. except for the special constant-effort and constant-flow sources. Such simplifications usually are highly desirable for two reasons: the essential character of the simpler model is easier to grasp.15. the effort eo for 0 = 0 is marked by a heavy dot. 4. INTERMEDIATE MODELING Model Equivalences The bond graph models developed thus far often can be simplified readily.242 4.3 CHAPTER 4. and the drag of a vehicle does not vanish when the speed approaches zero. for example.3. and nonlinear examples by dashed lines. The sources also are constants. a junction with only two bonds can be removed and the bonds connected. and a flow at zero pressure. Rather than directly employ a source with these characteristics. These replacements are particularly useful for straight-line characteristics. Further. Resistive loads with biased characteristics also have been employed. aiding design. It is essential to remember that each element can be combined only with other elements of the same type that happen to be arrayed on the same junction. 4. an impeller pump unit has pressure at zero flow. an important elaboration is considered which allows you to dispense with all one-port elements with characteristics that do not pass through the origin. you may re-direct the half-arrow on one of the bonds and compensate by changing the sign of one of its variables and the sign of the modulus of any associated one-port element. an unbiased resistance and a junction. has a torque at zero speed. however. Remember also that two junctions of the same type connected by a bond can themselves be reduced to a single junction. to construct the source by assembling more primitive elements: a constant-effort or a constant-flow source. and any subsequent analysis is likely simplified. instead of functions. or independent functions of time. including this point. First. that is do not pass through the origin.1 Thevenin and Norton Equivalent Sources and Loads This book has repeatedly represented sources with characteristics that are biased. An electric motor. it is usually better. The conversion from a bond graph to differential equations is thereby expedited.1. and a speed at zero torque. . The same approach also can replace compliances having characteristics that do not pass through the origin with those that do. for purposes of analysis or simulation. as you will experience in the next chapter. In the first example of a sink. Linear examples are shown by solid lines. for example. The most important reduction step has been treated already: the combining ’ of multiple R. as long as the two power halfarrows are directed similarly. The Thevenin equivalent sink is based on the resistance characteristic that results whenthe whole curve.

eO~Sf Norton R sources S.... . -~--~-1~ eo e=eo-e~ Thevenin -~ Se~ eo e~q q R e[ I s~eo qo qo 0 ~/~" Norton . [ilq~=qo-q~ e = S~ R Figure 4.4.3¯ 243 MODEL EQUIVALENCES I sinks/~/ e=eo+e~ ~---[eoW---s~= R q qo q=qR-qo10R qo ¯ .15: Thevenin and Norton equivalent sinks and sources .-~.

A heavy dot is drawn when the curve intersects the axis. The biased characteristic is reconstituted by adding the effort e0 to this resistance curve with its effort This summation. lgt’---~S ~ --- e0 C Figure 4. 4. virtually always benefit from these equivalents. also. Note again that the Thevenin and Norton resistances are the same if and only if they are linear. this is accomplished by bonding the element R to the flow source S)~ with a constant-effort or 0-junction.16. on the other hand. In the Norton version. INTERMEDIATE MODELING e eo e=e. again. A Thevenin decomposition can be applied here. defining the negative flow -O0. Flow summations are horizontal.t the origin. Nevertheless. The second case shown in the figure has the same characteristic as the first.. suggested by vertical arrows in the figure. as shown in Fig. it is possible to redefine the displacement variable of a compliance such that the characteristic passes through the origin. The Norton equivalent is an alternative that employs flow summation rather than effort summation.A passive resistance R with flow 0R results when this curve is shifted horizontally until the dot lies a. Effort summations are vertical. The biased characteristic is reconstituted by summing ~ with -00.+eo eo . Examples are shown in the bottom half of Fig. a resistive effort eR is subtracted from a source effort eo to get the total effort. In the Thevenin version. using a common-flowor 1-junction. but are distinctly different for nonlinear cases. A Norton equivalent does not exist. as suggested by the horizontal arrows. Compliance characteristics sometimes do not pass through the origin.16: Replacing a biased compliance with an unbiased compliance is shifted vertically until the point is at the origin.244 CHAPTER 4. since displacement sources are not defined. Se. and therefore need no Thevenin or Norton equivalents. In the bond graph.15. is achieved in the bond graph by bonding the unbiased resistance element R to an effort source. as always. 4. and are accomplished by a 0-junction. Most sinks have characteristics that pass through the origin. and accomplished by a 1-junction. a resistive flow 0R is subtracted from a source flow 00 to get the total flow. which is tantamount to . Sources. Note that the resistance elements for the Thevenin and Norton equivalents are the same for linear cases.

and the source version. This suggests combiningthe Thevenin and Norton structures. for example. The shifted characteristic employsthe variables e and 0.7. One way to combine them is illustrated in Fig.4. 4.3. that is if it never generates powerbecause the product e0 is . The new axes for the point labelled "b" are shownby dashedlines.2 Passivity istic* With Respect to a Point on a Character- The Thevenin and Norton structures can be viewed as translating the origin of the characteristics in the effort-flow plane. employsthe variables eR and 0R. shownin part (a). 245 MODELEQUIVALENCES eR ~R ~-R (a) original characteristic / ¯ el eo~ ~ . Note that the only difference betweenthe sink or load version.3. consider only howsuch a translation can be accomplished.17. shownin part (c).N~ ~ I . The original characteristic. Translating the origin of source and load characteristics to the equilibrium point at whichthey intersect can aid dynamicanalysis."4. 4. For the present.17: Shifting the origin of a source or sink characteristic along its path horizontal shifting of the characteristic. translation of both the effort and flow axes are necessary. shownin part (b). Aload or sink has been defined as passive if it lies entirely in the first and third quadrants.as will be suggested in Section 4. Consequencesregarding general dynamicbehavior appear later. lies in the signs of the flows.’ el I e~ 0 q~ e=e~-eo (b) origin shifted to point (c) source perspective o qe~ R R Figure 4. Informationregarding stability can result.If the origin is to be displaced along a characteristic.

The characteristic resides partly in quadrants two and four. 4. The property of passivity does not change as a result of this change in perspective. you first label either the effort or the flow on the left-hand bond with a symbol. The characteristic in part (c) of the-figure is identical to that of part (b). 4. This option simplifies subsequent analysis. If. or direct way. a point to the left or below point a or to the right or above point c is chosen as the origin.1 gives the equivalences. This is why two fluid vortices.is an oscillator.3 Truncation Bonded of to R. Transformers C or I and Gyrators Elements Frequently a one-port element is attached to a system model by a transformer or a gyrator. As a result. and deduce the equivalent modulus. the characteristic is said to be passive with respect to that origin. and a gyrator bonded to an inertance yields an equivalent compliance.2). as shownin part (b) of Fig. Thus.3. and therefore is said to be active or non-passive with respect to this origin. 2The gyrator is said to be a dualizing coupler. each inertance "sees" the other inertance as a compliance." interact in an oscillatory fashion. except that the sign of the flow is inverted by employing a source rather than a sink power convention to the bond for e and 0.17. since an inertance bonded to a compliance --. you employ the definitions of the transformer or gyrator and the one-port element to find the corresponding conjugate variable. Then. Finally. . assumes the power convention of a sink. on the other hand. Consider the given characteristic with point b as the origin. A transformer does not alter the type of the combination.246 CHAPTER 4. The combination of two elements can be replaced by a single equivalent element. however. Although a gyrator bonded to 2a resistance remains a resistance. the power convention of a source must be inverted before the test is applied. You need not memorize these equivalences. if two inertances are bonded together by a gyrator (11 ~G~I. passivity becomesa property of the point chosen as the origin as well as being a function of the shape of the characteristic. since you can reconstruct them readily at least two different ways. for example. a compliance remains a compliance and an inertance remains an inertance. In the first. The same conclusion applies to any point between points a and c. On the other hand.for example a mass attached to a spring -. This property can change when the origin is movedalong the characteristic. its modulus becomes inverted. neither of which has any "springiness. Table 4. and can enhance your grasp of the fundamental structure of a system. a gyrator bonded to a compliance yields an equivalent inertance. The test for passivity above. a resistance remains a resistance. this system would exhibit oscillatory behavior. INTERMEDIATE MODELING nowhere negative. you compare this result to the corresponding conjugate variable for the reduced case.The resulting plot is a left-toright flip of that given in part (b).

the gyrator requires the flow on the right to be e/G.1 " ~ so that R~ = T2R. namely ½C(G0) 4. Meshes of the types shownin Fig. For the second example. If you start with the effort. the energy stored in Y.1 Equivalences for One-Port Elements Bonded to Transformers and Gyrators Consider the following two examples.4 Reduction of Two-Pair Meshes Cascadedbonds interconnecting alternate 0 and 1-junctions to form a closed path is called a simple bond-graph mesh. 2. Alternatively.247 4. the compliancethen requires the effort on the right to be p/CG(recall that p =. Since the inertance I’ is defined by the relation O = P/I~. its value is I ~ = G~C. you start the procedureby labeling the effort on the left as e or the flow as 0. In the right-hand example. so that I ~ = G2C. and finally the gyrator requires the flow on the left to be p/GCG.3. Meshesoccur often. The resistance element then requires the effort variable on the right to be RTO. .and finally the transformer requires the effort variable on the left to be TRTO.2q and~R(Tq)-. simplifing subsequentanalysis.As a consequence.fedt).2. MODEL EQUIVALENCES Table 4. which are the first and the fourth entries in the table: e= T2R~ O T RT~ R TO e O CG2 = P~ G p/CG e/G C In the left-hand example. namely~lqlT~. equals the energy stored in C.the flow variable on the right must be TO. the powerdissipatedin the resistancesequalsboth~R1 . The second methodfor deducing the equivalences is based on poweror energy. 4. For the first example. you could start by labeling the effort variable on the left as e. and is recommended as a superior discipline.18 can be reduced to equivalent tree-like bond graphs.Since by the definition of R~ this effort variable also equals R’(L it follows that R~ = T2R.3.youstart by labeling the fiow variable on the left as 0. and proceed around the bonds in inverse order..

-o---~ ""~-o~ z~ Z~. INTERMEDIATE MODELING Z~ ~..~.18: Reduction of simple even bond-graph meshes .¢ Z~ 0 =g>--.1 "Z~ w. 0 1 ..248 CHAPTER 4.-o ~ "’~’o---~ Z2 (a) single meshes (b) nested meshes Figure 4.

.T. as shown in part (b) of the figure.18 are even. EXAMPLE 4. Basic Engineering. 4. These reductions are extremely useful. compliances. and odd if the number is odd. Nested even meshes each having four mesh bonds. 4. These equivalences apply to the junction structure regardless of whether the elements bonded to it are resistances. as shownin the left side of Fig. As a result.4. because they both equal the sum of the flows on the two 1-junctions. Brown. The same statement applies to the bonds with counter-clockwise power direction convention~ since the total number of mesh bonds is even. MODEL EQUIVALENCES 249 A mesh is called even if the number of mesh bonds which have the same power direction convention in the clockwise sense is even. Four-bond even meshes. reduce the graph to an equivalent tree-like structure. boundary bonds or something more complex. oddness therefore model markedly different systems. fortunately. 1972. their analysis is more involved. as shown. In the first case. because tree-like graphs are easier to analyze than graphs with meshes. v 94 n 3 pp 253-261Sept. inertances. sources or something else. All the meshes in Fig. can reduced to the tree-like forms shownin the right side of the figure. and show that an even mesh results. In the second case. in a compatible fashion. the flows on the left-most and right-most bonds are the same.10 Apply the standard steps to modeling the circuit below with a bond graph."Direct Application of the LoopRuleto BondGraphs. which could be resistances. always leaves the evenness or oddness unchanged. They tend to occur less often."J.3 This means that reassigning the power convention arrows and compensating by changing the signs of the efforts or flows.18 part (a). Two meshes that are identical except for the property of evenness vs. ASME Trans. aF.3. because they both equal the sum of the efforts on the two 0-junctions. It can be shown that such evenness or oddness is a property of the mesh. the efforts on the left-most and right-most bonds are the same. Then. can be reduced similarly. Odd meshes are not reducible in this manner. The figure employs the symbol Z to represent unspecified one-port elements. . as you will see later.

192).5 Transmission Matrix Reduction of Steady-State Models* Often a two-ported modelor sub-modelcomprises a chain-like structure with no energy-storage elements. 4.5. The transmission matrices for the shunt and series resistances are readily . Z7 1 -. since two of its mesh bonds have clockwise-directed power arrows and the other two are counterclockwise. INTERMEDIATE MODELING Solution: The standard procedure gives the following results after steps 2 and 4 (p.3.r. whichfor somepurposes is a significant simplification. and so the equivalencegiven there is used to get the final graph.250 CHAPTER 4. A commoncase is shownin Fig. respectively: after step 2 1 after ".6 (pp.in Fig.1 ---~-Z3 Z~ 0 Z 4 15 "--~-Z z~ 4. The transmission matrix concept is introduced in Section 2. step 4 ~ /6 1 /\ 1"~-0"--~ 1 "-~’0 Z1 Z. The 0-junction is first split into two 0-junctions in order to removeelement Z1 from the mesh.. The meshthen is in the form of the second meshgiven. 82:83).0 -~. 4..19. Such cases can be represented as a two-port element described by a single transmission matrix.18. Z~ The mesh within this graph is even.

3.2/R1] " A cascade of such elements of any length can be telescoped in this manner.~ l q2 R The cascade of two such structures in Fig.4.’ 0 ~ 1~ ~> -~--~-.3. q~ l~1-~2 ~2es+e2 l q2 eI e~ TWO-PORT. EXAMPLE4. e 1 ~1 e. Moregeneral steady-state couplers are discussed in Section 9.19: Reductionof a chain of junctions with stub elements deduced as ~1 e2 0 I~:] = I1}R 011 [~:]. 251 MODEL EQUIVALENCES ---4-~.11 Find an equivalent transmission matrix for the static two-port model . 4.19 can be reduced by multiplying their respective transmission matrices: l/R1 I+R.-~-ql q2 R 2 Figure 4.

INTERMEDIATE MODELING Solution: The combination of T. the resistances R.3.19: and at- e2 e3 e4 e~ ---~--~.-~--~ ~i ~2 ~ q3 q4 -4e~EQUIVALENT OVERALL2-PORT e. The central portion of the bond graph.6 G/T1R1 (t4 Summary Subsequent analysis is simplified if general sources and resistive loads are replaced by Thevenin or Norton equivalents. with its 0 and 1-junctions tached resistances. is now precisely in the form of Fig. as well as this central portion. 4. if transformers bonded to one-port elements are replaced by equivalent one-port elements.9 and/~3 is first reduced to an equivalent resistance R~. to be telescoped in one step: l/T1] [1/1~1 01] [~ ~141 [1/G = (I+R4/R1)/GT1 4.. in order to isolate the even mesh clearly: R2 The next intermediate graph implements the standard reduction of this mesh.2-PORT ~ G . if simple even meshes are replaced by three-like structures. Also the left-hand 0-junction is expanded into two bonded 0-junctions.~ and R~ are combined to give R4.--~ ~ q~ q4 The transmission matrix approach allows the transfomer and gyrator parts of the graph. and if steady-state portions of models . if compliance characteristics similarly are made to pass through the origin.T~---~--~. In the third intermediate graph.252 CHAPTER 4.

4.7 comprising junction structures and resistances are reduced to simpler one or two-port models. 2. These equivalences are first used in the next section to help find the equilibium states of models. 3. Reduce the graph one junction-structure element at a time. Follow the procedure suggested in Sections 4. Sketch the characteristics of the resistances. 253 MODEL EQUIVALENCES F Figure 4. 5. 2. The use of bond-graph equivalences and analytical reduction techniques have been presented toward this end.2.1 (p. 4. (a) draw a bond graph model.3. to a tree-like graph. Suggested Steps: 1. starting presumably at the right end. and . As a check you might reduce the electric circuit by some other means.1. 4. also. Guided Problem 4. Invert the procedure of Section 4. For the system of dashpots and a lever shown in Fig.7 This problem typifies simple analytical reduction of a model comprised of steadystate elements.6 and 4. The use of graphical methods for nonlinear structures is developed.7 for finding the junction structure.2. Reduce any mesh. .20: Guided Problem 4.45 (p. 192) to find the equivalent electric circuit. and note that the lever acts as a transformer. should your bond graph have one.(c) find an equivalent overall resistance.20.34 Represent the hydraulic power supply of Fig. PROBLEMS 4. 85) by (i) Thevenin and (ii) Norton equivalences. (b) draw an analogous electric circuit. The final result should be the desired answer. At each step find the new terminal equivalent resistance. The lever may be assumed to rotate only through a very small angle.

15 (p. 4. 70) comes a dashpot when the electric circuit is completed by an added resistor to give a total electric resistance R.41 Determine whether any of the meshes in the bond graphs of Figs. 4. represent this source with a Thevenin or Norton equivalent. 4.0. 45). (b) Resistors come in different physical sizes depending on the heat they need to reject. 231) with the well and pump. 4.10 (p.38 Show that the translating coil device of Guided Problem 2. Also. 4. 4. 4. 210) to a tree-like structure. in terms of the given parameters. Drawa bond graph for the system. . reduce the model to a simple equivalent inertia. Also.254 CttAPTER 4. with no transformer.6 (p. 217) and Example4. Also. add to your model the compliances of the well and the tank.36 Determine the criterion for a linear (constant) resistance to be active with respect to any point on the characteristic. determine the dashpot coefficient. INTERMEDIATE MODELING 4.s and the dashpot must tolerate a steady speed of 5 rev/sec? (c) State fu nctional advantage th is de vice ha s ov er a vi scous ro tary damper of the type pictured in Fig.1 N-m. 218) are reducible to tree-like structures. Also. The flow Qs into the well decreases linearly as the level of the well is raised. determine the dashpot coefficient. 4. and find its modulus in terms of If and r.8 (p. The gears and bearings may be presumed to be frictionless and massless.39 A permanent magnet DC motor can be made into a rotary completing the armature circuit with a resistor. dashpot by resistance b.37 The flywheel with rotational inertia If depicted below is driven through pair of gears with ratio r > 1. What power rating is needed if b -.23 part (d) (p. the resistance due to friction in the pump/motor subsystem. combine the two bond graphs to get a single model representing a positivedisplacement pump or motor with both friction and leakage. (a) If the motor satisfies the relation M = ai and the electrical of the armature curcuit is R.40 Reduce the bond-graph mesh in Fig.35 Continue Problem 4.6 (p. and an output flow Qt drawn from the tank. b. 2.

45 Showthat the bond graph below represents an oscillator. Relate all moduli to physical parameters. 1 C1 g ~ and find its C2 4. 4.t°r~ F S. natural frequency.1 (p. 203). (a) Modelthe system with a bond graph. 255 MODEL EQUIVALENCES 4. DC motor M=ai (ii) ~~C ~ flywheel ~ "-’~ M=o. DCm°. w~.44 A rotary spring with rate k. and (ii) considerations of poweror energy.46 A large charged capacitor drives an ideal DCmotor (neglect armature resistance and friction) after a switch is closed.43 Derive the secondand third entries in the right-hand columnof Table 4. 4.~g ~ (a) Modeleach case with a bond graph. Relate the new modulusto physical parameters. 247) using (i) direct interpretation of the elements.3. or a (ii) spring. . and flywheel with momentof inertia I are connected as shownbelow. and state whether or not it oscillatory in nature. Certain parametersare definedon the drawings.42 Reducethe mesh for the bond graph of Problem4.7 (p. The motoris connectedeither to a (i) flywheel. (b) Find an equivalent simpler bondgraph with only two elements. (c) Find the natural frequencyat whichthis systemoscillates following non-zeroinitial condition. massless and frictionless pair of gears with diameters dl and de.4. 4.

. find its" natural quencyin terms of the parameters defined.... as shownbelow... In version (ii). In version (i).48 A motor with a knowntorque drives a frictionless and leak-free positivedisplacement pumpthat in turn drives a system that includes a frictionless and leak-free single-rod-end cylinder and two valves (that might be adjustable)...]opentank (reservoir) (a) Define parameters and variables and draw an appropriate bond graph model.256 CHAPTER 4. determineits natural frequency in terms of the defined parameters. IH. INTERMEDIATE MODELLNG (b) If either (or both) of the systemsare oscillatory.. F. (b) If either (or both) of the systemsis oscillatory.. ~ t-. the shaft is connected to a rotational spring with constant k. Assumethat all parameters are known. 4.. the shaft connected to a flywheel with rotational momentof inertia I." .... and state whether or not it oscillatory in nature.F .47 The electrical terminals of a DCmotor/generator for which e = a~ are connected to a large inductor with inductance L..and that the load force F is known..._DC motor/generator ~’~-~ (a) Modeleach case with a bond graph.. valve dl ~va~ve piston/cylinder ~]l~-L--II-. 4. Gravity maybe neglected. Electrical resistance and mechanicalfriction can be neglected for purposes of the following questions: flywheel (i) ~ ..

’T "~ ancnorea ena or cable 9 sheaves top view showing pulley system (reducedscale): fixed shaft .3. T (a) Find a model for the system.4. 4. (c) Solve for ~ in terms of the knowninformation. The accumulator contains a free piston to prevent mixing of the oil and the air (or nitrogen). hyraulic fluid is forced out of the cylinder and flows through a flow restriction to an accumulator. One end the cable is anchored. A hook on the aircraft grabs a cable which is connected to this machine by pulleys.velocity. considering the tension T as the input variable and the velocity v as the output. compliance . The inertias of the crosshead assembly and the pulleys are important.free piston . which are not shown below. the pressures of the oil and air are virtually identical. v movable shaft tension.49 The energy of an aircraft landing on an aircraft carrier is dissipated in an "arresting engine" located below decks. the nine pullleys at the other end rotate about a commonshaft which is attached to a single large piston via a moveablecrosshead. As a result. 257 MODEL EQUIVALENCES (b) Write whatever relations for the elements that are not explicit in your annotated bond graph.~I~ ~diu~tahle 9 loops of cable -~ came common fixed shaft 9 sheavest attachedxto crosshead ~ension. cross-section side view: accumulator/. Define parameters and variables as needed. but losses due to mechanical friction can be neglected. (b) Reduce your model to give a single equivalent resistance. The restriction is adjusted continusously throughout the process in order to maintain the desired pressure in the cylinder and therefore the desired tension in the cable. The nine sheaves at one end rotate about a fixed shaft. The cable wraps nine times around the two sets of sheaves (pulleys) on the machine itself.

Rs/(R..butthe net resultshouldbethe same: R = RaR------2--9= Rt{R2 ÷T2[R3 +R.----~-R~. even if one exists.7 R. x X R~ ~1 R9 "-~-R =R8 ~ F--~O-’-’~R~x R8 RS=T~’R7 I t ~ 2 0 ~ 1 "--’-~-R " X Thisis left to the student. A model with one or more unstable equilibria might never settle downto a stable equilibrium.+R~)]} z R~ 9 +R 4. (c) Evaluate the compliance relationship assuming (i) adiabatic or isothermal conditions. or R3 ~ R3"-=~"6 1 ~R R7 ’ 1 R6 T R7 =R3 +R 6 R4 z~ 0 ~ 1 -----~R ~ 0 ~ I . P0. 4.258 CHAPTER 4. The bond graph above is tree-like.R 5/(R4 +Rs)] Equilibrium The equilibrium or equilibria of many models are stable. SOLUTION TO GUIDED PROBLEM Guided Problem 1. but oscillate in what is . and inertance. INTERMEDIATE MODELING Relate these parameters to your primitive parameters.. the formal procedure gives a meshwhich is best replaced by an equivalent tree-like graph. can be taken as given..~ "~-5 1 "---"-~- 0 ~R ¯ -----------------~0 "~-1 "---’~-R x 2. However. The initial pressure of the gas. should you draw the electric circuit on the right below.4 Rt +R2 +T [R3 +R. The resistance can be presumed to be a known function of the position of the cable. The state of such a model eventually settles down to one of these equilibria after time-varying excitation ceases.

as shownin part (h). meshes and parallel combinationsof elements.Single A steady-state model (no energy-storage elements) with a single unspecified source or boundary bond can be reduced to a Thevenin or Norton equivalent load regardless of its original complexity. R. It then proceeds to consider the effect of compliances. you maychoose to combinethe resistances R~and R3 to give R5. Analytical and graphical options for carrying out the reduction are illustrated below. Each of the pairs R2. Note that the force is common to both boundarybonds. Sucha reduction allows the location and stability of its equilbriumstate(s) to be determinedfor any source characteristic. The resulting tree-like graphs are shown parts (e) and (f). How inertances affect the’stability of equilibria is deferred until the next chapter. as shownin part (c) of the figure. The location of an equilibrium can be affected by compliances. as shownin part (i). and proceed as shownin part (d) by adding a new 0-junction so as to isolate the meshin preparation for its reduction. Carrying out the indicated steps is especially simple if the resistances are constants. The meshreductions are carried out with the aid of Fig.mayprovide greater insight.21. even if you plan to employan analytical methodultimately. R3 and R4. the latter is reduced to the former by the delayed combining of R2 and R3. on the other hand. Section 4. Either procedure employsstep-by-step coalescence of cascades. Alternatively.4. resistances. you maydefer this coalescence.18 (p 240). Thus. R is the resistance of an effective overall dashpot.4.R4 + R6 .1 Reduction of Steady-State Source. The stability of the equilibrium states. can deal directly with experimental data. R~ and R5 are combinedto give R6. however. a preliminary rough graphical analysis can be well worth the effort.3 abovedescribes most of the individual steps in a reduction. transformers and gyrators lead to efficient matrix methods. until a single source-load synthesis remains.4. is apt to be strongly affected by both inertances and compliances.but not by inertances. This section first addresses the equilibria of modelscomprisingonly nondynamicelements. As a first step. It is the methodof choice for linear models. Next. The analytical approachimplies the simultaneoussolution of algebraic equations for the various elements of the model. As a case study. Finally. 4. Case Study Models With a .Graphical methods. The two resulting 0-junctions are combinedin part (g). R6 have a commoneffort. 4. R--. 4. and offer particular advantagesin addressing complicatednonlinear modelsfor which analytical models require fancy methodsor create confusion regarding multiple equilibria. so 1 1 1 = +--. R3 1 1 1 R. 259 EQUILIBRIUM knownas a limit-cycle oscillation. R4 and are combinedto give the overall resistance. consider the system of dashpots given in Fig.the constancyof the moduliof the various sources.

~F 0 ~ (b) bondgraph R.260 CHAPTER4.21: Determinationof the equilibrium for a steady-state model . F R~ choicesg R~ ~lter::te [~0~1~0 ~ F ~ R~ ~0~ ~o ~ ~ [ ~4 R~ ~ 0 ~ R~ Figure 4. INTERMEDIATE MODELING Rl (a) system F F R~ F_~ 1 ~ 0 ~ 1.

The source-load synthesis. a constant force Fo is applied to the left end of the system. The characteristic for the particular R3 shown in the figure allows no motion until a threshold of force is exceeded. before again increasing. Vertical addition is employed in part (b) to form R6 from R1 and Rs. so R6 = R1 q-R5. If the resistance force of the overall equivalent dashpot exceeds Fo. to consider non-equilibrium states. and the right end is grounded. Nonlinear characteristics can complicate the algebraic approach considerably. but are they stable? To answer this question it is necessary. Example characteristics for the five original dashpots are plotted in parts (a). This case . Horizontal addition is employed in part (a) to form R5 from R2 and R3. This pattern is typical of some bearings and rubbing surfaces that become better lubricated as the sliding velocity increases. horizontal addition is employed in part (c) to form R from R4 and/~6. You may prefer to use graphical methods instead.4. In part (d) of the figure. R5 has a commonflow.22. impededonly by the inertia of the parts. The resulting characteristic for the overall system also has a dip-then-rise. (b) and (c) of 4. this is typical of dry friction.22: Graphical analysis of the steady-state model The pair R1. rapid leftward acceleration ensues. the force decreases for a range of small velocities. reveals three equilibria. All are valid. showngraphically. Finally. Then.4. as always. 261 EQUILIBRIUM (b) F F 3~ R~ R Figure 4.

non-iterative matrix methodscan be used. you may need to start with an initial condition somewhatclose to the desired equilibrium. with time as the independentvariable. the resistance force is less than F0. An inertance with a constant flow has zero effort.23 . in favor of a simpler alternative. the effort on the ¯ junction becomeszero. and therefore have a commoneffort. Whenever the objective is merely the determination of equilibria. A complianceat equilibrium has a time-invariant effort and therefore a constant displacementand zero flow. INTERMEDIATE MODELING is illustrated by the states associated with the arrows on the left and on the right. Occasionally. 4. This is done by adding either real or fake compliancesto 0-junctions or inertances to 1-junctions. the flow of each compliance element is zero.2 Alternative Steady-State Approaches Models to Reducing Not all very complex systems can be reduced completely through use of the above procedures. in place of the original algebraic equations. all inertances maybe removedfrom a bondgraph. The revised modelleads to differential equations. If there are multiple equilibria. Recourse can be made to iterative numerical procedures applied to the set of corresponding equations. this action leaves the flow state of the systemunchanged. and are not addressedhere. The situation is morecomplicatedfor compliances. 4. illustrated by the states associated with the center arrow. If. This analysis reveals that the center equilibrium is unstable and the outer two equilibria are stable. This alternative converts the static model to a dynamicor time-varying model. therefore. Routine numerical simulation then applies. 4. The compliance elements therefore can be removedfrom the graph. In the special case of linear systems. ~the solution maycontinue to oscillate rather than settle down to an equilibrium state.3 Removal of Elements for Equilibrium Equilibrium implies that each bond has a time-invariant effort and flow. Such approaches are the subject of textbooks and monographson algebra and numerical methods. rapid rightward acceleration ensues. Youshould then change the assumedcompliances or inertances until stability is achieved. It is a by-product of the modeling and analysis of dynamic systems.262 CHAPTER 4.4. whereasthe flows on the resistance elements are not. The solution usually settles down ¯ to a stable equilibrium state that is independentof the presence of the added compliances and inertances. Whichof the two stable states comesinto being depends on the history of the force and the initial velocity of the dashpot. so all other bondsattached to it also maybe excised. on the other hand. if an inertance is bondedto a 0-junction.4. whichis the major subject of this book.This situation is illustrated in Fig. The availability of this option sharply reduces your needto becomeproficient in the more exotic methodsavailable for reducing complexalgebraic or steadystate systems. Further. Whenboth resistance and compliance elements are bonded to a common 0-junction.

xe" equilibrium~:> .4.x~ ~?~ ~ x: i. requiring infinite force. The examplesof Fig. The force compressingthe spring equals the total applied force.~ ~ . 4.10 m.24. a constant displacementx3 . 4._~ R x~ ~ x. Thus.05 mand the pulley has a radius of 0. WhenM= constant._. so ~3 = ~72.. the displacement of each compliance would increase without limit.~’~-. this flow must be zero if the appled force is constant. Otherwise. 263 EQUILIBRIUM C (a) .~. Since the equilibrium flow is zero. The weight weighs 150 N. on the other hand. both resistance and compliance elements are bonded to a common1-junction. Youusually can tell whetherto delete compliancesor resistances by appealing to physical reasoning.23 are no exception. Problemsregarding the deflection of static structures._]~F~~. as plotted. the resulting equilibrium involves a steady flow. If all complianceis gone.4. the resulting equilibrium is static. when combinedwith the other steps for bond-graph reduction.~ R ~~-x~ R C~I. (b) ~ C x~ ~ 2~ ~. If instead all resistance is gone. Retention of the compliances then gives the proper displacements. this F is constant. ASa consequence.~R - x~-x: equilibrium C Figure 4. the .4. and this bond maybe removedfrom the graph. The winchhas a radius of 0. removal of the resistances from the graph leaves the equilibrium state unchanged. The motor and the viscous coupling havenonlinear characteristics. In the situation illustrated in part (b) of the figure. for example.4 Case Study with a Steady-Velocity Equilibrium The objective is to find the equilibrium state of the motor-driven mechanical system shownin Fig.23: Removalof compliancesand resistances for the equilibrium state part (a).x2 results. The model comprises some components with constant moduli and others with nonlinear characteristics. the power on the compliance bond is zero. Since both elements have a commonflow. the effort on each resistance is zero. disregarding the dynamicsand assumingthe weight does not interfere with the pulley before the equilibrium becomesmeaningful. 4. The bond graph given in part (b) of the figure includes a compliance and resistance Rc bonded to a common0-junction. eliminate either all resistance or all compliancefrom all or major portions of the bondgraph. The above erasures. are of this latter type.

load 5 0 ~coupling 0 100 200 300 ~. rad/s (d) givencharacteristics 0 I 0 " I I 100 I I I 200 300 ~.~ 1.. weight and ~ dashpoi winc(h (b) bond graph ~ motor (a) mechanicalarrangement 15 moment. rad/s (e) determinationof equilibrium Figure 4...ta’~RSae ’....ml( ~_...~...~"~Ulllbrlum )’. or N’mll // Jviscous (c) bondgraphfor equilibrium (motor) moment N.24: Case study for 9quilibrium with motion . 264 INTERMEDL4TE MODELING /~.~pulley M~ F r°as’r’n ¯ spring| .CHAPTER 4. Cr Ca.

4. by considering off-equilibrium states.5M (4. and therefore removed from the graph.n. The model of the system comprises a source for the motor. whereas equilibrium B is unstable. modeledas a rotational spring.26 is presumed to have the plotted torquespeed characteristic. as drawn in Fig. there are three equilibria. so when F = constant the velocity ~d = 0. Given W-~ 150 N.10b) gives the corresponding values of ~lm. Equation (4. as a result. and the corresponding values of ¢c picked off from the plot for the coupling. so Ca also can be removed. Part (c) of the figure shows the resulting reduced bond graph for equilibrium.F = T~. 2. 4.(W + TpM) = 0. Rd is removed. Should a particular given weight Wbe placed on the spring. The force-deflection characteristic exhibits a rise-fall-rise shape as shown. as usual. ’ = 2. The stability of these equilibria can be determined. This is similar to the stability analyses carried out in Chapter 2 for constant-velocity equilibria. The compliance Cd and the resistance/~d are bonded to a common 1-junction.25.. 4.10a) + 10M) = 7. 4.6 Case Study with Limit-Cycle Behavior The gear-head electric motor in Fig.10 = 10 m. a resistance for the load. a compliance for the spring and a 0-junction to recognize that all three componentsexperience the same torque. The torque-speed charactertistic plotted for the load that the motor drives is characteristic of a bearing heavily loaded in a transverse direction.5 Case Study Equilibria with Stable and Unstable Static Belleville springs are washer-like disks with a conical shape when they are unloaded. For higher speeds the torque again increases as viscous drag takes over.4. . Fitting a smooth curve through these points gives a "load" characteristic that intersects with the "source" (motor) characteristic to give a close estimate of the equilibrium state. given in part (b) of the figure. The three resulting points are plotted as heavy dots on the axes for M:.5 + 0. Examples of these. The analysis reveals that equilibria A and C are stable.~. The friction drops as the speed increases from zero and a lubricant begins to act.ent M= 0.0 Mm = T~. The compliance Ca now becomes bonded directly to the 0-junction. 265 EQUILIBRIUM compliance Cr has a fixed deflection (zero relative velocity).10a) gives the corresponding values of ~. The shaft connecting the motor to the bearing is presumed to have significant torsional flexibility.05 m and Tp = 1/0. The velocity ~2 equals ~1.0.4.4. and ~). indicate acceleration resisted only by inertia. and equation (4. -.105) Values of the viscous coupling mom. T~.5 and 5. since there is no significance to a two-ported 1-junction.05(150 (4. 4.0 N m are now chosen.

and then state 3. giving state 6. The endless cycling from states 6 to 3 to ~5 to 6 is knownas a limit cycle oscillation. Thus the spring winds up. any initial condition results almost immediately in the same limit cycle. It accelerates virtually instantaneously (no inertia. The state of the system at a moment in time therefore can be described by a horizontal line segment. Whenthe momentattempts to exceed that of state 3. It is still decreasing when it reaches state 5.d.¢2 now causes the twist angle ¢1 . so the spring rapidly unwinds and the torque decreases. however. If the system is started from rest. i ~ INTERMEDIATE fixed MODELING support (a) cross-section through centerline F W X (b) force-deflection characteristic and . again) to state 4. ¢2 is far from equilibrium. This equilibrium is unstable! The spring allows the motor to run at a different speed from the load. When~1 reaches its equilibrium value. This particular type of limit cycle often is called a stick-slip instability. the load finally is forced to break away.25: Case study of Belleville springs with stable and unstable equilibria The single equilibrium state is defined by the intersection of the source and load characteristics. At this point ~2 >> (ill. State 6 soon becomes state 2. as you can verify. and the same limit cycle results. W Figure 4. and so on: the cycle is repeated. ~1 jumps to the given value virtually instantly.¢2 to increase in time. (All inertia of the rotor and gears is being neglecte. with the state of the motor at one end and the state of the load at the other. at which point the load speed has no choice but to suddenly become zero. but requires that the torques at its two ends be equal.266 CHAPTER 4.stability analysis for applied weight. Thus the wind-up of the spring continues. the load still hasn’t budged. any small disturbance will trigger the instability. so." Should the system be started somehowat its equilibrium condition. . It is also known as a commontype of "chatter. at state 2. and the torque Mincreases. it will initially jump to the pair of states labeled as 1 in part (e) of the figure. In fact. at only a slightly smaller rate.) The velocity difference ¢1 .

4. 267 EQUILIBRIUM C frictional load (a) schematic M ~ or).4. Mvs (b) bondgraph ~1 (c) torque-speedcharacteristics of source and load (d) shaft spring characteristics M II II II II II II II II II t3.6 t3.4ts.26: Anunstable S-R-Csystem that develops a stick-slip limit-cycle oscillation .4ts.6 time (e) state diagram (f) timehistory Figure 4.

neglecting the effect of all inertias.1. ~M MO -~-~.~. Equilibrium 1 is seen to be unstable. Compare the result to the case in which the springiness is neglected and inertia assumed. . equilibrium 2 is approached. The horizontal line segments are replaced by vertical line segments. For inertia but no spring.268 CHAPTER 4.~)L > O. whereas 2 is unstable. Computer simulation and analysis will show how this inertia modifies and. If it is less than that of equilibrium 1.4.4).) The shaft and the load are connected by a spring coupling (rather than the rigid shaft of Problem 2. so the vertical coordinates of the two states are the same. the 0junction is replaced by a 1-junction and the C element by an I element. while both equilibria 2 and 3 are stable. These indicate that equilibria 1 and 3 are stable. taking care to get the right junction type: S . moment is plotted on the vertical axis. 28. p.4. The arrows. If the initial value of ~Mis between that of equilibrium 1 and the peak of the curve. which means that a line segM ment drawn between the two states is horizontal. ~( otherwise.12 An induction motor and load have the characteristics plotted below (which are the same as in Problem 2. Determine the stability of the three equilibria. equilibrium 3 is approached. If it is to the right of the peak. EXAMPLE 4. R [ C The momentis common. which indicate the direction of change. M [ ]~~~load Solution: Model tile system with a bond graph. with the inevitable load inertia added. are pointed /~ M upward if CdM/dt = ~M -. the system stalls altogether. they are downward. eliminates the limit cycle. INTERMEDIATE MODELING This problem is continued in Section 5. if large enough.

Onthe other hand. Anequilibrium state betweensource and load characteristics is knownto be stable if both characteristics are passive at that state.26 are passive. and the absenceof a limit-cycle oscillation.8 Summary The equilibrium states for bondgraphs that modelconstant velocity systems can be found analytically or graphically. Whenthese methodsbecomeexcessively awkward. . The equilibrium state is approachedasymptotically in time.4. The equilibria of modelswith both compliancesand resistances can be found by eliminating resistance elements that have zero generalized force. using numericalsimulation (as introduced in Section3. All points on the source characteristic in Fig. QUILIBRIUM 4. are assured if both characteristics are passive with respect to the equilibrium point. also.3. F.7 269 Necessary Condition for Instability or Limit-Cycle Oscillation* Thestability of an equilibriumpoint betweena static source characteristic and a static load characteristic. The tools necessaryto ~nakethis determination analytically.determination of the stability requires a detailed examinationof compatible non-equilibrium states. are developedin later chapters.4. and eliminating co~nplianceelements with a non-zero generalized velocity. but is beyondthe scope of this text. and yet respond to a large impulsive disturbance with a limit-cycle oscillation. 4.4. without simulation.2 (p. The graphical analysis presented with that simple exampleis sufficient to predict the limit cycle instability associated with any equilibrium point in this range. an equilibrium point could be stable in the presence of small impulsive disturbances. but the analysis reveals neither an instability nor a limit-cycle oscillation for an equilibrium in that range. Points in the positively sloped portion of the load characteristic with momentsbelow the break-away momentare nonpassive. 245). Otherwise. Whetheror not an equilibrimn point that fails this test is actually unstable depends on the magnitudes of any compliances and inertances that maybe present. In somecases equilibria with steady velocity result. static equilibria result. the non-passivity of any point on the negatively sloped portion of the load characteristic can be observed readily. Further.4.7). 4.leading to a solvable set of differential equationswith time as the independentvariable. In others.the simultaneoussolution of a set of nonlinear algebraic equations by iterative numericalmeansis a major alternative. The presence of an unstable equilibrium maysimply favor one or more stable equilibria. In other cases a limit cycle results. assumingthe modelis stable. A more straightforward approach adds real or fake compliances and/or inertias to the model. This passivity is defined in Section 4.

reservoir cm3/s / 0 cm3/s 5 10 15 20 P~. 4. of the adjustable valve.’~i~i ~ valve ~~k co on . A relief valve limits the pressure... several resistances and possibly a fixed-effort source. Guided Problem 4.8.27: The system of Guided Problem4.. Use this graph to guide the steps below. It also has a static or break-awayfriction of 60 Nmand an addedviscous friction of 5."~. partly for safety. Presumablyyou will include several junctions. and suffers an internal leakage of 0.relief valve and adjustable valve including the attached fluid line are plotted.Q~ 7--~_ ~/ Q~ adjustablel : .. This is probably the most critical step in your solution. as shownin Fig. MPa | //~ 0 5 10 15 Pl-P~.~ relief ( "--~.’il~/i:. The motor has a displacement D = 60 cm/rad. Youare asked to find and plot the torque-speedcharacteristics of the system as seen by the load.8 This problem typifies graphical reduction of a model comprising steady-state elements. treating the motor/pumpas a one-port source.. INTERMEDIATE MODELING P~ ~ ~ P2 Q2 valve ~1 ~ --~. Hydraulic oil is pumpedto high pressure by a motor-driven pump.2 cm/s for each MPaof pressure drop across it.0 Nm/(rad/s). Construct a bond graph model. (The drawinguses standard fluid-power symbols. Suggested Steps: 1. for various angular positions. The fluid in turn drives a hydraulic motor with an unspecified attached load.270 CHAPTER 4.27. and an adjustable "meter-in" valve can be set or modulatedto control the torquespeed characteristic seen by the load. ¢.. 2.) Characteristics of the motor/pump. 20 MPa Figure 4. Combinethe given characteristics of the motor/pumpand the relief valve to plot a source characteristic pertaining to a location betweenthe relief .

//(/////////////////// piston --~p. pushing a piston against a linear dashpot with modulus R. Subsequent specification of a load characteristic and an angle ¢ would give a specific operating condition. Combinethe results of step 2 with the characteristics of the adjustable valve to plot a family of source characteristics. Due to the simple transformer type of coupling. and the associated algebraic relationship. find the moduli of any new (c) Give the simplest bond graph valid for equilibrium. Modify the results of step 3 to refer to the flow which actually displaces the motor but does not leak past it. elements._. modulus R ] R ~ --~’P T=A i "~" R ’ C R’= TeR = A~-R (a) Model the system with a bond graph.50 Slightly compressible fluid enters a cylinder. this can be done by merely replacing the scales of the ordinate and abcissas. shownbelow. (b) Simplify the model as much as possible. .4. PROBLEMS 4. The result should be the desired family of source characteristics as seen by the load. Correct the characteristics of step 5 to account for the static and viscous friction on the shaft. angular velocity.4. each for a different angle ¢. 4. Convert the characteristics plots of step 4 to get torque vs.~ fluid. Determine the moduli of the elements in terms of expressed and implied parameters. pertaining to a location between the adjustable valve and the hydraulic motor. pressure P ~/~ areaA ~’//I///////////////// "-~i "-’-"~ C T "--~r- dashpot. EQUILIBRIUM 271 valve and the adjustable valve.

.. Theindicated radii are rw = 6 in. without slipping.e.10 in... against a restraining dashpot with resistance coefficient b = 0. ft/s ’’ winch I- 12 ~ -[ vO 4 8 12 displacement. motor ~ 200g.52 A gear motor (relatively slow speed).. non-zerovariables at equilibrium). (b) Drawa bond graph of the system.25 lb. drawa bondgraphfor the system.51 A motor drives a winch of radius 1. find the values of all constant.272 CHAPTER 4.. up a 30° incline at steady speed. ~ 51! 06’’’~’’’~’’’1’ 2 angularvelocity. rad/s ’’~~’’’’~ velocity. (b) Estimate the equilibrium angular velocity of the winch and linear velocity of the center of the roller.~- /I force..Z~ . rotates a winch which rolls a cylinder of weight W= 10 lbs. 4. (c) Estimate the moduliof any elements not described by the plots. (d) Find the equilibrium state of the system(i. (a) Define variables...s/in.5 in. INTERMEDL4TE MODELING 4. shownat the top of the next page.0 ft with spring and damperattached. and label its variables consistent with part (a). Thetorque-speedcharacteristic of the motoris plotted. and find the values of the parameters (moduli) therein.ft (a) Define appropriate variables. rl -. as shownbelow.. which both lifts a 50 lb weight and pulls a massless pulley of radius 3. and r2 ---. and identify and label the plots correspondingly.0 ft. Characteristics of the componentsare plotted below.. momenq~ ft-lblo01 force.

35.lb. Air drag coefficient: 0. Include hill-climbing and the possibility of acceleration. 273 EQUILIBRIUM 50 40 in-lb 30 motor~ "" "~I rider. The work can be divided into two phases: (a) Model the system. Get explicit constants or functions which detail the meanings of the various one-port elements in the model. In addition to.lb. Someinterpolation is necessary. plus a manual transmission. inclination of the road and vehicle acceleration. exclusive of the engine and accessories. rad/s 4. 6%at 90 ft. fuel consumption and needed power for the accessories are proportional to the engine displacement.65% of the total road weight. hill-climbing capability.0 liters weighs 240 lbs including accessories. Axle gear ratio: reduction of 2. 15%at 20 ft. The result should give the vehicle speed and engine torque which results from any specified gear ratio.s2/ft 2Frontal area: 15.4. Engine: The weight.lb engine torque. Vehicle weight: 2200 lb. An engine with a displacement of 2.53 Design problem. stating the engine displacement and the various gear ratios. You are asked to size the engine and propose the values of between three and five transmission ratios. acceleration and fuel economyare left to your discretion.6:1.0024 lb. available torque. Overall power train losses: 25%at 10 ft. W O0 I11 I 21 I 13 I 41 ~ 5~ ~ 6~1 ~. (b) The design phase of the problem includes your compromises regarding acceleration. fuel consumptionand relative cost. and has .4. 4. A four-passenger vehicle is to be given an engine scaled to an engine with a known displacement. describe the performance of the vehicle in a form that an informed consumer would want and can understand. hill-climbing. constant rolling resistance force equals 0. presumably with a bond graph.. Most parameters can be taken as given below.4 ft Tires: 1032 rev/mile. Factors such as vehicle loading. Density of the air: 0.

rad/s 500 4.. 4. " locus of lowest f~el consumption 0 100 200 300 400 engine speed.54 For the Belleville spring shownin Fig.source of fluid to this compliance. The jet entrains a constant rate of fluid. Describe howthe force is related to this plot. 4. the jet bends upward. cooling fan.23 (p. a pressure difference P~ . lO0 shaft torque.274 CHAPTER 4. at the bottomof the stand pipe (just above the jet).) varies linearly from 2 HPat zero speed to 6 HPat 80 mph. 264) that is non-passive.maximumtorque ~-’~solidlines: indicated values of BSFC // / . (a) Drawa bondgraph for the system. ft’lb 75 ~. if Ps < P it bends downward.56 A submergedturbulent jet of water discharges into a region of still water.. INTERMEDIATE MODELING the characteristics plotted below.)The powerto run the accessories (alternator. power steering and brakes. Also. ("Brake" refers to the meansof measurement. If on the other hand the jet is not bent upwardvery much. fromthe bottomof the stand pipe.the sumof the flows Q~÷ Qpacts like like a multivalued . sketch the stored energy as a function of the displacement. Ps. Also.24 (p.. 4. These flows are plotted. if the constant pressure of the still water. P. The solid lines represent constant values of the brake-specific fuel consumption(BSFC).. Qe. etc. with a correction for the weight of the water in-between.and the flow Qe acts like a (negative) .P causes a flow Qp to enter or leave the stand-pipe region through the gap betweenthe jet and the edge.55 Identify the portion of the load curve for the case study in Fig..Considerable upwardbending causes a return flow Q~ to be peeled off by the edge.. is higher than the pressure. 4...defined as poundsof fuel per brake horsepower hour. identify which portion of the load curve wouldbe unstable or producea limit-cycle oscillation should the source (motor) characteristic be modifiedto give an equilibrium there. passing below a stand-pipe with a sharp edge or lip as shownon the next page.. recognizing that the stand-pipe acts like a compliance. 265). ~..

275 EQUILIBRIUM (b) Replace the individual characteristics for Qr and Qp with their sum. (c) Determinethe equilibrium states. has the pressure-flow characteristic as plotted.) (d) Determinethe stability of each equilibrium state.~~OQ.rnp flow Qo accumulator valve x "1 valve characteristic ~ Pm ~ reservoir testing system am (a) It is desired to test this valve over as muchof its characteristic as possible. and identify which direction the pressure is inducedto changeas a result. constructed as shownbelow left. according to the bond graph model. (Postulate compatible non-equilibriumstate pairs in the usual fashion. p. without inducing an instability. The fluid is incompressible.) psi //~cross-sectional[// ~/~area A=20ft2 ~/standg~//’~ (1density p ~-//pipe "94 lb’s2/ft4) -4 . This valve can be used as the heart of a "hydraulic ram" (see Problem6.4. A pumpwith constant flow Qo and .O.. (Hint: there are three of them.4.57 A type of reverse check valve. ~pu. ~ ~ " ft3/s -8 _Q~ still -12 water ~ /i/i/i/i///i/ 4.16. 407).

(b) Removethe accumulator and repeat (a). 1 and 2. correspondingto different restrictions. Then.58 A motor. sketch-plot the angularvelocities of the motorand the load as funtions of time. Show your reasoning.drives a load through a spring coupling. on the plot.) (c) The orifice in tile valve has area 2~rrx/x/~. Hint: limit cycle behavior exists for a wide range of starting conditions.~. in which the charactistics the source and the resistance are a plotted. construct two or three useful sourcecharacteristics.276 CHAPTER 4. rad/s \ 4. Deducethe region of the (load) valve characteristic that can be tested stably with your flow source. N. and computeP. Choose a value of Q0 > Qmand plot this point on the Q axis of the plot above.. indicate whether the equilibrium is stable or unstable. system(i) system(ii) characteristics . In each case.~ and Q. (A tiny fluid inertance mains. and for equilibrium points. Assumingthe characteristics given belowand neglecting all inertia. The fluid has density p.m 2 I I 20 I I 40 I I I ~I I 60 \ 100 speed. INTERMEDL4TE MODELING the traditional accumulatoris used with an adjustable linear restriction placed as shownabove right. 4. wherer is a constant effective radius. The valve opening x equals xo whenthe spring (stiffness k) has no force. Find the characteristic of the valve in terms of the parameters given. ~ flexible shaft -~ 4 momem.59 Consider the two systems (i) and (ii) below.

Also. rad/s I I I I 200 4. Estimate the gear ratio in each case that produces the largest possible stable speed of the load. however.4.two different possible characteristics are plotted. A fan with the characteristic plotted belowsupplies air to a chamber of volumeV = 4 ft 3.61 A motorwith a given characteristic and negligible inertia drives a load with a given characteristic and insignificant inertia through a gear pair with negligible friction and inertia.4. Assumeatmospheric conditions. and permits the load to reach this speedfrom start-up. Determinethe stability of the respective equilibria.a 0 10 flow. ft3/s 20 4. Anorifice in the chamberexhausts the air to the environment. has a region of negative slope (due to heating and softening of the material)._. as shownon the next page. .62 A cylindrical workpieceis machinedon a lathe by a cutting tool fed axially by a screw drive at a constant velocity }. 3O torque. fi-lbs 20 10 0 I 0 I I I I I 100 ~1 I speed. and detail any limit-cycle operationthat results. plotted as a function of 1).60 Axial-flow fans and compressors typically demonstrate an S-shaped characteristic. estimate the equilibrium speeds. The tool is somewhat flexible. 15Pgage lOpsi 5- orifice fan t_____. The conCernis that strong vibration might result (tool "chatter"). Thecouplingis (i) rigid or (ii) highly flexible. The reaction force on the tool in the samedirection.so its feed rate ~ at the cutting edgecan differ from ~. 277 EQUILIBRIUM 4.

electric circuits. T. lbs 20 to01 k__~j ~ rotating MODELING workpiece 10 .66 Answer the previous question using kilograms but not Newtons. 4.. 5~ axial 30 force. in/s (a) Model the process with a bond graph.3 0. Use Newtons. what would it be? (There are several good answers. 4. F for force.32 in/s.. Label five rows as follows: R.63 Review question: Make a table of the dimensions of the moduli of bond graph elements.o 0 case i I I 0. k 40 ~ l/ force vs. ¯ = in/s. 2 = 0. INTERMEDIATE force vs. Label four columns as follows: mechanical translation. question using M for mass and eliminating F for 4.. incompressible fluid flow.) 4.. mechanical rotation. t for time and Q for electric charge.. . (b) Find the equilibrium drive forces for the three drive characteristics plotted by the dashed lines above: case (i).. Employthe symbolsL for length. (c) Determine whether each of the three equilibria is stable.) Hint: Your model should include but not be limited to a compliance... G. should any limit cycle exist.67 Answer the above question in terms of inches pounds force and seconds. I... Omit the electric circuit column.. show how the force and speed of the workpiece change. meters. 4. case (iii).09 in/s. (For example.64 Answer the previous force.65 Answer the previous question substituting SI units for the dimensions.. case(ii).2 case iii I I I I 0.1 case ii I I 0. a resistance and a junction. .. including only phenomena mentioned above. Show your reasoning.278 CHAPTER 4.. (d) If you were to add one more phenomenon or parameter to the model.. as given in Chapter 1. exclude inertias and any variation in the rotation rate of the workpiece. secouds and coulombs only. Assumethe transformers and gyrators involve no transduction. C.. Also. 2 = 0.4 velocity.

60 N m. 0.4. 5..0 N m/(rad/s) torque. from those of P1 above.P .(P1 .~ 0 4.8 1 Rr Ro RL So Rj R~: resistance to flow of the relief valve.4.MPa P2 -. to get Q3 6oo~ ’400 ~__150--’~. subtract the (horizontal) given in the plot for R. RL: resistance of the hydraulic motor to leakage flow. Subtract 5 10 15 20 P~-P~.P’2).. R~.2 cm3/(s MPa) R f: viscous frictional torque resistance Se: break-away frictional on the shaft. MPa values of P~ - .: resistance to flow of the adjustable valve. MPa QL = P/RL = P/0.~ 0~"~ cm~/s I ~oo~6"~--.2 from Q3 to get Qm: 400 ~. Therefore. 40O cm~/s 200 0 0 5 10 15 ~0 P~.1 .~~ 0 5 10 15 20 P~-P_. 279 EQUILIBRIUM SOLUTION Guided TO GUIDED Problem PROBLEM 4.

m ~ m m~ MN rad/~n~ ’ 300 600 900 1200 M~.Ry~ = Mi . Nm 6.~cma’MN-= 60 x 10-6P~_ x 106ma MNN = 60P.60 . Nm . therefore ~ = Q.~ rad/~g ~ 1200 M~.5.) 00° ~. INTERMEDIATE MODELING 1 1 rad/s = T = ~ = 6-~ cma/’~-~.~ N.~/60 and ideal torque: Mi = 60P.Se .280 CHAPTER 4.0q~ (Subtract torque loss from plot above.10 ~. M = Mi .

5. in Section 5. Apply the mandatory causal strokes for each effort or flo~v source (but not general source) that is present. Models are categorized accordin~ to their causal status. Finally. the discussion focuses on the basic category of "causal models. After these categories are identified." for whicha set of state-variable differential equations suitable for analysis or simulation is produced directly.1. methods for approximating nonlinear models by linear models are presented.4. In most cases the following steps are sufficient: 1. Section 5.1 focuses on the writing and simulation of differential equations for the simplest causal classification. Section 5. An operator notation is introduced that allows linear differential equations to be treated like algebraic equations. 5. 281 .Chapter 5 Mathematical Formulation The writing and simulation of differential equations is extended in this chapter from the simple cases given in Chapter 3 to the much broader class of models considered in Chapter 4. and whether they are linear or nonlinear.3 introduces certain special features of linear models that are identified by their behavior as well as their structure.1 Causality and Differential Equations The application of causal strokes to bond graphs leads to a categorization of models according to the types of equations that they produce. and an efficient simulation program for linear models is accessed.2 focuses on more difficult cases. along with reasons why such approximations may or may not be justified.1 Applying Causal Strokes A certain procedure must be followed in applying causal strokes to every bond of a bond graph to assure that a subsequent procedure gives a proper set of state differential equations. and Section 5.

