Flexures

Flexures
Elements of Elastic Mechanisms

Stuart T. Smith
University of North Carolina, USA

CRC PRESS
Boca Raton London New York Washington, D.C.
 Library of Congress Cataloging-in-Publication Data

Catalog record is available from the Library of Congress

Thi s book contains information obtained from authentic and hi ghly regarded sources. Reprinted material is quoted with
pc1mission, and sources arc indicated. A wide variety of references aJ'e listed. Reasonable efforts have been made to publish
reliable data and information. but the authors and the publisher cannot assume responsibilit y for the validit y of all materials
or for the consequences o f their usc.

Neither this book nor any pan may be reproduced or transmitted in any form or by any means. electronic or mechanical,
including photocopying, microlihning, and recording. or by any information storage or reuievnl system . without plior
permission in writing from the publi sher.

The consent of CRC Press LLC does not extend to copying for general distribution. for promotion. for creat ing new works.
or for resale. Specific permission must be obtained in writi ng from CRC Press LLC for such copying.

Direct all inquirie~ to CRC Press LLC. 2000 N.W. Corporate Bl vd .. Boca Raton. Florida 33431.

Tr adem ark Not ice: Product or corporate names may be trademarks or registered trademarks. and arc used only for
identification and explanation. without intent to infringe.

Visit the CRC Press Web site at www.crcpress.com

© 2000 by CRC Press LLC

No claim to original U.S. Government works

International Standard Book Number 90-5699-26 1-9
Printed in the United States of America 4 5 6 7 8 9 0
Printed on acid-free paper
Contents

Preface xiii

Chapter 1: Introduction 1
1.0 Origins 1
l.l Objective 2
1.2 Advantages of fl exures 2
1.3 Disadvantages of fl exures 3
1.4 Goals of fl exure design 3
1.5 Design considerations 5
References 13

Chapter 2: Essentials 15
2.0 Overview 15
2.1 Basic elasticity 15
2.2 Behavior of materials 17
2.2.1 Metals 17
2.2.2 Non-metals 20
2.3 Principal stresses and strains 22
2.3.1 Biaxial stress 24
2.3.2 Triaxial stresses 25
2.4 Non-principal stresses 26
2.4.1 Plane stress 26
2.4.2 Three dimensional stresses 31
2.4.3 Shear stresses and shear strain 32
2.5 Yield criteria 35
2.5.1 Ductile materials failure criteria 36
2.6 Fatigue 37
2.6.1 SIN curves 37
2.6.2 Effects of notches 40
2.6.3 Effects of mean stress 42
2.6.4 Damage assessment 42

v
CONTENTS

2.7 Bending of symmetric beams 43

2.7.1 Thebendingequation 43
2.7.2 Deflection of beams 45
2.7.2.1 Sign convention for bending moments 46
2.7.3 Moment, shear force and rate ofloading relationships 48
2.7.4 Singularity functions 49
2.8 Torsion 51
2.8.1 Torsion of a prismatic beam of circular cross section 52
2.8.2 Torsion of a prismatic beam of rectangular cross section 53
2.9 Mobility 54
References 56

Chapter 3: Rigid body dynamics 59

3.0 Overview 59
3.1 Generalized coordinates 60
3.2 Properties ofvariational operators 64
3.2.1 Commutation 64
3.2.2 Minima of a function 64
3.3 Hamilton's principle 66
Example I: A simple spring mass system 68
Example 2: Lateral vibration of a bar 69
3.4 Lagrange's equation 71
3.4. 1 Rayleigh's dissipation function 74
3.4.2 General use of Lagrange's equation 75
3.5 Linear systems theory 75
3.5.1 The simple, single degree of freedom, linear, spring-mass-damper
system 75
3.5.2 Some equivalent definitions of a linear system 78
3.5.3 Frequency response functions 79
3.6 Measuring the critical damping ratio 82
3.7 General linea r systems revisited 87
3.8 Multi-degree of freedom linear systems 92
vi
CONTENTS

3.8.1 Note on the graphical representation of frequency response

functions 98
3.9 General response function short cuts 100
3.9.1 Rayleigh's approach to the problem of computing generalized
frequency response functions 103
3.9.1.1 Example of a two degree of freedom system 105
3.10 Eigen analysis 106
3.10.1 Conservative systems 108
3.10.2 Systems with damping 112
3.1 0.3 Interpretation of complex eigenvalues and eigenvectors 113
3.10.4 Summary of primary steps in the derivation of the response
function of a linear multi-degree of freedom system 114
3.10.5 Example: A series, six mass vibration isolator 115
3.10.5.1 Results 118
3.11 Vibrations and natural frequencies of continuous systems 120
3.11.1 Strings 121
3.11.2 Longitudinal vibrations of a rod 124
3.11 .2.1 Longitudinal vibrations of a clamped-free rod 126
3. 11.2.2 Longitudinal vibrations of a free-free or fixed-fixed rod 126
3.11.2.3 Longitudinal vibrations of a clamped-free rod with a rigid mass
at the free end 127
3.11.3 Lateral vibration of a bar 129
3.11.3 .1 Hinged-hinged beam 130
3.11.3.2 The free-free or clamped-clamped bar 131
3.11.4 Lateral vibration ofbars with a rigid mass attached 134
3.11.4.1 Cantilever beam with a rigid mass attached at the free end 134
3.11.4.2 Hinged beam with a central mass, M 135
3.11.5 Vibration of plates 136
3.11.5.1 Vibrations of a circular plate 136
3. 11 .5. 1.1 Free vibration of a circular plate clamped at the perimeter 137

vii
CONTENTS

3.11.5.2 Rayleigh's method applied to a circular pl:lte with a

central mass 140

3.11.6 Vibrations of a rectangular plate 145

3.11.6.1 Free vibrations of a rectangular plate si1r.ply supported at the
edges 145
3.11.6.2 Fundamental frequency's of rectangular plates with other
boundary conditions 146
3. 12 Case study I: A simple two degree offreedom fl exure mecha nism 146
References I51

Chapter 4: Flexure elements 153

4.0 Over view 153
4. 1 Leaf type springs 153
4.1 . 1 The cantilever as a rotary hinge 154
4.1.2 The cantilever hinge subject to an axial compressive force 158
4.1.3 Combined axial and tangential loads 159
4. 1.4 Combined axial and tangential loads plus a moment applied to
the free end of a simple cantilever 167
4.1.5 Leaf type flexures for parallelogram flexure applications 173
4.2 Notch hinge 177
4.2.1 Theoretical considerations 179
4.2. 1.1 The leaf type flexure reconsidered 179
4.2.2 The circular notch hinge 180
4.2.3 Accuracy of stiffness estimates for a notch type hinge 183
4.2.4 The notch hinge of elliptic cross section 185
4.2.5 Compliance's of elliptic hinges in other axes 188
4.2.6 Results 191
4.2.6. I Finite element results 191
4.3 Other hinge elements 192
4.3. 1 The cross strip pivot 193
4.3.1.1 Center shift of the pivot 198
4.3.2 The cattwheel hinge 199
viii 4.3 .2. 1 Torsional stiffness of the cartwheel hinge 201
CONTENTS

4.3.2.2 Center shift of the pivot 202

4.3.2.3 Stresses m the hinge 203
4.3.3 The cruciform hinge 204
4.4 T wo axis hinges 206
4.4.1 A simple two axis hinge {y, B) 206
4.4.2 The two beam, two degree offreedom flexure (y. B) 208
4.4.3 The two axis toroidal hinge (e.,, e,) 211

4.5 Case study 1: Force sensor for contact probe ch aracterization 213
4.5.1 Toroidal notch type flexure 2 15
4.5.2 Two, stacked, notch type flexmes 217
References 218

Chapter 5: Flexure systems 221

5.0 Over view 221
5.1 The four ba r link 222
5 .1.1 The simple leaf type rectilinear spring 222
5.1.2 The simple notch type linear spring 226
5.1.3 The vittual center 231
5.1.4 Effect of the drive on the natural fi·equency 237
5.2 O ptima l geometry for th e rectilinear motion of components on a
simple linear spring 237
5.2.1 Simple linear spring mechanisms 238
5.2.2 Effect of axial strains in the flexure supports 242
5.2.3 Combined effects 243
5.3 Plana r mech anisms 244
5.3.1 Discussion of the simplifying assumptions for reduction of
mobility analysis to planar mechanisms 245
5.4 Dyna mics of ideal plana r fl exures (some common mechanisms) 251
5.4.1 The compound rectilinear spring 252
5.4.2 The double compound linear spring (including the lever driven
spring) 256
5.4.3 A coupled two-axis flexure 259
ix
CONTENTS

5.5 General model for dynamics of planar flexures 263

5.5.1 Coordinate systems 265
5.5.2 Notation 266
5.5.3 Transformations 268
5.5.4 Case study 1: The simple linear spring flexure 274
5.5.5 Comments on general planar analysis 278
5.6 General dynamics of flexures 280
5. 7 Sou rces of interesting fl exure mechanisms 283
References 284

Chapter 6: Hinges of rotational symmetry 287

6.0 Overview 287
6.l lntroduction 287
6.2 Coil springs 288
6.3 T he disc coupling (freedoms 8t, 8=, x) 291

6.3.1 The inner to outer rim disc coupling 293

6.3.2 The outer rim disc coupling 297
6.3.2.1 Angular stiffness 304
6.3.2.2 Axial stiffness 306
6.3.2.3 Torsional stiffness 306
6.4 Rotationally symmetric leaf type hinge (axial stiffness of the disc
coupling revisited) 308
6.4.1 Axial stiffness and maximum stress calculations for the rotationally
symmetrk leaf type flexure system 310
6.4.2 S implified equation for maximum stress 313
6.4.3 Assessment of approximate equations using finite element models 314
6.5 T he bellows as a flexure element 317
6.5.1 Torsional stiffness of the rectangular bellows 320
6.5.2 Axial stiffness of the rectangular bellows 322
6.5.3 'S' Shaped distmtion of the bellows 325
6.6 T he notch a nd leaf type hinge applied to couplings of rotational
sym metry 326
X
CONTENTS

6.7 Case study 1: A metrological three-axis translator for constant force

profilometry 327
6.7. 1 Principle of operation 328
References 331

Chapter 7: Levers 335

7.0 Overview 335
7.1 Introduction 335
7.2 Mechanical levers 336
7.2. 1 The rigid lever 337
7.2.2 Soft spring - stiff spring attenuation 343
7.3 Lost motion 345
7.4 Effects of levers on flexure dynamics 352
7.5 Case study I: A fine adjustment mechanism for motion attenuation 357
7.5. 1 Finite element analysis ofthe lever flexure 359
7.5.2 Mal)ufacture of the lever using wire electro-discharge machining 368
7.6 Case study II : Optical levers, galvanometers and the filar
suspension 369
References 372

Chapter 8: Manufacturing and assembly considerations 373

8.0 Overview 373
8.1 Manufacture 374
8.1.1 Conventional machining 374
8. 1.2 Electro-discharge machining 379
8. 1.2. 1 Plunge electro-d ischarge machining 379
8.1.2.2 Wire electro-discharge machining 380
8. 1.3 Lithographic etching 381
8.1.3. I Flexures produced using m icroeleclronic processing techniques
(MEMS) 382
8.1 .3.2 Lithographic etching of copper sheet 383
8.1.4 Electroplating (or electro-forming) 384
8.1.5 Diamond grinding 385
8.2 Assembly 388
xi
 CONTENTS

8 . ~.1 Assembly of flexures 389

8.2.2 Coupling the actuator to the flexure 393
8.2.3 Flexure mounts 396
8.2.4 Adding damping to flexure systems (extracting energy) 396
8.2.4 .1 Internal friction in solids 396
8.2.4.2 Adding damping to a tlexure 403
8.2.5 Effects of manufacturing tolerance on the stiffness of flexure
elements 406
8.2.6 Typical nexure drives 410
8.3 Machining and h eat treatment of some common fl exure materials 414
8.3.1 Steels 415
8.3.2 Beryllium copper 415
References 416
Author index 419
Subj ect index 423

xii
Preface

Flexures represent a broad range of compliant elements for connecting rigid

bodies. Flexure mechanisms are produced through the successive connection of
numerous rigid bodies, or 'links', and flexure elements, or ' joints', in such a way
that, upon application of an appropriate force, there will be defined motions of
one link relative to the others. Consequently, the goal of a flexure mechanism is
to maintain a precise geometric relation between links while simultaneously
providing sufficient compliance to accommodate relative motion in specific
directions. Precise geometric relations can only be maintained through suitably
applied constraints while relative motions require freedom. Because the required
constraints and freedoms are in different directions, such a requirement is, in
many cases, readily achieved. Such directionality can often be approximated
using elastic elements, the geometry of which results in high and low compliance
in the required directions. Though providing neither perfect constraint nor
freedom, relative stiffness in different directions can vary by many orders of
magnitude thereby producing a close approximation to the ideal. It is the design
of flexure elements and their integration to form useful mechanisms that
occupies this current text.
Being on the design of elastic flexure mechanisms, it is hoped that, by
highlighting current issues and deficiencies in knowiedge (possibly the authors),
this has broken the ground for further developments. In many applications, and
within limited ranges, flexures provide an inexpensive solution to problems
requiring ultra-precise motions and/or forces. In some cases, the results are
nothing short of spectacular. As a caveat, during the synthesis stage of design,
there is always a danger of premature 'fine focus'. Too close attention to this
book in isolation could induce 'near sight'. As has been the author's experience,
flexures are ~ot always the best solution to a fine motion or linear force problem.
It is the objective of this book to provide ideas to enable 'good' flexure design
and, possibly more importantly, an analytic framework to determine their
suitability for a particular application. If the results of such analysis push the
designer towards uneasy compromises, the chances are that alternatives may be
more suitable. If this happens, it is probably best to look back to the possible
solutions that were considered during initial brainstorming or reconsider the
design requirements. Because of their simplicity, it is sometimes difficult to drop
the idea of using a flexure in favor of more complicated alternatives. Some solace
can be derived from the knowledge that the degree of difficulty sometimes
represents a competitive edge in the marketplace.
This book provides the relevant mathematical tools and formulae for a broad
range of flexure elements and their subsequent combination to form useful
mechanisms. It is hoped that it will be of use for designers of precision
mechanical instruments be they engineers, physicists or any other discipline in

XIII
PREfo'ACE

which fine instruments are required . Though this book is intended to be of use to
professionals concerned with the destgn of precise mechanical mstruments and
machines, tt could also be used to compliment graduate courses on precision
mechamcal destgn and will certainly provide extensive background material on
flexu re design to graduate researchers. It is the queries of the latter that created
the mitial motivation for starting this work and I hope that tlus will address
some of the common, recurrent tssues.
For the assessment. o ptimization and simplification of designs, symmetry
has played probably the greatest part in the author's experience. Symmetric
designs often tend to nullify possible errors, simplify analysis, de-couple
vibration modes, reduce cost of manufacture and are inherently thermally and
dynamically balanced. In addition to the above advantages, symmetry is 'easy on
the eye' and can often be spotted at a glance. However, after a flexure has
distorted, the symmeh·y is perturbed. In reality, vibration modes and forces
originally considered to be independent are always weakly coupled and the
engineer must live with the consequences.
This is drawn from collaborative work with many colleagues and students
from around the world over many years. The support and faith of immediate
coworkers at both the University of Warwick, UK, and the University of North
Carolina at Charlotte, USA, has made this possible. In the production of this
manuscript it became necessary to divert time from other duties. Bob Hocken
without mention added a number of my more onerous burdens to his already
substantial workload. Special mention goes to Shane Woody for working
through many sections of the book, verifying some of the newly developed
theories and removing a number of errors and misprints. Vivek Badami, Eric
Coley, Jami Dale and Ashok Muralidhur were most helpful in providing case
study examples. Dr's Harish Cherukuri and Jimmie Miller both carefully read
the manuscript and, in many places, removed ambiguity and inelegant prose
while adding considerable insight to the theoretical foundations upon which
most of this book is built. Thanks also go to John Lovingood for both reading the
proofs and providing the majority of the figures.
These notes represent the ideas and opinions of the author and, at the
present time, much is left to say on this topic, possibly a second volume.
Although every effort has been made to eliminate errors, experience dictates
their insidious and inevitable presence in a work of this length. Ex tensive
references are included to enable the reader to verify this work using articles that
have been subject to the scrutiny of peer review. When detected, the author
would be most grateful for g uidance on errors or omissions.

xiv
 1 Introduction
Then he took up the bow; with his right hand,
he tried the string; it sang as clearly as
a swallow's note1.

1.0 Origins

Elastic deflections have been utilized in fine instrument mechanisms and precise
machines for certainly more than three hundred years (a lot longer if the bow is
considered a precision machine). While Galileo (1564-1642) sowed the seeds of
scientific investigation of the static response of a built-in beam to applied forces,
it was left to Hooke (1635-1703) and Marriott (1620 -1684) to recognize the basis
of linear elasticity. Subsequently, Augustine Cauchy (1789-1857), Barre Saint-
Venant (1797-1886) and William Thompson (1824-1907 later Lord Kelvin) played
a major part in the establishment of a rigorous framework upon which present
theoretical investigations are founded, for a more complete historical discussion
see Love, 1927, and Timoshenko, 1953. Possibly, an epocryphal moment marking
first uses of precision flexures in instrumentation began with John Harrison's
(1693-1776) development of a clock of unsurpassed precision for determination
of longitude, see for example Sobel, 1995. With Principia as a guide, Harrison
strove for most of his eighty-three years to produce a mechanism of temporal
accuracy matchin& as near as possible, the precision provided by the new
Newtonian description of the universe (at that time with unprecedented success).
Another period of note for which flexures can be identified as 'instrumental' to
scientific progress, was the development of galvanometric instruments for the
precise determination of the unit of electrical current (designers involved in the
development of these early precision instruments include Helmholtz, Joule,
Kelvin, Maxwell and Weber). Today flexure mechanisms in various forms find
use in applications at the limits of precision and can also be found in
commonplace consumer products. Examples at the extremes of precision include
fine positioning stages for a wide variety of mechanical measuring instruments,
probe microscopes, mass balances, step-and-repeat cameras and x-ray

1
The Odyssey of Homer, trans. Allen Mandelbawn, 1990, University of California Press, Berkeley and Los
Angeles, Ca, Book XXI, Greek [401-428]
FLEXURES

interferometers. Commercially, flexures are ubiquitous and can be found in

computer disk drives, compact disk players, coordinate measuring machines,
optical scanners, optical interferometers, dial indicators, Fabry-Perot etalons,
electronic lithography and almost any precision manufacturing machine, for a
condensed review of various mechanisms see Geary, 1954, 1960, Sydenham,
1984. Possibly the main reason for their success is because flexures are easy to
manufacture and provide smooth, friction free and wear free motion. In marked
contrast, friction induced errors often account for a major portion of the precision
motion errors of conventional mechanisms.
A review of the emergence of this specialized topic within the field of
instrumentation must await the attention of a more qualified science historian.
However, readers may derive some sense of the history of flexures within a
scientific context from some of the case studies in this text and also the book of R.
V. Jones (1911-1997) and references therein.

1.1 Objective

A flexure is usually considered to be a mechanism consisting of a series of rigid

bodies connected by compliant elements that is designed to produce a
geometrically well-defined motion upon application of a force. The objective of
this book is to cover design of flexure mechanisms for precise control of motion
or forces. In particular, techniques for analysis of flexure elements are discussed
in some detail with some issues pertaining to manufacture and implementation
covered in later chapters. Information on the design of the more conventional
types of spring, including material properties and selection, can be found in
more detail in the books of Carlson, 1978, and Wahl, 1964. These references
provide a comprehensive discussion of coil and other, more traditional, spring
designs and could be considered complimentary to this book. Before proceeding,
it is worth reviewing the merits and limitations of flexure mechanisms for
application in precision instrumentation and machinery.

1.2 Advantages of flexures

1. They are simple and inexpensive to manufacture and assemble

2. Unless fatigue cracks develop, the flexures undergo no irreversible
deformations and are, therefore, wear-free.
3. Complete mechanisms can be produced from a single monolith.
4. Mechanical leverage is easily implemented.
5. Displacements are smooth and continuous. Even for applications
requiring displacements of atomic resolution, flexures have been shown
to readily produce predictable and repeatable motions at this level.

2
CHAPTER 1: INTRODUCTION

6. Failure mechanisms such as fatigue and yield are well understood.

7. They can be designed to be insensitive to thermal variations and
mechanical disturbances (vibrations). Symmetric designs can be
inherently compensated and balanced.
8. There will be a linear relationship between applied force and
displacement for small distortions. For elastic distortions, this linear
relationship is independent of manufacturing tolerance. However, the
direction of motion will be less well defined as these tolerances are
relaxed.

1.3 Disadvantages of flexures

1. Accurate prediction of force-displacement characteristics requires

accurate knowledge of the elastic modulus and geometry/ dimensions.
Even tight manufacturing tolerances can produce relatively large
uncertainty between predicted and actual performance.
2. At significant stresses there will be some hysteresis in the stress-strain
characteristics of most materials.
3. Flexures are restricted in the length of translation for a given size and
stiffness.
4. Out of plane stiffness values are relatively low and drive direction
stiffness tends to be relatively high in comparison to other bearing
systems.
5. They cannot tolerate large loads.
6. Accidental overload can be catastrophic or, at least, significantly reduce
fatigue life.
7. At large loads there may be more than one state corresponding to
equilibrium, possibly leading to instabilities such as buckling or 'tin-
canning'.

1.4 Goals of flexure design

The goals of any flexure design are twofold:

1. To provide a precise displacement upon application of a specific applied
force.
2. To provide a precisely known force upon application of a specific
applied displacement.

3
FLEXURES

9.84

.
...-., 9.83 +-------------~
---=--!!!!:
- =;;...__-

g 982 +--------------~~---------
§ /
'D 9.81 +-- - - - -- -----,.,./
"'----- - - - - - - - -
1
G
9.8 ~---/_,.,c-.-------
/
0 9.79 +--~
-----,~::....._ _ _ _ __ _ _________
·;;:
~ 9.78 +-~=--·--------------
977+------~------~-----r-------r------~
0 20 40 60 80 100
Latitude (deg)

Figure 1.1: Local gravity acceleration at sea level, 0 degrees corresponds to either the
' true' north, or south, poles while 90 degrees is at the equator.

The former of these goals is surprisingly difficult. Were it not for the
convenient constant effects of gravitation and a somewhat tenuous, artifact
based, mass standard, things would be nigh impossible for the flexure designer
and many other scientific investigations. To this day, the simplest method for
deriving a 'known' force is to utilize the weight of a calibrated mass. In fact, if the
latitude, ¢, and height above sea level, h, are known, assuming smooth
topography, the local gravity acceleration in the direction of the Earth's center of
mass can be computed with uncertainty of less than one part in 10,000 from the
equation, Howard and Fu, 1997,

g =9.78024(1 + 0.0052884sin 2 ¢- 0.0000059sin 2 2¢)- 0.000003086h (1.1)

At sea level, the local gravity constant as a function of latitude is shown in

figure 1.1. Irrespective of the value of gravity acceleration, at a fixed point
relative to the Earth this will have a value that can be assumed constant for
extremely long periods of time. Ignoring growth of contaminant films, wear and
other 'decay' processes masses derived from solids can also maintain a relatively
constant value over protracted periods of time. However, for most applications,
a constant force would lead to a rather dull and unproductive machine, the
gravity wound' clock being an exception.
Alternative methods for the production of forces requires known values for
electromagnetic interactions that often include the requirement that electrical,
mechanical and thermal properties of system elements be known. Solving
Maxwell's equations for the system and computing Lorentz forces between
elements will enable precise computation of dynamically varying forces. In many

4
CHAPTER 1: INTRODUCfiON

cases, symmetry, simple geometric construction and for slowly varying forces,
the equations describing the system performance reduce to simple form.
The latter goal of flexure designs requires that the mechanism be driven by a
prescribed displacement Of course, a given displacement requires a specific
force. However, for this latter system, it is possible to use a displacement
transducer (direct methods such as capacitive, inductive or optical gaging or
inferred displacements from strain gages, pressure transducers, etc.) so that the
motions can be determined and feedback systems may be utilized to obtain
prescribed displacements. With closed loop control, the limiting precision of
achievable motions is changed from the performance of the drive to that of the
sensors and control systems. Because many drive actuators have undesirable
characteristics such as hysteresis, creep, thermal susceptibility and non-linear
stiffness, systems incorporating closed loop control of displacement are often
necessary to realize the highest attainable precision.
In some cases, the drive train may be sufficiently stiff so that less direct
means of inferring displacement are sufficient. Examples might include a stepper
motor drive of a feedscrew in which pulses to the stepper motor may be used to
infer displacement, high torque motor drives with end stop monitoring and
assuming constant velocity during motion, hydraulic drives in which pressure
differences may be used to infer forces. If such a system is sufficiently reliable
and performs to within the requirements, open-loop designs often represent an
economic solution.
In general, for any flexure system
The relative stiffness of the drive and flexure represents a measure of
achievable precision.
More precisely this can be stated as
The stiffness of the components of the mechanism in which force and
measurement loops coincide is a measure of the ability to maintain precision
while doing work.
Determination of flexure element and related component compliances and
the subsequent distortions, of which some may be undesirable, occupies a
significant fraction of this book. In many cases flexures are used to translate a
'rigid' body in a specific direction (i.e. linear motion or rotation about a fixed
axis), deviations of motion from this path are referred to as parasitic errors
throughout. For more general discussions the reader can consult, Rivin, 1994,
Smith and Chetwynd, 1992, and Slocum, 1992.

1.5 Design considerations

While there are no systematic rules for the creative process required for the
conception of a means for producing a mechanical effect to satisfy a need, often

5
FLEXURES

referred to as 'synthesis', a number of approaches and important considerations

can guide the designer to the best of the available solutions. In this brief section,
some general guidelines for the assessment of a design, and, in particular, a
flexure design, will be outlined. In the following brief discussion, a background
appreciation of the design process will be assumed. If the reader is not familiar
with some of the terms used in this section, more detailed discussions can be
found in Smith and Chetwynd, 1992 and Ertas and Jones, 1996.
Synthesis is the creative process that seeks the best solution for a specific
need. Important stages of this are
1. Recognizing a need
2. Establishing design requirements
3. Conceptualization and creativity
4. Assessment of feasibility
Although brainstorming should be an open-minded process, unless guided
by detailed and, where possible, quantitative requirements, it is unlikely that the
solutions will satisfy all of the needs and it is also likely that time will be spent
considering unnecessary issues. Establishing quantitative requirements is an
essential first step in the design process. It is also important that these
requirements are independent of a design. A frequent trap with any design is to
think of the solution and tailor the requirements to suit. Such an approach will
invariably result in poor design. As an exainple, the specification for the range of
forces to be measured in a particular application might be expressed by either of
the two following requirements
a. Forces ranging from 0.1 to 1,000 N must be measured
b. The cantilever of the force balance should be designed for applied forces
ranging from 0.1 to 1,000 N
While both of the above specify the required range, it is obvious which
requirement is independent of design. In many cases such a distinction is less
obvious. The temptation of the second requirement is that it represents less work
for the designer. Consequently, it might, incorrectly, be perceived that this is a
lower cost path! While the first is open to any solution, the latter can be
visualized, modeled and assessed for feasibility immediately. If the first stage of
synthesis involves analysis, it is unlikely that a design process has occurred.
Often, quantitative performance measures can be classified using the following
terminology
• Range
• Resolution
• Precision, a measure of machine performance not effected by calibration

6
CHAPTER 1: lNTRODUCfiON

• Accuracy, dependent entirely on calibration

• Errors, often split into random and systematic
• Traceability
• Repeatability/ stability
• Linearity
• Relevant standards and legal requirements
Analysis is the assessment of possible solutions using quantitative models.
This is used to determine feasibility and optimize parameters of the model. If the
model accurately represents the design, optimization of one implies the other. It
is rare for analysis to correct, or even compensate, for poor synthesis.
Conceptualization and creativity relate to the skills, imagination and
experience of the designer. The ability to apply knowledge, 'common sense' and
experience is the key to good design. Some concepts can help to broaden these
qualities. In particular, paradigm shifts are often the result of lateral thinking and
an excellent discussion of this is to be found in the books of Edward de Bono,
1971, 1974. Although a discussion of the creative process is beyond the scope of
this book, some.general approaches are worthy of mention.
Inversion can transform a feasible but impractical design to an elegant
solution. The conventional arch and the suspension bridge represent a classic
example of inversion. The former traditionally utilizes concrete as the material of
construction. Because of this, bridges were designed to produce predominantly
compressive forces to favorably exploit the materials properties. However, the
costs of such constructions become prohibitive for longer spans. Turning the
bridge upside down results in a design that produces predominantly tensile
forces thereby making steel wire a possible material of construction.
Reduction is also a powerful method for simplification. Often, designs tend
to evolve through successive integration of individual solutions. Generally, each
element of the solution is somehow connected to the other in the hope that, when
combined, the complete system will satisfy all requirements. During this process
it is common that elements have a 'base' and these are simply connected. Often
the two connected components could be produced from a single block of
material. Sometimes, a complex fabrication might be replaced by a mechanism
produced from a single monolith, as will be shown for many successive flexure
designs. Although it is rare that a complex design can be reduced to a single
component, it is often surprising how many components can be eliminated by
intervening ideas between the initial concept to the final product.
Probably one of the most important skills to be acquired by a designer is the
ability to sketch. An idea has little chance of being accepted unless it is effectively
communicated. In the same way that analysis should play little part in the initial
conception of a design, detailed drawings can also constrain the solution

7
FLEXURES

prematurely. While the ability to sketch is not a prerequisite for good ideas, it is
most frustrating to listen to a lengthy explanation of an idea only to realize that
either a simple sketch could have portrayed the concept in a fraction of the time
or, after much discussion, that a number of people have interpreted the
description erroneously. The latter of these scenarios is time consuming and,
sometimes, expensive. The key benefits of rapid sketching are, Jones, 1932,
• Sketches enhance the communication of ideas.
• Easily drawn sketches are readily modified or discarded, thereby
preventing premature focus on less than optimal designs.
• Clear, well-proportioned sketches can be rapidly converted to
engineering drawings.
Almost by definition, brainstorming is likely to produce many solutions to a
given problem. In practice there will not be time to undertake a detailed
feasibility study of them all. Often, the best strategy is to cut out the least feasible
using the concept of Ockham' s razor. This states that
When you have two competing theories producing exactly the same
prediction, the one that is simpler is the better.
Another complimentary proposition also applies
A simple solution almost alWtllJS exists.
Having selected a set of possible designs, analysis can be readily used to
assess feasibility based on known physical laws. Such analysis tools represent the
bulk of the material presented in this book There are, however, more qualitative
concepts that may be used to critically assess a design.
As mentioned in the preface to this book, symmetry confers many favorable
characteristics to mechanism performance. Some of the benefits of a symmetric
mechanism often include
a. Balancing of stresses can null distortion errors. Stresses about a line of
symmetry are often of equal and opposite value resulting in no net
displacement of points on the line.

b. Symmetric mechanisms are inherently balanced providing immunity to

inertial forces and heat inputs. This is true only for disturbances of
similar symmetry. For example, thermal expansion effects in a
mechanism will balance if the heat source is placed at the line of
symmetry. In practice this is likely to be the case if the heat source is
symmetric.
c. Symmetry often simplifies analysis. Not only can this lead to boundary
conditions such as zero slope and displacement that are easily modeled,

8
CHAPI'ER 1: lNTRODUCllON

symmetry often nulls some effects such that they no longer need to be
considered.
d. Forces and vibration modes become decoupled. Under such conditions,
each plane of symmetry can be analyzed independently. However, in
practice, it is very difficult to prevent vibration in one plane of
symmetry from coupling into other planes. To choose an example from
Rayleigh, 1894, try setting a guitar string in vibration so that the string
oscillates in a single plane. In some applications, it is also possible to use
broken symmetry to alter or suppress modes of vibration to the
designer's advantage.
e. Components are often easier to manufacture.
f. Symmetry can be spotted at a glance. Once recognized it is often most
informative to consider the consequences thereof.
Another important design concept is that of force and measurement loops
intrinsic to any instrument or machine (this discussion is adapted from notes
prepared by Derek Chetwynd). By definition, a flexure mechanism produces
defined displacements (be they linear, angular or combinations thereof) upon
application of applied forces. For most applications it is necessary, at some stage,
to measure the relative displacement of defined points on the flexure. Such
measurements might occur during calibration or be continuously monitored by
some integrated displacement sensor. Simplistically, most instruments and
machines and almost all flexures require application of forces and measurement
of relative displacements. As a consequence
Every force in a static system
• must pass from body to body in a closed path (Newton's third law)
• causes strain in those 'rigid' bodies
Every meaningful measurement requires a datum. This requires
• an arbitrary origin locked to a component of the mechanism
• a 'rigid' link from this origin to the gage probe
Both of these involve a structural loop. In some cases components of the
measurement loop may consist of an optical path either in air or along a fiber. In
either case the input and output points sources must be anchored to a frame and
it is this frame plus any changes in the optical path length that constitutes the
structural components of the loop. By definition
• Force loops carry loads and hold the system together and must necessarily
distort.
• Measurement loops carry datum information around the system and must
be rigid.

9
FLEXURFS

Errors may be introduced whenever components of the force and

measurement loops coincide so, where possible, keep them apart.
In general, optimal designs often conform to the following rules for loops
• Make loop elements stiff.
• Keep forces (including self-weight) small.
• Keep bending moments in the 'rigid' bodies (very) small.
• Keep loading (thermal or mechanical) as symmetric as possible.
• Maintain stable temperatures.
• Avoid temperature gradients.
• Put heat sources at points of symmetry.
• Keep fundamental resonance frequencies as high as possible.
• Keep loops small. This infers high frequency response, small bending
moments, low self-weight and fast thermal response.
Other considerations can have profound influence on the attainable precision
of a flexure mechanism. Of these, compensation and both cosine and Abbe errors
represent two powerful concepts for assessment of design potential.
A number of compensation strategies are commonly employed. These being
• Cancellation. This is achieved by adding equal and opposite effects and
often occurs as a consequence of symmetry. One method is to use
differential measurement for the elimination of common mode errors
due to random disturbances (vibrations, electrical interference etc.) or
systematic perturbations (inertial forces, thermal changes etc.).
Balancing is also a classic example of a cancellation strategy.
• Correction. Repeatable errors can be removed by providing a look-up
table to determine the true value from a measurement. Where
appropriate, fitting the input/ output relationship of a system to a
reasonable mathematical model can be used to enhance the accuracy of
an instrument or machine.
Often, measuring an artifact can be used to assess instrument errors.
However, such measurements contain two systematic errors, those of the
instrument and those of the artifact. Instrument errors are fixed to its frame
while those of the latter move with the artifact. In many cases, using the
instrument to measure the artifact in different orientations and comparing results
can enable separation of the two errors. A review of error separation techniques
can be found in Evans et al., 1996.
For linear d isplacement measurements, cosine and Abbe errors often
represent the dominant limit to instrument accuracy. These are caused by

10
CHAPTER 1: INTRODUCTION

a) probe
r-------------------~ \ \
'

~-------'----~) oB
P'.......f---1
x'

Perfect scale
Reading=x

b)

II II III III
X

c)

\J oB
II IIIIIIII a P'
x'

t
I

Figure 1.2: Illustration of alignment errors that can occur with linear displacement
measurement, a) perfect measurement, b) probe misaligned but with axis initially
passing through point to be measured, c) misalignment error with Abbe offset

differences between the line of action of the measuring probe and the motion of
the point to be measured. Conceptually, the models of figures 1.2 (a-c) can
represent these. Figure 1.2 (a) represents a perfect configuration. In this, motion
of a point P on a surface is to be measured. It will be imagined that this
undergoes a linear translation to point P' accompanied by a parasitic (i.e.
undesirable and, often, unknown), small rotation about this point of magnitude
o(). Clearly, if the probe is both parallel with the axis of motion and the axis of
measurement passes though the point P then the perfect relationship between
measurement, x, and true motion, x', is achieved i.e.

x = x' (1.2)

II
FLEXURES

In this case, parasitic rotations have no influence on the measurement. Such a

situation does not occur in practice. Figure 1.2(b) represents a less ideal scenario
where the line of action of the measurement is at an angle, B., to that of the
motion but its axis still, initially, passes through the point P. In this case, the
relationship between the scale reading and true displacement is

x =x'cosB. + x'sinB. sino(}

~x'(l - 8 f +B.oB)
(1.3)

For small angles, the error due to misalignment is of second order and in
many cases is relatively small. Equation (1.3) implies that the parasitic rotation
might compensate the cosine term. However, in practice, the sign of oB is often
arbitrary. Also, the magnitude of the parasitic rotation in comparison to initial
misalignment is usually small. Consequently, the first term of equation (1.3}
dominates and therefore this is often referred to as a 'cosine error'. More serious,
and more common, in addition to a misalignment there will also be an offset a
between the point to be measured and the line of action of the probe as shown in
figure 1.2(c). In this case, the relationship between the scale reading and true
displacement is

x =x'(cosB. +sin B. sinoB)- asinoB

(1.4}
~x'-a/iB

From equation (1.4) it can be seen that the offset a results in an error that is
directly proportional to the parasitic rotation. In many designs this represents
the dominant error. Ernst Abbe realized the importance of this and it is to him
that the alignment principle is named. Simply stated
When measuring displacement ofa specific point it is necessan; that the axis
of measurement should be both parallel to and pass through the motion of the
point
When the errors are repeatable and free from hysteresis, this can often be
removed, or at least reduced, by arranging for the calibration of the probe to
satisfy the alignment principle.
While most of the above applies to design in general, some problems specific
to flexures appear omnipresent. By definition, analysis requires a process of
abstracting a real world component to enable quantitative mathematical
modeling. In the process of reducing the complexity of a complete model to
something that can be solved within a reasonable time it is necessary to make
some assumptions. Often, the resulting model will still represent a reasonable
correspondence between predicted and actual behavior. However, there are two
commonly applied assumptions that should be used with utmost care. These are

12
CHAPTER 1: INTRODUCriON

1. Flexure elements are connected by infinitely rigid bodies, and

2. Flexures can be assembled with perfect fixtures that neither distort the
flexure elements nor introduce significant stresses.
In many designs requiring high precision motion such assumptions may lead
to significant discrepancy between modeled and actual behavior. The following
caveats should always be attached to the above assumptions
1. Rigid bodies are not rigid, especially when levers are present.
2. For small motions, fabricated mechanisms often show some hysteresis
due to tin canning, plastic deformations or insufficient fasteners.
Tin canning is often caused by curvature of leaf type flexures. Causes of this
can be either pre-existing bending of 'flat' plates as supplied by the manufacturer
(often due to rolling) or bending of the plates by the clamps. Microscopic plastic
deformation can, once again, occur as a consequence of the clamps. Either,
surfaces in the clamped region will experience plastic deformation of surface
asperities (microscopic plastic deformation) or there might be some initial
permanent bending of leaf springs (macroscopic plastic deformation).
Additionally, bolted joints may creep and move when subject to varying forces.
Some design considerations for assembly of fabricated flexures are discussed in
chapter 8. In general, if there appear to be discrepancies between the model and
flexure behavior, the above assumptions are likely sources.

References
De Bono, 1971, The Use of Lateral Thinking, Pelican Books
De Bono, 1974, Po: Betjond Yes and No, Pelican Books
Carlson H., 1978, Spring Designer's Handbook, Marcel Dekker, Inc., NY
Ertas A. and Jones J.C., 1996, The Engineering Design Process, John Wiley and Sons
Inc., NY, chpt. 1.
Evans C.J., Hocken R.J. and Estler W.T., 1996, Self-calibration: Reversal,
redundancy, error separation and 'absolute testing', Annals of the CIRP, 45(2),
617-636
Geary P.J., B.S.I.R.A. Research Report M18, (1954),
-B.S.I.R.A., Research Report R249, (1960)
Howard L.P. and Fu J., 1997, Accurate force measurements for miniature
mechanical systems; a review of progress, Proc. SPIE, 3225, 2-11.
Jones F.D., 1932, How to Sketch Mechanisms, The Industrial Press, NY.
Jones R.V., 1987, Instruments and Experiences, J. Wiley and Sons, London.

13
FLEXURES

Love A. E. H., 1927, A Treatise on the Mathematical Theon; of Elasticity, 4th ed., Dover
Publications Inc (1983), NY, chapter 1 provides an historical overview, the
rest of the book is for the serious scholar.
Baron Rayleigh, J.W.S. Strutt, 1894, Theory of Sound, Dover Publications (1945
edition), NY, §149, page 243
Rivin E., 1994, Stiffness in Design, ASPE tutorial notes.
Slocum A.H., 1992, Precision Machine Design, Prentice Hall, NJ.
Sobel D., 1995, Longitude: The Ston; of a Lone Genius Who Solved the Greatest
Scientific Problem of His Time, Walker and Company, NY, as the title suggests,
this is light reading but nevertheless interesting from a flexure design
perspective.
Smith S.T. and Chetwynd D.G., 1992, Foundations of Ultrapreciswn Mechanism
Design, Gordon and Breach, London, UK.
Sydenham P.J., 1984, Elastic design of fine mechanism in instruments, J. Phys. E:
Sci. Instrum., 17, 922-930, see also, Mechanical design of instruments, parts A
& B: Putting elasticity to use, Measurement and Control, 14, 179- 185 & 219-
227.
Timoshenko S.P., 1953, Histon; of Strength of Materials, Dover Publications Inc
(1983), NY.
Wahl A. M., 1964, Mechanical Springs, McGraw-Hill Book Co., NY.

14
 2 Essentials

2.0 Overview

Being concerned in this book with the achievement of precise motion from elastic
distortions upon application of controlled forces, an appreciation of stress and strain
relationships is necessanJ throughout. This chapter reviews some essential concepts
of elasticity and the various techniques for the prediction of stresses, strains and
subsequent distortions. Because reliabililtJ of flexure designs is a common concern,
overviews offailure criteria for brittle and ductile materials as well as simple fatigue
calculation methods are presented. In general, this chapter is a compilation of
relevant physical laws and a selection of the commonly used mathematical
techniques extracted, for the nwst part, from texts on elasticity and strength of
materials. Finally, flexure mechanisms are used to achieve a prescribed displacement
that can be characterized by one or more independent coordinates. Analysis of the
number of independent coordinates and their loci forms the subject of kinematics. For
any flexure mechanism, it is necessary to determine the number of independent
coordinates, or freedoms. Considering a mechanism as a kinematic chain of rigid
bodies subject to constraints imposed lnj the flexure elements, the number of
freedoms can be readily assessed using the concept of mobility analysis presented at
the end of this chapter.

2.1 Basic elasticity

Hooke (1635-1703) was the first to observe a linear relationship between the
extension of a wire due to a force applied along its axis and the proportionate
scaling with the length. Mathematically, from this simple relationship it is
possible to define the two parameters stress, a, and strain, s, given by

F
0'=-
A
1-/0 M
(2.1)
&=--=-
/0 /0

where F is the applied force normal to a surface of area A and l is the length of
the wire after the applied force and [0 is the original length, see figure 2.1.
FLEXURES

In practice, equations (2.1) apply only to thin

wires in tension under relatively light loads. It is
also necessary to have a more rigorous definition
of the quantities being measured. In the first of
the above equations, the true stress a,,. should
be considered as the ratio of tensile force in the
thin wire to that of the instantaneous cross section

rr
of the wire. Inherent to this equation is the
assumption that the stress is uniformly
l distributed across the wire (a reasonable
I I assumption in this instance). Under these
I I circumstances the true stress is given by

L
Applied force F
a = =- (2.2)
'"'• Instantaneous area A,

Unfortunately, even when measuring

Figure 2.1 : Experimental something as simple as a wire, it is often a lot
arrangement for the measurement more convenient to measure the dimensions of
of a thin wire due to an applied
load
the wire before testing and use the original area
for the calculation of stress. Being a more
pragmatic measurement, the computed ratio of force to area is known as the
engineering stress and is given by
Applied force F
a = =- (2.3)
•nz Original area of wire A.,

Similarly, when considering the engineering strain in the wire given by

equation (1.1), it is assumed that the length does not change significantly. For a
true assessment of strain it becomes necessary to integrate incremental changes
in length divided by the instantaneous length giving

(2.4)

Clearly, for small deformations, the distinction between engineering and true
stress and strains is not significant. Fortunately, throughout this book we are
concerned with the design of elastic mechanisms. For a flexure to be repeatable

16
CHAPTER 2: ESSENTIALS

and reliable it is necessary to keep stresses small and therefore we may

reasonably ignore this distinction between 'true' and 'engineering' stress and
strain and thus drop the subscripts.

2.2 Behavior of materials

Equations (2.1) to (2.4) above impose no limits on the possible values of stress or
strain. In reality, all materials have limits, the values of which define the strength
of a given structure. To evaluate these limits, it is common to apply a tensile or
compressive load of known value to a specimen of uniform cross section and
known length and measure subsequent extension. From such measurements it is
relatively straightforward to compute the engineering stress/ strain curves. In
general it is found that there are two quite distinct types of behavior. For almost
all metals there is an initial linear-elastic (note that elastic does not necessarily
imply linear) region in which the specimen will return to its original length if the
load is released. This is followed by a rapid extension with increasing load as the
stress is increased beyond a certain value. In this latter stage of deformation the
material has undergone an irreversible (or plastic) flow. The ability to plastically
deform without fracture is called ductility and is characteristic of many metallic
solids. Other materials that might be used for flexure applications, such as many
ceramics, exhibit brittle failure due to the growth of cracks in the plane of the
maximum tensile stress. Because of these differences, the responses to simple
uni-axial loads of metals and non-metals are discussed separately below.

2.2.1 Metals
Figure 2.2 shows a typical stress verses strain curve for a medium carbon steel
specimen. At low loads, it can be seen that there is a relatively linear region and,
upon unloading, the specimen will return very nearly to its original length.
Above a certain value (point A in figure 2.2) there is a relatively sudden plastic
transition, the upper yield stress, after
(j D which the apparent stress drops to a lower
yield stress that continues at a relatively
constant value for a finite range of strain.
At strains above point C the stress then
continues to steadily increase with a
relatively large proportion of plastic
distortion. This reaches a peak at D, which
is known as the ultimate strength after
which there is a decline in stress with
strain followed by catastrophic failure.
Experimentally, the variability of the
strain at which complete separation occurs
Figure 2.2: Typical stress-strain curve for a renders this value of little utility.
soft steel specimen

17
FLEXURES

400

-005 005 015 025 0 35 045

Milli strain

Figure 2.3: Typical stress strain curves for commercially available metal
alloys
If the steel is loaded beyond the lower yield stress region and the load is then
released, the specimen will tend to recover exhibiting a near elastic characteristic
but with a permanent offset strain at zero applied stress. Upon subsequent
testing it is often observed that the lower yield phenomena is absent resulting in
a smooth transition from elastic to plastic behavior. Such curves are also
characteristic of a large number of other metals and typical stress strain curves
for a variety of these are shown in figure 2.3.
To design reasonable flexure mechanisms, it is important to clarify some of
the parameters used to described the strength properties of metals. Possibly the
most common measure is called the yield stress. Generally, this refers to the value
of stress that must be applied to produce a certain percentage of permanent
engineering strain. Most suppliers choose to use the stress at 0.2% permanent
strain although 0.1% has been used in the past. Such a measure introduces a
quandary to the flexure designer interested in high precision mechanisms that
are usually required to perform with considerably better repeatability.
Consequently, it makes more sense to seek a value that represents a considerably
lower percentage permanent deformation. The proportional and elastic limit
represent two alternative and more promising parameters that are often quoted.
The proportional limit represents the maximum stress at which the stress-strain
characteristic does not deviate from a straight line to within the capability of the
instrumentation with which it is measured. Similarly, the elastic limit represents
the maximum stress at which no permanent deformation can be detected upon
removal of all applied stresses. In practice, for many metals, both values are
nearly the same and the two are often considered synonymous. Unfortunately,
both of these parameters are very dependent on the precision of the
instrumentation with which they are measured. One particularly sensitive
measure of the elastic limit is to apply a cyclic load and plot this continuously.
Clearly, if there is any permanent deformation each cycle will exhibit a closed

18
CHAPTER 2: ESSENTIALS

loop the area of which is a monitor of 'hysteresis' and thus permanent

deformation in the stress-strain characteristic. Using such methods, for some
materials the limiting shear stress for purely elastic behavior can be anything
down to one hundredth that of the yield stress, Dieter, 1986.
As a consequence of the above arguments, parameters such as the true elastic
limit are of little utility for most designs and alternative rule of thumb
parameters must be sought. For steels, it is known that below the endurance stress,
cyclic loads will not introduce significant fatigue damage. For a wide range of
steels the endurance limiting stress, S,, can be reasonably found from the
ultimate tensile stress, s.,, using the relationship, Shigley and Mische, 1989

_ { 0.504S,, S", < 1400 MPa

(2.5)
S,- 700MPa s., > 1400 MPa
Unfortunately there does not appear to be a straightforward relationship
between the ultimate tensile and yield strengths in simple tension, the latter
parameter being more commonly measured. However, it is always safe to assume
that the endurance stress is more than half of the yield stress and often greater
than two thirds of this value.
For other metals, the onset of plastic deformations in microscopic regions
within a stressed solid is again difficult to predict and has been subject to
extensive investigations over the years. Unfortunately, no simple conclusions can
be drawn from such studies. However, the stress at the first detectable onset of
plastic deformation in metals, often called micro-slip, is often dependent on the
grain size and can have values of anything from 25 to 75% of the macroscopic
yield stress commonly supplied by manufacturers, Brentnall and Rostoker, 1965.
Consequently, a reasonably conservative value of one quarter of the yield stress
might be more appropriate for ultra-precision applications requiring a high
degree of repeatability.
A measure of a materials capacity for the storage of strain energy is the
resilience, UR • This can be relatively simply derived from the work done per unit
volume in an element on the point of yielding given by

u = (}' y&y =!:!...

2
(2.6)
R 2 2E

Clearly, a good material would possess a high yield stress and low modulus
of elasticity. Work-hardened and heat-treated beryllium copper is a good
example of such a material. Resilience represents the ability of a material to
return energy after elastic deformation. Toughness is the ability to absorb energy
during plastic deformation. This could be calculated from the area under the
stress-strain curve between the strain at yield and at failure. However, the
variability of the failure strain renders this an unreliable measure and a more

19
FLEXURES

rigorous definition based on fracture mechanics is often adopted. For flexure

applications it is necessary to avoid irreversible deformation and, as a
consequence, behavior of materials in the plastic region is not considered further.
However, it should be noted that the toughness is a measure of the degree to
which an overstressed flexure might withstand catastrophic failure and might be
important for safety critical applications.

2.2.2 Non-metals
Considering the experimental difficulties associated with the determination
of the elastic properties of metals, brittle materials at first appear a lot more
amenable to a simple elastic analysis. Figure 2.4 shows the results from a bending
test on a simply supported beam machined from single crystal silicon, a
characteristic typical of many 'brittle' materials. For convenience, the ordinate
has been scaled to give an estimate of the maximum tensile stress on the surface
of the beam. In general, for many 'brittle' materials, the stress-strain graph is
extremely linear up until catastrophic failure.
Because of this near perfect linear, elastic behavior, such materials would
appear ideal. Certainly, for applications to flexures requiring extremes of
precision this is indeed so. Brittle materials tend to either exhibit superb linear
elastic properties or they break. As a consequence they can often be considered
the mechanical equivalent of the electric fuse. A spring constructed from a brittle
material can be trusted to either work as designed or not at all. In contrast, a
metal spring may undergo plastic deformation. When this occurs, the spring
characteristic becomes non-linear and hysteretic. Note also that the magnitude of
the hysteresis is dependent on the strain history. As a consequence, satisfactory
analytic models for the prediction of force-displacement characteristics for
springs undergoing plastic deformation have yet to be developed.
As usual there are catches, brittle materials are often difficult to both
manufacture and assemble/ fasten, will not tolerate accidental overloads and,
160
like metals, the point at which
catastrophic failure occurs is
,-... 120 somewhat difficult to predict.

!
For example, during testing of
80 silicon beams, the maximum
~
stress in bending varied from
E
r:n 40 around SO up to 400 MPa
dependent upon the surface
0~-----+--------+-----~--------~----~ damage introduced during
0 0.0002 0 0004 0 0006 0.0008 0.001 manufacture. However, the
Strain upper values of stress could
Figure 2.4: Stress versus strain curve for a single only be obtained after a more
crystal silicon beam measured in a three point than 80 J.1.ffi of the as-ground
bending apparatus surface had been removed by

20
CHAPTER 2: ESSENTIALS

1.2

~
~
......
08
0
?;> 06

:g
:-3
04

~ 02

0 50 100 150
Stress (MPa)

Figure 2.5: Cumulative probability of failure corresponding to a Weibull distribution

with m = 10 and a 0 = 100 MPa

chemical etching, Smith et al., 1991. It is now generally accepted that surface
flaws are responsible for the reduction in the strength of a broad range of brittle
materials represented by predominantly covalent bonded solids and many
ceramics. The effect of a particular flaw on the surface strength has again been
studied in considerable detail and can now be considered relatively well
understood, see for example Lawn, 1993. Unfortunately, the measurement or
prediction of flaw sizes is not possible for a vast range of engineering processes.
For applications where the breaking strength must be consistent, it is common to
follow up the rough machining of ceramics, usually grinding, with finishing
processes such as wet etching, lapping and polishing. The random nature of the
physical process tends to result in a random flaw size enabling use of statistical
approaches for strength prediction. For a large number of ceramic materials, it
has been found that the probability of failure for a given stress with a material
produced using a specific process will closely follow a Weibull distribution.
Consequently, the probability of failure, P, follows a cumulative distribution of
the form

(2.7)

where cr is the stress acting to open the crack (referred to as mode I type stress in
fracture mechanics), a 0 is called the scaling stress and m is the WEd bull modulus.
Figure 2.5 shows a graph of the above equation for m = 10 (typical for many
ceramics) and a 0 = 100 MPa. The probability density given by the derivative of
(2.7) is shown in figure 2.6. For example, if a failure rate of 1 percent is
considered acceptable the design stress can be either measured directly from the
graph or computed from the equation

21
FLEXURES

004

0
·;;; 003
s::
Q)

g 002
"0

:.0
"'g
.D
001
0..

0 50 100 150 200

Stress (MPa)

Figure 2.6: Probability density of the WeibuJl distribution for the parameters used in
figure 2.5

(2.8)

which for this example corresponds to a working tensile stress of 63 MPa, a value
not unreasonable for single crystal silicon.
Associated with the almost binary performance of these materials is the fact
that they can be survival tested after manufacture by applying the maximum
stress that they are likely to encounter during use. This is known as proof testing
and has the effect of eliminating specimens with unacceptably large flaws. For a
probability of failure during a proof testing PP, the new cumulative probability
distribution of the remaining specimens, P , is given by

P- P
P'= - - P (2.9)
1- Pp

If a suitable proof stress is chosen, by eliminating the low strength tail of the
distribution, such a procedure can result in a substantially reduced variability
with the loss of a small percentage of total production.

2.3 Principal s tresses and strains

Clearly, for solid bodies of arbitrary geometry and applied loads, to determine
the state of stress at a specific point in a solid, it becomes necessary to resort to
the more advanced mathematical theory of elasticity. It is not the intention of this
book to provide a text for such a study and therefore we shall restrict ourselves
to a brief outline of the relevant formulae and definitions of the terms contained
herein.

22
CHAPTER 2: ESSENTIALS

The linear relationship between load and strain first observed by Hooke
suggests a constant ratio of stress to strain for a given material called the modulus
ofelasticity, E, given by
(j
E= - (2.10)
e
This simple linear relationship is of enormous utility to engineers for the
estimation of stresses in solid bodies of homogeneous, isotropic materials when
subject to externally applied loads. It should be emphasized at this juncture that
the elastic modulus is a material property and, unlike measured parameters such
as stress and strain, will be prone to breakdown at extreme values, to be
discussed in sections 2.3 and 2.4. To extend this linear relationship between stress
and strain to three dimensional objects it is informative to discard the thin wire
model in favor of a vanishingly small cube of material subject to simple, uniform
stresses on each of its three orthogonal pairs of faces as shown in figure 2.7. In
the first of these figures, the cube has been selected at an arbitrary angle. Under
these circumstances the possible types of stresses applied at each face can be
reduced to a direct stress plus two shear stresses acting in the plane of the
element. The subscripts used to denote the shear stresses indicate the plane and
direction respectively and a counter--clockwise applied shear stress is considered
to be positive. The second of figures 2.7 represents the stresses on a cube that has
been oriented in the direction of the principal stresses. In this orientation the
shear stresses are zero. To emphasize the unique values of the principal stresses
these are denoted with the subscripts 1, 2 and 3 instead of x, y and z used to
denote three orthogonal stresses at an arbitrary orientation. Obviously, if the
b)

Figure 2.7: Stresses on the faces of an

elemental cube of material s ubject to
external loads, a) stresses on a cube of
arbitrary orientation, b) stresses on a
cube oriented so that the shear stresses
are zero and, by definition, the direct
stresses correspond to principal stresses.
Note that equilibrium requires r 11=•,,

23
FLEXURES

cube becomes vanishingly small, the stresses on each of the three opposing faces
will be equal. Considering the effects of either a tensile or compressive stress on
one face of the cube, it is obvious that there will be a corresponding strain in the
same direction. However, close inspection will reveal strains of opposite sign on
the other two faces. Experimentally it can be shown that these strains are
proportional to the applied stress from which it is possible to deduce the
relationship

2 Es E&
- u =-- = - -3 (2.11)
a1 a1

or

(2.12)

where the dimensionless material property u and is known as Poisson's ratio in

honor of its originator S.D. Poisson (1781-1840).

2.3.1 Biaxial stress

Because the relationships between stress and strain are all linear, it is possible to
apply the principle of superposition and state that the total strain is simply the
sum of all contributions. Consider the case where there are only two applied
stresses. The forces generating these stresses are in a single plane with zero stress
perpendicular to this. Such a state of stresses is common to many engineering
structures (particularly in flat plates and shells) and is known as plane stress
where it is usually assumed that the zero stress is that normal to any free surface.
In the presence of forces in directions 1 & 2 of figure 2.7(b) the three components
of strain are simply given by the sum due to individual stresses

a, al
& =-- u-
1 E E
al a,
& = - - u- (2.13)
2 E E
al a,
s =-u- - u -
3 E E

It is informative to consider the special case where the strain in any of the
directions is zero. Ignoring the trivial case in which the two stresses are both
zero, the first strain will be zero if the stress in its direction is exactly the same as
the product of Poisson's ratio and the stress in plane two and both stresses are of
the same sign. One can easily imagine this by considering firstly, compressing

24
CHAPTER 2: ESSENTIALS

one side of the cube in direction 2 say so that there is a corresponding expansion
in direction 1. Applying a stress, a 1 of magnitude va2 will then return this face
of the cube to its original dimension thus reducing the strain to zero. Similar
arguments follow for the strain in direction 2 while for the strain in the
unconstrained direction it is obvious that this can only be reduced to zero
through the application of equal stresses having opposite sign. In words, this
occurs when the strain due to compressive stress in one direction is canceled by
an equal and opposite tensile stress in the other.
Equations (2.13) can be relatively simply rearranged to yield the components
of stress in terms of the strains

(2.14)

Equation (2.14) can be useful when strain gages are being used to measure
the stresses provided that they are oriented in the direction of the two orthogonal applied
forces. The reason for this proviso is because we are presently considering an
element that is oriented in the direction of the principal stresses of which the more
general case will be discussed in section 2.5.

2.3.2 Triaxial stresses

Based upon the reasoning for the derivation of the strain due to biaxial stresses it
is a relatively obvious extension to write the strains due to a state of triaxial
stresses

(2.15)

with similar expressions for & 2 & & 3

Again, with some algebraic manipulation it can be shown that the explicit
relationship for principal stresses in terms of principal strains is given by

(2.16)

with similar expressions for a 2 and a 3

25
FLEXURES

2.4 Non-principal stresses

2.4.1 Plane stress

So far we have only considered an elemental cube of material subject to forces
normal to its surface. Assuming that the directions of these principal stresses are
not known in advance it is likely that the cube would be subject to forces at an
angle to the surfaces. Under these circumstances, it is possible to simplify the
problem by resolving these forces into components normal and coplanar with
each surface. The normal forces are given the prefix x, y or z while the planar
components of force when normalized by the area of the plane are known as the
shear stresses and are denoted by r,,, where i is the plane (i.e. the plane i =
constant) and j indicates the direction. Figure 2.7 shows the complete system of
stresses acting on an elemental cube. Although not of great utility in this book, to
be consistent, the shear stresses are considered to be positive when acting
counter-clockwise.
A complete analysis of three dimensional stresses and strains for arbitrary
applied loads tends to become rather involved and is adequately covered in such
texts as Timoshenko and Goodier, 1970, or Popov, 1976. For the purposes of this
book, the relevant formulae and techniques for the visualization of stresses
(Mohr's circle) will be introduced. The primary purpose of this section is to
indicate that any combination of stresses can be reduced, by suitable rotation of
the coordinates of the element, to an orientation at which the shear stresses are
zero! Such a reduction then enables the prediction of stresses likely to produce
failure as well as the likely
y
orientation of either shear

t deformation or crack propagation

r·---
<Yy T"' depending on whether the material
acts in ductile or brittle manner.
Ignoring, for the time being, both

-
the d irect and shear stresses in the z
<Yx o-, plane, it is possible to reduce our
analysis to a two dimensional
X
problem of a thin elemental shell
subject to stresses only in the x-y
.. I plane. Such a plane stress

.. approximation is applicable to a
,.,.

t (Yy
broad range of engineering
problems. A thin shell element
subject to plane stress is depicted in
Figure 2.8: Stresses acting on the faces of an figure 2.8. Clearly, if the element is
element for which stresses in the plane of the z not going to rotate, equilibrium
axis are zero

26
CHAPTER 2: ESSENTIALS

..
requires that

l ..
(j" (2.17)

rxy Cutting the element of figure 2.8 at an

t,
-r)IJC
arbitrary angle '' see figure 2.9, and
resolving forces results in the equation

Figure 2.9: Stresses acting on the faces of

a triangular element cut from that of figure
2.8 at an angle ¢

ads = u 1 sin(¢)dx + u" cos(¢)dy + T ,.y{sin(¢)dy + cos(¢)dx) (2.18)

However, noting that

dx =sin(¢)ds
(2.19)
dy =cos(¢)ds

equation (2.18) can be rearranged, with some trigonometric manipulation, to give

the direct stress in the explicit form

(2.20)

Again, with some algebraic manipulation, the shear stress at an arbitrary

angle ' can be shown to be
(j (j
-r = "2
-
1
sin(2¢) + r "Y cos(2¢) (2.21)

The angle at which the direct stress is a maximum or minimum can be

obtained from the equation

~; =0 =-{u " - u 1 )sin(2¢) + 2-rxy cos(2¢) (2.22)

from which we have

2-r
tan(2¢) = xy (2.23)
(j" - (jy

27
FLEXURES

Because the right hand side of this equation represents the stresses on the
element at in the direction of the fixed coordinate system, these represent known
values. Consequently, solving for the tangent term on the left-hand side results
in two values for 'which are shifted by n/2. Similarly, there are two angles at 90
degrees to each other at which shear stresses are a maximum given by

(2.24)

Before discussing the implications of this equation in more detail, it is

informative to investigate the values of this shear stress when the direct stresses
are at a maximum and vice versa. Equation (2.24) (' when the direct stresses are
a maximum) can be rearranged to produce

. (2¢) =-.:.._--'--'-
sm 2-r xy cos{2¢)
(2.25)
ux -uY

Substituting this into the equation for the shear stress (2.21) at this angle
gives

(2.26)
= 0!

This is particularly important for analysis of plane stress problems in that it

indicates that there will always be an angle at which we can orient our
coordinate system, so that, only principal direct stresses are acting on the
element. Under these conditions equations (2.10) to (2.16) can then be used to
convert between stress and strain and these will be used in the next section for
the prediction of failure of materials subject to three-dimensional stresses.
It can also be shown that the value of the stress normal to the plane of
maximum shear stress is the same in each plane and is given by

(2.27)

Note that the subscripts 1 and 2 in equation (2.27) indicate that these are
principal stresses. This leads us to a general principle for plane stresses, which can
be extended to three-dimensional analyses, and this is
For an elemental volume of material subject to any combination of stresses it is
possible to find an orientation in wlrich the element is subject to triaxial
principal stresses only

28
CHAPTER 2: ESSENTIALS

In 1882 Otto Mohr (pronounced 'more') showed that the above rather
complex trigonometric relationships could be graphically represented as a circle
in which the x- axis represents direct stresses and the y-axis represent shear
stresses. First of all squaring and then adding equations (2.20) and (2.21} gives

(O' - lt2(0'x+O'y )f +1' =(lt2(0'x-O'y)Y +1'~

2
(2.28)

which is an equation of a circle of the form

(2.29)

where x =cr and y = 't and

R =!_~(0'
2 X
-0' )2 + 41'2
y X)l
(2.30)

(2.31)

0' tension

Figure 2.10: Geometric representation of stresses on an element at an arbitrary

angle ¢J using Mohr's circle

29
FLEXURES

A graphical representation of this circle is shown in figure 2.10. This is called

Mohr's circle for stress. Clearly, this can be easily plotted for any plane stress
problem if two orthogonal values of direct stress plus a shear stress are known at
any orientation. Construction of this circle from such data is covered extensively
in undergraduate courses and will not be reproduced here. However, it is worth
indicating the general steps for the determination of the magnitude and
orientation of the principal stresses using Mohr's circle. These are
Determine the direction and magnitude of the applied direct and shear
stresses in the plane of interest
For the two faces (i.e. x andy planes) determine the nature of the stresses.
These being either tensile or compressive in the case of the direct
stresses and clockwise and counter-clockwise for the shear stresses.
Draw the axes for Mohr's circle. In this case, the horizontal coordinate
represents values of the direct stresses, with tensile stresses to the right
and compressive stresses to the left of the origin. In opposition to the
signs of the shear stresses, these are represented by the vertical axis
with clockwise (or negative) shear stresses being above the horizontal
line and counter-clockwise shear stresses being below.
Plot the two points corresponding to the coordinates (ux,-rxy ) and
(u. . ,-r....J
There will then be only one possible circle that bisects these two points,
draw this.
To determine the stresses at an angle¢ rotate around.the perimeter of the
circle an angle 2¢ in the same direction as on the element. Note that, if
the counter-clockwise shear is plotted above the origin, it would be
necessary to rotate in the opposite direction on Mohr's circle to that on
the element.

From equation (2.26) the principal stress occurs when the shear stress is zero.
Consequently, from (2.28) the principal stresses for this plane stress problem are
given by

(2.32)

Again from (2.28) the maximum shear stress in the plane under consideration is
given by

(2.33)

30
CHAPTER 2. ESSENTIALS

2.4.2 Three dimensional stresses

In the above analyses, although not stated, 1t is assumed that stresses applied to
the two-dunensional thin shell represent critical des1gn values. However, 1t has
already been stated that, for a general three-dimensional element, it IS possible to
determine an orientation such that the element ts subject to three orthogonal
pnncipal stresses. Under these circumstances it can be shown, Ford, 1977, that
three single Mohr's circles can be used to represent the stresses at an arbttrary
orientation within the element. The Mohr's circle representation of stresses for
the elemental cube of figure 2.7 is shown in figure 2.11. Although beyond the
scope of this monograph to analyze such a system of stresses, it is informative to
review the main conclusions from such an analysis. These are

t' max

cr,
Figure 2.11: Mohr's circle representation of three-dimensional stresses

For an arbitrary orientation within the element, the stress will be wtthm the
shaded region of figure 2.11
To assess maximum and minimum stresses m any element (including plane
stress) it is necessary to use values from the Mohr's circle of largest radius.
For example, in a plane stress problem in whtch the prmc1pal stresses are
equal cr1 =cr 2 , the Mohr's circle becomes a point. However, because
cr3 =0 it is relatively easy to visualize that the maximum shear stress in
the element is u 1 I 2 . It is obviously most important to keep thts point in
mind when assessing the stresses in a solid
For a plane stress problem, the zero stress normal to the plane of the two
dimensional element is a principal stress

31
FLEXURES

2.4.3 Shear stresses and shear strain

So far, because the major effort of the above analysis has been concerned with the
detemunation of principal stresses in which the shear terms are zero, the effects
of shear stress on the distortion of an element have been ignored. However,
shear distortion can be significant in a number of applications, in particular, for
bodies subject to torsion loading.
Following Timoshenko and Goodier, 1970, it is informative once again to
consider a two dimensional element subject to both linear and angular
distortions as shown in figure 2.12. It can be shown that the components of strain
on an element at any point in our Cartesian coordinate system are simply given
by the equations

tU
s =-
• at
&=-
ov (2.34)
)1 o/
eM
&
z
=az
-

ou
u+~

yt ~
Ov
v+ - dy
Oy

ou
u+-dx
Ox Ov
v+-dx
Ox
X

Figure 2.12: Distortions of an element subject to bia,Ual stress

32
CHAPTER 2. ESSENTIALS

u, v, w are the displacements of particles m the body caused by the strruns.

Through analogous reasoning, it can also be shown, Timoshenko, 1970, the
shearing strains, y (measured in radians) can be obtained from

6U 6\1
r ~= q, +~
6U Q.v
rxz =& +a; (2.35)
6\1 Q.v
r =&- +-q,
)'%

Constdering Mohr's circle once again, it is

obvious that the only condition in which it is
possible to have pure shear on the faces of this
element and zero normal stresses ts if the
average direct principal stress is zero. For our
plane stress element this corresponds to the
two principal stresses being of equal
u magnitude with one compressive and the other
tensile. It is also relatively easy to see that these
will be at 90 degrees to the element on which
we are considering the shear stress.
Figure 2.13: Stresses acting on the Consequently, an element under pure shear can
faces of a square element within a
solid subject to equal and opposite
be equally assessed by considering an element
direct stresses of similar magnitude that is at an angle of 45 degrees withm a large
shell subject to simple tensile and compresstve
stresses of equal magnitude, cr, as shown in figure 2.13. Again following the
treabnent of Timoshenko and Goodier, 1970, the shear stress on the element is
given by

(2.36)

From simple geometric considerations it can be shown that if y is the total

angular distortion of the square element then it can be shown

tr r) tan(tr I 4)- tan(r I 2) 1+ e 1

tan 4 - 2 = 1+ tan( tr I 4) tan(r I 2) = 1+ e2
(
(2.37)
1-(r 12)
~ 1+(r 12)

Because we are considering biaxial principal stresses

33
FLEXURES

(1 + v)u
s. =-
E
(2.38)
(1 + v)u
&2 = E
From equations (2.37) and (2.38) it is relatively simple to derive the
relationships for shearing stresses and strains m the form

2(1 + v) 2(1 + v)
r= E u= E "
(2.39)

Often it is useful to define the constant relationship between shearing strain

and shearmg stress using the notation

G= E (2.40)
2{1 + v)

from which

o"
r =- (2.41)

where G IS called the nwdulus ofngtdtty

Assuming that the element was initially square, the shearing strain is the
angle at which the elemental cube dzffers from its initial value of 90 degrees.
Equation (2.41) can be extended to yield the general relationship

'r xy
Yxy= G

r Y2 "
= ; (2.42)

Yu = "G
The importance of the above equations will become apparent when we
consider the torsion of prismatic beams m section 2.8

34
CHAPTER 2 ESSENTIALS

2.5 Yield criteria

As an introduction to the field of fatlure prediction, it is worth contemplating the
remarkable graph of figure 2.14 (reproduced from Spain, 1987). This shows the
change in the volume of a single crystal silicon specimen as it is subjected to
hydrostatic pressures up to 100 GPa, as measured using a fascinating instrument
known as a diamond anvil cell. These extreme pressures are achteved by
compressing the specimen between the flats of two dtamond anvils with a
metallic seal to prevent lower stresses perpendicular to the applied stress (for a
popular account of the development of this technology see Hazen, 1993).
Contrast this with the simple bending test shown in figure 2.5 indicating

1.0

:
\ I

)
I

dI :
I
08
v Tetragonal <P- Sn)

v.
~ 0

'-.,
l Sunple hexagonal
06
~-~exagonal
Intermediate
close-packed

phase ._____!_cc

20 40 60 80 100

GPa
Figure 2.14: RelatJve volume of silicon as a function of
pressure. Dashed line mdicates metastable state on slow
pressure to the metallic state. Dashed dot line represents
volume on release of pressure to a metastable phase
(reproduced from Spain, 1987).

catastrophic failure at an applied stress of 138 MPa! It is obvtous from these two
graphs that it is not simply the magnitude of stress that causes failure. Although
it has only recently been possible to demonstrate materials responses to such
htgh stresses, it has been known for many years that it is the devzation from
hydrostatic stress that results m failure. For ductile materials there are two
criteria for predicting the onset of yield, these are
1. For metals, failure occurs when the maximum shear stress exceeds that
which is observed to cause yielding in a sunple tensile test (Tresca's
criteria). For brittle materials, failure is predicted when the tensile
stress at the surface rises above a specified value.

35
FLEXURES

2. Alternatively, failure may be predicted when the dzstorhon strain

energy in an elemental volume becomes comparable to that present in
a simple tensile test at the onset of yield (von Mises criteria)
2.5.1 Ductile material failure criteria
Considering the case of metals, the maximum shear stress is simply obtained
from the radius of the largest Mohr's circle for a given element. Fortunately, the
value of the maximum shear stress is easily measured in the laboratory using a
simple tensile test. Such a test imposes a uniform, principal stress over the cross
section of the specimen with zero stress in the other two orthogonal planes.
Consequently, the maximum Mohr's circle passes through both the origin and
the yield stress. Clearly, under these conditions the maximum shear stress is

{2.43)

For arbitrary stresses, the failure condition can be obtained by computing the
largest difference between principal stress values. For example if 0'1 > 0'2 > 0'3
then fa.ilure will occur when

(2.44)

Alternatively, failure is predicted upon the condition

(2.45)

In general, this condition can be written

(2.46)

where 0'Y is the yield stress measured using a simple tension test
This theory of failure was proposed by Tresca in 1868 and often bears his
name.
An alternative failure criteria based on distortion strain energy appears to
have been first proposed by James Clerk Maxwell. Ignorant of this, the same
hypothesis was independently presented over subsequent decades by a number
of researchers including E. Beltrami, M. T. Huber, R. von Mises and H. Hencky,
(for a discussion of the development of this theory see Timoshenko, 1983). For
historic reasons, it is now more commonly associated with von Mises. Stated
simply, it is predicted that failure will occur when the dzstortion strain energy of
an element of material exceeds that at the onset of yield in a simple tensile test.
Based on this, it is necessary to consider the distortion of an elemental cube of
material. These distortions can be resolved into two separate components. One

36
CHAPTER 2: ESSENTIALS

corresponds to the average strain in each axis that results in a change of size of
the elemental cube. To get to the final distorted shape, it is necessary to add
additional strain energy corresponding to deviations of stress about the average
value (often called the deviatoric stresses) resulting in a distortion of the shape of
the element. Based on this criterion, failure is predicted to occur when the
differences in stress become equal to the yield stress in the relation

(2.47)

In many applications, beams or other components are subject to torsion and

bending loads. This introduces a simple direct stress plus shear and, depending
on whether the von Mises or Tresca criteria are used, failure is predicted when

von Mises (2.48)

Tresca (2.49)

Although both of the above results are identical for the prediction of yield in
a simple direct tensile test, there is a difference in the criteria for failure due to
additional effects of shear. For example, if a torsion test is applied to a rod of
material, failure due to shear is predicted to occur at the point at which

vonMises (2.50)

O"y
1'
xy
=2- Tresca (2.51)

Experimentally it has been found that the von Mises criteria more accurately
predicts the onset of yield while the Tresca criteria provides a more conservative
estimate of the allowable design stress. When subject to cyclic loads, ductile
materials can fail at stresses below the yield stress in an apparently brittle
manner, whence

2.6 Fatigue

2.6.1 SIN curves

It is the intention of this short section to provide a min.imal outline of fatigue
analysis. Consequently, it will not be possible to discuss the more intricate
methods that are required for full solutions to problems in which accurate
predictions may be critical to the safe use of a particular design.

37
FLEXURES

Although metals are known to fail due to shear under slowly increasing load
conditions, if a cyclic stress is applied to a material, it is possible that a pre-
existing crack in the material will steadily grow until it reaches a critical value
after which the component will fail catastrophically. Such failure is characteristic
of brittle materials and can be analyzed in terms of fracture mechanics. However,
for our purposes, it is more informative to present a brief overview of more
conventional engineering approach to the problem of fatigue prediction and,
more importantly, avoidance. More complete introductory discussions of this
topic can be found in Sherratt, 1982, Hertzberg, 1989 and Zahavi and Torbilo,
1996. Consider the behavior of a small specimen subject to a stress that cycles
slowly between equal values of tension and compression. For many commonly
used metals, it is found that the number of cycles to failure is dependent on the
magnitude of the applied stress. If the stress is relatively high in comparison with
the ultimate tensile stress, failure can occur after a few thousand cycles or less.
Components with such short lives are analyzed in what is called the Low Cycle
Fatigue region. For flexure designs, it is unlikely that a design will be subject to
stresses so close to the yield stress. It is more common to design with limiting
stresses for which the component will survive a large number of cycles to failure
(called the High Cycle Fatigue region) or a stress that is below a value that can be
endured indefinitely. Such a stress value for a particular material is known as its
endurance limit.
Unfortunately, the fatigue life of a particular component will be dependent
upon the nature of the applied load, the material of its construction and its
geometry. However, in many applications, a component is subject to either
simple direct stresses or it is possible to predict the major principal stresses.
Consequently, it is reasonable to apply simple tensile and compressive stresses of
magnitude S to a specimen and count the number of cycles, N, to failure. The
upper case S is used to denote the engineering stress as opposed to true stresses,
whic~ may be significantly different when measuring the ultimate tensile stress
s., . Plotting the engineering stress against cycles to failure on a log-log graph
often results in the plot shown in figure 2.15. In this graph it can be seen that
there is a linear portion between approximately 1()4 and 1()6 cycles after which
failure is not observed. Clearly, in the linear region of this graph or high cycle
fatigue, there is a relationship between number of cycles to failure, N, and stress,
S, of the form

(2.52)

where

(2.53)

38
CHAPTER 2: ESSENTIALS

a and b are material constants.

s 600

400

300

b) 200

100

0 ~-T--~~--~--r-~~
1 E+OO 1.E+01 1 E+02 1 E+03 1 E+04 1 E+05 1 E•05 1 E•07

s 1000

c)

Figure 2.15: Plots of S- N data on different scales, a) linear, b)

Log-linear, b) log-log scales
Looking at the S - N curves for metals other than steel it will soon be
apparent that this simple picture may provide faulty predictions. For example,
although aluminum does not have an endurance limit it is usual to quote the
stress at which it will survive 500 million cycles. Table 2.1 at the end of this
chapter lists the commonly used endurance limits for a number of metals as
provided in the Atlas of Fatigue Curves edited by Howard E. Boyer.
At this stage in the discussion it is worth spending a little time indicating
some of the errors that may be introduced when following this analysis route.
Great care has been taken to indicate that we are considering only the
engineering stresses on a specimen. A paradox becomes apparent when it is
realized that we are assessing a phenomenon in which a crack is gradually
growing within the stressed zone. As a consequence, as the crack grows the area

39
FLEXURES

supporting the applied loads reduces, thereby increasing the true stress in this
region. For small specimens, this will probably not have a significant effect but
care should be employed when analyzing components in which the stress
extends over more than a few tens of millimeters in any one direction.
Techniques are available using strain control that, potentially, overcomes this
particular problem. A discussion of strain-based analysis is outside of the scope
of this book Another related problem is that of accidental overload. If this
overload stress is beyond the high cycle fatigue limit a significant reduction in
life might ensue after which the analysis is no longer valid. Again, techniques to
cope with such scenarios do exist. However, the complexity of their
implementation means that it is often more economic to monitor for overloads
and replace components after such an event has occurred. Another, more
frequently encountered, and easily analyzed, fatigue life reduction mechanism is
discussed in the following section.

2.6.2 Effects of notches

Upon inspection of components that have failed by fatigue, it is often apparent

that this has occurred as a consequence of the growth of a crack from a region
where there is a sudden change of geometry such as a notch or hole. For some
pretty spectacular examples see Hertzberg, 1989. Clearly, this is due to the
concentration of stress in the region near to the notch. Reasonably, the effect of
the notch could be analyzed by calculating the maximum stress due to this
sudden change of geometry and using it as the basis for subsequent calculations.
For this purpose it is usual to define a stress concentration factor, K, , given by

K = Maximum stress in the notch

(2.54)
' A nominal stress away from the notch

For example, if a small hole is drilled into a wide flat plate that is subject to a
uniform tensile stress there will be a local threefold increase in stress at two
points on the periphery of the hole (i.e. K, =3 ). Clearly, it would seem
reasonable to measure the stress somewhere remote from the hole and simply
multiply it by the appropriate factor to predict the fatigue life. However,
subsequent tests on notched specimens often reveal that such an approach is
rather conservative and specimens will last longer than predictions based on the
stress concentration factor, see figure 2.16. Consequently, for notched specimens
it is usual to use a strength reduction factor, K1 , which is always less than or equal
to the stress concentration factor and given by the ratio of allowable stresses for
specimens with and without notches i.e.

K =Allowable range of stresses without notch (2.55)

1
Allowable range of stresses with notch

40
CHAPTER 2: ESSENTIALS

1000 s

- ----..
~

- -- ... ..
X...............
- s
100
-~ ~. ..~ .. - - SJkt

- - - - Sikf

X Sexptl

10
1.0E+OO 1.0E+03 1.0E+06

Average number of cycles to failure

Figure 2.16: Plots of S - N data for a specimen containing a local notch

Often the allowable stresses are simply given as the endurance limits for the
notched and notch-free specimens. Unfortunately, there are two problems with
such an approach. Firstly, the strength reduction factor is only valid if the curve
for the notched specimen is parallel to that for the notch-free case as shown in the
figure. In practice, the notches have little effect in the low cycle fatigue region.
The influence tends to increase with number of cycles as shown in the figure.
Consequently, the use of a constant value for the strength reduction factor is only
appropriate for predicting the limiting endurance stress in the presence of
notches. Secondly, the strength reduction factor is not a material property and
must therefore be experimentally determined for each design, a costly process.
Clearly, the strength reduction factor must be in some way related to the radius
of the notch, r, the stress concentration factor and some property of the material.
A further factor known as the notch sensitivity, Q, is given by the relationship,
Pilkey, 1997

1 K -1
Q= =-1- (2.56)
l+plr K,-1

where p is a material constant.

Consequently, using graphs of the notch sensitivity of metals as a function of
notch radius and stress concentration factor, it is possible to determine the
appropriate value for the strength reduction factor. A comprehensive collection
of stress concentration graphs for a broad range of geometry's and load
conditions can be found in Pilkey, 1997, and no respectable library is without a
copy.

41
FLEXURES

2.6.3 Effects of mean stress

In many applications, pre-loads due to fastening, internal residual stresses,
self weight or constant applied forces are unavoidable. Consequently, stresses
often vary about a mean value, given by

(2.57)

Under these conditions it is usual to use the allowable stress amplitude, Sa ,

as predicted from (2.52) for a given life as a function of the mean load given by

(2.58)

where S0 represents the equivalent stress at zero mean offset and is

commonly measured in standard fatigue tests, n is a constant with common
values being
1 for general use and this is known as the Goodman criteria
2 for a more accurate representation of experimental data for a
broad range of materials and is known as the Gerber parabola
1.5 is commonly used for steels
Sometimes, the ultimate tensile stress in the Goodman version of equation
(2.58) is replaced by the yield stress and this is known as the Soderberg criteria
which is also the most conservative estimate of the effect of mean stress.
It is assumed, and necessary, that stresses, even in notched regions, are
below the yield stress. Provided that the requisite conditions are not violated, the
above analysis will enable the computation of a reasonable expected fatigue life
for a component if the stresses are accurately known in advance. For a more
precise estimate of expected number of cycles to failure, a statistical approach
should be used to account for the scatter in experimental data and the reader is
again referred to Sherratt, 1982, or Zahavi and Torbilo, 1996, for a brief overview.

2.6.4 Damage assessment

Stress variations on a component are rarely going to be of uniform amplitude or
mean value. Consequently, it is necessary to assess the damage induced from an
arbitrary load cycle. To do so, it is first necessary to break the stress down into
individual cycles having individual values of mean and amplitude. From each of
these, it is then possible to determine the number of cycles to failure. Assuming
that each cycle contributes in a linear way to the total damage, we may compute
the damage of the ith cycle from the simple relationship

42
CHAPTER 2: ESSENTIALS

1
D =- (2.59)
• N,

where N, is the value for the number of cycles to failure for this particular
cycle
It is then a simple matter to sum over all cycles to produce an estimate of the
total damage, D,01 given by the Palmgren-Miner equation

Dror =L:-N,1
1
(2.60)

A value at which the component has reached the end of its useful life is again
in the hands of the designer although a total damage of 1 is reasonable and the
more conservative value of 0.3 also commonly used.
As a final word before moving on to new ground, it is important that the
counting of stress cycles is correctly evaluated. For example, there will often be
irregular stresses that vary rapidly about more substantial variations. Using a
routine that simply looks at adjacent peaks and valleys, larger fluctuations over
the longer time periods would be ignored. Such an approach would be
particularly erroneous because it would fail to include the stress variations that
induce the most damage. A more robust algorithm called 'rainflow counting'
should be employed to determine stress cycles from an arbitrary stress history.ln
this, cycles are identified using a peak detection routine that identifies the small
stress variations within larger peaks and valleys. Successive cycles are eliminated
from the stress history until the last cycle represents the maximum difference in
stress over the complete history. Again, more detailed explanations can be found
in the references.

2. 7 Bending of symmetric beams

2.7.1 The bending equation

So far, the evaluation of stress and strain has been restricted to a point on the
surface or within a component. A great many of the flexures presented in this
book will require the prediction of maximum stresses and strains as well as
deflections relative to a fixed frame. In particular, a broad range of flexures
utilizing deflection of slender, symmetric beams of uniform cross section can,
with relative ease, be either fabricated or machined from solid. Consequently, it
is necessary to review those techniques used for the prediction of stresses and
subsequent deflections of simple beam elements. Consider the slender beam of
figure 2.17 subject to a pure bending moment, M, , that induces a deflection
having a curvature, 1/ R. Clearly, this will result in a simple direct stress, u..,,

43
FLEXURES

acting in a direction along

the beam axis. For the beam
to remain stationary, it is
necessary that the total force
due to the distributed stress
y R is zero. Consequently, the
total sum of compressive

"
M stresses must equal the
tensile ones. Also it is
apparent that there will be a
compressive stress at the
upper surface and a tensile
X
stress on the lower one.
B D
Therefore, there must be a
Figure 2.17: A symmetric beam subject to a pure bending plane running along the axis
moment of the beam in which the
stress is zero, appropriately
called the neutral plane. If we were to take a cross section through the beam, the
neutral plane would intersect along a line in the z direction and this is known as
the neutral axis. For a symmetric beam (all the beams considered in this book)
planes parallel to the neutral plane represent contours of equal stress and,
therefore, strain. Looking at one such plane, a'a' in figure 2.17, it is readily
verified that the strain for a given curvature is given by

(R + y)oB- RoB y
&
JC-
-
RoB
--
- R (2.65)

Because this is a principal strain, the relationship between stress and strain is
immediately written

a
x
=E!...R (2.66)

Summing the forces across the complete cross section gives

(2.67)

From this we conclude that the neutral axis coincides with the centroid of area
of the cross section. It is also apparent that the total bending moment about the
neutral axis can be obtained through the summation of all products of stress,
elemental areas and the distance y from the neutral axis i.e.