(III) That part of the bond graph that remains without effort and flow annotations has. Annotate the flow side of the bond with pi/~i or. (II) Annotate the fiow side of the bond of each element Ci with ~i. but one or more compliance or inertance elements is assigned differential causality. 4. a functional designation ei(qi). if nonlinear. and each compliance and inertance element is assigned integral cuasality. It is better if you use more specific symbols than the generic qi and p~ to represent the respective generalized displacements and momenta.4. as causal inputs. This procedure produces one of the following outcomes: Every bond. since these designations may still be valuable to you. Also. The procedure is a modest generalization of that introduced in Section 3. (I) Annotate the input effort on each effort source and the input flow on each ilow source with some appropriate symbol.2 Some bonds are not assigned any causality. These models are called under-causal. These models are called over-causal. (See below for the latter. Apply integral causality to the bond of one of the remaining unspecified compliance or inertance elements. which become the state variables. a functional designation (ti(ei). annotate the effort side of the bond of each element Ii with Pi. Repeat steps 2 and 3 as many times as possible. Every bond has assigned causality. transformers and gyrators. if nonlinear. Before you start you should remove any variable designations that may already be written on the bond graph or. only the input terms from step I and . if the conjugate variable is of no interest.4. and place a circle around this designation. Similarly. has assigned causality. and place a circle around this designation. Differential Equations for Causal Models Once causal strokes are placed on every bond. Apply any mandatory causal strokes that follow from the presence of 0 and 1-junctions. They are more difficult to treat than causal models. They are the easiest to treat. Annotate the effort side of the bond with qi/Ci or. or re-formulated to become causal models.1.) 3. These models are called causal. start with a fresh unannotated rendering of the graph. MATHEMATICAL FORMULATION 2.282 CHAPTER 5. you mayplace a large cross (×) in the appropriate place to register your disinterest. a four-step procedure produces a complete set of state-variable differential equations. o 5. They can be treated directly.

which demandsthe causal stroke as shownin part (b) of the figure. 5. 5. Note that the circled terms are "bottled up. For example. It is driven by an effort (voltage) source. This key procedure is highlighted in the examples below. The bond graph has two junctions. integral causality is placed on the iner- .5." and do not contribute to any further labelling. following the dictates of the causal strokes and using the power-convention half-arrows to determine signs. Propagate these variable designations throughout the bond graph. Next. if the circled term is bonded to a junction.1: Electric circuit example the terms q~/C~ and pi/Ii (or their substitutes). (IV) Write one first-order differential equation for each circled Oi and ibi. The right side of the equation comprises functions of the input and state variables as dictated by the annotations and power-convention half-arrows on the bonds in the immediate vicinity of the circled term.1.1. the right side comprises the sum of the properly signed causal inputs of the proper type (effort or flow) the other bonds on that junction.3 Case Study: A Linear Circuit Consider the electric circuit shown in Fig. CAUSALITY AND DIFFERENTL4L EQUATIONS (a) circuit (b) bond graph with first causal stroke (c) causal strokes continued (d) step II completed (~p/l q/C~(~ I Qlp/l 1 C (e) junction causalities employed 283 q/C~Q C (f) final annotated bond graph Figure 5. The circled term appears alone as the left side of the differential equation.1.

which is one of the few that you should have memorized.2) The first term on the right side of this equation is positive because the power conventionhalf-arrow on the center horizontal bondis directed towardthe junction.~-~q. There are two dependent variables (p and q) and the same numberof firstorder differential equations. Takingthe derivative of equation(5.2). (5.1) The circled term dq/dt is written on the left side of the other differential equation. This is always the outcomewhenthe procedures are applied to a causal model. and the modelis seen to be causal. The first-order state differential equations nowcan be written (step IV).5. As a result. 150). and the causal inputs of the attached 0-junction provide the terms on its right side: dq 1 1 ~ = ~p.~q. (5.Comparingthe coefficients. substituting the right side of equation (5. and the efforts and flows on the bonds for the C and I elemeuts are designated as directed. and the causal inputs of the attached 1-junction provide the terms on the right side: dp 1 d--~ = e .41) (p. The second term is negative because the powerconvention half-arrow on the R bond is oppositely directed. forcing admittance causality on the resistance bond. Integral causality nowcan and must be placed on the compliance bond.5. (5. This completes the assignmentof causal strokes. and the causal strokes around the 0-junction dictate that the efforts on the center and right bondsbe labelled with q/C. is q = Ce (whichis different from the 1-junction case). since both energy-storage elements haveintegral causality. Step III starts by observing that the causal stokes around the 1-junction dictate the labeling of the flows onthe left and center horizontal bonds as p/I. as shownin part (c). The steady-state solution. as shownin part (f). MATHEMATICAL FORMULATION tance (inductance) bond. Part (d) of the figure also showsthe application of steps I and II for the writing of differential equations. This also forces the causality on the bondto the right side of the 1-junction. . gives d2 q I dq IC-~_ + -~-[ + q = Ce. as shownin part (d). which emerges whenthe two derivative terms in the differential equation are set equal to zero.1) for the resulting term dp/dt and rearranging. = 1/v/~-~ and the damping ratio is seen to be (R/2)V/~ (which happen to agree with the 1-junction case discussed in Section 3. the input effort is designated as e. The presence of impedance causality for the R element forces this effort to be written as q/RC. p. 149).3) This is a classical second-orderlinear systemof the form of equation (3. the associated flow is markedwith a ×. This leaves only the flow on the R bond unannotated. as shownin part (e) of the figure.284 CHAPTER 5. The circled term dp/dt is written on the left side of one of the differential equations. the natural frequency is seen to be w. the equations are solvable as a set.

The bond graph below has been drawn as the first step in the analysis. CAUSALITY AND DIFFERENTIAL EQUATIONS 285 EXAMPLE 5. Steps I and II of the procedure for writing the differential equations. as shown below (and in Problems 3. which also becomes the flow on the bond for R. . Zo ~ foundation I= m. In the present case. The causal stokes on the bonds for the 0-junction mandate that the flow on the vertical bond be written as izi -p/I. the foundation itSelf has a known displacement zi = zi(t). The effort on the bond for R becomes R(~. and the resulting displacement of the machine. and relate zo to the state variables. Since there is no choice in any of the causal stokes. as given in Section 5. is to be found. As shown at the top of the next page. 131). and finally the ettbrt on the vertical bond and the bond to the right of the 0-junction becomes q/C + R(i~i .47 (p. As the second step.5. R = b I Solution: The causal strokes are drawn following the four steps given in Section 5. Se.1. C = 1/k.1.1. the model is causal. also are shown below as annotations on the bonds for Sf. C and I: The output variable is related to the state variables p and q by zi = f ~i dt = (1/I) fpdt. Integral causalities on the bonds for C and I produce no causal conflicts at the junctions.1 A machine of mass m is isolated from its foundation by a spring and a dashpot. Zo(t).2.p/I). find a set of state differential equations.28 (p. and 3. starting with the mandatory strokes on the bonds for SI and Se.1. 158)). 3.i-p/I).39 (p. define state variables. the lower 1-junction and its causal strokes force the writing ofp/I for the flow on the bond to the left of the junction. and therefore also are required.

5.-~p dp 1 R dt . In part III. The system is re-drawn in part (a) of Fig. as shown in part (d) of the figure.4 Case Study: Nonlinear Stick-Slip The stick-slip system analyzed in Section 4. 282) does not apply. The source S is neither an effort source nor a flow source. as shown in part (e). The bond-graph model shown in part (c) the figure is the same as given before. the causal stroke on the compliance bond now dictates the annotation of the efforts of the lefthand and center bonds.mg. MATHEMATICAL FORMULATION C q/C 1 R(~-p/I) q/C+SR ~(~i"~ -p/I)l ii "p/I q/C+R(zl 7 zi p/I -p/I) I mg ~ 1 t-~--S~ I The two differential equations result from the respective causal strokes and annotations on the bonds for the respective 1-junctions: dq dt = ~(t) . and the causal stroke on the inertance bond dictates the annotation of the flows of the center and right-hand bonds. the admittance causality of the nonlinear source gives a function in the . 266) is nowrevisited. Step 3 recognizes the causal mandate of the zero junction. and the flow on the resistance is also labeled as ~R.4. and the given linear torque-speed characteristics of the motor and the resistance (frictional) load are reproduced in part(b).1. however. one proceeds to step 4 and applies integral causality to the inertance. except for the added inertia. integral causality is applied to the compliance. 4.286 CHAPTER 5. to give the situation shown in part (h) of the figure. Step I of the procedure for writing differential equations (p. Step II annotates both the efforts and the flows of the energy-storage elements.-~q.6 (p. which also requires the use of a 1-junction to register the fact that the load resistance and load inertance refer to the same angular velocity. Finally.2. Finally. The model is now too complex for its behavior to be deduced directly. the causal mandate of the 1-junction dictates the causal stroke on the final bond.26 (p. leaving the bond graph of part (g). the effort on the source is also labelled as Ms. so step 1 of the procedure for drawing causal strokes (p. it is included in part (g) of the figure. as shown in part (f). 5. 267).Tp + Rib(t) . 281) does not apply. No such mandate exists for the 1-junction. differential equations are needed. In step 2. introducing the effect of the rotational inertance of the load that was neglected in Fig. For convenience.

..5. t°rque I [ . ~" (a)sy-- frictional loadu 1~1 Cs S~ s 0 1~1 ~ 1 1~1 g ~R CR / g ~ C 1 (c) bondgraph .o...) ¯ rR C I (h) applicationof junctioncausalities /I C I (i) fully annotatedgraph Figure 5......’. ¯ MsvsCs lb 0 ~0 X 80 speed.Ms=..load turn -II ¯ [ ~...~... 287 CAUSALITY AND DIFFERENTIAL EQUATIONS force[ 0.. 0~11 MR =MR(~e.... ¢c/C r ¢c/ M = (/)c/C ¢c/C ~ 0 ----~-~--~1 ~ ..¢c)(~pj~p/l ¢c/C ~---~.2: Stick-slip systemwith load inertia included .p/I.i p/I=CR ¢c/C1~.:... rad/s (b) torque-speedCharacteristics S"-~-~ 0 -’----’~- S I’-’-~’~ 0 "~’t 120 1 ~R 1 "----’~R C I (d) first causalstroke C I (e) causal strokes fromO-junction (f) causalstroke for the inertance ¢c/CI~ (~p/l _ C l (g) final stroke... andstep SI.1.rR _ S~ s T.or I nex’ble shaf~ ~ 8~motor. I ..inertive ~ ’....

(5. A corresponding MATLAB function file follows.1 second. I=0. ~R2=0. ~SI=0. ~=0. ~c~£on ~=stkslp(t .0.1.s~.10 ft-lb. Ms = ¢c’/C. (5. equation (5.’/R(q~R ) for its effort.1.x(: . ~R0=5.lb. f (2) =x (1)/C-~ ign (x (2)) *gR0+~Rl*abs(x (2) /I) -~R2* (x The model can be simulated and the output shaft speed ~c = p/I plotted for 0.~) These equations are solvable.0000~. The writi.9 ft-lb Ms1 = M~ = 0.s M~= 0.1.ng of the differential equations (step IV) now can proceed. are given by Ms= M~o. presuming the functions ~s(M8) and Mn(~n) are known.5) gives an integrable state-space formulation.4) Similarly.x) ~S0=10. (5. C = 0. The circled term ¢c is set equal to the sumof the efforts on the 0-junction: dec = (hs(Ms) dt .lb).~S2) ) "2+~0/~$2-x (1) -x(2)/~. ~s = Ss(Ms). f (1) =-~S1/(2.6a) can be solved for ~s: +s- 2Ms~ ~k~] + Ms~ CMs~ (5. and the impedance causality of the resistance gives function in the form ]I.M~. 6b) Mso = 10. The particular value of I given (I = 0. M~= M~($~).~p.6b) equation (5.00001 lb-ft s ~) corresponds to virtuMly negligable inertin. (5.7) Substitution of equation (5. the circled term ~ is set equal to the sumof the flows on the l-junction: ~ = . ~RI=0.M~$~. however.~S2) +sqrt (SS 1/(2. since the only variables are the state variables ¢c and p.x]=ode23(’stks~p’ .00086.00086 ft-lb.lb Ms2 = 0.0056.1.4) and equation (5.0 ft.1 rad/(ft.9.00001)) [0 0]) . with the following commands: [t.2/.6~) M~= sig~(p)M~oM~ Ip/Zl + M~ (p/I)~. as shown in part (i) of the figure.~/c . The source ~nd load torque-speed characteristics.M~. ~ = dt I" (~. ~$2=0. They are not readily solvable analytically. from a resting start.0056 ft. All the efforts and bonds are now annotated.0.~) Noting that Ms = M = ¢c/C.s~.288 CHAPTER 5. MATHEMATICAL FORMULATION form q~s(Ms) for its flow. p/ Z = ~. as plotted. due to the nonlinearities of these functions.7) into equation (5. p~ot(t. M~0 = 5.

9100 lb. 289 CA USALITY AND DIFFERENTIAL EQ UATIONS Plots of load speedvs time: 120 .ft.s 2 [ rad/?OI/~/A A A A /~//~ /~ ~ 0.ft.s 40 rad/s 2O .s ~ 0. seconds Figure 5.6 0.08 t.1.0 .0022 and 20.4 0.s rad/~o 0i i 0 ]! II 0..10 6O 2I = 0.5.s I = 0.ft.0022 lb.0.04 II II II II II 0..ft..s 40 rad/s 20 0 5O 2I = 0....00001lb.. seconds / 0..0021lb.001lb..ft. ~ I = 0.02 I| 0.06 0.2 0.ft.3: Simulationresults for stick-slip model 1.0100lb..8 t.

+MRo _ 25. 5.1 are in the notations for the efforts on the bondsfor C and R: C aC~q.1 (p. reveals a lowerlimit-cycle frequencyand a reduced fraction of time spent sticking. Ms~ (5.0022 ft-lb s’~.37. but with a very small effect of the inertia. for comparison. A further increase in inertia increases the decayrate for Oscillations without any effect on ~equib.. It is instructive to plot one of the state variables of a second-ordermodel versus the other. the sticking portion is seen to have virtually vanished. however. ¢~q~i~ =~/Mso V Mm. phase-plane plots of the torque Ms = ¢c/C versus the angular velocity p/I are given for the various cited cases.~ ~(t) - . aC"q+b(~. This is called a phase-plane plot.p/I ~ X ~agrq +b(~i-p/[) 2 mg I The differential equations therefore become dq_ dt 1 i p.8) Thusthe instability has abruptly given awayto stability. substituting the nonlinear constitutive equations ec = av/-~ for the complianceand eR = b~ for the resistance. EXAMPLE 5.01 ft-lb s2 showslittle oscillation. Solution: The only differences from the solution given for Example5.b(z~ -p/I)~. The exampleof I = 0. shown next.4. which is ¯ . Whenthe inertia is increasedto I = 0.i-p/I)2[~.3.0021ft-lb s’~.290 CHAPTER 5. In the present case. shownat the top of plot of Fig.19 rad/s. the underlying source and load torque-speed curves are shownby dashed lines. The corresponding result for I = .2 RepeatExample5. Whenthe inertia is increased marginally to 0. 5.~ (equation unchanged) ( d--i =.~vq . In Fig.i. reveals a limit-cycle behavior almost the same as deduced graphically in Fig. Timeis supressed in these plots. but the directions of increasing time of the trajectories is indicated by arrows.0010 ft-lb s2. . 285). MATHEMATICAL FORMULATION Theresult.the simulation reveals that the shaft speed converges toward the equilibrium speed. 4.

010lb ft s 2 motor \40 80 ~ speed. .1. rad/s 80 \40 ~ speed. rad/s .4: Phase-planetrajectories for the stick-slip system 120 .0021 lb ft _\ I 2 S - \\motor 8 \ 2 0 ~20 0 0 ~ speed. rad/s 120 120 10 8torque 6 lbft 4 \\motor = 0. ra~s Figure 5.5.00001lb ft s’- 2I = 0. 291 CAUSALITY AND DIFFERENTIAL EQUATIONS 10 \\motor 8 torque 6 lb ft \\ I = 0. i 80 ~ speed.

~ t)~ q: (a) transformer or ~!--.5 Case Study with Transformers and Gyrators The causal strokes on the two bonds of a transformer must be aligned the same. 5. The shaft in turn drives a flywheel. that is be symmetrical with respect to the element. also as shown.6: DCmotor driving a flywheel through a flexible shaft 5.q/C~ ~ /~Tq/GC~e/G-R~Tq/G C-~Te/~RIT q/G C] ~ Rfl C I (b) bond graph wi~ a~omtedvariables Figure 5. that is be asymmetricalwith respect to the element. aligned oppositely. This follows directly from the definition of the transformer.5: Causal constraints for bonds on transformers and gyrators / ¯ ~ ~off~’~)gear reduction compliance C DCmotorwith fixed field note: bearings not sho~ [ flywheel. 5. and similarly for the efforts. These causal constraints direct the propagationof causal strokes through transformers and gyrators in bond graphs. specifying the flow on one side also specifies the flow on the other side. or vice-versa. The causal strokes on the two bonds of a gyrator must be.5. as shownin Fig.1. The electromechanicalsystemshownin Fig.~ q: MATHEMATICAL FORMULATION e~ ~G e.Either both flows must be specified and the two efforts follow directly. A DCmotor with a resistance in the armaturecircuit and a fixed field drives a flexible shaft througha gear reduction.292 CHAPTER 5. el ~T e. .ine~ia I (a) system / R~ R~p /1 ~/I~’ : ~ ~ e--~e-R~Tq/GC~Tq/C~ .~ q~ G~ q2 (b) gyrator Figure 5. The non-idealites of the motorare ignored.6 illustrates the use of causal strokes with transformers and gyrators.

Similarly.2. You should carefully see that this happens. Step I of the procedure for writing differential equations (Section 5.1. The resistance R1 has impedance causality. it equals the difference between the effort e and the effort(~). Placing integral causality on the inertance bond determines the final two causalities.-~q .10) . The efibrt (~) is the causal output effort of 1-junction. annotating the efforts and flows on all the bonds according to the dictates of the causal strokes. Note that the power convention half-arrows determine the signs of the terms. Finally. or dp 1 -p" dt . The effort (~) is a causal output from the transformer.--~ q ) . to equal the effort of the zero junction minus the effort of the resistance R2. from the causal strokes and power-convention half-arrows around the right-most 1-junction. First. p. The resulting bond graph model is shown in part (b) of the figure. the effort on all the bonds of the 0-junction are labeled as q/C. This flow is the causal input to the left-hand 1-junction. the causal strokes on the gyrator bonds dictate the flow (~). Next. The circled ~b is similarly seen. to equal the flow entering the 0-junction from the left minus the flow exiting to the right. The flow (~) then follows from the causal strokes on the gyrator. since this is the causal input to that junction. starting with the two horizontal bonds off the 0-junction. consistent with its impedance causality. its causal output must be labeled as R2p/I. The differential equations for the system now can be written upon inspection of the annotated bond graph (step IV).--i (5. The circled 0 is seen.1. CAUSALITY AND DIFFERENTIAL EQUATIONS 293 along with the inertia of all parts except the relatively massive flywheel. from the causal strokes and power-convention half-arrows around the 0-junction. and is completed by placing an × on the effort side of the bond for the flow source. Step II places the standard notations(~and q/C on the compliance bond. The springiness of the shaft is assumed to be important. it equals T times the causal input effort on the right side of the transformer.9) d--[ = -~.e . 282) places the effort e on the left-hand bond. like its cau~al output is the effort (~). You now proceed to label the remaining efforts. or dq T ( R~T ~ (5. since this is the causal input to that junction. however. the flow on all the bonds of the rightmost 1-junction are labeled as p/I. and flows. The resistance R2 has one of these flows as its causal input. You may find this practice helps you handle confusing situations. step-by-step. demandsparticular attention. Placing integral causality on the compliance determines the causalities of all except the two right-most bonds. while the causal strokes direct their content.5. and therefore is written also on the bond for R1. Dashed lines with arrows have been placed on the bond graph to underscore the meanings of the causal strokes and the consequent sequence of the determinations of the variables written next to each bond. andGand p/I on the inertance bond.-] p. Step III. the transformer converts this to the flow (~). following a sequence dictated by the causal strokes. and is represented by a compliance.

R r 1 S s (~. As always.~. these are placed on the respective bonds. The conjugate variables for the I and C bonds also are shown. The 1-junction allows integral causalities for the compliance and the inertance.294 CHAPTER 5.(~) ~ T ~I C Solution: The causal stroke on the left-most bond is mandated by the flow source. define state variables and write a set of state differential equations.. (Parameters such as I. The state variables therefore are the displacement q on the compliance and the momentum p on the inertance..) The result is .G . which therefore must be chosen.3 The bond graph model below comprises elements with constant moduli. R. all uncircled annotations must include no variables other than the state variables and input variables. with dots to indicate differentiation and surrounding circles as flags for their significance. not variables.7 given in Section 5. R S~I . MATHEMATICAL These equations are solved for particular lem 5. FORMULATION parameter values in a Guided Prob- EXAMPLE 5.3. Apply causal strokes. as represented at the top the next page. Note the constraint on causalites imposed by the transformer. G etc. and the gyrator forces the causality of the next bond.(t) ~1 1 -IT p/I r~I q/CI @ C The causal strokes nowdictate the writing of the remaining efforts and flows. are constants in this problem.

1. Thenumberof independentfirst-order differential equations. equals the numberof energy storage elements with integral causality only. two state variable displacementscan be defined.Theyalso are sufficient. The energy stored in any elements found to have differential causality is not independentof the energy stored in the elements given integral causality.~-~P ~qo(t). Specifically.~. One of them must be assigned differential causality because of the causal constraint of the junction. although sometimesthis reduction is not practicable. junctions. In contrast.as shownin part (b) of the figure: C =C1 +C.¢t).6 Models Reducible to Causal Form. 5.~-~q. as shown in Fig. because they both are proportional to the same effort.7.1. to determine the energy stored in the inertances and/or complianceswith differential causality. the displacements on the complianceswith integral causality and the momentaon the inertances with integral causality not only determine the energystored in their respective members. As a result. This assumes that the standard procedure which favors integral causality has been followed. as shownin part (c) of the figure. This situation is easily simplified by combining the two compliancesinto a single compliance.. simplifying the subsequent determination of their differential equations. transformers and gyrators sometimesprevent one or more of the compliances and inertances of a particular system from being assigned integral causality.. and two differential equations can be . Order of a Model The causal constraints for effort and flow sources.5.Git. The two displacements are not independent.(tJ-RplTI-qlC . 295 CA USALITY AND DIFFERENTIAL EQ UATIONS R s~i OplTS . These over-causal modelscar/ be reduced to causal form. collectively.G plTl -i olTl . that is the order of the system. 5. Consider two compliances bonded to a commmon 0-junction.~. Both can be assigned integral causality. consider two compliances bonded to a common1-junction. the two energy storages are dependent. .p/TI ~1 T -~----~I q/C C from whichthe differential equations can be written directly: dq 1 dp 1 R G + dt ..

and therefore is preferred. expressedas a function of this commmon effort or flow. only one independent differential equation can be written. It is not necessaryor even . it is possible to combinethe two compliancesinto a single compliance:1/C = 1/C1+ 1/C. whereaswith a 1-junction. MATHEMATICAL FORMULATION (a) first-order withdifferential causality ~C2 e=el+e2 C=CI+Q ~ e ~ (b) first-order. As you have seen before. If this is done. are summedto give the energy of the combined compliance. In the first method. the order of the system is legitimately reduced by one. In either case the combinedcompliancecan be found either of two ways. the combinationrules for inertances are the dual of those for compliances. Their energies. Parts (d) and (e) of the figure showpairs of compliancesthat are less obviously combinablebecause of the presence of transformers. The second method tends to be more powerfuland general.1 in Section 4. 5. You can say that the minimumorder of the system(with the combinedcompliance)is less that the nominalor actual order of the system whenthe compliancesare treated as distinct. which then is combinedwith C1. two differential equations maybe written.7: Combiningdependent compliances written. the compliance C2 is combinedwith the transformer to give an equivalent compliancewith modulusC2/T’2 (as noted in Table 4. nodifferential causality 1/C=I/CI÷I/C2 (c) second-order. as shownin Fig. The contribution to the order of the overall systemby these compliances therefore is two.8. The causal strokes reveal the same pattern as before. Similar considerations apply to two inertances bonded to a commonjunction. In the second method.3). the efforts or flows of the two compliancesare related to a common effort or flow.q 2C=C~+C2/T T 1/C=I/C~+T~/C~ C~ CI (d) first-order withtransformer (e) second-orderwith transformer 1 Figure 5.2. Nevertheless.3.296 CHAPTER 5. however: with a 0-junction.those for a 0-junction in one case act like those for a 1-junction in the other case. although the numbermaybe reduced to one by combiningthe two compliancesand the transformer into a single equivalent compliance.~’T~qC 2 ~.no differential causality ~eo q ~ T ~ C~ ~ .

noting particularly which variable is commonand which variable sums. CA USALITY AND DIFFERENTIAL EQUATIONS 297 (a) first-order with differential causality e=e. In some cases the reduction of an over-causal model to a causal model is relatively difficult to accomplish. starting with the mandated flowsource bond.8: Combining dependent inertances desirable to memorize these rules.5.4 Determine the order of the following model that comprises elements with constant moduli: Sy-. T~----"I2 ~ 0 e q ~T r. +e~ I=4 +I2 ~ ~ ~ (b) first-order.-------~-R C Solution: Causal strokes are placed. You have the option of foregoing this reduction. EXAMPLE 5.12 ~ I/I=I/II+I/T~I~ ~q l=Ii+T2l2’ TdI~ 11 (d) first-order with transformer (e) second-order with transformer Figure 5.1-----. You should work out each case by focusing on its variables. One of the three possibilities is . noI differential causality e 1/I=1/11+1/I2 (c) second-order.. no differential causality e ~ 1.1.~ I .~. Any two of the three energy storage elements can have integral causality assigned. and proceeding directly to the writing of differential equations. and utilizing the constitutive relations of the elements.2. This option is discussed in Section 5. but the third must have differential causality.O -----’--~- Il T-------.

The model is called under-causal whenever the procedure fails to assign causalities for all the bonds. assuming the standard procedure which maximizes this number has beeen followed. They are therefore not independent. following the special causal rules.298 CHAPTER 5.- T I~---~- I ~---~-R C As a result. This procedure annotates the efforts and flows on all the bonds. Otherwise it is still possible to proceed. and the number of differential equations and the order of the model equals the number of energy storage elements with integral causality. Previously written annotations may violate these rules. simple examples have been given.-~-~. Definitions of variables using different symbols can be reintroduced after the differential equations are written. The signs are determined by the power convention half-arrows. and should be removed or set aside whenever this happens. 5.7 differential Summary A four-step procedure for assigning causal strokes has been given which is sufficient for causal and over-causal models. Onedifferential equation of first order results from equating the derivative dpi/dt or dq~/dt for each energy storage element to the causal input of its bond. MATHEMATICAL FORMULATION l ~’differential causality Syl--~ O I~--. before the equations themselves are written. the over-causal models are those for which one or more differential causalities result. Even if an annotated effort or flow is functionally correct. . as described in the following section.1. which also addresses under-causal models. Every compliance and inertance element in a causal model is given integral causality by this procedure. if it violates the causal strokes it will frustrate the purpose of the causal method. Another four-step procedure is given for subsequently writing a set of statevariable differential equations for a causal bond graph. there are only two state variables and two first-order equations. The model is said to be of second order. The energy storages with differential causality in an over-causal model can be expressed as functions of the same state variables in terms of which the energy storages with integral causality are described. It is advisable to reduce over-causal models to causal form whenever the reduction is not very difficult to carry out.

the net flow that fills the right-hand tank and the flow which leaves the drain.1. Suggested Steps: Identify five flows: the input flow Qi~at the left. the net flow that fills the left-hand tank. and write a corresponding number of state differential equations. Other moduli should be expressed in terms of physical parameters such as areas.9: Two-tank system for Guided Problem 5. Apply causal strokes to the graph. using the proper priorities. and a resistive drain from the right-hand tank. Label each compliance bond with t~i and qi/Ci. Propagate the causal inputs Q~n. 5. and complete the graph. Draw a bond graph for this system. Use these flows to establish a junction structure for the bond graph. where the subscripts are chosen to distinguish different elements of the same type. but note that this could be done. CAUSALITY AND DIFFERENTIAL EQUATIONS 299 Figure 5. Is integral causality possible for all the energy storage elements? Identify the order of the system.1 Neededpractice in finding differential equations for linear models is provided by this problem. and may be identified by unspecified R’s. You are not asked to combine these equations into a single equation of higher order.5. lengths and fluid density. Fig.9 shows two fluid tanks with an independent supply of liquid to the left-hand tank. using the causal stokes and power convention arrows to direct your work. Place × for the effort variable associated with the input flow Qin. the flow between the two tanks. an interconnecting tube with both fluid resistance and inertance. and each inertance with ibj and pj/I~. Identify the order of the system. qi/Ci and pi/Ii through the bond graph until all efforts and flows are properly annotated. . Write the differential equations. Circle the terms Oi and ~bj.1 Guided Problem 5. The resistances depend on viscosity.

3 Even if your instructor suggests that you forgo carrying out this more sophisticated problem. Attempt to add integral causality violation of causal constraints. employing the causal strokes in the usual 7. and express it as a function of the generalized velocity of one or both of the remaining inertances. Add 1-junctions.10.300 Guided CHAPTER Problem 5. Define state variables and write the associated set of state differential equations. and observe the 3. It involves the determination of nonlinear differential equation models for the mechanical snubber shown in part (a) of Fig. Reapply integral causality (it works now). MATHEMATICAL FORMULATION 5.4 (p. Augment the model of the hydraulic system of Guided Problem 4. Suggested Steps: 1. An important way to eliminate differential causality is demonstrated. 5. Write the differential manner. and fits closely with the glass cylinder to give little leakage. and the corresponding state differential equations are found. equations. Write auxialiary equations for the outputs Q and ~ as functions state variables. which allows the differential equations to be found. Computethe kinetic energy of the offending inertance. augmenting the resistance there appropriately. which is available commercially. 2.2 Dynamics are added to a steady-state model in this problem. inertances and resistances the three inertances and three resistances. Guided Problem of the 5. and similarly express it as a function of the generalized velocity of one or both of the same inertances. 225) by adding masses or rotational inertia and resistances to the rotary actuator and the two pistons. in case these are of interest. An air valve acts like a small orifice for out-flow (which . Computethe energy dissipation of the corresponding resistance. to the bond graph to represent to these inertances. augmenting the value(s) of the inertance(s) appropriately. The piston is made with a low-friction coating. Use this calculation to transfer the removed inertance to one or both of these inertances. and prepare the energy storage bonds with the usual notations. you should at least study it and its given solution. Use this calculation to transfer the removed resistance to one or both of these locations. 4. 6. 5.

so significant details are proposedbelow. which are not addressed until Chapter 12. Note: the symbolsign(y) maybe useful. (a) The bondgraph in part (b) of the figure neglects the compressibility of the air. the compressionmaybe considered to be isentropic.~ v/(2/p)lP. piston eval spring in chamber I ~ 1 ~ Cspr (c) model with air compressibility Re( l one-way adjustable orifice (a) system 0 -----~ C~ Rva! Figure 5. (b) The bond graph in part (c) of the figure recognizes compression the air. the fluid leakage across the piston by a constant resistance Rp (based on laminar flow). Find an approximate differential equation model.Pal for valve flow. The mechanicalrubbing can be characterized by a dry friction force F0.) Further. and find an approximate differential equation model. the nominaldensity of the air by P0. the effective area of the orifice for in-flow by A~and the area of the piston by Ap.Youare asked to find a differential equation modelwhich can give the position x(t) in response to an applied force F(t). equals +1 when y > 0 and -1 wheny < 0. since the interest is in pressures P which do not differ radically from the atmospheric pressure Pa.3 in practice can be adjusted). the mass of the piston assemblyby m. (A more accurate modeling requires considerations of compressible fluid flow.1. .10: Guided problem 5. the spring by its rate and free length x0. The air compliance is more difficult to implementthan most nonlinear compliances. Annotate the graph with symbols. the effective area of the orifice for out-flowby Ao.5. but a one-wayvalve (check valve) presents a much larger effective orifice for in-flow. It is suggested that you retain the approximatior/Q . 301 CAUSALITY AND DIFFERENTIAL EQUATIONS F drive rod (b) model neglecting Rs (air compressibility . Asoft spring fits inside the cylinderto resist compression.

PROBLEMS 5. 5. using integral causality. Write the two first-order causal strokes. Use an orifice equation to approximate the flow through the valve. Use this relation to express the effort on the 0-junction as a function of state variables. and write the corresponding set of state differential equations. directly or by substitution. 3. inverse need to differential equations. Apply causal strokes. q and p. Makesure the only variables that appear on the right sides of the equations are.1 Systems are modeled with the constant-parameter bond graphs shown below. with guidance from the the solution of the model with compression of the air. 6.302 Suggested CHAPTER 5. but with the sign reversing with the direction of the motion. Note that the assumption of an isentropic process means that the pressure in the chamber can be written as P = Pa(q/x) k. To start strokes applies. so that you do not invert the equation for leakage flow. Write a computationally realizable function for this force adjacent to the bond. Then.4 (not to be confused with the spring constant). . Apply causal strokes to the bond graph of part (b). define state variables in the conventional way and annotate the conjugate forces or velocities on the I and C bonds with the standard computationally realizable functions. The resistance Rfric acts something like a force source. 4. where q is the portion of the piston displacement associated with compression of the air and k = 1. add the flow for the leakage across the piston. depending on the sign of the velocity. Write all three state differential equations. Then relate this flow to a state variable. Define state variables. apply causal to part (c) of the figure and adapt the prior model to the extent Note that the causality on the leakage resistance element is the of what you used for the incompressible model. MATHEMATICAL FORMULATION Steps: 1. 2. Also. and invert to find the pressure as a function of the total flow. relate the force Fair to the pressure and thereby to a state variable. the excitation variable F and the state variables x. 7. get expressions for the leakage flow in terms of this effo}t and fixed parameters.

3 Answerthe question above for the bond graph shownabove right.4 Find electric circuits that correspond to the bond graphs of (a) Problem 5.5.1 for the bond graph given below: I ein S~ O "~"7" C G"--"-~ R. S~ el-~ i" 1 ~ 0 ----~R S]. I R~ 5.~-~. Theelectric current is specified as a function of time. 0 ~ 1 "-~-R I C C I 5.’/ Answerthe questions of Problem5.6 Write the state differential equations for the modeldeveloped in Guided Problem4. in which the causal input is rC 0 rR (a) Applycausal strokes. 5.2 Answerthe question above for the bond graph shownbelow left.8 For the bond graph below.2 and (b) Problem5.1 and (b) Problem 5. 5.5 Find mass-spring-dashpot systems with x-motion only that correspond to the bond graphs of (a) Problem5.:3. choosestate variables and label all the efforts and flows accordingly.:3. 303 CAUSALITY AND DIFFERENTIAL EQUATIONS 5. 5.1. . (b) Write the correspondingset of state differential equations.

(d) Express. Define state variables and write a set of first-order differential equations in a form suitable for solution. considering the fluid pressure. 47) employed as the dashpot in the system of Problem 4. any time constants.12 Consider the mechanical system described in Fig.24 (p. 299). 5. Write the state differential equations for the resulting nonlinear system. 201). Relate this to the state variables and the input variable. with primary interest on the response of the second compSance. in terms of the given parameters.14 Write a set of state differential engine" of Problem 4. 5. and (ii) electric capacitors resistors. Continue to ignore the inertias of the moving parts. The nonlinear functions may be defined as unspecified functions. MATHEMATICAL FORMULATION (c) The output variable of interest is ~0. 5.10 A simple I-C (or L-C) filter connects a voltage source to a resistive load.24 (p. You may employ parameters of your own devising if they are defined explicitly. 4. filter 5. Find the corresponding state differential equations in a proper form for integration. Find a differential equation relating the load voltage eL to the source voltage es. 5. as shown below. 5. e. 5.49 (p. P.g.304 CHAPTER 5. equations for the aircraft carrier "arresting . Find a commondifferential equation model. natural frequencies or damping ratios that apply. Assume constant parameters.11 Continue Guided Problem 4.5 (p. 227) by defining a minimumset of state variables.9 (p.9 Find systems with (i) springs and dashpots. 264) and the accompanyingtext. e0. 257).13 A hydraulic shock absorber of the type described in Fig. 2. to be an independent excitation and the motions of the solid parts to be responses. ]~/m:/~//ra(~m) or ~)m:~)m(Mrn).2 part (c) (p. that are analogous to the two-tank system of Fig.

167. but the power a reasonable-sized battery with solenoid or electric motorcan deliver is vastly less than the instantaneous powerneeded to drive a large nail in one motion. Carpenters often use pneumatically driven nailers for framing.37. roofing and finish nails. for otherwise the recoil momentum is too great.valving and mechanism details are not to be addressed. 3.The questions regard the size of the accumulator.. (See Fig.5.25 inches in length and 0. The oil also pushes on a piston which.) This presents an excessiveresistance to the flow of oil that surges out of the accumulator to drive the nail. the gas pre-charge pressure. Battery-powerednailers have been considered. drives the nail. I~ is proposed to use a small motor-driven hyraulic pumpto force 0il into a small hydraulic accumulator. Youare asked to investigate the problem and makedesign recommendations. The nails are the large 16-d size. whena trigger release is actuated. not shown.1. The plunger is returned to its initial position by a small spring. ~ three-way two-position valve restrlcuom~ ~ mOt ~~ pis!onOr i trigger [ [ i [ k~ ) pump ~ ~ reservoir 1[ ~ nail The principal problemis that although sufficiently small commercialbladder accumulators exist. Details about the electric motor.022 lbs in .15 Design Problem. ~ accumulator . they have a vMvemechanismto prevent the gas pressure from extruding the bladder through the neck of the accumulator. poweredby a large air compressorthrough a rather large hose.. p. the minimu~n allowable effective area of the flow restriction in question. 3. pump. and the diameterof the hydraulicactuator. Springs weigh too much.. CAUSALITY AND DIFFERENTIAL EQUATIONS 305 5. The nail must be driven homein a few milliseconds. It is suggested that a computersimulation wouldbe an effective tool in addressing these questions. Some~neans is need to store energy for quick release. whichis the subject of this problem. The energy storage of a compressedgas is another. Someof the details are suggestedin the schematic drawingbelow.Kinetic energy is one possibility.

/’Q:dt~ ~porousplugdrain Integral causality can be applied to the inertance and both compliances without producinga causal conflict. MATHEMATICAL FORMULATION weight. }~ and p.V::.306 CHAPTER 5. the charging pressure of the nitrogen (the pressure when no oil is in the accumulator) should be at least one-third of the maximumpressure to be used. .1 Q i~ A. to prevent excessive flexure of the bladder.~ "-Vl-"~. is three. Theorder of the system. It has been decided that the accumulator can hold a pressure of up to 3000 psi. They are resisted by a force that grows linearly with the penetration of the nail into the wood. Thestate variables are Vx. therefore. for a total work of 100 ft-lbs. Energy lost because of spent momentumand orifice flow must be kept small so that the battery and electric motor not become too heavy.dt ~ A2 :~ . It is suggested that the inertia of the nail and the other moving parts is important enough to be included in any dynamic model. Problem 5.~ -.-2. SOLUTIONS TO GUIDED PROBLEMS Guided 1-3.

-~l pl T~ t dp__~_dt= I PT2. dp~ __ dt = 1p_ R. I’3 = (T3/T2)213.2 3. Combine this inertance with I2 to give the overall inertance x~_=I2 +(T~/T2)2X~. Combine R~ with R2 to get R~ = -~2 + ¯(T3/T2)2R3 6. CAUSALITY Guided Problem AND DIFFERENTIAL EQUATIONS 5.1.2 1-2. Compare the two bond graphs above to get 1 Q= T~---~.~. 1~-’-.5.2 + 1 307 . T = -~I~¢3 = l~I3(T3Q2)2 = ~I3(T3 ~ 2]T 2) z ~13x2 Therefore. . T~ F~..~ P2 7.1 T3 M~ 1 I~-l~ ~ differential RI causality forced R~ 1 "2 2 1.

I=m 3. Fftic = -sign(~.~II=QIA~. dx 1 pdt m dPdt = F .. P = P~(q/x) Therefore.~( a .A’~-~. with the expressions given in t[ above step substituted for Fair and Q.+(x . R.) + ~(P- P~). R)~.xo) . volumeflow: ~C ~IF. Q = Ao~(P- P.P) .~o) .1] and Q = Ao + RpAp’ P>0 Q =-di V --~--~ ~ F~.. ~ = -A. . (P.xo) sgp(p)Fo .3 1. F~=A~(P-P~)=A~-~+V + 2p P ~ ~ + IReAe ~ ~ (-RvA{. FI 1 MATHEMATICAL 2. p<0 ~ +V~ 4..~.~ + R~. Therefore.F~ir dq 1 Q dt . p<O The differential equations are as given below.=-d~ ’ p >0 p . I~ FORMULATION 5.-]P Ap .)Fo = sign(p)Fo C=l/k.P). Let ~ > 0 for upwardmotion.[ ~. Differenti~ equations (note that p is downwardmomentum): dx 1 p dt m _1 @ F d~ = ~(~ . h =F+k(x-x°)+F°+Av~+ V ~ + A~I] p< 0 k6.F~ =F+k(x-xo)-Fo-Ap ~+ V 2p ’ >0 A~I] 2 a.. F~ir = ApPa[(q/x)~ .Ffric .F~ir = F + k(x .308 CHAPTER Guided Problem 5.

Dashedlines with arrows describe the sequence followed. In many cases of both types. is forced to have differential causality. the derivatives of the state variables lie only on the left sides.~ is expressible in terms of the state variables for the other energy storage elements. step II proceeds with the placement of notations adjacent to the bonds describing integral causality: 0 and q/C on the compliance bond and ~5 and p/I1 on the inertance bond. is shown in part (c) of the figure. Also. the energy in I. Special software is available. since it does not have the required integral causality. The situation at this stage is shownin part (b) of the figure.1. the equations can be reduced to a simple set of differential equations. Then.11. Step I starts with placing the notations O0 and x on the bond for the flow source. 282) are used.. First. Third. 5. The inertance I~ receives integral causality. however. 5. As a result. An example is shown in Fig. The problem of a mixed set of generMly nonlinear differential-algebraic equations (commonly known as DAE’s) sometimes must be faced.2.2 OVER-CAUSAL AND UNDER-CAUSAL Over-Causal 309 MODELS and Under-Causal Models Over-causal models are described by fewer differential equations than there are energy storage elements. After the mandatory causal stroke is placed on the bond for the flow source.2 (p. as shown in part (a) of the figure. The propagation of the terms 0o. a method for treating these situations is described which approximates an under-causal model by a causal model. direct use of the bond graph produces equations in which derivatives appear on both sides of one or more of the first-order differential equations. 5. with careful interpretation. permitting use of standard methods for ordinary differential equations. the causal output of the resistance is written as Rq/GC. Second. the gyrator causalities are implemented by writing the flow on the 1-junction as q/GC. Case Study Differential equations can be found directly from most over-causal bond graphs. The same four steps given for causal models in Section 5. Under-causal models produce a typically more difficult problem: some of the equations produced are algebraic rather than differential. integral causality is placed on the como pliance. q/C and p/I onto the remainder of the efforts and flows. The causal output of the inertance 1.2 is written as 12 .2. This placement forces the causalities of all the other bonds. The following presses ahead. step III. As a result. or whether to press ahead with the existing graph.5. this flow is replicated as the causal inputs to both the resistance R and the inertance I2. as noted above. but the inertance I. the remaining effort off of the 0junction is annotated as q/C. This completes the step. The care is in properly implementingthe differential causalities in the third and fourth steps. The modeler must decide at this point whether to find an equivalent bond graph with only two energy storage elements. The fifth step is the rearrangement of the resulting differential equations to insure that. and a fifth step is added. the inertance I2 does not participate.1 Treatment of Over-Causal Models. Fourth. the model is only second order. as mandated by its impedance causality.

310 CHAPTER 5.O .~.~ ~q/u~ /’T~ kx~// (c) ~lly annotatedgraph (step III) Figure 5.------~t G b---. MATHEMATICAL FORMULATION I Sf~--.~.11: Exampletreatment of an over-causal model . ~o t C ~ (b) a~ersteplI R e~- T ~ I~/GC.~O.------~I]~=.--~ l b---~-l: I (a) bond graph with causal strokes I C 1 R Sf~..-I 2 q°q/CIq(q @ C I ~P/~q/C-.IGI-.-.

for example.13) comprise the state differential equation model of the system. This is because different state variables are employed.12) The difference is that the derivative term appears on both sides of the equation.5. There is one potentially serious complication. This term is dictated by the causal strokes. or I2(t/GC. the element I2 could have been given integral causality. completing step III. The two state differential IV).~P. The propagation of the state variables can be carried out even when the elements are nonlinear. the effort noted as e (to save space) now is written as shown to the right of the graph. This rare situation sometimes can be avoided by proper choice of the energy storage element given differential causality. The equation for ib is equations for the system now can be written (step dp 1 dt .11) and (5. . the signs of the sub-terms are determined by the power-convention half-arrows.~q" (5. although it does not depart from the procedure. Finally. rather than the element C. This appearance of the time derivative of a state variable is peculiar to the differential causality. You need to be especially careful in these cases to follow the dictates of the causal strokes meticuously. This problem is rectified by collecting the two terms together on the left side: I~ ) dq 1+~ 1 R (5.11)~ The equation for q differs in an important respect: dq 1 1 1 I: dq R q" d-~ = (~o .-~e = ~o . Fifth.~P G’~C dt G:~ (5. This model is somewhat more complex than the corresponding model which results from subsuming the dependent inertance element into the independent compliance element. the flow to the left of the gyrator is written as e/G.13) This final step can be viewed as a usually simple step V that is necessary whenever derivative causality is implemented. however: if the causal output of an element with differential causality is a nonlinear function of the derivative of a state variable. In the case considered above.2. as shown. Equations (5. it may be difficult or impossible to carry out step V. OVER-CAUSAL AND UNDER-CAUSAL MODELS 311 times the time derivative of the flow.

Since the question asks for the displacementon the compliance.(t) C rlGI ~ 0 --~--~R Propagating the variables according to the dictates of the causal strokes gives the effort on the right side of the gyrator.The result at this point is I S~ e.-~ei(t) I dq q’ G~C dt 1 R~ . The differential causality on the inertance tells you that its effort is I times the time derivative of the flow. but not both. the modelis over-causal.5 Find a differential equation relating the displacementof the compliancein the modelbelowto the excitation ei(t).written here as q. the flow on the left side of the gyrator. it is simplest to makethis the sole state variable by assigning integral causality to the compliance.at first. I el(t) Se’------~ C r G . as shownbelow. 0 -~-’--~R Solution: Either the inertance or the compliance element may be given integral causality. dq 1 dt -. and the flow on the inertance.312 CHAPTER 5. or I(t/GC: The completed annotations then become I C I?I/GC Iq/GC q/CIQ q/C q/C ei(t) ei(t)-l[l/GC S~ 1 q/GCr I G lei(t)/G_l~l/G2 ~ 0 -~-.~-~IRq/RC Thesole differential equation becomes. All elements have constant moduli. MATHEMATICAL FORMULATION EXAMPLE5.

1. consistent with the usual rules. For each degree.2. t). which is the simplest type of DAEto solve. The step is tentative. The numberof virtual energy storages added by step 5 plus the numberof implementationsof step 6 (below) can be called the degree of under-causality.5. The first option (it makes . Collecting these terms.14a). described below. in this eventuality the step is aborted in favor of a Step 6. or a virtual compliance is bonded to a 0-junction that has some strokeless bonds. 281-282)leave four bondswithout causal strokes. using other junctions with strokeless bonds. as shown in part (a) of the figure. 281-282) leaves somebonds without strokes. one algebraic equation results. and propagating the causal strokes as far as possible. t). As Step 5.2 Equations for Under-Causal Models Under-causal models result when the four-step procedure for placing causal strokes (Section 5. Steps 1-4 of the procedure for placing causal strokes (pp. (5.12. The author advocates a special procedureto handle these cases.14a) (5. The literal answerto the question is this equation with both sides divided by the constant within the parentheses.1.ghey forma set of differential-algebraic equations (DAE’s)that are said to be in semi-explicit form: dx d-~= f(x. y. Step 5 is repeated. y. In somecases involving a bond-graphmeshsuch a conflict appears at a junction. 5.14b) Case Study The algebraic equations (5.2. 0 = g(x.14b) potentially can be used to eliminate the variables y from equations (5. so as not to change the meaningof the model.2. a virtual inertance is bondedto a 1-junction that has somestrokeless bonds. These added elements normally will be considered to have zero moduli. DAEmodelsfor whichthis reduction is possible are called index zero models. An exampleis shownin Fig. Takentogether with the differential equations in the model. The newStep 5 adds either a virtual inertance to the 1-junction or a virtual complianceto the 0-junction. until all bondshave causal strokes or causal conflict prevents further application.3 Algebraic Reduction Method. 5. Step 5 is completedby applying integral causality to the addedvirtual element. pp. OVER-CAUSAL AND UNDER-CAUSAL MODELS 313 in which ~ has been written with the more formal notation dq/dt in order to emphasizethat this term appears on both sides of the equation. contingent on it not producinga causal conflict. 5. that is to reduce the DAEto an ODE(ordinary differential equation).

f../ -pZ " Rz .12: Linear under-causal exa~nple 4o-..:..[ ~p/l I (c) steps I-II of procedureto write the differential equations (d) after step Figure 5.314 CHAPTER 5. MATHEMATICAL FORMULATION C~ (a) steps 1-4 for placing causal strokes (b) step 5 with virtual inertance T T qo /.

. This fv is then substituted where it appears in the differential equations..165) EXAMPLE 5.5. All four bonds given causal strokes as a result.~p.15b) (5. a circled 0 is written on the effort side to represent its zero effort.00 fvT R2( ~)’ ~) q_ -~ R1 f~. (5.2[1 ~lo. (t°dt dP=R’2( (5. The equations for the two real and one virtual now written (Step IV) energy storage elements are dq d-~ = fv. can be solved for the flow fv in terms of the input variable 00 and the state variables q and p. The final result is the ODE dq T T q] dt . and after part III of the procedure is shown in part(d).15c) which.I + R1T2/R. The four-step procedure for writing differential equations for causal bond graphs (pp.. All moduli are constants.15a) fvT 0=~. 282-283) now is implemented. Iv. and a symbol fv is written on the flow side to represent its non-zero flow.15c) The virtual inertia has given algebraic equation (5. 315 OVER-CAUSAL AND UNDER-CAUSAL MODELS little difference) is implemented in part (b) of the figure.6 Write a differential equation for the sole state variable in the following model. since it is linear. on the effort side of the virtual inertance bond and pv/Iv on the flow side. The result after completion of step II is shown in part (c) of the figure. in place of writing a circled/). Specifically.16a) (5. with a critical difference that recognizes the zero value of the added virtual inertance.~C ’ dp _ T RIT 1 ] RIToo -[-p + -~q ¯ dt 1 + R~ T2 / R2 [ (5.2. the degree of under-causality is one.

and the effort./G~e~. due to air pressure underneath caused by a large fan. without direct contact.316 CHAPTER 5./G R~ Setting the encircled zero flow equal to ~he sum of the input flows to ~he O-junction gives ~he algebraic equation 1 Rx 1 1 which can be solved for ev.~ G 2~ e~]~ . different methods usually are needed.= = dt 5.2.T e~tr lei/G_Rle~. is written as ev: The annotations on all the bonds now can be completed to give the result e. If the craft is displaced downward~the restriction . however.13. which is the non-zero ratio of a zero displacement to a zero compliance. A nonlinear example is the ground-effect machine or GEM(sometimes called a Hovercraft) shown in Fig. A rigid skirt around the periphery constricts the outflow of air from the interior plenum. _ e~(t) ei-Rle~.. 5.4 Differentiation e~ R~/G ~ ~ 1/~ Method. Multiple linear equations also are solvable if they are non-singular. Setting the encircled ~ equal to ~he effort of ~he O-junction gives the answer -. MATHEMATICAL FORMULATION Solution: Only the left-most and the right-most bonds have mandated causal strokes. In other cases./G _ Se~ l ~. Doing the latter. You have the choice of adding a virtual inertance to the 1-junction or a virtual compliance to the 0-junction. This vehicle hovers over land or especially water.15c) is solvable for fv because it is linear. so the model is under-causal. or chamber./G i p’-~l Rae~. Case ~(t)- ]P " Study* Equation (5. the flow on the virtual compliance is written as a zero and is encircled.

13: Primitive GEM for vertical stabilization . kPa section A-A 2.0 pressure.Om 80 40 0 0 0.~ ~0~.¢-a--~.5 2.0 1.0 (a) systemand fan characteristic (b) system model (c)causal strokes after step 4 S (d) placementof virtual complianceand R annotation of bonds P P ~l-. 317 OVER-CAUSALAND UNDER-CAUSALMODELS 2OO top view A fan A 160 volume flow 120 m3/s lO.5 3.5. Qslr. y) Se P/T T~I~I Cv Figure 5.5 1.2.

(5. (5. as plotted. as described later.0×10 a4 = 1. increasing the pressure and causing the craft to rise. 282) also are included in part (d). step 5 (p..Qs. dt dy 1 ~ = ~P. to the 0-junction. The GEMpictured. a1=0. ~.17b) 2 where the pressure P is in N/m (Pa) and the flow Q is in m3/s. and y is the eleveition of the bottom of the skirt above the surface.rag. by adding a virtualcompliance C. and is neglected. does not correct for the tilting motions of roll and pitch.Q8 .19b). . where A = 40 m~ is the area of the vehicle.a~ P + a~P"2 P4. QI . and the resistance to skirt flow by a resistance. 282) ends without designations for the source and resistance bonds. The weight Of the craft is represented by the element Se. The fact that the resistance R is a function of y makesy a state variable.65 represents a discharge flow coefficient.a4 (5.17a) Qo =180. attention is directed toward vertical translational motion only.08m 5/Ns. 313) is invoked. so that the excess of the fan flow over the skirt flow. The fan is represented in the bond graph of part (b) by a general source. a~=2. The pressure-flow characteristic of the fan. can be }epresented by the equation Q f = Qo . Therefore. The relation is represented in the bond graph by the transformer with modulus T = l/A. and requires the inclusion of equation (5.19b) 1 0 = Qf . as shown in part (c) of the figure. as shown in part (d). Steps I-III for writing the equations (p. MATHEMATICAL FORMULATION for this airflow increases.alp + a2P2 . with a plenum depth of 0. m3/s.19a) (5.~-~p. (5. The effect of the compressibility of the air in the plenum is slight. The vehicle shown measures 4 x 10 meters. The coefficient Cd = 0.318 CHAPTER 5. p.a4P4 .8 meters. is a state variable.Lycd . and its mass by the element I = 4000 kg. Step 4 of the procedure for placing causal strokes (p. The pressureflow characteristic for the skirt flow can be approximated by applying Bernoulli’s equation: Qs = LycdI~P. produces the vertical velocity of the vehicle.~--~p = Qo . (In general. -SmT/N2s. The resulting equations become dp = PIT .19c) The momentum of the vehicle.23 kg/m3 is the density of the air. with its single plenum. state variables are contributed both by energy-storage . (5. For now. embellishments to the design are needed for these. and p = 1.482 × 10-12 m9/N3 s.18) The periphery of the GEMhas length L = 28 m.

2.2a2P + 4a4P3 + Lycd/ 2~-~ " (5. As a result. meters 0.) Equation (5. No other value has a legitimate meaning.19b). and P is not a true state variable. Algebraic equations become differential equations upon differentiation.0001 prevent division by zero) and P(0) = 2000 Pa. despite the minimumorder of the model being two.19c) theoretically can be solved for the pressure in the plenum. . An analytic solution of such a high-order polynomial equation is not known.14 for a simulation that starts at lift-off ("differentiation model"). y and P. The three dependent variables are p.2 ~~ 0. Note that the number that represents the fan flow at P = 0 (180 m3/s) was lost when equation (5. This differential equation can be used as the third in a solvable set of ODE’s. permitting only two independent initial conditions. the more difficult a DAEis to solve. seconds Figure 5. regardless of the method used. y(0) = 0.19c) cma be solved for readily for dP/dt as a function of p. The higher the index.19c) can be solved for dP/dt as follows: dR LCdx/~--~(dy/dt) + (1/TI)(dp/dt) d---~ = al . given the values of p(0) and y(0).Qs = p/TI. 5. and the result substituted into the differential equations.14: Three simulations of the lift-off of the single-plenum GEM elements and by resistances and transformers that are functions of displacements.19c) differentiated.3 elevation. the DAEmodel is officially of index zero.20) A solvable third-order ODEresults when this equation is combined with equations (5.1 0 AND UNDER-CAUSAL 0 mpliance) 1 2 time. The index of a DAEis defined as the number of differentiations theroetically needed to solve for the time derivatives of the variables represented by the algebraic equations. since the model has minimumorder two. elevation y is plotted in Fig. This value is mandatory. Therefore P is treated as though it is a third state variable. P. x and P.19a) and (5. in terms of the net fiow Qf .5. its effect is recovered through the use of the proper initial condition P(0) = 2000 Pa. other cases of the latter are illustrated in Chapter 9. however. The derivative of equation (5. OVER-CAUSAL MODELS 319 five times actual compliance actual compliance 0. The initial conditions are p(0) = 0. The derivative of equation (5.

Case Study Continued* Regardless of the index of a DAE. demonstrating that compressibility is of virtually negligible effect. even if this phenomenon is not of interest. A trade-off between accuracy and computingtime is characteristic of the methodof non-zero virtual complianceor inertance.19) are replaced by dp 1 -~ . dy 1 (5.LyCd. 5. A faster simulation with five times the proper complianceis included in Fig.4). on the flow side and q. It is barely different from the solution with Cv = 0. Simulations must proceed at ~ rate dictated by the need to have manytime steps per cycle of the fastest phenomenon in the model. this compliancewould have the approximate value Vo/kPo = 0.320 CHAPTER 5.. Po is the meanabsolute pressure (virtually 101 kPa) and k is the ratio of specific heats (1. resulting in an actual compliancein place of the virtual compliance.14..21a) (5. so any ODEsolver can be used. (5. Treated as a constant. 5.) Pressure excursions from the nominal movingequilibrium have very rapid dynamicsrelative to the basic motion of the vehicle. the annotations on the bond for the virtual compliance C~are changedto ~.14. due to the smallness of C. would recognize the compressibility of the air in the plenum.215) ~-~ = ~p. dqv d---~ = Qo-al--~q. P = qv/C~.265 × 10-3 mS/N. The execution can be speeded up by using a larger value of C. although it might be acceptable for somepurposes. but at a cost of inaccuracy. qv . the simulation is not excessively slow for this simple model. however. due to the meanelevation y of the craft)..21c) 3 2 a3 2~/-~_~ The principal questions noware what value should Cv be given. however. where Vo is the nominalvolumeof the plenum(a little greater than its minimum value. it so happens. Therefore./Cv on the effort side. Modelscombiningrelatively very r~pid and very slow phenomenaare called "stiff. like any real compliance. Executionof the simulationis relatively slow.it is possible to generate an approximate ODEsubstitute by imbuing the virtual complances Cv and inertances Iv (in the step 5 above) by small but non-zero values. and howsignificant an error does this complianceintroduce into the modeland its solution? A more accurate model of the system. and equations (5. . Eachalgebraic equation is then converted to an approximate differential equation. The result of a simulation with this modelalso is given in Fig.1~-~P.-C-~q~ ." Multi-step integrators typically are superior to single-step Runge-Kutta integrators for handlingstiff models. +a2(D C2v -~ .2. MATHEMATICAL FORMULATION 5.5 Methodof Non-ZeroVirtual Energy-Storages. (In absolute terms. The error is significant.CvTq~. For the current example.