44
CHAPTER 2: ESSENTlALS

(2.68)

Combining this with (2.66) produces the bending equation

a E M
-=-=- (2.69)
y R I

From this, we observe that the stress varies linearly with perpendicular
distance from the neutral axis of bending and that the curvature of the beam is
proportional to the applied bending moment and inversely proportional to the
flexural rigidihJ, EI. I is always the second moment of area about the neutral axis.
2.7.2 Deflection of beams

Rearranging equation (2.69) in terms of beam curvature gives

l M
-=- (2.70)
R EJ

For the coordinate system shown in figure 2.17, the curvature is positive and
can be calculated from the equation

(2.71)

For small deflections, the denominator of equation (2.71) is approximately 1

leaving the relatively simple relationship

(2.72)

Substituting (2.70) into the above gives the governing differential equation
for elastic deflection of symmetric beams

(2.73)

The notation explicitly infers the fact that we are considering the distribution
of moments about the z-axis along the axis of the beam. However, in future
equations it is assumed from the context of the discussion. Consequently,
subscripts will be dropped in future analyses.

45
FLEXURES

2.7.2.1 Sign convention for bending moments

In all of our subsequent analyses any beam will be assigned a right hand
coordinate system. To determine the sign of the bending moment, consider an
arbitrary position x along a cantilever beam subject to a concentrated moment at
the free end. It is relatively simple to determine the direction of the applied
bending moment at that position looking either to the left or the right. Imagine
that the observer has cut the beam at this position and is holding it together. Our
sign convention is
a bending moment is considered to be negative if it acts to induce a tensile
stress in the upper face of the beam
A simple example would be a cantilever beam with the left side fixed and
with a constant clockwise bending moment of magnitude M 0 at the free end.
Clearly this will be a negative bending moment from which (2.73) gives

d2y
EI-
dx2= -M0 (2.74)

Integrating this twice and substituting in the boundary conditions requiring

that both slope and deflection are zero at the fixed end on the left (at x = 0), after
a little rearranging, we obtain

(2.75)

where X= x/L
Equation (2.75) is plotted in
0 02 0.4 0.6 08 figure 2.18 in terms of the
Or---~~--~----r----T----, X dimensionless lengths and
·0 1 displacements.
·0.2
In fact, it is not necessary to
-0.3
integrate (2.74). It has already
·0.4
been shown that the curvature of
·05
a beam subjected to pure
·0.6
bending is constant and the right
Ely/M 0 L 2
hand side of (2.75) is simply the
Figure 2.18: Cantilever beam with a bending moment truncation of the expression for y
applied at the free end as a function of x for a small
portion of a large circle i.e.
2 2
1 M0 x
Ely = - R(l- cos(x I R)) ~ - - - - - (2.76)
M0 2

46
CHAPTER 2: ESSENTIALS

As a slightly less obvious example

consider the cantilever beam of figure
2.19. In this example it has been
assumed that self-weight cannot be
ignored. The self-weight is given by a
,1-----------' ---~- force per unit length, q0 , which is
x simply the product of gravity
L acceleration, density and area of cross
Figure 2. 19: Cantilever beam subject to self- section. For the purpose of computing
weight load the bending moment, the distributed
load has been lumped to a point load
at its center of action which is half way from the observer at x to the free end of
the beam. With gravity, the force due to self-mass will act downwards resulting
in a clockwise, applied bending moment to the right of x. Being of negative sign,
the moment at position x is given by

(L- x)
2

M(x) =- qo 2 (2.77)

Substituting this into (2.73) gives

Eld2y =- (L-x) 2
dx2 qo 2 (2.78)

Integrating (2.78) twice with respect to x

El (L -x) 4
- y=- +CI x+C2 (2.79)
qo 24

Substituting the boundary conditions of zero slope and deflection at the fixed
ends enables the
o 0.2 o4 os o.8 determination of the two
0
d constants after which the
0
•.;::J ·0.02 X expression for deflection is
(.)
Q
!;:l -004 given overleaf by
0
""0
<ll -0.06
<ll

-a0
0
·0.08

-~ .0.1
0
.§ ..0 12
""0
-01 4

Figure 2.20: Deflection of a cantilever beam Wlder a

unifonn vertical load

47
FLEXURES

(2.80)

In this instance it is less clear that the beam is sagging under its own weight,
but this is readily verified by looking at the plot of this equation in figure 2.20.

2.7.3 Moment, shear force and rate of loading relationships

In the above example for a uniform cantilever beam subject to its own self-weight
loading, it was assumed that the load was distributed evenly and, therefore, the
load per unit length could be represented by the constant value. Considering the
case of a constant distribution of load, q say, acting in the positive direction, it is
relatively simple to show that the total force acting at a position x is
L
V = Jqdx =q(L-x) (2.81)
X

Considering now the bending moment due to this shear force. Starting at the
fixed end and moving in a distance x, the bending moment is simply the product
of all incremental moments along the beam from x to the free end. Consequently,
the moment for a constant load per unit length is

M = (load at centroid)(distance from point x to centroid)

= q(L - x)(L - x) =q(L-x)

2 (2.82)
2 2
Comparison between (2.82) and (2.81) implies the relationship
L
M =- fvdx (2.83)
X

Equations (2.83) is generally valid for all beams and can be written in the
form

dM d 3y
V = - - = -El - 3 (2.84)
dx dx

and

(2.85)

48
CHAPTER 2: ESSENTIALS

In summary, the relationship between displacement, slope, bending moment

and shear force is given by
Deflection y
dy
Slope
dx
Eld2y
Bending moment
dx2 (2.86)

Shear force V =- dM =-EldJy

dx dx 3

Rate of loading q=-dV =Eld4y

dx dx 4
This leads to a generalized method for the solution of beam deflections in
terms of the load distribution and is outlined in the section below. It is important
to note that the equations for deriving moments, slopes and displacements do
not require any changes in sign when integrating the rate of loading equation
(2.85). Normally, for the purpose of deriving the shape of a beam, it is necessary
that the equation satisfies the boundary conditions at each end (or at each
support/ constraint). The simplest and often most usable conditions are those
having zero value (i.e. zero displacement ~ pinned, zero shear ~ free, zero
moment~ clamped).

2.7.4 Singularity functions

Straightforward integration of the moment distribution along a beam will result

in the deflection shape of a beam for any load conditions. However, in many
instances, the beam will be supported at discrete points, or subject to bending
moments from attachments occupying a short span of the total beam length.
Under these conditions it can be assumed that the load or bending moment
appears at a point along the beam. This can be included in the bending equation
and integrated to determine the displacement. Consider for example the simply
supported beam of figure 2.21 with a pure applied bending moment at mid span.
Being a clockwise moment of magnitude, M, it is obvious that the beam will be
bent in some sort of 'S' shape. A
problem arises when we try to express
the moment along the beam. More
_J
..
importantly, it is even more difficult to
L/2
.. ..I L/2 ... envisage, let alone model, detailed
mechanisms in these small regions.
Figure 2.21: A simply supported beam with a
For example, any pure bending
centrally applied pure bending moment moment acting at a point will induce
an infinite shear stress. Clearly, the

49
FLEXURES

brackets and connections around this point will locally distort, a discussion of
which is outside the scope of this text. A powerful method for solving such beam
bending problems was originally developed by an engineer named Macauley
and is sometimes named after him. Today, this is more commonly referred to as
the method of singularity functions. In this, a singularity function is defined so
that; it is always zero if its contents are less than zero, is undefined if its contents
are zero and is equal to its contents when they are greater than zero. A
singularity function is discriminated by enclosing it in sideways vee brackets, i.e.
(contents of singularity function are here). This makes it possible to describe a
broad range of load conditions that are easily integrated in compliance with the
rules outlined below.
Table 2 (p58) shows the various expressions for the rate of loading on a beam
due to a variety of applied loading conditions. It will be worth spending a little
time looking at the rather unusual properties of the singularity functions as they
appear in this table. When the singularity function is raised to the power 0, it
does not become unity nor does the integral of tJ'te integral of the singularity
function raised to the power -1 become a logarithm. Because of these integral
properties, the reader is advised to proceed with great care when using
singularity functions. Another small but important point, the keen observer will
have noticed that in the last row the expression for the singularity function at x ~
a becomes a regular equation, a point indicated by the change in parentheses.
Returning now to our problem of the simply supported beam with a central
bending moment, the rate of loading equation can be easily written as

(2.87)

Integrating (and noting that the -oo or even zero limit makes all singularity
functions zero)

(2.88)

Again, integrating

(2.89)

At the right hand support the condition for zero bending moment gives

(2.90)

From which

(2.91)

Equation (2.89) can theref~re be written

50
CHAPTER 2: ESSENTIALS

(2.92)

Taking moments about each end could have more easily solved for the
moment equation (2.92). Substituting the above into (2.86) and setting a = L/2
gives
d2
El- f = M(x)= M <x-L / 2 > 0 -Mx I L (2.93)
dx

Integrating twice for the vertical deflection

< x - L/2>2
Ely = M -Mx 3 !6L+C1x+C2 (2.94)
2

From the boundary conditions of zero deflection at either end the constants
can be eliminated to yield

(2.95)

Ely/ML
2 A dimensionless
0.01 deflection along the
beam is plotted in
0.005 figure 2.22 Although
a relatively simple
model, it is applicable
for the modeling of
-0.005 many flexure
mechanisms and
-0.01
variants on this are
Figure 2.22: Deflection of the simply supported beam shown in discussed in more
figure 2.21 detail in chapter 5.

2.8 Torsion

In addition to pure bending, elements within flexure mechanisms are often

subject to torsion forces. In fact, we have already seen that any generalized
applied force can be reduced to simple direct plus shear stresses. Additionally, if
a body is subject to an arbitrary but small force, this can be resolved into forces
and torsional moments. Using the principle of superposition, it is then possible to
analyze the distortions due to each load in isolation and sum these to derive the
complete distortion due to the original load. Consequently, we shall look only at

51
FLEXURES

applied shear stresses due to pure torsional loads and leave it to further chapters
to demonstrate the use of superposition to determine full solutions for practical
load cases. For the purposes of this book it will be adequate to restrict attention
to the analysis of shear stresses and subsequent distortion due to pure torsional
moments applied about the axis of prismatic beams of circular or rectangular
cross-section.

2.8.1 Torsion of a prismatic beam of circular cross section

In the case of a pure torsion applied along the axis of a beam of circular cross
section, the shear stress is zero along the axis and increases linearly to a
maximum at the outside edge resulting in an angular distortion of a square
element on the surface of angle

r= -
1'mtx

G
(2.96)

The angle of twist of the cross section, B, is simply the integral of the angular
distortion along the beam so that

() = f11ix
o
= 1' max L
G
(2.97)

Exploiting the linear relationship between shear stress and radial distance
from the axis r and the fact that the angle of twist of the cross section is a
constant, it is therefore possible to derive the shear stress on the cross section at
any radial position r by the equation

-r GB
-=- (2.98)
r L

Based on the above it is a simple matter to derive the torsional load to

produce this deflection given by

= 1tR
4
T GB = JGB (2.99)
2 L L

where R is the outer radius of the rod, Jis the polar second moment of area and T
is used to represent the torsional moment about the x-axis to distinguish it from
the bending moment, M , in the previous analyses.
Combining equations (2.98) and (2.99) produces the general torsion equation
for beams of circular cross-section
T GB -r
-=-=- (2.100)
J L r

52
CHAPTER 2: ESSENTIALS

For a beam of tubular section with inside and outside radii, R, and Ro
respectively, this equation can still be used with the modified polar second
moment of area given by

(2.101)

2.8.2 Torsion of a prismatic beam of r ectangular cross section

For a rectangular bar the shear distribution over the cross~section is not as
straightforward. Solutions for the maximum shear stress and applied torque are
derived using more advanced elasticity theory, see for example Timoshenko and
Goodier, 19701.
For a beam of thickness, t, and width, d, the maximum shear stress in a beam
of length L subject to an angular twist Bis given by

1
-rm"" = - - -
2-
G~

L
8G~

1f L
L~

ttsl,3,S,
~2------~
n cosh(nmi /2t)
(2.102)
=kGfA
L

where k is a constant dependent upon the ratio dft.

For a thin rectangular strip, the ratio d/t is large and the above reduces to

GfA
'i
miX
=L- (2.103)

Similarly, it can be shown that the torque T to produce an angular deflection

Bisgiven by

T = t d GB(1_19; ~
3 L
3
~tanh(
1f d n~l.ml,S,. n 2t
f nnd))
(2.104)
3
-k GBt d
- I L 3

Combining equations (2.103) and (2.104) to eliminate B, the maximum shear

stress as a function of the applied torque is given by

I Note that in the Timoshenko and Goodier reference eis in units of radians per meter (i.e. atL in our text)
and t =2a and d• 2b

53
FLEXURES

0 2 4 6 8 10 12
dlt
Figure 2.23: Graph of the torsion constants for a rectangular bar

(2.105)

For convenience, all three of these constants are plotted in figure 2.23. All of
the above series converge rapidly so that, for computation of reasonably accurate
values, it is only necessary to use three or four terms in any of the ·series
expansions.

2.9 Mobility
Flexure mechanisms considered throughout this book will be modeled as a chain
of rigid links connected by flexible elements that are relatively compliant to some
forces or moments and very stiff in other directions. Consequently, flexure
elements can be modeled as connections between the rigid bodies that impose a
specific number of constraints. In mobility analysis, each of these connections is
called a joint. For example, a leaf type flexure connecting adjacent links might
produce large displacements of the links due to applied bending moments
whereas all other forces may have little effect. In this case, the leaf type flexures
would be modeled as 'ideal' hinges providing a single rotation freedom while
constraining relative motion between the links in all other 5 directions (i.e. 2
rotations and 3 linear translations). For simple mechanisms, the possible motions
of the links are obvious. For example, a simple pendulum consisting of a rigid
link attached to a rigid structure by a leaf spring would be free only to rotate
about its 'hinge' axis. Hence the pendulum is said to have a mobility of 1. As the
number of links and joints increases, the subsequent mobility of the mechanism
is less easily assessed. Using generalized methods of kinematic analysis relatively
simple equations for mobility analysis are developed below.

54
CHAPTER 2: ESSENTIALS

Consider first the analysis of planar mechanisms. In this case it is apparent

that a rigid link in a mechanism has only three degrees of freedoms
corresponding to rotation in the plane and translation in x and y directions (for a
detailed discussion of this subject see the book of Phillips, 1984). Consequently,
the mobility can be determined by first of all separating each link in which case
the freedom of the complete ensemble is simply 3 times the number of links, n,
minus the one that is necessarily fixed. Joining each successive Link together
through a joint, i, having between 1 and 2 freedoms reduces the mobility by the
number of constraints, c, imposed. This leads to an expression for the planar
mobility, M3 , given by

M 3 =3(n- l) - fc, =3(n-1)- ±(3 -/,)

i.r.
t•l /=I
(2.106)
=3(n-j-l)+
tal

where n is the number of links in the mechanism, j is the number of joints and f is
the number of freedoms provided by the ith joint. In figures the value of J is
indicated as a superscript and should not be confused with the normal
mathematical convention for raising variable to a power. For example, a simple
parallel hinge consisting of four links with four single degree of freedom joints
(hinges) can be readily shown to possess a single mobility. To consider three-
dimensional motion, it is necessary to consider each link possessing 6
unconstrained freedoms leading to mobility M 6 given by

Lf.
J
M6 =6(n- j -1)+ (2.107)
ta l

Subsequent analysis of our simple linear spring mechanism reveals a

mobility of -2 implying that such a mechanism is over constrained. This is indeed
the case and this was obviated in the planar analysis by the intrinsic assumption
that all of the hinges had axes parallel and normal to the plane of motion. Due to
finite manufacturing tolerances, this will not be possible and there will be some
out of plane torsional moments about two or more of the hinges.
A single degree of freedom mechanism is obtained if three further
freedoms are added. This could be achieved by replacing three of the one degree
of freedom hinges with those having two. In many practical situations over
constraint is unlikely to produce stresses that are unacceptably large and,
consequently, it may be assumed that planar analysis is adequate for planar
mechanisms. This being the case, consider a mechanism in which each joint
possesses a single degree of freedom. This could be a slideway (not necessarily
linear), a simple single degree of freedom joint or a mechanism composed from a

55
FLEXURES

subset of links and joints. Clearly, under these conditions, the sum of freedoms is
equal to the number of joints and equation (2.106) reduces to Grubler's equation

M=3(n-1)-2j (2.108)

Equations (2.106) to (2.108) are used extensively in chapter 5.

References
Howard E. Boyer (ed.), 1986, Atlas of Fatigue Curves, Metals Park, Ohio: American
Society for Metals
Brentnall W.D. and Rostoker W., 1965, Some observations on microyielding, Acta
Meta//urgica,13, 187-198
Dieter G. E., 1987, Mechanical Metallurgy, McGraw-Hill, NY, chapter 4.
Ford, H., 1977, Advanced mechanics ofmaterials, in four parts ; part IV with the
collaboration of J. M. Alexander, 2d ed., Halsted Press, NY
Hazen R. M., 1993, The New Alchemists; Breaking Through the Barriers ofHigh
Pressure, Times Books, NY
Hertzberg R. W., 1989, Deformation and Fracture Mechanics of Engineering Materials,
John Wiley and Sons, NY, chapter 1
Lawn B. R., 1991, Fracture ofBrittle Solids, Cambridge University Press
Pilkey W.O., 1997, Peterson's Stress Concentration Factors, 2"d ed., Wiley and Sons
Inc., NY.
Phillips J., 1984, Freedom In Machinery, Cambridge University Press.
Popov E. P., 1976, Mechanics ofMaterials, Prentice-Hall, Inc., NJ
Sherratt F., 1982, Fatigue life estimation: A review oftraditional methods, J. Soc. Envir.
Engs., December, 23-30
Shigley J. E. and Mischke C. R., 1989, Mechanical Engineering Design, McGraw-Hill,
NY
Smith S.T., Chetwynd D.G. and Jackson D., 1991, Manufacture of large scale devices in
·single crystal silicon by high speed grinding, SPIE, 1573, 53-61
Spain I.L., 1987, Semiconductors at high pressure: New physics with the diamond anvil
cell, Contemp. Phys., 28, 523-546
Timoshenko S.P. and Goodier J. N., 1970, Theory of Elasticity, McGraw-Hill, NY (note
that in both Timoshenko and this book the total shear strain is used. Many other
texts use tensor, or index, notation that leads to symmetric matrix representations
that are often mathematically concise and computationally more efficient. However,
in tensor notation 2&1) =riJ i ¢j, see chapter 1 section 7 of this reference)
Timoshenko S.P., 1983, History ofStrength ofMaterials, Dover Publications, Inc., NY.
The discussion of distortion energy theory appears on pages 368-372.
Zahavi E. and Torbilo V ., 1996, Fatigue Design; Life Expectancy ofMachine Parts, CRC
Press, Boca Raton

56
 FLEXURES

Table 2.1: Strength characteristics of some metals

Material Condition Tensile strength Yield stress Endurance limit
(MPa) (MPa) (MPa)
Steel alloys (endurance limit based on 10 million cycles) I l
1015 Cold drawn - D-/o 455 275 240
1015 Cold drawn - 60% 710 605 350
1040 Cold drawn - 0% 670 405 345
1040 Cold drawn - 50% 965 855 410
4340 Annealed 745 475 340
4340 Quenched and 1950 1640 480
tempered (204 C)
4340 Quenched and 1530 1380 470
temperedj427 g_
4340 Quenched and 1260 1170 670
tempered (538 C)
HYI40 Quenched and 1030 980 480
tempered (538 g__
D6AC Quenched and 2000 1720 690
tempered (260 C)
9Ni-4Co-0.25C Quenched and 1930 1760 620
tem_l)_ered_(315 g__
300M 2000 1670 800

Aluminum alloys (endurance limit based on 500 million cycles) I I

1100-0 90 34 34
2014-T6 483 414 124
2024-T3 483 345 138
6061-T6 310 276 97
7075-T6 512 503 159

Titanium alloys (endurance limit based on 10 million cycles) I

Ti-6AI-4V 1035 885 515
Ti-6AI-2Sn-4Zr- 895 825 485
2Mo
Ti-5Al-2Sn-2Zr-4Mo-4Cr l 1185 1130 675

CC>{'Per alloys (endurance limit based on 100 million cycles) I I

70Cu-30Zn brass Hard 524 435 145
90Cu-10Zn 420 370 160

Magnesium alloys (endurance limit based on I 00 million cycles) I 1

HK31A-T6 215 110 60-80
AZ91A 235 160 70-90

57
FLEXURES

Table 2.2: Macauley functions describing the rate of loading along a beam
as a function of the load type

Load type Singularity function

Pure bending moment M acting at a position a M(x-a)-2 = 0 x:t:a

-
" 2
J<x -ar dx = (x- ar'

(x-ar2 = ±oo x =a
Concentrated force R acting at point a and commonly used R(x - ar ' = o x:t:a
to model a reaction force

-
X

J<x -ar' dx = (x - a) 0

1
(X- ar = ±oo X =a
Unit step at a corresponding to a uniform load to the right x<a
of this point (x-a)o=LO
x~a

" 0
J<x-a) dx=(x-a) 1
....
Ramp starting at a
(x-a)1 = {0·· x <a
x-a x~a

- r
" 1 (x -a)2
J<x- a ) dx
2
Parabolic load starting at a
(x-a) 2 = x<a
(x -a)2 x ~ a
" (x a) 3
J<x-a)2 dx- -
- 3

58
 3 Rigid body dynamics

3.0 Overview

The objective of this chapter is to introduce sufficient mathematical tools to enable a

generalized approach to dynamic systems with sufficient breadth to include solutions
for most of the situations encountered in the analysis of flexure systems.
Consequently, it is already assumed that the reader is familiar with the basic
concepts of dynamic systems and it is hoped that this chapter will provide a
comprehensive and concise approach to dynamics that will build on techniques
already developed during a typical undergraduate science based degree.
It almost goes without saying that the two most important characteristics of a
flexure system are motion repeatability and dynamic response. Synonymous with
dynamic performance is the appearance of the first mode natural frequenet; when the
flexure provided with an impulse which, in the absence of external energ1; sources,
will eventually deem;. By monitoring the periodic oscillation and the rate at which
the StJSlem eventually comes to rest, it is often possible to deduce the dynamic
performance of the flexure.
To a large extent the prediction offree oscillation offlexures undergoing relatively
small displacements with negligible energy dissipation will fonn the focus of the
proceeding pages. This simplifies the analysis in two important ways. First, the
small oscillations will, in general, not influence the overall geometry of the flexure
system. Subsequently the equations governing motion reduce to simultaneous,
second order, linear differential form. Second, because most systems do not dissipate
significant energ1;, the natural frequencies of the damped and undamped system are
likely to be comparable. Consequently, for many purposes in which the fundamental
frequenet; is of prime importance, a reasonable estimate can be determined from an
analysis of an undamped system. Such an approximation avoids the complexity
associated with forces due to energy dissipation, which, in all but the simplest forms,
considerably increases modeling complexity.
To derive the equations governing nwtion offlexure systems, use will be made of
the equivalent, energ1;-based approaches known as Hamilton's principle and
LAgrange's equations. Though abstract in their derivation, it will be found that these
readily produce equations of motion for both continuous and lumped models of
arbitrnn; complexih;. For StjStems involving more than two degrees of freedom
LAgrange's equation are of great utilih;. For small amplitude vibrations, the
equations of motion invariably reduce to simultaneous, linear, differential form that
FLEXURES

can be represented in matrix form. Solutions to these equation.s have formed a field of
study for hundreds of years and can be readily computed using eigen-analysis. To
this end, matrix techniques for analysis of multi-degree of freedom systems wiU be
presented.
In many instances, the equations of motion wiU become too complex to be able to
produce simple analytic expressions for the natural frequencr;. Under these
circumstances it may be adequate to estimate the frequency based on an assumed
deflection shape. This is known as Rayleigh's method and has been shown to provide
a reasonable estimate for the natural frequency provided that tire estimated shape
satisfies, at least most, and desirably all, of the boundary conditions. Rmjleigh's
method has two important features. Firstly, the solutions often resolve to a simple
formula from which numerical values for the first mode natural frequencrJ can be
computed using a calculator or spreadsheet. Secondly, although the solutions can
only claim to be approximate resulting in a slightly high value for the natural
frequency, for any reasonable guess for the mode shape, the errors are usually within
a few percent. In most designs, errors of manufacture and, in particular, assembly,
lead to significantly larger variability in frequency. Consequently, the 'approximate'
equations are alnwst always sufficiently close to the 'real' value for engineering
purposes.
At the end of this chapter, a case study considers a simple, two-freedom, planar
flexure. Using the tools of this chapter, it is shown how geometric and applied
damping parameters may be selected to produce a required frequency response.

3. 1 Generalized coordinates

In the following section it will be shown that all reversible mechanical systems
can be described in terms of a simple principle known, rather obscurely, as the
'principle of least action'. Before moving on to this, it is worthwhile reviewing
the concepts of generalized coordinates. First, consider a system of N
infinitesimally small particles, the position of each can be determined in
rectangular, or Cartesian, coordinates from the values

x,,y,,z, i = 1, 2, ... N (3.1)

Alternatively, we might wish to express the positions of our particles

through different coordinates of the form

(3.2)

For example, for a single particle, it might be more desirable to use spherical
coordinates. The relationship between the two systems is given by

60
CHAPTER 3: RIGID BODY DYNAMICS

x =rsinBcos¢
y = rsinBsin¢ (3.3)
z =rcosB
In general

(3.4)

For a system of free

particles 3N is the number of
degrees of freedom. Clearly, to
control the motion of a
mechanical system, there must
be some constraints. Consider
an idealized molecule made up
from two rigidly connected
atoms a fixed distance a apart,
m2 figure 3.1.
Figure 3.1: Two rigid point masses connected by a rigid Because of the rigid
massless link of length a coupling between the two
masses, there is an additional
constraint such that

(3.5)

Consequently, if five coordinates are known, the sixth can be readily

determined. Those having previously studied kinematic design would consider
this a six-degree of freedom rigid body in free space. It must be remembered,
however, that point masses do not have rotational inertial properties and
therefore rotation of the atoms about the axis of the link can be ignored. In reality
it would be necessary to know the rotation of any finite mass about this axis to
fully characterize this system dynamically (because these are point masses,
rotation about this axis involves no momentum). To save further labor over such
semantic terminology, upon the addition of a further atom to our molecule, an
additional coordinate would be necessary to determine the positions of all
particles and this would then become a rigid body with six degrees of freedom.
For engineering purposes, a rigid body will be assumed to consist of a large
number of rigidly connected atoms, therefore requiring six coordinates to
describe its position in space.

61
FLEXURES

Returning to our simplified rigid bodies, it can be stated that, in general, if

there are m constraints, the number of independent coordinates and, therefore,
degrees of freedom n can be determined from the equation

n = 3N - m (3.6)

Consequently, if the q's are to represent independent generalized

coordinates, equation (3.4) can be written in the form

(3.7)

In subsequent discussions, only holonomic systems will be considered.

Under such conditions, n represents both the number of degrees of freedom and
the number of coordinates necessary to specify the potential and kinetic energy
of the system. Exceptions to this rule usually involve rolling systems (e.g. a
sphere rolling on a smooth surface has three rotational degrees of freedom but
will require five coordinates to specify its position, Pars, 1961).
Let us now consider a single particle whose coordinates are given by

X= fx (q ,,q2,q3)
y =/ y(q .,q2,q3) (3.8)
z = f . (q ,,q2,q3)

To apply the principle of virtual work, consider an incremental displacement

of the coordinates to q, + &],, q2 + &}2 and q 3 + &}3 • This can be transformed to
an increment in rectangular coordinates using the equation

n=3 (3.9)

similarly for y and z.

In general for an arbitrary number of particles

(3.10)

62
CHAPTER 3: RIGID BODY DYNAMICS

o is used to indicate a virtual displacement. Although this must satisfy the

constraints of the system, it does not necessarily conform to the equations
governing actual motion. An actual incremental motion would be designated
dq,. Apart from this distinction, o is considered to act as a normal differential
operator.
Consider the necessary force, F acting on a single particle to impart a virtual
I

displacement OR . The work done, W, for such a displacement is

OW =F. OR = FJix + F/iy + FJiz (3.11)

From (3.9), this can be expressed in terms of the generalized coordinates in

the form

(3.12)

Consequently, it can be seen that the generalized force can be obtained from
forces is given by

(3.13)

Where each of the forces corresponds to that which is applied in the direction
of its subscript and results in a virtual displacement measure in the snme
direction as indicated in equation (3.12). As a further extension, let us assume the
work done can be expressed in terms of some differentiable commodity, U, that
is a function only of the displacements i.e.

OW = oU (3.14)

where

(3.15)

'Qifferentiating this

(3.16)

If this function is work being done by the system, it will be negative and is
denoted V given by

63
FLEXURES

V =-U (3.17)

Vis called the potential energy of the system and, from equations (3.12) and
(3.16), this component of the generalized force is given by

ou
Q =-= - -
av (3.18)
• c;q. c;q.
This is of fundamental significance in mechanics. A general response to the
question, 'what is force?' could be
'Force is a consequence of, and proportional to, a potential gradient and
always acts in the direction of steepest gradient'
This can be extended to different topics such as; heat transfer where
temperature is the potential, fluid mechanics where pressure is the potential and
even diffusion where the potential is concentration (although in this case the
'force' is due to 'random walks'!).
This generalized approach to modeling is often referred to as Fields Theory.
For our purposes, it is only important to recognize that
Q. is the generalized force acting to produce a c1umge in the coordinate q• as
is cottsistent with the definition itt equatiott (3.12)
Before moving on to consider mechanics from Hamilton's perspective it is
worth recapitulating some mathematical tools.

3. 2 Properties of variational operators

3.2.1 Commutation

It can be shown that, Lanczos, 1970

d(o/) = dy
0 (3.19)
dx dx

also

oJJ(x)dx = J<r<x)dx (3.20)

3.2.2 Minima of a function L = f(q,q,t)

Consider the general function

L =f(q,q ,t) (3.21)

64
CHAPI'ER 3: RIGID BODY DYNAMICS

a - = small increment

Figure 3.2: Deviation of path corresponding to a small increment in the

Lagrangian oL as system transforms from states A to B for a particle
Assume that the function remains a fixed relationship of the three variables
but undergoes incrementally small changes as it transforms between states A and
B caused by correspondingly small increments in the two variables q and q .
Consequently, the deviation from the actual path can be expressed by the
relationship

oL = f(q + T?(t),<i + iJ(t),t) - f(q ,q,t) (3.22)

Assuming that the system is a point mass (of unit value, although other
values can be included with little added complexity, Longair 1984) moving over
time between positions a and b corresponding to times t 1 and / 2 • This is
illustrated in figure 3.2 in which the function L is identified as the Lagrangian
and is the difference between kinetic and potential energies of the system ( L = T
- V). It is now possible to state Hamilton's principle
For the true, or qctual, motion ofa system, the time integral of the Lagrangian
of the system as it transforms from fixed state A to another B will always be a
minimum
Mathematically this can be expressed as

J J
6 Ldt = oLdt = 0 (3.23)

To fully appreciate this remarkable theory the reader is recommended to

read Feynman's enthusiastic account of his personal discovery of the simplicity
and universality of this principle, Feynman et al., 1964.
Proceeding, it is possible to expand the expression of equation (3.22) using
Taylor's (named after Brook Taylor, 1685-1731) series and, ignoring second order
and higher terms in 11 (which we assume is small anyway), (3.23) is given by

65
FLEXURES

(3.24)

To eliminate the fluxion of the function T/ we can integrate the right hand
term in parentheses by parts to give

'f-.
•if . =[if- .rJ]'' - f (if)
i]dt
12
d -. T]dt
dt
'• aj aj '• '• aj

f- (if)
(3.25)
=-
,, d
- - :-
,, dt aj
T]dt

It is clear from figure 3.2 that the deviation from the path 17 is always 0 at the
two end points, thus justifying the disappearance of the definite integral above.
Substituting (3.25) into (3.24)

- - -d (Cf))
0 fLdt = '!(Cf
12

----:- 1]dt
,, ,, cq dt aj
(3.26)
=''r( cq - dtd (oL))
,!l 8L aj 1]dt = 0

Because 17 is arbitrary, and, for conservative systems, the forces are always
either inertial or potential, Lagrange's equation for a system with no externally
applied forces is simply given by

(8Lcq _!!_(oL))
dt aj
=O (3.27)

3.3 Hamilton 's principle

Starting off with the original formulation of Hamilton's principle, equatjon (3.23)
can be expressed by the equation, Goldstein, 1980,

ofLdt = foLdt = f(bT- OV}it = 0 (3.28)

It is most important to realize that this analysis is derived without any

consideration of virtual work. The principle of least action can be considered as a
fundamental property of dynamic systems that will always follow the path that
minimizes the exchange between kinetic and potential energy.

66
CHAPTER 3: RIGID BODY DYNAMICS

Considering a system of masses constrained to motion in a single linear

coordinate then the work done in changing the kinetic energy of a system is

(3.29)

Or in the limit

(3.30)

where m, is the mass per unit length along the direct x.

Substituting (3.29) into (3.27) and multiplying through by -1 it can be seen
that

Jm,yaydx + OV =0 (3.31)

Alternatively, an expression for the kinetic energy may be written in the

form

T =- 12: mx ·2
(3.32)
2 ' '
'
For a small increment in x to x +ax equation (3.32) becomes

(3.33)

Substituting OF from equation (3.33) into the equation (3.28)

J(or - 6V)it = J(~m,x&-ov}t

(3.34)
= J(l:m,x d::- OV}t
'

r
Integrating the summation by parts yields

J(~m,x ~ - }t 5V =[ ~m,i& +( - ~m,x& - 5V}t

I
(3.35)
= ( - L:m,i&- OV)dt = 0
I

67
FLEXURES

Equation (3.35) is D' Alembert's principle, Lanczos, 1970.

Example 1: A simple spring mass system

As a first example, a simple spring mass system shall be considered. For such a
system, it would be far quicker to apply Newton's laws directly. However, this
will serve to illustrate some of the basic steps in the application of Hamilton's
principle.
As usual, the spring is considered to have insignificant mass while the mass
is rigid. Additionally, it will be assumed that the motion of the mass is
constrained to motion in a single linear direction x for which the potential and
kinetic energies are given by

T =..!.mx 2
2
(3.36)
1 2
V =- lex - mgx
2

Finite variations of the energy terms can be readily obtained

T +OT =..!.m(x +axf

2 (3.37)
~ T +miliX

Similarly,

OV ~ kxlix - mglix (3.38)

From Hamilton's principle the motion of the system is governed by

J(or - cW}it = f(miliX- kxlix +mglix}it = o (3.39)

Again, to eliminate the fluxion of x, the first term in (3.39) can be integrated
by parts so that

f(miliX - kxlix + mglix)it =[miaxt: - fmiaxdt- f(kxlix- mglix}:it

(3.40)
= J(-mi -kx +mg}lixdt =O
By definition, the boundary term of the integral vanishes at the two limits.
Because lix is arbitrary, it is necessary that the expression in parentheses in (3.40)
vanishes over aU times so that the equation that must be satisfied is

- mX - Icx +mg = 0 (3.41)

68
CHAPrER 3: RIGID BoDY DYNAMICS

This result will be familiar to anyone who has taken an elementary course in
mechanics.

Example 2: Lateral vibration of a bar

This is a more complex situation, but nevertheless can be similarly treated. The
beam will be considered to have uniform cross-section and be subject to small
displacements.
The potential energy in an element of length dx will be equal to the work
done in distorting it to a radius R by bending the element through an angle B.
Consequently, the potential can be obtained from

v = r )-BMdB= L El ds
2 2R R
(3.42)
2)2
~ ~~/ ~; dx
I (

where BM is the bending moment and ds is the length of the element, which is
considered to be equal to x in view of the straightness of the beam, and the radius
has been approximated for small slope.
For an element of length dx the kinetic energy is comprised from two
separable components, these being due to linear and angular velocities. For
planar motion

(3.43)

where j is the radius of gyration of the element, which is a constant for our
prismatic beam (i.e. a beam of uniform cross section), and m is the total mass of
the beam. For present purposes, it will be assumed that the contribution due to
angular rotation can be neglected, thus reducing the kinetic energy term to the
first of the integrals in (3.43).
Again, the variation of potential can be readily deduced from the equation

V +OV =~I J(?(~:&)rdx

(3.44)
~ V + EI ( ~; ~?)dx

69
FLEXURES

Integrating the above equation by parts twice, the variation of potential is

given by

(3.45)

After integration by parts, OJ' is

I
m Jyaydx (3.46)
' 0

Consequently, the variational equation of motion is

Because the boundary terms must separately vanish (3.47) can be split into
the equation which must be satisfied at all points along the beam

(3.48)

with the boundary conditions

(3.49)

For example, a completely free beam will have an arbitrary slope and
deflection at the free ends. Therefore, the boundary conditions given by equation
(3.49) are

[EI~; I =0
(3.50)

[E~ ~n: =0
The two terms in (3.50) correspond to zero bending moment and shear force
at each end of the beam. T~ provides 4 constraints from which a complete

70
CHAPTER 3: RIGID BODY DYNAMICS

solution for the motion of the beam can be obtained. From (3.49), it can be seen
that the boundary conditions depend on combinations of specified displacements
and three successive derivatives. Separately, these dictate the independent
conditions corresponding to displacement, slope, bending moment and shear
force. These correspond to pinned or clamped (zero displacement), clamped
(zero slope), pinned (zero moment) and free (zero moment and/ or shear)
constraints respectively. In general, conditions for which solutions can be readily
obtained can be drawn from any combination of these at either end of the beam.
Solutions of the free vibration of beams for these boundary conditions are given
towards the end of this chapter.

3.4 Lagrange's equation

D' Alembert's principle to can be extended to be expressed only in terms of

generalized coordinates. Using methods of calculus, it is possible to equate . the
inertial term with an equivalent formulation involving only the kinetic energy.
Dropping the vector notation from our previous analysis, we can write out
d' Alembert' s principle for N particles of mass m, can be written in the form

f (m, (x,&, + .Y,~. + z,&,)) =f(

~I ~
F,, &, + F,, ~. + F•.&,) (3.51)

For a system with n degrees of freedom, there will be the functional

relationships

x, = /, (q l>q2, ... q.)

y, =g,(ql>q2, ... q.) (3.52)
z, =h,(ql>q2, ... q.)

In general, a small increment in one of the generalized coordinates /Jq, can

be related to the Cartesian coordinates from the generalized relationships

(3.53)

and so on for y and z.

Similarly

. dx, q; . q; . q; .
x, = - = -q,+...+-q, + ...+ - q. (3.54)
dt cq, cq, t9.
and so on for y and z.
Differentiating (3.54) with respect to q, immediately gives

71
FLEXURES

(3.55)

Because the generalized coordinates are independent (3.54) can be

differentiated with respect to q, to produce

ac, a.:,
-=- (3.56)
aj, ltj,

If only the s11• coordinate is varied, the virtual work done by the system due
to the inertia of the masses will be given by

..
N
l:(m,(x,&, + ji,o/, +z,&,))
,
_f( (·· tZJ
t•l
-+ y.., -
- L...J m, x,if. tg,+ z
I
.. , -
tZJ s
az, ))JV.
tZJI
'-":ts (3.57)

= - Q,/Jq,

Canceling IJq,

f (-m, (x,..
Q, -_ L...J if. .. tg, + z,..
+ Y, az, ))
t• l ltj, itJ. itJ. (3.58)
= S,

which is intrinsic to the concept of virtual displacements.

Because equation (3.58) contains both sets of coordinates, it is of little use. To
eliminate the Cartesian coordinates system, the concept of kinetic energy, T, is
introduced. In Cartesian coordinates, for a point mass, this is defined by

(•2
x, + y,·2+z,•2)
N
T =L...Jm,
" (3.59)
1• 1

Differentiating Twith respect to q, gives

-.
iJl' =L...J
aj,
f( (. itJ.if.
t•l
. -+z
- m, x, -+y, . ,- ~.
itJ.
az~))
itJ.
(3.60)

72
CHAPfER 3: RIGID BODY DYNAMICS

Differentiating again with respect to time yields

-
dt
8/')- ~ ( (·· tZJIq;
d ( - - L-. m x - + y - + z -
ql
.. ~. ..
tZJI
at,))
tZJI

,(.X, ~ (!J + Y, ~ ( ZJ + z, ~ ( :J))

t• l I I t I

...+ t.(m (3.61)

=-S1 + L-. m, ( x,-

t• l
~(
. d -
df tZJ s
(cr. ) .
d (-
+ y,-
df tZJ1 d/ tZJ1
~.)
+ z,. -d ( -<11, )))

Equation (3.55) can be used to transform the last summation in (3.61) to give

!!_( ~} =-s. + ±(m,(.x, (if.)

dt Cti. tZJ,
+ y, ( 0:,} + z, ( ;i,, )))
t•l tZi. tZJ, (3.62)
= -S +
I tZJI
8/'
Equation (3.62) can be simply rearranged to yield

-S =!!_( 8/') - 8/' (3.63)

• dt Cti. tZi.
This has produced the desired effect of reducing the equation for the
generalized inertial forces in terms only of the total kinetic energy of the system
and the generalized coordinates. It is only a trivial step to introduce this back
into d' Alembert' s relationship in general coordinates, to produce

L:
n (
1• 1
Q -d-
I dt Ctis
8/')
( - --
lZis
8/') =o s =1, 2 ... n (3.64)

From (3.18) forces due to potential V, are given by

Consequently, the externally applied forces can be split into those due to
potential and externally applied forces yielding

L:" ((-
s• l
d(-8/') -tZJ1iJl'- +tZJ1fJV)- =Q1)
dt Ctj,
(3.65)

73
FLEXURES

This is called Lagrange's equation after its originator Joseph-Louis Lagrange

(1736-1813).
Because the forces of potential are functions of position only, it is possible to
define the Lagrangian L = T- V and write the above in the abbreviated form

2:
n ((d(OL)
••, dt Ctj,
OL) = Q•)
- - - it!.
- (3.66)

For systems undergoing free vibration, there are no externally applied forces
leading to the apparently simple equation

2:
" ((-
d (-
iJL )- -
•• , dt Cti.
iJL ) =o)
it!. .
(3.67)

This is, of course, identical to equation (3.27) based on Hamilton's principle

3.4.1 Rayleigh's dissipation function

In the above equations it has been assumed that work done during motion of
components of the system is due to changes in conserved quantities of energy.
However, in all engineering systems there will be additional work due to
mechanisms that dissipate energy. Generally, these involve frictional forces due
to relative sliding of solids or shearing of fluids. Unfortunately, in all but the
simplest cases this results in phenomena that cannot be simply modeled. By far
the simplest dissipation mechanism is that due to forces in proportion to the
relative velocity between two bodies. Under these circumstances, Rayleigh
demonstrated that there is a dissipation function, D, of the form

(3.68)

where the b's represent the damping force proportional to relative velocities
between coordinates i and j. Under these conditions the equation governing
motion of the system becomes

2: - 8]') -8]'- +iJD- +iJV)

d (-
n ((
- =Q ) (3.69)
dt Ctj
l•l it! Ctj it!
I I S S S

It is important to note that D is always positive as are each of its coefficients.

This damping mechanism is characteristic of energy dissipation due to laminar
shear of a massless, vist;:ous fluid and is consequently referred to as viscous
damping. It should also be noticed that the dissipation function has the same

74
CHAPTER 3: RIGID BODY DYNAMICS

form as the kinetic energy and this is certainly the most frequently assumed form
'in books at any rate' to quote Rayleigh.

3.4.2 General use of Lagrange's equation

Determination of the equations of motion of an arbitrary system requires the
following four steps
1. Decide on the most appropriate or meaningful coordinate system
2. Determine the potential and kinetic energy and the dissipation function
for an arbitrary displacement
3. Determine the generalized forces applied to the system
4. Substitute the above into Lagrange's equation
The advantages of this approach are due to the relative simplicity that the
scalar energy functions can be determined. There is also little room for error
which often happens when attempting problems using vector and/or free body
diagram approaches. The disadvantage of this approach is the difficulty of
selecting the 'best' coordinate system for the problem.
Once the equations of motion have been derived they can be integrated
either numerically or analytically to determine the subsequent velocities and
displacements for each coordinate.

3.5 Linear systems theory

3.S.J The simple, single degree of freedom, linear, spring-mass-damper system

Any arbitrary linear system can be considered as the superposition of simple

single degree of freedom systems (this statement will be justified shortly).
Consequently, an understanding of
the general characteristics of a single
degree of freedom spring-mas!r
damper system provides an intuitive
feel for the more complex behavior
and analytic techniques for handling
systems having multiple degrees of
m
freedom. To derive some more
complete understanding of general
linear systems, it is informative to
Figure 3.3: A single degree of freedom analyze the simple idealized spring-
spring/ mass/ damper linear system mass-damper model of figure 3.3.
The equation governing motion
q1 of the mass m due to an applied
force F(t) is

75
FLEXURES

mij1 + bq1 + kq 1 = F(t) (3.70)

Assuming that the applied force F(t) repeats after some time T (which might
be a very long while) we might consider it to be periodic and we can therefore
express it as a Fourier series of the form

F(t) =a 0
+ f (a, cos(m,t) + b, sin(m,t))
••I
(3.71)

ao is a constant over time and therefore only contributes a static load or

deflection. As long as the flexure to which displacements are imposed is
operating within its elastic limits, this will simply add to dynamic deflection and
any subsequent stresses. Because we are only considering dynamic effects, this is
not relevant to this study (this is not to say that it isn't important!) and will be
ignored in all subsequent analysis. So, in general, the input is assumed to be of
the form

F(t) = L(a, cos(m,t) + b, sin(m,t))

(3.72)
=L(F. cos(m,t + 9',))
l•l

where the magnitude of the forceF, and phase 91, at frequency m,can be derived
from

F, =~a,l + b,l
(3.73)
91, =tan-1 ~
a,

and

i7r
OJ, =T- (3.74)

Without any justification, let it be assumed that, after all of the transient
effects have decayed, the system will settle to a steady state described by a
solution to equations (3.70) and (3.71) of the form

X 1(t) = L(H1F, cos(OJ,t + 91:)) (3.75)

••I

where (for the time being) the H, 'sand 91: 's are also real constants

76
CHAPTER 3: RIGID BODY DYNAMICS

., I,. HOJ,b sin{OJ,I + ;; )

I
I
I
I
I
I
I
I
I
I
I

Hk cos(OJ,t + ;; )

Figure 3.4: Vector representation of the individual terms in equation (3.77)

Substituting equation (3.75) into (3.70) results in the rather cumbersome

expression

+
L[-OJ~ H,F,mcos(OJ,I IP) - OJ,b~~ sin(OJ,I 1p;) + +] (3.76)
,. , H,F,kcos(OJ,t + qJ, )- F, cos(OJ,t + qJ,)

Notice that the F, 's cancel so that we are left with the equation

+
L[H1 {-0J~ mcos(OJ,t IP) - ~,bsin(OJ,t +IP:) +] (3.77)
,. , kcos(OJ,t + !p, )} = cos(OJ,t + qJ,)

One method for a solution of the above equation is to draw each term in
equation (3.77) as a vector diagram as shown in figure 3.4. From this, and using
Pythagoras' theorem, the equation for the gain and relative phase difference
between the input and output can be determined from the equations

ln. I=((k-mOJn 2+(bOJY )"2 (3.78)

77
FLEXURES

¢,(m,)=rp,)- bm, )
• rp, =-tan -•( k-mm; (3.79)

These two equapons respectively represent the gain and phase response of
this system to an input force.

3.5.2 Some equivalent definitions of a linear system

The above analysis relies on four very important assumptions governing the
nature of this system. These are;
1. The system gain H. and relative phase lag ¢, are constant for any given
m,. H, and ¢, depend only upon the constant parameters m, b, k and
m, and are independent of the amplitude of the input.
2. The output is linearly proportional to the input at a given frequency.
3. The frequency at the output is exactly the same as that at the input and
depends only upon inputs at that frequency.
4. Alternatively, the above might be stated that the output at a given
frequency only depends upon inputs at that same frequency.
All of the above conditions individually define a linear dynamic system. At a
first glance it is not apparent that, for example, statement 4 is the same as
statement 3. It is worth spending a little time investigating why this is so. Let us
imagine a simple system with a sinusoidal wave at the input with an unknown
signal at the output. Clearly, if we were to amplify or attenuate this signal it will
emerge as a sinusoid, although of a different amplitude, of the same frequency
and will thus satisfy all of the above conditions. Let us now think of amplifying
the input but changing the gain a little with amplitude. If the input signal is not
changing sufficiently rapidly that we need to consider dynamic effects, this

=:;
~ 1----------~,c_ ___
0

input

Figure 3.5: Input/output relationship of linear and weakly non-linear

systems

78
CHAPTER 3: RIGID BODY DYNAMICS

might result in an input to output relationship shown in figure 3.5. Clearly, as the
input increases in amplitude the output starts to 'flatten' out. Under these
circumstances it is clear that the output will no be longer sinusoidal. There is
obviously another shape present. If this were of the same frequency, it could not
alter the shape of the signal but only change the phase. Therefore, the effect of
deviating even slightly from a linear gain is to change the frequency content of
the output which is simply a more involved way of restating condition 4.
Consequently, it is possible to use a black box approach to single degree of
freedom linear systems and represent any system by the diagram of figure 3.6.

x(t) =I IH(w,)A, cos{w/ +¢ +argH)

.....
1

H(iw) )51

Figure 3.6: Input-output relationship for a single degree-of-freedom linear system

It is also important to note that the relative phase shift between input and
output is independent of the phase of the input (condition 1). Consequently, to
derive system responses, it makes no difference what value is chosen for the
input phase. For mathematical convenience, the input phase will be assumed to
be zero.

3.5.3 Frequency response functions

Based on the above arguments it is possible to model the input force to a
linear system by the equation

F(t) =F0 cos(OJt)

It turns out to be a lot more convenient to express this in the complex

notation

F(t) =Re{F0 e"'" } (3.80)

where Re implies the extraction of the real part of the expression within the
parentheses.
Correspondingly it is necessary to seek a steady state solution of the form

x(t) =Re{H(ia>)F0 e'.. } (3.81)

The iOJ in parentheses indicates that H is a complex function of the input

frequency a>. Because it is a complex constant it must therefore be comprised of

79
FLEXURES

two constants (the real and imaginary values). Inserting equation (3.81) into the
governing equation (3.70) and canceling the complex exponents yields

Re{-m 2 mH(im)F0 (t) + imbH(im)F0 (t) + kH(i(J))F0 (t) = F0 (t)}

From this, the complex form of the transfer function, or frequency response,
is given by

1
H(im) =
k- mm 2 + iaJb

This can be rearranged into the more standard form

H(im) = 11 k . (3.82)
{()2) {()
(1- -(J) 2 +2~i -
n {() n

Multiplying the top and bottom of equation (3.82) by the complex conjugate
of its denominator gives

(3.83)

Substituting equation (3.83) into (3.81) yields the steady state solution

F. I k
x(t) = 2
° 2 Rel\f, C - iD)e' 01 }
C +D
= ~olk (CcosM+Dsin(t)(]
C +D 2
(3.84)
=.J F.0 tk rlc cos(M +tan -t (D I C))]
c2 +D2
=fF. iH(i(J))icos((t)( + rjJ((J)))

where

(3.85)

80
CHAPTER 3: RIGID BODY DYNAMICS

and

(3.86)

~ is known as the critical damping ratio which should be familiar to anyone

having attended an elementary course in vibrations. Substituting this back using
the equations

(3.87)

It can be seen that equations (3.85) and (3.86) are identical with equations
(3.78) and (3.79) as expected. Essentially, q is a measure of the closeness of the
damping value of the system to that for which the impulse response ceases to
contain oscillatory components. The critical value of the system damping at
which this occurs is b = be, at which point q = 1. In the absence of damping (i.e. b
= q = 0), the response to an impulse will result in a harmonic oscillation
continuing indefinitely. It should be emphasized at this juncture that critical
damping does not imply the optimal value. In practice, the 'best' value is going
to depend on the design application and can vary considerably. For example,
vibration isolators or tuning forks may require a very low value (q ~ 0.005 -
0.00001 or less) while averaging indicators could be designed for values q of 5-
100 or more. Control systems are usually designed to track or respond to an
externally applied input and for such designs it is common to choose a damping
ratio in the region of q = 0.4 to 0.8. Often, the compromise value of q = 11.J2 is
used as a design goal for optimal response. Not only is this within the range of
optimal response values, but it also happens to be at 45 degrees on a root locus
diagram,. see section 3.10. It is rare to require damping of q = 1 for two reasons.
Firstly, it is not a value that that provides the fastest response for a control
system and, secondly, it is a singular value. Therefore, errors due to tolerances in
production will move this either into or out-of this singular region.
Even in the absence of designed damping, all systems intrinsically possess
energy dissipation. These can arise due to a large variety of phenomena. These
include; scattering of elastic waves in solids at grain boundaries or interfaces,
internal energy dissipation as part of the materials properties manifest as a
hysteresis in the stress strain characteristic, see Lazan, 1968 and section 8.2.4.1,
contact with fluids or gases, etc. For simple mechanisms where no efforts have

81
FLEXURES

been made to a lter the energy dissipation, the damping ratio will often be in the
region~= 0.1 to 0.001.
The magnitude and phase between the applied force and output
displacement is simply the magnitude and phase of is determined directly from
equation (3.82). Because of its simplicity, this general method for deriving system
responses is adopted throughout.

3.6 Measuring the critical damping ratio

Before leaving equation (3.85), it is worth looking at some of the characteristics of

this second order response. In particular with many systems we are interested in
the damping ratio, ~. There are two ways of deriving this from the transient
response, these being the logarithmic decrement method or, alternatively, fitting
the decay characteristic to the transient solution of the second order equation
using a non-linear curve fit. From this it is possible to determine the two
parameters

(3.88)

The former of equations (3.88) is the inverse time-constant of the transient

decay envelope and the latter is the damped natural frequency (in units of
radians per second).
For steady state assessment, it is necessary to measure the response to an
input containing a continuous bandwidth of frequencies encompassing the
resonance. The resonance peak occurs at a frequency, (J), , , given by

(3.89)

For low values of damping it can be seen that the resonant, damped and
undamped natural frequencies approach the same value. Therefore the ratio of
the input to output displacements at the resonant frequency can be
approximated by canceling the left hand side of the denominator of (3.85) and
assuming the maximum amplitude occurs at a frequency coincident to the
undamped natural frequency to give

(3.90)

In principle, the above equation can be used to determine the damping ratio
simply by measuring the maxim~ amplitude at the natural frequency.
Unfortunately, this requires knowledge of both the applied force and the

82
CHAPTER 3: RIGID BODY DYNAMICS

stiffness of the system and is very sensitive to errors of the applied frequency. In
many cases either the parameters are not known or such a measurement is not
feasible. Alternatively, calculation of the damping ratio based on the width of the
resonance peak is often used. One method is to look at the width of the square of
the ratio (3.85) at half of the maximum height. Clearly, the square of the output at
any given frequency represents the power and therefore height of the resonant
peak at half of the maximum value is known as the half power point which is
given by

Q2 1
2) 2
(3.91)
2=( l -~ 2=( l - a 2)2 + - l
+ 4 ~2~ a
2

w2n w2n Q2

Expanding the denominator and rearranging yields a quadratic in a.2 given

by

(3.92)
1
~l± -
Q

Clearly there are two solutions corresponding to the frequencies either side
of the resonant peak at the half power points w1 and w2 say. Noting that these
frequencies will also be near to the undamped natural frequency, we can subtract
the two solutions to give

(3.93)

Consequently, it can be seen that, if a system has a relatively low damping

ratio, it is possible to determine an approximate value for this based on the
frequency difference at half the maximum square of the response function
divided by the frequency at the peak.
A final, and particularly powerful, technique for assessing the damping ratio
in the vicinity of resonant peaks can be derived from plots of the real and
imaginary components of the frequency response. For a single
spring/ mass/ damper system this is given by

83
FLEXURES

(3.94)

Clearly, C/E and -D/E represent the real and imaginary parts of the response
function.
Such a response function can be readily determined using modern signal
analyzers. For a simple single degree of freedom spring mass damper system, the
real and imaginary parts of the frequency response (3.94) are shown in figure 3.7.
From these two graphs it is clear that the damping in the vicinity of the resonant
peak can be determined from the characteristic shapes of these curves. The
simplest piece of information to extract is the undamped natural frequency. This
occurs at the zero crossing of the real component of the response function. To
derive more information about this system, it is necessary to assess the values of
the data at the peaks of each of these plots. For the real and imaginary parts of
equation (3.94) the positions of zero slope are given by

(3.95)

From the first of (3.95), the numerator is given by

3
E dC -C dE= 2w -Sq 2 ~+ 2ws _ 4w (3.96)
dw dw w2n w2n w6n w4n

Equating this to zero and solving this for w, the two frequencies at the peak
and valley of the real component of the frequency response are

(3.97)

From this equation it is relatively simple to show that

84
CHAPTER 3: RIGID BODY DYNAMICS

a)

Re{H(iw)}
6

4 11
~· f----~=0.05
2
.L:.

0 ~
I ·'"?
7
·2
~=0.5
-4
v
·6

b)
Im{H(iw)}
10
9
8
~= 0.05
7
6
5
4
3 +-- - · - - · -·-1- - -----. - .. - - -- - - ·-·- -
j~,
2 +-- - - ----lf- - " t r - - - -- - - - - - - - - -
1 1-----~ ~ ~ ·-·-·· - ---- ---- --- - · -

0
0 0.5 1.5 2 2.5 3
wl w,
Figure 3.7: Real and imaginary parts of the steady state frequency response
of a single degree of freedom spring, mass, damper system in which the mass
is subject to a sinusiodal force, a) real, b) imaginary part

(3.98)

Because the frequencies are exactly the same as those of equation (3.93), this
also provides the damping ratio in the same way as the haU power points. In this
case, the value for the damping ratio is independent of the actual values of the
peaks, a most important result. Additionally, assuming that the frequency
response is precisely determined, in contrast to the haU power point approximate
method, this equation provides an exact value for the damping ratio.

85
FLEXURES

Another method can be derived by substituting these frequencies back into

(3.94) after which the values of the real part of the frequency response are given
by

(3.99)

From which, the peak to valley distance of the real part is

(3.100)

At the undamped natural frequency (i.e. aJ =aJn), the amplitude of the

imaginary component of the frequency response is given by

(3.101)

Dividing (3.100) by (3.101) and rearranging, it is possible to compute the

critical damping ratio from

(3.102)

Note that this equation requires only the relative magnitudes of values of the
frequency response measurement. Consequently, it is only required that the
sensors used for measurement of the inputs and outputs have a constant gain (or
sensitivity) over the narrow band of frequencies about the resonant peak. They
s
do not need to be calibrated! Unfortunately, as tends towards around 0.5, it
becomes rather difficult to detect the peak values. However, reasonable values
are obtained for values of less than 0.4 which is typical of many mechanical
systems. For very low values of damping, it is sometimes difficult to discriminate
the values at the peaks leading, once again, to inaccurate results.
Before leaving this topic, it is worth mentioning that the peak of the
imaginary plot occurs at the undamped natural frequency, aJ4 , and not the
resonant frequency, aJrrs, given by

(l)d = aJn~~ - ~2
(3.103)
aJ,., =(l)n ~~ - 2~2

86
CHAPrER 3: RlGlD BODY DYNAMICS

At the peak, the amplitude of the imaginary term in (3.94) is given by

(3.104)

For small c; this, of course, tends towards equation (3.101)

3. 7 Genera/linear systems revisited

It is now possible to express mathematically the general input/ output

relationship for a linear dynamic system. Essentially, this represents any system
that can be described by a linear differential equation of the form

(3.105)

A more compact form of the above can be achieved using the convention
known as tensor notation. However, in most cases, the mechanical systems to be
analyzed result in simultaneous equations that can be expressed as matrices
having symmetric properties. Computational solutions to these have been
extensively developed and are easily implemented. For this reason, matrix
methods will be used in the following.
This general equation is for a system with i x j degrees of freedom. This can
be simplified to a single degree of freedom by eliminating one subscript and
making the other equal to one so that for a single degree of freedom system the
relationship between a single output and the input can be written in the form

"c -
dnx
£...... n dt"
n
="d-
dmy
£...... m dim
m
(3.106)

where c and d are real constants. This is the only equation that satisfies the
conditions for linearity given in section 3.5.
As a second example, consider the frequency response of a single degree of
freedom vibration isolation table. Assuming that we are interested in the
displacement response of the table to an input displacement of the foundation
y(t). In the absence of any applied forces to the mass, the equation governing its
motion is given by either

mX + b(i - y) + k(x- y) = 0 (3.107)

or

mX + bi + kx =by+ ky (3.108)

87
FLEXURES

This is clearly of the form (3.106) and, therefore, must be a linear differential
equation. Again to derive the frequency response in terms of the relationship
between x and y, an input may be assumed to be of the form

y(t) =Re{Y,e""'} (3.109)

with the subsequent response being given by

x(t) =Re{H(i(ll)Y,e""'} (3.110)

Substituting (3.109) and (3.110) into the equation of motion (3.108) and
rearranging we have

(3.111}

from which

H(iw) = X max = k + imb

Y, k- mw 2 + imb

l + i2~~ (3.112)
(lln
=---:;:----"-
aJ
(V 2
l - - 2 +i2~ -
{J) n (ll n

Skipping the previous steps, the magnitude and phase of the transmissibility
equation can be deduced directly from (3.112}

(3.113)

As a recap, the magnitude and phase of the transmissibility have been

derived from the general complex number relationship

88
CHAPTER 3: RIGID BODY DYNAMICS

z1 = a+ ib = r1 L81
z2 =c + id =r2 L82

~ = I!LIL(O,- 82)
z2 ;:;]

The transmissibility can be transformed to simple complex number form by

multiplying top and bottom by the complex conjugate of the denominator.
Subsequently, the expression for the magnitude and phase angle is

(3.114)

It is also noted that the relationship (3.114) is identical to that for the ratio of
forces transmitted to the foundation, F,, , by forces applied to the mass, F~, i.e.

X m.,. F,
- - =- (3.115)
Yo Fo

Plots of the transmissibility are shown in figure 3.8. It can be seen that
attenuation of the motion between the foundation and mass (in vibration
isolation the mass usually represents the platform that supports the instrument
or machine) only begins at frequencies (C) greater than .fi(C),. . The phase shift
initially tends towards a lag of 1r. However, as the excitation frequency increases,
so too will the forces due to viscous damping until eventually these become
dominant. Consequently, at very high frequencies input and output tend
towards a tr/2 phase lag, the more rapidly for higher values of damping.
Vibration isolation is always a compromise. On the one hand it is desired that the
vibrations be attenuated as rapidly as possible, indicating as low a value of
damper and natural frequency as possible. However, the resonant peak may
cause some problems and therefore impose a limit on the damping ratio.
Generally, vibration isolators with large mass, low support stiffness and
reasonably low damping are desirable. Unfortunately, this is further complicated
if there are resonant frequencies near to that of the isolator. This will be
discussed shortly.

89
FLEXURES

a) 12

-~ ::0.05

2
J; ~ ~:: 0.5
r-- ---..
0
0 2 3 4
(J) I (J)II

b)

o~--~~----~----~----,
-0.5 +---~'11-'------"----"'3_____;;4

,.... -1 i-----t.->~-- ~ = 0.5 - -

s -1 .51--------l~---======---
"0

j -2-f---~~---­
c.. -2.5 r------\\~......_==::;::::==--
-3
-3.5 ~ :: 0.05

Figure 3.8: Transmissibility for a single degree of freedom

spring/mass/damper system, a) magnitude response, b) phase
response(~ = 0.5, 0.3, 0.2, 0.1 and 0.05)

The above analysis is general to any single degree of freedom linear response
function and will result in an output response x(t) to an input of magnitude
F0 ((J)) at any given frequency, (J), that can be expressed in the form

(3.115)

A curious question with regard to the representation of a phase lag arises at

this point. For illustrative purposes let us assume that, at a particular input
frequency, the gain is unity and there is a negative phase angle of 1t/3 radians.
Plotting the input and output signals, figure 3.9 reveals that the output on the
time axis moves to the right and intuitively gives the impression of a phase lead.
However, if we place a point P on the input graph at time 11 we can see that the

90
CHAPTER 3: RIGID BODY DYNAMICS

1.5

-1 5
time

Figure 3.9: Representation of a cosine function with and without a

phase lag (tV =3.142 and the phase lag = 1r 13 )

output does not reach this point until a time f = (J I tV later. Consequently, a
positive phase angle corresponds to a phase lead and a negative phase change a lag.
The velocity of the output for the same input is simply obtained by
differentiating equation (3.115) with respect to time which is the only non
constant variable on the right hand side, so

x(t) = - F.,tVjH(itV)jsin(at +(J(tV))

= -F.,tVjH(itV~cos(at + (J(tV) -tr /2) (3.116)
= F.tVjH(itV)jcos(aJt + (J(tV) + 1r / 2)

Clearly from this equation it can be seen that the velocity of the output leads
the displacement by a factor of n/2. This transformation can be readily achieved
by defining the velocity frequency response H. (itV) by the relationship

H.(itV) = X;•x = itVH(itV) (3.117)

0

Correspondingly, it can be shown that the acceleration x for an input y is

related by the acceleration response function

H.,(itV) = X;ax = -tV 2 H(itV ) (3.118)

D

It can readily be shown that the acceleration leads the velocity by tr/2 and
displacement by a factor tr.

91
FLEXURES

3.8 Multi-degree offreedom linear systems

It has already been stated that, at a single frequency, the output of a linear
system is pr-oportional to the input and that the total output is the sum of
isolated responses to each input. This can be extended to an arbitrary number of
inputs and outputs. Using the Fourier series, each of the inputs can be
decomposed into a simple sum of separate frequencies. Based on the previous
analysis, we may assume that for each input there will be a finite linear response
at all of the outputs. Consequently, there will be a linear response function
linking each input to all of the outputs. This is represented diagrammatically in
figure 3.10 in which each output is independently and linearly related to all of
the inputs. The subscripts i and j relate to the outputs and inputs respectively.
Generally, for the mechanical systems being considered in this book, the total
number of outputs j will be equal to the number of degrees of freedom of the
system. Consequently, the response function H IJ (i(j)) relates the response x/t)
to a sinusoidal input y,(t), or F,(t) in many cases. The outputs from the system
can be simply obtained from the equation

(3.119)

where m is the number of inputs, n is the number of degrees of freedom of

the system and Ya are the inputs amplitudes (these can be forces or
displacements) for the coordinate i and frequency (j)t. In practice, the number of
inputs will nearly always be equal to the number of outputs. This corresponds to
there being an input at each degree of freedom of the system. Exceptions are rare.

F; (t) =f. F;, cos(OJ,t + rp

1•0
1,)
J•l 1• 0

[H(iw)] :

=t-o
F.,(t) i;F.., cos(OJ,t +q>.,1)
.. ., H

Figure 3.10: A 'black box' representation of a multi input/output linear system

At this stage, the reader is probably beginning to lose track of which
components of the frequency response relate to which combination of inputs and

92
CHAPTER 3: RIGID BODY DYNAMICS

outputs. Consequently, the meaning of individual terms tends to become

obscured by the density of the equations. In general, the concept is less complex
than the above mathematical representation appears. It is at this point that matrix
notation is more concise.
Normally, it is the gain between an input force to a coordinate j and the
subsequent output at coordinate i that is of interest. Fortunately, this can be
obtained directly from equation (3.119) expressed in matrix notation

Hn ((l)*) - H,,((l).J - Him ((l)k)

I I
{X((l).)}= H,,((l)k) H,m((l)k) {Y((l)* )} (3.120)
I I
Hnl((l)k) - HnJ((l)k) - Hnm ((l)k)

{in this text curly parentheses, or braces, indicate a column vector)

Each response function
H"((l)k) of equations (3.119)
and (3.120) gives the frequency
x, (t), F; (t) response for the output at j for
an input at i with all other
inputs being equal to zero. To
illustrate how various transfer
functions may be deduced,
consider the vibration isolation
table of the previous example,
only in this case with a further
platform mounted on top,
figure 3.11. In this particular
Figure 3.11: A vibration isolation table with a further
spring/mass/damper system mounted on the table example there are three
possible inputs, these being the
motion of the ground, and the two forces applied to each mass. Actually, this is
the same as a three mass system with one of the. masses being very heavy in
comparison to the others (it could, for example, be the Earth). The equations
governing motion of this system are

m1x1 +b1(x 1 -x2 )+ k 1(x 1 -x2 ) = F1(t)

(3.121)
m2i2 +b,(x2- x,) +b2(x2 - .Y) +k,(x2- x,) + k2(x2 - y) = F;(t)
The third equation is similar in form to the first of the above equations only
we can imagine that the mass term is so large that it dominates the whole
equation. For the first of our response functions it will be assumed that the input

93
FLEXURES

forces are equal to zero and our foundation is subject to a sinusoidal

displacement (after all, subsequent motions of the masses of the table are not
going to alter those of the Earth!) given by

y(t) =Re{ foe'ot} (3.122)

The responses are therefore given by

x1(t) =Re{H1y(ia>)Y0 e'O>t }

(3.123)
x1 (t) =Re{H2y(ia>)Y0 e'Q)t}

From equations (3.121)-(3.123), we can derive the two simultaneous

equations

- a> m1H 1y (ia>) +imb1 (H 1y(im)- H 2Y(im) )+k1(H,Y(ia>)- H 2y(im)) =0

2

1
- m m 1 H 1y(im) + imb1 (H 2Y(im)- H 1y(iw))+ imb2 (H1Y(ia>) - Y0 ). (3.124)
k 1(H 2y(iw)- H 1y(im) )+ k1 (H2Y (iw) - Y0 ) =0

This provides two equations and two unknowns. Dropping the iro terms in
parentheses for the frequency response functions H, the first of equations (3.124)
gives

(3.125)

Substituting (3.125) into the second of equations (3.124) and rearranging

yields

(3.126)

This can be rearranged further to provide the response functions

H 1y _ iw b1k2 +h2 k 1 - m1b2 w 2 )+k1k 2 -w 2 (m1k2 + b1b2 )

Yo- (m mw 1 2
4
-
2
iw m 1 b1a> + mb w
1 2
2
+ m 1b1m 2 -b2k 1 - b1k 2 ·)

-m (m1k 2 + m1 k 1 +m1k 1 +b1b 2 )+k1k 2

2
(3.127)
=_.,.:__:.....
A(a>)+ _iB(m)
__,_.:._
C(m) + iD(w)

94
CHAPTER 3: RIGID BODY DYNAMICS

and

k k - OJ b b +iOJ(b k +b k
2
--
H ly
=-=--=-----=--=---:....-=-=--"-=-
1 2 1 2 1 2 2 1)

Yo C(OJ)+iD(OJ)
(3.128)
= E(OJ) +iF(OJ)
C(m) + iD(OJ)

It is interesting to note that the numerators of equations (3.127) and (3.128)

not only contain lower powers of OJ than in the denominator but both the real
and imaginary terms generally contain smaller coefficients and will subsequently
change less quickly. Consequently, it is the common denominator of both
equations that is of interest. Before analyzing the complete response functions, it
is informative to look at the limiting case when the system has zero damping.
Under this condition and dividing through by the products of the stiffness of the
two springs, equations (3.127) and (3.128) simplify to

(3.129)

and

(3.130)

The most obvious thing about these new equations is that the complex
components of both numerator and denominator have disappeared.
Additionally, the denominator is a quadratic in m2 • Consequently, two
frequencies exist at which the denominator vanishes resulting in infinite
displacements at masses 1 and 2. If we denote these frequencies A- 1 and A. 2 , then
we may also rewrite the above equations in the form

(3.131)

and

95
 FLEXURES

(3.132)

These two equations are an indication that the transfer function of a system
of arbitrary complexity can be relatively easily represented if we have a
knowledge of the characteristic roots of the response function. This will be
looked at in more detail in the following section. However, for now, to avoid
misleading the reader, it is possible to make the statement that each term in
parentheses represents the product of a complex conjugate pair and this accounts
for the factor of two in the numerator. Finally, the values of the characteristic
roots of the denominator (i.e. A. 1 and A. 2 ) are frequently encountered in matrix
analysis and are called the eigenvalues.
Another fascinating consequence of equations (3.131) and (3.132) is that the
numerator of the former equation is equal to zero at a frequency, aJ given by

aJ = [£"
v-;;;; (3.133)

This corresponds to the mass of the intermediate table being stationary

relative to a fixed (or inertial) frame of reference while the second mass is
oscillating at its individual natural frequency as if the second mass where not
present. At first this seems to be a paradox because it is reasonable to ask how
the mass knew to resonate if the second mass isn't moving. The answer is that
the intermediate mass would move during the transient stage and will only
eventually settle to a stationary value after these effects have decayed (there is
always finite damping). Ultimately, in the steady state (which is the only
condition being studied), when the foundation oscillates at this frequency, the
upper mass will be applying an equal and opposite force on the upper side of the
middle mass to that force due to the distortion of the lower spring. In the absence
of damping, the two eigenvalues of this system are given by

m 2 +m ( -1 +-
--
k I k k
1 ±J
12 12 - 2 I 2
"1 •/1.2 - (3.134)

It is also worth noting that the product of these two eigenvalues is

96
CHAPI'BR 3: RIGID BODY DYNAMICS

Turning our attention now to the frequency responses of the two masses
having a force applied at each mass. For the first response we will assume F; (I)
and y are zero and therefore we have a single input of the form

F; (I)= Re{ F;e"'"} (3.135)

with the response

x 1 (1) =Re{H 11 F;e""'} (3.136)

Substituting equation's (3.135) & (3.136) into (3.121) and rearranging gives
the responses

H2y imb1 + k1
Ya =(m mm 1 2
4
-
2
im(m 2 b1m + m 1b2 m +
2
mbm
1 1
2
- b2 k 1 - b,kJ.)
-(J)
2
(m 1k 2 +m2 k1 +m1k1 +b1bJ+k1k 2 (3.137)
imb, +k,.
= C(m) + iD(m)

and

H 2y -m2 m 2 +im(b1 +b 2 )+{k1 +k2 )

Ya =(mmm 1 2
4
-
2
im m 2b1m + m 1b2 m + m 1b1m
2 2
- b2 k 1 - b1k 2 ·)

-m 2
(m 1k 2 +m2 k 1 +m1k 1 +b1b2 )+k1k 2 (3.138)
= -m2m 2
+im(b1 +b2 )+(k1 +k2 )
C(m)+iD(m)

Note also that, in the absence of damping, the displacement of the upper
mass can be zero if the input force is at a single frequency given by

(3.139)

Similarly, it is possible to derive the transfer function response to a force

applied at the middle mass
2
H 22 - m1m +imb1 +k1
=(m1m2m 4 2 2
(3.140)
F2 - im(m 2b1m + m1b2 m + m 1b1m 2 - b2 k 1 - b1k 2 ).)
-m2 (m 1k 2 +m2 k 1 +m1k 1 +b1b2 )+k1k 2

and

97
FLEXURES

H•2 io>b• + k•
(3.141)
F2 =(m1m2(i)
4
-i(i) m2 b1(i)
2
+m1b2 (i)
2
+m b -b k - b k
1 1(i)
2
2 1 1 2 ·)

-(i)
2
(m1k2 +m k +m k +b b )+k k
2 1 1 1 1 2 1 2

We can now write the general response matrix in the form

(3.142)

Out of interest, equations (3.140) and (3.141) are plotted in figure 3.12. The
eigenvalues, or the undamped natural frequencies, given by equation (3.134) are
computed to be 6.2 and 16.2 (rad s·l) respectively.
After having performed these rather lengthy and cumbersome algebraic
manipulations, an obvious question at this point is how to generalize this
analysis for systems having an arbitrary number of degrees of freedom and
inputs. Unfortunately, the answer is not as simple as one would hope.
Obviously, having obtained the general frequency response matrix (3.120), for
our particular example, given (3.142), the only remaining task is to determine the
frequencies and phases of the subsequent inputs. There are no simple,
generalized methods for the determination of the frequency response functions.
However, there are a number of short cuts outlined in the section below to enable
plots of the frequency response to be computed for linear systems of arbitrary
degrees of freedom. Also outlined are computational methods for calculation of
characteristic roots of the denominator {also called eigenvalues or, because the
frequency response shoots up to infinity for undamped systems, these are
sometimes called poles) and associated mode shapes (also called eigenvectors).

3.8.1 Note on the graphical representation of frequency response functions

Three representations of the frequency response are shown in figure 3.12. Others
could certainly be added. The objective of these graphical representations is to
present three pieces of data
1. The magnitude of the gain between the input and output
2. The phase lag between input and output
3. The frequency at which the above occur

98
CHAPTER 3: RIGID BODY DYNAMICS

Figure 3.12 a)

0.0001
0 10 20 30 40 50 60 70

b)

0.001

0.000001
0 10 20 30 40 50 60 70

In the first plot of figure 3.12, the gain is plotted directly as a function of the
frequency. However, the phase information is lost. Generally this should be
included as a second plot often placed directly below to enable both gain and
phase to be determined using a ruler or, at least, easily assessed 'by eye'. Often in
electrical applications, linear systems are designed in a modular form. By
insuring that the input impedance is high and output impedance is low,
connecting successive systems results in a straightforward passing of the output
from one system into the next. Consequently, the total gain of a system is simply
the product of gains for all sub-systems through which the signal passes while
phase shifts will simply add. Using a logarithmic scale for frequency and gain
enables simple assessment of the effects of adding sub-systems together. As

99
FLEXURES

e)

Hn(i(J))
0.04
0.02 p Re _ _ _ _ _ _ _ __ _ _ __

0 -
~0---....:> 20
' J.--.....:;.:.-~
-0.02 +--I 30_ 40
__:.:....__ 50 60 _ _:.70
:..:__ ..:..:..._

-0.04 +--+- - -- - - - - -- - - -
Im
-0.06 ~-~ - - - - -- - - - -- - -
-0.08

Figure 3.12: The frequency responses of the two masses of figure 11 to an input force at mass 2:
a) The response of mass 2, b) the response of mass 1, c) & d) are the corresponding polar plots, e)
real and imaginary parts of c) (parameters used for these plots are m1 =m2 =1 , k 1 = k 2 = I 00,
b1 = b2 = 20, 16, 6, 2,1 respectively

plotted, the effect of series connection of sub-systems can be simply assessed by

adding individual plots. If the gain is plotted in units of decibels this is called a
Bode diagram. Unfortunately, for many mechanical systems the effects of adding
further mechanical components will not correspond to a simple addition of
system responses. Under these circumstances this method is not valid. The polar
or Nyquist plots of figure 3.12 (c & d) indicate the gain and phase as radial
vectors. However, the frequency information is lost. Because there are three
pieces of information, these can only be represented in one diagram if it is three-
dimensional. Three-dimensional polar diagrams are sometimes used in which
the axes are the real and imaginary parts with frequency forming the third axis.
In such a three-dimensional plot, the frequency response wi11 be a string. The
main advantage of the 2-D polar plot is that the phase shift can be readily
evaluated irrespective of its value. For example, it can be seen from the figures
that the phase of mass 1 tends to a 360-degree lag at high frequency while that of
mass 2 tends towards 180 degrees. The real and imaginary plots of figure 3.12 e)
are really only plots of the string in the 3-D Nyquist plots when viewed
perpendicular to the frequency axis from the 'top' and 'side'. The advantages of
this plot are that the damping at the individual resonance's can be readily
assessed using equations (3.98) or (3.102).

3.9 General response function short cuts

100
CHAPTER 3: RlGlD BODY DYNAMICS

Figure 3.12(cont'd): c)

0
0.04
-o.01

-o.02

-0.03
8'
-0.04

-0.05

-0.06

d)

The first task is the derivation of a more direct method to obtain the numerator of
the transfer function. It is important to realize that, after substitution of our
assumed solution into our simultaneous differential equations, th a simple
simultaneous equation emerges of the form

[A]{H,} ={l} i = l.. ..n (3.143)

It is therefore possible to solve for each transfer function using Cramer's rule
given by the ratio of determinants

101
FLEXURES

lA, I
H. =IAI (3.144)

Unfortunately it is common to use the same convention for the determinant

of a matrix as for the magnitude of a complex number. Fortunately, these two
situations rarely cross paths and, based on the context of the discussion, it is
hoped that ambiguities will not arise. The upper determinant of equation (3.144)
corresponds to the minor of A with the ith column replaced with the input vector
containing a 1 in the position of the origin forcing function and zero's elsewhere.
It is worth reviewing at this stage the method of determining the determinant of
a square matrix. Taking a 3 x 3 square matrix of the form

(3.145)

The determinant of this matrix is obtained by summing the values of

successive sub-matrix determinants from the following four steps
1. Select a suitable row or column from the matrix (preferably one with lots
of zero's in it)
2. Select successive elements from the selected row or column and strike out
the row and column that it occupies
3. Find the determinant of the remaining matrix and multiply this by the
currently selected element
4. Finally multiply this by (-l)i+i. This is equivalent to multiplying successive
elements by alternating positive and negative number.
For example, selecting the first row for step 1 above, the determinant of the
above matrix is given by

(3.146)

This considerably simplifies the expression of the numerator of the transfer

function. For example, the above problem, figure 3.11, for the case of an input
force at mass 1 and ignoring motion of the ground, results in a simultaneous
equation that can be expressed in matrix form

- m1a> 2 + ia>b1+ k 1
[ - (ia>bl + k.)
or (3.147)
[A]{H} ={1}

102
CHAPTER 3: RIGID BODY DYNAMICS

Therefore, to solve for the numerator of H11 it is necessary to compute the

minor determinant

which is, of course, the same as the numerator of equation (3.138) only this
time more easily derived.
Similarly, the numerator for H 21 , also in accord with (3.139), is simply

(3.149)

Because we are considering linear systems, it is possible to derive the general

system response through consideration of individual ~esponses to a single input
at each coordinate. A rather striking feature of equation (3.144) is the fact that the
denominator will be the same for all of the frequency responses. As a
consequence it is possible to state the important lemma that
all denominators of the response function of a linear system are the same and
am be obtained from the determinant of the response matrix [A].
Clearly, this enables the derivation of a generalized approach to frequency
response calculation. Also it is apparent that the frequency responses computed
at any point to forces applied at arbitrary positions around the system will have
common poles. However, the number and frequency of zeros will vary at
different coordinates of the system.

3.9.1 Rayleigh's approach to the problem of computing generalized frequency

response functions
The use of determinants to derive a generalized form of the frequency response
that is both mathematically robust and readily computed did not escape
Rayleigh's notice. However, it must be acknowledged that, prior to Rayleigh's
exposition, a more generalized approach had been thoroughly investigated by
his "coach" E. J. Routh. To determine the frequency response to a generalized
dynamic system, it is necessary to return to the original rigid body approach to
dynamics. Assuming that we can split our system into discrete masses, springs
and viscous damping elements, through appropriate choice of the generalized
coordinates, q,, for small displacements, it is possible to write the energy
components of the Lagrangian in the form

103
FLEXURES

II

T = L:a,tj,2
•••
II II

D= LLbuqtqJ (3.150)
•• I J•l
II II

V =LLCuqtqJ
, • • J• l

It should be noted that

(3.151)

Substitution of these into Lagrange's equation immediately yields

a 1ij1 +b11 q 1 +b12 q 2 ... +c11 q 1+c12 q2 + ... = Q1

(3.152)
a2ii2 +b2,q, +b22q2 ... +c2,q,+c22q2 + ... =Q2

and so on for all other coordinates.

Assuming harmonic applied forces, and subsequent responses, the fluxions
in (3.152) can be replaced with im. Collecting all terms common to each
coordinate we are left with the simultaneous linear equation

e,,q, +e,2q2 + ... =Q,

(3.153)
e2,q, + e22q2 + ··· =Q2 etc.

where

e, =a,(im) 2 +b,(im)+c, (3.154)

Clearly, equation (3.153) can be expressed in matrix form

(3.155)

In the absence of applied forces, for a non-trivial solution it is necessary that

the determinant of e vanishes. It remains to link (3.155) with the general form of
a linear simultaneous equation given in equation (3.143). Denoting V as the
determinant of e, its minor is given by

104
CHAPTER 3: RIGID BODY DYNAMICS

av (3.156)
ae,.
A general solution to the simultaneous equation (3.155) is given by

(3.157)

etc

Setting all of the generalized forces to zero except for Q. say, the response of
the ith coordinate can be found from the equation

av
Vq, = -Q. (3.158)
oe,.
Rearranging the above, the frequency response function can be readily
recognized as

(3.159)

For those of a mathematical disposition, it is noteworthy that, by equation

(3.151), a principle of reciprocity exists for dynamic systems i.e.

H., (iOJ) = H ., (iOJ) (3.160)

The reader is left to show, once again, that equation (3.159) can be used to
reproduce the frequency response functions given by (3.140) and (3.141}.
The utility of equation (3.159) is readily apparent for systems of two or more
degrees of freedom. Once the equations of motion of a system have been
determined (using Lagranges equation, of course) it is no more than a collecting
of terms and programming a computer to evaluate the determinants for
individual frequencies of interest. From such a computation, the gain and phase
can be computed directly. An example of a six-degree of freedom system is
presented in section 3.11.5.

3.9.1.1 Example of a two degree of freedom system

For the two-degree of freedom system considered above, (3.147) gives

105
FLEXURES

-(icob. +k.)
- m2tV 2 +itV(b1 +h2 )+(k1 +k1 )
l
=i-m- tV.(tcobl+icob+ k.)+k
1
1
1 1 -(icob. +k.) I
-m1 tV 2 +itV(b1 +h2 )+(k1 +k2 )

From which the frequency responses can be computed from

I~ e•11

=I e••
e21
H,
e•1 1
e21 en

II., ~I
ell
H.,=
ell el2 I= H,
ell e12

~
11
'e

=J•"e21 e•1e22l"c
ell

H,

3.10 Eigen analysis

Fortunately, it has been the preoccupation of many mathematicians to be able to

express the determinant in the form of a series of simple products or,
alternatively, the sum of simple polynomials. Before looking at this technique, let
us back-track a little and return to a simple transfer function of the form

(3.161)

This can be factored to give the alternative form

106
CHAPTER 3: RIGID BODY DYNAMICS

(3.162)

Clearly, the two values for A. correspond to the roots or characteristic values
of the denominator of the frequency response. If we substitute the two
eigenvalues for a simple single degree of freedom spring/mass/damper system
given by

A.1.2 =- qa>n ± ia> n ~1- q2

(3.163)

we derive the frequency response function

H(ia>) = J..( 1
m a>?, - a>2 + i2qa>a>"
) (3.164)

This is exactly the same as the frequency response derived in equations (3.78)
and (3.95). Therefore we have two alternative forms of the frequency response
function in terms of the characteristic roots, or eigenvalues of the system. This
problem has been extensively studied and there now exist a large number of
software packages capable of computing the eigenvalues of a matrix. Although
beyond the scope of the present book, it is possible to produce the frequency
response of any linear system in the form, Newland, 1993,

(3.166)

Derivation of equations for d and c is a complex task and adds little to our
understanding of system behavior. There are two characteristics of the frequency
response that are of interest in design, these are
1. Values of input frequency at which the numerator becomes small
2. Values of input frequency at which the denominator becomes small
The first of these results in the magnitude of the frequency response tending
to zero and, as a consequence, these points are called zeros. The second condition
results in rapid increases, or spikes, in the frequency response and these are
consequently known as poles. It is the poles that result in large stresses and
deformations and their identification is of some concern to the designer. It is
clear from (3.162) or {3.166) that, for purely complex roots, the poles occur at the
eigenvalues of the system.
The remaining piece of the jigsaw puzzle for our multi-degree of freedom
systems is the computation of the characteristic, or, identically, eigen, values.