The MATLAB integrator odel5s is a variable-order integrator for "stiff" systemsthat you can readily use to handle 4a class of DAE’s.2. the offending virtual element must be removed. North-Holland. Petzold.7 Case Study With Meshes* Somebond-graph meshes.) Moregenerally. 3 These programswork well for most index zero and index one systems. however. eds. Petzold. Some different junction should be chosen. as described in Chapter 10.2.two new initial value ordinary differential equation solvers" ACM-SIGNUM Newsletters. . 1980.5. to retain the original state variable by ~selectinga different virtual element. "LSODE and LSODI. North-Holland. Campbell and L. Rolling and side-ways motions are omitted. however.L. Elsevier. 10-11. 248).R. pp.18 (p. you probably should avail yourself of a software package designed 1 and LSODI. (Care must attend treatment of the power-conventionhalf-arrows should they not be aligned initially in the mannerillustrated.2. v 15. Amsterdam. "Solving Index-1 DAE’sin MATLAB and Simulink. 41. an even or odd meshcan be re~novedby reformulating the modelwith different state variables. if this is possible. Occasionally. M.15 illustrates the problemand its solution. Scientific Computing.however. For simple ever~ meshesa sophisticated methodsubstitutes a tree-like equivalent bond graph. pp.S. Numerical solution of Initial-Value Problems in Differential-Algebraic Equations." A description of DASSL: A differential/algebraic system solver". They are beyondthe scope of this book. 1983. as given in Fig. no such cleverness is attempted for a complexbond graph with a mesh:the basic methodfor finding differential equations is applied directly to the given graph. Kierzenka. Step 5 Of the procedurefor designating causal strokes (the introduction of a virtual complianceor virtual inertance whena modelis under-causal) sometimes produces an uncorrectable causal conflict when a bond-graph mesh is present. 3K. and someindex two systems. 538-552.E. an apparent causal conflict can be corrected by switching the causality of an energy-storageele~nent from integral to differential. Stepleman et al. Extensive discussions of these formulas and the programsare available. so there are two "degrees of freedom": that 1L.whichfocuses on modeling rather than numerical methods. Shampine. In these cases.6 Commercial AND UNDER-CAUSALMODELS Software for 321 DAE’s The solving of DAE’sin general is a major enterprise. 65-68. Hindmarsh. produce special complications. Sophisticated or basic methodscan be used to overcomethese complications.Brenan." SlAMReview. Youmight prefer.1989. the proper virtual energy-storage element appended.W. based upon examinationof relations for the stored energy.R.F. 2A.C. pp. Reichelt and J. S.. R. For the present. explicitly for them. 5. v. Should you deal with themoften. 5. The classical modelfor the small-motiondyna~nicsof an assumedrigid vehicle shownin part (a) of Fig. 4.A. 4L. New-York. Twopresently popular packages are DASSL 2 These are based on multistep backward-differentiation formulas (BDF). or closed loops of bonds. OI/’ER-CAUSAL 5. 1999.and the resulting graph checkedfor causal compatability.

..~.- 1 R2 (Can add gravity force or weight. ’’’’~ S~ 0" (c) bond graph before choice of unspecified causality C~=l/k~ . MATHEMATICAL FORMULATION sc~ L2--~1 0 Tl / .15: Model of vehicle dynamics with bond graph representation ./.. "x"~" T2 ." S~ l p*/T3I~’ Tz (~IP~/1:. R~=b~. R2=b2 Figure 5.r L2 T3 = L1 +L2 ly2 0" (b) junction structure with mesh r~ o q~ /Cl~O 1-.322 CHAPTER (a) 5. C2=1/k2 . q2/C2~ 1-~.r~--0~ R~ 1 IT.) l¢=J ..-0-~. Ira=m.but this is just a constant and doesn’t affect the dynamics.

are not mandated by this causality. which are the vertical displacements of the two axles. respectively. b~. The suspension at each axle (springs.16. with parameters k~.5. These are T=1 ~ " ~mv~ + ~ J¢~.. where q~ and q2 are the extensions of the springs from their nominallength: q~ = y~ .23) 1 ~ 1 k ’~ (5. are moreconvenient. The causalities of manybonds. m. since the kinetic and potential energies are morereadily expressed in terms of these and their time derivatives.2. a simpler methodexists..and the pitch angle.22b) These lead directly to the junction structure of the bond graph shownin part (b) of the figure.indicating an under-causal model. Therefore. (5. including the mesh bonds. however. relative to somereference level. However. however. two generalized displacements and their derivatives are required to specify the state of the system. ycm. the time derivatives of these displacementswill be co~{sideredto be the excitations of the system. this approach is more complicated than necessary.24) ~ = ~k~q~ + ~ ~q~. (5.22a) +~~’L~ +L~ 1 ~ . Expressions for the kinetic and potential energies of the modelin terms of the defined displacements and velocities are written ~ the next step in the modelingprocess. These allow the bondgraph to be completed ~ shownin p~rt (c) of the figure. a virtual energy-storage element is to be bondedto one of the causally incompletedjunctions.y~l.Y~. . ¢. Althoughacceptable.the vertical displacement of the center of mass.25b) q2 = Y~. The two displacements could be yl and y. The ground under the front and rear axles has vertical displacementsyg~(t) and yg2(t). This conflict is correctabl~ by switchingthe causality of the I¢ elementto differential form. The vehicle has a knownmass. revealing that the modelis indeed fourth order. and knownmomentof inertia. ~ illustrated in part (a) of Fig. (5.a causal conflict results. tires and shock absorbers) is modeledsimply (and rather crudely) by a parallel spring-dashpot combination. ~bout a known center of The geometric constraints betweenthe velocities of the ~les and the center of m~s are L2 L~ Vc~= ~~L~+ (5. OVER-CAUSAL AND UNDER-CAUSAL MODELS 323 is. as Msoshownin p~rt (c). k2 and b2. as shown.L~ + Le(92 . Integral causality is applied successfully to all four energystorage elements. 5.25a) (5. If you try to appenda virtual inertance to either of the mesh1-junctions.9~). J.

16: Continuation of analysis from Fig.15 . q~I ~ 1 1 ----~- S: 0 -~.TpmI1~.1 R~ 0 (b) addition of virtual T3 compliance to a mesh bond ~[p~/l~ .’T" (a) additionof virtual inertia with attempedretention of integral causalities R| 5... ~0 ~ note: ’~.~-I 1 qil e2_ I. T1 q.~-- FORMULATION q2 .26) Figure 5. C~l Cv~ v" 11-~---L2 0 ql -.-¢-~1 RI MATHEMATICAL 10 __ IC~ 1 ~-=./T31¢. ~~Tle C~ T..CHAPTER 324 G ~-[[ 1~0.0 ~-~---~ 1 R2 1 SI I \ ~.--~ 0 ~ 1 s. ~e.+T./C~(~ ~Y 0 -.C./1~... 5.. (c) addition of second vi~al complianceto complete causality --~ +il~ Note:for~and~ R~R~C~ seeequations(5.. T3 ~’~[p -~_ . .=I 1.

OVER-CAUSAL AND UNDER-CAUSAL MODELS 325 As a second attempt.29a) (5 ¯295) TaIo I. but still leaves the bond graph under-causal.27).29b) to eliminate f2.26b) for f. 0 = A Tlpo (5. as desired: el = R2~g2 + Rl~91 + (TzR1 .26a) el + e~.. equation (5.26b) R~. and the efforts and flows on all the bonds can be annotated in the usual manner. which employe~ and e~ on their . and using the fact T1 + T2 = 1 (5. a virtual compliance is added to the upper mesh 0junction in part (b) of the figure.T3I---~’ ¯ f2=-yg’.5.#g~. (5.27a) Pm + T2fa im ¯0 =Tifi (5.2.26b) can be substituted into equation (5. as shown. This produces no causal conflict.) and noting equation (5. R~ q2 (5.T~R2)p¢ T3I¢ Finally. giving el = R2~g2 -- e2 q2 C2 TIR2P¢ T31¢ R2Pm (5¯31) Substituting equation (5.30)therein givestheeffortel in termsof the state v~riab]es. it is helpful to define the flows fl and f2 as functions of the state variables and the efforts el e2 of the virtual compliances: 1 fl : f2 . the four state differential right sides. a second virtual compliance is added to one of the remaining 0-junctions with incomplete causality.~_" The algebraic equations associated with the circled zeros on the bond graph are q~ eu 0 ---fl --~ii+ /~IC-~--~ .32) C1 C2" equations. as desired: ~2 _ T~R~p¢ T3I¢ R~pm q~ + ~ + ~- ~l~gl- (5.28) reduces these equations to pc T~Ie~ 0=~.26a) into equations (5.~ " Solving equation (5. substituting equation (5. are (RI + R2)Pm Ln ql q2 (5.) q~ e2 ~RI R~ C~ ~gi.C.30) To find the corresponding equation for e~. as illustrated in part (c)¯ Now. Thus. To reduce the clutter on the graph.the causal strokes are completed without conflict. (5.27b) Substituting equation (5.27a) for e2.28) gives the effort e2 in terms of state variables.

EXAMPLE 5.33d) The choice of variables above and the resulting differential equations are not the only ones possible.[~= causality (iii) C ’~qi’(t) 0"-~ 1 ~. saving effort at the cost of increased required competence./l differential .. Finally. The mandated causal strokes are as follows: (i) . over-causal or under-causal.33a) (5. 326 MATHEMATICAL FORMULATION (5.7 Determine whether the three mesh-type bond graphs below are reducible to tree-like equivalences.33c) (5. the meshes are odd and not reducible to tree-like structures. define state variables and write the corresponding set of state differential equations for case (ii). Then.CHAPTER 5. A recast model presented in Chapter 10 for this problem has no mesh.3ab) (5. determine whether each of the graphs is causal. Therefore.-" / 1 S~ 0 ~ (ii) ~ 1 )SY~ 0 ~I } I~ C } unspecified caus~ strokes R 1 ~strokes } u~quely specified I . i ~ (iii) C 1 C~ 1 1 1 ~ (ii) 12 I R Solution: Three of the mesh bonds have clockwise power arrows and one has a counterclockwise power arrow in all three cases.

p/l ~ e~) "~ l ~f C. The differential causality on an energy-storage element implies that its energy is not independent of the state variables for the other energy-storage elements. as shown below right.2../R-q/RC I. this time placing a virtual compliance on the left-hand 0-junction.+ n--~ .7’ which gives q Rqi(t) e~ = -~ + 4 The state differential dq 2e. so the model is causal.~R R The algebraic equation associated with the circled zero is 4e~. An attempt place a virtual inertance on the lower 1-junction produces a causal conflict._q/ci2e.2. several bonds have mandated causalities. as shown below left.. In case (iii). but regardless of the choice only one of them can have integral causality. as shown. 2q p 0 = q~ (t) .-~p.. The basic 4-step methodfor writing differential equations still applies. 2e. Time derivative terms appear on both sides .8 41" equations become -~O~(t) q_ 1 R = ev C ~q + -~?li(t) R .-q/C~ S~ T-_~.e.~-e~.. so the model is over-causal (and of first order). but requires special attention to insure faithful application. OVER-CAUSAL AND UNDER-CAUSAL MODELS 327 In case (i) the causalities of the two inertances could be inverted. Therefore.5. A second attempt. the efforts and flows on this bond graph are annotated in the standard fashion. Summary Over-causal models are those for which it is impossible to apply integral causality compatibly to all the inertance and compliance elements. The under-causal case (ii) is the most difficult to address.. In case (ii).~ 1 7 dp d-[ 5. so the model is under-causal. produces a satisfactory result.-h-. both energy storage elements have integral causality and no bonds are left unspecified. C q/C~(~ C I" 1 ~conflict S~ 0 V""~" -)’~~ "~1 0 "~1I 0 ~2e~’/R-q/RC+p/I’~O~’-~I ~’..

328 CHAPTER 5. Start by drawing a detailed bond graph for the system. Then. and defining the moduli of all elements. Care must be taken to employthe proper initial conditions fcir the added differential equations. Guided Problem 5. this step sometimes produces a causal conflict. Special software is widely available. and comparewith the first set. Then apply causal strokes. Suggested Steps: 1. called a DAE(differential-algebraic equation). In most cases these terms can be collected on one side.17 is used here to illustrate the two basic methods for handling dependent energy storages. find a bond graph which has only one inertance. The result is a set of differential equations equal in number to the number of energy storage elements with integral causality plus the number of resistances or inertanccs that depend on a displacement that is not otherwise a state variable. In a more accurate procedure. 5. If they are so solved. The smaller the moduli chosen. DAE’sare more difficult to solve numerically than ODE’s(ordinary differential equations). the more accurate the approximation but the more difficult the solution. Whena bond-graph ~nesh is present. MATHEMATICAL FORMULATION of the differential equation that involves an element with derivative causality. and note that differential causality appears for one of the inert’ances. Should this happen. and the equation solved for the derivative. using the standard method. or a virtual inertance to an affected 1-junction. algebraic equations within a DAEset are differentiated to give differential equations. a different junction must be chosen for appending a virtual inertance or compliance. because of the zero value of its modulus. A fifth step appends a virtual compliance to an affected 0-junction. as an alternative approach. Proceed to find the corresponding differential equations. the model comprises a set of one or more algebraic equations as well as differential equations. Alternatively. equations is simulated with a simple ODEintegrator. it is suggested that the angular . Otherwise. One algebraic equation is produced for each virtual energy storage element in a bond graph. the writing of differential equations can proceed normally.4 The system of Fig. Carry out the first part of the solution. also find its differential equations. The algebraic equations emerging from step 5 may or may not be solvable for the defined effort or flow in terms of the input and state variables. including the MATLAB integrator odel5s. Under-causal models are those for which the basic four-step procedure for placing the causal strokes (pp. 281-282) fails to define all of them. Simulation based on multistep BDF algorithms (backward-differentiation formulas) usually work better than single-step RungeKutta algorithms. it is possible to approximate a DAEby an ODEby conferring small non-zero values to whatever virtual compliances and inertances have been introduced by the procedure for placing causal strokes. and the resulting expanded set of differential.

Are the differential equations the same? Are they equivalent? Which set more clearly reveals the behavior? Guided Problem 5.4 velocity ~ describe one of the 1-junctions. its effort should be labeled as the product of the inertance and the time derivative of the flow. OVER-CAUSAL AND UNDER-CAUSAL MODELS 329 Figure 5. 5. Re-do the bond graph with only one of the inertances. Find the differential equations. and then write the differential equations. Whenthe GEMrolls to the right. write an expression for its stored energy in terms of its velocity.5.2. To get the correct sign for the force source.4 and 5. to the vehicle.18. as shown in Fig.17: System of Guided Problem 5. and that the two radii become transformer moduli. Note that the inertance bond with differential causality has no state variable.5.2. These pressures impart a correcting moment. You are asked to model the system with a bond graph . To find what this equivalent inertance should be. note whether the input work is positive or negative for your defined velocity.5 This relatively complex problem compares the the use of virtual energy-storage elements with the differentiation method in the treatment of under-causality. It extends the modeling and analysis for the ground-effect machine addressed in Sections 5. Or roll stiffness. The lower edges of these skirts are in the same plane as the external skirts. the resulting increase in the resistance to flow for the right-side skirt and the decreased resistance to flow for the left-side skirt produces a higher pressure in the right-side plenum than the left-side plenum. and compare to an expression for the sum of the original two stored energies also reduced as a function of the same velocity. Annotate both sides of all bonds. Roll stability is imparted to the GEMby introducing two internal skirts.2.

The effort required for you to carry out the details may not be justified.330 CHAPTER lO. 1. "tbu will see that the algebraic equations are too complex to permit direct reduction of the model to a minimal set of state differential equations. partly to determine the equilibrium elevation. 2. Definevariables omitting compliances. The physical compliances and five times these values are employedseparately.Om i 5. Each of the three plenums merits a 0-junction. the accuracies of the simulations and the computational time required. Note that the model is under-causal.18: GEMwith internal skirts plenum for Guided Problem 5. Apply the first four step. other parameters are as before or as shown in the figure. that ignores the small compliances of the plenums.5. The steps below and the solution at the end of the section outline the procedure and describe the solutions. the behavior following an initial roll velocity of 0. Draw a bond graph for the system. Carry out step 5 of the procedure for causal strokes. 3. which illustrate important conclusions. but does not modify the model.05 rad/s at the equilibrium elevation is simulated. The differentiation approach is more difficult and also increases the order of the model by three. The radius of gyration of the GEMin roll is taken as 1. The two models are simulated first with no roll angle. increasing the order of the model by three. 4. and proceed to develop the corresponding describing equations. Next. MATHEMATICAL FORMULATION ~ ght plenum ]-~-~ ~ enter : plenum i _.~left skirts~ top view section A-A ~-j A Figure 5. but it is instructive for you to establish a plan. The simplest approach confers non-zero values to the virtual compliances.5 meters. Suggested Steps: and parameters. The various results are compared with respect to the behavior of the physical design. . for placing causal strokes. Annotate the bond graph in the recommended manner. and there are 1-junctions and inertias for both elevation and roll.

Find the corresponding state differential equations in a proper form for integration. 5. and that the center pressure is higher than the others. . Write the correspondingset of state differential equations. 11.16 Continue Guided Problem4. and again compare. Carry out the respective simulations. 5. OVER-CAUSAL AND UNDER-CAUSAL MODELS 331 Writethe state differential equations and associated algebraic equations. Notethat the pressures in all three plenumsis non-zero. PROBLEMS 5. 10. with the time derivatives of the three pressures as the dependent variable.2. Find the differential equations for the derivative method.14 (p. Youmayemploy parametersof your owndevising if they are defined explicitly. considering the fluid pressure. Establish the initial conditions for equilibirum height and roll angle but the specified roll angularvelocity. to be an independent excitation and the motions of the solid parts to be responses. 227) by defining a minimumset of state variables. This equation is readily solved by MATLAB.5.5 (p. Note howmanyseconds your computertakes to carry out the simulation.17 Augment the bondgraph belowwith causal strokes. and comparewith each other and with the response in Fig. The simplest approach employs non-zero values of the virtual energy-storage elements. The algebraic equations cannot be solved analytically for the non-state variables in terms of the state variables. 319). A very small initial elevation is required to preventthe differentiation modelfrom trying to invert a singular matrix (which wouldmakeit an "index two" problem). Also find the initial conditions whenthe physical complianceis increased by a factor of five. Establish the initial conditions for lift-off (with no roll) for the three models. Carry out three-second simulations for lift-off. and write the correspondingset of state variable differential equations. P. define state variables.This is done by taking the derivatives of the algebraic equations. A3 x 3 matrix equation can be written.

p. 47) employed as the dashpot in the system of Problem 4.22 A hydraulic shock absorber of the type described in Fig. Write the state differential equations for the resulting nonlinear model.24 (p.12. identify state variables. 2.21 Answer the questions of the preceding problem for the bond graph above right.20 The bond graph below left has constant parameters.332 CHAPTER 5.----- R~ 12 5. except note that the input is 00 and the output is eo. and write state differential equation(s). 264). O~G~ R 1 e. (This is a modification of Problem 5. Check to make sure they agree with the results of part (b).18 For the bond graph given below left: (a) Apply causal strokes to find the order of the system.. MATHEMATICAL FORMULATION 5.24 (p.o 0 C 2 S~ eo 1 ~.19 Answer the questions of the preceding problem for the bond graph above right. (b) Find the set of state differential strokes. (a) Assign causality. assuming constant parameters. Sf 1 -. equations corresponding to your causal (c) Reduce the number of energy storage elements to a minimum and find the corresponding state variables and differential equations. (b) Find the output fiow 0o as a function of the input effort eo. 5.-~-G--. 5.. 305.) . 5.

Plot the angular velocity for the motor and the displacement of the center of the roller as functions of time until equilibrium is virtually reached.) can be approximated 5. The function M(~. 5. so as to eliminate the differential causality.5 rad/s. 272) has a rotational in.2. (c) Carry out the procedure illustrated in Guided Problem 5. OVER-CAUSAL AND UNDER-CAUSAL 333 MODELS 5. inertia J = 0. (b) Assign causal strokes to the bonds. . Comparethe result the single equivalent differential equation. 296).lb. starting from an initial motor speed of 1.5. 5.s 2 and the cylinder has a radius of gyration rg = 8 in. (d) Write state variable differential equation(s) for the model.7 (p. Note the presence of differential causality. 297). using the differential causality given. The electromagnetic torque on the motor may be left in the form M= M(~).8 (p. 5. 5.3 (a) Repeat part (a) of that problem to include dynamic effects.52 (p.24 Repeat the above problem for the model shown in part (d) of Fig.26 Augmentthe bond graph below with causal strokes. define state variables. (e) Simulate the response of the system.27 The bond graph model below comprises elements with constant moduli. State the order of the model..25 The motor of Problem 4. and write the corresponding state variable differential equation(s). 5.23 Write the state differential equation(s) for the model shown in part (d) Fig.4 to combine the energy of the element with differential causality with that of a different element.

(b) Sketch an engineering system for which the bond graph could be reasonable model. and compareto makesure they are the same. (b) Definestate variables. C. and relate the parameters of the bond graph to those of the engineeering system. and find a set of state differential equations. The source is characterized by a knownfunction es = f(0s) or ?Is = g(es). 5. and the right-hand portion of the modelrepresent a mechanical subsystem. and write a set of state differential equations. 4. MATHEMATICAL FORMULATION (a) Definestate variables. (Note that integral causality specifies the causalities of all the bonds. Find the differentia1 equations in terms of the samestate variables as in part (a). Suggestion: Let the source be electrical.t equilibrium in terms of the given parameters and/or functions.28 The bond graph below has constant parameters I.29 Answerthe preceding question for the bond graph given below.334 CHAPTER 5. makingthis a relatively simple problem. 5. T. except omit the transformer. (a) Determinethe values of e~ and Os a. (a) Definestate variables and write differential equations without changing the given bondgraph with its mesh. 5.) (b) Applya bondgraph equivalence from Fig. R.18 (p 248) to eliminate mesh.30 Consider the bondgraph given in Problem4.7 (p 202) for the effect of drain tube on humanhearing. .

2. The design pictured below employsfive plenumswith internal and external skirts all having their lower edges in the same plane.2.5. w2 = 8 rad/s and w3= 32 rad/s. OVER-CAUSAL AND UNDER-CAUSAL MODELS 335 5. Modelthis systemwith a bond graph.5 lacks roll stability and pitch stability.5 hovers over water that rises and falls uniformly according to the formula z = zo sin wt. Carry out simulations for frequencies wt = 2 rad/s.0m A~. The initial conditions can be for equilibrium conditions. all with the amplitude zo = 0. 5.33 The ground-effect machinediscussed in Sections 5.31 Repeat the problem above for the bond graph below. makingthis a more difficult problem. C 5. 10.2.2.4 and 5.4 and 5.05 meters. and the improvementinGuidedProblem5.32 Consider that the ground-effect machinediscussed in Sections 5.2. as a first step towardan analysis.5 lacks pitch stability. Note that integral causality does not specify the causalities of all the bonds.-~ top view A section A-A .

’2p dt.4 T~ ~- r2 C = 1/k dx T. dt .1/kr] T= J~b ~ + ~rny = (J + mr~’~)(b ~ = ~ ~ ~ . since (b . d* ] (b) Collecting terms in equation (b).’therefore. 1 i .~P .1 -~7¢ = r~mg . 1 These are the same equations.j + ~nr~p = -~p orp’= l + ---~-) p an(~ ¢ = x/r~. Note that ~ = pig.T~ \ v. Although the first method shows the structure in more detail.J÷ rnr~ dp’ = r~mg."~ ---- (c) kr~x + rimg Equations (a) and (c) are a completeset of state differential equations. C’ -. I=J+rnrl 1 de 1 . the second .336 CHAPTER 5. I~T~2 "~ dp T~ 1+ I~ / ~-~ = -~-x+Turng or -. methodis easier and shows the behavior more clearly. IVlATHEMATICAL FORMULATION SOLUTIONS TO GUIDED PROBLEMS Guided Problem 5. 242 V=~r~ =~-~¢ 2 therefore.I¢ p= -~P (a) dp Tx ( I~T~ dp d-7= c x .kr/ ¢ d---~ 1 .

I~ = mr~ = 9000 kg-m Repeating equation (5.4 = 40 ms. Qf(Pc) = Qo .0 x 10-amT/N2. = y ( L + -~ ) + ¢w (-~ + ~-66 ) The respective flows are Ql= Alcd~/~Pl and Q~= Arcd~P~.~w (~.s. From the given information.5 iCangularmotion(roll) left plenum~ plenum center t~ ~right plenum I vertical motion(heave) The area of the deck is . with flow Q~ = L~yCd~Pc. Tc~ = ~]~ = 0.100 m-~.V -- 1 (A/4)(aw/S) 1 m_a. and T~ = Tt~ = )-~ = 0.050 The width of the deck is w = 4m.2. a~ = 2. Q0 = 180 m3/s.al P¢ + a~Pf .482 x The center plenumhas a periphery of outer skirt of length L~ = 4m.08 m~/N.17).P~) . Outer skirt are of the left-hand plenum: At = ~. at = 0.s. Guided OVER-CAUSAL Problem AND UNDER-CAUSAL 337 MODELS 5. and -Tic = T.a4P~.(L + ~) . The flows across the internal skirts are Q¢~= L y - c~ (P~ . aa = 1.+ ~-~) Outer skirt are of the right-hand plenum: A. I~ = 4000 kg.5. ~.

s \ s~~ Thefoursta~ev~iablesp~.338 CHAPTER 5.g and~ giveNurdifferentialequationsin terms of the functionsdefinedin step1: d~ ~ = ~/I~ . MATHEMATICALFORMULATION / . p~.

P~)’ f¢ andF f~o f2 f~ f7 0 fS f9 = OQ¢l + OQt O(P~ . the pressure Pt is replaced by V~/Ct.y) .5.~ f =i)Q~\ OQcl’~ o~ o~~ +) ( \OQi0¢OQct-~)~ ~ ~-~ .y.¢) . ¢) .Pt.Q~.¢) .d’-~ 1 Q~.(Pc P~. V~and V~ are added to represent the three volumetric compressionsin m3.5.Qc(Pc.¢) .265 x 10-am~/N.y. [ dP~/dtJ f~ wheref = OQ~ O(P~.~b/Tl¢ .Q¢. In the differential equations of step 5. P~ is replaced by V~/C~and P~ is replaced by V~/C~.y.Qc~(P~ .P ~.Pt) OPt + 1 d(b + 1~ ~ I. OVER-CAUSAL AND UNDER-CAUSAL MODELS 339 --= --P~+ + dt T~u "1’~ dp ¢ 1 ~ p~ d~’~" = ~¢Pt + 77¢ There are also three algebraic equations: 0 = Qcl(Pc .2. y.~/Tt~ 0 = Qcr(Pc . ~ computedin Section 5.y.~/T~ 0 = QI(P¢) .Q~.T~--~V~ - 1 1 d--~. and C~ and C~ each represent one-quoter.y.2.Q¢~.Qt(Pt. The following three state differential equations replace the derivatives of the pressures: dV~ = Q.Q~(P~.= QI . IdP~/dt] =F-lf.¢) .¢) .P~.~ /T Thethree state variables V~. The total compliance is about 0.--T~uI~ p~ Note that C~ represents half of the total compliance.

0. The initial compressionof the center plenumis therefore Vc(0) = ~ x 0.1 0 0 ~five times nominal compliances --no compliances (differentiation model) ~ 1 2 time. although increasing the compliancesby five times gives a significant error. MATHEMATICAL FORMULATION Duringlift-off. At equilibrium.265 x -3 x 3000 = 0.399 ma. any other values gives an equilibrium roll angle different from zero. the flow leaving one of the side plenumsequals the flow entering.1703 m.m/s. The responsesare similar to those found in the absense of internal skirts.340 CHAPTER 5.0818 m 9-10. the proper initial pressure in the left plenumis 13 Pa greater than the 570 Pa.44. seconds 3 The lift-offs with and without the compliancesdiffer little. meters 0. This problem doesn’t quite arise in the simulation with the compliances~ since these compliances reduce the pressure decrease enoughto keep the pressures positive.3 elevation. Consequently. The final elevation is 0. although the loweredpressures in the outer plenums reduces the equilibrium height. so that only four initial conditions can be selected independently. so the pressure in the center is 3000 Pa. 0. Therefore. are 1392 Pa and 570 Pa. wherePs is the pressure in the side plenumand Pc is the pressure in the center plenum. respectivbly.265 -ax 1229. For this roll. Note that the minimumorder of the differential model is four. Therefore. for ¢ = 0 and any y. the pressures in the side plenumsbriefly drop belowzero because of the rapid rise of the craft caused by the high pressure in the center plenum. 6 = a. it is necessary to employsign and absolute value functions in computing the flows Qt and Qr and their derivatives. The final pressures in the center and side plenums. (These corrections can be found by trial-and-error. and the initial pressure in the right plenumis 13 Pa less. P8 Wheny = 0 the fan flow is zero. The roll rate of 0.2 0.05 = 450 kg. which is used as an initial condition for the simulation with roll. ( Pc=l+ 1+ =2.05 rad/s corresponds to an initial condition for the angular’momentumof 9000 × 0. The reasoning above gives the pressures in the sides as 1042 Pa. and in the side plenumsis ~ (0) = V~(0) = ~ x 0.) .

3. you choose two arbitrary functions. radians 0 -0. the use of special software such as DASSL can be justified. a weaknessin the performance.1 Superposition and Linearity Linear systems can be identified merely by their response to disturbances. Suppose that you can choose u(t) and observe x(t) for a variety of separate "runs.7. and single response or output.5. 5. Thethree simulations are nearly indistinguishable. ul(t) and u2(t).01 roll angle. as trated in part (b) of the figure. Increasing the compliances by a factor of five reduced the computationtime to 5 seconds.3 Linear Models and Simulation The property of superpositlon defines the class of models called linear. as in Fig. 0. it is not necessary to have any analytic representation. unlike for lift-off. seconds 5 The simulations for roll reveal significantly less dampingthan those for lift-off.19 part (a). The derivative model took only about 2 seconds. Then you choose a function u3(t ) which equals . LINEAR 341 MODELS AND SIMULATION 11.01 0 1 2 4 3 time. Operational notation allows these equations to be manipulated like algebraic equations. the use of script files and global variables. the relative ~dvantageof the derivative model wouldhave been even greater. Unlike nonlinear models they are generally amenable to analytic solution. 5. A MATLAB command that directs such a linear simulation is presented below.0001 with MATLAB’s ode23 in version 4. Bond graphs with elements having constant moduli are linear models.3. Linear models also can be represented by linear differential equations. with a single disturbance or input. Consider a "black box" system. after the statements above are clarified. Hadthe efficiencies of the computations been increased through. The reason is that the three pressures do not vary nearly as muchas in the case of lift-off. x(t). Nevertheless." First. despite the fact that each iteration required roughly twice the computation. This fact allows numerical simulation to be carried out considerably more efficiently than through the use of a general purpose algorithm such as the Runge Kutta integration presented in Section 3. it is hard to justify the effort required to carry out its programming. 5. and observe the respective responses xl (t) and x2(t). Thesix-second simulation for roll with the nominal compliancesrequired 32 seconds of computer time on a particular PC computer. using an error factor of 0. u(t).Whenevercomputer time becomesexcessive for the virtual energy-storage methodto give adequate accuracy.

~ time = ~(t) ~e (b) signals Figure 5._~ "~~esP°nsexflt)__.- response x2(t) excitationu3(t) =u~(t) +u2(t) . MATHEMATICAL FORMULATION (a) "black box" systemwith excitation and response excitation u l (t ).19: Property of superposition .342 CHAPTER 5..________ ~ rune exci~t~ion k~~~g--.

LINEAR 343 MODELS AND SIMULATION the sum of ul(t) and u2(t): (5. (5. which implies u~(t) = ul(t) + us(t) = 2ul(t). requires x3(t) = xl (t) + x2(t) = 2x~ (t).3. leading to the conclusion that the response to the excitation cub(t).37a) x2 (t) = xl (t). that is for sufficiently large signals. The simplest case of all eliminates time as an independent variable. is cx~(t). Now. doubling . (5.~ u .34) u3(t) = ul (t) + u2(t).3. the system is nonlinear. A plot of x versus u is a straight line through the origin. Outside of this domain.-~-~÷bm_~ d-~-~-ff_~ dx + al-~ + aox du +. that is if X3(t) : Xl (t) (5.=b~. the property applies if this relation is true for arbitrary ul (t) and u2(t).x within which superposition is satisfied and the system acts linearly. leaving an algebraic relation between the excitation and the response.5.(t). The property of superpositlon is said to pertain to the system if the observed response x3(t) equals the sumof xl (t) and x~. The nth order differential equation d’~x an-~ d~-~x + an-l =f(t) ~ + an-s d~-~x d-~i"~_2 + "’" dm u d"~ .’. For example. t. Systems often have a domain of u. represents a linear relation between the independent function u(t) and the dependent function x(t~) because each term satisfies superposition and therefore is linear.36) u2(t) ul (t).37b) The property of superposition. This is what is meant by linearity. this happens regardless of the size of u~ (t).÷b~-~-[+bou (5. if u~ produces x~.38) In words. More precisely. so in general x = ku where k --" x~/u~. any other relation violates superposition and mandates the name nonlinear: 5. superposition requires that the doubling of an excitation merely doubles the response at each and every time.39) in which the coefficients ai and bj are constants. where c is any constant.2 Linearity and Differential Equations You may be most familiar with linearity in the context of algebraic and differential equations. 2u~ must produce 2x~. (5. if it pertains.35) ÷ X2(t). A convenient special test employs (5. Further.

39) expresses a stationary linear model. equation (5. MATHEMATICAL FORMULATION x(t) doubles each term. The model becomes nonstationary if one or more of the coefficients ai become explicit functions of the independent variable of time.) The relation between input and output variables. (5. Whensuch a set of equations is linear and stationary. and therefore corresponds to doubling u(t). on the other hand. with all their conceptual and computational advantages. . nonstationary linear models are represented by the same equation with A = A(t) and B = B(t). for example. is apt to be rnuch more difficult than for a linear model of comparable complexity. Identifying a dynamic nonlinear model from observation of dynamic behavior rather than knowledge of the internal structure of the system. As you probably have noticed. the linearity and its property of superposition would be lost. (Matrix methods are largely deferred to Chapter 7.39). however.40) in which A and B are matrices of constants. t.344 CHAPTER 5. (The modern theory which goes by the name "chaos" actually is based on deterministic nonlinear dynamics. In matrix notation. by _coptra~s_t~ descri_be~ only the relation between single input variable and a single output variable of the system. and is the focus of the solutions developed in Chapter 6. Thu_s. characterization of most nonlinear systems in terms of their operational behavior is very difficult and usually impracticable. bond graphs in which the moduli of all the elements are constants directly produce linear algebraic and/or linear stationary differential equations. Were any term raised to any power other than unity. This incomplete model may nevertheless be all the modeler needs. The general excitation is the vector u(t) and the response is the vector x(t). nonlinear models may be deduced from direct considerations o] the physics almost as easily as linear models. the modelis still linear. the response to a given disturbance may for example vary wildly depending on the initial condition. The use of causality.. for example. Superposition still applies." but rather means that the dynamic characteristics are invariant in time. By comparison. which is sufficient for vibration analysis or classical control calculations. Note again that the term "stationary" does not i~nply "static. establishes the transfer function between these variables. A model described by a set of first-order differential equations such as equation (5. This book considers analytical solutions for stationary models only. it can be placed in the canonical or standard matrix form dx ~ = nx + Bu.) Despite this complexity.on a particular bond graph results directly in a set of first-order differential equations. since the behavior of all the state variables is represented. More explicitly. The fact that the shape or form of the response of a linear system depends only on the shape or form of the disturbance and not on its magnitude allows the system to be characterized simply in terms of its operational behavior.40) can be called complete. Equatio_n (5. characteristics such as natural frequencies and time constants do not change over time. This property of linear systems is developed in later chapters to give powerful analytic results.

I.3. G(S) and H(S): X = a(S)u.dr" Withthis notation.A)x = Bu.. the double lines indicated that the associated variables maybe vectors rather than scalars. Representingthese sumsby the vector Y(Q. (5. (5. S. Usually. 5Someauthors prefer to use the symbol D for the time derivative operator. is defined by S =.46) is represented by part (b) figure. Theseequations also are written in terms of the transfer function operators. The derivative operator.20. .42) (5. Equation (5. G(S).39) and (5. The result is called a block diagram. the modeler maybe interested in one or more weighted sums of the state variables and the input vector. although somepractitioners use them more broadly to represent dynamicmodels.-1S~-1 + . (arts n + a. where I is the unit diagonal matrix. [y(t) = Cx + Du. 5. Equation (5.3 Operator Notation The manipulationand solution of linear differentia t equations is simplified by ’~ the use of operator notation. bmSra + bm-~sm-1+ "’" + b~S + bo a(s) = a~s~ + an_~S~_~ +. y= H(S)u.40) become. (~’~) G(S) = (SI . Block diagrams are most widely used in the specialized domain of automatic control. The circle with the summationsign inside indicates that the output variable is the sumof the two input variables. equations (5.5.3.41) This classical representation includes as special cases y = x (C equals the unit diagonal matrix. The functions upon which they operate are indicated by an arrow pointed into the box.43) (5.. 5..44) (SI . + alS + ao)x = (bins m + bin_iS m-~ + ". ~ shownin Fig. Transfer functions commonly are placed in boxes.45) is represented by part (a) bf the figure.45) has been used to give the result in terms of the output variable y(t).46) wherein the last case equation (5.respectively. ¥’a-~-[oo x= G(S)u.A)-~B. + b~S + b0)u. which historically is known~ the Heaviside operator. and the functions whichresult are indicated by an arrow pointed out from the box.. H(S) = C(SI. and D = 0) and a single output (C and D become row matrices). one wants D = 0. LINEAR MODELS AND SIMULATION 345 In general. Mostuse in this text is restricted to control applications.A)-~B (5.

dt d-~ = 4p . .46) to the single-variable formsof equations (5.3. (5. 346 MATHEMATICALFORMULATION (a) scalar u(t) (b) vector u(t) .45) and (5.8 Find a transfer function relating the input u(t) to the output q(t) for the following differential equation.4 Transformation from State-Space to Scalar Form Oneuse of the operator notation facilitates the reduction of the state-space models of equations(5.= -2p + 2q + 24u(t). Solution: In operator notation.44) or (5.x(~ _~.40).9q + 4 + 2u(t). and give the correspondingsecond-orderdifferential equation: -. (S + 9)q = 4p + (4S + 2)u. giving the following equat. y(t) _ Figure 5.46) 5. ion from whichp has been eliminated: (S + 9)q 4(2q + 24u) S + 2 + (4S + 2)u.CHAPTER 5.43) or (5. these equations become (S + 2)p = 2q + 24u. The variable p can be solved for in terms of q. EXAMPLE 5.45). (5.20: Block diagrams of equations (5.39). algebraically: P-- 2q + 24u S+2 This result then can be substituted into the secondoperator equation.

S]q(t) = [96 + (4S + 2)(S + from which G(S)= 2 + 10S + 100 96 + 4S2 + 10S + 4 4S = S ~ +11S+10 2+11S+18_8 This transfer function can be interpreted to give the differential du d2u d’2q + 11-~ + 10q = 4-~.A)(SI .A) (from equations (5. .8 using equation (5.A) becomes which when substituted S+9 ’ into equation (5. dt---~.. The square matrix (SI . EXAMPLE 5.A)x = det(SI .47) Solution: The state-space Note that the derivative the derivative operator.A)-IBu by det(SI . The procedure 6 becomes even more streamlined if you choose to use matrix notation in addition to the operator notation.47) This is a vector set of scalar differential equations: the first in terms of x~. LINEAR 347 MODELS AND SIMULATION This equation can be multiplied by the denominator factor (S + 2) to give [(S + 9)(S + 2) . equations in matrix form are term 4 du/dt has been accomodated by the use of S. the second in terms of x2 and last in terms of x~.8.3.46)) gives general result det(SI . Notice that algebraic operations have replaced derivative operations. (5. which is a great convenience. gives the resulting dif- 6TheHeavisideproceduredescribed here is formalizedand extendedin the definitions and applications of the Laplacetransforml whichis introducedin Section 7.9 Solve the problem of Example 5. equation 10 + dt + 100u.47) gives The second of these two scalar equations identically ferential equation of Example5.2.5.A)-IBu. Pre-multiplying both sides of the equation x = (SI .

4Sd) whichapplies as long as m_< n . scalar transfer function).. if you have only one input.i The ss2tf can be read "state-space to transfer function" (which really means. den]=ss2~f (h. 1 (5. and employs A-- --an-i-an-2 1 0 0 0 0 . it dependsuponwhat definitions of are chosen. ..B.3.48) gives -d Xl = dt x2 x3 5. The argument± is an index designating which of possibly multiple inputs is to be considered.1. as in Examples5.9 above.6 -3 -4 0 1 Transformations Xl x2 x3 Using ~.. B.48c) (5.. EXAMPLE5.D.. D must be defined numerically before this command is issued. ~=[o21]x_~ . insert 1. C.. 5.C.348 CHAPTER 5. al 0 -ao o| ] (5.10 Find a state-space differential equationconsistent with the scalar differential equation du d3 x 3 d~ x~ dx 2 dt ~ + dt~_ + 4-~ + 2x = dt + u" Equation (5. The output argumentnum (whichcould be designated with any symbol)is a row vector of the coefficients of the numerator polynomial of G(S). The state-space matrices A. The output argumentden is a row vector .48b) C = [bn-1 bn-2 "" b0] D=0 (5.3.48a) .(t). The simplest schemeassumes x~ = x and an = 1. This is accomplished with the command [hum.5 Transformation MATHEMATICAL FORMULATION from Scalar to State-Space Form* It is possible also to transform a scalar differential equation into state-space form. x3 MATLAB MATLAB provides translation from the state-space format to the scalar format.. Theresult is not unique.8 and 5.

den] = ss2tf(A.B. by the MATLAB command Isim(num. which therefore is reported only once.C.61c) The MATLABentry A=[-4 1/2 0.0000 8.u. D]=tf2ss(num.4x~ + ~x.2X’2 + 2X3. Solution: (5.61b) (5.0000 t0." The argument t is a row vector . B=[1.0000 which indicates the transfer function 2S+8 G(s) = $3 10S2 + 29S + 20" The inverse transformation is given by the command [A . C--[O~ 0].1) gives the response IlUm = 0 0 1. dx." 5.D.0 1 -4].0] .C.0.11 Use MATLAB to find the transfer function from the input u(t) to the output x. Each has the same denominator. The respective numerator polynomials are reported as the rows in num.den) in which tf2ss can be read "transfer function to state-space.~ + u(t).3. EXAMPLE 5.4x3.5. [hum.7 Simulation of Linear Models Using MATLAB* Numericalsolutions to linear scalar differential equations are plotted.B. more than one transfer function is being requested.0000 2.2 -2 2. D=[0] .2 (t) for the third-order model dXl 1 dt .0000 20.61a) (5. dt dx3 d--~ = x2 . lsim stands for "linear simulation. 349 LINEAR MODELS AND SIMULATION of the denominator polynomial. If the matrix C has more than one row.3. with little effort on your part.den.0000 den = 29.t) Related options given below include the treatment of matrix differential equations.= 2Xl -.~ -.

the coefficients on the left side of the differential equation or the d4nominator of the transfer function are recognized by the statement den = [1 3 4 2]. u(t): ul = 5*sin(2*pi*tl/lO).t2 = [5.7. you specify a corresponding vector for the excitation signal.3~.t) from the single added command" .i:5]. The commands tl = [0:.. a process which is potentially both more accurate and more efficient than the numerical simulation scheme presented in Section 3.350 CHAPTER 5. EXAMPLE 5. the. which are given in the numerator of the transfer function format for the model. it must have the same length as t.10.t=[tl establish vectors for the two segments of time and the combined time duration of 0 to 10 seconds.den.) Exact analytical solutions of the differential equations are evaluated numerically. t_>5. u2 = zeros(size(~2)). Solution: You start by defining a vector of discrete times at which the solution will be found. ¯ Similarly.l:lO]. + dt 4-2x= dt + u’ dr--~ with the particular excitation (5sin(2~rt/10). A plot of the response results isim(num. and excercises that assumption. The argument u is a vector of the input values at precisely those same times. are recognized by the statement hum = [2 1] . 0 < t < 5.12 Secure a plot of the solution of the differential 5. Consider the range 0 < t < 10 seconds. Next. repeated here: equation used in Example dUx d2X 4 dx du~ 2 4.u. Superimpose a plot of u(t).u=[ul The coefficients on the right side of the differential equation. program assumes that the actual input for intermediate times is given by a linear interpolation of the immediately surrounding given values. MATHEMATICAL FORMULATION of the times at which the response is to be computed. (In some cases it concludes that a step-like zero-order-hold function is intended. u= 0.l:. Normally.

1 0 0.u.t) or you could use the commandwith the left plot (t. B.5 0 -0.D.C.5 -1 0 1 2 3 4 5 Time(secs) 6 9 10 The commandl sim handles matrix differential equations through use of the standard A.C.03.10).den.t.5 0. C.B.u.u.13 Repeat Example 5. EXAMPLE 5.D notation.D.x] and then type .D.12 using the state-space formulation (as given in Example 5.5 2.3.C.x] = lsim(A. The second case allows you to introduce non-zero initial conditions by first defining the vector x0.B.x0).0 1 0]. c = [o 2 ~]. [Y. B = [1. The options below include creating an accessible file of the results by adding a left side to the statement: [y. In the third case.t).B.5.X] = lsim(A.u/2. [y. Solution: The matrices A.0.B.C.y) side [y. D are defined with the commands A = [-3 -4 -2. The excitation u(t)/2 is added by the commands hold plot(t.x] = lsim(num. LINEAR 351 MODELS AND SIMULATION The plot below results.t). D=O The plot results from the command lsira(A.’--’) 3.u. the output vector y and the state vector x actually are identical.

Operator notation allows a linear differential equation to be treated like an algebraic equation. A model is described by the two state-space differential equations below.8 Summary Systems that exhibit a ~lomain of input and response variables satisfying the property of superposition are. MATHEMATICAL FORMULATION lsim converts the differential equation into a difference equation which for initial-value problems is accurate regardless of the size of the time interval. Operator notation also leads to analytic solutions using transform methods. MATLAB capitalizes on some of this with the linear simulation Commandls±m. but it applies only to linear models. (It does this using a matrix exponential.q dt Suggested Steps: 1. 5. including the elimination of unwanted variables. if the variables get large enough there must be a catastrophic failure.) Because of the assumed linear interpolation of u(t). as is developed later. Write the two equations with operator notation. of course. which is described in Sections 7.3. This expedites the transformation of differential equations from one format to another. The domain cannot be infinite for a real physical system. . This simulation is more efficient than a corresponding one using a nonlinear integrator such as a Runge Kutta algorithm. linear within that domain. Transfer function and state-space formats are of particular interest. which you can use even before you understand its mathematical basis. 2. dp -- = -2p-q dt dq = 7p . which is a nonlinearity of the most severe type. by definition.6 You will use operator methods in this problem to combine two coupled first-order linear differential equations with the objective of characterizing the dynamics of the model. Guided Problem 5. Determine its natural frequency and damping ratio. the time interval might have to be rather small to give acceptable accuracy. or q as a funciton of S and p. Solve one the the equations for p as a function of S and q.2. They may be represented by linear algebraic or differential equations. however.1 and 7. MATLAB carries out these transformations with little effort on the part of the analyst.3.3.352 CHAPTER 5.

Guided Problem 5. and one equation with dependentvariable p only. Finally. Then use operator methods (but not MATLAB) to accomplish the same objective. C = 0. Combine these algebraic equations to get one equation with dependentvariable q only.10) in the forms p = p(q. assuming that the input variable is the voltage e. 4. It also offers practice in linear simulation with Considerthe electromechanicalsystem of Fig. 292) and equations 5. p or q. 5. and check to see that they agree with the results of step 2. 2. Rewrite the original differential equations using operator notation.LINEAR MODELS AND SIMULATION 353 Substitute the result of step 2 into the other equation so as to eliminate either p or q. parameters are T = 0. Combinethese equations directly to get two single secondorder differential equations with dependentvariables p and q.2. substitute the results of step 1 into these equations to get two resulting differential equations of secondorder. C. D.6 (p.0Nms.2 kg m Rl=4ohmsandR2=l. and using MATLAB. each having only Onedependentvariable. respectively. Interpret the result of step 4 as a differential equation with dependent variable p or q. respectively. 294). Then.9 and 5.9) and (5. Place equations (5. find the matrices A. Use MATLAB to find the two second-order differential equations a third way. use ls±m to plot the shaft speed as a function of time for 1. G = 1. . 149) to determine the natural quency and the dampingratio. Cast the result of part 3 into the form f(S)p = orf(S )q = OThe function f(S) should be a polynomial. Suggested Steps: 1. dq/dt) and q = q(p. using both direct and operator analytical methods. Interpret the results of step 3 in terms of differential equations. respectively.0 second. Use the result of equation (3. The -1. assuminge = eosinwt with e0 = 5 volts and 03 = 8~r rad/sec. and that the output variable of interest is the angular velocity of the shaft.0 volt sec. Take the devivatives of the equations found in step 1. dp/dt).41) (p.7 This guided problemgives neededpractice in the reduction of state-space differential equations into differentiM equations with a single dependentvariable. Next.05 N-~m I = 0.10 (p. B.

t o ’i’’ time. using operator equa- methods (but not MATLAB). C and D. (c) Find the state-space matrices A. excitatlon. Then. You may annotate this plot by hand if you do not care to learn MATLAB labeling procedures. Employ the isimcommandto secure the resulting plot.354 CHAPTER 5.26 (p.000 seconds with an interval of perhaps 0. assuming that (~ = p/I and q are the output variables of interest.35 Consider the IRC model of Fig. . extract the coefficients A and B and evaluate these matrices numerically. check to see that the differential equation that emerges agrees with those of steps 2 and 4. Define a new vector of times from 0 to 1. B. respectively.005 seconds. 137). as represented by equations (3. Relate the desired output variables to the state variables p and q. 3. (b) Repeat part (a). 139):. Evaluate these matrices numerically. seconds 5. express the excitation voltages at these times as a row vector. (a) Combinethese equations to get single second-order differential tions with dependent variables p and q. Examine the data carefully and report whether there are any signs of nonlinearity.34 A system responds as shown below to the given excitation. seconds 7 ~ time. Use the MATLAB function ss2tf with the results of steps 5 and 6.32) (p. Are the results reasonable? PROBLEMS 5. so as to define the coefficients C and D. using MATLAB. MATHEMATICAL FORMULATION Place the original equations in the state-space matrix form.

0.36 Answer the questions of the preceding problem for the IRC model of Examples 3. Verify whether your answers are consistent with those of part (a).15 (pp.2) (pp. (d) Using lsi~.6 m. B.40 Consider the vehicle of Figs. 355 LINEAR MODELS AND SIMULATION (d) Use MATLAB to convert your answer to part (c) to single differential equations in ~1 and qc. assuming the output variables interest are the depths of the water in the two tanks.37 Answerthe questions of the preceding two problems for the electric of Fig.11 and equations (5. 5.30) and (5. (e) Characterize the dynamics of the model with the parameter values given in part (d) by finding its natural frequency and dampingratio.1 and equations (5.5.025 ft (c) Using the MATLAB command ls±m.14)) RI~. A2= 4 .13) (pp. L1 = 1. circuit 5. this gives (from equation (2.0 ft 2. 5. The values of the parameters m = 1250 kg.11) and (5. ks = 32 kN/m and R1 = R2 = 4000 Ns2/m. 313). (b) Find the equilibrium depths of the water in the two tanks for steady flow Qi : Qo = 3/s. C = 2 and I -= 3.39 Consider the linear model of the two-tank system of Guided Problem 5. In part (d). 5. 140). 299) with A1 = 2. C.0 ft above the equilibrium) with Qi = as above.1) and (5. 5.38 Answer the questions of the preceding problems for the example of Fig. Ap = ~ra : = ~r/ 4 in 2 and L = 20 ft. L2 : 1.3.s/ft ~ and I = 4pL/3Ap = 9480 lb-s2/ft 5. 149). using R = 1. (p. find and plot the depths of the water for 300 seconds when. 283-284). 324. kl = 30 kN/m. J = mr~" with r 2 = 1. Assuming quasi-equilibrium laminar flow of water at 70°F. 322.14 and 3. . at t = 0 seconds.1. 5.32) (pp. 5. Hint: Use equation (3:41) (p. Suggestion: First make sure the unexcited system produces no response.0 2. find and plot the depths of the water for 8000 seconds following an abrupt dumpingof 2. 325). 312. The resistance of the outlet is Rd -= 2RI:. (a) Evaluate the matrices A. = 8pL/~ra 4 = 7056 lb. 138.16 and equation (5.15 and 5. D.4 m. let I2 5.0 ft 3 of water into tank 1 (giving an initial depth 1. the input flow Q~ switches from Q0 to Qo[1 + sin(2~rt/100)].33) with equations (5.6 m2.

Comparingthis differential .dp -~ + c T~) + "5"e or.--~--p ÷ C (b) ~tt dp _ R~T2I dq d’-)q IT de (c) dt G~C dt I-~ + --~--~ dq _ CR~ dp d~-p (d) dt I dt + C-~ Substitution of equations (b) and (d) into (a) R~T:I dep~ [CR~ + cdP~ IT v =. the road is otherwise smooth. q = -(S + 2)p.5.a~----5~.41) (p 149). Let y~l = yo sin(2rt/T). d2p 5. 4.05 m and T = 0.9 SOLUTIONS Guided Problem TO GUIDED PROBLEMS 5.] -~ + 1+~ ] ~ = -~ ~ Substitution of equations (a) and (c) into (b) CR~ f R. The rear wheels pass over the same bump T seconds later.) 5. 6. ~ c -~ + C R ~ + --~-. -(S + 1)(S + 2) 7p. C.--i-~ "~) .2 seconds. y0 = 0. (b) Using the MATLAB commandls±ra.-q. + 3 ~ + 9p = O. Fromthe first equation. 3.3. (S2 + 3S + 9)p = 0. plot the res~)onses of the vehicle to a bumpwhich passes under the first axle and then under the second.T~I q=-i-t. D assuming that the input variables are ~gl and ~g2 and the output variables are ~cm and ~. (c) Using MATLAB. and 2(w~= 3 so that ~ = 0. (S + 1)q = 7p.I(CR: \-i. valid for 0 < t < T/2. Substituting the result above into the second equation.--drO-q-~+~ IT) +c ( R~T~Idq a~c ~ ~-~+-~-~] d~q ITde’~ . (This part can be done independently of part (b). find the scalar transfer function between the excitation ~gl and the response 9c. 2.equation to equation (3. Guided Problem 5.6 (S + 2)p -. Rearranging.7i idq + -~IT e (a) CR ~ q -.7 RiTeI p= -O~ q . w~= vf~ rad/s. B.356 CHAPTER 5.~. MATHEMATICAL FORMULATION (a) Evaluate the matrices A. Give the corresponding differential equations. collecting terms.

and substituting first.3. 357 AND SIMULATION terms. on the other hand.5. = [-3.2.2000 116. so C = [0 5] and D = 0.2S+116e ~= S or (S 2 + 8.] R1T2"~ q+ 1I Sq/C + R~/ I .TG Multiplying both sides by (S + R~/I) gives Solving the second equation for q.] 2 + CR’~ q + ~p = ~e G dt and ~q = 0 Solving the second equation for p and substututing into the first. response B = [.0]. [z.a.C. s +tN+ s+t +N The differential equations resulting from step 3 ~e the same as those resulting from step 2 (apart from the multiplicative factor IC in p~t 2).20 -5]. A_ = [-’23d The output variable is ~ = p/I. LINEAR MODELS or. = 4e(t).D) Z = 0 0 20. 4. ~ =[ l/C 2Substituting -R~IIJ + ~G e values into the two matrices. collecting IC-~ 3. D=O.2S + l16)p = 20Ie(t) above. into the ( o. eS+ -~U-~-. The MATLAB programming and its . is as follows: C = [0 5]. which agrees with the result .0000 8. 5. 0000 1.2 -5. + \’--’GT-- S + -~-.B.p] = ss2tf(A.0000 p = 2O This means that ~+8 .

6 0.4 ~.B. The simplicity of the characterization and analysis of linear models further motivates the modeler to overlook nonlinearities. MATHEMATICAL FORMULATION 7. most nonlinear systems act as though they were linear.D. Nevertheless. rad/s 0. seconds 1.4. An analytic solution to the nonlinear differential equation - Ao v ~ (5. however. 5. inverting equation (5.t) 0. The MATLAB programmingand response is ms follows: t=[O: . however.1 Case Study With Linearization of a Resistance Consider a fluid tank with uniform area At and a Bernoulli square-law orifice flow. so this approach is presented first. It shows that the flow is computed with the causality Q = Q(P).8 time.49) where Q is the volume flow through the orifice and A0 is its effective area (including the vena contracta). A model may be linearized before or after it is converted to differential equations. Further. as shown in Fig.358 CHAPTER 5. simulation would be the natural recourse.4 0. for small perturbations about some nominal state.-8.e. or changes in discrete steps. and therefore sometimes is preferred.2 0.0 Linearization Nature is nonlinear.C.4 0. This characteristic is plotted in part (b) of the figure. Methods for approximating a nonlinear system by a linear model are considered in this section. In most cases the latter procedure is preferred. The pressure at the bottom of the tank is related to the orifice flow by ¯ P = 21-0 ~00 ’ (5. 5.50) happens to exist if the input flow Qin is a constant. The meaning of linearization usually is clearer when it is carried out directly on the individual elements of the model. the property of superposition would apply if the . e=5*sin(8*pi*t) isim(A. It also allows a graphical alternative approach. The solution of the corresponding problem with a linear or constant resistance is muchsimpler.005:1] . A bond graph model of the system is given in part (c). as is typically the case for vibrations.49). This is not usually the case.2 r~ 0 -0.2 0 5.21.

359 LINEARIZATION P volume V pressureP tl~orifice a (a) system (b) resistancecharacteristic C P= WC 1R P ss *_ P*I~ ~---~--~0 LQ(p R (c) bondgraph (d) linearizedcharacteristic 4 3 s. Figure 5. ° linearized: ~-- 20%increase in flow (e) linearized bondgraph t/r (f) responsesto step changesin Qi.5.4..21: Linearization of a fluid resistance .