107
FLEXURES

3.10.1 Conservative systems

It can be shown that any linear mechanical system can be described by a
governing equation of 'free' motion in the generalized matrix form

(3.167)

Clearly, any solution to this equation will represent its transient behavior.
Again, a solution is assumed to be of the form

(3.168)

Substituting this into the above gives

..t2[m]{X} + ..t[b]{X} +[k]{X} = 0

(3.169)
= ..t2(m]+..t[b]+[k]

Clearly, there are only certain values for !.. which satisfy the above equation
and these are the eigenvalues of our system. As a first example, consider a simple
three mass system with no damping as shown in figure 3.13. The matrix
equations of motion for this system are easily derived and are given by

which is of the form (3.167) with [b] = 0.

Substituting (3.168) into (3.167) and canceling out the exponent gives

(3.171)

Multiplying through by the inverse mass matrix (which is a diagonal of 1/ m

terms if it is a diagonal matrix as for this example) results in an equation of the
form

[..t2[1)-[mr'ckJ]{x} = o
(3.172)
= (..t2(1)-[AJ]{X}

Apart from the trivial solution {X} = 0, solutions for ..t2 and {X} are called the
eigenvalues and eigenvectors respectively. Usually, if the system has more than
two degrees of freedom it is common to employ computer solutions for the
eigenvalues and eigenvectors based o~ the system matrix [A]. For equation
(3.168) to be valid, the eigeny~ue must be an imaginary number. Intuitively, this

108
CHAP'fER 3: RIGID BODY DYNAMICS

is obvious because, in the absence of damping, any transient motion will result in
a sinusoidal oscillation that will continue indefinitely. It is clear from the solution
(3.168) that this is only true if the eigenvalue is imaginary. Two eigenvalues are
possible for each solution corresponding to a positive and negative imaginary
number. This leads to an important conclusion that the solution gives two
eigenvalues that are complex conjugate pairs. Substituting these into (3.168) and
expanding for the ith degree of freedom and then and n+1 eigenvalues we have

x, (t) -_ X m eA• t +Xt+lne A•• ,, (3.173)

Associating the eigenvalues with the frequency of oscillation

A., =i(l),
(3.174)
A.n+l =-i(l),

and expanding (3.173) using Abraham De Moivre's theorem

x, (t) =X,, (cos (I) ,t + i sin(/),t) + X,+1 (cos (I) ,t- i sin (I) ,t) (3.175)

This gives the displacement of the ilh mass in terms of imaginary quantities!
Clearly, this cannot be representative of the real world and it is easy to show that
the corresponding amplitudes of motion, or eigenvectors, must, and do, also
occur as complex conjugate pairs for x(t) to be real.
For a non-trivial solution to equation (3.172), the determinant of the matrix
must vanish. This will reveal the roots of the characteristic equation which also
yield an expression for the denominator of the linear system frequency response
function.
As an example, consider the three mass system of figure 3.13 for the
parameters given by

x1 (t)

Figure 3.13: A simple three mass linear system without damping

109
FLEXURES

m1 = m 3 = 2m
m2 = m
kl = k4 = 3k
k2 =k) =k

]!x} lo}
After substituting the assumed solution and rearranging into the form of
equation (3.172), the matrix equation of motion for free oscillation becomes

(4k - 2mA.l )
-k (zk -- km;e) 0
-k =
[
0 -k (4k-2mA?)

Setting the determinant of the first term to zero will correspond to the
condition for a resonant frequency. In the absence of damping, this is the same as
the natural frequency. Expanding the determinant of the above matrix gives
2
(4k - 2mA.2 ) (2k -mA? ) - 2k2 ( 4k- 2mA? ) = 0

Expanding this further and dividing by 4m 3

yields

This can be factorized to give an expression for the determinant in the form

4m 3 (~ - -t2)(2 ~ - -t2)(3 ~ --t2) =0

= 4m 3 (~ - a)(~+ a)( g- i-t)(g +a)(g -i-t)(g +i-t)
(3.176)

Comparison of equation (3.176) with (3.162) indicates that the characteristic

roots can be found from the eigenvalues of the free vibration matrix [A] of
equation (3.172). In this particular instance, the constant cis simply the product
of inertial terms. As can also be seen, the eigenvalues occur in complex conjugate
pairs and in the absence of damping are purely imaginary. This system could be
represented by plotting the eigenvalues (or characteristic roots) on an Argand
diagram whereby it is obvious that they will all be situated on the imaginary axis
with no real parts. The values for the eigenvalues that satisfy equation (3.176) are
the characteristic values of the system anq these are given by

11 0
CHAPI'BR 3: RIGID BODY DYNAMICS

(3.177)

The frequency response of any mass to any steady state sinusoidal input can
be obtained from equations (3.162) and (3.177)

(3.178)

100

10

4
~ 0.1
a;_
0.01

0.001

0.0001
()) / ())n

Figure 3.14: Graph of the frequency response of the three mass system
of figure 3.13 (note that the resonant amplitudes have been suppressed)

It should be noted that the phase shifts by 1t every time the system goes
through a resonance and therefore it is more informative to plot the magnitude
of 1/IAI which for k = m =1 is shown in figure 3.14. As would be expected, the
magnitudes tend to infinity at the three natural frequencies. At frequencies
higher than the largest resonance, the amplitude drops off at rate approaching 18
dB per octave corresponding to the reciprocal sixth power law with frequency.

Ill
FLEXURES

Determination of the numerator of equation (3.178) is left for the following

example.

3.10.2 Systems with damping

In the previous eigenvalue analysis system damping has been ignored in order to
obtain the undamped natural frequencies of our system. These undamped
natural frequencies are also the eigenvalues and were found to be purely
imaginary. In reality there will always be some damping present and we must
therefore assume a system equation of the general form

[m]{x} +[b]{x} + [k]{x} = o

Again assuming a solution of the form (3.168), we seek a solution to the
equation

A?[m]{X} +A[b]{X} +[k]{X} =0 (3.179)

This can be rearranged to give

A[[m]{V} +[b]{V} +[k]{X} = 0 (3.180)

where

{V} =A{ X} (3.181)

which, by examination of equation (3.168), obviously corresponds to the

velocity eigenvector.
Equations (3.180) and (3.181) can be rearranged into the equivalent
eigenvalue problem

(3.182)

By defining

{Z} ={:} (3.183)

and

(3.184)

112
CHAPI'BR 3: RIGID BODY DYNAMICS

equation (3.182) can be written in the standard eigenvalue form

[A]{Z} =A.{Z} (3.185)

There are many software packages that will supply the eigenvalues and
eigenvectors for the [A] and {Z} matrices respectively. However, in this case both
the eigenvalues and eigenvectors occur in complex conjugate pairs. To see why
this is so, consider the following interpretation of these values.
3.10.3 Interpretation of complex eigenvalues and eigenvectors
From equation (3.168), it can be seen that if both A. and X are real, then, provided
that A. is negative, the solution for x(t) is an exponential decay. If the real part of
the eigenvalue (or characteristic root) is positive, the output will be exponentially
increasing with time which corresponds to an unstable system. This does not
occur unless there is some form of feedback or self-induced effects. When the
eigenvalues and eigenvectors are complex they will occur in conjugate pairs.
Therefore considering the ith mode transient response of the Jcth coordinate of our
system we have the solution

x.(t) =c"X"e).•' + ck+lX*+ 1eA.. ,,

(3.186)
=c"X"eA" + c;x;e.t;,
The constant c values depend upon the boundary conditions and, for the
purposes of illustration, these can be set at arbitrary values and unity is the
obvious choice. The eigen value can be described by a constant complex number
of the form

A., =-a+ iCtJ (3.187)

where a and CtJ are real positive constants. Inserting (3.187) into (3.186) and
expanding yields

(3.188)

Making X =A+ iB
x" (t) =e-(IJ (2A cos( ax)+ 2Bsin(ax))

=e-(IJ (2.J A 2 + B 2 cos(ax+ tan-•(B I A)) (3.189)

=2e IX" Icos(ax + LX")

-()II

113
FLEXURES

Consequently, the displacement of x is purely real as would be expected.

This can be represented on a phasor diagram as the sum of two equal vectors
rotating at equal velocity in opposite directions and each is shortening
exponentially at a rate

(3.190)

The reader is recommended to draw this. Plot the first vector rotating
clockwise with its tail at the origin of an Argand diagram (another name for a
real/ imaginary set of Cartesian axes). The second vector should be connected to
the tip of the first and rotating at the same rate in the opposite direction. Its tip
should be always touching the real axis and at the point given by equation
(3.189). This corresponds to the transient solution for the k1h coordinate of the
freely oscillating multi-degree of freedom mechanism.

3.10.4 Summary of primary steps in the derivation the response function of a linear
multi-degree offreedom system

The mathematical arguments result in a reasonably simple approach to the

computation of response functions. In general the response can be obtained
through the following steps
1. Derive the equations of motion using Lagranges equation or Hamilton's
principle.
2. Linearize equations by assuming small displacements.
Set up the equations of motion in the standard matrix form
[m]{x} +[b]{x} + [k]{x} = {f(t)}
3. Substitute in the trial solution

{x,(t)} ={x,eAI}

to derive simultaneous equation in {x} in the form

[A]{X} ={1}

4. Determine the numerator of the frequency response function for the

response of the k1h coordinate to input at j from the equation

where jA. j is the minor determinant of the matrix [A] with the J<th column
replaced with zero's, apart from a value of 1 at row j.

114
CHAPTER 3: RlGID BODY DYNAMICS

Compute the response at any frequency using the equation

H (i«>) =i!_ =_!_ oV

II Q, '\j 0e11

4. Finally, to determine the characteristic roots (eigenvalues) and mode

shapes (eigenvectors) of the system, rearrange the solution matrix in the
form of an eigenvalue problem of the form

(A.[/] - [A]]{Z} = 0

where

Use any computational mathematics package to determine the

eigenvalues and vectors.

3.1 0.5 Example: A series, six mass vibration isolator

Before moving on to continuous systems, an example of a multi-degree of
freedom vibration isolator is considered. The global interest in the detection of
the extremely low perturbations induced in solid bodies (the 'antennae') through
the effects of gravity waves propagated by the creaking together of distant
massive objects might provide some more clues to the nature of the universe to
which we are bound. Requiring the measurement of induced strains which could
be as low as 1o-2o (which would correspond to an extension of about one atoms
length in a bar that that is about the length of 1,700 earth's strung together), it is
probably an understatement to say that this effect is small. Clearly, it is not
acceptable to have our 'antennae' mounted on a platform that may be vibrating
due to ground and air-borne vibrations. Adding more spring/mass/damper
combinations to our isolator leads to successively higher and higher 'roll-off'
rates, or attenuation, in response to perturbations at frequencies well above the
highest resonance. It is for this reason that multi-stage isolators have received
considerable attention over the last decade. With this in mind, let us consider the
six-stage vibration isolation mount shown in figure 3.15. In our conceptual
example, we will assume that the final mass number 1 is the antennae and that
all inputs originate from the ground in the form of seismic disturbances. To
further simplify the analysis it will be assumed that each of the masses, springs
and damping elements are nominally identical.
Stepl

115
FLEXURES

Clearly, for this simple mechanism, the system

matrices are given by

m 0 0 0 0 0
0 m 0 0 0 0
0 0 m 0 0 0
[m]=
0 0 0 m 0 0
0 0 0 0 m 0
0 0 0 0 0 m

k -k 0 0 0 0
-k 2k - k 0 0 0
0 -k 2k - k 0 0
[k] =
0 0 -k 2k - k 0
0 0 0 - k 2k - k
0 0 0 0 - k 2k

b -b 0 0 0 0
-b 2b -b 0 0 0
0 -b 2b -b 0 0
Figure 3.15: A six-stage vibration [b)=
0 0 -b 2b - b 0
isolation mount
0 0 0 -b 2b - b
0 0 0 0 -b 2b

0
0
0
{F} =
0
0
Y,

where Y, corresponds to the seismic motion of the ground.

Step2
The numerator of the frequency response can be found from the determinant of the
following matrix

116
CHAPTER 3: RIGID BODY DYNAMICS

IAyl=
0 -(iOJb +k) 0 0 0 0
0 (-m(J) + iOJb + k) -(ia;b + k)
2
0 0 0
0 -(iOJb +k) -(ia;b + k) 0 0
0 0 -(iOJb + k) -(iOJb + k) 0
0 0 0 -(ia;b + k) -(iOJb + k)
fo(iaJb + k) 0 0 0 -(ia;b + k) ( - m(J) 2 + iOJb + k)

Step3
The determinant of this matrix is simply

IAyl= fo(iaJb+k)6(-l)s
6 4
= -Yo(-(aJb) +6i(OJb) 5 k+l5(aJb) k -20i(OJb) 3
2
e -t5(0Jb) e + 6i(OJb)k
2 5
+k6 )

Step4
This requires the reformulation of the matrix [A] for computation of the
eigenvalues and eigen vectors. This was carried out using MATLABTM for 1Jl = k
=1 and two values of b = 0.0, 0.4 and 0.1 respectively. The corresponding
eigenvalues were
Forb = 0

1.9419 -0.1885 ± il.9327

1.7709 -0.1568 ± i1.7640
1.4970 -0.112l±i1.4928
A.(b =0) = ,A.(b =0.1) =
1.1361 -0.0645 ± i1.1343
0.7092 -0.0251 ± iO. 7088
0.2410 -0.0029 ± i0.2411

- 0.7542 ± i1.7890
-0.6272 ± i1.656l
-0.4482 ± il.4283
A.(b = 0.4) =
-0.2582 ± i1.1064
-0.1006 ± iO. 7020
-0.0116 ± i0.2408

117
FLEXURES

It is worth checking that the product of squares of the eigenvalues yields the
appropriate value for the static response. In this case the output of the top mass
will be equal to the displacement of the base. It is therefore clear that the product
of squares of the eigenvalues should be equal to the ratio of products of stiffness
and mass which is equal to 1. A quick calculation will verify this.
StepS
The above values plus the product of masses (=1) have been inserted into
equation (3.178) and numerically computed using a spreadsheet that provides
the complex coefficients of the frequency response function.

3.10.5.1 Results

For zero damping the response function has six frequency's at which the
response will tend to infinity with a corresponding instantaneous phase shift of 1t
(i.e. a change in sign) as the frequency passes through each resonance.
Interestingly the eigenvalues and eigenvectors are useful for the identification
and visualization of characteristic mode shapes. The eigenvectors are

055 - 0.52 -Q.46 0.37 -o.26 - 0.13

0.52 -0.26 0.13 -0.46 0.55 0.37
0.46 0.13 0.55 -o.26 -Q.37 -0.52
{X}=
0.37 ' 0.46 ' 0.26 ' 0.52 ' -Q.l3 ' 0.55
0.26 0.55 -0.37 0.13 0.52 -Q.46
0.13 0.37 -0.52 -o.55 -0.46 0.26

The first three modes are plotted in figure 3.16. One can see the characteristic
shapes for each. The zero crossings are stationary over time and are the
characteristic nodes of
the system. The number
of nodes is directly

J
c.
related to the frequency
mode m. In this case,
for the mth mode there
~~~~~~~--~-=~~ will be (m-1) nodes.
1 +-~~~~~--~--~--~~----~-­ For a relatively low
] +-~~~--~~----------~~------ damping coefficient
corresponding to b = 0.1
mass the frequency response
function for the upper
Figure 3.16: First three eigenvectors showing characteristic mass is shown in figure
mode shape for the six-mass vibration isolator 3.17. The relatively high
Q value for the lower

118
CHAPTER 3: RIGID BODY DYNAMICS

log10 (X1 I Y) modes is readily apparent

2 r---~----~----~----~----. although at the higher modes
there is so much attenuation
that the final two peaks are
-2
difficult to discern even when
a logarithmic scale is used for
the magnitude of the
response. It is also interesting
that the attenuation occurs at
a relatively high frequency of
a little over 1.5 radians per
·101------..__--___.______.______.__ _ second (0.24 Hz).
0 02 0.4 0.6 0.8 For even higher damping
frequency (Hz) corresponding to a damping
Figure 3.17: Frequency response of the six mass isolator factor of b = 0.4, the frequency
forb = 0.1 response still shows peaks at
the first two fundamental
frequencies, but they are
almost impossible to discern
visually at the high mode
resonance's, figure 3.18. One
·1
method of visualization is
that of the polar, or Nyquist,
·2
plot in which the real and
imaginary components of the
response function are plotted
·5 on an Argand diagram,
figure 3.19. Interpretation of
~0~----~--~----~----~----~
o2 o4 oe oe this is simple, the magnitude
frequency (Hz) of the gain at any frequency
Figure 3.18: Frequency response of the six mass isolator on the plot is the length of a
for b = 0.4 straight line drawn between
the point and the origin. The phase angle is measured between the real axis and
the line from the origin to that point. At high frequencies, there is also a phase
shift of 1t between each successive spring/ mass/ damper stage. Consequently, for
the six-mass isolator, a phase shift of 6n would be expected. This is not readily
apparent unless the origin is magnified as in figure 3.20 in which it can be seen
that the Nyquist plot does indeed spiral around the origin the requisite number
of times.
It is interesting to assess the effects of adding damping to any system.
Choosing a six mass system where each mass is 10 kg, and with stiffness values
of 10,000, the eigenvalues for zero damping are

119
FLEXURES

4.450, 8.678, 12.470, 15.637,

18.019, and 19.499 (rad s·l)

Starting at zero damping

the eigenvalues all lay on the
imaginary axis. Increasing the
damping results in the
individual eigenvalues having
increasingly negative real parts
-12 with the imaginary component
reducing, see figure 3.21.
"1'!e' - --4' - - -_2. ___0. ___2.___4.___6.____,8
Eventually the eigenvalue will
real become purely real at which
Figure 3.19: Polar plot of the frequency response point the system is critically
of figure 3.18 damped. Interestingly, each
imaginary characteristic root traces out a
o05,, - - - - . - - - - . - - - . . - - - - . - - - - . -- ---, locus that is similar in form to a
single degree of freedom
0 system. For obvious reasons,
this plot of often referred to as a
roots locus.
-01

3.11 Vibrations and natural

-015
frequencies of continuous
systems
-02
In this section, equations for
-O~L.l-~0--0~1--0~.2-~0.~
3--0~4-~05 computing the natural
real frequencies of continuous
Figure 3.20: Amplified portion around the origin of the systems such as wires, beams
polar plot of figure 3.19 and plates that are common to
many flexures will be
presented. Invariably, Lagrange's equation and Hamilton's principle are used to
produce 'exact' solutions to these problems. Rayleigh's method, although only
approximate, produces reasonable estimates and often results in an equation that
can be computed on the 'back of an envelope'. In all of the following, it will be
assumed that distortions and, therefore stresses, are small. As a consequence, all
will behave in the manner of a linear system and therefore the principle of
superposition applies. Almost all of the following can be found in Rayleigh's
treatise, in most cases in a more compact and complete form.

120
CHAPrER 3: RIGID BODY DYNAMICS

3.11.1 Strings
15 Strings are probably the simplest
10 vibratory systems to analyze. In
5
most texts, this is discussed with
reference to plucking of the string
and, in some cases, analysis is
extended to include the response
to forced vibrations. Such
discussions are not relevant to
this text and the curious reader is
referred to the excellent
real
discussion given by Rayleigh ..
Imagine a string subject to a
Figure 3.21: Roots locus tensile force of magnitude To. In
the absence of gravity forces, the
string, when stationary will be straight. If it is then set into vibratory motion by a
force applied perpendicular to its axis, it will attain additional strain energy due
to the extension of the string in this deformed state. Assuming that the tension
does not change significantly the strain energy will be equal to the work done.
Choosing an element of length dx, the work done on the element as it increases
in length to ds is

(3.191)

For small slopes this can be reduced to

WD =To(dx-~(dxy +(dy)2)
(3.192)
~~(:rdx
The strain energy throughout the length of the wire can be obtained by
integration. Applying the variational operator to this integral gives

(3.193)

For a string of mass per unit length p', the kinetic energy of the string, T, is
given by

(3.194)

121
FLEXURES

From Hamilton's principle defined by equations (3.28) and (3.35), the

equation of motion to be satisfied at all points along the string is given by

(3.195)

Equation (3.195) is the one-dimensional wave equation with c being the

characteristic wave velocity. There is, of course, another independent wave
equation governing the vibration of the string in the z axis direction. For a
uniform string, this will be identical, with identical solutions. However, it should
be emphasized that the motion of a point on a string undergoing oscillations in
both directions will trace out an ellipse perpendicular to its axis. The general
solution to (3.183) is

y(t) = f(x-ct)+ g(x+ct) (3.196)

The above solution represents two arbitrary shapes travelling in opposite

directions with a velocity c. Shapes of the waves are determined by the arbitrary
functions f and g. For example, if z = x - ct then any function of z will satisfy
equation (3.195). Examples that the reader might try could be z", e' cos z, log z,
etc. If the two waves are of identical shape and amplitude, the subsequent
superposition of the two will result in a standing wave of the same shape.
A solution to equation (3.195) of the form (3.196) is given by

y(t) = C cos(m(x - ct)) + D sin(m(x + ct )) (3.197)

This can be rearranged to yield

y(t) = (Asin(mct) + Bcos(mct)XEcos(mx)+ Fsin(mx)) (3.198)

For a string held fast at each end, it is clear that E = 0.

Assuming a boundary condition of zero displacement at each end at all times
it is relatively easy to derive the condition

sinml = 0 (3.199)

From which

n1!
m=- n = 1, 2.. oo (3.200)
L

The characteristic frequencies of the string can be easily deduced as

122
CHAPTER 3: RlGID BODY DYNAMICS

n = 1, 2 .. oo (3.201)

Clearly, there are an infinite number of solutions for the free vibration of the
string. Equation (3.197) can be rewritten with separate time and displacement
terms.

y(t) = ~ --
f.tsm . (mrx)(
1
A, cos(mrct)
1
. (mrct))
- - + B, sm - -
1
(3.202)

Because the shape of the deflection curve is independent of time, as

expected, this represents a standing wave. Equation (3.202) constitutes a Fourier
series for an odd function that is periodic in l. Consequently, for any initial
displacement of the string, the constant coefficients may be determined and
subsequent motion of the string over time computed. Because the Fourier series
can be used to describe arbitrary distortions of the string, equation (3.202)
corresponds to the most general solution of this problem. Following Fourier's
original approach, the coefficients can be determined based on the initial
displacement and velocity along the string at time t = 0 given by

Yo= ..
L:A,sin(n1TX)
-
n•l I
(3.203)
1CC l:nB,sin-
Yo=- .. (n1TX)
I n• l J

Exploiting the orthogonal properties of equation's (3.203), the coefficients can

be determined by multiplying through by sin(n7TX//) and integrating over the
length of the string to give

Jy0sin(m1TX)dx = JfA,. sin(n1TX)srn(m1TX)dx

0 l o ,... I 1 l
LA,
=-
2 (3.204)
Jy sin(m1TX)dx = 1CC fnB,. sin(n1TX)sin(m1TX lAv
I r
0
0 I I l n•l

= tr:cn B
2 II

From which

123
FLEXURES

(3.205)

For example, for a string plucked a distance rat a position measuring b from
one end, it is can be shown that

(3.206)

Note that both coefficients can be zero if the point at which the string is
plucked corresponds to a node. Hence, if the string is plucked at a node, that
particular mode will not be excited. For example, a guitar string plucked in the
middle will produce a sound devoid of all even harmonics, a less pleasing sound.
The above analysis can be applied to all systems governed by a wave
equation.

3.11.2 Longitudinal vibrations of a rod

Instead of the Hamiltonian approach of previous sections, use of Lagrange's

equation can sometimes provide insight into the nature of vibrations of
continuous systems. In the case of longitudinal vibrations of a homogeneous,
prismatic beam, it is possible to split the beam into longitudinal elements of
length dx. Bach element will have a stiffness and mass given by

EA
k =-
dx (3.207)
m = pAdx

where A is the cross sectional area of the rod [m2], p is the density of the material
[kg m-3], and x is a distance measured along the bar. During longitudinal
vibration, each element will undergo a displacement u measured relative to an
inertial frame. For an arbitrary motion, the potential and kinetic energy in the bar
is given by

1 ..
V =- ~)(u, - u,_1 ) 2
2 1•1
(3.208)

124
CHAPTER 3: RIGID BODY DYNAMICS

Substituting equation's (3.208) into Lagrange's equation (noting that there

are two terms involving u, in the expansion for the potential energy), the
equation governing free motion of the bar is

pAiidx + k(2u, - u,+1 - u,_1} = 0

pAiidx + !: (2u, - u~+ 1 - u,_1 } = (3.209)

.. E (2u, -u,.1 -u,_1 } _
u+p (dx)2 - 0

In the limit as dx tends to zero (3.209) becomes the familiar wave equation

(3.210)

Here c= (E I p)Y. is the velocity of sound along the thin bar.

Assuming that the bar is undergoing free oscillation, in the absence of
damping, it will assume a continuous vibration of constant shape f(x) varying
harmonicaily with time. With this assumption, it is possible to immediately write
the solution

u =CJ(x)cos(m,t + fp) (3.211)

Substituting (3.211) into (3.210) and canceling cos(m,t} on both sides gives

(3.212)

For which the solution for f(x) is

(3.213)

To determine the natural frequencies corresponding to longitudinal

vibrations of a bar, it is necessary that the constants in (3.213) be chosen to match
the boundary conditions. Three examples that apply to some flexure designs are
given below.

125
FLEXURES

3.11.2.1 Longitudinal vibrations of a clamped-free rod

The boundary conditions require that, at all times, there is zero displacement at
the fixed end while the free end will be strain free i.e.

u = O@x = O
Ou
- =O@x=l
ox
From the first of these conditions /3= 0, while from the second we have

n = 1, 2 .. oo (3.214)
= 2n - {I
v-;;
(J) ln
• 21

Note that the period of the fundamental frequency corresponds to the time
required for a pulse to travel the bar four times.

3.11.2.2 Longitudinal vibrations of a free-free or fixed-fixed rod

For a bar in space, each end will be correspondingly free from strain giving the
boundary conditions

Ou
- = O@x=l,x=O (3.215)
Ox

Substituting these into the second of (3.213) gives /3= n/2 and also

(3.216)

From this the natural frequencies of the bar are

(J)
n
= nn fp{I
/
n = 0, 1, 2 .. oo (3.217)

A general solution for the free vibration of the bar is

u(t) ~co{mrxX
=~ - - An co{nnct) . (nnct))
. - - + Bn sm - - (3.218)
1 1 1

126
CHAPTER 3: RlGID BODY DYNAMICS

Comparing this with equation (3.202), it is apparent that Fourier

decomposition can be used to deduce the magnitudes of the coefficients A and B.
Note the change in summation limits in which a value of zero for n corresponds
to the position of the bar.
When both ends of the bar are fixed, the boundary conditions require that
the displacement u be zero at both ends. Substitution of these conditions yields
the same equation (3.217) for the natural frequencies of the bar. Rayleigh (section
154), points out that the general equation for displacement is given by (with
symbols changed for consistency with the present text)

u(t) =&0 x~
l+ ~sm - - An cos(nm:t)
. (n1TX)( . (nm:t))
- - + Bn sm - - (3.219)
1 1 1

Where &0 is the residual strain in the rod, which is, after all, over-
constrained.

3.11.2.3 Longitudinal vibrations of a clamped-free rod with a rigid mass at the free
end
For a rigid mass, M, attached at the free end of the cantilever, the force due to the
product of stress and area must provide the requisite acceleration to the mass.
Mathematically

(3.220)

Substituting this into (3.200) gives

-AERcon~co{con~/)= -MRco; sin(con~/) (3.221)

Noting that pAl is equal to the total mass of the rod m, equation (3.221) can
be rewritten in the form

(3.222)
=CtanC

Clearly, for any given ratio of masses there will be an infinite number of
solutions to equation (3.222). The first two solutions for a number of different
mass ratios are tabulated overleaf

127
FLEXURES

m/M 0.1 1 10 100 00 (M= 0~

Ct 0.311 0.860 1.429 1.555 n/2
3.173 3.426 4.306 4.666 31t/2
0.311 0.866 1.519 1.707 1.732
0.473 1.111 1.498 1.563 1.5707

Assuming that, during vibration at the fundamental frequency, the rod

extends linearly to a distance oat the free end, it is possible to derive equations
for the kinetic and potential energy

(3.223)

21 2

Using Rayleigh's method these two can be equated to yield a approximation

for the fundamental mode natural frequency given by

-
ti)~- (3.224)

Using this simplified equation a slightly high frequency is predicted as can

be seen by comparing rows 2 and 4 of the above table. For appreciable masses
attached to the bar, the difference is, for most purposes, negligible.
For small masses attached at any point x along the bar, it is reasonable to
assume that the mode shapes will remain unaltered. Under these conditions, the
kinetic energy for any mode is given by, Rayleigh, 1894,

(3.225)

Consequently, it can be seen that each natural frequency will reduce in

proportion

(3.226)

128
CHAPTER 3: RIGID BODY DYNAMICS

For a mass attached to the free end this becomes

M)-112
(
1: 1+ -
m
(3.228)

Compare rows 5 and 2 of the above table to assess the relative error of this
assumption on the first mode natural frequency as a function of the mass at the
free end.

3. 11.3 Lateral vibration of a bar

The equation governing motion for lateral vibrations of a bar has already been
derived in section 3.3 using Hamilton's principle and is reproduced here for
convenience

(3.229)

Again, assuming that the vibration shapes are independent of time, it is

possible to write the solution to (3.229) in the form

y(t) = f(x)sin((i)n/ + (/J) (3.230)

Substituting this into (3.229) yields

d• f(x) = m(i); f(x) =a 4f(x) (3.231)

dx 4 lEI

where m is the totnl mass of the bar and

(3.232)

Assuming a general solution of the form

f(x) =CeAr (3.234)

The roots of equation (3.229) are

(3.235)

Consequently, the four solutions to (3.235) are

129
FLEXURES

..t = a~.h + iO =±a ,±ia (3.236)

Substituting (3.236) into (3.234) and rearranging, the solution to (3.229) can
be expressed in the form

f (x) =C 1 sinax+C2 cosax+C3 sinhax+C4 cosh ax (3.237)

Clearly, the constants C must be determined from the boundary conditions.

In general, the case of a beam being either free or hinged or clamped at the
ends will be considered. In the first of the following examples, all steps in a
general solution will be indicated. Thereafter, for brevity, only major components
of a solution will be presented. The readers are once again referred to Rayleigh to
see how it is rigorously pursued and Timoshenko and Young, 1955, for detailed
discussions.

3.11.3.1 Hinged-hinged beam.

For the hinged beam, at all times the deflection and bending moment must be
zero at each end i.e.

f (x ) = 0& af(x) @ x = O,x =I (3.238)

ax

For x = 0

f(x) = 0 =c2 +c.

af(x) = 0 =-C2 +C4
(3.239)
Ox

This is only satisfied if both constants are equal to zero. For x =l

f(x) =0 = c, sinal+ c3sinh a/
(3.240)
af(x) = 0 =-c. sinal+ c) sinhal
ax

Because sinhal cannot be zero, then C3 must be zero. The only non-trivial
solution that satisfies all boundary conditions is

sinal = 0
or n = 0,1 .. . c:o (3.241)
al =mr

Consequently, the natural frequencies of the hinged beam given by (3.232)

are

130
CHAPTER 3: RIGID BODY DYNAMICS

(3.242)

From the above, it can be deduced that the mode shapes for this bar, being
successive integral sine waves are the same as those previously derived for a
string. However, the frequencies, or tones, for the bar increase by the square of n
as opposed to linearly in the case of a string.
A general solution for the bar is

f B, sin(n1tX)cos((mr ) ~ E~ 1)
2
y(t) = (3.243)
, .1 I ml

Again, exploitation of the orthogonality of this equation will enable the

computation of the coefficients for B, based on the initial distortion of the bar.
3.11.3.2 The free-free or clamped-clamped bar
For a bar 'floating in space', the conditions there will be an absence of both
bending moments and shear forces at each end. Mathematically

fi f(x) = fl f(x) = O@x = 0 x =I (3.244)

Ox 2 ox3 '

The second and third differential equations for f(x) are

From the first of (3.244) we have

-C2 +C 4 =0
-c. +C3 =O
The second condition gives

c. (sinha/-sinal)+ c2(cosha/-cosai) = 0
(3.245)
C1(cosha/-cosal)+C2 (sinaJ + sinhal)= 0

131
FLEXURES

Having two equations and two unknowns, it is possible to solve for the two
constants in (3.245) after which these can be substituted back into the general
solution for f{x).
From equations (3.245) it is possible to express the ratio of the constants by
two equations

C sinh a/ -sin a/ cosh a/ -cos a/

- -2= =----- (3.246)
c1 cosh a/ - cos a/ sinh a/ + sin a/
Rearranging (3.246), compatibility requires that

(cosh al- cosa/Y =sinh 2 al- sin 2 a/

or (3.247)
cos a/ cosh a/ = 1

Equation (3.246) can be solved to provide an infinite number of roots, the

first two of which are

a,I = 4.7300408
(3.248)
a 21 =7.8532046

Values for each natural frequency can then be calculated from equation
(3.232)
Based on the above analysis, the general solution for the free-free beam is
given by

~((sina"l-sinhaiXcosa"x+cosha"x}+} ( )
y(t) = £... . . n COS (t)nf (3.249)
n·• ..(cosh a"' -cosa"zXsrna"x+sinha"x}
In the case of a clamped bar it is necessary that the deflection and slope are
zero at both ends. Substitution of these boundary conditions results in the same
compatibility equation (3.247) as that necessary for a free-free bar. As a rather
surprising consequence (to me anyway) the natural frequencies in each case are
of the same value. However, it is obvious that the mode shape corresponding to
each frequency will differ. Again, because (3.249) corresponds to a series of
orthogonal functions, it is possible, although rather tedious, to determine the
constant coefficients of (3.249) using the techniques of Fourier series, see for
example Timoshenko and Young, 1955.
Case 4: The cantilever or clamped-free bar
For a clamped-free bar the compatibility equation is given by

cos a/ cosh a/ + 1 = 0 (3.250)

132
CHAPTER 3: RIGID BODY DYNAMICS

To determine the natural frequencies for any of the above configurations it is

necessary to calculate the coefficients of a ,l. For convenience these are tabulated
below.

Tabulated values for the coefficients a ,l

i Camped-clamped Clamped-free Hinge-hinge
Free-free
1 4.7300408 1.875104 3.141593
2 7.8532046 4.694091 6.283185
3 10.9956078 7.854757 9.424778
4 14.1371655 10.995541 12.566371
5 17.2787596 14.137168 15.707963

Finally, for want of a better place, the effective mass of a beam, when it
vibrates with an's' shaped deflection curve of the form given by equation (4.76)
is assessed using Rayleigh's method.
Case 5: Effective mass of a bar vibrating with an 's' shaped deflection curve
Many flexures consist of two leaf-type spring connected at each end by two rigid
platforms. Application of an appropriate force will result in the linear motion
between the two platforms. Under the correct conditions, and for small
distortions, the platforms will move parallel to each other in a near rectilinear
path, see chapter 4. Bach of the leaf springs will undergo an's' shaped deflection
with zero slope at each end. From the beam bending equation (see also section
4.1.5), the deflection perpendicular to the axis of a beam is given by

(3.251)

Assuming that this represents the fundamental frequency mode shape, the
kinetic energy of an element along the beam is

(3.252)

Integrating equation (3.252), the total kinetic energy for a single leaf spring is

133
FLEXURES

(3.253)

Equating maximum potential and kinetic energies for this system, the first
mode natural frequency for s springs can be approximated by

2
= k
0) ---;:~-- (3.254)
n M + L0.37m

where k is the stiffness of the complete platform.

3.11.4 Lateral vibration of bars with a rigid mass attached

Although extension of the above analyses can be used to determine equations for
the lateral vibrations of bars having rigid masses attached, these are rather
cumbersome, see the comprehensive paper of Maltbaek, 1961. In contrast, for
these examples Rayleigh's method produces simple expressions for which the
errors usually deviate less than 1% from the 'true' value. These and many more
examples can be found in Rayleigh, 1877, or Timoshenko and Young, 1955.
3.11.4.1 Cantilever beam with a rigid mass attached at the free end
As a reasonable guess, the deflection curve for a massless cantilever beam with a
concentrated mass, M, on the free end will satisfy the appropriate boundary
conditions. From simple bending theory, the deflection along the beam can be
written in the form

(3.255)

The maximum kinetic energy of an element along the beam is given by

(3.256)

Integrating over the length of the beam yields

134
CHAPTER 3: RIGID BODY DYNAMICS

Consequently, the total kinetic energy for the beam and attached mass is

(3.257)

The maximum potential energy is readily calculated

(3.258)

Equating equations (3.258) and (3.257) an estimate for the fundamental mode
natural frequency is given by the equation

lV 2 = - - -k_ _ (3.259)
n M +0.2357m

Interestingly, it can be shown, Maltbaek, 1961, that the error between this
approximate equation and the 'true' value is never greater than 1%. For almost
all engineering applications this is more than adequate.
3.11.4.2 Hinged beam with a central mass, M
Choosing the static deflection curve for a massless beam with a central force of
value Mg, the kinetic energy of an element dx at a distance x from one end is

(3.260)

Integrating over the length of the beam yields

(3.261)

135
FLEXURES

Following the previous analysis, an estimate of the fundamental mode

natural frequency is given by

l()l = ___k_ _ (3.262)

n M + 0.4857m

Again, irrespective of the relative values of these masses, (3.262) is

sufficiently close to the 'true' value for most engineering applications.

3.11.5 Vibration of plates

As might be expected, when departing from two to three-dimensional problems,

the mathematical difficulties are correspondingly enhanced. In fact, it appears
that the solution for vibrations of a circular plate was a harsh testing ground for
some of the best mathematicians of the early 1800's, the full solution finally
yielding to Kirchoff in 1850. Of course, Rayleigh's succinct discussion of the
vibration of membranes and plates remains one of the more complete accounts of
the theory to this day. Solutions to the vibrations of a rectangular plate remain
intransigent and will be discussed only in brief.

3.11.5.1 Vibrations ofa circular plate

From the theory of plates, the potential of an element of a distorted flat plate can
be expressed in terms of the two principle curvatures, r

(3.263)

D is the flexural rigidity given by

(3.264)

and t is the total thickness of the plate.

For small displacements, u, in the z-axis direction, the total potential energy
in the thin plate can be obtained by integrating equation (3.263) over the surface
giving

(3.265)

where V represents the two-dimensional Laplacian

136
CHAPTER 3: RIGID BODY DYNAMICS

(3.266)

The total kinetic energy of the plate is readily obtained from the integral

(3.267)

Application of Hamilton's principle (eventually) yields the equation

governing free motion at all points on the plate

-
D n4
v u-
Z +u=
.. 0 (3.268)
pi

where Z represents body forces on the plate. Assuming that these are
gravitational and, therefore, constant, this will not affect the vibration
characteristics of the plate and will be dropped henceforth.
Again, for convenience, the constant in (3.268) can be reassigned as

(3.269)

Assuming a solution to (3.268) of the form

u(r,B,t) = f(r,8)e 1"'-' (3.270)

Substituting (3.270) into (3.268) gives

V4 / =k;nJ (3.271)

where

(3.272)

3.11.5.1.1 Free vibration of a circular plate clamped at the perimeter

To analyze the vibration response of a circular plate, it is convenient to express
equation (3.268) in polar form (the r and 8 in (3.270)) in which the Laplacian
becomes

(3.273)

137
FLEXURES

Assuming a separable solution for the radial and tangential terms in (3.270)
the solution for f will be of the form

f(r,B) = g(r)h(B) (3.274)

Because r and B are independent of each other, equation (3.275) can only be
satisfied if both are equal to a similar integer constant, n 2 say. From the right
hand side of (3.275) we have

(3.276)

for which the solution is readily obtained as

h = Acos(nB - <p) (3.277)

A and <pare constants that depend on initial conditions. For symmetry, it is

necessary, and convenient, that n is an integer. From the left hand side of (3.275)
the solution for the function g can be obtained from

(3.278)

From which

g = BJn(k,""r)+CJn(ikmnr)
(3.279)
=BJn(kmnr) + C/n (kmnr)
Jand I represent Bessel functions of order n while the constants B and C can
be determined from the boundary conditions. In addition, there will be an
infinite number of admissible values for k the determination of which provides
the natural frequency for that mode.
From equations (3.270), (3.277) and (3.279) the complete solution for
displacement of the plate is given by

(3.280)

where A,&, ytand <pare constants.

138
CHAPTER 3: RIGID BODY DYNAMICS

The two integers, n and m correspond to the number of radial and

circumferential nodal lines on the plate respectively. Based on the value of n
there are an infinite number of discrete frequencies determined from the Bessel
functions in the above. For flexure applications, the circular plate represents a
diaphragm. In most cases the plate will be clamped at the perimeter for which
the boundary conditions are

oul
uIr• R -- Or -0
- (3.281)
r• R

Because these conditions must be satisfied around the complete periphery of

the plate over all time, dropping the subscripts on k, substitution of (3.281) into
(3.280) gives

Jn(kR)dl"d:!) - J.(kR)dJ~~kR) =0
or (3.282)

It can also be seen that the constant 'I' is given by

(3.283)

Equation (3.282) provides an infinite number of roots, (kR)'"" corresponding

to each value of n. From (3.269) and (3.272), the natural frequencies for each
mode are given by

(3.284)

The first 16 roots of equation (3.282) are listed in the table below

Values of the roots (kR)Iffll corresponding to equation (3.272)

m/n 0 1 2 3
1 3.196 4.611 5.906 7.143
2 6.306 7.799 9.197 10.537
3 9.440 10.958 12.577 14.108
4 12.577 14.108 15.579 17.005

The complete solution for vibration of the disc is given by

139
FLEXURES

I« 0 1 2 3

0
0 CD EB ®
1 @ @) @ 00
2
(@ ~ ~ ~
3
0 •••
Figure 3.22: Nodal lines corresponding to different
mode shapes for free vibration of a circular plate

.. ..
u(t) =LLA,,. cos(nB- q>)(J,.(kmnr) + V'mnl,. (k.,,.r))cos(a>,.,t- &..,,.)
n.O m• 1
(3.285)

The constant coefficients A..,. are dependent on the initial deflection u(O).
Again, because (3.285} represents an orthogonal function, it is possible to
determine values for each coefficient using the methods of Fourier series.
Normalization coefficients for the series expansion to determine A,.,., were first
recognized by Bessel. For a symmetric mode deformation both n and I are zero
and the Fourier-Bessel series of equation (3.285) is, Hildebrand, 1976,

(3.286)

To visualize each of these modes, figure 3.22 shows the nodal lines. Figures
23 show isometric views of the first four modes represented in figure 3.22.
3.11.5.2 Rayleigh's method applied to a circular plate with a central mass
Determination of the natural frequencies for different plate configurations
involves solving complex differential equations. For many flexures, it is the
fundamental frequency that limits performance. As a consequence, Rayleigh's
method can be used to derive simple approximate formulae. In this section, the
clamped plate with and without a central point mass is analyzed.
Case 1: Circular plate damped at the perimeter

140
CHAPTER 3: RIGID BODY DYNAMICS

a) b)

4
x10
~

t.
:'ll2
lo
-2
I

v ·I •I y
X

c) d)

00$ 0.02

t i
:l5 :'ll
0.01

I
0 0

1~.01
0
~.D$ .,02
I I

y ·I ·I
X

Figure 3.23: Mode shapes for a clamped circular plate a) m =0, n =0, b) m = 1, n =0, c) m =0, n
= l, d) m = I, n = I

As a reasonable guess for the mode shape, the static deflection due to a
uniformly distributed load is

u=-q- (R2 - r2Y

64Dt (3.287)
=6(R2 - rlY
where q is load per unit area and o is a constant.
Converting equation (3.265) to polar coordinates, the potential energy of a
circular plate with deflections symmetric about the center (i.e. contour of
constant height being circles) is given by

(3.288)

Substituting (3.287) into (3.288) gives

141
FLEXURES

R
5
V = 2 nD5
2
J((R r - 4R r +5r )+v(R r -4R r +3r )}ir
4 2 3 5 4 2 3 5

0
(3.289)
1ClJ2s 5 2R6
=----
3

As would be expected, the term involving Poisson's ratio corresponding to

energy of twisting vanishes. The total kinetic energy in the plate can be
determined from the integral
R
T = trpl{J);5
2
J(R 2 -r 2 Yrdr
0
(3.290)
rrpt{J)25 2Rio
= n
10

Equating (3.289) and (3.290), an estimate of the fundamental mode natural

frequency is given by

(3.291)

Compared to the ' true' value of 10.21, the approximate method is a little over
1% higher.
Case 2: A clamped circular plate with a central mass
Assuming that the dominant static forces on a plate are those due to a central
point mass, M, the subsequent deflection curve can be represented by the
equation

(3.292)

Substituting this into (3.288) and ignoring the term involving Poisson's ratio,
the integral can be rewritten in the form
0
V = trD4R 2 5 2 J(e.; 2 e 2~ +.; 2~ +e 2 ~)i.;
(3.293)

where

142
CHAPTER 3: RIGID BODY DYNAMICS

Alternatively, the potential can be deduced from

Similarly, the kinetic energy can be computed from the integral

(3.294)

Equating the maximum values for potential and kinetic energy, an estimate
for the lowest mode natural frequency is given by

a>" = 4Ji;~=~
R
M+ -
7
m
7
M+ - - m
2.3 3 2.3 3 (3.295)

= ~ M +~.13m
At low values for the central mass, M, the frequency equation becomes

(3.296)

In view of the initial assumption that the mass of the plate is insignificant in
comparison, the error is still less than 10%. In fact the boundary condition of zero
slope at the center of the disc is also violated.
If the central mass is relative small with respect to that of the plate, it may be
assumed that the shape is relatively unchanged from that assumed for the plate
of case 1 above. Consequently, the potential energy given by (3.289) will remain
unchanged while the kinetic energy becomes

(3.297)

143
FLEXURES

~
D rn EJ] [IIJ
0 1 2 3

0 .

I
B t=tj till ffiE+

2
§ ~ ~ m§
3
~ mg m
Figure 3.24: Nodal lines corresponding to different mode shapes
for free vibration of a rectangular plate

Consequently, for small values of the central mass, M, the fundamental

frequency is given by

(3.298)

It can be readily verified that (3.297) become equal to (3.291) when the central
mass becomes vanishingly small.
If the radius of the central mass cannot be neglected, an approximate formula
can be obtained for the case where the mass of the disk is considered negligible.
In this case, the stiffness, k, of the disk is, Den Harto~ 1952,

(3.299)

where a = R I r and is accurate to within 10% for values ranging from 2 to 5.

In this case, the fundamental mode natural frequency can be computed from

= ["k
lV
n v"M (3.300)

144
CHAPTER 3: RIGID BODY DYNAMICS

3.11.6 Vibrations of a rectangular plate

Exact solutions for the free vibrations of rectangular plates require some
mathematical rigor. Fortunately, there is one exception.
3.11.6.1 Free vibrations of a rectangular plate simply supported at the edges
By substituting back into equation (3.271), it can be readily verified that a
solution for the vibration of a rectangular plate is given by

(3.301)

where a and b are the lengths of the sides and, yet again, the constants A can
be obtained thorough Fourier decomposition of the initial conditions.
Correspondingly, the natural frequencies are readily shown to be

(3.302)

A representation of these various mode shapes is shown in figure 3.24. To

visualize the schematic representations of figure 3.24, isometric views of the first
four modes are shown in figure 3.25.
a) b)

004 0015

loos

j::
l

0
I

y 0 0 X X

.,o. Figure 3.25: Mode shapes for

free vibration of a rectangular
f,

r
plate with its edges simply
l supported, a) m = 1, n = I, b) m = 1,
n =2 (form =2, n = 1 interchange
x andy in figure), c) m =2, n =2

y X

145
FLEXURES

3.11.6.2 Fundamental frequency's of rectangular plates with other boundary

conditions
Investigations of plates having other boundary condition are best approached
through the successive refinement of Rayleigh's method commonly called the
Ritz method. Using this technique, equations for predicting natural frequencies
for many mode shapes and boundary conditions have been investigated, Blevins,
1985, Shockley, 1995. Considering equation (3.257)

(3.303)

Values for k 2 corresponding to a variety of boundary conditions are

tabulated below

Table of values for the coefficiente in 3.303 for a s uare plate

Boundary conditions k2
Clamped at one edge free 3.494
elsewhere
(cantilevered plate)
Clamped all around 35.99
Clamped at two edges free 6.958
elsewhere

3.12 Case study 1: A simple, two degree of freedom flexure mechanism

This case study has been chosen to illustrate some of the techniques introduced
in this chapter, most of which will be repeated throughout the rest of the book.
Figure 3.26 shows a notch type planar flexure that consists of two, parallel,
simple four-bar Jinks. A simple planar mobility analysis indicates a total of 2
freedoms (==3(7-1)-2(8)) which are assigned the coordinatesq1 &q2 • It is readily
apparent that, for small motions, these correspond to linear displacements of the
two platforms. It is desired that, through suitable selection of the notch hinge
thickness, t, the fundamental mode undamped natural frequency be greater than
30 Hz. Identical viscous dampers are to be applied to each platform as shown.
Determine the damping force necessary to produce a critical damping ratio of
value 0.2 at the fundamental mode natural frequency.
As a first step in any dynamic analysis, it is necessary to determine the
equations governing motion for the system. To do so it is necessary to compute
the kinetic and potential energies and the Rayleigh dissipation function in terms
of the independent coordinates of the mechanism. Using local coordinates for
each rigid body, the kinetic energy is given by

146
CHAPTER 3: RIGID BODY DYNAMICS

70
I.
.--- -

Depth= 12

Figure 3.26: A stacked, simple linear spring providing two, parallel linear freedoms
/1 = 22.5, 12 = 25, /3 = 14 = 45, (all dimensions in mm)

However, the two angular coordinates are not independent and, noting that
/3 =14 , can be expressed in terms of the linear coordinates by the equations
() = q, - q2
l l
3

Also, assuming that the support legs can be considered as long thin rods, the
values for their second moment of mass about the centriods can be computed
from the equations
m/2
1=-
12
Ignoring the mass of the notches, the kinetic energy can be rewritten to give

147
FLEXURES

2 2
T =.!.[m
2 l l
q·2 +m2 q·2
2
+ m3 fq· -q·2 )2+ 2m3 (41 +42)
6 \:I 2
+ m.
6 2
q·2 + 2m4
( 4
2
2) ]

Expanding terms in parentheses, rearranging and noting that, in this case,

m3 =m4 finally yields

The potential energy can be obtained from the equation

Again, expressing this in terms of the linear coordinates gives

Finally, the dissipation function can be readily seen to be

D = -b(. 2 ·2)
\ql +q2
2

Substituting the above into Lagrange's equation gives the two equations
governing motion

2m3 ) .. 2m3 .. b . 4k 4k Q
( ml + - - ql +--q2 + ql +-2 ql --;:q2 = 1
3 3 ~ ~
(a)
2m3.. ( 4m3 ) 4kb. 8k Q
--q~ + m2 +-- q2 + q2 --2-ql
..
+ -2 q2 = 2
3 3 /3 /3

For a notch hinge, the angular stiffness k is given by equation (4.91),

reproduced here

(b)

148
CHAPI'ER 3: RIGID BODY DYNAMICS

In the absence of damping and externally applied forces, the equations of

motion in matrix form become

(c)

Harmonic motions of each of the coordinates can be assumed of the form

q1(t) =Re{A,e'dl·'}
(d)
q2 (t) =Re{A2 e1"'.,t}

Where the constant coefficients of A correspond to the eigenvectors.

Substituting (d) into (c) and rearranging gives

(e)

The undamped natural frequencies can be determined from the roots of the
determinant

(f)

Expanding the terms in this determinant gives a quadratic in a {= w;) given

by

(4k -a(m +2m3))(8k

[2
3
3 I
-a(m +4m3))-(4k
/2
3
3 2
+a2m3)(4k
3 /2
3
+a 2m3)
3 /2
3

=0

For the dimensions shown, the mass of the platforms is 7 times greater than
that of the individual support legs. This can be used to further reduce the above
equation to

149
FLEXURES

(g)

Dividing through by the first term in the above gives

From. which, the lowest value for a is given by

2
834) (834) - 576
a = k ( 571 571 __ _
571
z;m3
--~-_;,___:....;..._____:_

2 (h)

: : 1 - k- = a» 2 = 355530
sz;m3 n

From equation (b), the thickness of the notch hinge is given by

(QED) (i)

Substituting equation (i) into (b),

400
the stiffness, k, of each individual notch
and mass of each support le~ m3 , is
200

k = 1.91 (Nm rad·I]

Im 0

·200
m3 =0.00672 [kg]
..oo

~oo
=~~~ ..oo
oo~~~~ ~~-3~
oo~-2~
oo~-~too~~o
Computation of the damping
Re coefficients is not as straightforward
and requires a computational
Figure 3.27: Roots locus for the two degree approach. Modeling the first mode
of freedom flexure
natural frequency in terms of a simple
second order system, the optimal damping coefficient can be obtained from a
roots locus of the corresponding eigenvalue. Consequently, it is necessary to
develop the complete system equation of free motion given by

150
CHAPTER 3: RlGID BODY DYNAMICS

(k)

From section 3.10.2, this can be written in the form of an eigenvalue equation
with the eigen matrix given by

(I)

Eigenvalues of this matrix will comprise two complex conjugate pairs, the
lowest (or gravest) mode natural frequency corresponding to the pair with the
lowest complex part. Modeling this as a single degree of freedom second order
system, the ratio of the real to imaginary parts of the eigenvalue can be
represented by the equation

Re(A.) = ~ =0.204 (m)

Im(A.) ~~ - ~2

Figure 3.27 shows a plot of the eigenvalues as a function of the damping

force. Equation (m) is satisfied at a value of

[Ns m·l] (QED)

If desired, the frequency response functions for this system can be computed
using the techniques outlined in section 3.9.

References
Blevins R.D., 1985, Formulas for Natural Frequency and Mode Shapes, Krieger
Publishing Co., FL
Den Hartog J.P., 1952, Advanced Strength of Materials, McGraw-Hill, NY, p131
Feynman R.P., R.B. Leighton and M. Sands, 1964, The Feynman Lectures on Physics,
vols I- III, Addison Wesley, Reading, Ma.
Goldstein H., 1980, Classical Mechanics, Addison-Wesley. An advanced level text,
reading this text is like playing chess. Likewise; L. A. Pars, 1961, Treatise on
Analytical Dynamics.
Hildebrand F.B., 1976, Advanced Calculus for Applications, Prentice-Hall, Inc., NJ.
Lanczos C., 1970, The Variational Principles of Mechanics, Dover publications Inc.,
NY. Chpt's 1-5.
Longair M.S., 1984, Theoretical concepts in physics, Cambridge University Press.

151
FLEXURES

Maltbaek J.C., 1961, The influence of a concentrated mass on the free vibrations
of a uniform beam,]. Mech. Sci., 3,197-218
Newland D.E., 1989, Mechanical vibration analysis and computation, J. Wiley and
Sons Inc., NY.
Rayleigh J.W.S., 1894, Theory ofSound, vol. I, Dover Publications. Quote is to be
found in section 96 on page 130 of this edition. I must confess to a rather
biased view of this treatise. Rarely a month goes by in which I do not consult
this text to see how it is really done. Unfortunately, in the discussion of the
vibration of plates the symbol h is used to represent the half thickness. Great
care must be applied when translating this to an equation in which the total
thickness is used.
Shockley W.F., Vibration of systems having distributed mass and elasticity, in
Shock and Vibration Handbook, 4th edition, ed. Harris C.M., McGraw-Hill, 1995
Timoshenko S.P. and Young D.H., 1955, Advanced Dynamics, McGraw-Hill, NY.
Timoshenko S.P. and Young D.H., 1955, Vibration Problems in Engineering, Van
Nostrand Co. Inc

!52
4 Flexure elements

4.0 Overview
Flexure devices consist of compliant elements that connect 'rigid' bodies to form a
mechanism. Fortunately, flexure elements almost invariably consist of either leaf
type springs, notch hinges or cross strip pivots (with a variant known as a cartwheel
pivot). Ultimately, flexures will be integrated into instrumentation attd machine
mechanisms to provide defined motions upon application of a force. Consequently,
analysis of the behavior of the simple leaf spring under the influence of combined
forces is discussed. From this, it is found that there is the possibility of buckling
under compressive axial loads and stiffening in the presence of tensile loads. While
the former of tltese represents an unwanted instability, the latter can be tolerated in
most applications. Design strategies that favor the latter are presented. Another
common element is the notch type hinge. This am be relatively easily manufactured
by machining away circular sections either side of a solid body to produce two
separate bodies connected by the thin, compliant web between the circular cut-outs.
Approxitnate equations for computing the compliance of this type offlexure element
were first presented by Paros and Weisbord in 1965 and are still used to this day.
Both the leaf and notch type hinges are singular manifestations of the more general
elliptic hinge for which the equations of Paros and Weisbord have been expanded.
Cross-strip and cartwheel hinges are also common flexure elements for which a
relatively comprehensive analysis is provided. All of the above are often used for
single degree offreedom flexures. Elements having two degrees offreedom or greater
have not been subject to such scrutiny. Some of the two axis mechanisms amenable
to analysis are presented at the end of this chapter.

4.1 Leaf type springs

While there appears no natural hierarchical order to the three different flexure
elements presented in this chapter, it is mainly for historical reasons that the leaf
spring is discussed first. It is possible that the reason this element has been in use
for so long is its ease of manufacture and assembly. Primarily, leaf springs can be
either fabricated by clamping a thin plate between two rigid bodies or, with more
difficulty, machined out of a monolithic block of material. In the former
FLEXURES

manifestation, leaf springs have been used in fine instrumentation for physical

l
apparatus since the onset of the industrial revolution and probably before.

FY

! FY Fx

-3
I
I
II

Fx
0__..:; s
L

ly
X

Figure 4. 1: A cantilever beam with an arbitrary, in-plane force plus bending moment at the free
end

4.1.1 The cantilever as a rotary binge

Possibly, the simplest flexure mechanism is the cantilever beam. For the beam of
figure 4.1, the bending moment due only to a force, FY, is

d2
El ___}'_ =F (L- x) (4.1)
dx2 y

For the purpose of computing the maximum stress, the maximum bending
moment, Mmox = FYL, can be directly substituted into the bending equation, see
chapter 2.
Integrating equation (4.1} twice
2
dy= F ( Lx- -x
El - ) +C
dx y 2 I

(4.2)

Substituting the boundary conditions that both slope and deflection are zero
at the clamped end (x = 0), equations (4.2) can be solved to give

Eldy =F
dx y
(Lx-~)
2
(4.3)
Ely= FY( L x; -:)

154
CHAPTER 4: FLEXURE ELEMENTS

For relatively small displacements, the slope (} and deflection oat the free
end of the beam are given by

F L2
(} =-y-
x• L 2El
(4.4)
FL3
0 =-y-
x-L 3El

From the above, the angular and linear stiffness of the cantilever are

(4.5)

Clearly, if the cantilever is to be used as a hinge, it is necessary to know

where the pivot point is. Although the pivot point will vary with deflection, it is
apparent that, if the deflections are small, so too will the variations of the pivot
point. Defining this as the intersection of the tangent to the free end of the
cantilever with the unloaded beam axis, it is clear that the pivot point, o, in figure
4.1 is located from the free end of the undistorted beam a distance, s, given by

oy
s =-- ~ - L
2
(4.6)
tanB, 3

Equation (4.6) implies a stationary pivot point located a distance of one third
of the way along the axis from the fixed end of the cantilever. Before moving on
to· more complex load conditions, it will be informative to look at a few
applications of the leaf type hinge. Four different flexure mechanisms are shown
in figure 4.2. In all figures, the stationary link of the flexure mechanism is
shaded. The first two mechanisms are variants on tlle simple linear spring to be
discussed in chapter 5. Application of a force to the moving platform will
produce distortions resulting in a parallelogram displacement of the four bar
linkage. For mechanisms in which the direction of applied forces may vary, it is
often desirable to use more than one hinge as illustrated in figure 4.2 c). In this
diagram, vertical and horizontal forces are each carried by the corresponding
hinge. Commonly, such a requirement is typical for a flexure that is required to
support varying weights, for example, a spring balance or truck stop weigh
platform. Construction of the crossed-strip hinge provides a challenge to both
designer and manufacturer. A simpler construction is shown in the hinge of
figure 4.2 d). The crossed strip hinge is discussed towards the end of this chapter.

155
FLEXURES

b)

==·''"
c)

IF, F,
~~::::!=_=-----11-
~
weight

Figure 4.2: Four flexure mechanisms utilizing the leaf type hinge, a) four-bar parallelogram
flexure with hinges under compressive axial load. b) four bar flexure with hinges under tensile
axial load, c) two-hinge pivot with elements crossed (the cartwheel hinge), d) separated two
hinge pivot (reproduced from Thorpe, 1953)

Figure 4.2 also illustrates another feature common to flexures, that is, the
forces are often applied to a rigid body remote from the hinge. As a consequence
it is often necessary to assess the effects of this on the subsequent distortion of
the hinge. Fortunately, such an analysis has been carried out by Thorpe, 1953, in
which various cases of hinges subject combinations of z-axis moments and axial

Figure 4.3: Mathematical model for a cantilever hinge with a rigid body attached to the free
end subject to a tangential force

156
CHAPTER 4: FLEXURE ELEMENTS

and tangential forces are presented. For the case of a single tangential force
applied to a rigid body attached to the free end of a cantilever hinge, figure 4.3,
the deflection A is given by

A =F LV2 +Ll +L2/ 3)

(4.7)
Y y EI

The distance from the 'free' end of the hinge to the effective pivot point is

IL+2L2 / 3
s =--_:__ (4.8)
L+21

Finally, the slope and deflection at any position from the stationary body to
the end of the rigid body attached to the free end of the cantilever can be
determined from the equations

(4.9)

At the free end of the rigid body (i.e. x = L + 1), equations (4.9) become

F
y = y- (L +l}'
3EI (4.10)
dy = Fy (L +IY
dx 2EI

Setting the length of the rigid body l to zero, it can be readily verified that
equations (4.7) to (4.10) reduce to the equations for the hinge with a tangential
force applied at the free end, equations (4.2) to (4.6).
Clearly, for many applications, bending moments about the free end as well
as axial and tangential forces will be present. Therefore, it is necessary to assess
the effects of these combined forces on the stiffness of the hinge. In the following
analysis, only bending moments about the z-axis will be considered. It is also
noted that, to determine the angle of twist and shear stress due to a moment
about the x-axis, equation (2.102) and (2.104) of section 2.4.9.1 can be added to
the stress due to the other loads. In all cases it will be assumed that the principle
of superposition is applicable.

157
FLEXURES

4.1.2 The cantilever hinge subject to an axial compressive force

Fred Easbnan investigated the effects of axial loads on the stiffness of a cantilever
beam in 1937 for application to the design of fine balances. In particular, it was
observed that
With the aid of ordinary beam theory, it is a simple mntter to compute the
stiffness of a flexure if it is subjected to a little or no load, but if it is heavily
loaded in compression, its stiffness 10ill be grently reduced and may be even
zero or negative. Negative stiffness indicates instability resulting from column
action (i.e. buckling). On the other hand, if the flexure is heavily loaded in
tension, its stiffness increases and may become many times it's no load
magnitude.
In fact, the balance design presented by this author represents an attempt to
use this 'tunable' stiffness·to create a free hinge to compete with the knife-edge
hinges currently in use. The victor in this
competition between the flexure and knife-edge
I pivot remains unresolved.
0 ' - --.- For small axial loads, the compression or
extension of a straight beam can be readily
computed from

L o = F:L
X EA
(4.11)

where A is the cross sectional area of the beam

For larger loads there will be a significant
bending moment due to the deflection of the
beam itself from the central axis. For a
compressive load as shown in figure 4.4, Eastman
Figure 4.4: A cantilever beam
subject to an axial compressive has shown that the stiffness expressed as the ratio
force plus bending moment of the bending moment, M ., required to induce
an angular variation, e. is given by

K0 u =M-
' ' e.
K; (K;)
• = El -• cot L
EI
_x
EI
=-aJcot(aJ)
El
L (4.12)
El
=- a1coth{a1)
L

Where the first equation should be used if the loads are compressive while
the second equation corresponds to a tensile axial load. From the first equation in
(4.12) the stiffness reduces to zero at a compressive load given by

!58
CHAPTER 4: FLEXURE ELEMENTS

aJ = L~ = 2r = ~
or (4.13)
2
7! El
F =--
x 4L2
This corresponds to the familiar Euler buckling condition. For zero axial
load, it can be shown that both of equations (4.12) reduce to

K _ M , _ El
O,M , - 8 - L (4.14)
•

Figure 4.5: A cantilever beam subject to combined tangential and tensile forces at the free end

4.1.3 Combined axial and tangential loads

Consider a cantilever hinge with a rigid rod attached at the free end. At the end
of this rigid rod, both axial and tangential forces are applied, figure 4.5. Different
solutions result if the axial load is tensile or compressive.
Case 1: Cantilever beam subject to combined tangential and tensile axial load
For this case, the deflection y at an arbitrary position x is, Thorpe, 1953,

As in the previous section a dimensionless axial load rating is defined by

(4.16)

159
FLEXURES

at which equation (4.16) becomes

Y =fl{Cl COSh(a>-Lx) - 1
(l+C )sinh
a>l+IIL L
x)
. ( a>- + ( - 1
l +C)
- X-C1}
L+l
(4.17)

where

(4.18)

The displacement at the free end of the attached rigid body can be
determined from the equation

(4.19)

The distance from the 'free' end of the hinge to the effective pivot point is

1
s= -I (4.20)
cl a> si.nh(a>)-(1+ cl )(cosh(a>)-1)
L L+l

Differentiating (4.17), the slope at any position xis

-dy =ll {a>C

dx
-
L I
s inh(a>-LX) -lL+l
- -cosh( a>-X) + (l+CI)}
+C1
L
--
L+l
(4.21)

Setting the length of the attached rigid body to zero (i.e. l ~ 0) corresponds
to the case where the forces are applied to the free end of the cantilever. In this
case, the constant C1 becomes

(4.22)

After which, the deflection along the cantilever is given by

(4.23)

160
CHAPTER 4: FLEXURE ELEMENTS

0.2

0.15
g
·s
c=
0.1
0
0.05

0
0 0.2 0.4 0.6 0.8
Position along beam

Figure 4.6: A cantilever beam with tangential and tensile, axial forces applied to the free
end F,c'= 0, 0. 1, 1 & 10 N. Dimensions are; width = 0.1 m, length = l m, thickness =
O.ot, tangential force = 1 N.
and

= FY (w - tanh(w))L
0 (4.24)
Fx w

Substituting (4.24) into (4.23) and rearranging yields

y =~~ (tanh(w)cosh(w~)-sinh(wz)+w ~ -tanh(w)) (4.25)

Plots of equation (4.25) for increasing axial load are shown in figure 4.6. It
can be seen that, as the axial force increases, the deflection reduces. For
comparison, equation (4.3) has been used to plot the deflection curve
corresponding to a beam with no axial load.
Differentiating equation (4.24) with respect to the tangential load, the
stiffness of the cantilever is given by

(4.26)

From figure 4.7, it can be seen that the stiffness varies almost with the 3/2
power of the applied axial load. For the purposes of computing the maximum
stresses in the cantilever, the maximum bending moment occurs at the base of
the cantilever and can be readily obtained. This is not so for the following case.

161
FLEXURES

20
........
16
8
eU)
U)
12

t
Q)

b 4
&:;
:a
0
~ 0 2 4 6 8 10
Axial applied force (N)
Figure 4.7: Plot of the axial stiffness of a cantilever beam as a function of tensile axial
force. All other parameters are as given in caption for figure 4.6.

Case 2: Cantilever beam subject to combined, tangential and compressive, axial

load
When the axial load is compressive, the situation is complicated by the
possibility of buckling. In this case, the sign of both loads has been reversed so
that the tangential force is acting in the negative y direction. The equations
corresponding to those used in case 1 above are

(4.27)

(4.28)

(4.29)

1
s= I (4.30)
C2 a> sin(a>)+(l-C )(1-cos(a>))
2
L L+l

(4.31)

162
CHAPTER 4: FLEXURE ELEI'viENTS

In this particular case, the maximum bending moment does not occur at the
root of the cantilever, but at a position along from the rigid support given by the
relationship

(4.32)

The magnitude of the moment can be obtained from

(4.33)

Again, it can be shown that, as the rigid member attached to the free end of
the cantilever reduces to zero length, the above equation becomes

(4.34)

0
= FY L(tan((())- (()) (4.35)
y Fz (()

(4.36)

(4.37)

In this latter case, the maximum bending moment occurs at the fixed end of
the cantilever.
Equation's (4.35) to (4.37) contain some initially unusual characteristics. It
will be recalled that the tangential force is applied in a direction opposite to the
direction of deflection. Clearly, if the axial force is zero, the deflection of the
beam will proceed in a negative direction. This can be seen in figure 4.8 when
axial loads are other than 10 N. However for an axial load of 10 N there is a
deflection in the opposite direction, the magnitude of which is less than that for
zero axial load! This can be explained if it is assumed that the beam is initially
deflected in the positive y direction. After this, the axial load is applied and the

163
FLEXURES

0 .15 --------------------
0.1 +-------------------------------~~-----
0.05 +---------------------=-_...::::;..______________
E

!
u o+---~~==~==~--~---~---.
-0.05 +-------------___:::....;;;;:;~;;;;;:----------------
].
0 -0.1 +--------------------~~~~---------
-0.15 4--------------------=:::.""'=:::-__:::~;;:----

-0.2 +--------------------__..::.-...:--~

·0.25

Position along beam

Figure 4.8: Deflection along a cantilever beam with tangential and compressive, axial forces
applied to the free end F.. = 0, O.l, 1 and 10 N. Dimensions are width = 0. 1 m, length = 1 m,
thickness = 0.01 , tangential force = 1 N.

original deflection forces reduced to zero. Clearly, under these conditions, there
wiJl be a residual bending of the beam in the positive y direction. Now, applying
the tangential force, it is obvious that the beam will be restored towards the
original unstressed shape. If, however, either the tangential force is large or the
axial force is small, this restraining force will push the beam past the origin (y =
0) and continue to distort the beam in the opposite direction. Under these
circumstances the bending moments due to both of the applied forces will be of
the same sign and the deflections due to each will add. If, however, the beam
were originally deflected in the opposite direction, the deflections due to each
force would always add. Consequently, it is not possible to determine the
deflection of a cantilever subject to such loads without first knowing its
deflection history. This state of affairs is further exacerbated if the beam is first of
all subject to an axial force of sufficient magnitude to induce buckling. The
critical load for buckling is obtained from equation (4.35) corresponding to the
point at which the deflection becomes infinite or from (4.37) at zero stiffness i.e.

or (4.38)

Under these conditions, at this axial load, the cantilever is in a symmetric

state of instability. Consequently, it cannot be decided in which direction the

164
CHAPTER 4: FLEXURE ELEMENTS

6
........
7a 4+-~~---------------------------------

e~
2+-------~~------------------------
~ 0
~ ·2+---------------------~~-------------
t

~ :+---------------------------------~~
-8
Axial applied force (N)

Figure 4.9: The tangential stiffness of a cantilever beam as a function of the

compressive axial force. All other parameters are as given in the caption for
figure 4.6

beam will deflect. Because of this, cantilever flexures subject to significant axial
loads are rarely driven using applied forces. Instead, positive drives are used and
the forces in the flexure are then determined from the prescribed deflections.
Unfortunately, to produce a positive drive implies adding an element of infinite
stiffness in the drive direction. Methods for achieving an approximation to such a
drive coupling will be touched upon in chapters 7 and 8.
Figure 4.9 shows the variation in stiffness as a function of the applied load
using the same parameters for the previous plots. The point at which the flexure
has zero stiffness may be of great utility for some design purposes.

Figure 4.10: Mathematical model for a cantilever bean subject to combined tangential and
compressive axial loading

To illustrate the situation when the tangential force is first applied, consider
the simple cantilever beam subject to the applied loads shown in figure 4.10.
These applied loads are the same as in the previous example with the sign of the

165
FLEXURES

tangential load being reversed. Because this solution is not provided in the paper
of Thorpe, a more complete derivation will be presented here. The bending
equation to be satisfied is given by

(4.39)

For convenience, this can be written in the form

(4.40)

With A - D being constants, the general solution to this equation can

immediately be written as

y =A co{aJ ~) + B sin(aJ ; ) + Cx + D (4.41)

Based on the boundary conditions of zero slope and deflection at the free
end, equation (4.41) becomes

(4.42)

For there to be zero bending moment at the free end of the cantilever, the
second derivative must vanish giving the condition

_ Q)(z2L3 +aJ2oy )
tan(aJ) - 3 2 (4.43)
Lz

Substituting (4.43) into (4.42) and rearranging, the deflection along the beam
is

(4.44)

The deflection at the free end is given by

(4.45)

It can be seen from the above equation that for axial loads less than that for
Euler buckling, the displac~ment is always positive. For other cases

166
CHAPTER 4: FLEXURE ELEMENTS

corresponding to the load conditions for the flexures of the type shown in figure
4.2 (b), the readers are recommended to consult the paper of Thorpe, 1953.

Figure 4.11: Cantilever subject to zero end slope constraint plus axial and tangential loads

4.1.4 Combined axial and tangential loads plus a moment applied to the free end of
a simple cantilever
For a simple cantilever beam subject to forces both perpendicular and along the
axis of the beam there will be an additional moment due to the deflection of the
beam itself, see figure 4.11. Under these conditions, the bending equation can be
expressed by

d 2y
El- = -M 0 +Fy-Fx
dx2 X )'
(4.46)

Following Plainevaux, 1956, it is convenient to introduce the variables

·x = !_
L
4y2 = F:L2
EI
2 (4.47)
FL
({J = -)1-
Ef
ML
mo= m0
It is noted that r = (J) / 2 where (J) is the axial load variable of the previous
section. In all analyses, the value of r is a direct measure of the influence of the
axial load on the stiffness of the beam, for an example see case study 1 of chapter
7. Equation (4.46) can be rearranged to yield

167
FLEXURES

(4.48)

A solution to this problem is given by

y = C1 cosh2yX + C2 sinh2yX + _!;-(m0 + 92X) (4.49)

4y

Differentiating (4.49) gives an expression for the slope along the beam
dy 1 dy 2y . 2y rp
-
dx= -
L-dX= CI - L sinh2"X+C
,. 2 -cosh2"X+-
L ,. 4y 2 (4.50)

Inserting the boundary conditions of zero slope and deflection at the

clamped end of the beam, the general solution for the y-axis deflection is
obtained

y= Lm~ (1 - cosh(2rX)) + L~ (2rX- sinh(2yX)) (4.51)

4y 8y

and

dy = ~ {1 - cosh(2yX))- mo sinh(2yX)
dx 4y 2y
(4.52)
d~ m0 rp .
- 2 = - -cosh(2yX) - -sinh(2yX)
dx L 2yL

Equations (4.51) and (4.52) can be used to determine the relationships

between bending moments and both vertical and horizontal forces for a given
axial load and end deflection. Consider the cantilever undergoing an 's' shaped
deflection with the boundary condition that the beam is constrained to have zero
slope and magnitude 8 at the free end. Substituting this into equations (4.51) and
(4.52) results in the simultaneous equations

8= Lm~ {l - cosh(2y))+ L~ {2y-sinh(2y)) (4.53)

4y 8y

dy =~{l -cosh(2y)) - mo sinh(2y) (4.54)

dx 4y 2y

Equation (4.54) can be rearranged in terms of either m0 or rp to give

168
CHAPTER 4: FLEXURE ELEMENTS

rp cosh(2r) - 1
m =-
0 2y sinh(2y)
(4.55)
sinh(2y)
2
fP = mo r cosh{2r)-l

These can be substituted back into (4.53) and solved for each parameter in
terms of applied loads and displacements. To see this, using the first of the above
equations, (4.53) can be rearranged in the form

o,. = ..!f!_[(cosh(2y)-1)
2
-sinh(2 )+ 2 ]
L 8y
3
sinh(2y) r r
2 2
= ..!f!_[cosh (2y)- 2 cosh(2y)+ 1- sinh (2y) + 2ysinh(2r)J ( . )
4 56
8y 3 sinh(2y)
= _!f!._[l - cosh(2y) + ysinh(2y)]
4y 3 sinh(2y)

Making the substitution

l - .cosh(2y)] = - tanh(2y) (4.57)

[ Stnh(2y)

Equation (4.56) can be rearranged to give the relatively simple expression

(4.58)

This can be further rearranged into the more familiar form

r
3
F,. 12£/( )
(4.59)
o,. =7 3(r - taoh(r))

The term in parentheses represents a stiffness factor due to the tensile load
applied along the axis of the beam and this has been plotted in figure 4.12. From
this graph it can be seen that the stiffness increases slowly for a tensile applied
load and vice versa when the axial load is compressive. In the absence of axial
loads, the analysis is considerably simplified and is presented in the following
section. For the zero load condition, the stiffness is given by the multiplier of the
term in parentheses. Consequently, the graph in figure 4.12 represents the
deviation in the stiffness as a function of the dimensionless axial load.

169
FLEXURES

2.5
stifthess factor

-2 -1 2
-0.5
y

Figure 4.12: Stiffness, bending moment and height factors for a simple cantilever
beam subject to an axial load and constrained to have zero slope at each end
Similarly, it can be shown that the bending moment at each end of the beam
can be obtained from

M0 L oy2r 2 tanh(r) (4.60)

E1 = L r - tanb(r)
Again this can be rearranged into a form readily comparable to that from
simple beam bending theory to give

M 6EI( r tanh(r) ) 2

8; = IF 3(r - tanh(r))
0
(4.61)

Again, the variable in parentheses is the dimensionless bending moment

factor, see figure 4.12. This varies less strongly than the stiffness factor.
Finally, to complete our analysis, it is possible to solve for the change in
height of the flexure, o ~ for a given deflection in y and axial load given by,
Plainevaux, 1956

o.. = ~(oy ) (5r~(3-tanh 2 r)-3tanhr])

2

(4.62)
L s L 12(r - tanhrY

Comparing with equation (4.83), the term in parenthesis on the right hand
side is the effective increase in height loss as a function of the tensile axial load
and this is also shown in figure 4.12.
lf the beam is subject to a compressive load, the bending equation is subject
to a change in sign so that it is necessary to seek a solution to the equation

170
CHAPTER 4: FLEXURE ELEMENTS

(4.63)

This is similar to the previous example only, in this case, the solution
involves circular instead of the hyperbolic functions of the previous analysis.
Again solving for the applied bending moments, forces and axial compression
yields

F, I2EJ( y
3
)
(4.64)
oy =I! 3(tan(r)- r)

tan(r) )
2
M0 6EI( r
T; =IF 3(tan(r)- r) (4.65)

ox= ~(oy) (5r[r(3+tan r) - 3tanr]J

2 2

(4.66)
12(tan(y)- r)
2
L 5 L

Again these are plotted in figure 4.U. For this case it can be seen that here is
a considerable reduction in the stiffness of the flexure with compressive load and
this goes to zero at a value

r= vffi
{FJ!
= 1.57
or
(4.67)

This is the familiar Euler load for buckling of beam that is pinned at each end.
In practice, for application to precision mechanisms, the dimensionless load
factor should always be kept as small as possible. It can also be shown that (4.66)
tends to negative infinity at y = 1t leading to the Euler buckling condition

1r
2
EI
F 5.-- (4.68)
4L2

This is a result more familiar for the prediction of cantilever stability with no
applied moment at the free end.

171
FLEXURES

In fact Plainevaux presents solutions to the bending of a cantilever in which

the 'free' end is constrained to a tilt angle e. Such a case corresponds to the free
cantilever with a moment applied to the free end. Under these conditions the
bending moments at each end of the beam are different, as has been indicated in
figure 4.12, and the complete equations for a tensile axial load are

r tanh]
2
ML tan { rcoth(y)+
_A_= r - oy2r tanh r
2

EI r- tanhy L r- tanhy
MOL =tanlycoth(y)- rltanhr]+ oy 2y2tanhy
EI VL r - tanh r L r -tanh r
F,L = o 4y 3 _ tanB 2y 2 tanhy
2

EI Ly-tanhy y-tanhy
(oy) (r[r(3- tanh r)- 3~rl) +(oy) tan(J.!.
ox =
2 2
r[r(3- tanh r) - 3!anhr]]+
2

L L 4(r - tanh(r)) L 12 v 4(r - tanh(r))

cothr- r(t + coth
2
r) r[r(3- tanh 2 r)- 3tanhr ]]
+tan 2 (B + - 2 -
{ 16y 16(y-tanh(y))
(4.69)

(4.70)
Similarly, for a compressive axial load

172
CHAPTER 4: FLEXURE ELEMENTS

MAL= tanJrcot(r)+ y tan r ] -

1
2y tanr oy 1

EI VL tanr-r L tanr-r

MoL =tanJycot(y)- r,tanhr]+ oy 2r,tanr

EI VL tanh r - r L tan r - r

- =o 4y 3
1 2
~L 2y tany
- - tan 8----'-------'--
El L tan r-r tanr-r

~ =(oy) (r[r(3 + tan r)- 31tanr]) +(oy) tan(B)[_!__ r[r(3 +tan r)- 32tan r]l +
2 1 2

L L 4(tan(y)- r) L 2 4(tan(r)-r)

cotr - r{I +coer) r[r(3 +tan 2 r)- 3tany ]]

+tan 1 (B)[ + - 2 -
16r 16(tan(r) - r)
(4.71)

Similarly, these can be expanded in series to give

(4.72)

It can be easily verified that by setting (} to zero in equations (4.69) - (4.72)

reduces them to the simpler versions (4.59)- (4.66) already derived while setting
y to zero results in the original simplified expressions for a beam with no axial
load.
4.1.5 Leaf type flexures for paraUelogram flexure applications
Because most mechanisms require either a pure rotation about a fixed point or
linear translation in a known direction, the arcuate displacement of the cantilever
is of little utility. Instead, following the analysis of Smith and Chetwynd, 1992,
consider the cantilever of figure 4.13 in which the applied force is transmitted to
the beam through a drive bar which, as a consequence of its geometry, imposes

173
FLEXURES

F,~
Figure 4.13: Cantilever beam subject to a tangential force plus bending
moment at the free end
an add itional bending moment. For this particular leaf spring, the bending
equation is given by

d 2y
El -
dx2= Fy s- Fy (L - x)
(4.73)
= Fy(s+x - L)

Again, integrating twice, substituting the same boundary conditions as

above and assuming small displacements
2
E/0 =Fy((s - L)x+ x )
2

Ely = F ( (s - L)-
x2 x3)
+-
(4.74)

Y 2 6

It is clear from the first of equations (4.74) that if s < L then it is possible for
the slope 0 to be zero at a point along the beam other than at the clamped end.
Such a condition is governed by the equation

s= L- x/2 (4.75)

Consequently, the condition for zero slope at the free end is re$idily found to
occur when s = L 12. Substituting this condition into equations (4.74), the slope
and deflection at the free end of the beam are given by

174
CHAPTER 4: FLEXURE ELEMENTS

2
Lx+x-) =0
E/8x• L = FY ( - -
2 2
3 2 3
(4.76)
E/u
:J' X•
L = EI§ = Fy ( -x6 -Lx
- )
4 =-F-
L
y 12

The second of equations (4.76) can be reananged into the more familiar form

(4.77)

where K 6y F.• is the linear stiffness of the cantilever beam when used as a
linear spring and the two subscripts represent the displacement and component
of generalized force respectively. Each of the subscripts has a subordinate
indicating the coordinate in which they are acting. In many cases, the context of
the discussion renders such notation unnecessary. Equation (4.77) corresponds to
the term on the left of the parentheSes for the more complex situation in which
there is also an axial load, see equation (4.59).
It is a simple matter to determine the bending moment applied to the end of
the beam for a given deflection. Consequently, the bending stiffness is given by

(4.78)

Again, equation (4.78) can be compared with equation (4.61) that includes
the effects of an axial load.
At first glance, it appears that we have a perfect mechanism producing a
pure, rotation free translation. However, it is obvious that for the neutral axis of
the beam to maintain it original length, the actual motion of the beam will be an
arc with an accompanying displacement along the x-axis. For small deflections
the incremental length dL. of the beam along the neutral axis can be obtained
from the integral

dL· = ~dx
(4.79)
= l + (~r( x • _ ~3
4
+L2:2)dx
Noting that from equation (4.77)

(4.80)

175
FLEXURES

500

,....,
~ 400
§
·c 300
·;"'
>
i 200
·u
:X:
100

0
0 2 3 4 5 6
Platfonn displacement (mm)

Figure 4.14: Comparison between theoretical model, equation (4.83), and parasitic motion
for a simple leaf type hinge measured by R.V. Jones, 1951

the square root term in (4.79) can be expanded giving

~ =l+(;IY(xs4 -~3 +L2;2)

. -i(rir(
x44- ~3 + L:2r .. (4.81)

Substituting the approximate expression in (4.81) into (4.79) and integrating

eventually yields

L.l':jL+-Y
38
SL
2
(4.82)

Because of the zero slope at each end, it could be argued that the beam is
straight at these points. Consequently, it is reasonable to assume that the length
of the neutral axis would be restored to its original value by introducing an x axis
displacement, ox corresponding to the height deviation, or foreshortening, of the
flexure of magnitude

382
0 =--y (4.83)
X SL

176
CHAPTER 4: FLEXURE ELEMENTS

Typically, the displacement of such a flexure would be less than 1% of its

length corresponding to a parasitic displacement of 0.6%. In a study of a range of
linear translation flexure designs, Reginald V. Jones studied this, among other,
so-called 'parasitic' motion error. Results for a simple flexure of length 45 mm
and thickness 0.56 mm subject to rather large deflections of up to 6 mm are
presented. This was the more complex simple linear spring mechanism to be
discussed in chapter 5. The height error measured in this study is plotted in
figure 4.14 with the theoretical predictions also included. Even for these
relatively large deflections, the theoretical prediction is reasonably close to the
experimental results.
To assess more fully the magnitude of parasitic errors in these simple
elements, it is necessary to investigate the influence of forces applied in other
directions. However, an arbitrary load applied to one end of the beam can be
resolved into three orthogonal forces and torsional moments. Because only small
deflections are usually considered, the principle of superposition will apply.
Consequently, in many cases distortions and stresses due to each force can be
considered in isolation and the effects summed to yield the resultant.

4.2 Notch hinge

Next to the leaf type flexure, the notch hinge is probably the next most popular
element. It is for this reason that this element will be discussed in some detail.

z
Figure 4.15: Hinges of elliptic geometry, a) the circular hinge (E =1), b) the elliptic hinge, c) the
leaf hinge (E = oo)
This section presents closed form equations, based on a modification of those
originally derived by Paros and Weisbord in 1965, for the mechanical compliance
of a simple monolithic flexure hinge of elliptic shape, the geometry of which is

177
FLEXURES

determined by the ratio & of the major and minor axes. It is shown that these
equations converge at & = 1 to the Paros and Weisbord equations for a hinge of
circular section and at s::::) oo to the equations predicted from simple beam
bending theory for the compliance's of a cantilever beam. These equations are
then assessed by comparison with results from finite element analysis over a
range of geometry's typical of many hinge designs. Based on the finite element
analysis, stress concentration factors for the elliptical hinge are also presented.
From finite element analysis and experimental data, it has been found that
predictions for the compliance of elliptical hinges are likely to be within 12% for
a range of geometry's with the ratio Px (= /5.ax) between 0.06 to 0.2 and for
values of & between 1 and 10, Smith et al., 1997.
Most flexure systems may be divided into two broad categories, notch and
leaf type hinges. Making two holes with a small separation between them to
form a circular notch, or web, produces notch hinges. It is this thin web which
serves as the flexible element (Figure 4.15(a)). Leaf type hinges typically consist
of a slender member connected at each end by two rigid bodies to provide a
compliant coupling (Figure 4.15(c)). The leaf can be monolithic or fabricated by
clamping each end of a thin-strip. Because of its relatively high off-axis stiffness',
the notch hinge is correspondingly more immune to parasitic forces.
Concentration of the stresses near to the thinnest portion of the notch also results
in a localization of strains therefore providing a well-defined axis of rotation. As
a direct consequence of this, high local stresses limit the deflection of the notch
hinge. The leaf type hinge distributes the deflection over the length of the hinge,
thus lowering stress and allowing greater deflection for a given hinge length. If
axial forces are present, the effective pivot point is not localized and moves along
the leaf hinge as it deflects. While each type of hinge is frequently used, there are
many applications where the optimum geometry is likely to be intermediate
between the two extremes. Up to the present time, the designer has had only
these two extreme options available. The reasons for these two common
geometry's may be discerned by looking at them from a manufacturing
standpoint. Traditionally, leaf type hinges have been fabricated by clamping a
thin strip (the leaf) at its ends or by machining the leaf out of a single piece of
material where the leaf thickness permits this. The notch hinge has been
traditionally produced by drilling and reaming (or jig-boring) two closely
spaced holes to produce the hinge. Thus the circular notches and parallel beam
flexures are constraints imposed by the manufacturing process. With the advent
of CNC milling machines and, in particular, CNC wire electro-discharge
machining (WEDM), hinges of arbitrary shape can now be readily produced.
In this section some of the merits and limitations of monolithic elastic hinges
for use as a single rotational degree of freedom mechanism are considered. In
particular, a hinge formed from a web that would remain if two elliptical holes
were machined from each side of a rectangular bar is assessed, figure 4.15.

178
CHAPTER 4: FLEXURE ELEMENTS

Clearly, the hinge geometry is related to the ratio, &, of the major to minor axes of
the ellipse which ranges from 1 for a circular notch to infinity for a leaf type
spring. An assessment of the stiffness and induced stresses in such a hinge when
subject to a pure bending moment has been performed by comparison with
results from continuum mechanics, simple bending theory, finite element
analysis and experimental measurements. In reality, hinges are likely to
experience combinations of both bending and shear stress, the magnitude of
which will depend upon the geometry of the complete flexure mechanism and
the nature of the applied loads. Although, in theory, it is possible to produce
pure rotation or rectilinear motion by correct application of applied loads, in
practice, undesirable or 'parasitic' forces and moments about other axes will
always be present. Subsequent off-axis distortions of an elliptical hinge can be
calculated using the equations for compliance in the other axes, also presented in
this section.

4.2.1 Theoretical considerations

Formula for the stiffness of the leaf spring and circular notch are presented in the
next two sections. For comparison between simple bending theory and finite
element modeling, an expression for the bending stress in a circular notch hinge
has been derived from the continuum mechanics solutions first presented by
Ling in 1952.
4.2.1.1 The leaf type flexure reconsidered
From simple bending theory, the angular stiffness of this type of hinge is given
by the equation

M EI
Ktih = - = - (4.84)
(} 2ax

where 2ax is the length of the hinge, E is the elastic modulus, I is the second
moment of area about the neutral axis, M the bending moment and B the angular
deflection about the neutral axis. This rather unusual definition of the length of
the beam has been chosen for comparison with notch hinges of circular and
elliptic geometry.
Ignoring stress concentrations at the outer edges at each end of the flexure, at
the onset of yielding, the stress in such a hinge for a given maximum angular
displacement, Bm••' can be derived from simple bending theory

Et Et
O"y = -(}max = -Bmax (4.85)
2L 4ax

Rearranging equation (4.85), the thickness of such a leaf spring is

179
FLEXURES

4a
I = - - x - 0" (4.86)
EBmax y

Note that the above equations are independent of the depth, b, of the spring.
At this angular deflection, the maximum stiffness for a given deflection
(which is also proportional to tlte maximum strain energy that can be stored in
such a hinge) is given by

(4.87)

This illustrates the well-known design rules that for a given material and
length of leaf spring, the stiffness scales linearly with its depth.

4.2.2 The circular notch binge

For the notch hinge of figure 4.15(a), an approximate solution for the angular
compliance was first presented by Paros and Weisbord, 1965 (replacing the
symbol R used in this paper with ax), and is given by

--=-
B,
KlllMr M,

1+ {3 + 3+2/3+/3 ]~1 - (1 + /3 - )2 +
2

3 [ I J [ rl r (2f3+ /31) r (4.88)

1
= 2Eba; 2{3+/3 [ 6(1+/3) ]tan-~(~2 + /3 (r- {3) )
2
( /3+/32)3'2 13 ~l - (l+f3 -r)2

For full semicircular notches that are considered throughout this section, the
dimensionless parameter y is given by

D 2ax +t t
r = - = - - = l+ - =l+/3 (4.89)
2a 11 2a11 2a.

whereupon the full expression for the hinge compliance in (4.88) reduces to

180
CHAPTER 4: FLEXURE ELEMENTS

(4.90)

3
= 2Eba~ f(P>
where f3 is the dimensionless factor representing the hinge geometry and f(/l)
is a dimensionless compliance factor
For small values of {Jequation (4.90) can be simplified to produce

K ~ 2Eba~ (2/J)S/2 2Ebt 512

O:Mz 3 3n- - 9;ral/2
(4.91)
X

This derivation has been obtained by simple integration of the bending

equation for elemental strips of the hinge. This is not a continuum mechanics
solution and relies on the slightly erroneous assumption that the principal
stresses at all points are normal to the yz plane. Consequently, this equation
cannot be used to calculate the maximum stress that occurs at each of the outer
surfaces of the thinnest part of the notch. Using Fourier integral methods and a
technique referred to as promotion of rank, Ling, 1952, has obtained a full
solution for the stresses in a notch hinge subject to pure bending and this, in
tum, has been experimentally validated by the photo-elastic studies of Frocht,
1935, and Goodier, 1941. This solution is reproduced in graphical form in
Peterson, 1995, in terms of a stress concentration factor K, representing a
multiplication so that, for an applied bending moment M, the true stress, cr, can
be calculated from a nominal stress using the equation

(4.92)

Over a very wide range of values for fJ, the stress concentration factor, to
within better than 2%, is given by (see figure 4.16)

K, = e;) -9/20

t ) 9/20
=( l + - 0 < /3 < 2.3 (4.93)
2ax
= (I + Pt20

181
FLEXURES

2.4
2.2
2
-+- Kt
1.8
~ - Fit
1.6 -+- Kt tension
1.4
1.2

0.2 0.4 0.6 0.8

2RI(2R+t)=1/(l+l3)

Figure 4.16: Stress concentration in a circular notch hinge as a function of 2ax l(t + 2ax)
(from Ling, 1968, Pilkey, 1997). For interest, the stress concentration for a tensile axial load
applied to the notch is also included.

Many designs often require a specified angular displacement and therefore

the applied bending moment is unknown. In this case, the approximate
compliance (4.91) (to be assessed shortly) is used to estimate the bending
moment for a given angular displacement B. Substituting this into equation (4.92)
provides an estimate of the stress at the outside edge of the thinnest portion of
the hinge given by
20
a = 4Ea; (l + Pt20 8 = E(l + !3r' 8 (4.94)
f(/3)1
2
/32 f(/3)
At a first glance, it would appear that the stress increases with a reduction in
thickness, t. However, a reduction in the web thickness also corresponds to a
reduction in /3. Because the geometric function in the denominator increases
more rapidly than the inverse square of 13, the stress as given by equation (4.94)
reduces with thickness (and therefore /3) as would be expected. At the onset of
yielding, the maximum thickness of the notch for a given deflection of the hinge
can be determined from the equation

t 2= - - (l + /3)9'2o - B
4Ea;
f(/3) (jy

or, from equation (4.91)

182
CHAPTER 4: FLEXURE ELEMENTS

(4.95)

Substituting equation (4.95) into (4.91), the maximum stiffness for such a
hinge is given by

(4.96)

Comparison of the above with equation (4.87) indicates different parametric

relationships for the maximum stiffness for the two extremes of elliptic hinge. A
generalized approach to the optimal hinge geometry is necessarily dependent
upon the design specification in terms of the load capability, displacement range,
allowable hinge volume, operating environment and other application specific
constraints. The ratio of the maximum hinge stiffness between equations (4.87)
and (4.96) is

4 2
Maximum stiffness of notch 37r (aY )
Maximum stiffuess of leaf spring ~ 8.19 K,s E 2 B
~..!..:2.(~)2
KsI EB

Based on a maximum stiffness design criterion, it can be seen that the

optimum hinge will depend on the magnitude of the required hinge distortion.
Clearly, dependent on the magnitude of the hinge distortion, an optimal
geometry is likely to be intermediate between these two hinge types and this
provides the motivation for the formulation of equations for predicting stiffness
values for an elliptical hinge geometry.

4.2.3 Accuracy of stiffness estimates for a notch type hinge

The difference between stiffness values predicted from the full (this will be
referred to as 'exact' in the following discussions) and the 'approximate'
theoretical formula given by equations (4.90) and (4.91) as a function of the
parameter f3 is shown in figure 4.17. The approximate equation (4.91) produces a
value that is lower than that of the full theoretical formula (4.90) with an error
that increases nearly linearly to 8.25% at a value of f3 = 0.3. In an effort to assess
the validity of these equations, Smith et al., 1997, analyzed a range of notch hinge
geometry's using finite element analysis.

183
FLEXURES

The true value for the stress

10 -
9 L concentration factor is taken to
8 / be that produced from
~ ~
.L_
L continuum analysis. Based on
~ 5 / this, it was found that for more
/
i5 4 L than 1000 elements the stress
~ 3 /
2
/
differences were less than 1%.
1
0
/ As yet, complete solutions for
0 0.1 0.2 0.3 0.4 the prediction of hinge stiffness
have not been produced.
Using meshes of between
Figure 4.17: The percentage difference between the full 1,000 and 2,000 elements, the
stiffness (4.90) and values predicted by the approximate stiffness values for a variety of
equation (4.91) notch hinges of varying p were
computed and compared with
the theoretical predictions given by equations (4.90 & 4.91). The results of this
analysis are shown in figure 4.18. Each model predicts a different value for the
stiffness of a hinge. Assuming that values derived from finite element analysis
represent the 'true' stiffness, it is apparent that the approximation to the precise
formula is the more accurate predictor for stiffness and that both equations
predict a stiffness that is too high. The percentage error between the true stiffness
and values predicted by the approximate equation (4.91) is shown in figure 4.19.

450
400
_,....... 350 +-- - · - - - - - - -
lo
e 3oo
~k(ex)
~ 250
-k(appr)
Cll 200
~ 150
~~~fe!:"l
(/.) 100
50