(5. Further. 2. which also is designated by an overbar: Q. A linearization now will be carried out. Q = ~ + Q*.51) The nominal volume is the equilibrium state for a particular flow rate.54) . 5. The linearization corresponds to replacing the curved characteristic for the resistance by a straight line. The linearized characteristic is described by P* ~. or changes.53) R* = ( d-~Q ) ¢2= ~. and how accurate the solution needs to be.360 CHAPTER 5. the pressure at the bottom of the tank then has its nominal value. This volume is represented as the sum of a nominal volume. MATHEMATICAL FORMULATION model were linear. V*: dV* . and measuring its slope. The water tank has a single state variable. T = R*C.52b) where Qi*~. The value of be estimated by drawing the tangent with a pencil.15) (p. The usual choice is a tangent to the nonlinear characteristic passing through the equilibrium point.22 part (b) (p. the input flow equals the output flow. 5. Can an approximate linear model be substituted for the nonlinear model in order to gain these advantages? The answer to this critical question depends on how greatly the depth of the water varies. and the behavior of the linearized model compared with that of the nonlinear model. V*: y = v + y*. (5.Q~n -R*c dV* V* or ~---~- V* * + = ~-Q~n. as shown in part (e) of Fig. Note that at equilibrium.~n(t) to be assembled from a sum of the responses to simple components of that excitation.R*Q*. (5.21.52c) (5. V. and a perturbation volume. ~ dt . The slope R* is called the linearized or perturbation or tangential resistance. as given in equation (5. This graph gives a linear differential equation in terms of the perturbation state variable. which is very different from the nonlinear or chordal resistance as defined in Fig. (5.49).52a) P = ~ + P*. V. the volume of water in the tank. Q* and P* are the perturbations. The linearized or perturbation variables and parameters can be placed on a special linearized bond graph. The non-equilibrium values of these variables are designated as Qin = ~ + Qi*~.21. in the state variables relative to some nominal condition.53). The three pressures and three output flows are indicated in part (d) of Fig. from equation (5. designated as P. (5. 44) and equation (2. 46). A linearized model describes the behavior of perturbations. enabling the response to a complicated excitation O. It also can be computed mathematically by evaluating the derivative. as shown in the figure.

are significantly different.5 (p. for nonlinear effects that are large enough to require abandonmentof the simplicities of linear analysis. however. the results of the linear model are only of qualitative usefulness. The solutions for the 100%increase. can be expanded in a Taylor’s series expansion as follows: Y = f(~) + -~xdf x-~. (5. 145).21.55) The solution of the linearized model for the 20%increase in flow is seen to be close to that of the nonlinear model.(x .5.~ = constant. = df -~-+ 2 ~=5 x*2 higher-order -~x ~=~x* + dx d2f terms.57) It is convenient to designate some nominal values of x and y as ~ and ~. as given in Section 3. (5.59a) y = y + y*. 5.5S) The actual values of x and y are designated as x = g + x*. 5. (5. which deals largely with very small pressure and velocity perturbations. namely V* = (1 .e-t/r)~-Q~ n. Nonlinear models with small disturbances about some equilibrium act virtually the same as their simpler linearized models predict. (5. (5.4.59b) The differences x* and y* between the actual and the nominal values are the perturbations. equation (5.higher-order terms. (5.54) for Qi*.56) if continuous in x. such that ~ = f(~). The linear solutions are plots of the analytic solution of equation (5.4. for example. The nonlinear solutions are found by simulation. Thus. agreement would be better for smaller perturbations. With this notation. on the other hand.60) . The engineering analyst must excercise vigilence.2 Linearization of a Function of One Variable The function y = f(x). (5.~) + ~d2f x=~ (x -2~)2 ~.57) becomes y. 361 LINEARIZATION This equation is in the standard form for a first-order model. linearized models are used almost exclusively in the field of acoustics. Solutions to the cases in which the input flow is increased by 20%and by 100%from initial equilibrium values are plotted in part (f) of Fig. The same situation commonly applies regarding mechanical vibrations. in most cases.

in that case.22 (p.61) The linearization of the resistance relation applies this more general result. The derivative can be found either analytically or graphically (by drawing the tangent to the curve at the nominal point. 5. Another application is to linearization of a compliance relation between a generalized displacement x = q and a generalized force y = e.22 for the special case of a gravity fluid compliance for a tank of nonuniform area. The relation between the volume of the liquid in the tank and the pressure at its bottom is reproduced here for convenience: V = 7~tan2~ (~9)3 . ~7. the derivative (df/dq)q=. x = Q and y = P.4 equals the reciprocal of the linearized compliance.22: Linearization of a compliance This relation can be approximated.14 Linearize the compliance and the resistance of the conical tank of Example 3. df a : -~x ~=. in which x equals the volume of liquid in the tank. 165) about some nominal volmne. This linearization is pictured in Fig. The latter procedure does not require an analytical expression for the nonlinear function. or linear. P. MATHEMATICAL FORMULATION *1 _p* C P V Figure 5. or l/C*. term. by truncating the series after the first. V. at least for small values of x* and y*. and y equals the pressure at the bottom.362 CHAPTER 5. In this case. but can proceed based on a plot created from experimental data. EXAMPLE 5.~" (5. The resulting linearization can be written y* ~_ax * . and measuring its slope).

the products of its conjugate efforts and flows are not the true powers.3 (~7-~tan’~(~) C* The corresponding linearized (5.4. resistance (cdA) 2- becomes.CdA \ntan2a) EXAMPLE 5.15 Write the linearized differential equation corresponding to the results of Example 5. Qin. and identify the time constant that characterizes the response to changes in the input flow.5.49) and cdA -. 363 LINEARIZATION Solution: The slope of the characteristic at a nominal volume ~ is pg 1 _ dP ~ 3 1/3 ~-~7/" v=V. Solution: The compliance and the resistance of the combined system are at equilibrium if the flow into the tank.36) and (3. from equations (5. equals the flow out at is bottom. A linearized bond graph for perturbations about this equilibrium is as follows: \ u IflCR*" IR* Although this has the same structure as the bond graph for the actual variables. dt This is a linear first-order differential equation of the form of equations (3. The graph gives dV* 1 Q~n C’R* V* . 145-146) with the time constant .37) (pp.53). Such a graph sometimes is called a pseudo bond graph to underscore the distinction.14 above. Q. and are not even proportional to the true powers.

23: Types of discontinuities 5.in Fig. however.) On the other hand.23. The issue faced at zero pressure drop for the orifice and in part (c) of Fig. for precisely zero pressure drop and flow the linearized pressure-vs. You could well decide that this characteristic comprises an essential nonlinearity. A characteristic whichsuffers a significant discontinuity in value at a singular point clearly demonstrates an essential nonlinearity at that point.23 is even more subtle: di scontinuity of a second derivative. as shownin part (b) of the figure. which is a kind of borderline essential nonlinearity. The discontinuity in the second derivative disappears. and linearization about the origin becomes fully justified for srnall disturbances. is no longer negligible compared with the effect of inertia. The actual characteristic for a fixed orifice would resemble the plot shown. 5. that is a condition for which linearization is not justified. which implies no resistance whatever.24. The resistance to flow (slope of the characteristic) never goes below some value associated with "creeping viscous flow. The calculation in Example 5. If a discontinuity of value or of slope is small enough. Discontinuities in the third or higher order derivatives are not considered essential nonlinearities. This is correct. . the linearization still cannot be carried out formally by taking the derivative. 5. (a) value MATHEMATICAL (b) slope FORMULATION (c) curvature Figure 5." Details can be found in texts on fluid mechanics or hydraulics. it suffers discontinuity in a f ir st derivative at the point. This occurs when linearization is performed for perturbations about the state of zero pressure drop and zero flow.4. Thus. (This assumption is indeed commonly made in acoustics.3 Essential Nonlinearities There is one case in which the model for the resistance to flow of an orifice may not be considered accurate within acceptable bounds for any departure of the arguments from their nominal values. If. one might choose to overlook it in favor of some linear compromise.364 CHAPTER 5. 5.49). A possible resolution of the question for the orifice results from use of a more accurate model of the orifice flow. if one is truly interested only in extremely small flows one could choose to neglect the orifice restriction altogether. that is about the origin of the characteristic plot. This happens because when the flow is very small the effect of viscosity. instead.14 gives R* = 0 or l/R* = c~. This case is illustrated in part (a) of Fig.-flow characteristic is horizontal. which is neglected in equation (5. the resistance departs from zero very rapidly as the flow increases.

5.62) can be linearized about the nominal values x = 5.as noted before.~)+ -~uOf ~=~_ (u - (5. The area of the ven~ contr~ct~ is considered a function of somevalve displacement.4.4 Linearization of a Function of TwoVariables The function y = f(x. contrary to the normalusage in this bookbut consistent with the valve industry.63) Defining perturbation variables as before. although the more commondanger is the overlooking of significant phenomenathrough excessively crude abstractions. is an art. ~) by retaining only the first or linear terms in a Taylor’s series expansion: Y~-~ +O-~xx=5(x . Resistances not infrequently vary in response to Changesin somedisplacement. someapparent essential nonlinearities are consequencesof our modelingabstractions rather than the physics. g. The adjustable hydraulic vane.5. (s.64) It is applied belowboth to single resistances and entire differential equations. as re~resented in ~ig. 365 LINEARIZATION Bernoulli model O Figure 5. Modeling.4. labeled here a~ x: ~ = ~(~. this result can be written ~* ~ c~z* + c~u*. Such a valve is essentiMly an adjustable orifice. cu = (g. is an example.~) . Such cases are perhaps best understood in graphical as opposedto algebrNc terms. u) (5. Note that flow is plotted on the vertical ~is and pressure on the horizontal axis.24: Correction of Bernoulli modelfor viscous effects Generalizing. u = ~ and y = ~ = f(5.2g. c~ = .

respectively. The family of characteristics may be approximated in a region about some nominal pressure difference Pd. equally spaced straight-line characteristics. . are plotted in the figure for three equally spaced values of x.366 CHAPTER 5. valve displacement 5 and associated flow Q by a set of parallel. x).63) can be appled to give ~. Often an engineer will elect to start an analysis with a quick. assuming the model A = Ao(x/xo) 1"~. This plot could be refined by adding any number of added characteristics for interpolated or extrapolated values of x. These represent a linearized model. Where to set these bounds.67c) x-~’~ p~=_~ x=’~ For most cases this linearization is accurate within any desired bound for sufficiently small departures of x. as shown in the figure. MATHEMATICAL FORMULATION X P~ O_ -= "--0 pa=p~-p2 Figure 5. Q.67a) (5. it is of the form Q = Q(P~. Mathematically. and when satisfied with the general nature of the result. Pd and Q. = P~ X* ---- X -- ~. equation (5. Pd and Q from the nominal 5. refine the analysis with a relatively expensive-to-use nonlinear model. 1 _ OQ R* k =~ OPaP~=-f ~ ’ (5. rough and simple linearized model.25: Characteristics of variable orifice (valve) and linearization The pressure-flow characteristics. ultimately is a matter of engineering judgement.67b) (5. (5. Mathematically. + kx*.66) which is a function of two variables. that is how much inaccuracy to accept in order to reap the benefits of linearity.

Finally.476 The differential equation for the speedof the boat simplyrepresents that ~heacceleration equals the ra~io of the net thrus~ (F1 .Fa) ~o the effective inertia: d~ _ F~ -Fa .6k + Fa = 17.3 (p.2) = 5. F ~ 600 [ 0’0 10 20 30 40 ~ 5~0 ~ 60 This plot helps yourepresent the linearized characteristics with linear algebraic equations.5. a set of equally spaced parallel lines can be drawnto approximatethe thrust for other values: 800 100 ~c~~9 force.5) = 17.z-0.6(~ .46. workingfairly well for speeds between42 and 52 ft/s. Then. The boat and its contents weigh 2500 pounds.0480 P + 8. .59P . for howlarge a changein thrust wouldyou judge that the linearization gives a reasonable representation? Solution: The first step is to drawtangents to the drag and thrust curves at the given equilibrium point directly on the. and ~he percent throttle be P: Fa = 324 + 5. LINEARIZATION 367 EXAMPLE 5.11 ft/s dt m The linearization applies exclusively in the region wherethe boat planes. assuming a virtual inertia 50%higher than the mass of the boat and its contents to approximate the effect of the water which is accelerated along with the boat.248k + 0.16 Linearize the drag and thrust characteristics of the boat described in Guided Problem2. It totally fails to represent the decrease in drag associated with the onset of planing. given plot.ll.27.59(P . Incorporate the results into a linear state-space modelof the boat.3(~ . 52) about the equilibrium condition at 45 feet per second. Let the thrust be Fa and the drag be Fa.4.70) .11.3 k . both in pounds force.

in which ~(t) = constant and the derivative d-2/dt = O. the variables x and u are decomposed into two parts: x(t) = ~(t) x*(t). . in most cases. but this does not upset the linearization.68) The term on the left side equals a function of two variables. (5. with its nonlinear resistance and nonlinear compliance.69a) u(t) = ~(t) + u *(t).368 CHAPTER 5. although the more general case is hardly more involved. the engineer is interested in perturbations about an equilibrium or steady state.5 Linearization of a First-OrderDifferential Equation The differential equation for a first-order and excitation u(t) can be written dynamic model with state variable x dx d--i=f(x. This has been done above for the example of the conical tank and orifice.82). MATHEMATICAL FORMULATION 5. is to find the nonlinear differential equation and then linearize.Ax* + Bu *.695) The novel feature is that x and u are functions of time. A simpler and more reliable approach. Of . they represent some nominal "trajectory" in time about which the linearization is taken. The terms x* (t) and u*(t) are generally small variations or perturbations of x(t) and u(t) about their nominal values Z(t) and ~(t).~)= (5. Thus.4.64) to give the linearization dx* -d~ ~.71) One approach to linearization linearizes the individual nonlinear ele~nents before constructing the differential equation.70) The second step applies equation (5. All time-dependent phenomena are therefore left to x* (t) and u* (t). The first order of business is to solve for the which corresponds to the given ~. as always.u(t)). For the case of a constant equilibrium. so this equation is in the form of equation (5. this implies solving the algebraic equation f(~. (5. B = ~u Of ~=~ " A = ~x ~:~ (5. In general. must satisfy the original nonlinear equation. Usually. This is the only case considered in detail below. (5. These nominal values. This approach now is illustrated. and can be linearized in the same way.

(5. the linearized differential dV* ~ ~ Qi~ 6 Ao~( equation becomes ~5 ~ 3 tan2 )~/~V*. that are considerably smaller than the mean volume. 362) and 5.4.15 by linearizing resistance and compliance elements individually.72) The discussion that follows drops the final argument t.74) .71).5. u(t).~).15. (~ tan~ O01/6 and compare the result to that found in Example 5.73b) The equilibrium solution to equation (5.37) with the time constant r = Ao~ - It agrees with the results of Example5. (5. V*(t).6 Linearization of State-Variable Differential Equations Linearization of a higher-order model usually is best carried out on the individual first-order differential equations in state-variable form.17 Return to the conical tank and orifice of Examples 3. which is dV A0x/~ (3V)1/6. 5. (5.4. LINEARIZATION 369 EXAMPLE 5. Solution: The equilibrium solution corresponding to Qin = -0 is the _ 0 = Q . This is a first-order linear differential equation of the form of equations (3. d--i = Qi. restricting consideration to stationary models. and is reasonably valid for perturbations in the volume.A0 ~t~ a Using equation (5. In this case the equilibrium and perturbation variables are X:~+X * ~ (5. 5.14 (p. 165).72) 0 = f (~. V. or dx d~.36) (p 145) and (3. Linearize directly the nonlinear differential equation.73a) u = ~+ u*.15.22 (p.= f (x.

EXAMPLE 5. The particular mass of interest compresses the spring 0.76) The matrices of derivatives that comprise A and B are called Jacobian matrices.u)+ x*+ Of u* +higher order terms.= O+ = f(x.100 meters whenthe applied force F is zero. k3 = 340. find a single linearized differential equation relating F = F* to x*.370 CHAPTER 5.kl5 - ~3.18 A mass rests on a spring which in turn rests on a rigid foundation. The spring satisfies the nonlinearstiffening relation Fs = kl X + k3x3. (5. kl = 2000 rmN/m. determinethe mass.75) This result can be cast in traditional form as follows: (5. Solution: The bond graph gives the state-space model d-~ = F+mg-klX-kaz The equilibrium condition for F = F* = 0 is 0 : 0 + mg. Modelthe system. MATHEMATICAL FORMULATION The Taylor’s series expansionof equation (5. k3 a " . and determine the linearized natural frequency.000 N/m (in order to support a range of massesand applied forces without bottoming out).73) dx dx* _ _ Of -.

that is with no excitation.~rq 1- (5. m = 11.5.1) + 40.77) represent an autonomoussystem.= -q + sinp dt d-~ = p.5 5. A secondregion in state space is virtually surroundedby the line designated as sepm’atrix B. so that d~2x*_ 1 dp* _ kl4:3k~-~ 2x. This point and this region are called meta-stable.26. This region is unstable. A phaseplane plot of the system is shownin Fig.1) 3 = 24.77) has three equilibria. + dt 2 m dt m The linearized natural frequencytherefore is ’~ o)~ ~/kl + 3k3~~" -~/2000 + 3 × 40000(. to achieve a picture of the overall behavior. Trajectories in state-space are given by solid lines with arrows that designate direction with increasing time. 371 LINEARIZATION from which !(k1~ + k3~ 3) = 2000(. Anyinitial state in this region producesa trajectory that approaches and then encircles point Pl. since no trajectory ever reaches the point itself.4. using various initial states on the boundaries of the plot. each of a different type. ql in the clockwisesense. 5.44 rad/s = 1. The two approaches are developed and comparedbelow.5 kg 9. -3k~ + F*. using simulation.7 Case Study ria With Three Different IF*.1) m 24. Lookedat as functions of time. All trajectories starting above the line designated separatrix A sweep across the state space.000(. Types of Equilib- The second-order system -. The entire behavior depends on whatever initial state is imposed. The behavior can be studied globally. Equations(5. the behaviorsfor states close to equilibrium can be determined using linearization. both p and q oscillate virtually sinusoidally.81 g Thelinearized differential equation is m= 2-kl ~ p* p. With muchless effort. and continue toward infinite values.4. remaining above the separatrix. but rather encircles it indefinitely. All trajectories in the third and remaining . These trajectories were found by simulation of the equations using the standard MATLAB routine ode23.820 Hz.

MATHEMATICAL FORMULATION Figure 5.372 CHAPTER 5.26: Phase-planeplot of systemwith three different types of equilibria .

without any need for extensive numerical computations such as those that produced the trajectories and separatrixes shown in the figure.4.~ to become dp* . there results d~_Z* =~. ~ = 1 for the first equilibrium into equations (5. d--i.4886.77) linearize about any nominal state state trajectory ~. Equations (5. q3This point and region are stable. Their intersections represent the equilibrium points.79).sin~. end at point P3. In addition to points Pl. dt dq_~_*= p.81) This equation describes an oscillation with natural frequency wn = 1 and damping ratio ( = 0. Substituting the values ~ = ~r/2. ~ = 0 for the second equilibrium into equations (5. 5. Substituting the values ~ --= 0.77) give ~ -.79).26 by dashed lines.ql = ~r/2. 0.-0. which lies between separatrix A and separatrix B. there is point p.60757. Labeling such points as ~ and ~. 1 and P3. This is an unstable equilibrium point.78a) ~ = n~(1 .~/2). (5. 373 LINEARIZATION region.82) .795) (You are urged to derive these equations yourself.). because any initial condition represented by a point near but not at it produces a trajectory that heads away from it.q* + (cos~)p*. there results dp*-- q* dt dq* _= p. + ~r(~ . q:~ = -2. Equilibrium points satisfy the conditions dp/dt = 0 and dq/dt = O. (5. equations (5.5. The absense of damping renders this equilbirum meta-stable.= . " dt Combining these two equations to form a second-order differential (5.1)q*. The behaviors in the local neighborhoods of the three equilibrium points can be determined by linearization. ~ = 0. *d~-p dt ~ +p* = O. dt dq* p. dt (5.80) equation in p*.79a) (5. (5.-q*.~*’ (5.) This result now will particularized to give linearizations about the three equilibrium points. noted above.78b) These two equations are plotted in Fig.

2 ! ~/(Ms1/2Ms2)°" + Mso/Ms2 - alp*1~M~. 5. as shownin Figs. there results dp* -.___~ p .0503q*.7) substituted. (the use of operators methods as above is suggested).374 CHAPTER 5.79). Could this stability have been determinedwithout carrying out the nonlinear simulation? The nonlinear modelis represented mathematicallyby the differential equation (5.8 Case Study: Stick-Slip System With Inertia* The stick-slip systemaddressed in Section 5.~¢c’ + 2. whichin this case is stable. (5. (5. (5.. The secondterm.5. exhibits rather exotic limit-cycle behavior whenthe load inertance is small. MATHEMATICAL FORMULATION In operator notation these become (S .q*.86) Combiningthese equations. MR~eq~b.83) Combiningthese two equations to give an equation in p* *d2p * = 0 or dt ~-(~r-1) 2 [S2+(~r-1)S+(1-~r)]p The solution is p* = p~e-2"ss41t +(1-~r)p* = 0. gives *d: P (5.184 rad/s.SSb) . 286) is linearized below order to demonstrate the methodand someof its advantages and limitations.4011p* = 0.p. . Substituting the values p = -2.1.3. ~ = -0.= p .4 (p.85) where p~ and p~ are arbitrary constants dependent on the initial conditions.4886. This has a natural frequency of 1. with its positive exponent.= -0.2 and 5. (5. (5. 5. 1 .4) with the algebraic equation (5.60757 for the third equilibrium into equation (5.84) d_~_~_~ + p~e0"7425t. however. dt d* q .167. It is rock stable. (S + ~)q* = p*.the state convergeson the equilibrium point.4.88a) (5. Whenthe inertance is large. -~.87) dt 2 + 5.~p . Thelinearization of these equations is d¢5 1 -(1/CMs~)¢~: dt -.5) with the algrabraic equation (5.1)p* = -q*. Recall that the system.1297d_~t + 1. but no oscillations appear because it is overdamped with ~ = 2.6b) substituted. and differential equation (5.reveals the instability.__[_-~ -.7943p*.

u(t) = 0). There is no excitation (i.14 Hz = 1/0. (9. = O.1010 s -1 .90 Hz = 1/0. These values can be seen to agree with the corresponding plots in Fig. The damping ratio can be either positive or negative.44) (p.0022 ft lb s "~ that gave a stable simulation. 375 LINEARIZATION Substituting all the given parameter values except for the inertance I gives d¢~ _ 26. if.9~17) (5.92) dt--~-. as outlined in Section 5. Rememberthat the solution of the linear equation is a sinusoid oscillating within an envelope with the factor e-.0567/I)p*.4 and given in equation (3. Therefore.0021 ft-lb s z is 63.47) (p. and a damping ratio. What happens if the linear simulator :~sira is used for this system? The results that correspond to the three largest inertances considered before are given in Fig.91) This is a standard linear second-order differential equation with constant coefficients. a positive value 6f ~ indicates stability. A scalar second-order differential equation with either of the state variables as the unknowncan be found several ways. 149). 347) is a one-step method that gives -~ dt det(Si_A)¢~. 151). I .4 (p.0567 ~. -~-+ -~b c=0. As discussed in Section 3.0987 s -1 and for I = 0. can be characterized in terms of a natural frequency. and a negative value indicates instability.3.26.+ n~.3.89a) (5. The borderline case.93) which is between the value of 0.4.27.5.(1/I)p*. depending on the magnitude of I.e. which for I = 0.Sgb) formulation therefore gives If.6.lT_O.lb-s (5.5.0022 ft-lb s e is 62. since the envelope decays to zero in the first case and explodes to infinity in the second.17 . wn.41) (p.0021 ft lb s 2 that gave an unstable simulation and the value of 0. Thus the natural frequency is ~/I.17¢~. as discussed in Section 3. dt The state-space (5.0.w. dt dp_~*= lobby: + (0. dec (5.2 rad/s = 9.5.+ 2. Thus the model indicates instability for small values of I (note the instability of the original model with I = 0) and stability for large values of I. 5.3 and given in equation (3._d2¢~+(26. as follows: d ¢c .00217 in.~"t. 5. The linear model is inaccurate for large perturbations from . for which ff = 0 is given by 0. Equation (5.~--~.7 rad/s = 10. 346).O__~67)d¢~.

.s .0 ...0022lb. 4o / 21 = 0.ft-s 40 .6 0... / | :FAAAAAAAA 2I = 0.4 0.010lb..376 CHAPTER 5. seconds Figure 5...s -20 K 0 0..27: Linear simulation of stick-slip system 1.2 0.ft.002t lb..8 t.ft....... MATHEMATICAL FORMULATION 21 = 0.

Coding of the MATLAB ls±m command for the case with I = 0. and families of curved characteristics by familes of equally spaced straight lines. C. A=[-26.and the power dissipation function :P in Taylor’s series. and drop all . 0022.17 -l/I. Graphical linearization is carried out by replacing curved characteristics by straight lines. motivating the linearization of nonlinear models. nonlinear resistances and compliances can be linearized directly. before differential equations are formulated.u.1001) x0=[0.0567/I] .D=0. Linearization usually is applied best upon the set of nonlinear state differential equations. which is located at the point of tangency of the two characteristics and may be different from the origin of the full variables. Graphs that act much like bond graphs may be constructed using the linearized variables. The linearized characteristic passes through the origin defined by these incremental or perturbation variables. u=zeros (i. The graphical option is especially useful when you have experimental data rather than an analytical model.B. The only substantive advantage of the linear simulation is its computational speed and efficiency.0022 ft-lb s 2 can be as folows: I=O.xO) plot(t. or even their incremental variations. The results for the second and third cases also also inaccurate at first.t. A third way to linearize a system is to expand the energy storage functions 12 and 7. with linear or straight-line characteristics. when the perturbations from equilibrium are large. but becomeincreasingly accurate as the oscillations decay.-25. by replacing their nonlinear effort-flow or effort-displacement characteristics. which is essentially all there is in the first case: the result merely certifies that the system is unstable.4.001:1] . respectively. a comparison of the graphical representations of the nonlinear and approximate linear characteristics conveys a sense of how the error imposed by the linearization varies with the span of the changes in the state variables. B=[0. t=[O:. LINEARIZATION 377 equilibrium.4. lO 0.x]=lsim(A. Even if you perform a linearization analytically.0]. but such pseudo bond graphs do not represent the actual power flows and energy storages.5. Alternatively.19.C=[0 1/I].D.I] . [y. Analytical linearization of a term in an algebraic equation or of a differential equation is carried out formally by truncating a Taylor’s series expansion after the linear or first-order terms.9 Summary Most nonlinear systems behave as though they were linear for sufficiently small disturbances about some equilibrium.y) 5. which can be quite important for very complex models.

including the inertance the disturbance. and . . Linearize the following state differential equations about the equilibrium state. Apply causality.28 plus a rotational inertia of 0. and define the state variable.05 kg. Note that the equilibrium value of dp/dt is zero. [1 + (a + bq)~’]~t "~ = el ..f v~ Suggested Steps: 1. The variables are p and q.0 sin(40 t) N. the coefficients a. Label these as p and ~. MATHEMATICAL terms of third and higher order. Drawa linearized bond graph for the system.cq + d(a + bq)p dq d-~=ep. Guided Problem FORMULATION in Guided Problem 5. 379. The drive may be considered to be inflexible. Perturbations are imposed on the load torque equal to M = 2. to leave. Suggested Steps: 1. Subtract the terms of the original differential equation. determining numerical values for the resistances. You are asked to make a close estimate of the resulting perturbations of the angular velocity.m2.) 3. Find the equilibrium state or states by setting the two time derivatives equal to zero. Guided Problem 5.. 4.8 This first linearization problem employs the straightforward analytical procedure on a set of differential equations. 2. .10 on p. 5. ] are constants.9 This second linearization problem employs graphical linearization with a subsequent mathematical analysis. Truncate the expansions to the linear. Linearize the steady-velocity characteristics of the motor and the load. Expand each term in a Taylor’s series expansion. however. An induction motor drives a load that has the resistive characteristic plotted in Fig.m. from the truncated expansion. with p and ~ substituted for p and q. a very commonprocedure. 2.378 CHAPTER 5. This method is illustrated 5.only terms proportional to the incremental variables p* and q*. terms in p* and q*. (Derivatives do not always have a zero equilibirum value.

This text has not yet addressed the solution to differential equations with sinusoidMexcitations. The answer you want is the magnitue of the simusoidal motion ¢*.28: Guided Problem 5. A simple pendulum has a momentof inertia I about its center of mass and Io -. The differential equation is only of first order. and p~. (b) Linearize this characteristic. rad/s 40 Figure 5. Guided Problem 5. (a) Neglecting dissipation.5. for the gravity- and plot on the same axes as the nonlinear (c) Write the nonlinear and linear state-based differential equations. Substitute an assumedsolution of interest p~ sin 40t + p~ cos 40t into the differential equation. which equals + Checkthat the resulting perturbations of the angular velocity are not large enough to impart significant error to the linearization.9 Write the differential equation. however. and determine the coefficients pl*.10 This third linearization problem is intended-to show how linearization can be carried out in terms of energy relations. and sketch a plot of this characteristic. where m is its mass and d is the distance between the centers of rotation and mass. and the part of the solution you want is so simple that the most primitive method taught in your course on differential equations can be applied. . characteristic for comparison.4. 379 LINEARIZATION 0 10 20 30 ~. find a nonlinear characteristic based compliance.I ÷ md~ about its center of rotation.

A flow coefficient Cd can be applied. as shown in Fig. 5. Group the variables and parameters such that a dimensionless group proportional to the flow Q is a function of a dimensionless group proportional to the pressure drop P and a dimensionless group proportional to x. 3. (b) Linearize the characteristics about some non-zero operating pressure drop and flow. representing the area of the orifice as the area of the segment of the cylinder of length x. 5. using the angle of departure from vertical. The resistance to the flow of oil through a hydraulic spool valve is associated with an orifice in the shape of a very short segment of a cylinder of diameter d. Compute the torque M = dl)/d¢ and plot M versus ¢. Write the state differential equations. using as state variables either the standard pair (p. An arbitrary maximumor reference pressure Ps can be defined to make this possible. ¢. Carry out the linearization two different ways. Guided Problem 5.611.29. MATHEMATICAL FORMULATION Steps: 1. Sketch the pendulum. The result gives the linear relation between M = d)2/d¢ and ¢. and find its gravity energy V relative to its minimum for the vertical equilibrium. directly linearize the relation between Mand ¢. Sketch the characteristics using appropriate dimensionless coordinates.380 CHAPTER Suggested 5. (a) Find the volume flow rate Q through the valve as a function of the pressure drop P and the spool position. Addthe resulting linearized characteristics to your plot. do a ~aylor’s series expansion of the function V = V(¢). 2. . First. as the displacement. The fluid can be considered to be incompressible with density p. Suggested Steps: Assumethe Bernoulli equation for the flow. 4. q~). and plot. q = 4) or an alternative (¢. For the second approach. Drop terms higher than the quadratic. x. which has a theoretical value of 0.11 The last guided problem involves linearization in two variables. and viscous effects can be neglected.

P.= kx + ax find its linearized complianceC* for perturbations about the displacmentz = 5. ~. .67) (p.29: Guided Problem 5. Notethe limited-area of applicability. 5.*-x etry "~+Ap . Find its linearized resistance R* for perturbations about a nominalvelocity ~. Equation (5.41 A dashpot obeys the relation F = a~3/2.5.42 A "stiffening" spring satisfies the force-displacementrelation ’~ F . Sketch-plot the results with the first group above as the ordinate. Figure 5. Drawa set of equally spacedparallel lines that correspondto the linearized characteristics. 4. 381 LINF~ARIZATION . 5. 366) can be adapted to the dimensionless groups. PROBLEMS 5.11 3. Carry out the linearization about someoperating point Q.4. the second group as the abcissas and the third group as the parameter which distinguishes three of four different curves.

0 in.43 A vehicle weighing3000lbs. 15 torque. Youare asked to find howthe speed of the vehicle increases with time.. (a) Estimate the initial and terminal thrust forces. The thrust force then is increased suddenly by 10%. MATHEMATICAL FOR_MULATION 5.25 in. (d) Solvefor the velocity as a function of time. lbs. (b) Linearize the drag characteristic. ~0- 2000 speed. 6O drag force. and estimate the terminal speed.-lb 10 1000 (a) Determinethe equilibrium speed. sketch-plot the result.44 A series-wour/d DCmotor has the torque-speed characteristic plotted below.and maintained at that level. rpm 3000 .lb at its equilibriumspeed. It drives a load with a total massmomentof inertia of 0.-lb-s 2 and steady-state torque of 5. 40 2O 0 I 0 I 20 I I 40 I I 60 80 speed.fl/s 5. is driven at a constant speed of 50 ft/s. in.382 CHAPTER 5. (c) Drawa linearized pseudo-bondgraph and write a differential equation describing the transient behavior. The drag characteristic is plotted below.

State the values of all constants.45 Briefly describe under what circumstances the following situations give rise to essential nonlinearities: (a) coulomb(dry) friction (b) fluid checkvalve (c) electrical diode (d) finite-sized fluid gravity tank of uniformarea (e) a simply mountedcompression spring 5. dq 2 x/~ p2 4pq u dt 7r dp 10 cos q dt 5.0 in.. m3/s 3V. noting the values of any time constants natural frequenciesthat exist.4.= -6v~ + 3u dt 5. (d) Sketch the speed vs. for ti~ne t > 0.47 Linearizethe followingset of state differential equations about the smallest positive equilibrium correspondingto ~ = 2. (c) Drawa linearized bondgraph and write a linear differential equation that whensolved approximates the transient change in speed. : C Q. m as . Determine its final speed.48 A system has been modeled using a bond graph and characteristics shownbelow. ~*(t). 5. dx -.lb.383 5. LINEARIZATION (b) At time t = 0 a brake applies an additional torque of 1.46 Linearize the following differential equation about the equilibrium corresponding to ~ = 4. time.

At this point you may leave the source and load charateristics as unspecified functions of specific variables.49 The model described by + 2x = u(t) dt~-~ + 2 has a nominal input u(t) = ~(t) = 2 + (a) Find the nominal response. 5.4 (pp. 316-319) and Fig. areaA (a) Model the system with a bond graph. 5.50 The system with an axial-flow fan and tank given in Problem 4. (b) Linearize the differential equation for the responses x*(t) = x(t) -~(t) to the disturbances u*(t) = u(t) -~(t). .13 and modeled by equations (5. (f) Determine the borderline value of 5. 277) is. (c) Linearize the resulting differential the equilibrium state. Hint: ~ = a + bt . (b) Write a set of differential equations for thee model. ~(t). equation(s) for disturbances about 5. fa~n L ~ m[~ orifice tube. (d) Determine the natural frequency and damping ratio of the linearized model.2. as pictured below.17) .384 CHAPTER 5. Consider the case of orifice #1. (e) Show that a small value of L produces instability (and by inference limit-cyle oscillation).(5. Take particular care with the signs. modified by placing a duct of length L = and cross-sectional area 3 × 3 inches. MATHEMATICAL FORMULATION (a) Annotate the bond graph as needed and find the state equation(s).19).60 (p.51 Consider the ground effect machine (GEM)described in Section 5. (c) Linearize the model of part (b). differential (b) Determine the equilibrium state(s) of the system if P0 = 50 2. and a large value of L produces stability.

56 (p. I -2 W = 5 lbs -1 1 2 3 k=21b/in. . equations for perturbations about the equi- (c) Determine the natural frequency and damping ratio of the vertical (heave) behavior. and estimate the corresponding incremental or linearized resistance. of the equilibrium 5. 5./s ~ (a) Draw a bond graph for the system. Inertial effects in the standpipe may be neglected. for the upward flow for the region -2. (c) Linearize the friction characteristic for the equilibrium state of part (b). lbs. 274).5 (b) Represent the dynamics of the system by a bond graph.52 A memberwhich moves at a controlled speed ~ pushes one end of a linear spring.14. the other end of which pushes a weight at velocity 3. (b) For the case & = 1 in/sec. Comparethese results to the corresponding nonlinear simulation in Fig. 2friction force. the outer two of which are stable. allowing the linearized characteristic to substitute for the actual characteristic. neglecting the effect of the inertia. R*. The net upward flow in the standpipe has been plotted as a function of the deviation of the pressure just above the jet from its nominal value.53 Consider the standpipe and fluid jet of Problem 4.Ps < O psi. (a) Linearize the characteristic P. Three equilibrium states have been found. 5. determine whether the corresponding equilibrium value ~ = 1 in/sec is stable. I I 4 I I I I 5 6 ~. and describe the character of the behavior. 385 LINEARIZATION (a) Determine the equilibrium values of the state variables y and (b) Linearize the differential librium.4. (d) Determine whether the inertance affects the stability state.5. in. The weight slides over a horizontal surface with the friction force as plotted below.

~ _ ep -~ q*. Compute the transient and steady-state response of the level of the free surface. SOLUTIONS TO GUIDED PROBLEMS Guided Problem 5. 4.cqo + d(a + b~)-~.m. rad/s Fromthe linearized characteristics 40 superimposedon the plots above. using the model of part (b).1.s load: M* = -b~*.s .386 CHAPTER 5.8 1. MATHEMATICAL FORMULATION (c) The jet is suddenly turned on.9 ’ ’ I ~. If~-~=0and ~-=0. After the equilibrium terms are substracted. Guided Problem 5. a = 3. following a quiescent state. dp" ~ ~ -ca* + 2d(a + b~)~p* + db~2q dq* ~ . ~ = -~ db]~ ¯ ~ db]~ ] dbf~. C2 or q + dbf2 Therefore. motor: M~. ~+~=0.94 N. then and ~= 0 = el .db~:q dq*a d~ Since ~ ~ = 0.~ I o[or 40 ’ N. this gives [~ + (~ + ~)~] *.m. +d(a + b~)2~p* + *. b ~.m 20 01 0 I ~ 10 ~ I 20 30 ~.86 N. = -a~*. or +f e2 -- ~.

-~madO -mgdO d-~ =.+ 41.024 so that the magnitude of the sinusoidal response of ~ equals X/(0. -~.40p~p2= 2. ~1 = (dM)_~_ o=o=mgdc°sO=mgd..Io or dO -~ o z . mgd~ 0 z/2 t? 3.= M" .---i--p 3.6(p~sin 40t + p~ cos 40t) = 2 sin from which 40p~+ 41. M = -~ actual . . giving p~ = .. -~.p~ sin 40t) + 41. + ~ .6p* = 2sin40t. . 387 LINEARIZATION MOTOR LOAD ! dp* a. The suggested substitution gives 40(p~cos 40t .(-~’) = M* .4.693 rad/s.0..l~ dT = mgsinO 2.cosO) linearized M]characteristic/~_.0.025)~ + (0. = 0.= mgy = mgd(1 .05 = 0. T = m#d 1 . 5..6p~---.10 7.024)~/0.a(~" .025 andp.-~. Fromthe plot it can be seen that the magnitude of the perturbations ~* are so small that the linearization introduces negligible error..b. alp* Therefore.5.6p~ . Guided Problem 5. ( ) dT M = --~ = mgdO p/ o = -~ = Co dO 1 -~ Top ~~for small 8..theref°re’C°= lmg-~ O~ 04 06 4.1 + -~. 41. 4.

respectively. Q. Q ~r ~~ =x/x. Zm --~0 there results Q0= 0. The area of the orifice Bernoulli’s equation.388 CHAPTER.. 5. 0.2 00 AP/Ps )Q = qo + O(A-~/po AP/ps ~ ( AP .5. See the set of parallel lines superimposedin the plot above (step 3).Po eo.11 1..8 0. ~0. .707.424.~o x AP For the special case -. = 0.6 0yz~--~..~Trd x.. 2x~ 0(A~Ps) Po. The dimensionless flow rate is CdX.~ V Po where xm and P0 are the nominal maximumvalve opening and the maximum applied pressure.6 and = 0. 5.4 0. according to = cdAV/~ where Ap is the pressure drop between the inlet and outlet channels. and Cd = 0.5= 0. is is A = x~rd.~o\ "P-: OQ J + ~)Po.424. .= 0.611.6 ~ ....~o =1x~1 (oQ) O(~) = = ~(0. The volume flow rate. Guided Problem MATHEMATICAL FORMULATION 5.6)~ = 0.

Classical or direct methodsof solution are given. despite being a small subset of all possible models.Chapter 6 Analysis Part 1 of Linear Models. Application of the property of superposition is used to find the response of a system to complicated inputs.that is differential equations in terms of a single dependentvariable.1 Direct Solutions of Linear Differential Equations This section reviews the solution to the general nth-order scalar linear ordinary equation with constant coefficients.. Thedirect solution of matrix differential equations is deferred to Chapter 7.39) and 389 . Second. along with the formal use of transform methods.3. which is given as equation (5. 6. There are two reasons. Later. Linear modelshave alwaysenjoyed a special status in engineering. regardless of whether mathematics is employed. This chapter deals with scalar differential equations. Several of the calculations are expedited by MATLAB.4. The significance of this fact has not waned. predictable ways. linear modelscan be analyzed mathematicallywith relatively little difficulty. the behavior of linear modelscan be characterized in vastly simpler ways than almost all nonlinear systems. Anextensive study of the response of linear models to sinusoidal inputs is based on complexnumbermethods. Before the advent of the computerthis fact forced linear modelsto the forefront. Sucha formulation alwayscan be deduced from a state-variable formulation by using the methodsintroduced in Section 5. but the present ubiquity of simulation routines weakensthe motivation. First.well engineered systems ought to behavein simple. so linear behaviorremains a frequent object of design. operatorbased methodsusing an expansion of the transfer function are seen to be more efficient for high-order models.

Its solution. plus any particular solution of the full equation. The homogeneousdifferential equation can be written in an algebraic form through the use of the operator notation introduced in Section 5....~_~S~-~ +. all the coefficients and the solution are zero.2) The existence and uniqueness theorem states that this solution is unique.4) This is an nth order polynomialin S whichpossesses n solutions.1) The consideration of impulses as disturbances or input excitations maybe new to the student. ANALYSIS OF LINEAR MODELS. designatedas xp(t). provided that precisely n independentconditions are specified.1. Whenthe excitation f(t) is non-zero..3) Since the general Xh ~ O.1): x(t) = ~h(t)+ ~p(t).. the homogeneoussolution forms a part of the complete solution.3.. (6.390 CHAPTER 6. Most commonly. + b~-~ (6. and usually are non-zeroevenif all the initial conditions are zero.. This solution contains undeterminedcoefficients equal in numberto the order of the system. In this case the undeterminedcoefficients dependpartly on the excitation.. as indicated by Equation (6. Their values are determined by an equal numberof conditions that must be specified to make the solution unique. Someresults are given without derivation. + a~S + ao = O.2(0). Theseconditions most commonly comprise values at the initial time: x(0). that are labeled here as S1.1 The Homogeneous Solution The homogeneous solution is the complete solution whenthere is no excitation f(t). If all these conditions are zero. PART repeated here: dn-lx d~-2x dx an~Z+a~-ld~-~_~+a~-~d-~ + "’" + ~1~+ ao~ = f(t) d~-lu =b . or roots. d’~-lx/dt’~-~It=o.2). S~.. + a~S + aO)Xh = 0.1 derivatives... du + bou..3 (p. .. I (6. It is necessaryto find these roots. equals the general solution to equation (6. 6. 345): (a~S~ + a~_~S~-~ + a. you are referred to almost any mathematicaltextbook on linear differential equations for a rigorous development. S~. these conditions are specified at time t = 0 and are called initial conditions. the following characteristic equation results: ~n an + a~_lS~-~ + a~_~S~-2 + . The most common initial conditions specified are the values of x and its first n . I (6.~d~udtm + bm-~ ~ +. regardless of whetheryou use the direct solution techniqueor a fancier transformtechnique.. The homogenousequation is the differential equation with f(t) = O. designated xh(t).

~) c~.12b) there results [Xh = e-at(C~COSW~t-. DIRECT SOLUTIONS OF LINEAR DIFFERENTIAL EQS. Real roots $1 and of a second-order equation give the solution Xh = cle -t/~l -t + ITs.8) between the two forms of the coe~cients can be seen to be The homogeneous solution becomes Xh = c~e(-a+j~")~ (+ a-j~")~. The simplest example is the first-order differential equation. ~ ~ -1/S~.[ its roots are. (6. sinwat). 146).-1/~-. This happens ff and only if the coefficients c~ and c~ are also complex conjugate: 1 (6.J (6. A real unrepeated root Si contributes the term cie s~t to this solution.48) (p. this solution must be real. > 0.14) . 391 The homogeneous solution comprises a sum of terms associated with the individual roots.~2Wn = -a ¯ jwd.10) To be meaningful physically. c2e T1 ~ -1/S~. 151). If the second-order characteristic equation is written with the familiar notation (s Is + (6.44) (p. (6.j sin~. Second and higher-order charateristic equations often have complex roots. The relations (6.13) equivalent forms (6. when the dampin~ratio ~ < 0. 153). giving IXh = ce-t/’. its negative reciprocal Ti has the more palpable dimension of time.1.2 = -¢wn ~ j~ .6) which is consistent with equation (3. ~ < 1 or w. which come in complex-conjugate pairs.12a) (6.c. for which the only root is $1 = -ao/al =. where ci is an undetermined coefficient. 8~. Alternative with conceptual and plotting adv~tages are Xh = ce-at COS(Wdt+ a) = -at si n(walt + ~) (6. c~e (6. With the use of the identities ej" ~ cos~ + j sin~. e-i~ ~ cosa .] The result agrees with equation (3.:= ~(c~¯ ~c~).38) (p. Recall that the symbol called a time constant.6. while Si h~ the dimension of inverse time.7) =0.5) This solution is consistent with equation (3.

100x = 2 + 3t Solution: The characteristic equation is S5 +9S 4 +43S3 + 71S ~.2. ANALYSIS OF LINEAR MODELS. which have five’undetermined coefficients. resort to a numerical solution is necessary. similar to the case with repeated real roots. and the pair of roots S = -3.100 = 0. No general analytic solution exists for a fifth-order polynomial.15) which corresponds to equation (3. 0000 -2.0000 +4.0000i -3. + C~mt’~)e The example of a second-order equation with two equal roots $1.[ + -~ + 43-~-f + 71~-~ dt .ssolution of the fifth-order differential equation d2x d3x d5x 9 d4x dx 24 dt-.392 CHAPTER 6. and the model is unstable.0000i contributes (c4 sin 4t + c5 cos -3t 4t)e which.also can be written as c~ sin(4t +/~)e -3~.0000i -2. the homogeneoussolution can be written in either of the equivalent forms Xh = c~et + (c2 + c3t)e -2t + c4e-3t sin(4t + ~). A repeated pair of complex roots (which doesn’t happen very often) produces a combined contribution of the form [c~0 sin(wrist + ~o) + C~ltsin(wd~t + ~) +’’ -at. Collecting these terms. 151). S~. Xh --~ oe as t --~ oo. 0000 TherootS = 1.24S. 0000 i.0000contributes thetermc~e~ to the homogeneous solution.1 Find the homogeneou.0000 contributes the terms (c2 + c3t)e -2t. repeated m times. EXAMPLE 6.. PART 1 A multiple real root of the characteristic equation.2 = -w~ (corresponding to a = -w~. . The pair of roots S = -2. c~t)e (6. Xh = c~et + (c2 + c3t)e -~t + e-~t(c4 sin4t + c5 cos4t). Since this root is positive.. +. To solve by MATLAB.47) (p. for which Wd = 0 and ~ = 1) gives Xh = (c~ + -~"~. you enter and receive noindent>> roots(J1 9 43 71 -24 -100]) -3.0000 -4.0000 =h 4. contributes to the homogeneoussolution the terms (c~o + c~lt + cet 2 slt.

1. 393 DIRECT SOLUTIONS OF LINEAR DIFFERENTIAL EQS.1. As pictured in Fig.6. however.1: Unit step waveform area = 1 lIT area = 1 area 0 0 t 0 t Figure 6.1. If there is an input. The unit impulse function 5(0 approximates a sharp pulse disturbance. 6.1. is defined as u~ (t) for t < 0 {01 for t > 0 (6.2: Unit impulse waveform A methodfor determining the undeterminedcoefficients is given in Section 6. followingthe discussion regarding the particular solution. 0 time Figure 6.2. it is defined as the limit of the square pulse of width T and integral (area) 1 as T -~ . The unit step function us(t). 6.the determinationof the coefficients must be deferred until the particular solution is addedto the homogeneous solution. Shouldthere be no excitation or input disturbance.6.16) Step changes of other amplitudes or dimensions can be represented by multiplying the unit step by a constant.2 Singularity System Inputs A family of singularity functions that is zero for all time t < 0 is particularly useful. pictured in Fig. 6. the proceduredescribed there can be applied directly.

rampsand impulses to be represented as a sum of singularity functions.3. (6.the time derivative of the unit step can be considered to be the unit impulse.4. 6. T-. 6. Conversely.394 CHAPTER 6.19) The unit rampfunction uT(t).(t) = --~. .0) as the argumentof the singularity function.t* in place of t (or t .o ~r(t)= 0 for t < 0 lIT O<t<T 0 for t > 0. This possibility considerably enhancesthe usefulness of singularity functions. PART 1 0 time Figure 6. although discontinuous functions are not differentiable in the formal sense: 5(t) = dus dt (6. ANALYSIS OF LINEAR MODELS. The property of superposition then can be invokedto represent the response of a particular systemto such a function as the sumof the responses of the individual singularity functions. This allows functions composedof a series of steps. exampleis shownin Fig. (6.20) The rampis the integral of the unit step. and the unit step is the derivative of the ramp: ur = us(t) dr.21) OO A singularity can be located at a time t = t* different from t = 0 by writing t .18) ’_ ¢J(t) dt =us(t). u.3: Unit ramp waveform ~(t)= limiT(t). pictured in Fig.17) The time integral of the unit impulse is the unit step: (6. is defined as increasing linearly with derivative (slope) 1 from the origin: t <0 ur = t0 for fort>0 (6.

6. The more general case of complexconjugate pairs .3 Exponential and Sinusoidal Inputs The exponential input u(t) = st occurs naturally as theoutput of a fir st order process and as componentsof the outputs of higher-order processes.5 ur(t) 1.as before.22) which is plotted in Fig. -0.~ = ½(cr =l: jci) will be assumed. as indicated in Fig..~’~=0..5. The special case of pure imaginary values of s. (6.. just as the complexconjugate pairs S and c are required to contribute real componentsto the homogeneous solution of equation (6. whereasnegative real values of s producefunctions which decay exponentially. e ~’ ~ ~~. Therelation c1.(t-l) .. DIRECT SOLUTIONS OF LINEAR DIFFERENTIAL EQS. imaginaryor complex.1.13)..ci sin wt. 395 0.5 ur(t-2) 0 1 2"". The coefficient s can be real.05 0 2 s 4 =-0.05 s=O ]~/-- s=-O.5 ur(t-2) Figure 6.6.6.2 6(t-2) .5: Real exponential waveforms 6. written as s = =l=jw. 6.0.5 u...~.gives the sinusoidal waveform u(t) = cr cos~t .1.4 6 8 l 10 ~ 12 Figure 6.4: Time-shifted singularites and waveformsynthesized therefrom 2 r plots u/c l 1 0 ~ ~- of .. 3 4 ~". Complexvalues of s and c must appear in complexconjugate pairs to produce a real function u(t). Positive real values of s produce functions which grow "exponentially" in time.

envelopeae """-’---..cl sin ag Figure 6. as shown in Fig. particularly whenthe concern is vibrations. Virtual sinusoids with amplitudes that grow or decay exponentially in time also are important.23) This waveformcan be viewedas a sinusoid oscillating at frequency w within an envelope of amplitude ~eat. Power-lawfactors occur naturally in the outputs of modelswith repeated roots. 6. . where n is a positive integer.jci)e =eat (cr cos wt .1.6: Sinusoidal waveform "--. (6. as noted above.c~sin wt). ANALYSIS OF LINEAR MODELS. 6. PART 1 cr cosag .396 CHAPTER6. The excitations or inputs to engineering systems frequently are assumedto be sinusoidal. since they often emerge from commonsystems with singularity inputs to becomethe inputs to other systems.. is used as an input to a system.7i Waveformfrom complex conjugate pair s = a =l: jw gives 1 1 u(t) = -~(cr + jci)e (~+~)t + ~(c~ .4 Power-Law Inputs Occasionally a memberof the power-lawfamily u = ct n.7.qsin 6)0 ~ Figure 6._ e~ (cr cos a~t .

25b) Thusc equals the sametransfer function G(s) as given by the operator notatiort of equation (5. substutite themfor Xp in the differential equation. 345) to represent the general linear differential equation with constant coefficients.an-18 n-1 + " ¯ " ¯ ao "t. + bo)uoest. This correspondsto the secondentry in Table 6.1. (an8 n q.5 DIRECT SOLUTIONS OF LINEAR DIFFERENTIAL EQS. and evaluate the coefficients by satisfying the resulting equationfor all times. particularly for sinusoidal inputs. it might as well be the simplest possible solution. except that the original argumentand derivative operator S is replaced by the exponential coefficient s.25a) (6.1 term in f(t) 1 st e nt Method of Undetermined Coefficients corresponding terms in Xp c st ve Cntn q. All that needsto be done is to affix initially undeterminedcoefficients to each of the applicable terms. (6. (6. 6. c = binsmq.1. [ [xp(t) = G(s)uoe = G(s). The Method of Undetermined 397 Coefficients This is the classical methodfor finding the particular solutions to linear ordinary differential equations with the various types of inputs f(t) detailed above. Carrying out the methodof undeterminedcoefficients by substituting these into the differential equation (equation (6.6.45) in Section5.24) +ao)cuoe Solving for c and thus zp(t). it is employedbelowextensively.¯ " ¯ -~ Clt -~" CO sin wt or cos wt or cos(~t + #) ca sin wt + c2 cos wt or c cos(~vt + # + The particular solution for the step input can be seen to be xr = constant for t > 0.Cn-1tn-1 ~.3. The forms of the terms in the particular solution that correspondto the various input terms are given in Table 6. .1. This fact is very useful.1)) and performing indicated differentiations. Table 6.1.3 (p.an-18 n-1 -~’" ~t = (bins m + bm-lsra-1 +. The particular solution can be any solution to the completedifferential equation. The exponential input u(t) = uoest gives the form f(t) = foest.’. the particular solution is of the form xp(t) = cu0est.bin-1sin-1 + "’" +bo an8 n q. t.

24(2c2t + cl) . equation and its roots are S~ + 2S+ 5 = (S + 1) 2 + (2) 2 = 0. Solution: The characteristic equation for the given initial x(0)=0.3 Solve the following second-order differential conditions: d~ x 2 dx dt ~+ dt +Sx=3(1-e-2t). This case is illustrated below.1. 6. The particular solution itself should at this point have no undetermined coefficients. The result is x~ = -O. with n = 0 and n = 2. clt + co) = 2 + 3t from which -100c2 = 3. $1.2 = -1 =t:j2. The undetermined coefficients in the homogeneouspart of the solution are resolved through the introduction of the same number of independent conditions.0144t . In the most commoncase these conditions are given at t -.0.24c~ . ANALYSIS OF LINEAR MODELS.100(c2t 2 + 2.03t 2 + 0. so that the homogeneoussolution is Xh = e-t(c~ cos2t + c2 sin2t).046056.2 Find the particular 6. 5(0)=0. PART 1 solution to the differential equation given in Example Solution: The terms on the right side of the example. are of the power-law form t ’~. EXAMPLE 6. such as a final time.0 and therefore are called initial conditions. EXAMPLE 6. . if it is nonzero. Table 6.c2t2 + Clt + Co.100ct = 0. Substitution of this solution into the differential equation gives 71(2c2) .100co = 2.1 gives the form of the particular solution as follows: Xp -.398 CHAPTER 6.1. 142c2 . namely f (t) = 2+ ~. -48c2 . Whenone or more conditions are specified at a different time. the procedure is essentially the same.6 Application of Initial Conditions The undetermined coefficients in the homogeneoussolution must be found after the complete solution is formed by adding the particular solution.

Substitution of these values gives the unique result x = 0.7 0 Solutions 1 to Impulse 2 3 4 t 5 6 Inputs The impulse function f(t) = 6(t) is not included in Table 6.3 2 3 e_2~ = 0.6(1 e-2~). giving simply G(0) = 0.and time ~ = 0+.8 0.e-2~).6. 5 s + 2s + 5 8=-2 Note that ul is in the proper exponential form with s = 0. along with its homogeneousand particular components.c2)sin 2t] + -~ [(2c2 dt Substitution of the initial conditions into these equations specifies the coefficients Cl and c2: 0=c~+0 0=2c2-c~+1. d__~_x = e.e-2t . it is useful to say that its duration is betweentime t = 0.1. The excitation is zero for t > 0+.e-t sin 2t). This is plotted below. 0.6(1 .4 6.6. which meansthat the total solution has identically the same form as .6 for its response.2 or c1=0 or c2=-0.6(1 . Thus the complete solution and its time derivative are -2~. the particular solution can be found from equation (6.25b) as follows: xp = 3G(0) . since it acts in a special way.3G(-2)e -2~ .Cl)cos2~ .The entire excitation lasts but an instant. Notethat the given initial conditions are satisfied. and that virtually any other initial conditions also could be satisfied by different weightingsof the two components. x =Xh + Xp = e-~(Cl cos2t -+ (2cl c2 sin2t) + 0.1. DIRECT SOLUTIONS OF LINEAR DIFFERENTIAL EQS.1.4 -0. 399 Treating the input as u(t) = ul (t)+ u2(t) with ul (t) u2(t ) = -e -2~.

26) = c. For the first-order case.~_~ = c. this integration is _ dt +rJo- xdt= c~I(t)dt. (6.27) which gives + T Note the use of the fact that the integral across a unit impulseis 1. This gives x(0+) =or ~(0+) = -acr . Theintegral term in equation (6.27) must be zero unless x were infinite over the vanishingly small interval betweenthe limits of integration. as given by equation (6.13) and repeated here: x =e-at (cr cos Wdt-. >_0 . The jump x(0+)-x(0-) can be determinedby integrating the full differential equation betweenthe times t = 0. (6.~dC~ + t . Solution: The complete solution (for t >_ +) i s t he s ame as t he homogeneous solution.and t = 0+. dx/dt+ (1/T)X = C~(t). 0 (6. ° Onlythe highest order derivative contributes to the left side of this equation. ANALYSIS OF LINEAR MODELS.4 Find the response of the general underdampedsecond-order model to an impulseof amplitudec. with the result that the impulse imparts a non-zero value to only this term: dt.ci sin Wdt). Since x(0-) is by definition zero.28) For the nth-order case. whichgives -t x /r = . the result is x(0+) = c.400 CHAPTER6.ce t > +. (6. for both general initial conditions and for zero initial conditions.30) EXAMPLE 6. The effect of the impulse is to impart an initial condition for this particular solution at time t = 0+ that is different from the condition at time t = 0-. whichit is not. PART 1 the unforced or homogeneous solution.

then x2----’dr (6.n. plot the unit impulse and step responses of the model d3x d2x dx dt a + 3~.~(0-) -.31) Thus. if the excitation u2(~) of a systemis the time integral of the excitation u~(t) of the samesystem.’ "’.c." ¯. EXAMPLE 6. 6. 6. Therefore. for example.x(0-). and then computingits derivative. traditionally written 9(t). similarly. for example. equals the time derivative of the reponse to a unit step. then the response x2 (t) is the time derivative of the response x~(t). That is.8 Differentiation and Integration Properties If an excitation u2(t) of a systemis the time derivative of an excitation ul(t) to the samesystem. bo of the numeratorpolynomialand the coefficients an.1. (6. and then simply call the routine of interest. taking care not to overlookany discontinuities. +) .9 Step and Impulse Responses Using MATLAB MATLAB automatically computes the step and impulse responses of linear models. a0 of the denominatorpolynomial.32) Thus.6. the response of a given model to a unit impulse input. dul if u2=--~. The programchooses its own(default) scales for the axes (which you can override). h(t) = fg(t) Theseproperties allow you to extend a knownsolution to certain other cases. Inversely. 9(t) = dh($)/d~.+ 102~ + 100x =50du dt + 200. the response of a given modelto a unit step input equals the time integral of the responseto a unit impulse.1. . the response x2(t) is the time integral of the responsex~ ($). Rather than go through the analysis above to find the response to an impulse. 401 DIRECT SOLUTIONS OF LINEAR DIFFERENTIAL EQS. Fromequation (6. -~[ax(0-) + ~(0-) Should the initial conditions x(0-) and ~(0-) be zero. since the impulse is the time derivative of the step. you have the option of finding the response to a step. c~ = x(0-).5 Using MATLAB. traditionally written h(t). if u2=/Ul(t)dt then x2(t)=Ixl(t)dt. x(0+) -.1. the solution plifies to x = (1/w4)e-a~ sin w4t.30). Since only the highest-order derivative changes across the impulse. Youenter the coefficients b.

5 1 0 0 0.5 2 2.402 CHAPTER6.5 3 3. >>step(num.5 3 3. ANALYSIS OF LINEAR MODELS. you enter >>nam= [50 200].me (ram) 4..5 4 2.S 1 1.5 2 O O. >>den = [1 3 102 100].5 1 1.5 . den) >>±mpulse(num.5 4.den) ~r~ 2 1.5 4 . PART 1 Solution: The transfer function is 50S + 200 G(S) = $3 3S2 + 102S + 100" To get the plots below.

Carrying out this promiseis the subject of most of the remainderof this chapter and Chapter 8. 05 : i0]. x. 2. It also illustrates an important phenomenon.10. Guided Problem 6.1 This straightforward problemreviews the classical methodfor solving linear differential equations. whichwill superimposesubsequent plots. Use these to write the homogeneoussolution with its two unknownconstants. .e.95cant) "t <0 t > O. Tospecify 201values of time evenly distributed over ten seconds. and exponential and power-lawwaveformshave been featured as simple system inputs with readily calculable particular solutions. Thesewaveformsare particularly useful. the times used stored in a vector t. The particular solution can be found.1. Suggested Steps: 1. because they can be combinedto represent muchmore complicated waveforms. for the cases of primary interest. unless you enter hold on. The first case addressed in detail. the output values are stored in a vector y and the state variables chosenby the routine are stored in a matrix x. DIRECT SOLUTIONS OF LINEAR DIFFERENTIAL EQS. Singularity waveformsincluding impulses. Find and plot the solution x(t) to the followingdifferential equation. using the methodof undetermined coefficients. Valuesof t specified in advancewill be used if t is used as the third argumentof the command. Further options to control the details of the plots and secure data files for further use are available. for example. subject to the initial conditions x(0) = 0. 403 Pressing the enter key will erase the present plot. Note that the equation can be simplified by employingt ~ = wnt rather than t as the independentvariable.6. den) or [y. ~(0) i d2x ( 0 w-~dt--~ + x =~ sin(0.10 Summary Theclassical representation for the solution of a linear differential equationwith constant coefficients comprisesthe general solution to the homogeneous equation (with zero right side) plus any particular solution to the completeequation.t] =impulse(num. If you substitute [y. den). 6.1. i. in the following section. den. Write the characteristic equation and find its roots. step (num.201) or t= [0 :. steps and ramps. The homogeneous solution is found using the roots of the characteristic polynomial formedfrom the coefficients of the various terms. is that of sinusoidal excitations. t] =step (num. The final step in finding the unique solution to a particular input is the determination of the coefficients of the homogeneous part of the solution throughthe use of the initial conditions. you enter either t=linspace (0. t). x.

4. 6. Substitute the initial conditions into the total solution to evaluate the unknownconstants. Se e---£-~ I 1 ~ 0 ~R C (a) Definestate variables and find a set of state differential equations.5 Solve the preceding problemwith the coefficient 4 replaced by 6.4 Solve the followingdifferential equation with the initial conditions x(0) = and 5(0) = 0.1 (p. ANALYSIS OF LINEAR MODELS. d2x 4dx dt ~ + dt -b 8x = lO 6.6 Solve the preceding problem retaining the coefficient 6 but changing the coefficent8 to 9.2 Answerthe precedin~question substituting u(t) 10sin2t. Plot the results using MATLAB or someother software package~ The result demonstrates a "beating" phenomenonbetween motion at the natural frequencyand motionat the slightly different forcing frequency. 6. Sketch-plot your answer. Sketch-plot yo ur answer.1 Solve the followingdifferential equation with the initial condition x(0) = and the forcing function u(t) 10. 6.7 For the bond-graph model below.404 CHAPTER 6. . 397) to find the particular solution. wheree0 is a constant. which is perpetuated because of the absence of damping. PROBLEMS {}. 5. d’-~ +5x = u(t) 6. Use the methodof undeterminedcoefficients as summarizedin Table 6.3 Answerthe preceding question substituting u(t) 10(1 6. (b) Find the step response of the displacement on the compliance for e = eous(t). PART 1 3.

(b) Representthe input u(t) as sum ofsteps. (c) Sketchthe response of the systemas the superposition of the responses to the individual steps in part (b).10 An input to a system is as plotted below: u(t)21 / I 0 1 2 I t 3 (a) Representthis waveformas a sumof singularity functions. (b) Find the response to the first-order system dx/dt + x = u(t).11 Carry out the above problem for the input plotted below: I I 0 I 2 4 t 6 .6. {1. of the system.8 A particular system responds to a step excitation with a ramp response. (a) Carefully sketch and label the step response. 405 (c) Find the responsefor e = e0 sin wt valid for t 6.1. DIRECT SOLUTIONS OF LINEAR DIFFERENTIAL EQS.9 The excitation u(t) and impulseresponse g(t) of a linear systemare plotted below. h(t). 6. Characterize the dynamicsof the system generally. 6.

. which gives a force equal to 67r#rv and applies with considerable accuracy for Reynold’s numberNR= 2pvr/# < 2 and can be adequate for 2 < NR< 10. PART 1 {1.s/in 2 (water at 70°F). 6. 6.14 Find F(t) for the system shownbelow with x = xous(t).due to the inertia of the sphere and someof the surrounding fluid. The terminal velocity sometimesis not reached quickly.. k k/ F(t) ~~x 6. dl = 1 in.s.406 CHAPTER6. and # = 2.s/m2.9 x 10-6 lb. (b) Find the time constant for the motion. as shown. ANALYSIS OF LINEAR MODELS.13 The combination of two springs and a dashpot shownbelow are hit with an impulse (perhaps from a hammer)with f F dt = Z lb. .15 One way to measure viscosity is to observe the velocity v at which small spheres settle in still fluid. in terms of the symbolsgiven.12 A spring pushes a piston whichforces a viscous liquid through a tube. (a) Find the effective resistance of the systemwith respect to the motion of the piston. Assumefully developedflow. Find the resulting motion. Evaluate for k = 50 lb/in.13 cm and density Ps = 7840 kg/m (steel) settling in a fluid of density p = 1200 kg/m3 and viscosity # = 0. AssumeStokes law. 3Consider a sphere of radius r = 0. L = 40 in.05 in. however.20 N.. d2 = 0.

PL= pghL check valve Find the quantity of water pumpedto the load tank in one cycle. The latter approximately equals the mass of fluid that occupies a volumeequal to one-half that of the sphere. using MATLAB. (d) Find the correspondingdistance. 160).52. Water passing through the reverse check valve is dumpedinto a stream at that level. 407 DIRECT SOLUTIONS OF LINEAR DIFFERENTIAL EQS. which includes the mass of the sphere plus the "virtual mass" of the fluid that also is accelerated. whereupon the pressure rises and opens a simple check valve to a load tank with a constant high pressure PL.16 A penstock such as in Problem3. {1. 275) to form what knownas a hydraulic ram. neglecting losses. A drawingis shownbelow. Checkthe Reynold’s numberto makesure Stokes relation applies.6. . After the flow has stopped.5 (pp. Determinethe characteristic values numerically. is terminated in a reverse check valve such as given in Problem4. Water accelerates through the long tube until its velocity is so great that the valve slams shut.17 Guided Problem5.1. The equilibrium values given in step 9 of the solution maybe used. simple systemis preferred. Also. and the cycle repeats. and the time at whichv reaches 95%of the terminal velocity. find the cycle time. (a) Find the resistance with respect ot the velocity (b) Find the effective inertance. pumpingwater to that tank. the check valve closes to prevent downward flow. (c) Find the time constant. but smaller. in terms of the parameters given aboveand in Problem3. the velocity v(t) starting from rest. 329-331) finds nonlinear models for a groundeffect machine(GEM)with roll stiffness. the reverse check valve opens automatically.52 (p. 6. The device is attractive whenno electrical poweris available and a cheap.57 (p. Linearize the modelwith the actual compliancesfor roll motion.

d2x~ = -(0. xpl = 1 .95t’).256sin(0.2 -. Therefore.1 d2x dt.2 Sinusoidal Frequency Response Periodic excitations of natural and engineered systems abound.=l=j.0 = csin¢.~ -~.744. d2 x 4.95t’)]. With t’ 6. 1 2 = 10.~ -. The characteristic 3.95t’) + + ¢).= 0 -.256 si n(0.95t’).408 CHAPTER SOLUTION Guided OF Problem 1. x -.744 cos(0. therefore. ~. = c ÷ 9.95t’). Aerodynamic and water wave forces vibrate bodies. Substituting into the differential equation.95t’) therefore.95t’).95t’) 2. c -. It will be shown in Chapter 7 . At t’ -.95t’).-9.95t’) ÷ x~2 cos(0.256.c cos t’ ÷ 9.744 sin t ~ ÷ 10.95t’). xp2 cos(0.sin(0.744. ¢ = 0. OF LINEAR homogeneous solution = sin(0. is xh = csin(t’ xp Xpl si n(0.0. dt. acoustic signals comprise music and speech. Since f(t) MODELS.95t’)] -. dx At t’ = 0+. PROBLEM = sin(0. giving x -.95)2[x~1 sin(0. xp2 = 0. ANALYSIS GUIDED PART 1 6. The simplest representation of excitations such as these is sinusoidal.256 sin(0. which has roots sl. The resulting t’ > equation is s 2 ÷ 1 = 0. [-(0.-9.95) giving x = csin(t’ + c~) 10.x = 0 t’ < 0 = w~t.(0.95t’) ÷ x~ cos(0.csin(t’ ÷ ¢) . and rotational imbalance shakes machines.95) 2 ÷ 1][sin(0. X/Xol© -2O 6.9.

p. All that is neededis stability and a pinch of damping. 6. with s = ~jw: xp(t) = 1 [eJ~G(jw)e ~t -j + ~t] e-~ZG(-jw)e (6. whichmeansthat the real parts of all the roots of the characteristic equation are non-positive. systems often are characterized by howthey respond to sinusoidal inputs.6.36) where ¢ ~ gG(jw) is called the phase angle of G(jw).33) can be recast as the sum of two complexconjugate terms: u(t) = 1 [ej(~+~ ) ÷ e_jC~t+~)] ~uo = + . ~d therefore contributes a term of the form of Equation (6. The tacit assumption is made below. there is no point in studying any particular solution.25b) (p. that the systemis stable. G(-jw) = ~G(j~)le (6. 409 SINUSOIDAL FREQUENCYRESPONSE that any periodic waveformcan be decomposedinto a sum of discrete sinusoidal oscillations at multiples of the base frequency. on the other hand.35) G(jw) is called the frequency transfer function. This section is concernedwith the response of linear modelsto the sinusoidal excitation [u(t) = u0 cos(wt + ~). usually decays quickly.33) The focus is on the particular solution. therefore. The homogeneous part of a complete solution. Further.1 The Phasor Method The sinusoidal excitation of equation (6. Eachof the terms is exponential. and any waveformcan be decomposedinto an infinite sumof sinusoidal oscillations.Withoutstability. The terms G(jw) and G(-jw) are complexconjugate. 391). 397) to the particular solution.33) and (6. which wouldquickly be overwhelmedby the exponential growth of the homogeneous solution.2. whichoften is called the steady-state response since it persits as long as the input persists. "Frequencydomain" methodsform the basis of the Fourier and Laplace transform methodsdeveloped in that chapter. which has as its tangent tan¢ = Im[G(jw)] Re[G(jw)] (6. G(j~) = ~G(j~)~e~¢. J (6.12).2.34) follows from trigonometric identities (equations (6. The equivalence of equations (6.37) . so that they can be written -~¢. Physical experimentsare carried out using electrical function generators and mechanical shaker tables that sweepan excitation over a range of frequencies.

ANALYSIS OF LINEAR MODELS. is knownas the phase lag.6 Find the steady-state response of the modelwith the transfer function 50S + 200 G(S) = $3 + 3S2 + 102S+ 100 to the excitation u(t) = 2 sin(10t). and then sketching the phasor.35) becomes xp(t) = ~]G]uo[ej(~°t+~+~)+ e-J(~t+~+O] (6. ¢(w) =/G(jw) = tan -1 [Im[(G(jw)] [ Re[(G(jw)]j (6. 6. Examples of the complex number G(jw) are drawn in Fig. or 180° < ¢ < 270°.39) Thus. if both ~3[G(jw)] and ~[G(jw)] are negative. knownas the phase shift. Youare urged to find the quadrant of phasors by first determining the signs of the both terms. For example. IG(jw)[ is called the amplitude ratio or magnitude ratio gain. Be warned that manysimple calculator and computer routines automatically assumethe first or fourth quadrant.8: Determination of the quadrant of G(jw) Equation (6. PART 1 a =Im[G(]w)] b =Re[G(]w)] ~a>Oa>O[N~ b<~ >0 aR<eo Re FORe ~Ol ~e a<o~/~ first quadrant second quadrant third quadrant fourth quadrant Figure 6. x~(t) is sinusoidal with an amplitude IG(j~)I times the amplitude of the excitation. The phase angle of x(t) is advancedrelative to that of u(t) by the angle ¢(w). EXAMPLE 6.410 CHAPTER 6. Most commmonly this angle is negative.38) or xp(t) IG(jco)lUo cos(wt + ~3+ ¢).8 as vectors or phasors in the complexplane. ¢ lies in the third quadrant. It is essential that ¢ be recognized in its proper quadrant. .

the response to sinusoidal excitation at manydifferent frequencies is observed and plotted.4 0. in a clear.09 The particular solution.2. therefore.09 6 4 2 0 -2 0 6. A shaker table ofter is used to producethe sinusoidal disturbances on mechanical systems.300)+2(1020- 1000)I/(200)2+(500)2 = V ~-~-~-U¢~ = 2.2. is xp(t) = 2 x 2.2 0.tan-1 = 68. 411 SINUSOIDAL FREQUENCYRESPONSE Solution: At the specified frequency.8 1. visual way.358 sin(10t. with no use of modelsor mathematics.679 ¢ = tan-1 \200] . Plots of the magnituderatio or gain and the phase angle as functions of frequencyconveyto the analyst not only the meaningof a frequencytransfer function.679sin(lOt. 50(j10) 200 2++ IG(jl0)I = (jl0) 3 + 3(j10) 102(j10") + 100 I I 200+500j = ](100 .6. whereas the denominator is a phasor in the second quadrant.20° .174. Both linear and logarithmic axes have been used widely for the plots.2 0.106.09°) = 5. The information can be deducedfrom direct experimenton the physical system. Note that the numeratorof G(jl0) is a phasor in the first quadrant. °106.09°).0 Bode Plots Magnituderatios and phase angles expressed as functions of frequency serve to characterize a linear transfer function completely. An exampleusing logarithmic axes for the frequency and the magnituderatio and .6 t 0. but also the very character of the dynamicmodel. The phase angle of G(10j) equals the angle for the numerator minus the angle for the denominator.106.29° °= -106. Plots of u(t) and x(t) are given below.

rad/s agree with the calculated values.412 CHAPTER 6.20 log lo m] (6. You should check to see that the magnitude ratio and the phase angle as plotted at 10. abbreviated "db.6 above. and the logarithm of a square gives a multiplier of 2. widely available software. the decibel scale. as you will see later. Power is proportional to the square of the sound pressure (or the product of pressure and velocity.0 10 o). 6." is ~ " [ db -. The particular Bode coordinates have become the standard for system analysts interested in dynamics and control. The use of the multiplier 20 rather than the more natural 10 appears to be a carryover from the prior and continuing use of decibels to represent the power of an acoustic wave or an acoustic field. rad/s 100 Figure 6.9: Example Bode plot a linear axis for the phase angle is given in part (a) of Fig. Finally.9. The decibel scale has become a conventional representation of the logarithmic magnitude ratio.1 ANALYSIS OF LINEAR MODELS.employs Bode coordinates. 0. This is largely because they allow a model and its transfer function to be decomposed into components and reassembled graphically. Such a representation is known as a Bode plot. as we have seen). PART 1 1. Further. including MATLAB. the use of the logarithmic axes allows small but important amplitudes to be seen clearly despite the presence of vastly larger amplitudes at other frequencies.40) Both the m and the db scales are shown in part (a)of the figure. 1Some older references use 10 log10 m. The transfer function chosen is the same as in Example 6. . If the magnitude ratio is m.

. Solution: >> hum = [50 200]. den) directs that the calculated values of frequency. The equation 1 d2x 1 (6.39) is simplified xp(t) = G(jw) u(t).3.42) w--~ dt --~ + x = Uo cos(wt + 8). 413 SINUSOIDAL FREQUENCYRESPONSE EXAMPLE 6.41) Note that G(jw) can be negative. >> bode(num.den) Compare theresulting plotwiththatofFig. For example.9: 0 10 Frequency (rad/sec) "1 10 0 10 Frequency (rad/sec) 101 = 10 The alternate command[mag. which corresponds to a phase angle of +180 At any particular frequency.w] = bode (hum.phase.3 Models Without Damping Modelswithout dampinghave only even-ordered derivatives in equaiion (6.7 Use MATLAB to produce a Bode plot for the model of Example 6.6. the response is said to be "in-phase" or "out-ofphase" with the excitation.. .2. that is a~ = 0 and b~ = 0 for i = 1.. >> den = [1 3 102 100].200). G(jw) is real. magnitudeand phase be stored in the vectors w. magand phase.5. It also lets you specify the frequencies. wn = --~. 6. (6. respectively.to specify 200 values of frequency with even logarithmic spacing over the band from 1 to 40 rad/s. you enter w=logspace(1. °.1) (p 390).2. no phase angle is needed.. As a result.40. The second-order IC modelis a simple example.6. and equation (6.6.

44) Note that xp(t) is in-phase with u(t) for (w/wn) < 1 and (180°) out-of-phase for w/~. give a second Bodeplot for F/F~. Solution: A bond graph for this system.8 A machinecontains an unbalanced rotor whichproduces a vertical component of excitation force Fe = -ml (d2/dt2)[r cos wt].41) yields the solution Uo ~ zP = 1 .(u/un) cos(wt+ fl). and is supported by a spring with rate k in order to reduce the vertical force F transmitted to the floor.43) Equatioa (6. Also. where wnis the natural frequency.~x. dx 1 These equations can be combinedas follows: ICd2x 2dt = cd.414 CHAPTER6. .~) > 1. PART 1 gives a(S) 1 + (6. (6. The machineis constrained to movevertically. Give a Bodeplot for F normalized to (divided by) the fixed force mlrw~n. gives the state differential equations ~ dPdt mlrw coswt . p. = Cmlrw 2 coswt - x. ANALYSIS OF LINEAR MODELS. such a force promotes undesired noise and vibration. EXAMPLE 6.

the center of gravity of the masses ml and ms remains virtually fixed. The force on the floor is F -~ Xp -. this ratio approaches unity. or the natural frequency is reduced well below the excitation frequency~ this ratio rapidly decreases toward zero. As the frequency of the excitation is increased well above the resonant frequency. It shows how the amplitude of the force varies with the applied frequency.~-- 2 mlrw cos The solid line plotted below represents the nondimensional ratio of the amplitude of this force to mlrw~.1 o. . As the excitation frequency is increased well above the natural frequency. this gives 2 rw ~ xp.6. w.I 1. normalized force on floor 0. At high frequency.1 Cml .0 :~/~o~10 The dashed line represents the ratio of the amplitude of the force on the floor to the excitation force mlrw2. The design objective therefore is to make k so small that the natural frequency is considerably below the actual excitation frequency. or is shut down.(~/~n) cos wt.( 0.2. SINUSOIDAL 415 FREQUENCY RESPONSE With uo -. You may have noticed a brief vibration when you looked at an auto engine being started or stopped. The vibration can be minimized by passing the engine speed through the resonant frequency quickly. the amplitude of the force on the floor therefore is proportional to the resulting fixed amplitude of x times k.1. the resonant frequency is traversed. Whena shaft-driven machine that is operated aboye its natural or resonant frequency is started from rest. which grows rapidly with frequency. An unacceptable resonance can result.

then. This is because there is one geometrically independent position variable. such a model would be fourth order. however. Models without damping. A model with two degrees of freedom has two geometrically independent positions. is said to have one degree of freedom. produce the associated magnitude Bode plot to show what happens for other excitation frequencies. are always of even order.8. PART 1 A second-order mechanical model. EXAMPLE 6. since there would be two momentumstate variables in addition to the two position state variables. To get a particular result. but is degenerate in the sense of having only one momentumvariable. the state differential equations corresponding to the four energy storage elements are dx 1 . still has two degrees of freedom. made nondimensional by normalizing by rmlw~ as in Example 6.416 CHAPTER 6. Show that this force vanishes at one frequency of excitation. A third-order mechanical model.7. and the mass ratio 12/11 is 0. a first-order mechanical model still has a degree of freedom but also is missing an inertance. because of a neglected inertance but a considered resistance or damping. as the following example with two degrees of freedom demonstrates.2. consider only the special case in whichthe natural frequencyc~n-. ANALYSIS OF LINEAR MODELS. such as described above. In general.9 A vibration absorber comprises an added mass and spring on the system modeled in Example 6. z cos cot F~ = m~rco me (~gca Solution: From the bond graph of the idealized model with no damping. Also. Find a transfer function between the input cos wt and the output force. Similarly.1/x//x//x//x//x//x//x~ equals1/x/7"-2-~2.

0000 0 1. Therefore. This transfer function has as its input the force Fe = -rml w2 coswt. all others should be eliminated. bode(hum.0].0000 which means that $2+1 G(S) = S4 2.-1 -.C. when w/w~ = 1).6. so that regardless of the units of the individual parameters. retaining the unspecified values for the parameters. +S The MATLAB coding continues: num=[1 0 1 0 0].1. None of these are nearly as easy as using the MATLAB function ss2tf.t and frequency considered as the nondimensional w/wn. This can be done various ways. changing it to S4 2.2 0 0]. 2S~ + 1 -w 2 + 1 G(jw) = w4 _ 2. C=[1 0 0 03.2 0 c=[1000].2 0 0. which increases in magnitude with frequency. B=[0.D) The response is num= 0 0 1. The MATLAB coding to get the transfer ¯ ’ B= ¯ ’ D=0.0000 den = 1. I2 = 0. the results apply to any values of the parameters that satisfy the given restrictions.B. C1 = 1.0 0 1 -5.0 . the following specific numbers can be assigned: I1 = 1.~000 0 1. SINUSOIDAL 417 FREQUENCY RESPONSE The dependent variable of primary interest is x.0. which requires specific values of the parameters. with or without the use of operator notation. The matrices needed for ss2tf are A= 0 -1 0 0 1 -0. Because of the restrictions given. [num.den) .0000 0 ~. which can be accomplished by multiplying the numerator by S2. D:0. function from Fe to F is A=[0 0 1 0. These give wn = 1 tad/s.2. and C~ = 5. the G(S) needs to be multiplied by the ratio (w/w~) 2.2.2w2 + 1 The force therefore vanishes when w = 1 (or in the general interpretation.2 0 0. den] =ss2tf (A. the results can be interpreted with time considered as the nondimensional w. whereas the requested Bode plot is to have the constant-amplitude -rmlw~ coswt as its input. With this interpretation. to give a single fourth-order differential equation.

makingit easier to retain unspecified parameters. since G(~Wz)=O. one at 80%of Wnand the other at 125%.Therefore. to give G(S) = and G(S) = k/S. The model G(S) = 1IS is said to have a pole at the origin. The transfer function G(S) k(1 ± S/wz) is described as a zero at S =~:wz. The two resonances can be spread further apaxt in frequency by increasing the mass I~ (and decreasing the complianceC2 correspondingly). a value of k different from zero simply shifts the magnitude curve up or downand leaves the phase .418 CHAPTER6. 4O I I i I I I I i I I I I I special case for ~o~= ~ F 20 note: decibels ding segments -20 ire °ut °f Phase I -40 O.1 1. the idea of the vibration absorber has merit only whenthe excitation frequency is fixed or nearly fixed. sinc e G(0) = 0. with someannotations.) The phase angle is a constant 90° in the first case and a constant -90° in the second case.4 Models Comprising a Single Pole or Zero The transfer function G(S) = is sai d to have a z er o at theorigin. Similarly. For the square magnitudeformat used. and in the limit of zero mass the behavior (and the system) wouldbe identical to that of Example6. if the mass were reduced. The existence of two resonant freqencies is revealed. since it becomesinfinity whenS = 0.0 co/a4 I I I I I 10 The vibration absorber is analyzed in the alternative style of classical vibrations in AppendixB. would merely shift the magnitude plots up or down. 6.10. This style somewhatsimplifies the analysis for the restricted class of modelsto whichit applies. Youshould verify that the substitution S ~ jw gives these results. is given below.8. Multiplyingthe transfer functions by a constant factor k.2. As before and for all transfer functions. PART 1 The resulting magnitudeplot. The phase plots would be unchanged. ANALYSIS OF LINEAR MODELS. the slope of the magnitudecurve is +1 in the first case and -1 in the second case. Bodeplots for these cases are given in Fig. it wouldbe ±20 db/decade. the separation woulddecrease. (With the decibel format. 6.11 for the special case of k = 1. Bodeplots for this modelare given in Fig. 6. leaving their slopes unchanged.

01 °90 0. (w/wz) (6. the particular solutions of unstable systemsare of no interest.6.1 1 10 ¢o.12.46b) Note that only a plus sign is used in the transfer function. is given in Fig.This is a critical difference.45a) (6.2.45b) . The reason is that a minussign wouldrepresent an unstable system. 419 SINUSOIDAL FREQUENCYRESPONSE 20" 1£ db 0 -2O 0. l°gl° V/1 + (w/w~)2’ (6.tan-~ (-~ ./G(jw) = +tan-l (~z) Note that the magnitudeplot is the sameregardless of whether the plus or the minussign is used. and decreaseswhenit is minus. 6. . The transfer function G(S) = k/(l+S/wp) is described as a pole at S = since G(-wp)-~ ~. Theseplots follow from the substutution S = jw: logxo [G(jw)[ = log~o [1 + jw/wz[ = log~o V/1 + ~. Onthe other hand. again for k = 1. Its Bodeplot. the phase angle increases whenthe sign is plus. these plots follow from the substutution S = jw: I 1 loglo IG(jw)l = lOglo 1 + j-w/wp[ 1 _ = l°gx° ~/1 + 2(w/w~) / G.46a) (6.i(jw) = . rad/s 100 G=S OO -90° G=I/S Figure 6. rather than the + sign used in the transfer function for the zero. Again. As noted before.10: Bodeplots for a pole or a zero at the origin curve unchanged.

(+ slgn). PART 1 20 .I. (..~ linear approximation.0 10 phaseshift -" left-half plane ° 30 .11: Bodeplots for single real non-zero zeros . ° 90 ° 50 .420 CHAPTER6.C O.10 magnituderatio db " 10- 0 .sign) ° .50 _900 Figure 6. asymptotes 1. ANALYSIS OF LINEAR MODELS.

2. The high-frequency asymptote therefore is the same as a pole or zero at the origin. The low-frequency asymptote corresponds to the neglect of the term w/wz or W/Wprelative to 1.. however..12: Bodeplot for single real non-zero pole The magnitudeplots for the modelswith the single pole and the single zero are top-to-bottom flips of one another. for the common plus-sign case... and for high frequency is +90°. SINUSOIDAL FREQUENCYRESPONSE 421 asymptotes 0 . The two asymptotes intersect at the break frequency. 10 "" l~:ar approximatio~ Figure 6..6.. The high-frequency asymptotecorresponds to the exact opposite: neglecting the term 1 relative to the w/wz or w/wp.11 and 6. The asymptote for low frequency is 0°.. The magnitudeplots for the real non-zero poles and zeros can be approximatedby asymptotes. The greatest departure between the actual value and the value of the nearest asymptoteis 45°.. This fact enables you to closely approximatea break frequency if you are given the plot (e.O. a common situation for an engineer.. Asymptotesalso exist for the phase angles. -10 - -20 . The phase asymptotes represent the actual phase more poorly than the magnitude asymptotes represent the actual magnitude. 1 O.... experimental data) rather than the equation.....1wi < w < 10wi is a straight line (in the semi-log . again at the break frequency. scaled so that it equals 1 at the break frequency w = w~ or w = Wp. A muchbetter approximation for the frequency range 0.1.0 ~. corresponding to the neglect of the corresponding terms. 6. it produces a horizontal straight line.. 1. as shownin Figs. So are the phase angles.g.12 by dashed lines. where it equals the ratio v~ or 1/~/~.0 db " magn~ :\/ . The greatest departure of the actual magnitudefrom the closest asymptoteoccurs at the break frequency.

At precisely the break or natural frequency. As with the first-order poles. which can be deduced simply from equation (6.tan-’ \w. PART 1 coordinates of this plot) betweenthe two asymptotesat the two respective frequencies a decade below and a decade above the break frequency.2.)=. the straight-line approximationused for real poles applies. For low dampingratios. but observe that the drop in phase totals 180° and is relatively abrupt for small dampingratios and gradual for large dampingratios. with the total phase changebeing double the 90° for a single pole. ANALYSIS OF LINEAR MODELS. Lowand high-frequency asymptotes are drawn as dashed lines. This leaves simply a horizontal line with value 1. 6. The dampingratio ~ strongly affects the behavior only for frequencies near the break or natural frequency. Althoughcases with ~ _> 1 are included. The highfrequency asymptote neglects all terms in any sum except those proportional to the highest powerof w.47) has a pair of complex conjugate poles wheneverthe dampingratio ¢ < 1. The Bodeplot variables are I 1 log10 IG(jw)[ lo g10 1 . these cases are better computedby factoring the transfer function into a product of two real poles.13.12 by dashedlines. These linear approximationsare represented in Figs. 6. The phase curves are harder to sketch with accuracy (you will likely refer back to this plot for this purpose).~) 2 + = -l°g~° zc(j. Since the case of ~ = 1 corresponds to the product of two equal real poles.5 Models ros Comprising a Pair of Complex Poles or Ze- The modelwith the transfer function k 2G = 1 + (2~p/Wp)S + (S/wp) (6.. the low frequency asymptote corresponds to the neglect of all terms in any sum except those proportional to the lowest powerof w.422 CHAPTER 6. .48a).(w/w. This leaves a straight line with a slope of -2 which intersects the low-frequencyasymptoteat the break frequency.. Use of this fact plus the asymptotesallows you to sketch the magnitudecarve for a particular dampingratio with adequate accuracy for most purposes.. 6. the magnitudepeaks or resonates at approximatelythe break frequency. as described later. regardless of ¢. which equals the natural frequencywnregardless of ~. the magnitude ratio equals 1/2~.~] \ wn ] Plots for various values of ~ are given in Fig.11 and 6.

0.13: Bodeplots for second-order poles 1011800 . 423 SINUSOIDAL FREQ UENCY RESPONSE - ~-0.25 0.1 0.1 mag -2O ~ .2.05 ./ ~ - o0 ° -90 -4O ~ 1 +2 IS~ton+(S/a~n) o 10 Figure 6.6./ .

they contribute a term porportional to e-t sin(3t + f~) to the homogeneous solution. Zm are the roots of the numerator polynomial. the phase increases by a total of 180° rather than decreases.49) The denominator of G(S) is the characterisitic polynomial.. ANALYSIS OF LINEAR MODELS. It allows the homogeneoussolution to contain a constant (or c e°t). The length of the chord from ¯the origin to either pole is the natural frequency wn = V~d . Again.. the minus sign in the exponent is associated with the presence of the pole in the left-half of the S-plane. For this reason. 6. this relationship is the same as you saw before with first-order zeros. Imaginary and complex poles and zeros come in complex-conjugate pairs. the values z~.49) still applies. The numerator is factored similarly. Wd. and because such second-order zeros occur relatively infrequently. and the values Pl. The logarithm of the magnitude of this transfer function therefore equals precisely minus the logarithm of the transfer function above.. (S -. Sn-N of the characteristic equation. and are known as the zeros of the transfer function. The pair of poles labeled 3a and 3b in the example S-plane lie in the left-half plane. in which the abcissas is the real part and the ordinate is the imaginary part. Individual poles and zeros can be real. In this case. Should there be one or more zeros at the origin. The phase angle plots for the zeros are similarly top-to-bottom flips of the phase angle plots for the corresponding poles. and zeros by (0).. The pole labeled 1 in t~is exampleis real and negative. PART 1 A system with a pair of complex-conjugate zeros rather than poles has a transfer function that is the reciprocal of the transfer function above.2..424 CHAPTER 6. On the other hand.zl)(S z2). if any. Consequently. It contributes a term to the homogeneous solution proportional to e-2t. known as the poles of the transfer function. The pole 2 is at the origin. the magnitude plot for the complex pair of zeros is exactly the top-to-bottom flip of the magnitude plot for the complex pair of poles.Zm) G(S) -.14.. ¯. Poles are indicated b~ crosses (×).Pn-N. They often are represented in an S-plane plot. 6. Right-half-plane poles produce instability.. The imaginary parts +3 give the damped natural frequency.6 Factorization of Higher-Order Models Transfer functions of second and higher-order models are profitably follows: k(S . The integer N represents the number of additional poles with zero value (or at the origin).’". the case of the zeros permits the damping ratio to be negative without causing an instability. imaginary or complex numbers. are precisely the roots $1. as illustrated in Fig. and the real part -1 gives the exponent a = -~wn. but left-half-plane poles do not. a separate plot is not given here. the phase plot is identical to that of the pole with positive damping.SN(S _pl)(S_ P2)"" (S --pn_N)" factored as (6.. their presence in the left-half plane is associated with the minus sign in the exponent and the stability of their contribution to the unforced behavior. equation (6. but N becomes a negative integer.

den) The "tf2zp" can be read as the "transfer function to zero-pole" transformation. Should you start with the state-space format A. the command [z.k+~..¯ ..~. B.. The dampingratios ~zd+~. unlike poles.p~ are real zeros and poles. the non-zero poles and zeros are interpreted in terms of postive frequencies w~i = m k l-I(1 ± s/~:~) II [1 + ~=~+1 G(S) = ~=~k n-N (6.ss2zp(A. D. those labeled 5a and 5b are complex with conjugatewith negative real parts.50) s l-I(1 + Slo: ) H i--~l i=k+l The notation l-I~=l(1 + S/wzi) implies the product (1 ± S/w~i)(1 ± S/wz2) ¯ . Zeros..k] --.B.) .C.14: ExampleS-plane plot The zero labeled 4 is real and positive.’".p. zj and p~. ~n-Nall represent subcritical behavior.2. (1 ± S/wzj).6. as was done above for the underdamped second-order model. MATLAB will factor a transfer functipn with the numerator and denominator specified in terms of the polynomials numand den with the following command: [z. with plus signs being used for left-half plane zeros and minus signs for right-half-plane zeros. SINUSOIDAL FREQUENCYRESPONSE 425 ImS 0 5b -I × 3b Figure 6.~zm and Cp.p. Also. can exist in either half plane without affecting stability. They are then described in terms of its dampednatural frequency and dampingratio. C. For most purposes it is convenient to replace the factors for a complexconjugate pair by its real product..’’’.. they are less than 1. The parameters z~.k] = tf2zp(num.D..

but cancellation of an approximately pole-zero pair covers up an instability.0000 -1. C=£0 1 0].k] = ss2zpA.1 gives the response -4 -5.0 1 -4].2 -2 2.7 Bode Plots for Higher-OrderModels* Bode plots for models with any number of poles and zeros can be assembled readily from the Bode plots of the individual poles and zeros.426 CHAPTER 6. poles and exactly.D. The ordinate of the magnitude Bode plot equals the logarithm of the magnitude of G(jw).10: scribed by Find a factored ~ = transfer -~2 function which of for the model de- x+ with the output y = x2 (t). Solution: The MATLABcoding A=[-4 1/2 0. These but MATLAB is not that smart (yet). Cancellation of an approximately equal left-halfproduces little error.0000 -4. 6. zeros rarely cancell plane pole-zero pair equal right-half-plane the factored result 2(s+ 4) G(S) = (S 1) (S + 4)(S + example has a zero and a pole for which S = -4. From equation (6.p. [z. this becomes J loglo IG(j~)I = log~o k + ~ log~o I1 ¯ Jw/wzil i=1 + ~ log10 tl . B=EII0. ANALYSIS OF LINEAR MODELS.B.0000 2 which indicates Notice that the should be canceled.Y log Ij~l .46).03.2. or log10 IG(jw)l.C. The index ± is an integer that indicates possibly several scalar transfer functions is being requested. D=[0]. EXAMPLE 6. In practice. PART 1 produces the same result.(w/wzi) 2 + j2~iw/w~il i=j+l .

2. dz a2 -~ + al ~ -t- aox = fo cos(wt q-/~). Solution: The transfer function becomes Y0 G(S) = a3S3 q.magnitude and phase plots equals the suin of the ordinates of the componentpoles and zeros. logIG(jx)l=lOglok-lOglo 1+ ~11 -l°gl° z. The magnitudeplot also is shifted vertically by loglo k. 427 SINUSOIDAL FREQUENCYRESPONSE The ordinate of the phase Bodeplot is the angle of the phasor G(j~). 1+ .a2S2 + alS -t- ao = (1 q.S/Wl)(1 -t- kSIw2)(1 The Bodeplot variables are.~ L[1. (6. and sketch the complete Bodecurves.oTl+(~) ZG(jw)=-tan:’(~l)-tan-’ (-~2)-tan-l(-~3). (6.) .53) EXAMPLE 6..51) shows that the ordinates of both the. ~vz= 50 rad/s and w3= 200tad/s. Then. plot the asymptotic approximation for the magnitude curve and the approximation for the phase curve. -log. These summationsapply to the asymptotes and linear approximations as well as to the exact curves. Decibelssumdirectly.~ Z(1+ jwl~p.(~/~. with mathematical symbol /G(jw): L[I1. The two examples below regard the third-order model d~ x a3 -~ q- d~ x .(~/~:i)~ + 5j2¢:~1~] LG(jw) = ~ L(1 :t: jmlw:i) i=1 i=lq-j k n-N i=1 i=k+l . w2and wa.. in light of the equations abovefor individual poles.11: Find analytic expressions for the Bode plot variables in the case of three real roots with break frequencies ¢ul.)~ + j2~. for the special case ofw~= 10 rad/s. Equation (6.N 2 .~l~.52) Theseequations expedite the determination and interpretation of Bodeplots.].6.

. -90° for frequencies greater than ten times their break frequencies..N. Between the first and the second break frequency only the asymptote for the first pole is sloped. the slope of the asymptotic approximation to the magnitude curve at any frequency equals -1 times the number of poles that have lower break frequencies.000 ~ ~ = (S+ 10)(S+~0)(S+200) o(~) °lO . their sum has slope -2.magnitude OF LINEAR MODELS.. ANALYSIS -20 ~. Above the third break frequency all three poles have asymptotes with slope -1. The sum of the asymptotes for the magnitudes of the individual poles is shown by a dashed line. ~ lO ~ 10 ~0~70° The Bode plots above are for the given values of ~.... The linear approximations to the three individual phase curves also are shown... so their sum has slope -3.. A generalization can be made regarding the magnitude asymptotes: in the absence of zeros...... PART 1 O° \. ~ and ~ and ~ = fo/ao = 1..._.... Recall that these approximations equal 0° for frequencies less than one-tenth their break frequencies.. ..~.. and therefore the sumof the three has the slope of -1. The sum of these three approximations can be seen to approximate the actual phase shift rather nicely. and a straight-line interpolation in between...~ linear iappr°ximati°ns : db -40 ~3 zGO’~o) °-180 100. Below the first break frequency all the asymptotes equal 1.. For frequencies between the second and third breaks both the asymptotes for the first two breaks have slope -1.428 CHAPTER 6. including poles at the origin..

0.065. set wl = 10 tad/s. plot the asymoptotic approximation for the magnitude curve and the approximation for the phase curve. SINUSOIDAL 429 FREQ UENCY RESPONSE EXAMPLE 6.2.log~o 1 + ~-~ log~0 ZG(jw) 1.6. Then.2 = 100 rad/s and ( = 0. and sketch the complete Bode curves.(/~2)S (S /~2)~]" + The Bode variablesbecome log~o Ig(jw)l = loglo k .12 Find analytic expressions for the Bode plot variables for the case of one real root at Wl and a pair of complex roots with damping ratio ( at w2. Solution: The transfer function can be factored as k G(S) = (I+ S/~I)[I (~ .(~-~ + \ w-’-~/ =-tan -1 ~ -tan -~ 1- (w/wz) ~] " The case with the given parameters and k = I is plotted below: "~agnitude ~\ -20 db -4O ° -180 -60 -8O °10 101 102 ~’270° 10 .

The model resonates about this natural frequency w2. The resonance nowappears at 10 rad/s. or -270°.90 o .13 Repeat Example6. ANALYSIS OF LINEAR MODELS. PART 1 The magnitude asymptote for the model has the slope -1 for frequencies betweenwl and w2. The resulting plots are given below.13. and the slope -3 for frequencies above w2. The total phase shift at infinite frequency is the total numberof poles times -90°.430 CHAPTER6. located at w = w2. because the dampingis fairly small. the presence of the first-order pole in this case might be neglected for practical purposes (as also noted below). about one-half percent error results from the effect of the break frequencywl. Solution The procedure is the same as in the previous example.11 and 6. The phase drops precipitously 180° in the vicinity of w = w2. In general. and the first-order pole appears at 100 rad/s. Note that the same result applies to the other third-order models in Examples 6. or two poles. The linear approximation to the phase curve can be seen to apply reasonably well for w < w2. The value of the magnitudeat the break frequency is almost exactly a factor of 1/2~ = 7. the lowest-order singularies tend to be the most important. EXAMPLE 6.69 greater than the intersection of the asymptotes at that frequency. o0 ° . In the present case. the drop in slope of -2 results from the double break.180 .12 for the case in which wl and w2are switched.

In case (b). (b) and (c). which can be seen in both the impulse and step responses. The time constant in this case is so relatively short. and decay with the damping ratio ~ = 0.2 rad/s 15. which is the same as a second-order pole with critical damping. that it effects the responses imperceptibly.01 seconds. and a natural frequency is the break frequency for a complex conjugate pair of poles. In case (a) (Example 6. . this model could be simplified by omitting the real pole (as also noted above).14 Sketch-plot the Bode curves for the fifth-order model 200.2. Solution: This case includes a pair of identical first-order poles.12) oscillations occur at the damped natural frequency w2v/i -.065.1. case (c) (Example 6. (Example 6.63 Hz.15. and decay according to the damping ratio ( = 0. 6. only r = l/w2 = 0.11). Step and impulse responses for the cases of Examples6. There is also a firstorder pole at low frequency and a second-order pole at high frequency. SINUSOIDAL 431 FREQUENCY RESPONSE A time constant in a homogeneous solution or an impulse or step response is the reciprocal of the break frequency for a real pole.(2 = 98. The double break at w = 10 controls the rise in the first one-half second. EXAMPLE 6.13). therefore.1 second. the dominant time constant is rl = l/w1 = 0. The resonance at about w = 45 produces the small rapid oscillation superimposed on an otherwise smooth response.000 a(s) = (s + 1)(s + lo)2(s~2000) Also.1 second is similar to that of case (a).11. the oscillations occur at the lower resonant resonant frequency of virtually 10 red/s. For manypurposes.12 and 613 are given in Fig. These plots can be secured readily through the MATLABcommands step and impulse. but the pole at the lowest frequency produces the dominant relatively slow exponential delay. The impulse response shows the effects of all these poles. 6. the effect of the time constant r = 0. with the respective designations (a).6. use MATLAB to plot the impulse response of the model. In addition.

12 I I I I 0.~. 6.4 t 0.3 0.13. 0’1 ~] ~.3"~ ~.432 CHAPTER6. PART 1 15 0 1. .11 for Example6.5 1C -5 2 for Example6.4 0 for E: ~-nple6.11.15: Impulse (g(t)) and step (h(t)) responses for the modelsof Examples 6.13 I 00 1 I I 2 t 3 Figure 6.2 0. ANALYSIS OF LINEAR MODELS.1 0.0. and 6.5 I t 0.12.2~ ~J 0.

SINUSOIDAL 433 FREQUENCY RESPONSE. den = [1 26 2225 42700 240500 200000].4 0. bode(hum.8 0.000 (s+ 1)(s+ ~o)~(s%ss+ 10° lift lOt co 10~ MATLAB coding that gives the complete Bode plot decibels) ~s well as the impulse response is as follows: (although hum = [200000] .2.den) 0. -90o ’1 phase ~ -6o ~)= G:S’ ZG(jeo) °-270 200.2 0 0 1 2 t 3 4 with .6.den) impulse(num.6 0.

The magnitude asymptotes and phase approximations also given in the figure themselvesare sufficient to determinethe transfer functions.15 The Bodeplots below correspondrespectively to the following transfer functions with zeros: 0.01(S2 4.2S ÷ 1’ Gb(S) S24. EXAMPLE 6. . the right-half-plane zero contributes an increasing phase as the frequency is increased.that is if the zero is movedan equal distance into the opposite half-plane.100) 1 -4. you can use the phase plot to determinethe signs of the real parts of zeros. from -2 to -1 in the first-order case and from -2 to 0 in the second-order case. Therefore. Modelswith no zeros in the right=half plane are called minimumphase-lag models. both d= signs should be +.0.0.0. On the other hand.2 S ÷ 1 20[ 10: G(/~.$2 4. Solution: In both cases the behavior in the vicinity of w = 1 is caused by the identical poles (in the denominator).2S 4.434 CHAPTER6. Both zeros have break frequencies of w -. A transfer function with a zero produces a magnitudeBode plot that is unchangedif the sign of the real part of the zero is changed. I dbOI 10 o0 -20 I -401 10" 10-1 ° 1~ 101 ~ 10~ -I 10 ° 10 ~180 101 ° (~ ~0 Resolve whether the 4. It is possible to deduce the transfer function of a knownminimum-phase-lagmodel from its magnitudeplot alone. Therefore. whereasthe left-half-plane zero contributes a decreasingphase (or increasing phase lag) as the frequency is increased. A modelwith one or more right-half-plane zeros is called non-minimum phase-lag.1S Ga(S) . PART 1 Zeros always contribute an increasing value to a magnitudeplot as frequency is increased. In both cases the breaks cause the slopes of the asymptotes to increase.10.signs should be plus or minus. ANALYSIS OF LINEAR MODELS.

mptotes I -90’ appr°ximati°n’~ G~(S)~. 6..typetwo and type-minus-one models (with the factor S in the numerator) are not uncommon. The presence and number of poles or zeros at the origin is readily identifiable by the slope of the magnitude curve as the frequency approaches zero. . ... and shining the magnitude result vergicNly to account for the coe~cient ~. These are su~cient to give the complete transfer functions. lO’t I ~ ~~’~. where N is the net number of poles at the origin.16...I 10"~ .2...~ . The models presented previously have been type zero.... The generic designation is type N. and that the model with the second Bode plot has a single zero at the origin. which is -N or -20N db/decade. corroborating the conclusions.. ~ti~~ ~ magNm~ ~ ~]~ -180_ q 10-~ ~ ~ -~ ~’" "~ phase -270° 10 ~ 10 10 ~ ~ ~ 10 104 . directly leading to the indicaged conclusions. ~rther. I0° I0~ I0~ ~ N~lu~m~he slope of ~he magnitude curve ~ ~ ~ 0 is -1 in the first c~e and +1 in ghe second. The magnitude ~ymptotes and the ph~e approximations are also shown above. it is -N x 90 EXAMPLE 6. Type-one models (with the factor S in the denominator) are also fairly common. but others are rare. phase angle at zero frequency is directly affected also. The ph~e of a second-order pole or ~ero that h~ low d~ping . A summ~y of the Bode ~ymptotes and line~ ph~e approximations for first-order poles and zeros is given in ~ig. Note that the °. the phase angle for ~ ~ 0 is -90~ in the first case and 90~ in the second...6. SINUSOIDAL 435 FREQUENCY RESPONSE Models often are classified according to their numberof poles or zeros at the origin.16 Verify that the model with the first Bode plot below has a single pole at the origin..a~. Recall that an approximation any higher-order model results from appropriately summing these ~ymptoges ~d approximations.

ANALYSIS OF LINEAR MODELS. 6.16. o magnitudeasymptote: 1 phase approximation: ’~g oi _90 o. and note the discrepancies between the approximation and the actual curves using Figs.436 CHAPTER 6. This process is an example of what is called the experimental identification of a dynamic model.11.12 and 6. You would also wish at least to approximate the resonance or anti-resonance in the magnitudeof such a case.16: Summaryof Bodeasymptotes and linear phase approximations is a partial exception.lcoo olOco ° lOO ~ _~’1 I 90 1+s/69 o lO l 1~ I’S/~v ~ 1 ] [ 10 O. since knowntransfer functions can be plotted readily by software programsincluding MATLAB. PART 1 10 key: a. drawthe associated asymptotes and the phase approximations. called the pole-zero excess. It is possible to estimate a transfer function from a Bodeplot drawnusing experimental data.1 ]1 ~ ~ ~ ’ i I lOco o ~ 90° ~ ~ 0 j °90 ~ °0 °_90 Figure 6.Rather than spread its 180° phase change over two decades of frequency. it is better to concentrate the change in an abrupt jump at the break frequency.lcoo co o ! i/~ 1 . The slope of the magnitude asymptoteat infinite frequencyfor any transfer function equals -1 times the numberof poles minus the numberof zeros. by noting the factor of 1/2~ at the break frequency. Youguess the type and locations of the break frequencies. . 6.15 and 6.J ~ 90° ~ °0 1 10 I+S/~o 1 ° -90 o. however. which is not included in the asymptotes.13. An understanding of the asymptotes and the discrepancies mayhave its greatest Utility in the interpretation of experimental data. You then makeadjustmentsuntil a satisfactory fit results. as suggested in Examples6.1 O.

437 SINUSOIDAL FREQUENCYRESPONSE EXAMPLE 6. N = 0. This meansyou have a type-zero model. magnitude ratio 1...6... The magnitudestarts to rise noticeably for frequencies lower than one-quarter of this.... 423).. observe that you also have a horizontal asymptote at high frequency. Puting this information together.. Youalso knowthat there has to be at least two zeros in the model(because of the pole-zero excess).0 0"10.... Next. Finally.. 6. whichshowsthe telltale rapid drop of well over 90° in the vicinity of 2 rad/s. which doesn’t happen for sharp resonances.2.. as can be verified by examination of Fig.. rad/s 100 °90 phase 0o °_90 °-180 Solution: First. The most prominent feature of the magnitude plot is an apparent resonance at about 2 rad/s..... 10 .1 1.. the pole-zero excess is zero.. Youconfirm the presence of a double pole by examinationof the phase plot..13 (p... however.01 0. you annotate the magnitude plot as follows: .0 10 ~o. and thereby write the associated differential equation. or the numberof zeros equals the numberof poles... examine the magnitude ratio and observe that the lowfrequency asymptote is horizontal. Approximatethe associated transfer function.17 The Bode plot below might represent experimental data. which only can happen because of a zero there...... observe that the phase rises at low frequencies..

. . [.... and to take advantage of the fact that at their respective breaks the deviation of the actual curves from the intersection of the first-order zeros is v~. rad/s The break frequency for the zero is merely guessed..?...~ .. which is precisely what the second-order pole demands.r..438 CHAPTER 6.. as also shown by a dashed line above.. Further.r. The result is: magnitude ratio 1. as shown. you try an asymptote with slope -1 before this pole..01 °90 phase 0o -90 ° °-180 10 1.... in any case it is followed by an asymptote with slope +1.... 0.~.. (~.. or about 3 db. You now make adjustments to the asymptotes to make their intersection more precisely align with the resonant frequency. plus something more for being near the second-order pole...e. for otherwise the magnitude curve would not bend from a sharply downward slope toward zero slope..ze....0 a). You note with pleasure that the two sloped asymptotes have a difference in slopes of -2... ANALYSIS OF LINEAR MODELS... 0’ 10..1 1:0 10 100 to...01 ...... rad/s 100 zero’in’e’ft’-l~i ~ ’plan~~:::~’..zeroin right-halfplane second-order po~le.0 O. Therefore... the intersection of these asymptotes is close to the peak of the resonance at about 2 rad/s. PART 1 magnitude ratio 1. The break for the second zero must occur at some frequency above the resonance.0 ~ second-order polefirst-prd.

Transport and wave-like phenomenaare among those often approximated by this model. imagine a sine waveof frequency w which also is delayed by T seconds. because the pole-zero-excess is zero. and the frequencytransfer function is G(jw) = -jTw. so that ~ = 0. The phaseincreases in the vicinity of the lowerzero.17 is one of manythat would arrive at the same answer. 6. Thusthe phase shift. Not e Msothat your transfer function gives G(joo) 0. The more general pure-delay operator becomes G(s) = e -Ts. The differertces betweenthe actual curves and this approximationare consistent with the plots of Figs.13 (pp. whichdictates the right-half plane. The information is redundant.55) . If the frequency is one cycle in T seconds. etc. SINUSOIDAL FREQUENCYRESPONSE. To find its transfer function. 420. (6.54) This is representable by a phasor of unit length and angle Tw. since the pure delay neither amplifies nor distorts the signal.6. The sum of these. (6. Onthe other hand. 423). you should be able to identify the wholepicture. is shownby dashed lines. The ratio of the values is 5:1. Thepresence of one zero in the right-half plane is confirmedby the fact that the overall phase change is -180°. as with a picture puzzle having several pieces.1.2. The damping ratio can be estimated by comparing the value of the magnitudeat the resonanceto the intersection of the two asymptotesthere. whichequals 1/2¢. If for examplethe frequency is halved. 6.2. the phase shift is -360° or -2~r radians. 439 Your value of k = 0. as plotted. If you understand the individual effects. equals Twradians. whichin turn could be used to solve for the response of the systemto any other input: du d2u 1 8 d2x + 0. the phase shift is halved. plus the 180° phase drop approximatingthe pole.2~-~ + 4x -~ dt = -0"2~ 5 + " dt + 2u" The procedure used in Example6. Were this a miniumum-phase-lagsystem.5 is determined so that G(jO) 0. and the impulse response is a impulse delayed by T seconds. 5. for which the step response is a step delayed by T seconds. the final phase would be zero.8 The Pure Delay Operator* Imagine a pure-delay model. so it mustbe in the left-half plane.2. There is one ambiguity left to be resolved: the sign of the zeros. You should not attempt to discover one procedurefor all cases.11 and 6. a confirmation. The resulting transfer function can be interpreted to give the following differential equation. in general. The phase approximations associated with the zeros are shownabove by dotted lines. or ~ = 2~r/T rad/s or Tw= 2~r tad. the phase change due to the higher-frequencyzero is negative.

(6. often is chosen for the magnitude. There are an infinity of such zeros.2IT’ [Gl(jw)[ = /Gl(jw) =-2 tan-X (-~-~). so their frequency transfer functions are real and their poles and zeros come in plus-and-minus imaginary 2The Control SystemToolboxof the professional version of MATLAB includes a command l~ade whichgenerates Pade approximantsof any order. That is.60Tjw + 120 (Tjw) 3 + 12(Tjw) 2 + 60Tjw ÷ 120 ’ (6. it must have zeros in the right-half plane. PART 1 so that x(t) = e-Tsu(t) = u(t -.17. 6. The transfer function G(S). called decibels (db).57a) (6. each matched by a pole of the same magnitude in the left-half plane.575) (6. Plots of the logarithm of the magnitude ratio IG(j~)I and of the phase shift versus the logarithm of the frequency are called Bode plots.440 CHAPTER 6. The Pade approximant with third-order numerator and denominator is G3(jw) -( Tjw)3 + 12(Tjw)2 .56) Since the magnitude of the pure delay operator is 1 while the phase lag is non-zero. ANALYSIS OF LINEAR MODELS. it must be a non-minimum-phase-lag operator. so the response of engineering systems to these inputs is a major concern.2. Models without damping have only even powers of S. gives a complex number G(jw) called a frequency transfer function. Its magnitude equals the ratio of the amplitude of the sinusoidal response x(t) to the sinusoidal input u(t) at the frequency w. A crude approximation for a pure delay phasor is offered by the combination of a single right-half-plane zero in combination with a single left-half-plane pole of equal magnitude: 2IT Gl(jw) _jwjw +. (6. when evaluated with S = jw.58) which also has unity magnitude for all frequencies. .57c) This is the crudest of a series of approximations to the pure delay in the form of ratios of polynomials known as the Pade approximants (pronounced Pahday). The phase angles are reasonably valid only for a restricted band of frequencies which increases with the order of the approximant2.9 Summary Sinusoidal excitations abound in the physical and engineering world. identical to the exact phasor e-jTw.T). 6. as indicated in Fig. The quantity 20 loglo IG(jw)l. Its phase angle equals the phase shift between the sinusoidal response and the input. Transfer functions for systems described by ordinary linear differential equations are ratios of polynomials which can be factored to reveal poles and zeros. in fact. including by the bode routine of MATLAB.