0 0.05 0.1 0.15 0.2 0.25 0.3

Figure 4.18: Comparison of stiffness calculations for a semi-circular notch hinge of

dimensions R = 10 mm, b = 1 mm, E = 207 GPa using equations (4.90), (4.91) and finite
element analysis

Finally, the errors between maximum stress calculated from the continuum
mechanics solution and the values derived from both finite element analysis and
using the bending moment calculated from the approximate theoretical equation

184
CHAPTER 4: FLEXURE ELEMENTS

(4.91) due to the angular displacement from FEM results are plotted in figure
4.20. There is agreement between
14 the continuum mechanics and
12 +-----------~~ ~---- FEM models to better than 2%
10 +-----------~~ -------- over the complete range of 13 from

J~ 8 +-------~/~
6 .f----
---------­
/_,..1111/F~----
0.05 to 0.3 while the approximate
formula introduces an error that
4 +----~~-------------- is slightly, but not significantly,
2+--.r
~--------------- larger than the stiffness error of
o +-~--~----r-----~--~ the same analysis.
0 0.1 0.2 0.3 0.4

4.2.4 The notch hinge of elliptic

Figure 4.19: Percentage error between stiffness cross section
predicted from finite element analysis and the
approximate formula ( 4.91) for the data shown in the
Considering the elliptical hinge
previous two figures geometry of figure 4.21, the
height y at a position x on the
hinge axis is given by

(4.97)

0 0.1 p 0.2 0.3

0
-2

..
§
-4
-6

!
-8 ~ Errr calc/FEM

-10 - % Errorcalc/app
Cl)

~ -12
-14
-16
-18

Figure 4.20: Errors between maximum stress calculated from the continuum mechanics
solution and the values derived from both finite element analysis and using the bending
moment calculated from the approximate theoretical equation (4.91)
Following the analysis of Paros and Weisbord, 1965, an approximate
expression for the compliance of the elliptical hinge can be obtained by splitting
the hinge into thin vertical strips and integrating the bending equation to give

185
FLEXURES

Figur e 4.21: Parametric model of the elliptic flexure hinge

() = [~dx
' a EJ, (x)
(4.98)
_ [ 12M, dx
- a Eb(2y(x)r

Substituting equation (4.97) into the second of equations (4.98) yields

() = [ 3M, dx (4.99)
2)1/2]3
(aY+ Y2) - aY 1- :;
r a [ (
2Eb

Using the substitution x =ax sin(), equation (4.99) can be rearranged to yield

() = 3M.ax 3
"J 12
cos() d()
~
3
• 2Eba 12 ( (
1+ }lza)l -cos())
)
Y -"
(4.100)
=c. J cos() 3 dB
(C2 - cos B)
The second equation in (4.100) is identical to the integral used in the paper of
Paros and Weisbord except for a multiplication factor of the ratio of major to
minor axes & = /ay' and with aYreplacing the notch radius R, and PYreplacing
0

186
CHAPTER 4: FLEXURE ELEMENTS

400~------------------------------------~

400+-------------------------------------~

_,...... 300 +--------------------------------H~

~ 300 +-------------------------------rT-~~
! 250 t-----------------7'7£7¥-~
~ 200 +-- - - - - - · - - -- ------------::..s"'-fiY::....__-7"-
~ 150+---------------------~~~~~~~~
~ 100+---------------------:;~~~~~~------
50
0~--~~~~~~--~--~--~
0 0.05 0.1 0.15 0.2 0.25 0.3
f3x
Figure 4.22: Calculated value of stiffness for an elliptical hinge for ratios of major to minor
axes of the ellipse of 1 (circular notch), 1.2, 1.6, 2, 3, 10, 100, oo (leaf spring). Parameters
for calculation are a• = 10 mm, b = I nun, E = 207 GPa.

p. Therefore the compliance equation (4.90) can be replaced by a new expression

for an elliptical hinge given by

(4.101)

where

(4.102)

Because the length of the hinge, 2a., is of importance in most designs, it is

more convenient to express equation (4.101) in the alternative form

(4.103)

Setting e = 1 the above equation is the same as that for a circular notch hinge.
As the major to minor axis ratio is increased, this corresponds to the hinge
becoming more elliptic until, as e tends to infinity, it can be easily shown that the
stiffness converges to that for the leaf type given by equation (4.84). If a. and P.
are maintained at a constant value, then as t increases, the stiffness value

187
FLEXURES

equation (4.103) transforms from a circular notch to a leaf type spring while
maintaining a constant thickness at the center. This has been plotted in figure
4.22 for values of£ from 1 to 100 with the leaf spring formula of equation (4.84)
included to represent the value of stiffness with£ at infinity. For P, greater than
0.06 there is less than 10% difference between the stillness values for e = 100 and
the cantilever formula of equation (4.84). A vertical line through the graph of
figure 4.22 represents values of stiffness at constant hinge thickness as the major
to minor axis ratio is increased. This is plotted in figure 4.23 for p, ranging from
0.02 to 0.2.

160

.-. 140 13.

-] 120 -+-0.02
a 100 - 0.08
b 80 --6-0.12
3 60
- 0.14
-11-0.18
~
en
40 -+-0.2
20
0
0 2 4 6 8 10
£

Figure 4.23: Stiffness variation of an elliptic beam of length 20 mm and depth 1 m.m for a
range of values of P,

4.2.5 Compliance's of elliptic hinges in other axes

In this section, the formulae
20 for compliance of elliptic
18 P,=o.os hinges in other axes are given
16 in terms of the appropriate
14 modifications to the Paros
.-. 12 and Weisbord equations. The
.s 10
C!l.
rotational compliance about
/~ 0.1 the y-axis, see figure 4.15, is
r_,-------0.15 ____
00

w : given overleaf by
4 y-- 0.2
2
0+-----~--~----~-----r----~
0 20 60 80 100

Figure 4.24: Geometric stiffness factor sg(sftx)

188
CHAPTER 4: fLEXURE ELEfvfENTS

(4.104)

As the ratio & tends to infinity, the factor eg(ef3x) converges to f3x , figure
4.24, which corresponds to the stiffness formula for a cantilever as deduced from
simple beam theory.
The linear compliance corresponding to tension or compression of the hinge
along its axis is

_ I_ =~ = _!_g(sf3) (4.105)
KfitFx ~ Eb X

In the z-axis, the bending compliance is given by

(4.106)

The second function in parentheses in the above equation can also be shown,
see figure 4.25, to converge to the correct value of 2/3" I 3 as e tends to infinity,
again corresponding to the formula deduced from simple beam theory.
The shear compliance's in they and z-axes are identical and can be computed
from the equation

(4.107)

189
FLEXURES

Comparison with
14
equation (4.104) verifies that
this also produces the
~.. = 0.05
formula for a simple shear
element as e tends to infinity.
0.1 The above equations for
the compliance of an ellipse
0.15 all converge to the stiffness
of a notch hinge for e = 1 at
which value fJ x =fJY =fJ and
0+---~~---r----~----~--~ a.. = aY = R. Under these
0 20 40 60 80 100
conditions, it is possible to
use the simplified equations
Figure 4.25: Geometric stiffness factor &h(efJ.. ) of Paros and Weisbord
which, for small values of 13,
are given by the equations

(4.108)

F 2Ebt 512
(4.109)
K lkF)' = B: = ~1 - /3 2 9ttRm

(4.110)

(4.111)

(4.112)

190
CHAPTER 4: FLEXURE ELEMENTS

4.2.6 Results
14
12
Both finite element analysis
and direct experimentation
10
_. have been used to assess the
8
- . - - - - - - - - ---- stiffness predictions of the
~ 4
6
above equations. Based on the
~ assumption that stresses
2 - - - - - - - - - ---• computed from finite element
0 analysis correspond closely
-2 (i.e. to better than 2%) to
-4 predictions based on
0 2 4 6 8 10 continuum mechanics
solutions. An assessment of
Figure 4.26: Percentage stiffness error for the elliptic stress concentration factors
runge as a function of e: for values of flx = 0.06, 0.12, 0.2 for the elliptical hinge is also
presented. It should be noted
that the figures in the following sections give the percentage errors between
theoretical and expected values. Typically these turn out to be within a few
percent and represent values that are often adequate for most engineering
purposes.
~.2.6. 1 Finite element results

Errors between stiffness values predicted from finite element results and the
fu ll theoretical equation (4.90) are shown in figure 4.26. In view of possible error
mechanisms, for the hinge geometry's considered the theoretical models were
shown to produce stiffness
1.01 values that are within 12 %.
1.06 Because, as the hinge tends to
a leaf spring, the equations
1.05
converge to those derived

-
~
1.04

1.03
from the bending equation,
the accuracy of predictions
will correspondingly
1.02 converge. Equivalently, it can
1.01
be stated that the circular
notch hinge represents the
1 worst case error between
0 2 4 6 8 10
theoretical and finite element
models.
The stress concentration
Figure 4.27: Stress concentration factors for the elliptic
factors have also been
hinge flx =0.06, 0.12, 0.2
assessed and the results are
shown graphically in figure

191
FLEXURES

4.27. In the graph for the concentration factor for p~ = 0.06 one of the data points
appears to be slightly erroneous. However, it is apparent that the difference is
only a factor of less than 1 % and could easily be due to errors in analysis and
data processing. For the circular notch hinge, this data converges to within 2 % of
the 'exact' figure predicted from equation (4.93). In all cases the increase in stress
is less than 10 %.
Experiments with aluminum hinges showed the relative errors between
measured and theoretical results were less than 10 % and within 12 % of the
finite element analysis, Smith et al., 1997. These errors are considered to be within
acceptable limits for values of E between 1 and 10 and represent a reasonable
confidence limit for most design purposes.

4.3 Other hinge elements

Although the notch-based hinge is relatively simple to implement and integrate
into more complex mechanisms, it suffers from a limited deflection range.
Historically, the more compliant cross strip pivot has been used for hinges
requiring larger deflections. Such a hinge is relatively simple to construct,
consisting of two or more flat strips each attached to a fixed base at one end and
the moving platform at the other, figure 4.28. As can be seen in figure 4.28, each
strip is mounted at an angle to the other, normally 90 degrees, and the hinge
SIDE VIEW

u
a) - - - -- t - - - - __ Jlinge
axis

b) c)
r---~----~--~----~

Figure 4.28: The cross strip pivot. The center line of intersection of the flexures represents the
hinge axis, a) the simple two strip pivot, b) the symmetric, four leaf pivot, c) the symmetric
three leaf pivot (center Ieafis twice the width of the strips either side)

192
CHAPTER 4: FLEXURE ELEMENTS

pivot is assumed to be about the axis of intersection between the strips. Figure
4.28(a) shows the simplest design in which the hinge is produced by clamping
two successive strips of equal width. Although this may be satisfactory for many
designs, it has a relatively low torsional stiffness about the vertical axis. Being
symmetric about the vertical axis, a more desirable hinge is shown in figure
4.28(b). This consists of two pairs of strips. The strips at each end, being parallel,
form the first pair while the two in the center form the second. Each of the two
pairs are arranged at an angle as for the simple pivot of figure 4.28(a). Because
the deflections of each hinge pair are identical, it makes no difference to the
operation of the spring if the central pair is joined thus leading to the equivalent
three-strip hinge shown in figure 4.28(c).
Moving the two strips together so that they effectively cross through one
another produces a more compact, and symmetric, design. Clearly, to maintain
contiguity it is necessary the strips be joined along the hinge axis. For obvious
reasons, this is referred to as a 'cartwheel' hinge, Howells, 1995. In the past, there
has been less interest in such a design because of the difficulties associated with
producing such a shape. However, modern manufacturing techniques such as
advanced welding processes and wire electro-discharge machining readily
produce such shapes, see chapter 8. The geometric similarities between the
simple cross strip and cartwheel hinges result in similar performance
characteristics. A comparison indicates the former to offer larger deflections
while the latter provides a more stable position of the pivot axis. Each are
considered separately in the following.
4.3.1 The cross strip pivot
Although these have been used for many years, there have been relative few
studies of this particular mechanism. In particular, Youn~ 1944, undertook an
experimental evaluation of the four-strip pivot, the results of which were used to
corroborate the theoretical investigations of Haringx, 1949, upon which the
following analysis will be based. At around the same time as Haringx, Wittrick,
1948, 1951, carried out a similar analysis that, where the spring geometry and
load conditions were similar, produced exactly the same results. Wittrick's later
paper extends this analysis to an investigation of the influence of moving the
point at which the pivots cross. In this work, which is not included in this
section, it is shown that the center shift of the pivot under load can be reduced by
selecting the point at which the strips cross. Although not experimentally
verified it is demonstrated that a cross strip pivot arranged so that the strips
cross at a point 87.3 percent (=1/2 + ..f5/6) of the distance along each strip will
maintain the center of pivot more accurately under the influence of applied
loads. Wittrick also demonstrates that a hinge produced from strips that cross at
one end will always remain stable in the presence of tensile forces. Such a pivot
is presented towards the end of case study 1 in section 7.5.

193
FLEXURES

In most applications, the cross strip

hinge will be subject to applied forces
that can be resolved into an axial load
and a pure bending moment as shown
in figure 4.29. Applying a pure bending
moment to the moving platform will
result in the deflections shown in
figure 4.30. Because the ends of each
beam of the pivot will undergo similar
displacements it can be seen from this
figure that for small displacements, the
angular deflection of the upper
platform, B, is given by
Figure 4.29: Mathematical model for the
symmetric, cross-strip pivot 2oy sin a 2oy
B= =- (4.113)
Lsina L

or

0y = LB (4.114)
2

where L is the total length of the cross-strip and ~ is the deflection of the strip in
a direction perpendicular to its axis.
For a pure applied couple, M , and no other applied forces, it is relatively
easy to show that the angular stiffness of the cross strip pivot is given by

(4.115)

where n is the number of strips and I is the second moment of area of each
strip about the neutral axis.
Equation (4.114)
implies that for small
M. displacements the pivot
point coincides with the
( initial intersection of the
flexures irrespective of
the angle of intersection.
Figure 4.30: General mathematical model for a single However, for small
flexure of the cross-strip pivot angles of intersection, the
position of this
intersection becomes both incr~asingly sensitive to deflection while the effect of

194
CHAPTER 4: FLEXURE ELEMENTS

manufacturing errors make it difficult to predict the exact position of the

intersection upon assembly. Relatively large intersection angles, 2a, will be
considered throughout.
The general applied loading for each beam is represented in figure 4.30.
From this, taking bending moments to the left of an arbitrary position x, the
bending equation is given by

d 2y
E/ dx2 = Mo - F;,x + ~y (4.116)

Again, defining

X =!_
L
4y 2 = ~£2
EI
2 (4.117)
FL
({J =-)1-
EI
M L
m = -0 -
0 EI

equation (4.116) can be rearranged in the form

(4.118)

For boundary conditions of zero slope and deflection at the fixed end, the
general solution is

y= Lm~ (cosh(2yX) -1)- L~ (sinh(2yX) - 2yX)

4y 8y

and

-dy =-m0 sinh(2yX)-

•
- ({J 2 ( cosh(2yX) - 1)
dx 2y 4y
(4.119)
d2y mo ({J .
- 2 = - cosh(2yX)- -sinh(2yX)
dx L 2yL

These equations are similar to those of (4.46)-(4.52) and are identical with the
parametric equations (4.69) and (4.71). For this isolated flexure the total bending
moment M, applied at the end of the beam is given by

195
FLEXURES

(4.120)

which can be written in dimensionless form

(4.122)

Substitution of equations (4.114) and (4.120) into (4.69) leads to the relatively
simple result

B
M,(L) r
El = tanh(y) + r
2
(4.123}

In series form this can be written as

(4.124)

which is different to the solution given in the paper of Haringx. For compressive
loads through either a similar procedure or by the substitution

r = ~F~~2 = ;~-it (4.125)

Substituting (4.125) into (4.123), a solution is immediately given by

B
M,( ElL) = tan(y)
r
- r
2
(4.126)

or in series form

(4.127)

Equations (4.123) and (4.126) have been plotted in figure 4.31. It is clear from
this plot that for relatively low loads the simple equation for a cantilever stiffness
(4.115) can be used with little error. For a high compressive load equation (4.126)
will tend towards infinity (at y = 1t) again resulting in the Euler buckling
condition

(4.128)

196
CHAPTER 4: FLEXURE ELEMENTS

Not surprisingly, this

also corresponds to equation
(4.68). From the series
expansion of (4.127) or the
graph of figure 4.31 it can be
seen that the bending
stiffness drops to zero at a
value of

y
r ~ ~ ~ 0.866 (4.129)

Figure 4.31: Plots of the moment equations (4.123) and It can be seen from figure
(4.128) showing the effect of dimensionless axial load, y, on 4.31 that for the relatively
the normalized bending stiffuess of a single flexure of a broad range of loads
cross strip hinge
between y = ± 1/4 the
. bending stiffness will remain
within 10 % of the nominal value with zero axial load.
It is normal to determine the stresses from simple bending theory. To
account for the effect of the axial load, it is possible to determine the maximum
stresses due to each individually and simply add these to obtain the resultant.
These can be readily obtained from the formulae
Et
0' =-8 (4.130)
b 2L

F
a =--L (4.131)
" tb

from which the maximum total stress that occurs at the outside edge of the
flexure is given by

a = Et B+ ~ (4.132)
2L tb

At the limit of axial loading (i.e. y = 1/4), the force is given by

F = El2 = Et 3b2 (4.133)

" 4L 48£

Assuming an upper limiting stress 0' max the maximum angular deflection
()miX is
I

197
FLEXURES

2L t
(4.134)
24£
(} = - (j --
max Et max

In the above, the influence of stress concentrations at the junction of the

flexure and rjm have been ignored. Some issues are deferred to chapter 8, in
which practical implementation and fabrication are discussed.
4.3.1.1 Center shift of the pivot
It is of interest to assess the stability of the pivot axis during deflection. For a
pure bending moment, the flexures are distorted to arcs of a circle. For a total
angular deflection, (}, the shift of the pivot, 8 P is always towards the moving
platform and its magnitude for a simple cantilever is given by, Haringx, 1949

8P = 2sin(B I 2) - cos((} I 2) (4.135)

L (}

For flexures crossing at an angle 2o:, the deflection is modified to

8P =_ l _ (2sin(B I 2) _cos({} 1 2)} (4.136)

L cosa (} 'J
It is common to use a cross strip pivot in which the flexure bisects at 90
degrees. In this case, expanding (4.136) into a series and, assuming the angular
deflection to be small, this reduces to the approximate equation of Howells, 1996,

8p .fiB2
-=-- (4.137)
L 12

Equations
0.12 (4.136) and
Q (4.137) are
0 0.1
'::l
~
!-Howells ' plotted in
0
.... 0.08 1-Haringx l figure 4.32
~p_ l
(I)
(I)
with the
-a0
Q)
0.06
experimental
·~ 0.04 data of
4)

.§ 0.02 Young, 1944,

0
included for
0
comparison.
0 10 20 30 40 so 60
It can be
Angular deflection (deg.) seen that
there is a
Figure 4.32: Center shit\ of the pivot point of a cross strip pivot (a = tr/4) reasonable

198
CHAPTER 4: FLEXURE ELEMENTS

correspondence with the approximate formula even for relatively large angular
deflections.
4.3.2 The cartwheel binge
The cartwheel hinge is similar to a cross strip pivot. Both consist of a pair of leaf
type flexures, usually at right angles to each other, with the point of intersection
being considered the pivot point.
Cartwheel hinges, see figure
4.33, differ from cross strip
pivots in that the flexures
coincide at the hinge axis and are
joined. Such a geometry is both
symmetric and, unlike the cross
strip pivot, amenable to
manufacture using wire electro
discharge machining. Because of
this, the cartwheel hinge can be
produced from a single
pivot monolithic piece of material. For
some designs, the cartwheel
hinge may be integrated within
complete monolithic
mechanisms and therefore
reducing some of the problems
Figure 4.33: The cartwheel binge
associated with assembly.
Following Howells et al., 1996,
because of the symmetry, it is possible to analyze a single spoke and apply
superposition to assess the
characteristics of the complete flexure.
In the absence of radial loads, the
deflections and force will be the same in
each spoke. Consequently, the angular
deflection of one spoke will be half of
the deflection between the moving
platform and base. Looking at the forces Rsin(B/2)
applied to a single spoke with no axial
load, figure 4.34, it is possible to write
the parametric equation
Rcos(B / 2) ~ R
(4.138)
Figure 4.34: Parametric model for analysis of
the cartwheel binge. B is the angular
deflection and this is usually small.

199
FLEXURES

where R is the half-length of a leaf spring, =L/2.

For zero slope and deflection at the hub, this can readily be integrated to
yield

ay x2
EI-
dx= FRx-F
y y-
2- MAX (4.139)

(4.140)

The boundary conditions at the rim (i.e. x = L) are

! lx•R= ~ (4.141)
Y]x•R =Rsin(%}
Again defining the dimensionless parameters

FR 2
rp =-y-
El (4.142)
MAR
m= - -
E/

equations (4.139) to (4.142) can be combined to yield

B= rp -2m
(4.143)
sin(%) = ~-;
Solving for m and rp finally produces

(4.144)

(4.145)

Assuming small angular deflections these can be simplified to give

(4.146)

200
CHAPTER 4: fLEXURE ELEMENTS

M ~ EI B= 2EI B (4.147)
A R L

where L is the length of the flexure spring from fixed to moving rim (i.e.
diameter of the cartwheel). For small angular deflections, B, the approximate
equations (4.146) and (4.147) into (4.138) can be inserted into (4.140) to yield the
complete set of equations for a single spoke of the cartwheel hinge

d2y =
dx 2 R
.!.[2- 3x]o
R
(4.148)

dy
dx
.!.[2x-
=
R
]o3x2
2R
(4.149)

(4.150)

From these it can be readily shown that the maximum bending moment,
Mmox, occurs at the hub of the hinge and is given by

(4.151)

There is a point of inflection, corresponding to zero bending stress, at a

position 2/3 of the distance from the hub to the rim.
4.3.2.1 Torsional stiffness of the cartwheel hinge
Because of the symmetry of the hinge, it is possible to determine the torque
required to rotate one of the leaf springs and then double this for the complete
cartwheel. The moment applied about the pivot for a single spoke must be
capable of producing both the bending moment along the beam plus the applied
force i.e.

M= FR+M.~ (4.152)

For the two combined spokes, the compliance's of each flexure will add
resulting in one half of the torque. However, there are two sets of these springs
and the stiffness of these will add. Consequently, the total torque is given by

(4.153)

201
FLEXURES

From this, the angular stiffness, kM8 , of the flexure is

4El
kMu = - (4.154)
R

4.3.2.2 Center shift of the pivot

Following similar arguments for the derivation of the height deviation for a
linear flexure, it shall be assumed that the length of the neutral axis along the
spoke remains unaltered upon deflection. We can replace the original spoke with
a new one of length R. and integrate along the x-axis to its end point, which will
be foreshortened to R• cos(B/2). Consequently, for a single spoke we can
calculate the length of the flexure, s, using the line integral

(dy)2
= ft+- - .. dx
R' 1
(4.155)

02 dx

Substituting equation (4.150) into the above and integrating yields

(4.156)

Consequently, neglecting terms in B of fourth order or above, the length of

the spoke will remain unchanged if the new radius of the rim is shortened by an
amount equal to the second term in parentheses in the last of equations (4.156).
Taking into consideration the similar motion of the pivot towards the fixed rim
and that the two springs intersect at an angle of 7t/2, the normalized
displacement of the central pivot is given by

op .fiB2
-= - -
R 30
or (4.157)
o
-
p
=--
.fiB2

L 60
Comparison between the second of equations (4.157) with the pivot motion
for a cross strip hinge given in (4.137) indicates that the cartwheel hinge

202
CHAPTER 4: FLEXURE ELEMENTS

y ~o.7&1- -1.46x+ 1.89

2 -+-kl
R' s0.98
- - 2kl
1.8
• tto..ells
.,; 1.6 - Poly. (Ho>.dls)

1.4

1.2

1+-----~--~----~----~----+---~
0 0.2 0.4 0.6 0.8 1.2
rlt

Figure 4.35: Stress concentration factors for a cartwheel hinge as a

function of the normalized fillet radius r/t. Curves represent values for a
simple cantilever with fillet from Frocht, 1951 and finite element results of
Howells, 1996

represents a factor of 5 improvement in the stability of the center of rotation. It is

also noted that the pivot point moves towards the stationary platform (or rim)
which is in the opposite direction for that of the cross strip pivot.
4.3.2.3 Stresses in the hinge
From the above simple analysis the maximum bending moment in the springs
can be used with the bending equation to provided an estimate of the maximum
stress given by

u = Mmaxt = Et e (4.158)
max /2 R

However, it is clear from the design that there will be a radius at the
conjunction of hinges with an associated stress concentration. Assuming a radius
ranging from somewhere between 1/3 and equal to the thickness of the hinge, it
is reasonable to assume that this can be considered closely equivalent to a beam
undergoing a pure bending moment with a fillet at the fixed end. Under these
circumstances it is possible to use a graph of the stress concentration factor k,
shown in figure 4.35. For the cartwheel hinge, each fillet will be subject to
stresses imposed from the moments of two spokes. Consequently, it is tentatively
suggested that the resultant stress concentration will be twice that for a single
beam and this is also plotted in figure 4.35. Finite element results derived from
the paper of Howells et a.l. also indicate that the true value will be somewhere
intermediate between the two values for small fillet radii and converges to
within a few percent as the radius approaches the thickness of the hinges. From
this graph, it can be seen that the stress concentration reaches a relatively
constant value of 1.2 in the region at which the fillet radius is equal to the hinge
thickness and this probably represents an optimal value. In fact, in the researches

203
FLEXURES

of Howells et al., it was found the stress increases with fillet radius above this
region. It is thought that this may be a consequence of the modeling. In this,
cartwheels were used with spokes having a relatively low thickness to length
ratio of 0.25/0.015. Consequently as the fillet radius becomes large, it will
effectively shorten the spokes resulting in stiffening of the cartwheel. Analysis of
stresses was performed by applying a fixed angular displacement and recording
the maximum Von Mises stress. Clearly, the effective shortening of the spokes
will increase the stress for a given distortion.
4.3.3 The cruciform hinge
It is occasionally desirable to provide a hinge with mounts displaced along a
common rotation axis. One method of achieving such a mechanism is to
construct the cruciform type hinge shown in figure 4.36. Such a hinge has
relatively low torsion rigidity along its z-axis while maintaining the high
resistance to bending about the two axes in the xy plane. In many civil
engineering structures in which a high torsional rigidity is desired, such a design
would, of course, be disastrous.
Equations for predicting the behavior of
such a mechanism do not appear to have
been investigated. However, to achieve a
rough estinuzte of the torsional stiffness of a
prismatic beam with no constraints at the
ends, it is possible to apply the principle of
superposition and consider this to be
equivalent to two beams at rights angles. For
a single beam, we have already derived the
relationship (2.104)

M= k Gt3d () (4.159)
I 3L

A reasonable estimate for the constant

Figure 4.36: The cruciform hinge that will not depart by more than 4% (see
figure 4.37) is given by the simple expression

(4.160)

Consequently, the torsional stiffness of the cruciform beam can be estimated

from the equation
3 4
K = 2( 1- 0.582t) Gt d - 0.4l 8 Gt (4.161)
II,M , d 3L 3L

204
CHAPTER 4: FLEXURE ELEMENTS

4.5

3.5
........
"$.
........ 2.5
tE
..... 1.5
g 0.5
0

Ul
-0.5 1.5 2.5 3.5 4.5

-1.5
d/t

Figure 4.37: Percentage error between torsion constant and the simple curve fit of
equation ( 4.166) as a function of depth to thickness ratio of a rectangular beam
This can be further rearranged to give

K
B, M,
=2(dt - 0.582) Gt
3L
4
-0.418 °3L
14

(4.162)
=(!!.- 0.373) 2Gt4
t 3L

The bending stiffness about the x and y axes are identical for a 'square' cross
and can be obtained from simple bending theory

Elyy E(dt 3 +td3 - t 4 )

K s,M, =L = 12L
3
(4.163)
Gt 4 (l+u{ -d + -d -1)
t
=---------------
(3

6£
Temporarily defining the dimensionless parameter z which is a function of
the slenderness of the two strips in the cross section and always greater than
unity

z =-dt (4.64)

The ratio of the bending stiffness about the x or y-axes to the torsional
stiffness about the z-axis is given by

205
 FLEXURES

Ko,M, (1 + u)(z + %
3
- 1)
(4.165)
K8,M, = 4(z - 0.373)

25 Figure 4.38
shows that for
20 values of djt of 6 or
0
more, the stiffness
'l:j
~ 15 ratio is greater
than an order of

en
~ 10
magnitude.
The objective
5 of this simple
analysis is to
illustrate the
2 3 4 5 6 7 8 essential feature of
dlt such a design. In
Figur e 4.38: Ratio of bending to torsional stiffness for the crucifonn reality, the beam is
hinge likely to be fixed at
each end to rigid
platforms and, as such, will be subject to constraints, which are n ot included in
the above model. Rigid end constraints will not effect bending of the cruciform
but are likely to increase the torsional stiffness and thus reduce the ratio
predicted by equation (4.165). Additional work is required to derive a more
complete understanding of this mechanism.
4.4 Two axis hinges
The following sections discuss elements to provide connections between rigid
bodies with high compliance in two desired axes. As before, the axes providing
compliance are considered to be the freedoms of the flexure element.
4.4.1 A simple two axis hinge (y, 0)
Attention has so far concentrated on the design of hinges to provide freedoms in
one rotation coordinate w ith constraints in the other five. In a large number of
applications it is desired that rigid bodies be connected by elements that provide
combinations of linear and rotational freedoms in two, three or more axes. A
relatively simple 'two axis' hinge, consisting of a slender beam fixed at each end,
is shown in figure 4.39. Translation is provided by distortion perpendicular to
the y-axis while rotation, B,, is achieved through the application of a pure
bending moment at the midpoint. Constraint along the axis of the beam, the x-
axis, is due to the axial compressive and tensile forces either side of such a load.
Equations for the stiffness at the center of the beam in the plane of the drawing
are

206
CHAPTER 4: FLEXURE ELEMENTS

k = 192EI = 16Ebt 3
6,F,. Ll Ll
l6EI 4Ebt 3
ko,M, = L =----y;- (4.166)

k _ Ebt
6,F,- L

L
y

Flexure depth = b
L

Figure 4.39: The simple, single beam, two-axis flexure

In the derivation of the second of equations (4.166), the singularity function

method for solution of the bending equation is of utility. In practice, it is not
possible to apply a bending moment at a point as this would induce an infinite
shear stress resulting is discontinuous slopes and deflections at that point.
Ignoring these difficulties, it is possible to solve the beam bending equation by
matching slopes at the point of application of the moment. Unless the beam is
slender, and great care is taken to avoid large stress concentrations at the point of
application of the moment, it is unlikely that the predicted stiffness will be
realized. Surprisingly, the maximum bending moment in the beam turns out to
be one half of that applied, see also Young, 1989. It is possible to derive formulae
for the stresses from simple bending theory. For a linear deflection in the y axis
direction,

(4.167)

In terms of the design stress a r , the maximum deflection in the absence of

other applied forces, moment or stress concentrations is limited to

207
FLEXURES

0 = CTyL2
(4.168)
"- 12Et

Similarly, for the beam subject to a centrally applied moment, M , , the

maximum moment in the region at which it is applied is given by

(4.169)

In this case, the maximum angular deflection is limited to

B = CTyL (4.170)
..... 4Et

For many applications, it is the ratio of relative stiffness values that provides
a measure of the immunity to parasitic forces. For linear displacements, the ratios
of stiffness in the desired y-axis to those in x and z are given by

k,51,
.-..!....L
(I )
=-
2

k6,F, b

=(4/)
2
(4.171)
k6,F,
k6,F, L

Typically, ratios of thickness of the beam to the depth will be of the order of
one tenth or greater resulting in immunity measured in relative displacements
per unit force of 10 to 100:1.
4.4.2 The two beam, two degree of freedom flexure (y, B)
Another common form of two-axis hinge designed to provide a single translation
plus rotation is shown in figure 4.40. This mechanism consists of two cantilever
beams connected to separate rigid bodies at one end and joined together at the
'free' end. Consequently! a y-axis displacement applied to the moving end will
cause each of the beams to distort in the familiar 's' shape analyzed in section
4.1.5. Because the cantilevers are series connected between the rigid bodies, each
will be subject to the same force and, therefore, displacement. The stiffness of the
complete flexure will be half of that for each cantilever i.e.