Real poles occur in all first. The opposite changes occur for right-half-plane poles and left-half-plane zeros. the transfer function corresponding to a given Bodeplot can be estimated by approximating these asymptotes and phase approximations. models . Conversely. T~o I 10 12 Figure 6. for a zero it breaks upward by ÷1. while complexzeros cause notches. As the frequency rises past a pole the slope of the magnitude asymptote breaks downwardby an added -1. then. Each pair is describable in terms of its natural frequency and dampingratio. and contribute exponential behavior that is either dampedor explosively unstable. Fourth-order models. it is convenientto leave their polynomials in unfactored quadratic form.6. as does right-half-plane zeros. Models with damping include real poles and/or complex poles. Bodeplots can be sketched quickly by first drawingtheir straight-line asymptotes and phase approximations. and is associated with a degree-offreedom. Complex poles and zeros comein complexconjugate pairs. This applies even to poles and zeros at the origin.2. Each left-half-plane pole reduces the phase asymptote by 90°. 441 SINUSOIDAL FREQ UENCY RESPONSE o0 phase °-180 ° -360 °-540 0 I I I I 2 4 6 8 ~. As a consequence.17: Phase lags of a pure delay and Pade approximants pairs. The numberof resonant frequencies that the poles represent equals the numberof degrees-of-freedomand one-half the order. third and other odd-orderedmodels. although only zeros can be in the right-half-plane without causing instability. two natural frequencies and two dampingratios. Complex poles cause resonancesin the vicinity of their second-order break frequencies. have two degrees of freedom. although if a dampingratio exceeds unity it is preferable to factor the associated second-orderpolynomialinto two first-order polynomials.

Determinea transfer function betweenthe input variable. Note that its input is the time derivativeof q(t). The objective is to comparethe amplitude of the steady-state displacement of the block. q. 4.2 with only left-half-plane singularities are knownas minimum-phase.18.:~Convert this to a frequencytransfer function. 6. 7o Presumingthe mass is given.442 CHAPTER6. q = q0 sin Suggested Steps: 1.18: System for Guided Problem6. that would(i) keep the amplitudeof ql to within ten percent of that of q.2 This is a classical elementary probleminvolving the undampedvibration of the single-degree-of-freedomsystem shownin Fig. 3. ql. observe the approximate range of spring constants. Model the system With a bond graph. Write the state-variable differential equations for your model. Either sketch the Bodeplot correspondingto the transfer function. 6. or use MATLAB. ql. 2. Guided Problem 6. PART 1 q=ql sin ~ Figure 6.and vice-versa. The pure delay function includes an infinity of right-half-plane zeros. . Its approximationshave a finite numberof such zeros. as a function of the momentum. (ii) wouldkeep the amplitude of q~ to less than ten percent of that of q. the phase angles follow directly from the magnitudes. and q~. 5. k. Express the output variable of interest. Combinethe differential equations to give a second-orderequation in terms of a momentum. Modelswith right-half-plane zeros are non-minimum phase. with the amplitude of the displacement of the bottom end of the springs. ANALYSIS OF LINEAR MODELS.

The mechanical system shownin part (a) of Fig. and plot its magnitude and phase in Bodecoordinates. which equals mlle.) 2. 4. 7. Find a bondgraph modelfor the system. Youshould be left with two equations in terms of the two momenta.2. Plotting is particularly simple if MATLAB is used.1 (p. both the use of MATLAB and detailed manual steps are indicated below. 347) to get a single fourth-order equation in terms of one of the variables. C. and substitute this into the other equation to get the first momentum as a function of the input excitation. Steps 5-8 below complete this part of the problem. If you wish to use MATLAB for finding the scalar transfer function of interest. use the MATLAB function ss2tf to find the answer.d/dt. 203. Solve one of the equations to get the second momentum as a function of the first (and the operator S).1. a different approachis suggested. SINUSOIDAL FREQUENCYRESPONSE Guided Problem 443 6. D of the state-variable formulation. kl = 25 N/m. Find the frequency transfer function for the displacementof the massml relative to the exciting force.25 N s/m. Then. p. k2 = 100 N/m. Multiply the two most complexalgebraic equations by S. and write the four first-order differential equations in the standard fashion. If you wish to practice manualdetermination of the scalar transfer function. B. Since the matrix operations required are awkwardwithout the use of appropriate software. and substitute the other two equations to eliminate two of the four variables. Substitute jw for S. unless matrix software is used. (This is done in the solution to GuidedProblem4. Combining the four first-order differential equations into a single fourth-order differential equation is the mostdifficult part. start by convertingthe differential equations to algebraic equations. your interest is in the momentum of the first mass. ms = 2 kg. 8. 198) is excited with the force F = F0 sin wt. find the elements A. R = 0. Suggested Steps: 1. The parameters are as follows: rnl = 1 kg. .3 This problemregards the use of the phasor methodto find a frequencyresponse. Write pl/F as a ratio of polynomialsin S. Applycausal strokes. At this point it is possible to apply equation(5.6. The result is the frequencytransfer function G(jw). Convert this to the ratio xl/F by noting that Sx~ = p~/m~. using the operator S =.47) (p.

420-423). is usually considereda very difficult problem. rad/s 100 400 Figure 6. Incorporate this improvementinto your sketch whereverit applies.13 (pp. Estimate the asymptotes of the magnitudeplot.12 and 6. 436).16 (p. (a) sketch-plot an approximation for the corresponding Bode phase plot.39) (p. noting that their slopes must be whole integers. A Bodemagnitudeplo~ for a modelis given in Fig.1 0. 6.12 and 6. if done analytically in an introductory course in vibrations.11. 6. PART 1 0.001 ]0 ~o. 2. 6.4 This is a basic problemin the identification of a linear modelfromits frequency response. or use the bode command of MATLAB as illustrated above. and (b) estimate the correspondingfrequency transfer function. Identify the break frequencies from your magnitudeasymptotes. you can apply equations (6. 3.444 CHAPTER6. You have nowcompleted what. 410). Note the sloped straight-line approximationfor the phaselag of first-order poles and zeros given in Figs.4 9. Guided Problem 6.19: Bode plot for Guided Problem6. . 6. The deviations of the curves from the asymptotes can be appreciated with the help of Figs. 6.19. Assumingthe model is minimumphase.11. ANALYSIS OF LINEAR MODELS. To compute and plot the frequency response. Suggested Steps: 1. and start the phase diagramby sketching its corresponding horizontal asymptotes.

20: Bode plot for GuidedProblem6. Three stable models have the same Bodemagnitudeplot given in Fig. Identify the dampingcoefficient from the degree of inverse peaking.01 10 ~ rad/s 100 r’-q~6360° Figure 6.20. Can a non-minimum phase lag exist? Comparewith the given plots for the three cases.2. Suggested Steps: 1. Note that a secondorder zero has precisely the same form in Bodecoordinates as the corresponding second-order pole except for an opposite sign. Estimate the corresponding transfer functions. but different phase plots as shown. 6.6. and drawconclusions. 2.1 ° -180 ° -270 0.5 This problemillustrates the effect of right-half-plane zeros on the frequency response of linear models. Predict the phase angles using the results of step 1. 445 SINUSOIDAL FREQUENCYRESPONSE. Recognize any ambiguitywhich nevertheless results in a stable system. Completeyour phase estimate with a smooth curve. 5. Write the transfer function or functions. . Estimate the asymptotesof the given magnitudeplot.5 4. Guided Problem 6. assumingall poles and zeros affect the plot. ° 90 ° _90 0. Usethe break frequenciesyou have already identified to estimate the transfer function. which the existance of the plot assumes.

446 CHAPTER 6. through four springs as pictured below (bearings not shown). 1 ft. the magnituderatio and phase angle of the steady-state response of the modelgiven in Fig. 6. 6.lb.20 A shaft with angular velocity ~ and applied momentMdrives the central memberof a torsional vibration absorber that has a momentof inertia Id = O. Checkto see that they agree with the plots. directly from the given transfer function. (c) Find a single differential equation relating $ to M. Estimate the actual case by focusing on the differences betweenthe phaselag of the remaining unknownsystem and the phase lags of the two knownsystems.19 Showfor what ratios of parameters the torsional system shown below corresponds to the tuned vibration absorber shownof Example6. 6.s. and the length a is 0. (a) Modelthe system with a bond graph. This memberin turn drives. Eachspring has stiffness k -. 412) for sinusoidal excitations at 1 and 100 rad/s. (d) Find the steady-state amplitude of ~ in response to.9 (p.500lb/ft. (b) Evaluate the parameters of your model.an Mthat~ oscillatory at frequencyw. Additional phase lag with no effect on the magnitude can result from symmetricplacementof one or more pairs comprisinga zero in the righthalf-plane and a pole in the left-half-plane. Sketch-plot this amplitudeas a function of w. Does this system act as an ideal vibration absorber at somefrequency? . as in the Pade approximants for a pure delay.0 ft.18 Evaluate. ANALYSIS OF LINEAR MODELS. 416).a flywheel of momentof inertia I = 1. PI~OBLEMS 6. A pure delay itself is included as a possiblility.9 (p.25 ft. PART 1 3.lb.

The use of similar steps is suggested. and compareto the corresponding modelwithout the tube. The objective is to choose springs and the dashpot so that the motion of q produces little motion of ql.1 ma peak-to-peak is drawnfrom the transducer for a signal of 4 volts peak-to-peak at the 5 Hz. 30. 39. The signal is received by a recording and display instrument that requires negligible signal current. as are the standard resistors of 10.) Find and sketch the Bode magnitude asymptotes for the model. that reduces the unwantedsignal by a factor of at least 10 without reducing any of the wantedsignal by more than 40%.) . 6. 68. its input impedance is very high. (Your may use MATLAB’s ss2tf. 33.23 An instrument transducer produces a signal that contains an unwanted periodic noise signal at 60 Hz. 91 kf~ and all factors of 10 lower and higher.22 Find the steady-state force F(t) for the system shown below with x = x0sin 6. 75. 27. The part of the signal of interest lies below5 Hz. 56. 82. 202) and Problem5. Briefly discuss the anticipated consequencesthe tube has on hearing. 62. 10 and 20 mf are available. if it exists.30 (p. 51.e. 47. using whatever knowledgeyou may have about hearing. 442).7 (p. which ignores dissipation.6. 12.Find the transfer function from the input Pin to the output Pe/Ie. speculate on the waythe inner ear compensatesfor the natural response. Also. 6. 22. 43.2. SINUSOIDAL FREQUENCYRESPONSE 447 6.2 (p.while ¯ keeping the current small. as shownbelow.24 Consider the classical case of sinusoidal position excitation of a massspring-dashpot system shown on the next page. 334) in which effect of a drain tube placed through an eardrumis modeled. The instrument is inaccurate if more than 0. 5. Standard electrolytic capacitors of 2. except that a dashpot has been introduced.21 Return to Problem4. (Note that this problemis similar to GuidedProblem6. A simple potential solution employsan RCfilter. 15. R Find a design. i.

the Bodeplot for the particular case which I = R2C/4 (The frequency Wn= 1/x/7-~ can be used as normalizing frequencyto give a dimensionlessfrequencyratio. Also. Suggestion: Consider the nature of the Bode plots for large and small amountsof damping. (c) Combinethe differential equations to get a second-order differential equation with a momentum as the dependent variable.) . which comprises a chamberwith fluid compliance. ANALYSIS OF LINEAR MODELS. (e) Sketch.b. I. as shownbelow.448 CHAPTER 6. q(t). duct Q~ ~ -~ Helrnholtz resonator az ~-~ (a) Modelthe system with a bond graph. Write the correspondingfrequencytransfer function. (c) Determinewhether b should be large or small if the frequency is (i) muchsmaller than the natural frequency of the system. In order to reduce the propagation of acoustic pulsations in the flow. and the output. an exhaust pipe from an engine is an example. (Hint: Two junctions needed. C. Q1.25 A duct leads from a acoustic source to a resistive load. PART 1 q =qosln ~ (a) Model the system with a bond graph. Also. relate the output flow Q2 to this momentum and Q~. in order to get a transfer function between Q1 and Q~. (d) Relate this momentum to the input flow. (b) Writedifferential equations. 6.) (b) Determinea transfer function betweenthe input. describe a design strategy k as well as b can be chosen. attached to the duct by a narrow entrance with fluid inertance. a Helmholtzresonator is attached. or plot using MATLAB. (ii) about the sameas wn(iii) muchhigher than wn. wn. ql (t).

following the steps given for the first of these problems.5). 6.6.5 seconds." Yourtask is to determinethe stability of a system comprising a mass connected to ground through such a "spring. 6. it can be described as a "spring with a lag.9 (p. 412) is excited by the samedisturbance as the two previous problems. or is in phase with the displacement. (b) Find the particular solution xp(t) valid for 0 < t < 2.5sec (a) Find the homogeneous solution xh(t).4sin |-’" t| \5] =0 0 < t < 2. 6. and use these as initial conditions for solution for t >_ 2.5) and :~(2. following the steps given for the second of these problems.27 Find an approximate numerical solution to the previous problem using the :[s±m commandof MATLAB. Evaluate the undeterminedcoefficients using the knowninitial conditions x(0) -. (Use of the phasor methodis suggested. (a) Find the response analytically. 6. but are not mandatory. The factor 1/j = -j describes the fact that the force lags the velocity by 90°. 6.26 A system with the transfer function I) G(S). (b) Find the response numerically.5 seconds.28 The systemrepresented in Fig.or lags the velocity by a little morethan 90°. SINUSOIDAL FREQUENCYRESPONSE 449 (f) Use the aboveresults to suggest the design of a muffler that would effective over a range of frequencies. .29 The impedanceof a spring is 1/CS in terms of the operator S.2. and 1/Cjw in terms of the frequency w.0. 5~(0) (d) Find x(2.$2lO0(S +2S+10 0 is excited by the disturbance u(t) -.) (c) SumXh(t) and xp(t) to get the completesolution valid for 0 < t < 2." and what effect is producedby adding a dashpot. Sometimes a servo system is designed to imitate a spring.5 seconds.5sec t _> 2. but produces a force that lags the displacementsomewhat. The following steps are intended to help.

can invent a symbol. (c) Transformthe result of part (b) into a function of (d) Write the differential equation for the system in operational form. Express the answersin decibels (db).s/ft 3. (e) Use the result of part (d) to find the characteristic values that determinethe stability. (b) Evaluate the ratio for (i) 100 Hz (deep bass) (ii) 2000Hz (soprano). Neglect the effect of motion of the plaster on the incident soundpressure. Pt/Pi. (g) Repeat part (f).8.450 CHAPTER6." (b) Characterize the "spring" by its impedanceas a function of frequency. instead placing the dashpot between the mass the spring. and its material dampingis virtually nil. choose some reasonable assumption to represent the small added phase lag. Repeat step (d) and (e). The acoustic impedancefor plane wavesentering a large space such as a roomis a resistance of magnitude2. PART 1 (a) Represent the system (without the dashpot) with a bond graph. whichare defined as 20 loglo(Pt/Pi ).30 You are asked to determine the transmission of sound through a wall comprisinga single sheet of plaster 3/8 inch thick. whichrepresents a ratio of pressure changeto velocity change.67 lb. assuming mechanicalconnection betweenthe two layers. using plaster sheets and steel supports. to represent the "spring. (f) Modifyyour modelto represent the addition of a dashpot betweenthe mass and ground. Thespecific gravity of plaster is 1. . Defendyour design in terms of strength and other practical considerations as well as its acoustic properties. Do your results suggest the acoustic significance of the mechanicalconnections conventionally used to strengthen the wall? (d) Design problem. using the result of part (c). (Note that humanscan detect loud sounds despite attenuations of 50 db and occasionally even 100 db. 6. (a) Modelthe wall so as to permit estimation of the ratio of the transmitted to incident sound pressure.) Whatwouldrock music sound like on the quiet side? (c) A double wall is constructed with the same plaster sheets and a inch air space in between. whichis very thin compared with the acoustic wavelengthsof interest. Repeat pa~ts (a) and (b) above. and neglect the effect of any mechanicalsupports for the plaster. Designa practical non-load-bearinginterior wall for sound isolation.such as Z. ANALYSIS OF LINEAR MODELS.

451 6. SINUSOIDAL FREQ UE‘NCY RESPONSE. 435).32 Sketch Bode plots for the models described by the following minimumphase and nonminimum-phase transfer functions. Find and plot their responses to a unit step input.2.16 (p. 6.0 10 frequency.1 1. 1 G~(S)= ~(I+ TS).33 Find the transfer functions for the two systems characterized with Bode plots in Example6.6.31 Sketch the Bodediagramfor the modeldescribed by the following transfer function: S2 + 20S G(s) = $3 4S2 + 104S + 200 6. Note that magnitudeasymptotes and straightline phase approximations have been drawn already. rad/s 100 .34 Estimate the transfer function for the system that produces the Bodeplot below. G2(S) = ~(1-TS) 6. 100 magnitude ratio 10 0.

........ o-250 .. PART 1 6. °: _-100 -2000 : ~"~ . ANALYSIS OF LINEAR MODELS. I0 I000 ~.36 Answerthe previous question for the Bodeplot below......0 ° I00 ..: ~ I . rad/s I00 _ ..452 CHAPTER6.35 Answerthe previous question for the Bodeplot below. . 0ilO0 o-150 01 01 I I I IIIll 1 0 I I I Illlll\ I I I lllll 100 ea rad/s 10~.. ~ 6.. 1..

Showthat the latter requirement gives the following frequency at the minimumpeak: tip = 2/(2 + r).~)2 + ~’4~’. and evaluate the effective mass.37 A simple untuned viscous damper replaces the spring of the tuned vibration absorbers above with a dashpot. so the dampernever gets over or under-extended. and find the ratio of the maximum displacementat the roof level to the amplitudeof an effective sinusoidal force applied at that level. where ~ = (w/wn)2.= b/2v~(kM). It is proposed to place a heavy mass on the roof that can moveback and forth on rollers. (b) Addthe inherent internal dampingwhich gives a dampingratio 0. whichhas a period of ten seconds. 453 SINUSOIDAL FREQUENCYRESPONSE 6.01. Also.38 A tall building sways excessively in the wind. The building swaysin its first modeof vibration. (b) The best design maybe the one that minimizesthe peak value of the force ratio found in part (a). . and which is constrained by a hydraulic damperas shown: ~per A soft spring (not shown)tends to center the mass. (a) Model the system with no internal dampingand no added mass and damper.(1 +r)f~] ]1/2 and =r m/M. (c) Employthe first derivative condition noted aboveto establish the followingrelation for the associated dampingratio: ¢ = v/r/2(1 + r)(2 + 6.2.The stiffness of the building referencedto the lateral motionat the roof is 4.6. without materially affecting the dynamics. (a) Showthat the ratio of the force on the support to the force F Fosinwt applied to the mass Mhas the amplitude 2~kxo[ = fl(1 ~ + 4~2/r . as shownbelow.0 x 10s lb/ft. . the derivatives of the force ratio with respect to fl and with respect to ¢ should vanish. For this. showthat the minimum value of the peak is 1 + 2/r. n2 = k/M. showingthe need for a heavy mass m. the wind force produces a synchronouscomponent.

02 is desired (doubling the inherent damping).~p.]q . and report the method by which it was chosen.(w/u. Tuned and untuned damper designs are possible. Specify the parameters of a damping system for the building of the previous problem. and that the motion of the mass relative to the building not be too great. ANALYSIS OF LINEAR MODELS. S 2 + ~ p(t) ( 1) = -~q =q-qc=q-CSp q~=q-CSsz:Cl/1Cq= IG(Jw)l [ ’ ] = 1 . Describe the behavior of your solution quantitatively.C (l(t) In operator 4. Don’t confuse q with qc = q.ql. Solve for the ratio of the displacement amplitude of the top of the building to the amplitude of the applied force.): 1 8~¢T/1C q. 0 --~ C dqc 1 d--~ = O(t) . Use the information given and whatever judgements you can defend. and write the corresponding differential equations.2 I 1-2.39 Design problem. SOLUTIONS TO GUIDED PROBLEMS Guided Problem 6. An equivalent damping ratio (for a building with no damping system) 0. 6. ~ = --~ q= S~-~-~-IC. and 0. I~=qoea cos s~ d2p ldqc 1 [ ~ = dt C dt . ql 1 ] - form.03 is excellent. PART 1 (c) Introduce the added mass and damper.454 CHAPTER 6. There is concern that the added mass not be too heavy for the structure to support.

SINUSOIDAL FREQ UENCY RESPONSE 455 0"1. should be less than about 0.3row Guided Problem 6.3. k < 2. the frequency square of the ratio. Therefore. 1 d~ 1 dp~ 1 ~ ~ [ql] ~ p~ ~ I2 ~ R dq~ d ~ 1.~ q~/C~ 2/C2 +e~/It-pz/12x 0 . l~rom the solution to part (a) of GuidedProblem4. P~p2 /I2 1 1 = ~ [ -~/c~ L 1/C~ 1 1 0 0 o -1/C2 0 R R 1/I~ -RH~ R/h 1/11-1/I~] R/I~ -R/h ~ + [ql] . (w/w. (w/wn)2.2.) 2. k > lOmw (ii) For the amplitude ratio to be less than 0.25 N s/m ez) = R(p~/ll-p2/1 ~ F 1.3 1.~)2 must be greater than about 3.1 (p.--’~~.I meshreducti°~-"/~[L~°l/I’ I.~ ~ .1.0 is likely of no practical use.o~ (i) For the amplitude ratio to be within ten percent of 1. 0. 203): rx~ i-~x2 F 1~ = tn I = 1 kg CI = 1/k~ = 1/25 m/N 12 = rth = 2 kg C2= 1/k2 = 1/100 m/N ~ R = 0.o I. (The small range of alternative possible values at slightly above 1.1.~ 0 o. 1 ~-~-~C2 c~ ~ x(x~ P2/~. Therefore.6.

5 7.5 8.255’25).B.62. 375. solve the second equation above for p2 = p2(pl): (R/11)5"-}. combining terms.255’ + 25 p2 -.25 sothat OF LINEAR MODELS.~.125S + 62.5 S2 +0.125jw-t.1250 62.456 CHAPTER 6.D) gives the response 0 1.125J C--[0010]andD=0.J Pl __-. ANALYSIS 0 0 -100 A= 0 -25 25 Lety=pl--[0010]x.S2 + (R/I2)S + 1/I~C1 + 1/I2C~pl : S2 p~ + 0.2500 The accuracy of the second coefficient in the last line above can be increased b: typing p(2).375S a ÷ 12.1/I~CI 0.5"f S2 ÷.5(jw) 2 ÷ 12.0010 0.0000 0.5S ÷ 1250 (jw)2 ÷ 0. G(jw) = (jw). PART 1 .0875 0. 375(jw)3 ÷ 87. 0.25S ÷ 25 S~ ÷0.25 .1255’÷ 12. Therefore.0004 0.125S + 62. x_~: p__L~: F SF S’~ ÷ 0.~ 0.125S+62. 1 1 Sx2 = 1 1 R R 1 1 (1 R R 1 ) 1 RS RS or.5 Substituting this into the first equation above gives I (0.0125 1. R ( $2+ 12 +~+ S+ = (5 6.p] =ss2tf (.5jw ÷ 1250 . the MATLAB command[z. whichgives the response 0.C. Since it is desired to keep pl.125 | and B --.5000 0 p = 1.0e+003 * 0.5)(0. With these values. 0 -.

SINUSOIDAL The FREQUENCY 457 RESPONSE MATLAB commands hum=J1 . bode(num.2.001 2 straight-line ° °90 approximations 10 ~.\ 0. rad/s 100 °180 400 . Problem IIII ~ 10 Frequency (rod/sec) 0 10 10-~ I I I I I IIII~ 10 6.6.5 1250].375 87.5].125 62.4 1 80\ 20". den=It . which shows two modestly damped resonant frequencies: Guided 1.den produce the screen display below.1 : "\~ 50 0.46875 12.

6. since this is the ratio of the asymptotesat w -. Guided Problem 6. Third. a nd drop t o -180° at t he complex pole pair (80 tad/s).19. however.0. the system would be unstable and the frequency transfer function never could have been measured. ANALYSIS OF LINEAR MODELS.-4.5 1. The behavior near the second-order pole pair is estimated with the help of Fig.5. there is a second-order zero at 10 rad/s. which meansthat the numerator of G(S) has a factor of S~ + 2~10S + (10) 2. shownin dashed lines. but the phase angles do not. a horizontal asymptote is drawn to locate the resonance at about 80 rad/s. not the -3 associated with three poles and no zero. The corresponding asymptote in the region immediately above 10 rad/s has a slope of 1. The pole. the asymptote for high frequency is horizontal at the magnitudeof 0. the peak of the resonace is at a factor of roughly 4 above the intersection of the corresponding asymptotes.5)(S2 + 20S ÷ 6400) Note that the number6400 is w~. The actual phase curve nowcan be estimated. the asymptote for low frequencies has a slope of -1. Five features of the given magnituderatio are evident. 1600(S + 20) 5. ¢ -.458 CHAPTER 6. The straight-line approximationsfor the first-order pole and zero. otherwise. The intersection of this asymptote with the first asymptote locates the zero at about 20 rad/s. Then.1. Attempting this combinationby itself produces an asymptote for the high freqencies whichis far belowthe data. Therefore. and to refine the phase plot. The transfer function is very close to G(S) = (S -I. Finally.0. A goodapproachto locating all the poles and zeros starts with approximating the asymptotes immediately after the first pole and after the complexpole pair. as shownby a curved straight line. spreads each rise or fall over two decades. Second. This transfer function overlooks the possibility of pole-zero combinations for which the magnitudescancel. The intersection of this asymptote with the asymptote in the region immediately above the zero gives a pole with break frequency 50 rad/s.w2 + j2w G(jw) = 2jw(jw 50) The ± in the numerator recognizes the possibility that the zero could lie in the right-half plane.. The substitution S jw can be used to check the magnitude plot and possibly refine the transfer function. These facts identify the presence of a first-order zero. The horizontal asymptotes for the phase drop down90° at the first pole (5 rad/s. This gives a dampingratio of ~ = 1/(2 x 4) = 1/8. the value of 1/2¢ is 5. the number1600 is chosen to give G(jO) = and the coei~icient 20 for the damping equals 2¢w. First. Fifth. on the other hand. . 2. The sumof these approximations is given by solid straight-line segments. which meansghat G(S) has a factor of 1IS. rise back to ° at t he zero ( 20 tad/s).). PART 1 The magnitude plot clearly indicates the presence of a first-order pole at roughly 5 rad/s and an underdampedsecond-order pole pair at approximately 80 rad/s.A closer look also reveals that the asymptote for the upper frequency range has a slope of -2. Putting these five conclusions together gives 82 ± 2S + 100 G(S) = 2S(S+ 50) so that 100 . recognizing the value of ¢. must lie in the left-half plane.10 to the actual magnitudeat that frequency. the value of [G(S)[ at w = is 1. Fourth.

The partial fraction expansion also is a tool in the application of Laplace transforms. If the phase difference betweenthe two cases is examinedas a function of linear frequency. so the transfer function becomes G(S) 6. . In this subsection. it can be seen to increase at a steady rate of 0. a powerful formalism closely related to the present techniques. and the minussign gives the proper phase for case (b). but a gradual discrepancy apprears shortly above the notch or antiresonance which appears to accelerate for higher frequencies.tan-l ( 5_--Ow Note that a negative denominatorfor the argumentof an inverse tangent places the angle in the second or third quadrant. This awkwardness can be circumvented by decomposing the model into a sum of first and secondorder models. analytic functions of time are found for the impulse response. A check of a few values shows that the plus sign in the numerator gives the proper phase for case (a).reducing drudgery.3.2. This reveals a pure delay of T = 2~/400 seconds.3. finding the impulse response of the simple component models. TRANSFER FUNCTION EXPANSION 459 ( ±2w / G(ju~) = tan-l \ ~-~--~ ] . The response of a model to this excitation then can be found by multiplying the transfer functions of the special system and the model. Laplace transforms are treated in Section 7.6.3 Transfer S2 + 2S + 100 -(~/2oo)s 2s(s + 50) Function Expansion The direct method for finding the response of a high-order model to an impulse or step requires tedious application of the initial conditions. A complex excitation signal usually can be considered to be the response of a special system to an impulse excitation. applying the partial fraction expansion to the product. Apparently some magnitude-cancellingpole-zero pairs exist. The decomposition is called a partial fraction expansion. Almost all time responses to transient inputs of linear models can be found this way. whereasa negative numerator places the angle in the third or fourth quadrant. The algebraic details that establish the coefficients can be found with the aid of MATLAB. and summingthe results. and in the following subsection analytic responses are found for the step and other excitations. each of which has a relatively simple response. Details follow. on the other hand.9 degrees per radian/second. reaching 90° at 100 rad/s and 360° at 400 rad/s. 6. The commandls±m extends this capability to more general disturbances.1 Impulse Responses You have seen that the MATLAB commands ±repulse and step produce plots and/or data files for the impulse and step responses of linear models. The phase for case (c) is virtually identical to that for case (a) for low quencies.

6514i -10. You alternatively can use the MATLAB command residue. 431.433).den) produces the response r = 0. (s.5000 .2838 -10.0234 0.3.0000 = k [] ’ trans- . for which 200.500S + 200.59) m expands as follows: ÷ rj+l ÷ . The partial ANALYSIS OF LINEAR MODELS.000 G(S) = $5 26sa + 2225S3 + 4900S: + 240.5000 + 44.0.p.00721 ..0234 -1. which requires that you have previously defined the numerator and denominator polynomials of G(S).0072i -2. [r.k] = residue(num.0000 -1. in the standard way.60) ~ Analytical methods for carrying out this expansion are given in Section 6.44.p )2 ÷ rj+m-1 (6. PART 1 fraction expansion of a transfer function G(S) with distinct poles is G(S).6514i -2.l S .14 (pp.p.000 Solution: The coding hum = 200000.rl S A pole pj of multiplicity Is rj tpl - ~ r2 ÷’"+~+rnkiS).. den = [1 26 2225 42700 240500 200000] . hum and den.2370 + 0.460 CHAPTER 6.18 Use MATLAB to find the partial-fraction expansion of the fifth-order fer function of Example 6.0000 -i0.8401 1.3. EXAMPLE 6.~ I (6.

61) A pure imaginary pair of poles is represented by the special case with Pr = 0.1)! r S trteP~ ~ (S . r g(t).pl) (m .5 .andthe corresponding impulse responses arelistedin the right-hand column.r/(S-pl).Pr + p~j _ 2rr(S . (6. Thepossible termsarelistedin the left-hand columnof Table 6.1. 400).28).1)! trn_lePl t rr + rij rr S .ri sin(pit)] . They can be verified by direct substitution into the differential equations that they represent. Itsimpulseresponse canbe found.Entry#2 is merelythe special caseof Pl = 0. It is helpful to notethe identity S: . #5 and #6 represent multiple real poles.pr) 2 + p~.the polePl equalsminusthereciprocal of thetimeconstant.Entry#I represents an identity.2838 10.44. Entry#7 represents complex-conjugate pairof polesin theformthatresults physically.as a special caseof entry#4.Pr .2.5 + 44.2 Transfer Functions G(S) and Associated Impulse Responses ImpulseResponses. there is no remainder. 461 TRANSFER FUNCTION EXPANSION The term k is blank. thatis the formwhichhas onlyre~ltermsin the impulse response. Entries #3.2ripi (s .Pij +S .7(p.0072j 0. As notedin Section 6.p~)2+ 2ep’t [r~ cos(pit) .8401 1. The mostusefulentryis #4.0072j G(S) -S ÷ 2. asgivenin equation (6.6514j + S + 2.Pr) . t >_ 0 rS(t) r rtm-1 r (m .2370 2 S ÷ I0 (S ÷ 10) S ÷ 1 Eachtermin a partial-fraction expansion contributes separately to theimpulseresponse. Thus.2prS + (P~r + P~) =.3.0234 + 0.6.(S .thisgivesthe impulse response rep’t.6514j 1. T. Table 6.0234 . 0.again.0.pl) r m (S .

#4.0005± O.0. The numerator coefficients change. 6.14. designated g(t).2.20 Find the step response for the model of Examples 6.18 using the results of the latter and Table 6.63) If you want to find h(t).2 Step and Other Transient The unit impulse is the time derivative 6.0005± 0.462 CHAPTER 6. as noted in Section or us(t) 1 = -~ 5(t). a zero is added to the end of the list of polynomial coefficients in the denominator. The replacement is accomplished by multiplying the denominator polynomial in G(S) by S. as noted in Section 6. h(t) = G(S)us(t) = -~ 5(t). and adds a zero as the new zeroth-order coefficient.14 (pp. This increases the order of each term in the polynomial by one. you can first find g(t) and then integrate over time. #5 and #4 in the table: g(t) =2e-2’st [0. 431.2370e Nearly all the terms in expansions of any order are addressed in the example above.1. 2368 i. however: 0.0. Check to see if it agrees with the plot in Example6.0234 cos(44.2838e-l°t .1. There is again no remainder.0072 sin(44.8 (p 401). Solution: First. The resulting poles are unchanged. therefore.(t) Responses of the Unit step.0001 . Solution: The respective terms in the impulse response are given by entries #7. so very few situations require additional information or calculation.10. except for the addition of p -.1.19 Find the impulse response of the fifth-order model of Examples 6. As a result. 2370 1.14. (6. 433) and 6. In operator notation.18 and 6.6514 . EXAMPLE 6. is therefore the time integral of the impulse response. 1. designated h(t). PART 1 EXAMPLE 6. ANALYSIS OF LINEAR MODELS.6514 t) . (6.0001 + 0. ~(t) = S u.19.8401 te -l°~ + -~.3. 0840 -1.0. 6.62) The step response.2. More simply. as follows: den = [i 26 2225 42700 42500 200000 0].0000 . you can replace G(S) with G(S)/S and carry out the process for finding the impulse response as detailed above.

0005 sin(4. respectively. a table of "Laplace transform pairs. youintroduce a factor (S+ 1/~)tothe denominator ofthetransfer function andproceed asifsolving foranimpulse response. The two transfer functions are cascaded.6514t)] ÷ 0.6.21 Find the transfer function G~(S)that gives the same response whenexcited by a unit impulse ~(t) as a modelwith transfer function G(S) gives when excited by u(t) = Uoe-~/~. than those given in Table or Figure 6.2370e-~ ÷ 1. Solution: Accordingto entry #4 in Table 6. 6. the given input is the unit impulse response of a systemwith transfer function = u0 The relation between the impulse and the response becomes z(~) = ~(S)~(~) = C~(S)C(S)~(~) S ÷ I/T Tocompute theresponse.3.1.21 by the transfer function G~(S). as presented in Chapter 8. Excitations other than steps also can be related to an impulse by transfer functions. ~ > O.2368e-l°t ÷ 1.2. This procedureis developedfurther through the use of the Laplace transform in Section 7. The sine and cosine terms happento be essentially negligible.and in turn is related to the output variable x(t) by the transfer function G(S). TRANSFER FUNCTION EXPANSION 463 The complete step response nowcan be asssembled as h(t) =2e-2’5~ [0.2. In particular.21: Cascading two transfer functions to relate .0001cos(4.0840re-mr . Their product then becomesan overall transfer function G~(S) between(~(t) x(t) EXAMPLE 6. relates transfer functions G~(S) or G~(S) to a considerably greater variety of signals u(t) or x(t). therefore. It has broad utility in the consideration of control systems. The excitation u(t) is related to the impulse ~(t) in Fig." given in AppendixC.the response of the second to an input impulse rather than its actual input .0.6514t) .

0.1. can be expanded with no remainder. if necessary. for which $2 = -1. They also can be found analytically. for whichS~ = 0. (The operator S is written as a lower case s. multiply G(S) by S + 1 and set S = -1: s]i_.(-1)(-1+ The same procedure gives ra = 0. ’ In the simplest cases the polynomial D(S) has distinct real roots.Si and then setting S equal to S~. EXAMPLE 6.m_~(S + 1)G(S)= sli_. They can be determinedby multiplying both sides of equation (6. D(S) can be divided into N(S) to give a new N(S) with the same order as D(S) plus a remainder polynomial in S. A transfer function G(S) that is a ratio of polynomials in S.59) by S .464 CHAPTER 6.] This gives rl- 2 (0+1)(0+5) ’ . r~(s+ 1)] s+5 J" .2. This first step.59) (p 460).22 Evaluate the residues rl. ANALYSIS OF LINEAR MODELS. i. Thecoefficients ri are called residues. without formal study of the transforms.4. as described presently. -" --+-~+ r~. + r3e Solution: To find r~. PART 1 6. To find r~.3.e. is presumedbelow to have been executed already. r2 and r3 of the transfer function r2 G(S)= S(S+21)(S 5)= rl~+ ~ + S +~’ r3 which has the impulse response g(t) = rl + r2e-t -5t. as in equations (6.3 Determination of Partial Fraction Expansions Partial fraction expansions can be found through the use of the MATLAB function residue. as described above. In the unusual cases where the order of N(S) exceeds that of D(S).m_~ S(S+ 5) This gives 2 r2 .) Youmayuse this table now. with the order of N(S) being no higher than that of D(S). and the remainderk(S) is zero.5. G(S) N(S)/D(S). multiply both sides of the equation by S and then set S -. 6.0: 2 (r r2S r3S "~ lim SG(s) = lim s-~o s-~o(S+ 1)(S+-5)-s-~olim\1 -~~--~ ~-S -[.

463). It is simpler to get these coefficients. find c by multiplying both sides by S + 2 and then setting S -. multiply both sides of the equation by the denominator of the transfer function (so the left side of the equation becomes the original numerator) and proceed to evaluate the coefficients. however. without the check. permitting c to be found anyway. 4-4+5 Then.12. the same procedure can be applied to get the real and imaginary terms rr and ri of entry #7 in Table 6. by leaving the sum of the complex conjugate terms of the transfer function in the alternative real form. . 465 TRANSFER FUNCTION EXPANSION Whena pair of distinct roots are complex conjugate.2 (p. which gives 2rr--a--2.6.44 + 39 -3.2rrPr Pi The complete impulse response becomes -t.-2: c- 2O . + 5si n2t)e g(t) = -2t + (2cos2t 5. multiply both sides of G(S) by the denominator: 5S2 + 22S+ 39 = (aS + b)(S 2) + 3 (S2 + 2 s + 5 ) = (3 + a)S2 + (2a + b + 6)S + (25 + 15).3. This procedure will provide one redundant check on the results for each real root previously evaluated. A comparison of the coefficients of the polynomials in S gives a = 2 and b -. A compromise strategy starts by evaluating the residues for any distinct real poles using the method above.23 Find the partial fraction expansion and the unit impulse response for the model with the transfer function 5S2 + 22S + 39 G(S) = 2 + 2S+ 5)( S + 2 Solution: Expand G(S) as follows: 5S2 + 22S+ 39 aS + b c + 2 (S + 2S+ 5)(S + 2) (S 2 + 4 S + 2 Next. (Had c not been found first. 461). 2ri- -b . Then.2 (p. EXAMPLE 6.) The part of the impulse response corresponding to the second-order pole is found using entry #7 in Table 6. this comparison would have given three equations with three unknowns. plus a check.

Therefore.2. ANALYSIS OF LINEAR MODELS. The formalismof these transforms. The impulse response (entries #4 and #6 in Table 6. The cascading of transfer functions can be generalized to address a variety .466 CHAPTER 6. Systeminputs other than impulses often can be treated as the output of a second system with an impulse input.4 t >_O. is given in Section 7. a~ = 0. The work is expedited through the use of Table 6. along with a broader class of applications. 2a~ + a2 = 2. The product of the associated transfer function with the transfer function of the systemitself transforms the problem into one with an impulse input.60) (p. G(S) is expandedacdording to equation (6. al + a2 + a3 = 3.(l+t~)e 6. from which a~ = 1. Multiplying by the denominator. (The only exceptions are models with more than one repeated real root or a repeated complexpair. 460): S2 + 2S + 3 a3 a2 (S + 1)3 = (S + 1)----~ + (S + 1)----~ + (~ + 1----~. (S + 1)3G(S) 2 + 2S + 3= a3+ (S + 1)a2 + (S ÷ 1)2 al = aiS2 + (2a~ + a2)S + (a~ + a2 a3).3. Summary The transfer function of most high-order models can be represented as a sum of transfer functions of first-order and second-order modelsthrough a partialfraction expansion. EXAMPLE 6. a3 = 2. p. a~ = 1.2.24 Find the partial fraction expansion and the unit impulse response for a modelwith the transfer function 3G(S)= $2+2S+3 (S+1) Solution: First.2.) Thus the step and impulse responses of most high-order models can be found without having to determine the coefficients of the respective terms by the relatively laborious procedureof substituting the initial conditions into the general solution and its several derivatives. 461) is g(t) = t2e -t + O + e-t -t = . or a corresponding table of Laplace transforms such as given in AppendixC. except table entries #5 and #6 are used instead of #7. PART 1 The same methodcan be used in the case of multiple roots.

4. PROBLEMS 6.6.2 (p. or fraction ex- 7. 2. 6. A modeldefined by the differential equation d2 x 2 dx dt "-~ + dt + 8x = u(t) is excited by the disturbance u(t) = 3(1 .2. 5. again using Table6. so its contributionto x(t) is that of a step response. Find the correspondingoperator Gul (S) by which G(S) is multiplied in treating the problemlike an impulse response. Note that the other term in u(t) is a decaying exponential. 6. 461).e-t/2). Suggested Steps: 1. Guided Problem 6. Find an analytic expression for x(t).41 Find partial fraction expansions for the following: . Sumthe operators from steps 2 and 3 to get G~(S). Assemblex(t) from the information given by steps 6 or 7. Multiply G~(S) by G(s) to get the equivalent transfer function G~(S) which treats the problemlike an impulse response. using Table6. as is introduced in Chapter 8. 3. 461).2 (p. 8.6 This is a basic problemin which the temporal response of a linear modelto a disturbance is found using operators and a partial fraction expansion. Use the MATLAB commandresidue to implement a partial pansion of the modifiedtransfer function. TRANSFER FUNCTION EXPANSION 467 of problems regarding the control of engineering and physical systems.3. Write the differential equation in operator form.40 Verify the equality of the complexand real expressions for G(S) given in entry #7 of Table 6. Find the corresponding operator Gu2(S) by which G(s) is multiplied to give an equivalent transfer function for an impulseinput. Notethat the first term in u(t) is a constant. Determinethe partial fraction expansionanalytically. Determineits numerator and denominator polynomials. and deduceG(S).

e-2t). x(O) = O. x(0)= 0. input. (d) Find x(t). and initial conditions d3x d2x "3 dx dt 3 + 3-~y + -~ + 2x = 6(t).42 For the model.44 Find the solution to the second-order problem d2z 2dz dt "-T + dt + 5x = 3(1. the procedure used in Guided Problem6. 6. .(O) (a) Identify G(S). substituting present section. (c) Performa partial fraction expansion of G(S). u(t) = sin(0. Alternatively. 398-399).45 Revisit Guided Problem 6. The following steps are suggested: the methods of the (a) Theinput signal. Determinethe transfer function Gu (S) for that model. Y. 6. PART 1 1 (a) G(S) = (S 1)(S + 2)(S +~ S+ 1 (b) G(S) = S(S 2)(S+3) (c) ¢(s)= (S+1)2(S+2) (d) G(S)= 2S2 + 6S + 7 (S + 2)(S2 + 2S + 2) 6.e-t/T).95wnt)for t > 0 but u(t) = 0 for t < 0. ANALYSIS OF LINEAR MODELS. ~(0) using the techniques Of the present section. whichis treated in Example6.1 (p. (b) Factor G(S). 403).6 can be used. ~c(O) = O.43 Find the transfer function G~(S) that relates an impulse input 6(t) the function u(t) c(1 . Hint: See the preceding problemfor help in finding Gu(S). 6. Compareyour solution to the classical methodapplied to this problem in Example6.21. Hints: This u(t) is proportional to thetime integral of e-t/T . can be considered to result from an impulse exciting an undampedsecondorder model.468 CHAPTER6.3 (pp.

5 4+ = 2. The step response is the time integral of the impulse response. [r.6458i -0. soent ry #4 in Table 6. 6. TRANSFER FUNCTION EXPANSION 469 (b) Find the transfer function G(S) for the given system.5 9 4 0].6.5. PROBLEM 6.0000+ 2. 3. 461)appli es with r = .3.0318i -0. (d) Use the result of part (c) to find the response. x(t).2 (p.3 an p = -0.5.6 1 ~ + 2S + 8" S 2.0. (c) Employa partial-fraction expansion on G~(S). Therefore. 1. G(S) so G~I(S) = 3/S.5) 1 1. (S~ + 2S + 8)x = u(t). 12 G(S) = (S 1)(S 2 + S + 4) SOLUTION Guided Problem TO GUIDED function below to u(t) te -3t.d) produces theresponse 0. n = 1.k]=residue (n.6458i -I.5 S( S+0. 0194.4138 0.2. You may use MATLAB. The MATLABprogramming 5.5 4. 5).46 Determine the response of the system with the transfer the given input excitation.0318i 0. au(s)=GuI(S)-I-Gu2(S)= S+0.p.3750 p= -I. 3 3 1.0194+ 0. u~(t) -3e-t/~.5 G~= ~+2S+8 S S~+0.5S3+9S2+4S+0" S .0000. and multiply it by Gu(S) to get the equivalent transfer function G~(S).5S 6. Therefore.5000 0 k= [] 1. d=[1 2. G~(S) = -3/(S 0.

-0. 8.4138e-°’St ÷ 0.5 = S2q-2Sq-8 + (0q-0-bS)(0q-0. since fewer are required to form a decent approximation.5) gives 12 3 ~+0.0194 cos(2. Convolution is a process of applying superposition.5)-~(s - 1. the present section becomes an optional development that leads to both formal methods of analytical solution and numerical approximations that are readily implemented using MATLAB.6458 come from the factoring S2 + 2S + 8 = (S 1)~ + (2.55 q. from which the responses to each individual input step or pulse is found.375 6.3. 1.05) o. 341-343) and 6.6458t)] .1099)S~ ÷ (0.5)+ 8](-.5 Setting the coefficient of the S3 term equal to zero gives a = 0. If the successive steps . as a meansto determine the response of a linear system to an arbitrary input.1 (pp.1 Decomposing Signals into a Sum of Steps An arbitrary function of time f(t) can be approximated by either a sum of pulses or a sum of steps.22.5 ~ 7.2.4 Convolution* The property of superposition has been emphasized repeatedly. Finally.0388.1293. 6.5a -b b q.0318.0318 sin(2.5) aS+b[ ]1 + [(-. most notably in Sections 3. 5. The response to a unit step or pulse is then determined. The values of rr and ri are rr = a/2 = 0.0. and setting the coefficient of the S term equal to zero gives b -.If. These values satisfy setting the coefficient of the S~ term equal to zero.(0.6458t) .5 aS ÷ b 3/8 12/29 S~+2S+8 S S+0. 148-149). you can restrict study to Sections 6. If you understand this concept.05)~ + 2(-0. x(t) 2e-~[0. (S 2q_2S_bs)S(S. From entries #2. these responses are summedto give the response to the original input. #4 and #7 in Table 6. you would merely like a more thorough grounding in superposition.1. Steps may be preferred. PART 1 1.1 and 6.5.5 Multiplying both sides by (S2 + 2S q. rather. if an infinite number of steps or pulses is used.6458)2. 6.2.4. If a limited number of steps or pulses are used.0647)S-b 1.5s)+ ~+28+8)(s+0. without the detailed analytical and numerical methods that usually can be replaced by other means.0.0.0. as suggested in Fig.0388)Sa q.4.5= (aS +5)(S ~_(S ~+88) +28 = (a .4.8)S(S ÷ 0.b0. the decomposition may be approximate. The values of p~ = 1 and pl = 2.0. in which the input signal to a linear system is decomposed into a sum of steps or a sum of square pulses or impulses equally spaced over time. These values agree with the results of MATLAB found in step 7. the decomposition becomes exact. whichserves as a check.0194 and ri = (-b/2 rrp~)/pi = 0.5) ANALYSIS OF LINEAR MODELS.2 (pp.470 CHAPTER 6.

67) .e. the entire step-function approximationfor t > to can be written f(t) ~_ f(to) ~-~A fkus(t -. it is assumedthat u(t) is zero for t < 0. Either a crude step-wise approximationor an exact respresentation of u(t) can be used.(t ](t) = ](to) T)dT. CONVOLUTION* 471 ~_~(t) ~ ~ df~ ~ ~. so the upper limit of integration can be reduced from ~ to t if desired. In the case of the approximation. In both cases. becomesinfinitesimally small). the steps could be uniformly spaced with a constant interval AT. Alternatively. . multiplying and dividing the summationterm by AT(so as not to change the equation) gives Ark . In this case. (6. u(t) ~ Au(Tk) us(t -.4. Af~ (negative) ~ Figure 6. their various amplitudesare designated by Ark and the times at which the jumps occur are designated by Tk. k.6. Tk)AT. 6.T) = 0 for T > t).22: Approximation of a continuousfunction by a series of step functions are distinguished by a subscript.Tk) k (6..Tk) (6.f(to) + ~ -~--~us~t (6. the ratio A flAT becomes the slope df/dT and the summationbecomesan exact integration: ft? rt df ~ u.4.64) k The steps in the figure are placed at varying intervals of time.65) k Nowas AT -+ dT (i. f(t) ~.66) The interval t < T < ~x~ contributes nothing to the integral (since #l(t .2 Discrete Convolution Anarbitrary input or disturbance u(t) is to be imposedonto a general stationary linear system as represented by its operator G(S). as appropriate considering the varying slope of f(t).

06 0. (6.800 0. u(t) ~.140 -0. The system is excited by the signal u(t) as also plotted.032 0. sec h(t) 0.150 0.30 0.016 -0.04 0.012 -0.60 OF LINEAR MODELS.3 unit step at origin t.015 -0.98 0. plotted as a solid line and tabulated in the right-most column of the table.3.82 0.18/-.u(Tk)AT JT(t -.196 0.988 1.12 0.46 1.38 0.656 0.24 0.004 1.200 0.472 CHAPTER 6.25 Consider the system with the response to a unit step.320 0.032 0.04 seconds and discrete convolution to approximate the response. PART 1 Applied to Example 6.008 -0.14/-.141 0.40 0.768 0.34 0. Therefore.131 0. An alternative discrete convolution is based on decomposing the excitation signal into a sum or sequence of square pulses of amplitude u(Tk)AT.098 0.800 0.080 0.022 -0.16 .004 1.00 0.160 0.58 1.28 0.026 0. but in view of the sluggishness of the step response its approximateness is not bad for most purposes.858 0.04 0.157 0.00 ANALYSIS Equation (6.120 0.54 1.793 0.14 0.48 0.56 0.704 0.88 0.200 0.80 .200 0.23. 6. These are shown by dashed lines in part (b) of the figure.019 -0.148 0.006 -0.00 0. x(t). as measured experimentally (perhaps by throwing a switch) and plotted in part (a)of Fig.975 0.008 0.137 -0.69) .000 The key idea is that the linear operator G(S) satisfies superposition.3.123 -0.20 0.44 0. The times T~ and amplitudes the steps Au.020 -0.26 0.32 0.140 -0.104 -0.000 1.Tk) k (6.176 0.019 -0.006 0.800 0.160 -0.008 0.02/.16 0. The response to the sum of steps is simply the sum of the responses to the individual steps: x(t) ~_ Ek Au(Tk) h(t I .Tk).118 0.800 0.02 .42 1.6s) EXAMPLE 6.907 0.08 0. h(t).16 0.118 0.50 1.784 0.40 0.056 -0.192 0. This is done graphically in part (a) of the figure.084 -0.128 0.134 -0.214 0. which is the product of the height and the width of the kth pulse.00 0.52 0.752 0. Use step decompostion of the input with At = 0.10/.480 0.020 8tim 0.669 0. Tk) also are given in the first numerical row of Table 6.14 .20 0.94 0. The use of only five steps may seem rather crude.160 0.001 -0.064 0.460 0.18 0.003 -0.96 0.60 0. The sum of these responses is the approximate system response. The next step is to find the responses to the five componentsteps.10 0.064 0. Solution: The first step is to approximate u(t) by a sum of steps.154 0.592 0.22 0. and are given in the five associated columns in Table 6.200 0.06/.74 0.132 -0. Table 6.25 Au(Tk)h(t -.018 -0.164 0.68) time ~ sea 0.Zk) for TI~IAu(Tk) .36 0.955 0.115 -0.00 0.200 0.02 0.

23: Example6.u(t) 0.6 ¢ //The dashedcurves represent the responses 0.1 -OT:2"-S ]] 0?3 074 time. seconds (b) response Figure 6.8 0.0 0.6 0.2 0 0 0. 473 CONVOLUTION* 1.4.4 0. seconds 0.4 time.5 0.2 0 0 0.1 0.25 075 0.0 ~ 0. x(t) shed curves) 1.2 ~ 1.6 (a) informationgiven 1.8 citation and stepwiseapproximation.2 systemresponse. 0.6 -- .4 /~~/// of the five componentsteps.2 0.6.3 0.

5648 0.50 0 . (6.04 .sec g(t) .04 . Thesums the responses at a given time t represent the completeresponse x(t) at that time.50 .06 .42 . centered at time Tk.22 .10 .2088 .Tk)U(Tk)AT.54 0 . these are shownin the plot by dots and in the table by the right-most column.9816 0.06 .10 .02 2.18 .004 .0204 .60 .22 .25 .0232 .10 .02 .22 .06 .18 .11 to give .0102 0 values sum (horizontally) .5 4.06 .04 .18 .034 .0714 .11 .01 .0408 .10 .11 . is defined as gT(t -.04 .32 .044 .17 .2244 .24 .17 .22 .3400 0. The responses to the individual pulses are plotted with dashed lines in part (b) of the figure.07 .04 .10 . Valuesof g(t) are listed just underthe figure title.34 4.18 .3 Discrete Convolution by MATLAB The MATLAB commandcony(p. For time convolutionbased on pulses.4 Equation (6.07 . Their times and areas are listed in the first two numericalrows of the maintable given in Table6.30 .04 .24 and the sameexcitation u(t) as repeated in the figure.38 0.25 2.0036 1. p represents the vector of values of Au.02 .18 .1276 . p represents the vector of . the overall responseis x(t) ~_ Z gT(t .22 .9314 0.17 .01 0 0 0 .10 .26 Applypulse convolution to the signals of Example6.70) k EXAMPLE 6.5 .04 .02 .T~) .04 .02 .11 .44 . 6.18 .0 0.0000 The response to a single pulse of unit amplitude.02 .04 .26 .56 .06 .36 .22 .10 0.0696 .002 0 0 0 0 .46 0 .0200 0.11 .4.04 . pulse area (2nd line) .07 .8830 0.5 1.0612 .10 .9650 0.22 .07 .18 .28 .30 .1160 .1734 .04 .0116 0 0 .58 .25. and listed in columnsin the table.10 . 6.11 .04 .08 .O20 .036 .10 . sec (ls_t line).11 .22 . using the unit impulse response g(t) plotted in part (a) of Fig.022 .18 .0408 .07 .{)6 . Solution: Fifteen square pulses as represented in the figure serve to approximate u(t).75 1.02 .18 .18 .46 .02 . q) carries out discrete convolution of the signals p and q.75 1.17 .10 .0020 1.12 .012 .9960 1.42 .5 .8256 0.18 .22 .1972 .04 .04 .4. ANALYSIS OF LINEAR MODELS.1122 .04 .04 Au(T~)g(t .17 .54 .26 for unit pulse centered at t = 0 t.40 .18 .26 .22 .70) Applied to Example 6.16 .14 .38 .1836 .22 .0464 .008 .014 .18 .14 .004 .008 .07 . Therefore.62 .0464 .06 .0232 . which must be vectors of equal length.10 .7308 0. and q represents the vector of values of h.5 5.17 .1360 0.5 .17 .11 .Tk).0204 .2552 .20 .17 .0002 1.22 .04 . Use the sametime intervals.07 .04 .10 .06 .34 .1020 .10 . PART 1 Table 6. For time convolutionbased on steps.17 .58 0 center of pulse T~.52 .474 CHAPTER6.0812 .48 .

2 systemresponse.6.2 0 0 0.4.5 0. seconds Figure 6.4 0.1 0.6 (a) informationgiven 1.2 0.8 0.0 ed from dashedcurves. CONVOLUTION* 475 0.2 0 ~ 0 0.6 .1 (b) response ~_.4 time. x(t) 1.3 0. seconds 0. g(t)/lO ~ 0.24: Example6._ 0.solid line is a smoothedinterpolation) 0.4 ~ ~ I "~lx°ne-tenth the response to a unit impulse.2 0.26 0.5 0.4 time.3 0.6 ~ 0.

25 2.4600 0.02 1 1 1 1 1 1 1 1 1 1 1]*. For purposes here.75 1.8000 0 Columns 22 through28 0 0 0 Columnn 29 0 0.0036 1.2 .5 5. all the remaining numbers should equal 1.88 .7928 0.04.9748 Columns 15 through21 1. whereas the real function continues at unity. The The computation produces a vector for the output that is twice the length of the procribed vectors.1600 -0.0000. x=conv (udt. produces the response = X Columns 1 through 7 0.16 -.5 1.16 .9076 0.8 .14 -.9884 1. The coding udt=[. The MATLABcoding du = [. The vector can be truncated to exclude the spurious tail by entering x = x(l:lS).0080 0.0200 0 0 0 0 0 0 The first 15 numbers are correct.82 . the vector of values of g.2 1 1. g=[2.26 using MATLAB.4 .8584 0. EXAMPLE 6.6 .5 4.04 . h = [. and q represents order of p and q can be reversed. ANALYSIS OF LINEAR MODELS.476 CHAPTER 6.27 Carry out the step-wise Solution: convolution of Example 6.0036 1.98 1 1 1 1 1].0 0. g) gives virtually the same response as the previous step-wise case.25 0 0 0 0].16 1.6688 0. and plotted by entering plot (x). This is because the vector h is truncated at 15 numbers. PART 1 values of the pulse areas uAt. x = conv(du.96 .0640 0.28 Carry out the pulse-wise Solution: convolution of Example 6.25 using MATLAB.9552 0.5 4.02 0 0 0 0 0 0 0 0 0 0].5 0.0004 -0. the second half of this vector is spurious.0000 0. . and should be discarded.2144 Columns 8 through 14 0.74 . EXAMPLE 6.75 1.h) in which du is really Au. but do not.5 0.94 .

These key equations are knownas the superposition or convolution integrals of the time functions h(t) and u(t). .4. and let du = -h(t . wh ich means th at the output equals th e convolution ofthe step responsewith the time rate of changeof the input.72) u dv = uv -Iv du.76) These express the convolution g(t) u(t). for a causal system. 6.72) gives x(t) ~o g(t T) u(T) dT-f_t-to g(~)u(t~) dT.70) for uniformly separated discrete times.72) becomes (6. The discrete convolutionbased on square pulses.T) (6. 472) gives the pair of equivalent equations x(t) = h(t .T) dT = -g(t .73b) dv = u(T) dT.6. The first term on the right side of equation (6. (6. whichvanishes becauseu(to) was given as zero. so equation (6. Recall the formula f (6.75) h(t . Equations (6. as given by equation (6. Thusequation (6. respectively.4.t .T) u(T)Itto =h(O)u(t) . can be seen to be the approximationof equation (6. which re presents th at th e output equals the time convolutionof the impulse responsewith the excitation signal.73a) (6. and. gs(O) 0.71) can be integrated by parts to give an alternative pair convolution integrals that are morewidely used. often written h(t) fi (t).68) (p.4 477 CONVOLUTION* Convolution Integrals For an exact analysis you can let Au -~ du and let T becomecontinuous.T.T) i~(T) dT = (6.71) h(v) i~(t The dots over the u’s in these expressions imply differentiation with respect to T and r -.h(t .to) = 0.76).

4. Thesesignals are plotted in parts (a) and (b) of Fig.) 6. In the example. It can be perceived as a function whichweights the contributions to the present state of the various parts of the excitation signal as a function of the age of those parts.5 Summary The superposition property dictates that the response to a signal comprising a sum of componentsignals equals the sum of the responses to the individual componentsignals. t >O. PART 1 EXAMPLE 6. that is howlong before the present time they occured.the older the segmentof excitation u(T) considered. . the summationsbecomeintegrations knownas convolutionintegrals. Thefirst function g(t-T) can be movedright or left to give any desired time t.478 CHAPTER6. the commondummytime T. Use a convolution integral to computeits response to the input signal u(t) = 1 . since g(t . The functions g(t-T) and u(T) are plotted vs.T) decays monatonically as T is reduced backwardin time from the current time t. their product also is shown.29 A simple system is described by its impulse response: g(t) =-at. in a process knownas discrete convolution. Solution: Fromequation (6. and the input signal directly in place of the changesin the input signal.-bt. 6.76). An alternative discrete convolution employs the impulse response instead of the step response. t > 0. ANALYSIS OF LINEAR MODELS. (The other three convolution integrals can be interpreted similarly. the less it contributes to the response x(t) (regardless of the function u(T)).c) of the figure. the response = -ot e(~_b)r~ t= 1 X--7] a(a-b) The meaningof convolutionis presented visually in part (.25. The shaded area of the product is the value of x(t). and summing the responses to each of these steps. In both cases the result approachesthe exact solution as the steps are made smaller and smaller. Approximateimplementationof this property can be carried out by decomposinga disturbance into a sum of discrete steps.

4..6. 479 CONVOLUTION* u(t) 1/b t (a) input signal (c) computation 1/a t (b) impulse response 1~ .25: Exampleof a convolution integral .. u(T) -)~ I/b T product of abovefunctions: 0 area = x(t) g(t-T)u(T)l~=fg(t-T)u(T)dT T Figure 6..

26: Excitation signal for GuidedProblem6. sec Figure 6.0 seconds. Integrate the unit impulse response to give the unit step response. PART 1 I I t.480 CHAPTER6.0 and duration 1. Analytical convolutionis intended.O 0 0. Guided Problem 6.27. 6. Find the responseof the systemto the individual steps of the input signal. Suggested Steps: 1. The system is excited with a double-step signal as shownin Fig.5 I t.27: Excitation signal for Guided Problem6.0~ t~2.7 This is a mandatoryproblem involving superposition which can be done graphically. Find and plot the response of the system. ANALYSIS OF LINEAR MODELS. Suggested Steps: 1.F~ i ~ul t. Since the excitation is not a finite sumof steps. the superposition must be carried out either by representing the excitation approximately by a . Since the excitation comprises steps. 6.7 1.8 Guided Problem 6. A system G(s) responds to a unit impulse with a square pulse of amplitude 2.0 _~. Figure 6. Superimposethe responses of step 2 to give the complete response. 3. without mathematics.26. Find its response. it is most convenient to work with the response of the system to steps.8 The system in the previous guided problem is excited by the signal shownin Fig. 2. sec l. sec.

Doesit makesense? PROBLEMS 6. You have the choice of computing the convolution g(t) * u(t) or h(t) * i~(t). 00 T t (a) Find the response of the system to unit st ep. 6. sec . Takespecial care of the limits of integration.51 Use MATLAB to plot an approximate response x(t).49 Find and plot the solution to Problem6. If youopt for the latter.47 The response of a particular linear systemto a unit impulse is 2 sin 3t.6.7 (pp. as given below. 481 CONVOLUTION* sum of small steps or or square pulses. find ~(t) and h(t).40 (p.48 Carry out suggested step 9 of GuidedProblem5.71) and (6.50 Do part (b) of Problem 5. t _< 8 seconds~ the modelwhich has the step response h(t) and the time derivative/~(t) of the input disturbance u(t). Let 0 < t < 10. 355) using the MATLAB command rather than lsim.4. Set up one of the fourconvolution integrals as given in equations (6.46 (p. it(t)l 0 0 2 I I 4 t. 469) numerically. Choosethe integration so as to get an exact answer to the problem as posed. or by carrying out a convolution integration. 353-354) using the MATLAB commandcony rather than isim. (b) The system is excited by the square pulse shown below. assuming zero in itial conditions. 6. Sketchthe factors in the integrand versus the variable of integration. using the MATLAB commandcony. 8. Find the smallest value of T for whichthe response of the systembecomeszero for t>T.76) (p. 477). 6. Completethe integration and sketch the result.

t_<0 0 < t < 0.5 h(t .52 Find the exact analytic solution to the preceeding problem. O<t<l 1.5) 0.5 s:x2 = 0.5 {°.5 Responseto step att = 0.5 <t < 1.5h(t)= 0. x(t)=xl(t)+x2(t)= x2I O00. g(t) 2[us(t) . SOLUTIONS TO GUIDED PROBLEMS Guided Problem 1.s ~ 6. O<t<l t_<0 2.7 g(t) at= 2t. 3. 2t-0.i sin~rt) t < 2 is given.5. t<0. The function l (t . As in the previous problem.8 1. 0.5 Guided Problem t .482 CHAPTER 6.5.5<t<l l<t<l. h(t)= f 6. from which . / 2. t.0.0.5 0.5 1.ur(t . t_>l.5 y 1 1L’~ "~ ’ )5t.1)] an d h(t) = 2Jut(t) ur(t-1)] where ur(t) is a ramp with slope 1 whichstarts at t = 0. Response to step att--0: x~=0.1 ’ t_>2 . 2.cos~t) t <2 0 t>2 2. For the choice x(t) = g(t . t ÷ 0. ANALYSIS OF LINEAR MODELS.~(t) = - t ~2(t) J"½(1. t_<0 t.T) ~(T) u(T) case shown: l<t<2 ¯ t-1 1 t 2 ~ r ~ . 0.5 t _>1. PART 1 {}.5.

~(t) = ~ + 2.~ T--sin~t ~ ~ 1 2 5 + ~cosrt =tfor2<t<a. CONVOLUTION* fort _<1.6. 1 2 ~(_~ =~-t forl<t<2. 483 t 1 (T- x(t) = 2.4. ldT = 2 1 Plot of ~he resulg: T--sin~ t2 = 2 1co sa-t] + +cos~t) 2. ~sin~rT)dT dT= dT+ [1+cos ~T + 2. + x(t)= ~(t)= l 5 2.1dT cos~t .~ = -~ +3tfor ¢ ~ a.

Periodic signals are more conveniently represented in Section 7.4. The superposition property is invoked.3 and 6. Certain matrix-based computational methods also are developed. This is awkward for many signals.3 by the decomposition of complex total behavior into modal components. A method to convert bond graphs directly to transfer functions is given in Section 7. Matrix methods for addressing linear inodels are extended in Section 7. Th. called a Fourier series. A twist on this latter transform produces the Laplace transform. called a Fourier transform. disturbances that are zero for t < 0.1 by a discrete sum of sinusoids. The response of a linear model to such a signal is reconstructed by summingthe individual responses to the individual steps or impulses. The section starts by presenting signal-flow graphs. as before: the response of the linear model equals the sum 485 . Then. which are closely related to block diagrams. Instead of steps or impulses. Excitations other than sinusoids are treated in Chapter 6 as discrete or continuous sums of impulses or steps. the signal is decomposed into a sum of sine and cosine waves. which often is called frequency-domain analysis. each of which behaves very simply. which is developed in Section 7. but is logically an optional topic.1 Fourier Analysis An arbitrary signal is decomposedinto a sum of steps or impulses in the timedomain analysis emphasized in Sections 6. 7.is method is a considerable convenience for those who use bond graphs extensively. with a single time constant or combination of natural frequency and damping ratio.Chapter Analysis Part 2 7 of Linear Models. both are widely used by control engineers. pulse-like signals are treated as a continuous sum of sinusoids. including the basis of the MATLAB commandls±m. A different approach is now introduced.4.2. The Laplace transform is ideal for finding analytic expressions for the responses of linear models to transient excitations.

This is an even function of time. The frequencyof the periodic function is (7. 7. The more harmonics that are added.486 CHAPTER 7.) Thesecoefficients are plotted in part (b) of the figure.~ sin(nw0t)]. A signal which is periodic in time comprises a sum of sinusoidal wavesat discrete frequencies. A non-periodic signal comprises a sumof sinusoidal wavesat all frequencies. or by an equivalent sumof phase-shifted cosine waves: u(t)= ao + ~ [2a~ cos(nwot) + 2b. Their sum.1 Fourier Series Consider a periodic function of time u(t) with period T. and the first three harmonicsare plotted in part (c).2) These sums are knownas Fourier series. The general formula for computingthe coefficients an results . called a Fourier series. 7. such as that shownin Fig. The sumof the first seven harmonicsalso is shown. 7. PART 2 -T -T/2 0 T/2 T 3T/2 2T time. t Figure 7. and all sine wavesare odd. plus the time-average term.1.1) = Anysuch function can be represented by a sumof sine and cosine waves.tan ¯ (7. (All cosine wavesare even. that is symmetricabout time t = 0.~ in order to use a Fourier series. called a Fourier transform.2. n----1 -1 + 0n =.1: Periodic signal with period T of the responses to the individual sine and cosine waves. Youneed to evaluate the Fourier coefficients an and b. u(t)= ao + ~ 2c~ cos(nwot + On).1. or anti-symmetricfunctions of time. As a result. is comparedto the actual square wavein part (d). the closer the sum approaches the square-wave function. ANALYSIS OF LINEAR MODELS. only the Fourier cosine series coefficients are non-zero. The exampleof a particular square waveis plotted in Fig.

2: A particular square-waveperiodic signal . three harmonics.7. ti me av erage =1) 2 ~.sum ~seven harmonics. 1 2 3 ..x -T/2 4 0 ~ 2 -T/2 ~’/~"~ (. t (c) the first three harmonics -sumof ~lrst .~. of the first plus time plus the time.~ : t ’ ~ 910 1~ "1~1 " /’1 (b) Fourier coefficients (for r T/4..... FOURIERANALYSIS 4~ time 487 average I time.. av~rage~. t 3T/2 (a) time function a. t Figure 7.¢ ¢’~/~time. average j average exact signal 0 T/2 T (d) sumsof harmonics 3T/2 2T time..first h~rmonic T/2 T 3T/2 2T time.1..

and integrating over one period.2a) to k. their integrals over the time interval T equal T/2.3) give (7. PART 2 from changing the index n in equation (7. you replace the cosine with a sine. have an average value of 1/2. ANALYSIS OF LINEAR MODELS. All terms on the right-hand side of these equations equal zero. equations (7. The terms that vanish are all sinusoids with zero meanvalue integrated over an integral numberof cycles.4b) on the other hand.5) comprise the Fourier series pair.4a) sin2 (nwot) " "~ [1 . To computethe coefficients bn.488 CHAPTER7.~) Equations (7. except for those that integrate the square of a cosine or sine function. which exist only whenk = n. multiplying both sides cos(nwot). The functions 1 cos2(nw0t) = 311 + cos(2nw0t)].cos(2nwot) (7. (7. .2) and (7. The respective integrations are The period chosen here is the symmetricinterval -T/2 < t < T/2. The multiplied functions are said to be orthogonal over the time interval. As a result.

4T 2~rn T sln~-~--) 4~" sin(~rnT/T) =-¯ T ~rnv/T ’ n>l - bn =0.7) whichgives the explicit relationship betweenthe coefficients Unand an. Substituting T = T/4 gives the Fourier series u(t) : 1 + ~ sin(Trn/4) cos(27cnt/T).2b) to . (7.1. 7. where~ /G(jw) is the angle of the phasor G(jw). 7.2 to generate the Laplace transform. then. Un=an -jb. (7.1 Find the Fourier series expression for the square-wavesignal addressed in Fig. The property of superposition. with w = nwo. requires the response to the Fourier series of equation (7.~= T--!J-r~2 dr.~. This morecompactrepresentation of the Fourier series pair is used belowto generate the Fourier transform. it is necessarythat U_. Eachterm in a Fourier serier is of this form. In particular. be the complexconjugate of Un.1.2 Response tion of a Linear System to a Periodic Excita- Recall that the response of a linear systemwith the transfer function G(S) to an input signal u(t) = cos(wt + 8) is x(t) = IG(jw)l cos(wt + ~). and in Section 7. u(t)= ~ Vne 1 fT/2 Un=~ J-T~2 u(t) e-J’~°tdt. Solution: Equations (7. bn. ~n/4 An alternative expression for the Fourier series pair employscomplexnumbers: jn~°t. U-n= an + jbn.2. 489 FOURIER ANALYSIS EXAMPLE 7. n >_1.6) Since u(t) is a real function.7. n >_1.5) give 4cos a.

1. the equivalent result ~ XneJn~’°t. 7.2 and Example7. (7.dcos(nwot + a~). and it can be seen that their sum plus the meanvalue comprises the bulk of the completesolution.2 Find the steady-state response of a linear systemwith the transfer function G(S) 2 . from the given G(S). (7.490 CHAPTER7. IG(2~rjn/T) l 2 V/[5 .8). from equation (7.4.PART 2 ¯ (t) = Xo(O) + ~ ~ IX. Its mean value is G(0) × 1 = 0. Solution: The response. x(t) =n=-~z Xn = a(jnwo)Un.(27rn/T) ~ " The response x(t) is plotted below for the case T = 5 minutes. The magnitudesand phase angles of the transfer function contained in these equations are.6). am = /G(2~rjnt/T).9) EXAMPLE 7. .tan -~ 5. For the morecompactnotation of equation (7.8) am = O~ +/G(jnwo). nmO IX~] = IG(jnwo)lc~. x(t) = G(0) x 1 + 2 si n(~rn/4) n=l ~rn/4 IG(2~rjn/T)lcos(2~rnt/T an). Note that the constant time term is treated as a cosine of zero frequency.$2+6S+5 to the square-waveexcitation of Fig. Eachof the first three harmonicsalso are shown. ANALYSIS OF LINEAR MODELS.(27cn/T)2]2 ~’ + [12:rn/T] 127cn/T ] /G21(2~rjn/T) = .

7.= A~ --~ 0. In this case.1. T. (7. The angles should be in the third quadrantfor n >_ 2.10b) The discrete frequencies nwo becomethe continuous frequency w. If you let T ~ ~. 27r (7. 491 FOURIERANALYSIS response sumof first three harmonicsplus mean meanvalue second harmonic o third harmonic -0.2 -0. min Care must be taken to insure that the phase angles of the individual harmonicsare identified in their proper quadrants. any signal can be so represented.2) becomes u(t) = lim ~ ~ U~e~’~ .1.so it is necessary to add (or subtract) rr to the calculated angles in these cases. Equation (7. however.10a) ~o = -~. nwo ~ ~. 7.4 0 4 5 time. Use of a standard inverse tangent routine to evaluate the phase angle might place the angle for n = 1 in the fourth quadrant and all the other angles in the first quadrant.3 Fourier 1 2 3 Transform Anon-periodicsignal can be viewedas part or all of a single cycle of a periodic signal with an extremely long period.

in Section 7.12a) is known as the inverse Fourier transform of U(jw). X(j~) = G(jw)U(j~). This limitation will be remedied. u(j~)= ~:[u(~)]. .12): x(t) = ~’-~[X(j~)].~) The response x(t) of the linear system G to the input u(t) can be found by ~ modifying equation (7. however. A similar proof is given in Section 7. ~ .2.~3) This computation of U(fi¢) is known as the Fourier transform of u(t). you should know that many commontime functions have transforms that are given in widely available tables. The transform applies only to "pulse-like" signalv.9) after the fashion of equation (8. relieving the analyst of analytical drudgery.2. 2This result also can be proven by applying the convolution integral. noting that G(jw) ~[g(t)].T) (7. for example. 0". ANALYSIS OF LINEAR MODELS. 1Energy or power is usually proportional to a signal squared. ~olim t.12a) (7. and the computation of u(t) using equation (7.492 CHAPTER 7. by employing a modification of the Fourier transform known as the Laplace transform. is known as the frequency or spectral density I . (7. Although the Fourier transform is very powerful in addressing manysignals and the responses of linear systems to these signals. Should you ever wish to do so. which is the content per Hz (per cycle per second).. The formula for U(jw) as a function of u(t) is found from equation (7. A~/2~ l = ~li~o(U. Indicating the Fourier transform with the symbol ~’. where g(t) is the impulse response.2 for the Laplace transform. and for oscillatory functions that do not decay. This text does not ask you to solve problems analytically using the Fourier transform.12b) Here U. the method fails for step functions and the responses of most systems to steps. it has a serious limitation because it is not defined (becomes infinite) if f_~]u(t)ldt = oc. and thus ]U(jw)] 2 is known as the energy density spectrum. Thus.6b): U(j~) (t) dr. u(t) = Y--~[u(j~)]. PART 2 or u(t) = 2~ J_~ U(jw)= ] f U.

These voltages are connected to the input terminals of an analog-to-digital converter ("A/D converter").~). The work of a dynamic analyzer can be performed on a general purpose computer cot~pled to an A/D converter. can compute the fast Fourier transform and its inverse for data provided to it. A sampling theorem limits the number of possible frequencies fn = ~n/2~ = niT for which Un and Xn can be estimated to N/2. as measured by instrument transducers. The details are beyond the scope of current interest. As an example. Numerical values also can be displayed and digitally read out. or alternatively their magnitudes and phase angles.2 part (b) and 7.3). as given by equation (8. the transform then becomes an approximate Fourier series.4 493 FOURIER ANALYSIS Digital Spectral Analysis* The availability of an extremely powerful experimental tool known as the dynamic analyzer or spectrum analyzer is sufficient to justify learning the concepts of the Fourier transform. which generates N digital values to represent each of these voltages at equally spaced times over an interval of time. T. or 512 for N = 1024 or 1024 for N = 2048. The analyzer can display the real and imaginary parts of U(2~rjfn) and X(2~rjf.1. with the fundamental harmonic frequency lIT. one voltage represents an excitation signal u(t). By using a dedicated computer chip these calculations are completed a small fraction of a second.211 = 2048. particularly when high speed is not critical. In practice. T. most dynamic analyzers offer a choice of special automatic "windowing" modifications to minimize errors that result when a non-periodic signal overlaps the time boundaries of accepted data. . and a second voltage represents a system response x(t). a rate of 256. appearing to give it an abrupt start and stop. as described below. 3 The true Fourier transform would show peaks at the fundamental frequency of 1/5 cpm and multiples thereof. The 3Commercially available dynamicanalyzers do not usually take data at such a slow rate. which is much more efficient than a direct approximation of equation (7. At the upper end.000 Hz is common. also. The signal need not have any periodicity.1024 and 800 for N -. Software is widely available. The computation is more generally called a discrete Fourier transform (DFT).1 and 7. since the remaining harmonics are increasingly susceptable to error. using an algorithm called a fast Fourier transform (FFT). the duration of the measurementsis finite. 7.2048 are found. With its typically 400 or 800 points or steps per signal.000data points per secondand a maximum display frequency of 100. In its basic form.1. In practice only perhaps the first 400 for N -. 7:2 and Examples 7. In a basic analyzer. MATLAB.~ = nit Hz. N = 2l° = 1024 or N -. assume that the square-wave signal of Fig.7. so this duration also can be viewed as the period.7). This fact is exploited below to allow the FFT statements in MATLAB to closely estimate the results plotted in Figs. 7. of a periodic signal. The Fourier series for each signal is then computed. the displays resemble continuous functions.2 has a period of 5 minutes and is sampled over a time of T = 200 minutes. and partly justifies the development above. as a function of the frequency ].3. On the other hand.

however. The first ten harmonics appear properly. . On the other hand.4 part (a). A very important special case is the impulse (of zero width).2 -0.2 1:2. gives a transform that is essentially continuous. is different from the ideal shown in part (b) of Fig.3: Display on the dynamic analyzer for the square-wave signal values between the peaks would be zero~ The actual screen display for N = 1024 and lIT = 0.8 ReU 0. cycles/minute Figure 7.2. 7.:4 frequency. 7. a non-periodic signal with a single peak. twenty would appear (up to cpm) if N = 2048. to the maximumfrequency of 400/T = 2 cpm. such as shown in Fig.6 0. PART 2 0. such shown in part (b) of the figure.3. The principal difference is the appearance of much smaller components at many other frequencies. 7.4 0. the signal is an even function of time so that the imaginary part of the transform is zero. These result from the discretization of the signals.494 CHAPTER 7. as shown in Fig.005 cpm. ANALYSIS OF LINEAR MODELS. Note that in the case shown.

’ 10 .2 0.6 ~ 0.1.4/ 0. seconds 6 (a) pulse in time domain 0. ’ frequency.. ~ -90 -135 -18o ° .. FOURIERANAL~rSIS 495 0.3~ 0 0 1 2 3 4 5 time.1 0.4: Exampleof a pulse and its Fourier transform .2 °o ’ ~ ’ .4 0.8 0.7. 0 -45 ’ ~. rad/s (b) pulse in frequencydomain Figure 7. ~ ’ ~ ’ .

an impulse contains all frequencies equally.496 CHAPTER 7.instrumented with an accelerometer to give an electrical pulse whenit strikes an object. ANALYSIS OF LINEAR MODELS.15b) gives the frequency transfer function G(jw)- X(jw) V(jw)" (7. the factor ej~t -~ e° = 1. In practice.3 Find and plot the Fourier transform of the unit impulse ~impulse~(t) (a) time domain t Solution: Equation (7. The results are meaningfulfor each frequency fn that has an experimentally significant content U(2rcjf~). PART 2 EXAMPLE 7. the computations [X(27rjfn)l" IG(27rjfn)l =[U(27rj f. excites all frequencies belowsomelimit associated with its actual non-zero duration.13) gives U(jw) = e-~tS(t) Since the integrand is non-zeroonly for the instant at whicht = 0. For a hammerblow.175) are madeand displayed. so the result is V(jw) = (b) frequency domain-716(t~] note:( _7is positivePhase anglereal)is zero % Thus. Equation (7.16) If the analyzer is used to measure x(t) and u(t) synchronously. the integral of a unit impulseis identically 1. which approximatesan impulse. for each of the 400 or 800 frequencies. a powerfulmeansfor testing mechanical vibration. (7.(2r jf~). and a hammerblow.17a) / G(27rj fn) = (~z (2rj fn) .~)[ (7. Further. This makesthe combinationof a dynamic analyzer and a hammer. for . it also can computethis ratio.

2 and Examples 7.6366 -0.0909 + + - 0. 7. Digital spectral analysis. The remainder of the numbers up to the 1024th are redundant: the complex conjugates of the 2nd through the 512th numbers.0028i 0.1. and are of little significance. The discrete Fourier transform is the same as an approximate Fourier series for a periodic signal of period equal to the duration of the data vector u.00281 The small imaginary components correspond to a time shift of one-half a sampling interval.3001 -0.00281 0. The second element in U is for the frequency lIT cyles per second or 2~r/T radians per second.ra) return the discrete Fourier transform (DFT) of the vector u. One cycle of the square wave of Fig. the third is for the frequency 2IT Hz. and so on up to the 513th which is for the frequency 512/T Hz. U is padded up to the length m with zeros. The discussion above serves merely as an introduction to the possibilities. The commandU=fft (u) returns the Fourier series. the 514th is the complex conjugate of the 512th. techniques and limitations.9003 0 -0. If m is less than the length of the vector u. could be entered as 1024 data points evenly distributed over the T = 5 minutes as follows: ul=4*ones (i.00281 0. 7.m) return the inverse transform. for otherwise a much slower mixed-radix algorithm is used. which means the fundamental frequency lIT Hz. is a powerful tool for examining the frequency content of time-based signals. The first element in U is for zero frequency and therefore is real. and the 1024th is the complex conjugate of the 2nd. U is simply truncated at m elements. with a complex conjugate pair of numbers for each frequency.1800 0 0.1000 0 -0.0692 + + - 0. they can be eliminated . for example.1286 0.0818 -0.1 and 7. They are in inverse order. showing its special virtue (aside from ease ofuse). then.7. this means all frequencies. It can predict the responses of systems. 497 FOURIERANALYSIS example.2.5 Fourier Analysis Using MATLAB* The commandsU-fft (u) and U=f~t (u.00391 0. and determining experimentally the transfer function of a linear system or process by monitoring its input and output.1273 -0. so if the integer m exceeds the length of the vector u.0028i 0. The instruction U(I: 16)/1024 displays the first sixteen elements as follows: 1. The commands±fft(U) and lfft(U. Nevertheless.0000 0.00391 0. 128) u2=zeros (I .1.768) u=[ul u2 ul].2122 0.0039i 0. This choice is recommended.00281 0. It is theoretically impossible to compute more than half as many frequency components as there are data points. If ra is a power of two.0028± 0.0600 + - 0. with each element ~nultiplied by 1024.00391 0.00281 0. a radixol fast Fourier transform (FFT) algorithm is employed.

498 CHAPTER 7.512 frequencies means that there can be 511 complex values of U (the second through the512th. The theoretical limit of 1024 + 2 -. x(t). Gl=freqs (num.2.w) Thecomplex conjugates of the secondthrough the lastvalues. however. PART 2 with the instruction U=real (U). up to at least four significant digits for the first several Fourier harmonics.2 results from the inverse Fourier transform command Pl0t(real(ifft(V)) The inverse Fourierroutineproduces verysmallspurious imaginary partsdue to round-off errors.1 and part (b) of Fig.2 are found as follows: hum=[2]. you should eliminate the imaginary part of the 513th value of G: G(513)=real G(513) Nowyou are ready to compute the Fourier series for the output signal: X=U. in inverse order. and the 514th through the 1024th) plus two real values (the first and the 513th). a plot consistent with the complete response as plotted in Example7. 7. ANALYSIS OF LINEAR MODELS.Shouldyou ever find a non-trivial imaginary partof an inverse transform. . The function G(jw) must have 1024 elements corresponding to the same sequence of frequencies as U. explaining the use of the sub-command real. for i=1:513 w(i)=(i-l)*w0. Theplotting routine requires realnumbers. Therefore. den=If 6 5]. To be completely consistent. end The 512 values of G(jw) for the case addressed in Example 7. you can construct a frequency vector w: w0=2*pi/5 .den. Fina/ly. *G. areneededto complete thevectorG: for i=1:511 G2 (i) =conj (G1 (513-i)) end G=[GIG23 . lookfora mistake. The Fourier series can be multiplied by the frequency transfer function G(jw) to get the Fourier series of the ouput signal. therefore. The result agrees with that of Example 7.

The result approximates the Fourier transform for a non-periodic signal.6 Summary A periodic signal can be represented by a Fourier series comprising a sum of sine and cosine waves. what emerges is the lower-frequency signal of interest. The analyzer can process both excitation and response signals simultaneously. 7. The saw-tooth disturbance plotted in Fig. Find the responseq( t ). by passing it through a "low-pass filter".9.1. The response of a linear system to a periodic signal is the sum of the responses to the individual sine and cosine waves. and a Fourier series for a periodic signal with period equal to the duration of the data train. The criterion for its existence is that the time integral of the absolute value of the signal over its entire duration be finite. for example.7. such as pulses. It then can compute and display the ratio of the two. ruling out step-like functions and oscillations that do not decay. Suggested Steps: 1. A broad class of aperiodic signals. can be represented by a Fourier transform. MATLAB has a capability in this regard.1 This problem provides a basic experience in computing a Fourier series for a periodic signal. so that the ouput signal might look quite different from the input signal. A discrete Fourier transform is computed for a train of numbers that represents a signal measured over a period of time. This is the frequency transfer function. The frequency of each component is an integral multiple of the frequency of periodicity. Guided Problem 7. Find the average value of the disturb~ce e(t) and the consequent constant term in the response of qc(t). 499 FOURIER ANALYSIS 7. High-frequency noise often is purposely stripped from a measured signal. and then finding the periodic response whenthis signal is applied as the excitation to a linear first-order system. The Fourier transform is the same as the Fourier series except it includes a continuous distribution of all frequencies rather than the multiples of a base frequency. Digital spectral analysis may be carried out by a special instrument called a frequency or spectral analyzer. p 142). The different frequencies are apt to be filtered quite differently. .1. or on a general-purpose computer. expressed as a function of frequency. G(jw).5 acts on a system modeled as dq T-~ +q = Ce (which corresponds to the RC system of Guided Problem 3. The response of a linear system to such a signal is therefore the sum (or actually the integral) the responses to the individual frequency components.

~. Assemble the response of the system to the complete disturbance by combining the results of steps 3 and 4. Suggested Steps: 1.2 This problem both illustrates basic concepts regarding the use of Fourier series and transforms and explores a very practical appliction of the dynamic analyzer.) . but the difference is not significant for the low frequencies shown. Evaluate the Fourier coefficients an and bn for the disturbance using the function e(t). A perfect impulse would have a constant magnitude spectrum.5: Disturbance for Guided Problem 7. Write the Fourier series using the cosine form with its coefficients cn and phase angles fl. Find the response of the system to a cosine wave of arbitrary phase angle. Also.500 CHA’PTER 7. 5. The acceleration or force signal from this hammer and the signal from an accelerometer or other motion transducer attached to the object under test are sent to a dynamic analyzer. The actual signal decays with frequency and goes negative at a particular frequency. Guided Problem 7. 0 ANALYSIS T OF LINEAR MODELS. Find an analytic expression for e(t) valid for at least one complete period of the disturbance. The analyzer displays the Fourier transforms of these two signals as shown in Fig. evaluate T. The spectrum of the response of a linear system to a perfect impulse is proportional to its transfer function G(jw). sketch-plot the transfer function (which the analyzer would do for you if asked). 7.1 2. you would likely benefit from carrying out a similar experiment. as described in the text. 3. Should you have access to a dynamic analyzer with an instrumented hammer. Discuss briefly where the analyzer with this excitation is prone to give large errors in G(jw). PART 2 2T time t 3T Figure 7.6. 4. Describe qualitatively in what way the actual excitation pulse differs from a perfect impulse. Show that a square pulse of duration T seconds could produce a hammer signal similar to that given. (The actual pulse likely would be rounded. Experi: mentally one attempts to generate the impulse with an instrument6d hammer.

501 FOURIERANALYSIS accelerometersignal o0 phase °180 I instrumented unOer/ test k. There are no scales given for the magnitudeplots.. log amplitude ~ ~ hammers~gnal: acce1 rom~eter eome analyzer dynamic phase0°1 °180 . and sketch-plot the magnituderatio. 4.2 2. Notethat the log of a ratio of twoquantities equals the difference betweenthe logs of the quantities.6: Guided Problem7. 3. The interest presumablyis in the dynamiccharacter of the response. Hz Figure 7. so the best you can do is sketch-plot G(S) to the same scale with no absolute amplitude specified. The phase of G(j~) equals the difference between the phases of the numerator and the denominator. that is the natural frequencies and relative magnitudesof the various resonances. Note that large errors in a computedratio occur whenboth the measured numerator and denominator are small.og amplitude 0 200 400 frequency. Sketch.7.1. .

Find the response of the system to the symmetric square wave shown below. (b) Find the transfer function of the system. plotted below.1 A linear system is excited by the signal u(t) and responds with the output variable x(t). (c) State which frequency component predominates in the response x(t). (e) Assemble the response x(t). 7. is the periodic signal (valid for all t) u(t) = 2 cos(3t) + cos(9t).502 CHAPTER 7. PART 2 PROBLEMS 7. 0 1 2 3 t 4 5 (a) Find G(jw) for the system. or the corresponding differential equation. ANALYSIS OF LINEAR MODELS. and the excitation. Let T = 2T. (d) Evaluate /G(jw) for the frequencies appropriate to this problem. . (b) Evaluate IG(Jw)lfor the frequency(ies) appropriate to this problem. (a) Evaluate the first three non-zero Fourier series components of u(t). The system is described by a~ x dx dt ~ + -~ + 81x = u(t).2 A linear system responds to a unit step excitation with the output 2(1 e-t/T).

and find the response to this zero-frequency component. Plot these emd their sum. 7. and setting the two states equal to one another in order to evaluate the actual state at the beginning and end of each cycle. neglecting any transient behavior associated with the initiation of the force. Carry out this procedure. (b) Write expressions for the magnitude and phase angle of the nth harmonic component of the response.2.4 Find the steady-state Fourier transform response of the second-order system 1 d2x 2~ dx w~dt ~. 489). as shown below. Which component is the largest. and compare the result to that of part (d).1. (d) Find the first three Fourier series components of x(t).-~ + x = u(t) to the square-wave disturbance u(t) shown in Fig. Find the consequent velocity as a function of time.7.1 (p.2 (p. AssumeT = and ~ = 0. (a) Note the time-average value of the disturbance.~-.5 Find the response of the problem above using the fast Fourier transform capability of MATLAB. using the given step response to find the consequent state at the end of one cycle. 503 FOURIER ANALYSIS (c) Find the response of the system to a sinusoidal excitation. (e) Since the signal comprises repeated upward and downward steps equal amplitude. and why? (The use of MATLAB is suggested./ %. 7.) (d) Plot the sum of a least the first overall response. 487) and described with the Fourier coefficients found in Example 7. five harmonics to approximate the 7.+~. . . 7.3 Find the response of the problem above using the fast capability of MATLAB. A /’ VtV"V.6 A mass-dashpot system is excited by a triangular periodic force. 7. either sketch-plots or the use of MATLAB are acceptable. (c) Plot the first five harmonic components for one cycle. the exact solution to this problem can be solved by assuming an undetermined initial state.

The calcuation can be simplified by noting that F(t) is an even function. (d) Find the response of the system to the nth harmonic. Note finally that the even numbered harmonics vanish. assuming both can be represented by ratios of polynomials. Sketch these terms and their sum. so the interval of integration can be limited to positive values of time. (a) Characterize the difference between the two transfer functions. using the fast Fourier transform capability of MATLAB. (b) What causes the difference? (c) What categorical nitude Bode plots? difference would you anticipate between the two mag- . 7. With one of the specimens the phase angle between the excitation and the response is 0° or +180° at virtually all frequencies. You may wish to use a table of integrals. a = 1 N and b = 10 N s/m. (b) Deduce G(S). since this reduces the number of summations if you intend to make a sketch-plot. and only the cosine terms are non-zero.504 CHAPTER 7.7 Plot the velocity ~ of the problem above. (e) For T = 1 s. m = 1 kg. 7. otherwise.8 An instrumented hammer strikes two specimens which are freely suspended and instrumented with accelerometers. determine A and B. for the values given in part (e). With the other the corresponding phase angle varies continuously with frequency. Note that very few terms contribute significantly. PART 2 (a) Write the differential equation with 2(t) as the state variable. Sum these responses to get an expression for the overall response. Represent this terms of a phase-shifted cosine or sine wave rather than separate sine and cosine waves. which in this case are scalars. with abrupt jumps in between. evaluate the magnitude and phase of the first harmonic of 2 and all others that are as large or larger than one percent of the first harmonic. ANALYSIS OF LINEAR MODELS. integration by parts can be used. (c) Find the Fourier series representation for the applied force F(t).

(b) If damping is neglected there are two time functions which the data could be describing. l~rom equation (7. and particularly the effect of radial as well as bending stresses.1 Problem PROBLEMS 0 T 2T 3T time~ Theaverage value of the disturbance is seen from the plot to be eo/2.7. dt+~ O< t < cos ~ dt .9 An accelerometer placed on the rim of an axisymmetric bration absorber. ’° ~ ~ tcos T2 ~-T/2 = -~t. e(t) 3.1. Also. (c) (extra credit) Explain. the cause of the non-fundamental harmonic revealed in the display. and state why the ambiguity exists. qualitatively.5). accelerometer rim hub ~sk ~F=Fos~ ~ (a) State if and why the data indicates linear or nonlinear behavior. suggest any practical implications this might have for the absorber. [Hint: Consider the nature of the spring characteristic between the hub and the rim. produces below right on a spectrum analyzer when shaken with a purely at frequency w0. 2. shown below left in cross-section. Find and sketch-plot these accelerations. 505 FOURIER ANALYSIS 7. e(t)= eo eo ~-t+eo. Does the natural frequency of the absorber vary with the amplitude of its motion?] SOLUTIONS TO GUIDED Guided 7. Both vertical and horizontal axes have linear disk-and-rim vithe display shown sinusoidal force scales. -T<t<0.

Ceo is qc = ~. ~ eo e(t) = ~ (~) 2e~osin 2 t 7rn -- dqc v-~. PART 2 sin + -T/2 "0 q_. ~oi ~’~(~_~)+~oo (~) dt b. = T~ cos ~ ANALYSIS -- OF LINEAR MODELS.sin(--Trn)]---This answer can be anticipated since e(t) is an odd function of time and a~ represents only the cosine or even parts of the signal.tan-r(T~). then c qc= ~+1 qc = ~ +~=~ cos[~+ ~ . Result for r/T = 1: 1 2 tiT 3 .+ qc = Ce.~ [2~_~sin (2_~_~t)] -T/2 =eo(~[cosQrn)--cos(--Trn)]+4~n[Si~(Trn)--sin(--Trn)]) + 2~n[Sin(0)-. ~1 + (2wriT/T) e cos + 90° + ¢~ .ta~-~(~)]. e0 = -~ + The steady-state ~ 2eo --zrn (~) 2 t cos --+90 ° . = ~ ¢-~/e tsin .=T~ ~ -~cos ~ ~T/esin -- -- -T/~ dt T ~cos =eo(~[sin(~n)-sin(-wn)]-~[cos(O)-cos(-~n)]) = {-eo/wn nodd 0 n even Therefore.506 CHAPTER 7. If e = e~ cos(wt + ~ .

~_~lu(t)e-~tldt < ~. Hz Transform The Laplace transform is a formal extension of the operator methods that have been used. THE LAPLACE Guided Problem 507 TRANSFORM 7. The Laplace transform is used in Chapter 11 to aid the solution of partial differential equations. precluding the use of the Fourier transform of u(t). however. The Laplace transform can be addressed at different levels. If. you should gain a deeper appreciation of the transform and its relation to other methods. the Fourier transform of the product more likely does exist. This is virtually always true for signals that are zero for t < 0. If rather you start with Sections 7.2 100 The Laplace 200 300 400 frequency.3 and continuing most particularly in Section 6.3. which is the traditional . starting in Section 5. You will satisfy the mini~numrequirement for practical use by starting with Section 7. or T = 1/300see.1 Development from the Fourier Transform* The integral f_~lu(t)ldt often doesn’t exist.2. it is shown to be adaptations of the Fourier transform and the convolution integral. which it converts to algebraic equations.3. that is. Crude model of pulse: U(jw) e-J’l dt = 2 ] cos wt dt = .2 1. It is used to aid the solution of ordinary linear differential equations. Problems with non-zero initial conditions are treated. the function u(t) is multiplied by e -~t. which it converts to ordinary differential equations. 7.2.1 and 7. as you saw in the last section.7. as well as the class of excited systems with zero initial conditions addressed in Section 6.2. at least for some range of a.2.2.2.sin ~t) d --T/2 This resembles the given spectrumfor the hammerif wt/2 = ~r for w = 2~r x 300.3. 0 7.

Therefore. ANALYSIS OF LINEAR MODELS.1g): This is a key result that you will employto computethe response of linear systems to explicit excitations. ~hich is included in the study of f~nctio~s of a complex wri~ble. it is beyond the scope of your present need. Two-sidedLaplace transforms are used infrequently because of convergence problemsthat often arise whenu(t)¢ 0 for t < 0. (7. The Laplace G(~). There is a great deal that can be done without this knowledge. u(t) = ~_~(e + j~)e~t~ = U(~ + j~)e(~+~td(j~).18) or U(s) t)e-Stdt =_£[u(t)]. usually involving residue theory. and the special G(S) used for exponential functions st e (including sinusoidal signals whenS = j~). (Z. The response of the linear process G(~) to the input u(t) nowCanbe generalized from equation (7. Partial fraction expansions also help.19) is defined. This book employsthe simple expedient of a library of commonly occuring pairs of functions of time and their Laplacetransforms.20) or. is knownas the transfer function of the system. (7. The inverse formula is u(t)e -~t = ~-~[C(a + jw)]. s = e + j~. Although some readers mayknow about this subject. This is knownas the two-sided Laplace transform of u(t). ~or the cases of usual interest in which ~(t) = 0 for ~ < 0.508 CHAPTER7. PART 2 transient signal. from equation (7. the function U(a + jw) = :~ [u(t)e -~t] = (t)e-~te-J~tdt. .12a).21) or ~(t) 2~ J~_~ This is integration in the comple~plane. unfortunately is not simplified. however. they reduce to the commonone-slded Laplace transform. equation (7. 2~j ~ (7. like the Heavisideoperator G(S) used repeatedly before.22). They are related intimately. it is merely necessary that a be greater than somefinite positive number. in table form. which is the only Laplace transform referred to in most textbooks: The inverse formula.

that is excluding outputs that occur before their causal inputs.24).. respectively.2. with the addedobservation that G(s) is the Laplace transform of the impulse response. 509 THE LAPLACE TRANSFORM ~region of d( J $ integration . Presuminga causal process. as suggested in Fig.7..2 Development from the Convolution Integral* Equation(7.2. ~ and ~.L--Z. the product is computed. The objective is to derive the Secondof equations (7... the convolutionintegral for the response of a system with impulse response g(t) to the excitation u(t) can be written as x(t) = u(t .7: Changingthe order of integration in equation (7. J . The definition t ~ ~ + ~ gives The key idea nowis to changethe order of integration. \ t Figure 7. This gives . 7.taking care to use dummytimes.28) 7.4) g(~) (7.7.. any symbol could be used. Therefore.25) Note that ~ is a dummyvariable. The one-sided Laplace transforms of the excitation u(t) and the response to a unit impulse g(t) are.24) (and its counterpart for Fourier transforms) also can be derived from a convolution integral.

or you represent that history solely by its effect on f(0) and the time derivatives f’(0). A set of transform pairs sufficient to ham die linear differential equations with constant coefficients is given in Table7. f"(0). Note that the transform is unaffected by the history of f(t) for t < 0. s = a + jw. reduces the need for partial fraction expansions. 27r3 J~_j~ (7. ANALYSIS OF LINEAR MODELS. models .510 CHAPTER 7. Therefore.()g(()d~) The integral in the brackets { } above equals x(t). 7. Either you restrict its use to functions that are zero for all t < 0. whichis not addressed here since you need to knowonly a restricted set of easily computedtransforms. 7.3 Definition and Inverse The Laplace transform F(s) of a function of time f(t) usually is defined as F(s) =.2.31) This transformexists (does not becomeinfinite) for all physically possible (finite) signals if a has a real value larger than some minimumnumber.4 The Derivative Relations The Laplace transform of the time derivative of a function is related to the Laplace transform of the function itself in a special way that can be found by an integration by parts: 4The pairs given on the last page of Appendix C apply to distributed-parameter such as in Chapter 11.2.32) Its direct evaluation is a majortopic in the study of the functions of a complex variable. (7.25). PART 2 G(s)U(s) = fo~ ( fote-Stu(t . G(s)U(s) = fo~e-St x(t)dt = (7. A moreextensive set of Laplace trangform 4pairs.£[f(t)] = f(t)e-Stdt.1. as indicated by equation (7. several of these are derived below. The inverse Laplace transform is written as f(t) = ~1I ~+~F(s)e+~tds. given in AppendixC.30) establishing the objective.

ab)/bw] ~b ---. ¯ e-~t f(t) +1 0 f(t .2..T) F(s + a) e-T~g(s) unit impulse (f(t) unit step us(t) 1/s 1 1 n+l 1/8 (n = 1.ab)] Jt=O ...2...tan -1 [bw/(c -.~f(t) F(s) s’~F(s) s" -lf(0) _ s’ -2f(O) ~F(s) f_t~f(t)dt ¯. 2.7.¢) = Ae-at sin(wt + ¢) A = ~/b 2 + [(c J~ ~ d (-1)n d-~F(s) tnf(t) ltn n! ~.1 A Short Table of Laplace Transform Pairs Laplace transform time function.2ab)/w] (bs + c)/[(s + ~ + ¢ --.3. t _> 0 af(t) aF(s) f(t) +g(t) ~ d dt. .(n = 1.) ne -~ cos(~t .tan-][(c .~t e.) " -at e n 1/(s + a) at 1/(s + "+~ a) ~. 511 THE LAPLACE TRANSFORM Table 7. 3.

5 Singularity Functions and Discontinuities It is also important to have at your disposal the Laplace transforms of a handful of key functions. fn-l(0-) are all 7. equation (7.. (7.0.36) The lower limit of integration in equations (8. s often is called the time derivative operator..38) . Applying this property again to find the Laplace transform of the second derivative. (7. 6 and 8. PART 2 =~ Sfo~e-stf(t)dt + e-Stf(t) 0 = sF(s) . Thus. a description which is most apt in the most commonapplication for which f(0) = 0.. but it is essential that all terms in the equation be interpreted consistently.35) operator: (7.33) As a result. for the nth derivative..fn-l (O) in th e g eneral pairs for first and higher-order derivatives as given in the table. since then the values f(0-).. as you have seen.option.31) and the transform pair (7. if(0-). This time is ambiguous in cases in which the function jumps from one value to another at precisely t -.36) is given as t = 0. d2 f(t) L:[--d~] (7. and the one approached from positive time as occuring at time 0 +.sn-2 f(O) . (7..34) = s£[~]-f(O)=sg"F(s)-sf(O)-](O).or 0+ in a given equation. . The two values can be distinguished by referring to the one approached from negative time as occuring at time 0-.(dn-l f The inverse of s can be considered to be the time integral - 8 ~ )dr. In practice one may choose to use either O.f’(0) . there also is f(0).!t)] [ a~ j = 1-us(O).37) 0 The unit impulse 5(t) is the time derivative of the unit step. /dtn-1)t= o. ~[~] = snF(s) . Whenone deals with functions f(t) that equal zero for t < 0. It can be viewed as a ibr~nalization of the operator S used in Chapters 5.f(O).33) gives £[5(t)]= £[dus. Thus.512 CHAPTER 7.sn-l f(O) .2. The transform of the unit step us(t) is £[us (t)] = fore -st dt = -8 1e-St ~=8"1 (7. it is simplest to employthe 0. ANALYSIS OF LINEAR MODELS..

41) For the product ~ d = Finally. whichgives/:[5(t)] = 1.39) Theuse of the 0+ option in effect transfers the role of an excitation to that of an initial condition. 7.T)] = foo°~e-~tf(t . is that either is correct. Z. (7. here. £[f(t . Whichdo you use? The answer. Examination of the defining integral reveals that the impulsethen resides entirely outside of the integration interval. however.513 7.T)dt ---.2.40) For a product of the exponential decay and a function f(t).stdt= e-(a+~)tdt _ _e_(~+~)~ oo_ (7. This result maybe checkedby direct calculation: + 0 (7. In the present case the use of 0+ gives £[(~(t)] = 0. for a function shifted in time by the delay T. as long as you choosethe sameO.e-TSF(s)¢ where f(t) = if t < 0. again..1. whichisn’t very useful. [e-~t:(t)] e-(~+"~tI(t)dt = F(s + (7.2. u~(0-) and us(0+) -. In a particular case. whichin somecases maybe helpful.6 Other Key Relations The Laplace transforms of other key functions given in Table 7. For the exponential decay e -at ~.or 0+ in the Laplace transform of all terms. To include the impulse in that interval you use the 0option.1 are computed -a~. one maybe more useful than the other. THE LAPLACE TRANSFORM But what is u~(0)? At time t = 0 the function jumps from 0 to 1. e [ ]= e -ate .43) .

48) [ dx(t)~ : Ax(t)+ Bu(t). ] Takingthe Laplace transform of both sides gives sX(s) . To address the linear state-variable formulation. Therefore..45) from which X(s)- F(s) p(s) (7.-~ +. 7.1) \dt ]t=o + a..44) an ~--~ + an-~ ~ +"" + a~-~ + aox : f(t) becomes (ans’~ + a... (dU-lx/dt~-l)t=o are zero. In this case.a X(s) =~~L(a’~sn-~ +""+a~)x(O)+(ansn-~.8 (pp.. also. The other major situation is the initial-value problemin whichf(t) = for t > 0.2.x(0) = AX(s) + (7. 439-441).. The Laplace transform of the response of an initially quiescent system to a transient excitation is especially simple to find. ~(0).1.47) In almost all cases of interest. since whenmultiplied by the Laplace transform of a function of time it gives the Laplace transform of the same function delayed in time by T.. recall the matrix notation: (7.49) . is addressed subsequently.~_ls n-~ +. because all of the initialvalue terms x(0).44) gives 1 [ +a~-~s~-2 .PART 2 The function e -Ts is knownas the time delay operator. The time delay operator maybe viewed as a generalization of the frequency pure-delay operator presented in Section 6.. The Pade approximantsintroduced there apply here. very few of the terms within the square brackets above are non-zero.46) wherep(s) is the familiar characteristic polynomial. The second step. + a~)~(0)(artsn-3+’" +a~)~(0) +.2. usually the terms are best found by direct application of the third entry in Table 7. evaluating the ouput variables themselvesfrom the transforms. the Laplace transform of equation (7.~ \ dt-~-~Y-~ ] t=0J n-3 8 (7. + als + ao)X(s) = F(s). (7. the Laplace transformof the scalar differential equation nd d’-lx dx (7.514 CHAPTER7. ANALYSIS OF LINEAR MODELS.7 Finding Laplace Transformsof Output Variables Thefirst of twosteps in solving a differential equation using Laplacetransforms is to find the Laplace transforms of the desired output variables. Note that the switch in the lower limit of integration from -T to 0 requires that the function be zero-valued for t < 0. with S = s. + (a~s + an .

EXAMPLE 7. x(t).1.5 (p 397) to represent the ratio of the response x(t) to the excitation est. using s = jw. using the Laplace transform table in Appendix C. y(t) = Cx(t) + Du(t).A)-IB + (7.A)-Xx(0).3 (p 345) merely to represent differential equations.7.5~) In this case.52) The matrix H(s) is simply a more generalized transfer function which allows output. G(s) and H(s) are identical to the operators first introduced in Section 5. y(t). These facts are closely related. A more general case traditionally variable is defined as value problem.2.A)-~B. Setting B = 0 or u(t) = 0 leaves an initial leaves a forced response problem. X(s)= G(s)U(s) G(s)= (sI + (sI . and setting x(0) = is cited in the literature. in the language of most elementary textbooks in automatic control which do not employ matrices. Note that the expressions for G(s). The elements of the matrices G(s) and H(s) scal ar tran sfer func tions or.A)-Ix(O) (7.4 Find the response of a system with the scalar transfer function 2 a(~) (s + 1)(s+ (which is the factored version of the transfer function used in Example7. The scalar G(s) also is noted in Section 6. except for the use of in place of S. H(s)_ = C(sI . Solution: The Laplace transform of the unit step is 1 u(~)= 8 . p.2. The output (7. to be different from the state vector.3. 515 THE LAPLACE TRANSFORM Solving for X(s). Y(s)= H(s)U(s) + C(sI . The subsequent development extended this result to frequency response. 490) to an input unit Step. just "transfer functions".50) .

c+ ~(0)+ax(0)sin~dt] [ EXAMPLE7. ANALYSIS OF LINEAR MODELS. ~(o) 1 -4 Solution: From equation (7.0. 400-401). 608) gives :~(t) = x(O)cos~t -at .5 Use Laplace transforms to find the response x(t) of the general underdamped second-order model to an impulse cS(t). 4 . 0. . with the coefficient values a -.2(0). pp. PART 2 The Laplace transform of the response x(t) is the product of G(s) and U(s).50).~.[sI - s/2+2 2 (s+4) s+4 2s+8 s:+6s+7 AI = s ~ + 10s~ + 29s + 20 = (s + 1)(s + 4)(s .1 and b = 5.516 CHAPTER 7. Se-t + -5t.(~ + ~)2~ [(~ + ~)x(0) +c + ~(0) +~( which from items #18 and #19 in Appendix C (p. Carrying out the indicated operations. X(s)=~[ 1 [ s2 +6s+6 2s+8 lg~a) 2 p(s) =-.2a~.(0) a = ~.A)-~x(0). 1 x(~). x(s) = (~ . or 2 X(s) = s(s 1)(s + 5) This is proportional to transform #15 in the table of Laplace transforms in Appendix C.4. Solution: The Laplace transform of the differential equation gives [(s + a)"~+ ~]X(s). Solving for X(s).6 ~ Find the behavior of the state vector of a third-order autonomous linear system described by its matrix A and initial conditions x(0): A= -2 2 1 . The corresponding time function therefore is x(t) 0. in the presence of general initial conditions x(0) and 2(0) (same case as in Example 6.~x(0). 1e EXAMPLE 7. 1 .

they are very special. The usual result comprises terms of first and second order only.1 and on the first page of the table in .D. You are urged to review the material at this time.8 Partial l~raction Expansions Laplace transforms comprising ratios of polynomials so as to represent ordinary linear differential equations with constant coefficients can be expandedin partial fraction expansions to reduce the complexity of the individual terms.2. is x(t)=x(0) From the present perspective the cancellation of so many nu~nerator and denominator factors to give so simple an answer seems rather an incredible coincidence. 461) of transfer functions and their associated impulse responses should be recognized also as a table of functions of time and their Laplace transforms.3. One rarely chooses initial conditions corresponding to a mode of motion accidentally. a great simplification. 7. Although that discussion preceeds the formal introduction of Laplace transforms. it applies directly to them.B. in Section 7. Partial fraction expansions are presented in Section 6.3 (pp. Table 6. from either the tenth item in Table 7.den]=ss2tf(~.2. All but the final entry in that table can be found in Table 7.1 (p.±) gives the numerator and denominator polynomials of the matrix H(s) = C(xI-A)-IB+ D given in equation (7. The coefficients of X(s) above result from setting B equal to the vector x(0) and D equal to the null vector.. Thus. X(~) = 1 +6s+5 0 ~+6~+5) -(s 1 (s+l)(s+4)(s+5) - s+4 ~(0) (s+l)o(S+5) _(s+l)(s+5)l z(o).1 or item #5 in the table of Appendix C. Later.4. the coefficients of the nine polynomials in p(s) above are given (as well as the denominator coefficients) by setting C equal to the unit diagonal matrix and D equal to the null matrix.7.C. this problem will be shown to correspond to an excitation of a "mode of motion" of a particular system. The inverse transform.52). The MATLAB co~nmand [num. the Heaviside S merely is replaced by the Laplace s. 459-470). THE LAPLACE 517 TRANSFORM Carrying out the indicated matrix product.

The same problemof a second-order Laplace transform.23 (Section 6. 7.1 (entry #7) relates the residues and poles of general second-order transfer function or Laplace transform to a simple analytic expression of the correspondingimpulse response or function of time. The initial value theorem is lim y(t) = ~li~m~[sY (s)]. rather than the moredifficult determination the entire function. To see the contrast. y(t). y(o~).or 0+ according to the choice employedin the original Laplace .3. It relates nicely to the output of the residue operation of MATLAB. y(0). and a final value. EXAMPLE 7. however:there are two sinusoidal functions of time where only one is needed. the following examplehas the same transfer function or Laplace transform as Example6.2. ANALYSIS OF LINEAR MODELS. 465) finds the values 2rr = and 2ri = -5. 5sin2t)e Althoughall three answers are equal.23 (p.9 Initial and Final Value Theorems Sometimesyou maybe satisfied with a quick calculation for an initial value.518 CHAPTER7. The final entry in Table 6. p. Thesevalues also can be used as a partial check on the complete response. which are the most commonlypublished forms. are treated in yet slightly different waysin items #8. the versions with a single sinusoidal term reveal their meaningmoreconcisely.8°). A slightly different approach to the same problem is taken in the final entry in Table7. You have choices. 20 and 21 of Appendix C. f(t) = -2t + x/-~e-t si n(2t + 21. 18.PART 2 Appendix C.68. The approachin Example6. It suffers from one practical complication. 9.53) in whicht = 0. or subsets of it.1 gives f(t) = 3e-2~ + v~-~e-t cos(2t .7 Find the inverse Laplace transform of 5s2 + 22s + 39 2s + 12 G(S) -~ F(s) ~ + 2s+ 5)( s + 2) (s -~ + 4+ --’s+ 2 3 Thefinal entry in Table 7. 465). t~o (7. from which f(t) = 3e-2t + (2cos2t + -t. resulting in a single equivalent phaseshifted sinusoidal function. 19.1.2°).

] =0"4’ which agree with the x(t) found in the example. is behavior (no poles EXAMPLE 7. Another class of exceptions. The inverse Laplace transform of the transfer function itself is the impulse response of the model. (Signals involving pure delays are exceptions. which assumes stable outside the left-half plane).6. from examination of its Laplace transform. Software packages such as MATLAB dispatch most analytical drudgery.2. ramps and sinusoidai signals that commenceat time t = 0. The Laplace transfer functions for lumped models are ratios of polynomials in s.2. 519 THE LAPLACE TRANSFORM transform. The same is true for the Laplace transforms of most of the simple excitations signals considered herein.) In these cases the defining integrals for the transform and its inverse can be avoided by the expedient of table look-up.s+l)(s+5) x(0): )] :0.2. illustrated in Guided Problem 7. . Both problems with excitations and with non-zero initial values are treated by finding the Laplace transform of the desired output and then computing its inverse. A transfer function G(s) is algebraically equal to the function G(s) found earlier for the ratio of the response of the system to the exponential excitation est . This fact permits simple determination of the Laplace transformof a differential equation. The final value theorem. as noted in Section 7. using a partial fraction expansion if necessary.8 Apply the initial and final value theorems to the X(s) of Example 7. such as impulses. the Laplace transform of the response of a linear system with zero initial conditions equals the product of the transfer function and the Laplace transform of the excitation. In particular.4: 2 X(s) = s(s 1) (s + 5) Solution: 2 lim[( s-~ s+l)(s+5 2 x(~x~)=lim s-~0 (.4. steps.10 Summary Laplace transforms reduce ordinary differential equations to algebraic equations. Initial and final value theorems permit rapid evaluation of a time function at t = 0 and t = ~c. Multiplying a Laplace transform by s is equivalent to taking the time derivative of the corresponding time function. respectively. is developed for distributed-parameter models in Chapters 11.7. 7.

2.31) (p 510) and integrate. Suggested Steps: 1. Substitute this function for u(t) in equation (7.unlike the other Laplace transforms considered in this volumeof the text (except for the pure-delay operator). Use the delay operator sum of the two ramps. Find the Laplace transform of the function av~.5 This problemillustrates the application of the transform of a pure delay.e-bt).4 The purpose for this optional problem is to suggest that Laplace transforms apply to a muchbroader class of signals than is considered in this chapter.520 CHAPTER7. 3. Computethe Laplace transform of a single ramp. e -Ts to assemble the Laplace transform for the . The fact that G(s) is the Laplace transform of the impulse response of the system also follows directly from the superposition property. Notethat the result is not a ratio of polynomials.PART 2 The Laplace transform can be viewed as a Fourier transform modified so that it can treat nearly any transient signal. whichis one of the most com~nonlyencountered in engineering and science. Find the Laplace transform of the truncated ramp: f(t)l Suggested Steps: 1. Suggested Step: 1. Represent the function as the sum of an upward sloping ramp and a delayed downwardsloping ramp.3 Find the Laplace transform of the function a(1 . Guided Problem 7. 2. Guided Problem 7. Guided Problem 7. Lookup the integral that defines the transform in a table of definite integrals. ANALYSIS OF LINEAR MODELS.

7 This problem uses the matrix formulation.1. C and D. Set this up in two ways: as a forced response. Then. Find the Laplace transform of x(t) given the values x(0) = 1. Showthat the first of equations (7. equation using the formula 2. and note A from equation (7. Do not refer to the Laplace transform table in Appendix C.52) (p.52) gives the same result as it does for the forced response. and as an initial value problem. Suggested Steps: 1. A partial . Use the table to evaluate x(t). Suggested Steps: 1.6 You are urged to start your experience solving differential equations by the Laplace transform method with a straight-forward scalar initial-value problem. Substitute the function for £[F(t)] and the initial values as given above.2. 5. 536) for the responses of the three state variables to an impulse of force F(t) of magnitude I. Take the Laplace transform of the differential for dnx(t)/dt n given in Table 7. 3. 3. Note the existence of a repeated root. Consider the example d3 x d’) x 8 dx dt 3 + 5-~ + -~ + 4x = F(t). Solve for ~:[x(t)] 4. Solve the third-order system example of equations (7. p. For the initial value problem. fraction expansion allows the use of simpler tables. Use a partial fraction expansion to find terms in the forms given in Table 7. for F(t) being a unit impulse. 2. x(0) = 2 x(0) = -3.515).7. find x(t). Define the matrices B.90) (in following section. THE LAPLACE TRANSFORM Guided Problem 521 7. find U(s) and compute Y(s) from equations (7.to 0+.90). For the forced response.1. 4.89) and (7. find x(0) by integrating the differential equation over the infinitesimal time interval 0. and compares the procedure for a forced response with that for an equivalent initial-value problem. Guided Problem 7. Find the inverse transform of Y(s) using one of the tables.

PROBLEMS 7. 3. Multiply this by G(s) to get the Laplace transform of the desired response. 2. Find the Laplace transforms of the step and its response using the table in Appendix C. Verify this with the initial value theorem. A partial may help. A linear system gives the response 3(1 .8 This problem offers a comparison between the response of a simple system to a transient sinusoidal excitation and the response of the system to a steady-state sinusoidal excitation (or frequency response).522 Guided CHAPTER 7. Find the desired response from its transform. fraction expansion 4. t >_ 7.0 of amplitude 2. Suggested Steps: 1. Find the response of the syste~n to the sine wave 4sin(3t) which starts abruptly at t = 0.10 Find x(t) given ~ + 4k + 3x = h + 2u and u(t) = ~(t). 7. Problem ANALYSIS OF LINEAR MODELS. 7. Find the Laplace transform of the sine wave. The response cannot suffer a discontinuity at t = 0. The system is identified by its response to a step excitation. and interpret its magnitude and phase angle relative to the excitation. PART 2 7.11 Repeat the above problem given u(t) = us(t). k(0) . however.(s 2) 3 7. Find the result for t -+ ~x~. Note that the final value theorem fails. 5.10 of X(s) .0.12 The system ~ + 2k + 5x = 0 has initial Find x(t). Compare these to the IG(jw)l and /G(jw) the frequency response technique.14 Find the Laplace transform of F(t) defined by x(0) = 2.e-2t) when excited by a step signal at t -. Find the transfer function G(s) which is their ratio.13 Find the inverse transform conditions 2s"~ + 8s ~. and therefore must be zero there. since the response has no asymptotic value as t --~ ~x~. Compare the result for t -+ ~ that found from a steady frequency excitation using the frequency response techniques.

SOLUTIONS TO Guided Problem 7.3 GUIDED PROBLEMS U(s) = ~o °°a(1 .4 (p. 445). 7. t>_O 7.21 The square pulse of the above problem excites a system with the transfer function G(s) b/ (s + a) .18 Estimate the impulse response in analytical form for the system characterized in Bode form in Guided Problem 6. 0 7. t < = tsinwt. > ~ equation x(0) 7. 7.5 (p. 444). t < 0 O<wt<~r wt. = = sinwt.15 Find the Laplace transform of F(t) for a single sinusoidal pulse defined by f(t) = O. 7.19 Estimate the step responses in analytical form for the three systems characterized by Bode plots in Guided Problem 6.7. 523 THE LAPLACE TRANSFORM ~(t) = o.2. Find t he r esponse. Fi nd and pl ot th e re sponse. 0 7.e-~t)e-~tdt(~s 1= a l_~_~_~t~e_st[ + s +b ] Io =~a (1 1 s s~-b .17 A secoffd-order system with damped natural frequency w and time constant 7 is excited by a transient signal f(t) = -at.16 Find the solution of the differential k+ax=bsinwt.20 Find the Laplace transform of the square pulse plotted below.

¥(8)][~(~)] 2-3. Equation (7.(t) . b = -6 and check: 40 .89). Presumingall three state variables are of interest. Guided Problem ANALYSIS MODELS. r] =residue (aura. Fromequation (7.p.) Guided Problem 7.524 CHAPTER 7. ~ + dt 2 + -~ -= 5(t) [dx(s).2"~ .7 1. F(s) = -~-(1 .4 (s+ l)(s -+-2 a b c s+l (s+2) 2 8+2 4.D=0 .ur(t T) a/T 2.4 a U(s) OF LINEAR avffe-~tdt = ~s ~r (This result is given in AppendixC as transform pair #36. (x 3 + 582 + 8s + 4)X(s) (s 2 + 5s+ 8)x(0) + ( s + 5)~(0) + ~ 2 + 78 + 16 "X(s)(s~ +58+ 8)1 + (s + 5)2+ (-3) X3 "~.(6t + -2t 9)e Guided Problem 7. C= 1 0 .s~(0).88 ~.j or solve by hand =a=10 s2 + 78 + 16= 10(s + 2)2 + b(s + 1) + c(s + 1)(s + = (10 + c)s 2 + (40 +b + 3c)s + (40 +b + 2c) Therefore. Can use the MATLAB’command [c .~x(0) +s[sx(~).82x(0).5 1.-T~) e Guided Problem 7.90) gives A -2 1 .582 "~.~. For the first ramp. Fl(s) = 3. f(t) = (a/T)[u.(0)]+ 5 [dx(8). x(t) = lOe-t . c = -9.den).6 d3 x 5 d2 x 8 dx 1.~(0)]+~ [. as follows: L [ Is2 +7s+ 16] ~. PART 2 7.18 = 16 x(~) s+l10 (s+2)6 ~ s+29 5.6 .

c~ries out a p~gial fr~tion exp~sion. Thus:p~(t)=~_~ 16. pair ~aa ~d the time derivative of pair ~g ~e necess~y. la +4s+16)+(s+ 6 b=-~. ~he developmentbelow c~ries ou~ ~he p~tial fraction expansions by hand (which is relatively uneconomicalin engineering practice): sa + Ss~ + 20s +16 ~-~ s ~ + 4~ + 16 6= ~(s~ Therefore. c : .~ 16 _~ 9~ -2~ sin [2v~t . after which the ~sociated time funcgion can be found more readily. ~inding p~(t) the same wayis more awkward. Alter~mtively. Transformpair in the ~able of AppendixC ~hen gives p~(~) directly. 1 P-~ s~ +3s+8 8 .[u(t)] -Y(s) = H(s)U(s) -~ s 0 s+2 6 s 3 p(s) = ~ +5s~ + 20s + 16 -~ 0 = H(s)= ~(0) .tan-l(3v~/16)] 96 18 13 13 78 13 =6 .~)-~x(0) = same ~ ¯ he denominatorof the Laplace ~ransforms for p~ and pb can be facgored into ghe product of a first-order ~ermand a second-order term.~. a summagionof terms from pair ~a2. Y(~) = e(sI . ~he use of the MATLAB commandIt.7.2. THE LAPLACE TRANSFORM 525 u(s)= .because of ~he numerator term s~.p. r] ~re~id~ (~. 3 22 b = . s~+as+18= ~(s l)(bs+c) 24 c=-~+ [ ~+ s~l ~ +4s+160= 6 18 13 13 check: 6s+18 ] (s+2)2 +12 (s+l)(bs+c) : Therefore.

Alternatively. (t) Guided Problem e-st sin(2vf~t + tan-1 2x/~) 7.0 the origin. x(t) -~ 12 -~ sin " [3t-tan-l(~-)] The frequency response method gives ~0 uo]G(j3)~ = 4 ~ 12 = ~+22 = ~. ~ansform pair ~32 can be used directly. Ls~o [sX(s)] -. U(s) = s~ + 3~ .8 4x3 2. 36 b=-~. 324 2 x 72 468 (check) 36= ~+ 1~= 1~ ¢=tan-l( ~)=-tan-l(~ x(O) = Ls-~ [sX(s)] = 0 Note that the time function above agrees. 36 -2t 12 -) x(t)=~-~e. however. The result is an enhanced understanding of the behavior of linear models as well as the development of certain practical computational methods.3 Matrix ior* Representation of Dynamic Behav- The previous chapters and sections have made a restricted use of matrix concepts.1 (p 511). the following p~tiM ~action exp~sion (which could be found using MATLAB) can be used: k bs+c X(s) = ~ + s~ + Therefore. X(s) = G(s)U(s) 3~)(s + 36 3. In this section. PART 2 e-t - Finally. 12¢= AG(j3) =-tan-l( ~ x(~) = xo sin(3t + ¢). whicha~ees with the result above. c = 2 and w = 3. with a = 0. This is not x(oc). q(t) = -~p. . 7. matrix exponential solutions to linear models and the modal behavior of these models will be developed. despite frequent use of matrix notation and MATLAB commands. since there are two poles on As t -~ ~. as before. ~om Table 7. 72 c= ~.526 CHAPTER 7. The linear state differential equations are. ANALYSIS OF LINEAR MODELS. ÷~sin(3t÷¢).

7.58).. Again. It is important to note that regardless of the magnitude of At. on the other hand. which can be called a state transition matrix..55) with a linearly varying excitation U -~ 110 -}.59) . The general simulation schemes.57) 2.55) The special case of the stationary model. general simulation routines such as the Runge-Kutta method presented in Section 3. 2.Ult. 3! This equation can be used to evaluate the matrix exponential.1 The Matrix Exponential The general solution of equation (7. for which u(t) = 0. Mathematics texts give various ways for evaluating functions of matrices in general and the matrix exponential in particular.56) expan- e At = I + At +9~A2t 2 + IA3t3 + .5S) In this application. this process entails no inherent error (beyond round-off error). The second use employs a fixed time interval.. 7. but there are much better approaches that do not require evaluating many terms and do not suffer a truncation error. By contrast. Then. (7. (7. it is important to recognize two uses of the matrix exponential. After this. x(2At) is found ~he same that is by premultiplying x(At) by the same numerical matrix exponential.3. The obvious use is to give analytic functions of time for the various states in x(t). C and D are constants. x(At) is found from equation (7. is assumed.7 suffer an error that grows rapidly as the time increment is increased. .3.. x(3At) equals x(2At) premultiplied by the matrix exponential. assuming the matrix exponential is evaluated properly. for which A. which can be chosen arbitrarily: x(t + At) = eA~x(t). (7. is treated first. A particular method for finding the matrix exponential is developed below that also elucidates the behavior of the system.. It has the same series sion as a scalar exponential: (7.55) for the homogeneouscase u = 0 can written as x(t) = eAtx(0). B. The autonomouscase. the matrix exponential eAAt is evaluated numerically. First. At.3. 7.. in which eAt is known as a matrix exponential.3.. have the advantages of handling arbitrary forcing functions as well as nonlinearities. (7. y(t)= Cx(t) + Du(t). to evaluate x(nAt) for n = 1. MATRIX REPRESENTATION OF DYNAMIC BEHAIqOR* 527 dx -~= Ax(t) + Bu(t). This process may be repeated indefinitely.2 Response to a Linearly Varying Excitation Equation (7. however.

3. Substitution of equation (7. from carrying out the integrations: x(t) (7. This is the basis of the MATLAB function ls±m. gives in general [x(t)----~ixi(t)=~ieS~txi(O). / equation (7.60) ~T" The Laplace transform of the equation (7.61) The desired result follows =A-l(e At-I)Buo+A-2(eAt-I)Bul-A-xBult+eAtx(0).3. ] in which si is an eigenvalue that satisifies Ip(s) = det(sI - the characteristic A) = 0. Each term in this summation.49). The Laplace transform of equation (7.3 Eigenvalues.7 (pp. Note that the assumption of a zero-order hold corresponds to ul = 0.50) (p. This can be useful in treating systems that have sample-and-hold control signals.56). is given by equation (7. the inverse transform of equation (7. PART 2 freqnently is employed to represent linear interpotation of an excitation between known data points u(tk).63) assuming that the roots si of the characteristic polynomial are distinct.62) The accuracy of :~s±m can be seen to depend on the fidelity of the linear interpolation of the excitation between the beginning and end of each time step. and to represent them accordingly.59) U(s) uo + = --s (7. called a mode of motion. 349-352). For more information.64) [(siI .55a) (same as equation (7. in which t’ and t" are dummyvariables of integration. p. ANALYSIS OF LINEAR MODELS. satisfies the equation (7.65) .48). Whenthe excitation is constant or varies precisely linearly. The command:~s±m also has a feature that is not well documented: it attempts to identify inputs that involve zero-order holds (step-wise excitations). the use of which is described in Section 5. and its solution for X(s) is given by equation (7. but also is a potential source of error. there is virtually no error.528 CHAPTER 7.60) into this result gives the transform that has its inverse transform x(t) /o /0’// e At Buo dt ’ + eAt’’ Bul dt" + eatx(0).A)xi = 0. see the Help instructions that accompany MATLAB. 7. for example. (7.50). Eigenvectors and Modes Rather than giving the solution to the autonomous or homogeneous problem in the form of equation (7. (7.

. For a root repeated m times.69a) LQuJ V3 A~ = -R~C~ A~3 = O.69) correspond to equation (7. this generalizes to est [x~(0) + V~lt + v~.63) are unchanged. which has three independent energy storage elements and thus is of third order. Q~]r somehowmust be removed. 7. consider the system shown in Fig. The eigenvalues customarily are assembled in a square diagonal eigenvalue matrix. Equation (7. Vao can be chosen such that they represent equal levels of water in the three tanks. without interaction with one another. S: (7. ~27). (7.68) The other terms in equation (7. + Vimt’~]. (7. as shown. " Aa~ = O.3. The volumes ~o. the Q = [Q~. Q~. Q~o.. . (7. This is accomplished by defining constant values ~o = [Qm. depending on the initial conditions. ~o.~t2 +. 7.8.4 Case Study: Three Fluid Tanks As a case study. ~o.63) reveals that the individual modes of motion can be superimposed in any proportion. equation (7. . gives the state equations d V2 dt V3 An = --~ + =A V2 + .67) where v~(0) is a vector of constants. ~o] r such that ~hus all changes in time reside in the & terms.66) If one or more of the roots of the polynomial are repeated. MATRIX REPRESENTATION OF DYNAMIC BEHAVIOR* 529 and has its corresponding eigenvector.7. x~.69b) To make equation (7.63) produces fewer than the requisite number of terms. Aa~ = R~C~ Rb¢ C~ -~ .~) (p.3. It can be very instructive to grasp the significance of the individual eigenvalues and eigenvectors. Applying integral causality in the usual manner. (7. For a double root s~ the associated terms are eSt[x~(O) v/ t]. Qao]r and Vo = [I%.

ANALYSIS OF LINEAR MODELS.0050 frS/lb Cb = 0. PART 2 a3-~ Rc I water: pg = 62.312 ft 3 A~ Co = ~-ff-= 0.0 lb min/ft (a) system and parameters l Vl/C 0 Co R~ S~2 Ra.0100 ftS/lb Rc = R~c ~ = 50.8: Three-tank system Rc .248 ft 3Ac = 0.4 ~ lb/fi 3Ab = 1.0 lb min/ft ~ R~ = R~ = Ra~ = 100.624 ft A~ = 0. R~ R~ (b) bond graph with causal strokes Figure 7.530 CHAPTER 7.0200 ftS/lb Cc = 0.

6 time.. O0 0. V~/V.8 1.2 1.--_.first experiment andmode ~% 4 3 2 . op-.2 1. third experimentand mode 2.0 (d) response.2-..2 1. MATRIX REPRESENTATION OF DYNAMIC BEHAVIOR* 1 [~. secondexperimentand mode 2 ~ V..v/vl’ .8 1.6 time.4 I I I I I 0..6 2.-. _~l . minutes 2.8 I I I I I 1. 0 5 0. minutes (c) response...2/Vii 1 0 -1 -2 ~" 0 I I 0... minutes (e) response..0 531 .4 I I 0.3.0 time...~V.4 0. .-..7..-.

73) The immediateinterest is the special case of AQ= 0.72) The equilibrium equation 5nowcan be subtracted from equation (7.25. The roots of the characteristic polynomial therefore are 81 : -4. systemthe three time const~ts ~e 0.20 minutes. Theessential characterof the dyn~icscanbe seento be representedpartly by three different exponentialdecayrates. wherev hasthe dimensions of time and is called a time constant.532 CHAPTER 7. The solution can be found from equation (7. PART 2 as indicated in part (a) of the figure. Thesolution to equation(7. .h AV + AQ. howdo the three levels approachtheir equilibrium with its equal depths if they start with different depths? The numerical parameter~sin the figure give the matrix A and characteristic polynomialalready used in Example7. Q. The corresponding depth of the water is V. dt (7.1 + LtS()J LV LSJ. Equatio~ (7. In the present." however.oRa = Q~oRb= Q3oRc.73) introduces no additional error.69) to give the desired result dAV .".For simplicity it is assumed herethat the first elementin the normalizedeigenvectoris unity.. .respectively.75) also is characterizedby three eigenvectors given in the squarebracketson the right side. (7. which is proportional to ~’~o/Ca = V2o/Cb = }~o/Cc.5 (p.as noted before.63): I-1. ~ discussed in Section 5. Decayrates of linear models. and ~ a result the model of equation (7.69) is "linear bi~ed.) Thenormalizedsecondeigenvectorof SThe approach is consistent with linearization. Thus. 1. s2 = -1.4. ANALYSIS OF LINEAR MODELS.Q2o= ~Q3o.o/Aa = V~o/Ab = V~o/Ac. 516).74) The order of these roots is arbitrary. (7.0 and0. Eacheigenvectorcan be factored into a product of a scalar and a normalizedeigenvector. (Later it will be seenthat MATLAB makesa different choiceto get aroundthe possibility that the first elementis zero.71) 0 = AVo + Q0 (7. In particular. or 1 Q10---. traditionally are expressed by e-~/~. s3 = -5 min-’.

V2(t).V21/VI~ V22/VI2 V231V~a = L v31/v~l va2/vl~. you removea bucket of water of volumeVll from the third tank and dumpit into the first tank. -1]. but affords an opportunity to check the results.76) The first of the three equations imbeddedhere requires V22(O)/V. Note that the entire responseinvolves only the first time constant. the three tanks are at equilibrium.78). the vector [V1(t. leading to the simple result for the x(/) of that problem. MATRIX REPRESENTATION OF DYNAMIC BEHAVIOR* 533 the three-tank problem. This establishes the initial condition yl~. 528).3.insparallel to its initial direction in state space. Repeating the sameprocedure for the two other normalizedeigenvectors and assemblingthe results into a modalmatrix. Va3/VIa 6 -2 .8 part (c). V3(t)]T rema.~e-4t + V. The second is redundant.79) which. 7. for example.2(O) The third then gives V32(O)/Vl~(O)= 2.77) Withthis result equation (7. Suddenly. respectively.~e -~ + --2 V~3e-SL (7. This same initial condition was given in Example7.80) whichis plotted in Fig.5 (p. This is the behavior of the first mode.75) can be written V~(t) = V3(t) V. gives P-. Initially. 1 2 2 (7. and decays at the rate e-4~ dictated by the first eigenvalue. as follows: = 0. s2 = -1. 516). all at the samelevel with the rate of flow out of each equal to the rate of flow in. showsthat V12and V13equal zero.).Vll 1 (7. P. 0. into equation (7. can be found by substituting the second eigenvalue. whencomparedwith equation (7.64) (p. (7.78) 1 Imaginethe following experiment. The subsequent behavior must be vdt)/ = Va(t) e-4t . The cancellation of the numerator and denominatorfactors of X(s) is nowexplained. and the ratios of the three excess volumesalwaysretain the proportions [1.7. . y2(0)/ = v3(0) 1 (7.

83) and the resulting z(t) into equation (7.84): each component zi(t).2(t)= v..78) can be generalized (7.534 CHAPTER 7.85) z(0) = P-ix(0).-1/si. . n.82). 82... six units into tank 2.83) and (7.82) represents tr ans]ormation of variables bet ween z(t ) and x(t) The variables z(t) are very special.56) (p. gives the final general result x(t) (7..87) l e*’ = PEstP-’.j ..: e t (7. l/~(t)] T remains parallel to its initial direction in state space. Equation (7.z(0).83) ~ .82) z(t) = Es. the resulting behavior would be -~. PART 2 If as a second experiment you suddenly dump one unit of water of arbitrary volume V~2into tank 1. ANALYSIS OF LINEAR MODELS.81) which is plotted in part (d) of the figure.. ]v~:~ = y(0) [ ~]~j ’1 16 0 a -3 ~:(0)/ t~(0). Inverting equation (7. Est = (7. 0 .12e (7. i 1. and two units into tank 3. s2t e . as revealed by equations (7. 527) reveals that the matrix exponential can be computed by evaluating (7.. ¯ . [ V~ ~ .84) ¯ Equation (7. A comparison of this result to equation (7.86) = PEstP-~x(0). v.82) and setting t = (7. and behaves as directed by the second eigenvalue.-. employing only its unique time constant. equation (7. The f~test decay would result from a third experiment thut estnblished conditions for the third modeof behavior. ] For the three-tank problem. Substituting this result into equation (7. that is the decay rate e-t. the vector [t~ (t). ~-i -. This is behavior of the second mode.85) gives .. Thus z(t) is the set modal variables. which is plotted in part (e) of the figure. V2(t). act s independently of the other components.

3.9.5 Case Study With Complex Roots The three-tank problem is rather special in that its roots are all real.3v2(0) + 6v3 (7. 24 L 6v1(0) .88) whichcan be substituted into equation (7. 7. The methodabove is applied next to a problem with complexroots. which has three independent energy storage elements and is thus of third order.78) to give the general result for that problem.3. Applyingintegral causality in the .9: Third-order system example [ 16V1(0) +0.7. 7. MATRIX REPRESENTATION OF DYNAMIC BEHAVIOR* Ra Se 535 C= 1 ~"=1/8 (L~ Ia~’:~-’~ 1 p~/l~ 0 "--’--~ y~ T pb/l~ q/C~ ~ C R~ I~ = 1/6 Ra=Rb=I/3 (a) bond graph 2.0 (b) responseto impulseor initial condition Figure 7.SV3(0) = 1/2v1(0 ) + 3v2(0)2V3(0) . Consider the system shownin Fig.

93) The first of the three equations imbedded here requires x~3 = -1.94) and the inverse of P is _ 1 -jv/-~ 26 +jx/~ 6.5 seconds. but affords an opportunity to check the results.jv/-~ 6+jx/~ -1. s~ = -1.90) [oo equation (7.89) The numerical parameters in the figure give A= so the characteristic 0 -2 -8 -6 (7. ANALYSIS OF LINEAR MODELS. (7.536 CHAPTER 7. and setting xl~ equal to unity: 0 8 1 6 -1 x12 2 LX~3 = 0. respectively.91a) or p(s) =_ 3 +5s~ + 20s + 16= (s + 1)( -~ + 4s+ 16)= 0. The roots of the characteristic (7. these eigenvalues tell you the essential character of the dynamics of the system. p_~ _ Adj(P) det(P) 1-2- j/v~ 2jx/~ .92) where j is the unit imaginary number. Wd = 2v~ rad/sec.2jv~ -l+2jv~J (7. The second then gives x~2 = -1. as represented by the decay rates e-t and e--~t with their time constants of 1. (7. Even if you don’t continue the analysis further.95) . (7. as shown. into equation (7. PART 2 usual manner. Youfind the first normalized eigenvector by substituting the first eigenvalue.91b) polynomial are s2.0 and 0. gives the state equations -dt = yRb/I~ --1lib -1/Ia 1/C -1/RaC (7. (7.3 = -2 ¯ 2jv~. The third equation is redundant.64).65) becomes 0 -1 s + 2 -1 6 s+3 i p(s) = 0. and the damped natural frequency. -1. Repeating the same procedure for the two other normalized eigenvectors and assembling the results into the modal matrix gives P= -1 l + j/v~ -1 -2 + 2jv~ The determinant IPI is -2jx/-~.

For most practical situations above the order two.73) over the infinitesimal time interval from t -. you find the matrix PEst. as can be seen by integrating the differential equation (7. you could forgo computingthe first and third columnsof p-l.0 to t = 0+.96) Equation (7.96).jx/’~ ¯ + jx/SJ (7.99) are desired.±nv(P). MATLAB can relieve the analyst of most of the drudgery above.7. Whenanalytical results such as eqtiation (7.e-2t[6 cos(2v~t) . (7. s~t 2jvr~)e (7. recommended.9.97). however. it is possible to avoid having to deal with the complication of complexnumbers. 7. you can verify that they satisfy the initial conditions given by equations (7. .14~/~sin(2v~)] (7.3.) Next. consider as the initial condition the result of an impulse of the force F(t) whichoccurs at the time t = 0.jv/-~)e jx/-~)e (-2+eiv%t + (7 .jx/~)e [ = Z__ ] 12e 26 L 12e-~ . saving significant effort if hand calculation is used. Whennumerical as opposed to analytical results are satisfactory. the calling of the MATLAB inversion command. As a partial check. the separate columnsof which represent the separate modesat time t: eSlt *’t PEst = -e s’t e eS2t (1 + j/v~)e ~ (1 (-2 + 2jv~)e *~t (-2 - es3t j/v~)e ~3t . where Z is the magnitudeof the impulse or time integral Z = fF dr.(6 . The immediate effect of the impulse excitation appears only on the state variable p2. pb(0+) = Z. q(0+) = 0. whichby definition includes the entire impulse. An optional methodis presented next. given by equation (7.98) Multiplying the vector of modalamplitudes.85) nowgives z(0) = p-1 = ~o .99) These results are plotted in part (b) of Fig.97) (Note that if you knew in advance that only pb(O+) would be non-zero. This time integral is a step of amplitudeZ and can be designated as Zus(t).(6 + 14jx/~)e(-2-~jv%t J z ~ -~e-t + e-*t[~ cos(~t/+ = i~ 6e-t-t + e-~t[Tc°s(2~/~t) v~sin(2~/gt)] . this matrix gives x(t) = PEstP-~x(O) (-2+~jv%t + (6 + jv~)e(-2-2jv~)t (-2-~jv%t ] [ -12e-t -t + +(7 (6 +. The force is designated as ZS(t).14jx/~)e(-~+’~jv~)t . The result is p~(0+) = 0. y(0). 6e . MATRIX REPRESENTATION OF DYNAMIC BEHAVIOR* 537 As an example.

3. In general.86) (p matrices as follows: x(t) (7. on the other hand. with the signs of the second eigenvector.100). . saving effort. For example. handles complex eigenvectors and modal and other matrices directly. This modified modal matrix is readily inverted to give p~l 13118 1 3vcg =-5 -66 ~ -11 .103) In general. and the terms for each complex conjugate pair appear as a 2 × 2 array. Modified Method ANALYSIS for OF LINEAR Complex MODELS~ PART 2 Eigenvalues* The complex conjugate nature of the terms in the modal matrix permits them to be coded with real numbers.~t = e -2t cos(2~/~t) -e -2t sin(2x/-~t) e -2t sin(2v~t) e-2t cOs(2V~t) (7. (7. the modified modal matrix for the third-order system above is Pm = -1 1 1/ -1 -2 2~/~ . so it is reproduced without change from equation (7.101) ~a4) can be modified to apply to these = P. and given in the form e-st coswt e -st sinwt -e -at sinwt e-c~t coswt where a is the real part and w is the imaginary part of the associated eigenvalue. are placed in the third column.102) to the third-order system now gives the modal decomposition and the function x(t) relatively simply. The key result of equation (7. (7. complex eigenvectors are paired off.100) J The first eigenvector or column is real.6 CHAPTER 7. The second and third eigenvalues are complex conjugates one another.538 7.~EsmtP~lx(0). and computes only the inverses of real matrices. for the present problem. centered on the diagonal. The modified form has the advantage of giving analytic solutions directly in terms of real functions. the exponential terms for the real roots appear on the diagonal.102) where. the real parts being placed in one column and the imaginary parts being placed in the adjacent column to the right. their commonreal parts are placed in the second column and the imaginary parts. Application of equation (7. Somesoftware automatically reports eigenvectors in this modified form. the matrix Esmt is Es. further motivating its use. MATLAB.

3. The equation of motion becomes d y2 pc = ~ Llk~ 0 1/m -L2k~ L2/J 0 I y2 pc (7. consider the vehicle dynamics of Figs. (The dampingis reinstated in Section 7.843 m (c) heave mode Figure 7. 322. 5.3. 324).) The reduced model shownin Fig.7 Case Study: Vehicle Dynamics As a final problemregarding autonomousmotion.10: Undampedvehicle modes 7.7.3. MATRIX REPRESENTATION OF DYNAMIC BEHAVIOR* 539 (a)system (b) pitch mode ¯ ¯ 4.16 (pp.1o4) .15 and 5. 7. which has a fourth-order modelwith two degrees of freedom.11 below.10. The excitation (the toad bumps) is removedto make autonomous.and the dampingis removedto clearly reveal the modalbehavior.

si. Werethis not true the state variables themselves wouldbecomenon-real.0008[ 32000 s 51200 0 s0J’ (7. and note that the second and fourth eigenvectors must be the complex conjugatesof the first and third.108) (7.96s2 + 3456= (s 2 . however.0008 si 32000 s~ 51200 0 -0. P3i. Therefore.3274j ’ 15853j (7.110) which. Equation (7.88)(s2 . P4~]T satisfies 0 30000 -42000 -0. L.540 CHAPTER 7. 008 si P2i /p3. P2i.107) (7. whichis impossible./ = O. (7. kt = 30 kN/mand k2 = 32 kN/m. whereuponequation (7.113) .8047 .8047 -3274j -]5853j 1 " -1. whensubstituted into the third equation.109) J LP4iJ The fourth scalar equatio.4 = ±8. L1 = 1.2 = 1.5195 P = | 6724j -6724j L-2221j 2221j 1 -1.0008 -0.1o5) and p(s) = I ]sI. The resulting modalmatrix is P3i I = 1 1 0.934j.478j. s o 0 30000 -42000 sI-A= -o.2 = ±6.n embeddedabove gives P2~= (42 000 - s~m4i)/51 200. respectively.6 m.112) gives third.48.08) = 0.112) 1250si + 7~m4i.110) gives the second element and (7.5195 0. PART 2 The parameters are given as m= 1250 kg.4 tn.106) or p(s) = s4 + 119.111) The first equation gives (7.ooo80. (7. J = 2mr with r 2 = 1. It is simpler to computeonly the first and third eigenvectors this way.6 m2.71.0007] s -0. (7.A = s4 + --1 m 1+k2+ L~kl --2+L~_k2~s + 2 r m2r2 (L~ + L2)2 = 0.111) can be used to find the fourth element in all four eigenvalues. yields P4i = -(1250s~ + 56 250)/(0. ANALYSIS OF LINEAR MODELS. s3. (7. Theith normalizedeigenvector[1.25s{).

21513 6. from which ejwlt Est -: [i 0 0 jw2 0 0 ] 0 0 ’ w2=8.3882 0. (7. This requires that a node (point with zero velocity)be located between the two points. with a 0° phase difference.43026 0 ’ 0 .21513 -6. the second complex conjugate pair of roots combine to give a mode of motion in which the amplitude of y~_(t) is 0.10 part (b). MATRIX REPRESENTATION OF DYNAMIC BEHAVIOR* 541 and the corresponding modified modal matrix is pm = O.8047 times the amplitude of y~ (t).114) These matrices reveal that the first complex conjugate pair of roots combine to give a mode of motion in which the amplitude of y2(t) is 1. 534) (7.117b) -6 0 0 1. with a 180° phase difference. as shown in Fig.478 rad/s. as shown in part (c) of the figure.117a) -5 0.5195 times the amplitude of y~(t).22352 -0.961j × 10 -1.2875 x 10-~J The matrix exponential caa be found ~om either equation (7.87) (p.4376j × 10 -~ -5 0.9753j x 10 2.3.8047 0 0 -3274 " -15853J (7.9753j x 10 -2.3922× 10 -0.9526jx 10 -5 .118) .3882 0.961j × 10 1.4376j × 10 0 -0. 7.11176 -0.77648 0.43026 0 0 0 -0. 95 0 6724 -2221 -1. suggesting the designation pitch mode.~E~tP~ ~.7.9526j x 10-5 ] -5 -5 0.5905x10 P~I = 0.sinw~t cosw~tJ The inverses of the modal matrices are -5 ~0.115a) (7. 538): e ~t = PEstP -~ = P.116b 0 0 cosw2t sinw_~t " ] [ coswlt sinw~t 0 0 .21513 0.1951×10 -4 -4 -0. (7.102) (p.21513 -0.934 tad/s. 0 -j~lt e 0 0 0 0 j~2t e 0 0 ] 0 0 e-j°~2t (7. (7.116a 0 -sinw~t cosw~t 0 O0 Esmt = (7. suggesting the designation heave The eigenvalue matrix is jw~ 0 S = 0 0 [ 0 -jw~ 0 0 6. This requires that a node be located at some distance from the vehicle.115b (7.11176 -0. Similarly.

5 1 1.8sinwlt5221sinw2t Pm pc i) / J L 3543. 0 0. consider as an initial condition yl(0) = 1. The. PART 2 4 4.0208 0.0332x10 -4 -0.seconds 3.04623 0.5sinwlt + 1724.6sinw2t /’ J which is plotted in Fig. coarseness of the time interval itself produces no error in the calculation.4 605.74506 eo.68565 -2905.7436x 10 0.pre-multiplication . Successive . pro(O) pc (0) = 0.04333 -2788. y2(O) = O.~A = I 0. =/-°.120) 731.7263 x -4 10 0. .9 0.5 3 time.4°338c°swlt (7.of the state vector by the constant matrix exponential above gives the highlighted points.66 -4 0.11: Motion of undamped vehicle This square matrix can be used either to find analytic functions of time in responses to given analyti c excitations.0.3 L 1104.40338cosw2t[ y2.542 CHAPTER 7. or it can be used as a state transition matrix to compute the responses at times nat numerically. 7.119) Further. As an example of the latter.7622 -4 -0.6685 J /" (7.5 5 Figure 7. solution + 0. consider At ---.5 ANALYSIS OF LINEAR 2 2.1 sec. This gives the analytic.6265x 10 -0.7037 x -4/ 10 -0.5 MODELS.11. This gives 0.

0000 -0.0000 0 0 0 -5. It is simpler. l*~).6667 -0. Equation (7. . the modal matrix corresponding to equation (7. Finding this result is expedited by determining the vector z(0) from equation (7.105) (p.2339 + 0. 533).2611 -0. 534) allows the determination of the analytic response as a function of t.85) by invoking z0=±nv(~i)*x0.108) is found by typing z0=£nv(P)* [0.The procedures are illustrated here by application to the three problems above. the sum of the squares of their elements is 1. The diagonal matrix es~ can be found for the specified value of t = 0.0000 0 0 0 -1.78) (p.86). merely to enter expm(. It accomodates the use of a zero value in this row. 1.1637£ + 0.3.1666 -0.0000 The three eigenvectors are normalized to unity length. in which the elements in the top row were arbitrarily chosen 1.7. 540) but normalized to give real unit tors is given as 0. Once the matrix A is entered. the commande±g(A) then gives the eigenvalues. The response is the modal matrix 0.1 . however.99) (p 537) would be directly in the first or complex form of the equation. you enter [P.0675± .87) can be found by entering P*ES*£nv(P). standard notation is used.9370 -0.1 by entering ES = ex!0ra(. 7. Note that expm(X) returns the matrix exponential x.0].0.(p.9044i The normalized vector z(0) of equation (7.0.1637i .5774 -0. For example. but would have entailed extra calculation if done manually.6667 -0.0.1666 -0.5774 -0.7071 -0. Construction of the analytic solution of equation (7.2339 . This is different from the modal matrix of equation (7.3333 0. The corresponding matrix exponential of equation (7. whereas exp (X) returns with = f or e ach of t he e lements i n X. MATLAB treats systems with imaginary or complex eigenvalues and eigenvectors the same way.1562 -0.9044± -0.S] = e£g(A). To get the modal matrix as well as the eigenvalues.7071 -0.0675i + 0. amongother advantages.S). The initial state x(0) = x0 must have been defined first.3.8 MATRIX REPRESENTATION Application OF DYNAMIC BEHAVIOR* 543 of MATLAB Eigenvalue problems are simply treated with MATLAB.3123 and the diagonal eigenvalue matrix -4.2611 -0.5774 -0. that is.

In thiscase. for example. taken from x(t) = uoIG(jw)l cos(~t +/~ + ¢) (7. t(k) = 0. B~ C.120).1 and 6. 397). Thispermitsthe timeinterval for the five-second run aboveto be reduced to 0. theasterisk should be leftoutof theplotstatement.55) (p.l) = [1.0007. The highlighted points in the plot of Fig.3. 7.005seconds. 0 .544 CHAPTER 7. ’*’) Notethatonlythefirsttwoof thefourelements in x arechosenforplotting. seconds’) ylabel (’yl (blue) and y2 (green)’) plo’6(t. the response.x(l:2. This is an advantage for finding the motion of the vehicle as expressed in equation (7. t(1) = for k = 2:51 x(:.9 Response to Exponential and Frequency Excitations The response of scalar differential equations to exponential and to frequency excitations was addressed in Sections 6. 7.11 can be computed through the repeated use of the numeric matrix exponential of equation (7. x(t) ~ G(s)uoe For frequency excitations equation (6.0 42000 -51200 0 0]. taken from equation (6.2.122a) ¢(w) = tan-l ~[G(jw)]}~(G[jw]" (7. with the only changes being that the scalar transfer fuction . :).1 * k. The studentversionof MATLABpermitsindividual matrices to have up to 8192 elements. end xlabel (’time.k) = ES * x(:. The points and plot can be programmed in MATLAB as follows: A = [0 0 . 527). but might be preferred as a meansof finding analytic results directly in the form of real functions.119).39) (p. D matrix formulation of equation (7. ES = expm(O.0].25) (p. PART 2 The use of the modified modal matrices with real elements is not necessary.-30000 -32000 0 O.122b) These results can be generalized to the A.0008.0. x(:. 410). st.0008 -. the response. morethanenoughto givevirtually continuous curves.ofcourse. ANALYSIS OF LINEAR MODELS.l* A). For exponential excitations of the form uoes~.0.121) of the form uo cos(wt +/~).0008 . (7.k-l).

4 (You might wish to verify the added terms by placing bonds Se F1 and Se F2 on the respective lojunctions for ~ and ~2.000 51.104) except for the addedforcing terms: d y2 =A Y2 PC PC whereA is given by equations (7.104) and (7. plot the magnitude of the transfer functions betweenthese forces and the two displacements.A)-IB = p~ 0 s = ~1 Adj p(s) [ 30.0007 -0.4 (s I .) Since all four state variables are considered to be the output variables of interest. 545 MATRIX REPRESENTATION OF DYNAMIC BEHAVIOR* or Laplace transform G(s) is replaced by the array of transfer functions H(s) of equation (7. is excited by the force the forward axle and the force F2 on the rear axle. Example 7.7. F1= Flocos(~t).52) (p.105) (pp. and carrying out the standard procedure. The matrix transferfunction H(s) becomes.i 00 32. C -. L-42.52) (p. and s becomeseither the coefficient of the exponentor the frequencyjw of the excitation. H(s) = G(s) 1 [Adj(sI- A)]B -0. where F2= F2ocos(~t+ ~). 539.0008 s 0 0 s 1 -1.105).0008 -0.000 ~.I. using equation (7. Solution: The equation of motion is the same as equation (7. 540). y2.0008 0. D = 0 and H(s) = G(s). Further. Find the matrix transfer function H(s) that relates the state variables yl. with parameters given by equations (7. -/51 /52 -1.104) and (7.200 " 1 . Pmand pc to these forces.7.3.3. 515). o1 = o .9 The undampedVehicle discussed in Section 7.

206j).478 and 6.88 + 638 / ’ L -1.968~ + 3456.0020882+ 0. ¢~j is :1:180°. Oij =/ Hij (jw). resulting in a real number. The resulting displacements are y2(t) = IH~IF~ocos(wt + Oe~)IH~.0 10 frequency.0017882+ 0. 540) as the imaginary eigenvalues. Hij = Hij (jw).000328 ’~ ] -0.IFzo cos(wt + ~ + 01e). (Thereis no difference between +180° and -180°. A phase angle jumps by 180° wheneverthe frequency passes through either a pole or a zero.88 1.546 CHAPTER7. since there is no damping.483 .6s3 + 728 J ’ p(s) = 4 +119. the phaseangle ¢~j is zero.108) (p.If this numberis positive. and need not be found.206 rad/s (s = =t:7.A)-1 or Adj(sI .934j and s = =t:8. if negative. .045 rad/s (s = +8.A) do not affect the result. rad/s 100 The two natural frequencies. ANALYSIS OF LINEAR MODELS.0003282 1 | 0. which were found in equations (7.045j) and of G2_~are at co = 7. The result is the 2 × 4 matrix F0.76.934rad/s. The zeros of GI~ are at w = 8.) The magnitudesof IHI~I and IHell are plotted in Bodecoordinates below: -100 . because of the zeros in the matrix B. PART 2 The first two columnsof (sI .478j.108| H(s) = p-~ s3 +76. Only even powers of s appear in the numerator and the denominator polynomials of each Hij.1152 -0. The eigenvalues or poles are s = +6. appear as resonant frequencies with infinite amplituderesponses. w = 8.

The Bode plots comprising the new magnitude ratios ¢ij for i = 1.3.0144s + 0. MATRIX Example REPRESENTATION OF DYNAMIC BEHAVIOR* 547 7. Hmand H22 are 1 [ 0.2k2 1/m -L1/mr’2 2L2/mr 1/m | -(L2R~ .1152 .R2 + L~R1)/mr 2 J With the given values of the parameters.L1R~)/mr 2 | " -(R1 + R.2)/m -(L2R2 . Solution: The matrix A is modified to become 0 A = 0 -kl Llkl 0 0 -k2 -L.o into the model of the vehicle above.(L~. rad/s 100 phase 1.0.A = 30 000 32 000 s + 6. Determine the modified matrix of transfer functions for the special case R1 = R2 = 4000 N s2/m.0144s + 0.7. this becomes ’ s -0. to make it more respectible.04 from which the values of HI~. db l -150 1.56s ~ + 892.0008 sI .0 10 frequency.4 ’ s ÷ 9.44s 3 + 177.00032s~ ] H(s) = ~ L -0.o020ss2 + 0.4 [_-42 000 51200 0.0008 0.8s + 3456.00032s2 o.10 Introduce the damping resistances R1 and R.0 10 frequency. and make Bodeplots of the positional responses to the force on the first axle.0.1082 J ’ p(s) = 4 +15.64 -0.L1R1)/m .rad/s 100 .00178s~ + 0. 2 and j = 1 are as follows: IHij[ and phase angles -5O gain. H~_~.

045=t: 6. 7.559 -1. Thevarious scalar transfer functions still have zeros.7. s -. also. The transfer functions still have poles. For example. i.4 = -4. Solid arrows are drawnfrom the poles to the point on the imaginary axis.10 Representation in the s-Plane It is customaryto plot the poles and zeros of a transfer function in the s-plane. Now. s3.12: Pole zero plot for H~(s) for the dampedvehicle . whichcorrespondto the eigenvalues sl. Oneapplication for an srplane plot is to visualize the response to an exponential disturbance u = uoe"~. 7.e.jw.2 = -4.2 = -3. also. The resonances are sufficiently dampedto be almost unnoticeable. fromthe zeros to the same point. whichrepresents the particular sinusoidal disturbance of interest.954j. whichhas the real part of s as its abcissas and the imaginary part of s as its ordinate. -jl0 IRes.548 CHAPTER 7.139j. which is the convention for such plots. Dashedlines are drawn. the numeratorof Hll vanishes for s = zl. Such a plot is given in Fig. This is illustrated in the plot for the special case of frequency reponse. ANALYSIS OF LINEAR MODELS. whichare either real or complex.161=t: 6. Poles are indicated by x’s and zeros by O’s.3.12 for the H~ of the exampleabove. PART 2 Theseresponses are clearly preferable from the viewpointsof comfort and safety.179j.the polynomials in the numerator and in the denominator of the transfer function can be factored into products of first-order polynomials: Im. s --jlO Figure 7.

7.3.

MATRIX

REPRESENTATION

OF DYNAMIC BEHAVIOR*

(8 - zl)(s z2)
Hll (8) = (8 - 81)(8 - 82)(8 - 83)(8

549

(7.122)

The solid arrows, then, represent the denominator factors, and the dashed lines
represent the numerator factors, both with s = jw. Considering the arrows
as vectors 1.~, I/~,...,lt
as shown, the magnitude and phase angle of Gu (jw)
becomes

ZHu(jw) = Z~ + Z~ - AV~ - Z5 - AV3 - AS.

(7.123b)

It is possible, then, to determine the magnitude ratio by measuring the lengths
of the vectors and multiplying and dividing as indicated, and to determine the
phase angle by measuring the angles of the vectors and adding ~d subtracting
as indicated. The frequency w c~ be changed and the process repeated.
It is more efficient to compute the transfer function algebraically, and MATLABdoes it automatically. It can be very instructive, nevertheless, to be able
to look at an s-plane display of poles and zeros and visualize the form of the
frequency transfer function.
7.3.11

Summary

An unexcited linear system with nonequilibrium initial conditions exhibits behavior that can be decomposed into a sum of simple modal behaviors. The
dynamics of each mode is represented by its eigenvalue. The shape of each
mode, that is the proportions of its component variables as represented by its
eigenvector, remains fixed in time. Thus the contributions of each mode to the
total behavior depends on the initial conditions. Standard matrix methods for
treating the details have been presented and illustrated.
Whencomplex roots
are present, you can choose to use either real or complex matrices. MATLAB
accomodates complex matrices, whereas certain other software packages use real
matrices only.
Initial value problems can be addressed in a more routine if somewhat less
insightful way using Laplace transforms, as presented in the previous section.
Cases with repeated roots are best handled this way, so details of their modal
behavior have been omitted here.
The simplest approach for finding the autonomous or forced behavior of a
linear system employs the numeric matrix exponential. These solutions obscure
modal behavior, but particularly
in heavily damped systems the modal decompostions may be of minor interest.
MATLAB
allows the matrix exponential to
be requested directly, or indirectly through the use of the commandlsim which
treats excitations that vary linearly in time over each time interval.
Exponential and frequency responses for a model represented by a matrix
transfer function are found in essentially the sa~ne way as a model represented
by a scalar transfer function. The matrix transfer function is simply treated as
an array of scalar transfer functions.

550

CHAPTER 7. ANALYSIS OF LINEAR MODELS, PART 2
R= I0~
I = 200 ~h
C = 50/~f
E= 5.0V
Figure 7.13: Guided Problem7.9

Guided

Problem

7.9

The first guided problemdeals with a second-order system, and can be treated
analytically without unduedifficulty so as to reinforce the key concepts.
The two-wayswitch in the circuit in Fig. 7.13 replaces the battery at t = 0
with a conductor. Find the subsequent voltage e(t).
Suggested Steps:
1. Represent the after-switch circuit with a bond graph.
causality and define state variables in the usual way.

Applyintegral

2. Write the state differential equation; find the matrix A.
3. Solve IsI- AI = 0 to get the characteristic polynomial.Solvefor its roots.
4. Writethe general solution for the charge, q(t), of the capacitor.
5. This problem is simple enoughthat you can avoid having to evaluate the
eigenvectors and modalmatrix, etc. Instead, specialize the result of step
4 by applying the knownboundary conditions q(0) = q(~)
6. Find the remainingConstant by noting that the initial current through the
inductor is E/R and p(O)/I, and p(0) is related to q(t).
7. Sketch-plot the resulting e(t) = q(t)/C to see if it makessense.
Guided Problem 7.10
This probleminvolves a fourth-order, ~wo-degree-of-freedomvibration model.
It is treated analytically, and therefore maybe too ambitious for your needs.
The beating phenomenon
that occurs is highly instructive, however,So at least
you ought to examinethe given solution. As an alternative, you could apply
somereasonable set of parameters, makingsure the spring is weak, and find the
behavior using MATLAB.
Twootherwise simple oscillators are not infrequently weaklyinterconnected
by a compliance. One exampleof this is two identical pendulumsconnected by
a weakspring, as shownin Fig. 7.14. Considering small motions only, examine
the modalbehavior of this systemfollowing arbitrary initial conditions.

7.3. MATRIX REPRESENTATION OF DYNAMIC BEHAVIOR*

551

Figure 7.14: Guided Problem7.10

Suggested Steps:
1. Drawa bondgraph, label variables, and find the consequentdifferential
equations. Evaluate any inertances and compliances. If you write a differential equation for the energy storage associated with the spring, note
that it is not independentof the other differential equations and does not
introducea fifth state variable.
2. Convert the result of step 1 to give the matrix A.
3. Find the characteristic polynomial,using IsI - A1 = 0.
4. Solve for the four characteristic values, whichcan be written as j~l, -jwl,
Jw2, -jw2.
5. Find the four eigenvectors, letting mli -- 1. Assembleinto either the
standard complexmodal matrix or the modified real modal matrix.
6. Find the eigenvalue matrix Est or Es,nt.
7. Compute the matrix product PEst or PmEsmt. Each column of this
matrix represents the behavior of its respective mode.Therefore, the general solution can be written as the sumof an unspecified coefficient times
the content of each column. If you used the complexmatrices to get the
general solution, the sum of the first two columns and the sum of the
other two columnsmust be real; this meansthat the unspecified coefficients must be complexconjugate pairs. Combinethese pairs of columns
to get the functions in real terms (sines and cosines). This awkwardstep
is unnecessaryif youuse the real matrices from the start.
8. Interpret the results. Describea motionfor whichonly the wl terms exist.
Repeatfor the w2terms. Now,note that if the spring coupling is weak, wl
and w2are almost the same. The sumof two sine or cosine wavesof nearly
equal frequency can be represented by a sine or cosine waveof frequency
equal to the average of the two frequencies modulated in amplitude by
an envelope that is a sine or cosine wavewith a frequency equal to the
difference betweenthe two frequencies. This is sometimesdescribed as a
beating phenomenon.

552

CHAPTER 7. ANALYSIS OF LINEAR MODELS, PART 2

m=l kg
k=400 N/m
b=40 Ns/m
F0=10 N

Figure 7.15: Guided Problem 7.11
Guided Problem

7.11

This frequency response problememploysa familiar simple model, but asks you
to treat it by starting with the matrix format.
The prototypical vibration problemwith a single degree of freedomcomprises
a mass, spring and dashpot excited with a force F(t) = Fo sin wt, as shownin
Fig. 7.15. Find the responseof this systemin the form x(t) = Xo(W)sin[wt+¢(w)]
and sketch the correspondingBodeplot for the parameterslisted in the figure.
Suggested Steps:
Drawa bond graph, write the state differential equations and place in
standard matrix form. Find the matrices A and B, using symbols rather
than numbers.
Extract the scalar transfer function G(s) that relates the force of amplitude
F0 to the displacementof amplitude xo from the moregeneral matrix G (s)
given by equation (7.50) (p. 515).
3. Substitute jw for s in G(s), and find the magnitude IG(jw)l and phase
angle /G(jw) as functions of the frequency w and the parameters.
4. Substitute the values of the parametersand employthe value of Fo to find
the functions xo(w) and ¢(jw).
5. Evaluate these functions for several values of w in the range of 1.0 to 100
rad/s. Plot in Bode coordinates. You may prefer to use MATLAB.
Guided Problem 7.12
The modelneeded for this second freqency response problemis of higher order;
as a result, the use of the state-space matrix methodsis more efficient than
scalar methods.
Consider again the two weakly interconnected pendulumsof Guided Problem
7.10. A small horizontal excitation force F(t) = Fo sinwt is applied to the left
mass. Find the response of this mass (¢1) and represent it by a Bodeplot for
the parameters L = 0.5m, a = 0.15m, rn = 1.0kg, k = 8.0N/m and Fo = 1.0N.

7.3.

MATRIX REPRESENTATION

Suggested

OF DYNAMIC BEHAVIOR*

553

Steps:

You already have the matrix [sI-A] and its determinant, and B is a vector
with a single non-zero element. Thus only one element in the inverse of
the matrix [sI - A] needs to be found to determine the transfer function
G(s). Determine which element this is.
2. One possibility is to solve this problem using MATLAB.
Otherwise, follow
the steps below.
3. Find G(s) using the determinant p(s) = IsI - AI and the adjoint matrix
Adj(sI - A).
4. Substitute the values of the parameters and substitute
the magnitude and the phase angle of G(jw).

jw for s; evaluate

Plot FoG(j~) in Bode coordinates, including a frequency band starting
at least one decade below the natural frequencies and ending at least one
decade above.
PROBLEMS
7.22 The mass-spring system shown on the left in Problem 7.23, known as a
Wilberforce spring, exhibits both translational
and rotational motion. The
parameters include the radius of gyration, rg. The static equations for force and
torque include a weakcoupling coefficient, k~¢, which is typical of coil springs:
M = k¢¢ + kxcx ;

F = kx¢¢ + k~x

(a) Model the system in standard matrix form.
(b) Find a modal matrix and describe the individual

modal motions.

(c) Find analytical expressions for the response given initial deviations
from equilibrium of x = 0.1 m, ¢ = 0, ~ = 0, ~ = 0, and the values of the
parameters assigned in Problem 7.23 below.

7.23 Address the problem above using MATLAB
with the initial
conditions
given in part (c) and the parameter values given at the top of_the next page.
(a) Do parts (a) and (b) numerically.
(b) Find the numerical state transition matrix for a time step of 0.1 second.
(c) Plot x(t) and ¢(t) for several cycles.

554

CHAPTER 7.

ANALYSIS

OF LINEAR MODELS, PART 2

kx = 400 N/m
k¢ = 0.400 N m/rad
kx¢ = 1.00 N/rad
rn -- 1.00 kg
rg = 0.031 m

7.24 The double pendulum shown above right can be modeled as two pointmasses of weight 1 lb at the ends of two massless wires each of length 10 inches.
The angles ¢1 and ¢2 should be assumed to be small, permitting a linear analysis. Answerthe same questions as stated in the preceding problem, substituting
~1 and ~b2 for the displacements x and ~b, respectively.
7.25 A five-story building is assumed to rest on a rigid foundation, but each
floor can movelaterally with a relative deflection per floor proportional to the
shear force. Model the system with a bond graph, define state variables, find
the matrix A, compute the natural frequencies of vibration, and describe the
positional modal shapes. Use of MATLAB
is suggested. The stiffness per floor
is 1 × 10s N/mfor all five floors. The mass is concentrated in the floors, and
equals 5 × 105 kg per floor. Damping may be neglected.

7.26 The building of the above problem is subject to a sinusoidal lateral force
of amplitude 2 × 106 N at various frequencies, uniformly distributed over the
height. Youare to investigate the deflection of the top floor.

(a) Evaluate the matrices A, B, C and
(b) Give a magnitude Bode plot of the response.
scales for this plot may require special attention.)

(Getting reasonable

7.27 Repeat the above problem if, instead of a wind force, an earthquake shakes
the ground laterally with a sinusoidal amplitude of 0.2 meters.

7.3.

MATRIX

REPRESENTATION

OF DYNAMIC BEHAVIOR*

555

7.28 The building of the above problems has a damping coefficient per floor
of 1 x 106 N.s.m. Find the decay rates and damped natural frequencies of the
individual modes.
7.29 Repeat Problem 7.26 assuming the damping added in Problem 7.28.
7.30 Repeat Problem 7.27 assuming the damping added in Problem 7.28.

SOLUTIONS TO GUIDED PROBLEMS
Guided

Problem

1.

q/C
",I-.--:-r~-~
O-,~-~--~.I
aq/~c T.~.P
q/C~)
C
1