(4.172)

208
CHAPTER 4: FLEXURE ELEMENTS

Figure 4.40: The two beam, two degree of freedom flexure

To prevent angular distortion about the z-axis while deflecting in the y-
direction, a moment must be applied given by (remember that for a total
displacement o>' each beam will distort half of this)

(4.173)

For angular deflections about the z-axis, each beam will bend as a simple
cantilever with half total angle. Consequently, the bending moment required to
induce an angular deflection B, is

M = EI B,
' L 2
(4.174)
3
k _ El _ Ebt
D.M . - 2£ - 24£

Clearly, for a combination of displacement and rotation, the contributions

from (4.173) and (4.174) must be appropriately summed.
Assuming that axial forces are small (i.e. 1 - cos(x/L}~::~cosh(x/L}- l), the
stiffness of this element in the x-axis can be approximated by the angular twist at
the 'free' ends of the cantilever given by

(4.175)

From simple beam bending theory, the angular deflection at the end of the
beam is given by

(4.176)

209
FLEXURES

Upon such a deflection, there are two components of deformation that

contribute to the displacement between the rigid bodies in the x-axis direction.
The first is due to the combined equal extension and compression giving a
displacement

o = 2L F = Lt2 F (4.177)
" Elbt • 6El "

The second contribution is due to the lever arm of length s and is given by

(4.178)

Again, assuming small deflections, equations (4.176) and (4.178) can be

combined to yield the axial deflection

o = s2L F (4.179)
" El •

The total deflection is given by the sum of equations (4.177) and (4.179) from
which the stiffness is

(4.180)

The separation between the cantilever beams, s, can not be less than the
thickness t and the ratio sf twill typically be in the region of 1.5 - 2. It is therefore
reasonable to ignore the term on the left in the denominator after which the
stiffness is approximated as
2
k _ El _ Ebt t
&,F, - s 2 L - 12L ( ; )
(4.181)

In terms of a design stress <Yr , the maximum deflections for such a design
are

f) = 4<YrL (4.182)
zmax Et

o = 2<YrL2 (4.183)
ymax 3Ef

210
CHAPTER 4: FLEXURE ELEMENTS

Equations (4.182) and (4.183) have been derived with the assumption that the
other displacements are absent. For combined linear translation and rotation, it is
necessary to sum the stresses for both and reduce the range accordingly.

Table 4.4.2.1: A comparison of the stiffness formulae and maximum possible

e ections £or th e srmple
dfl I one and two beam, two-d egree o ffreed om hin\ge
Hinge parameter Two beam Single beam Ratio
hinge hinge Two beam hinge
Single beam hinge

k 6, F, 6EI Ebt 3 192E/ 16Ebt 3 1/32

-=-- --=---
L3 2L3 L3 L3
k O,M, El
-=--
Ebt 3 16EI 4Ebt 3 1/32
--= --
2L 24L L 3L

El Ebt(')' -Ebt
~e-r
k6,F,
s'L = 12L ~ L
12 s
2 2
o y max _2CTr_
L CT
_ Y L_ 8
3Et 12Et
B,max 4CTyL CTyL 16
Et 4Et

Table 4.4.2.1 compares the stiffness and maximum deflection formulae for
the one and two beam, two-axis hinge (reminder, .two-axis implies that the hinge
provides freedoms in two coordinates). In this table it is assumed that the
geometric parameters are comparable, i.e. the lengths and thickness can be
compared on a one is to one basis. From this simple analysis, it can be seen that
the stiffness in the parasitic, x-axis direction is to the square of the ratio of t/ s.
Clearly, there is greater advantage in keeping s small thus favoring wire electro-
discharge machining as a method for manufacture.
Torsion stiffness and shear stresses about the x-axis can be computed from
the equations given in section 2.8.2 assuming the hinge to be a prismatic
rectangular bar of length 2L.
4.4.3 The two axis toroidal hinge (Bx ,By)

A common technique for producing a two axis rotary hinge is to use a thin rod.
In practice this can be achieved by simply drilling holes in both rigid bodies,
inserting a rod and clamping at each end. To obtain two axes with equal
compliance in both axes it is only necessary that the rod joining the rigid bodies

211
FLEXURES

is symmetric. Consequently, either a circular or square rod may be used. Because

of this symmetry, each axis can be considered separately as a simple cantilever
for which the previous analyses are applicable.

Figure 4.41: The two-axis circular, or toroidal, notch hinge, 1 = thickness at

thinnest portion of the web

Analogous to the notch is the two-axis circular hinge of figure 4.41. Turning a
circular notch into a solid rod can produce such a hinge. The equation for the
angular stiffness in the freedom axes can be approximated from, Paros and
Weisbord, 1965

MY _ M, ER 3 (2p")'h _ Et 7 12
- - - 1':$ --- (4.184)
(}y 8, 20 20RY,

For a force applied at the end of the hinge, the corresponding angular
displacement can be obtained from

(4.185)

The linear stiffness in the axial direction is

F, _ ER(2p)~ _ Et 312
(4.186)
8, - 2 - 2RY,

Although shear compliance's are alluded to in the paper of Paros and

Weisbord, 1965, it is not readily apparent how to derive such formulae.
Additionally, there have been no systematic experiments to verify these

212
CHAPTER 4: FLEXURE ELEMENTS

equations. Being derived from the same base equations used for the single axis
notch hinge, it would be prudent to budget for errors of around 20%.
Stresses can be computed from the stress concentration factors provided in
the book of Pilkey, 1997. For bending of the hinge about either rotary axis (}x or
(}Y the maximum stress can be computed from

(4.187)

where the stress concentration factor for bending is given by

2 3

k,b =3.032-7.431( 2p )+t0.39o( 2P ) -s.oo9( 2 P )

l +P l+P l+P (4.188)
t
P=-
2R

A torsional moment about the x-axis will produce a maximum shear stress
given by

= k I6Mx
f" _.,
·~
II
1lt 3 (4.189)

where the stress concentration factor for a pure applied shear is given by

=2.000 - 3.sss( 2p ) +4.89j 2P )

2 3

k., - 2.365( 2P ) (4.190)

I+P \.I+P l +P

4.5 Case study 1: Force sensor for contact probe characterization

In contrast to the more common requirement for a mechanism to provide
specified displacements, this case study represents an application in which the
deflections of a single, two-axis flexure are used for the detection of applied
forces, Periera et al. 1998. This work was motivated by a desire to determine
contact forces between scanning probes as used in coordinate measuring
machines and the surfaces of the component being measured, see, for example,
Bosch, 1995. A coordinate measuring machine is an instrument that is used for
the precise and accurate determination of the geometry of engineering
components. Typically such a machine consists of three, mutually orthogonal
linear motion slideways connected together in series. A touch sensitive probe is
attached to the third slideway. In a scanning mode, the contact probe is
maintained in contact with the surface at a nominally constant force while it

213
FLEXURES

Figure 4.42: Digital image of three two-axis, dynamic force sensors shown with the
mounting threads upwards. The central flexure consists of two simple notch hinges with
orthogonal axes, the other two are toroidal two-axis flexures
moves around the component being measured. Subsequent displacements of the
three axes are monitored and their paths used as a measure of the geometry of
the component. However, at high scanning speeds sudden changes in geometry
will result in large accelerations and the actual contact forces may vary
considerably. To measure these forces, the monolithic flexure based sensors of
figure 4.42 were constructed.
Three features of these monoliths are
• the mounting thread
• the flexure and
• a rigid cylinder
In normal operation the mounting thread screws into the base of the
measurement frame with the flexure connecting between the base and an upper
cylinder (note that the sensors are upside down in the figure). Two capacitance
gages are attached to the rigid frame with the upper cylinders being used as the
target electrodes. The capacitance gages are mounted in the frame with their
measurement axes perpendicular to each other and both are radial to the axis of
the flexure. To measure forces, the scanning probe is contacted on either the
inside or outside surface of the cylinder at a prescribed height. Applied forces
will cause a deflection of the cylinder in a direction parallel to that of the contact.
Theoretically, the vector of this displacement can then be determined from the
outputs of the two gages.

214
CHAPTER 4: FLEXURE ELEMENTS

For this study, two flexure designs were chosen. One was the toroidal hinge
discussed in section 4.4.3 while the other consisted of two, stacked, notch type
hinges, see figure 4.43. Relevant design parameters for each will be discussed in
the following two sections.
0.0000
D 30.00
I

<-- FORCE
45. 00

86. 00

Figure 4.43: Dimensions of the two-axis force sensor

4.5.1 Toroidal notcb type flexure

The first design was a circular notch type flexure, which is a commonly used
two-axis flexure. An advantage of using such a flexure is that for an applied
radial force the magnitude of deflection is independent of the radial direction.
Because tool marks on the notch could be a source of stress concentration the
notch was turned on a lathe from one end with a sinall feed.
In all analysis, the sensors are made from aluminum and this was assumed
to have an elastic modulus, E of 68 GPa and a yield stress, O"r, of 255 MPa. For
the design of this case study the notch radius R is 0.012 m and it has a thickness
at the center of 0.005 m corresponding to a value f3 = 0.208. The axial distance, l,
from the center of the hinge to the applied force is 0.048 m. During operation the
maximum force from the probe is expected not to exceed 5 N. For the design to
remain elastic, it is necessary to determine a maximum acceptable stress.
Assuming that the elastic limit has a value of around 30% of the yield stress and
including a safety factor of 2, the limiting stress becomes 39 MPa. Referring to
Table 1 of chapter 2, it can be seen that this is very close to the lo"Yest value of the
endurance stress for aluminum alloys. Typically, calibration is an infrequent
exercise and, at this design stress, fatigue is therefore considered unlikely. From
equation (4.188) the stress concentration factor is 1.16. Consequently, if the

215
FLEXURES

bending stress is predicted from the bending equation, the limiting stress should
be further reduced to 33.6 MPa.
Application of an applied radial force at the position on the axis as shown,
will be modeled as a combined force plus a bending moment at the end of the
hinge. Applying superposition, the resultant deflections will be considered to
comprise a linear sum of the force and moments in isolation. From section 4.4.3,
the angular and linear stiffness of the toroidal hinge is given by
M 1112
- • = E--=274 [Nmrad·1]
()t 20R 112

(N m·l)

Hence angular rotation, ()M, due to the moment is

()M = force x force offset distance x angular compliance

(rad]
= 0.000875

Correspondingly, the linear deflection due to the applied moment is

ou = force offset distance x ()u [ml

= 42xl0-6

Considering now the effect of the linear force component, the angular
rotation, ()F, is

()F = force x linear compliance

[rad)
= 2.14xl0-4

From this, the linear deflection, oF, is

oF= force offset distance x angular rotation
[J.lm]
= 10.27 xl0- 6

It is clear that the deflection due to the bending moment exceeds that due to
the applied force by more than a factor of four. Applying superposition, the total
deflection, 0, is simply the sum

[J.lffi]

216
CHAPTER 4: FLEXURE ELEMENTS

In practice, the linear force will impose a shear stress in the hinge that is
likely to be insignificant in comparison to the direct stress due to the bending
moment. Consequently, this is considered to adequately represent the maximum
stress in the hinge given by the bending equation (2.69). Noting that the second
moment of area at the thinnest part of the hinge is tr/ 4 I 64, the bending stress, a ,
is

a = Mt 121 = 32Fl I tr/

3
=19.6 [MPa]

In view of the low estimate of the elastic limit and the safety factor, this
flexure is expected to be linear for this application.
4.5.2 Two, stacked, notch-type flexures
In this design, the two-axis compliance is provided by two simple notch type
hinges aligned with their axes perpendicular and in the plane of the overall
flexure axis. A consequence of this stacked design is that the height from the
applied force to the centers of the two respective notches is different.
Consequently, if the two notches are of the same dimensions, the compliance at
the point of the applied forces will depend on the direction of the applied force.
Also there may be a problem with cross coupling between the two axes if the
notches are not coincident both with each other and the cylinder.
The main advantage of this type of design is that it could be manufactured
quickly by boring the two notches on a mill. For this design the same material
and a scan force of 5 N are used as for the previous example. As can be seen from
figure 4.43, the notch has a web thickness, t = 3 mm, a radius, R = 8 mm and a
depth b = 19 mm. From the former of these two, the dimensionless factor fJ =
0.1875 and the linear and angular stiffness can be obtained from equation's
(4.108) and (4.109)

F 2Ebt 51 2
-L= kF9 = r:--;;2 =6.41x10 4 [N rad-1]
B. 9trR.3' 2-vt- f32

M 2Ebt 512
__!.. =kMIJ = =504 [Nmrad·1]
B, 9trR 112

To determine the compliance's for forces applied about the compliant axis of
each of the notches, it is necessary to identify the upper (referred to as notch 1)
and lower notch (notch 2). The distance, 11•2 , from the center of the notches to the
applied force for each of the notches is

/1 = 0.051
[m]
/2 = 0.070

2 17
FLEXURES

Proceeding as the previous example, the maximum bending stresses in the

two notches are 1.8 MPa for notch 1 and 7.5 MPa for notch 2. Clearly, both are
well within the design limits for this material. The stiffness, k, at the point of
application of the force in the direction perpendicular to the axis of notch 1 is

For notch 2

[Nm-t]

For an applied force of 5 N, the maximum deflections in the two

perpendicular directions are 30 and 54 J.Ull for deflections of notches 1 and 2
respectively.

References
Bosch ].A. (Ed.), 1995, Coordinate Measuring Machines and Systems, Vol. 42 of
Manufacturing Engineering and Materials Processing, Marcel Dekker.
Eastman F.S., 1937, The design of flexure pivots,]. Aero. Sci., 5(1), 16-21
Frocht M.M., 1951, Strength of Materials, The Ronald Press, NY, 237-238
Frocht M.M., 1935, Factors of stress concentration photoelastically determined,
Trans. ASME, ]. Appl. Mech., 57, A67
- Frocht M. M., Guernsey R. and Landsberg D., 1952, A photoelastic
reexamination of notched tension bars, ibid .., 74, 124.
Goodier ].N., 1941, An extension of the photoelastic method of stress
measurement to plates in transverse bending, J. Appl. Mech .., 63, A27-A29
(1941)
Haringx ]. A., 1949, The cross strip pivot as a constructional element, Appl. Sci.
Res.; series A, Al, 313-332
Howells M. R., Duarte R. and McGill R., 1996, Properties of the cartwheel type
flexural hinge, personal communication
Jones R. V., 1951, Parallel and rectilinear spring movements, J. Sci. Instrum., 28,
38-41
Jones R.V., 1970, The pursuit of measurement, Proc. lEE, 117(6), 1185-1191
Ling C. B., 1952, On the stresses in a notched strip, Trans. ASME: Appl. Mech., 74,
141-146
Ling C. B., 1968, On stress concentration factor in a notched strip, Trans. ASME:
Appl. Mech., 90, 833-835

218
CHAPTER 4: FLEXURE ELEMENTS

Paros J.M. and Weisbord L., 1965, How to design flexure hinges, Machine Design,
151-156 (Nov.)
Pereira P.H., Muralidhar A., Hocken RJ., Miller J.A. and Smith S.T., A two-axis,
static and dynamic force characterization device, patent application in
progress, Apri11998
Pilkey W.O., 1997, Peterson's Stress Concentration Factors, John Wiley and Sons,
New York (1974), pages 104,122,128
Plainevaux J. E., 1956, Etude des deformations d ' une lame de suspension
elastique, Nuovo Cimento, 4, 922-928 (In French).
Smith S.T. and Chetwynd D.G., 1992, Foundations of ultraprecision mechanism
design, Gordon and Breach, London.
Smith S.T., Badami V. G., Dale J. S. and Xu Y., 1997, Elliptical flexure hinges, Rev.
Sci. lnstrum, 68(3), 1474-1483
Speake C.C., 1987, Fundamental limits to mass comparison by means of a beam
balance, Proc. Roy. Soc. Lond., A414, 333-358
-Keyser P.T. and Jefferts S.R, 1989, Magnetic susceptibility of some
materials used for apparatus construction (at 295K), Rev. Sci.
Instrum., 60(8), 2711-2714
-Flanders P.J., 1990, A vertical force alternating-gradient
magnetometer, Rev. Sci. Instrum., 61(2), 839-847
-Quinn T.J., 1992, The beam balance as an instrument for very precise
weighing, Meas. Sci. Techrwl., 3,141-159
Thorpe, II, A. G., 1953, Flexure pivots - Design formulas and charts, Product
Engineering, Feb., 192 -200
Wittrick W.H., 1948, The theory of symmetrical crossed flexure pivots, Aust. f.
Sci. Res., A1(2), 121-134
Wittrick W.H., 1948, The properties of crossed flexure pivots and the influence of
the point at which the strips cross, The Aeronautical Quarterly, 2, 272-292
Young W.C., 1989, Roark's Formulas for Stress and Strain, 6th ed., McGraw-Hill
Book Company, NY.
Young W. E., 1944, An investigation of the cross strip pivot, J. Appl. Mech., 11,
A113-A120

219
5 Flexure systems

5.0 Overview
Previous chapters have concentrated on individual flexure elements and techniques
for the analysis of stress, strain or, equivalently, forces and subsequent deflections.
In this chapter, flexure elements combined to produce complete mechanisms are
assessed. Static and dynamic performance of these systems is evaluated using
methods developed in chapters 2 and 3. A particularly important mechanism in
flexure design is the so-called Jour-bar link and, as such, this occupies the opening
discussions. Not only is this commonly used in its simplest form, many of the more
complex mechanisms can be constructed through appropriately combining successive
Jour bar linkages. Throughout this book, flexures are commonly modeled as a series of
infinitely rigid links connected by massless flexure elements. For applications
requiring precise rectilinear motion, it is necessary to consider the effects of the finite
stiffness of the links. In particular, the moments imposed by the flexure elements
often result in significant angular distortion of the links connecting them, sometimes
resulting in unacceptable tilt errors. To illustrate this, an analysis of platform
distortions in a Jour bar link is presented plus some discussion on the optimal
position of the location of the drive to reduce parasitic tilts at specific locations
around the mechanism. Following this, mobility analysis of general planar
mechanisms is discussed, in the process returning to the ideal link/joint models.
Using examples of some of the more common mechanisms, Lagrangian analysis for
the development of closed form analytic models is demonstrated. Analysis of static
and dynamic properties of these mechanisms provides a background to the following
section that discusses generalized approaches for dynamic analysis. D'Alembert's
principle is used to derive equations of motion for flexure mechanisms of arbitranJ
geometry and, for small motions, these are expressed in matrix form. Analysis of
planar mechanisms using matrix techniques is then introduced and illustrated with
a case study examination of the four bar link connected by flexure elements for which
compliances in all freedoms are included. Correspondence between the generalized
and simplified models of the preceding sections is demonstrated. Finally, the
transformation matrices necessanJ to extend this analysis to three-dimensional
dynamic models are provided.
FLEXURES

5.1 The four bar link

Already demonstrated in chapter 4, for small displacements, the four bar link
with equal length supports provides a simple mechanism for achieving an
approximate rectilinear motion. Probably the two most commonly encountered
forms of this are the notch and leaf-type mechanisms shown in figure 5.1. As
stated in section 4.1.5, if the upper platform of the parallel spring mechanism is
perfectly rigid, the optimal drive should be positioned mid-way betWeen the
moving and stationary platforms. When the moving platform has finite rigidity
additional considerations may be necessary, some of which are outlined in
section 5.2. Issues associated with the manufacture and assembly of these types
of flexure are discussed in chapter 8. To introduce the reader to some of the
analytic methods for assessment of linear spring performance, the following two
sections consider the leaf and notch type mechanisms respectively. In many
cases, similar analytic models can be developed for mechanisms comprised of
either flexure type. Because of this, the reader can assume that where analysis is
not included for one type, it is safe to follow the procedures for the other.
S.l.l The simple leaf type rectilinear spring
For the simple leaf type flexure mechanism shown in figure 5.1, the total stiffness
in the direction of motion, x, is given by

k
6, F,
= 24EI
£3
rz =2Eb(.!_)3
L (5.1)

where I refers to the second moment of area of a single leaf spring about the
neutral axis.
Equation (3.254) can be used to estimate the lowest mode, free natural
frequency corresponding to linear motion in the x-axis direction given by

k
-' _ =
24Efzz
(U n, = I - - - -6,''26
F, (rad s·l) (5.2)
M+ - m
35

where M is the mass of the moving platform and m corresponds to the mass
of each flexure.
There will, of course, be other frequencies that might be significant. In
general, most of these are likely to be higher, see section 5.5. Possible exceptions
correspond to the torsional mode natural frequency about the y-axis plus
cantilever bending about the x-axis. For the simple spring as shown, assuming
that the upper surface of the moving platform is normal to the gravity field (i.e.
horizontal), an observer on this surface would denote these yaw, BY, and roll, ()~ ,
respectively. Rotations about the z-axis correspond to pitch (B, , in the present

222
CHAPTER 5: FLEXURE SYSTEMS

a)

L z

lox SIDE VIEW

b)
A s o s
G

L
..
c F..
L
D L•

-
--

Figure S.l: Two common linear spring mechanisms, a) leaf spring based flexure, b) notch
hinge flexure

example). An estimate of the natural frequency for bending about the x-axis can
be obtained from the beam bending equation for a cantilever and Rayleigh's
method for the determination of the natural frequency of a cantilever with a
mass attached at the free end, equation (3.259)

223
FLEXURES

k 6 ,F, 6EI:rx
1--....:.....:.-- =
m
(Q = (5.3)
"•· 33
M + 2- - L (M +~m)
3
140 140

(Note that the total mass, 2m, and stiffness, 2k, for both flexures is included in
this calculation while the second moment of area is computed for a single leaf)
Clearly for small values of m relative to the platform mass, the ratio of the
two natural frequencies can be approximated from the equation

(5.4)

In most applications, b is much greater than t and, in such as case, the

assumption that equation (5.2) corresponds to the lowest mode natural frequency
is valid.
The stiffness in the z-axis direction in response to a force applied at the upper
platform can be computed from

(5.5)

Considering the linear elastic properties of flexures, and for small distortions
in comparison with the beam dimensions, it is assumed that the principle of
superposition applies. Therefore, the subsequent distortion as a consequence of a
combination of forces can be computed separately and, thereafter, added
vectorially. Another assumption is that of isotropic elastic properties. In most
(but certainly not all) applications, the spring is likely to be constructed from
readily machinable materials such as steel or aluminum which are invariably
considered isotropic.
The torsional stiffness about the z-axis, is found by applying equal tensile
and compressive axial forces to the flexures which are spaced D Apart, therefore

(5.6)

224
CHAPTER 5: FLEXURE SYSTEMS

Torsional stillness about the y-axis is more difficult to compute since several
effects combine. Rotation of the platform about this axis will cause equal and
opposite deflections in x and z and each spring w ill twist about its y-axis. All
these deflections contribute to the resisting torque. The torsional stillness of each
spring about its longitudinal axis is, Timoshenko and Goodier, 1970,

k. =k Gbt3 (5.7)
8,My I
3£

where G is the modulus of rigidity and k 1 is a shape factor that can be

obtained from equation (2.104), section 2.8.2.
For smaii rotations the deflections at the end of each beam are given by

D
=-sin() DBY
~ _ _
&
z 2 )I 2 (5.8)

&
X
=-D2 (1 - cos() )I
DB
)~_
4
Y_ (5.9)

The end forces to cause these deflections can be reinterpreted as couples on

the spacing of D and De respectively. For the x direction, equation (5.1) gives
exactly twice the linear stillness since we are only considering one beam at this
point. However, in the z-direction, the platform rotation constrains the springs
into the s-shaped configuration, and so, the related stiffness is twice that given by
equation (5.5). The torsional resistance as a result of these deflections in these two
axes is

(5.10)

Since By is small and t<b, the contributions from the second term in equation
(5.10) can, in most cases, be neglected. Therefore, the overall torsional stiffness
for the two-spring flexure may be taken as

~ bt[Ed
2 2 2
k = k· + 2 k.. D + 2k1Gt ]
(5.11)
O,M, O,M, OyM, L 2£2 3

Noting that G=E/2(1+v)~E/2.6 for typical values of Poisson's Ratio, v, that

b>t and that, provided lr-:13t or more, k1 will be of the order of 0.9, it can be seen
that the first term in equation (5.11) will dominate. Substituting the above values
and rounding to simple numbers gives

225
FLEXURES

(5.12)

Commonly, D and d are roughly equal and b is almost always greater than
2.5t, so it seems unlikely that there will be a contribution of as much as 10% extra
stiffness due to twisting of the flexures. Potentially more significant is the effect
of shear deflection in the z-direction. This will be proportionally larger than is the
case for simple cantilever bending. Taking the simplest model of shear using a
shape factor of 1.5 corresponding to a rectangular cross section, the deflection of
the beam in shear is

t5 = l.SFL (5.13)
• btG

where F is the shear force, which is equal to the deflecting force in this
application. This deflection is additional to that predicted for bending so the
effective stiffness is reduced. The s-shaped deflection stiffness in the y-direction
for each spring will become

(5.14)

If the ratio of length to breadth becomes significantly low, then this

expression should be used as the first term in parentheses in equation (5.12).
Even if this ratio falls as low as 4, the correction still only contributes about 20%
and will be compensated a little by the torsional stiffness of the springs.
Consequently, for basic design modeling, it is generally reasonable to use the
simpler form shown in equation (5.12).
Note that, using this approximation,

(5.15)

There is considerable scope, through choice of the geometric parameters, for

varying the relative values of linear and torsional stiffness to match a desired
stiffness ratio. A case study in which this torsional compliance is exploited to
create a two-axis actuator can be found in Smith et al., 1994.
5.1.2 The simple notch type linear spring
A dynamic analysis of the simple notch type mechanism is presented and this
provides an introduction to t:qe application of Lagrange's equation to flexure
mechanisms. The objective of ·the simple spring is to provide an approximate
straight-line motion mthe x-axis direction. For perfect hinges rotating about their

226
CHAPTER 5: FLEXURE SYSTEMS

mid points· and, assuming small displacements, the upper moving platform, A,
will approximate rectilinear motion. It is also apparent that the support legs, C &
D, will rotate so that the rectangular centers of the four hinges transform to a
parallelogram, see Figure 5.1(b). Considered to be a single degree of freedom
system, a single coordinate is all that is necessary to completely describe the
motion of all points in the flexure. If x represents the linear displacement of the
platform A, the total kinetic energy of the spring is the sum of linear motion of
this platform plus the sum of the kinetic energy due to rotation of the two
identical support legs, C and D, about the centers of the lower, stationary hinges.
As a first step in the determination of the Lagrangian, it is necessary to derive an
expression for the total kinetic energy of the flexure. In this case, being
convenient, two coordinates representing the displacement and rotation are
temporarily used

T =MA
- x·2 +
2
2(IciJ;)
--
2
(5.16)

However, these coordinates are not independent. For small deflections, the
angular rotation of the flexure can be approximated from

f)
' =sin(-=-)
L. R: .!__
L. (5.17)

Therefore (5.16) can be expressed in terms of the single coordinate x by

(5.17)

Changes in potential due to this displacement are considered to occur

primarily in the necked portion of the hinges. In reality, there will be some
distortion of the support platforms. Assuming a high compliance of the hinge in
comparison with the supports, in most designs, it is reasonable to neglect these
distortions. Because of the symmetry of this mechanism, all hinges will undergo
identical angular deflections. Ignoring shear compliance in the x-axis, the
component of strain energy due to pure bending is simply

(5.18}

227
FLEXURES

For this single degree of freedom system, Lagrange's equation can

immediately be written in the form

(5.19)

Substituting (5.17) and (5.18) into (5.19), the equation governing motion of
this flexure is given by

2/c )·· 4ko,M, -

( MA + L.2 x+~x- Fx
or (5.20)

The left-hand side of the latter of equations (5.20) is immediately

recognizable as a simple, homogeneous, second order differential equation of the
form

(5.21)

In its simplest form the natural frequency for the notch type linear spring is
given by

(5.22)

For the notch type hinge, the angular stiffness can be approximated from
equation (4.91) i.e.

2EbtSI2
Ko,M, ::: - -t&R--.,..,.1'2::- (5.23)
9

Determination of the second moment of mass, or moment of inertia, is not so

straightforward. However, if it is assumed that the support legs can be
adequately modeled as long, uniform, thin rods and the mass of material in the
notches is ignored, the second moment of mass about the centroid of the support
legs is

228
CHAPTER 5: FLEXURE SYSTEMS

I' = Mc L2 (5.24)
12

To determine the second moment of mass about the pivots of the lower
hinges, it is convenient to use the parallel axis theorem from which

I = Mc L2 + McL"l
12 4
2
Mc L Mc (L+2RY
= - - + ----'::....!...----'- (5.25)
12 4
= McL2 +Mc(LR+4R2)
3

Substituting the two above expressions into (5.23) yields

(5.26)

Equation (5.26) can be simplified further if it is assumed that R << L after

which it corresponds to equation (4.36) in the book of Smith and Chetwynd,
1992.
The total 'static' linear stiffness of the flexure can be obtained by setting the
acceleration to zero, and is

(5.27)

Although, theoretically, for a single degree of freedom system there can be

only one mode of vibration, in practice there will exist an infinite number of
frequencies and mode shapes. It is important to identify some of the frequencies
that may be of comparable value to that in the most compliant direction i.e. the x-
axis. For most designs, the second lowest natural frequency corresponds to the
flexure oscillating as a pendulum with the pivot being at the mid-point of the
lower notches and rotating about an x-axis at the center. This is probably more
easily visualized by looking at the end view of the flexure in figure S.l(b). For
this particular mode of vibration, the top platform will be moving about an arc in
the z-axis direction, the aforementioned pivot point being at the center of the
lower notches.
Yet again, to determine an approximate equation for the natural frequency it
is necessary to make a number of assumptions. Firstly, it will be assumed that the

229
FLEXURES

second moment of mass (moment of inertia) of the upper platform about the
upper notch hinges is small in comparison to its.second moment of mass from
the x-axis pivot point in the lower notches. Consequently, it will be expected that
the inertial forces on the upper notches about the x-axis and also linear forces in
the z-direction will induce negligible distortion in comparison to the lower
notches. Based on this assumption, it is possible to model the complete linear
spring as a rigid body attached to the base by the two notches.

s --+---.-

L z

Figure 5.2: Lumped model of the simple notch type linear spring for
computation of the fundamental mode frequency of oscillation about an axis
passing through the centroids of the two lower notches

When oscillating in the x-axis direction, subsequent motions in the y-z plane
do not effect the x position of any part of the flexure. Consequently, it may be
assumed that the natural frequency of vibration computed previously, which
involved predominantly x-axis motion of elements in the flexure, will have little
influence on the mode considered here and vice versal. Based on these
assumptions, the simple spring can be modeled as comprising two long thin rods
with a mass attached at the free end, see figure 5.2. The kinetic and potential
energy corresponding to this mode of vibration are given by

(5.28)

1 In reality, both modes of vibration induce motion of components of the spring in the y direction.
As a consequence, there will be some coupling of these two modes and this will have an effect on
the natural frequency. In all likelihood this effect will oft times be small.

230
CHAPrER 5: FLEXURE SYSTEMS

From equation (4.111), the bending stiffness of each notch is given by

(5.29)

Substitution of equation (5.28) into Lagrange's equation readily provides a

solution for the natural frequency

(5.30)

The numerator of equation (5.30) is the effective bending stiffness of the

flexure about the x-axis in response to a moment applied to the platform.
5.1.3 The virtual center
So far, it has been assumed that all four-bar links are comprised from a
rectangular distribution of hinges, therefore resulting in a parallelogram motion
between the upper platform and base link. A generalized mechanism modeled as

Instantaneous center ab

I~
I \
I \
I
Instantaneous I
motion of link C

L.
cd

Figure 5.3: The general four-bar link

231
FLEXURES

four rigid links connected by single freedom pivots is shown in Figure 5.3. Of
interest to this section is analysis of the motion of the moving platform relative to
the base and the dynamic characteristics of symmetric forms of this mechanism.
Before discussing the generalized four-bar mechanism, study of a single link
will reveal some important principles of rigid body motion. Consider a rigid link
free to move in space, upon this are marked two, fixed points, A & 8, figure 5.4.
Also indicated by arrows at points A and 8 on this link are the instantaneous
displacements. Clearly, if an observer were to be positioned at A but with a fixed
orientation relative to a stationary link then, although the observer would be
moving with A, the point 8 must always remain a constant distance away. If the
link has any rotation then point 8 can move relative to A. However, the
constraint of constant length imposes only tangential relative motion. Therefore
Theorem I; the relative motion between any two points in a rigid body is
always perpendicular to a straight line connecting them.
Velocity
Diagram Instantaneous
center
\Velocity of B
0
relative to A

O+tr/2
Velocity of
A relative a
too

Moving..../
link
Motion of point A

~h[~ ~~~~~'!~nk,
Figure 5.4: Geometry of generalized, planar, rigid body motion

In many ways, the above statement might be an obvious definition of a rigid

body. However, the above theorem can be extended to include a statement about
the nature of the motion relative to a point fixed in the fixed (often referred to as
the 'stationary') frame. If it so happened that the rigid body is rotating about the
fixed point, o, then it is obvious that the lengths between oA and o8 must be
constant and also that their motion will be perpendicular to the line connecting
these points. Consequently, a second theorem can be postulated
Theorem II; if a rigid body moves about a stationary point, o, in such a way
thnt any two .arbitrary points, A & B, on thnt body each describe an

232
CHAPTER 5: FLEXURE SYSTEMS

instantaneous motion perpendicular to the lines oA and oB, the point o lies on
the instantaneous axis of pure rotation of the rigid body.
In the case of planar mechanisms, the point o is the instantaneous pivot of the
rigid body. Referring back to the four-bar link of figure 5.3, it is obvious that the
pivots connecting each end of link A to links C & D will be moving
perpendicular to the axes of links C & D. Extending the axes of the links C & D
they eventually cross at the point marked ab. Clearly, at this point the two ends
of the link A will also be moving perpendicular to the line connecting them and
therefore by theorem II above, the link A is instantaneously rotating about ab.
The notation ab for the instantaneous center indicates the rotation of link A
relative to the stationary link B. Consequently, for small motions, it is possible to
use a four bar link to create a virtual center of rotation. This can be of great
practical use for designs in which a pure rotation about a virtual pivot is
required in the instance where the pivot point is to be vacant to enable the
placement of specimens etc. Also, by theorems I and II the instantaneous center
of motion for the link of figure 5.4 is shown.
The concept of the instantaneous center is not new and the theory of
mechanisms has been developed for over one hundred years (for introductory
texts see Keown, 1921, or Mabie and Reinholtz, 1987). In general, for a planar
system of u links connected by single freedom joints (either slides or pivots
although only the latter are discussed in this book) there will be n - 1
instantaneous centers relative to each link. However, at each location there are
two centers. Consequently, the total number of locations, N, of instantaneous
centers is given by

N = n(n-1) (5.31)
2

For the four bar link of figure 5.3, there are 6 instantaneous centers
(sometimes called centros in older texts on kinematics of mechanisms). Four of
these, be ca ad bd, occur at the pivots. As already shown the other two can be
found at the intersection of lines co-linear with the axes of the two connecting
links that provide the motion constraints. The construction for the instantaneous
centers ab & cd is shown in figure 5.3. The astute observer will realize that any
centro is located on a straight line with two others which complies with
Kennedy's more general theorem
Theorem III; for three independent bodies in general plane motion the three
virtual centers lie on a common straight line
Grubler's equation (2.108) for a planar mechanism immediately gives a
mobility of 1 for the four-bar link connected by single freedom hinges.
Consequently, in view of the fact that complete motion is defined by a single
coordinate, one may erroneously be lead to the belief that geometric analysis

233
FLEXURES

would be correspondingly simple. Unfortunately, given the lengths of the links

between hinges and one of the angles between two links, the remaining two
angles can only be determined from transcendental equations. To see this, the
vector equation that must be satisfied if the mechanism is to connect at the pivots
can be expressed in complex form by the requirement that

(5.32)

An equally valid form is given by the conjugate functions (these are merely a
multiplication of -1 in the imaginary direction)

(5.33)

Multiplying (5.32) by (5.33) eliminates the angular term in BA to give

(5.34)

Expanding the complex exponents by De Moivre's theorem produces

l~ =1; + l~ +I~+ 21
8 0 1 cosB0 - 218 /c sin Be-
(5.35)
2lcl 0 (cosBc cosB0 +sinBc sinB0 )

Consequently, if the lengths of the links and one of the angles are known, the
other can be determined from (5.35). Assuming angle Be is known and using the
standard t substitution

t =tan( 8;)
1- t 2
cosB0 = - - (5.36)
l+t 2
. B 2t
sm o = - - 2
1+ t

the angle B0 can be determined from the quadratic

(5.37)

where

234
CHAPTER 5: FLEXURE SYSTEMS

a= z; - I~ - /~ - I~ + 2/ /c cos Be+ 2/ 1 8 8 0 - 2lclo cos Be

b =4lcl0 sin Be (5.38)
c= z; - I~ - I~ -z; + 21 /c cosBc + 2lsf 8 0 + 2lclo cosBc

- , -- - -- - --

I
,,,.,.
I

. \
I I \

I
I~
I \ 1!- 20 =2a
I I \
I I \
R• 1 I \
I I \
h I A I \

Figure 5.5: The trapezoidal four-bar link

Fortunately, most flexures are symmetric with links A and B parallel so that
the joints form a trapezoid. In this case the added symmetry of the trapezoidal
four-bar link shown in figure 5.5 greatly simplifies analysis. In this case, the
height, h, of the centro from the base is simply

(5.39)

The radius of rotation, R., of the link A is given by

(5.40)

Because of the common motion at the moving pivots (in this case study, the
pivots are assumed to be at the centers of rotation of the flexure hinges), the
relationship between the rotation of R. and the support links is simply

235
FLEXURES

(5.41)

The second moment of mass about the instantaneous center, J, can be

obtained using the parallel axis theorem

(5.42)

I Ao is the second moment of mass of platform A about the point mid-way

between the two moving pivots.
Using (5.41) and (5.42), the total kinetic energy for a virtual velocity, oB, is

r =(Hi·)' }o' +lc

r
(5.42)

=(~(\~ '
1
+ lc }o'
Correspondingly, the potential energy is simply

v {~)<oo +Oa)' +{~Joo)' k[(1 +;. )' +l](oo)'

= =
(5.43)

=k[G:)' +l]<oo>'
In the absence of applied forces, Lagranges equation immediately gives

(5.44)

From this, the undamped natural frequency, wn, is

(5.45)

236
CHAPTER 5: FLEXURE SYSTEMS

Often, it might be reasonable to assume that the links can be modeled as thin
rods for which the second moment of mass about the center is simply m/2 I 3 . For
a mechanism with a rectangular distribution of hinges, equation (5.45) reduces to

(5.46)

Replacing s with L' , this is identical to equation (5.22)

5.1.4 Effect of the drive on the natural frequency
So far, it has been assumed that the drive applied to the flexure asserts no
influence on the natural frequency. In reality, drives often add significant
stiffness to the mechanism. Consider, for example a piezoelectric or micrometer
drive. This must be contacted with the mechanism via a coupling (maybe
including a lever mechanism, see chapter 7) of finite stiffness between the output
displacement and actuator, kd, see chapter 8. If electromagnetic drives are
employed, this additional stiffness can, in principle be reduced to zero, Smith
and Chetwynd, 1990. For closed loop systems the added compliance can range
from large negative to large positive values over limited ranges of frequency.
For a single degree of freedom system undergoing small displacements, the
natural frequency can always be expressed in the form

2 k.
((), = - (5.47)
m.

Assuming that the drive adds little mass, (5.47) is simply modified to include
the drive stiffness, kd, so that

(5.48)

5.2 Optimal geometry for the rectilinear motion of components on a simple

linear spring
In many applications, it is desired to produce small, rotation-free motions.
Examples include displacement of an optical wave front by translating a plane
mirror, small motion of a mechanical measuring scale etc. Attaching a mirror to
the platform of a flexure is an obvious solution for the former application. When
it is necessary to achieve precise rectilinear motions through use of a flexure
stage, the distortion of the moving platform and other 'rigid' links may be
significant. As an illustration, this section discusses the optimal design of a
simple linear spring mechanism and the effects of drive location on the

237
FLEXURES

rectilinearity of motion of the upper platform when induced elastic deformations

cannot be neglected.
5.2.1 Simple linear spring mechanisms
Lateral distortion of a prismatic beam due to applied bending moments can be
obtained using Macauley's method, see section 2.7. A mathematical
representation of the flexure analyzed in this section is shown in figure 5.6. From
this, the moment equation along the beam can be written as

(5.49)

L X

Figure 5.6: Bending moments applied to the moving platform of a simple linear spring

Integrating the bending equation, the displacement of the beam is given by

x2 x3 M 2
-EIPy=M12 - - -3 [x -a] +Cx+C
- +R16 (5.50)
2 I 2

The expression in square parentheses is the singularity function and defined

such that it only exists for positive values, see section 2.7.4. For a simply
supported beam, the boundary conditions correspond to zero deflection at each
end. Consequently, for x = 0 it is simple to show that C2 = 0. At the other end, x =
LP, the constant C1 is given by

)2
C= -M 3 ( L-a - MLP- - RL~- (5.51)
1 2L p I 2 I 6
p

In this particular beam configuration, the applied bending moments originate

from the distortion of the leaf spring type flexures plus the moment due to the
applied force from the drive. Assuming that displacements are small, the slope of
the platform at the connections to the applied loads is ignored. Under these
conditions, it is possible to calculate the magnitude of the applied bending

238
CHAPTER 5: FLEXURE SYSTEMS

moments from the reactions of the simple flexure spring, for which the relevant
parameters are shown in figure 5.7. The bending moments, M1 at each end of the
platform are given by the equation

(5.52)

tp
-.----

s
a

L, F
-'-

- t-tl

Figure 5.7: Parametric model of a simple linear spring mechanism

These moments are introduced from the horizontal force externally applied
to the platform. The moment arm of length s combined with the drive force itself
introduces a bending moment, M 3 at a distance a from one end. This moment
can be expressed in terms of the displacement of the platform

24Eir
M3 = -
L3 -o (5.53)
I

Clearly, these two applied moments given by equations (5.52) and (5.53) are
related by the simple equation
4s
M 3 = -M 1 = kM1 (5.54)
L,

239
FtBXURES

where k =4s I L1
The reaction at the support, R1 can be obtained by taking moments about the
opposite end of the beam and equating it to zero to yield

(5.55)

Equations (5.54) and (5.55) can be substituted into (5.50) and (5.51) to yield
2 3
x +M-(
-EI y= M1 - 1
k-2 - -1 [ x-a]2+C x
x ) -kM (5.56)
p 2 Lp 6 2 I

C
1
= M I [ -2Lk
p
(L - a
p
)2 ---(k
LP
2
-2)-LP
6
l
(5.57)
= M,[kLP+ ka2 -ka- Lpl
I 3 2Lp 6

In many instances, it may be desirable that the slope at either end of the
beam is zero. The slope at an arbitrary point x along the beam can be computed
from the equation
2
-EJpdx
- M 1 ( k-2)x
dy= M x+- --kM [
x-a ] +CI
I L 2 I
p

2 2
(5.58)
(k - 2)x kLP ka LP]
= M 1 x+----k[x-a]+ - + - - -
[ LP 2 3 2Lp 6

l
At the left-hand support this is given by

2
- EI -dy= M [kLP
- +ka LP
- - ka--
P dx I 3 2£ p 6
(5.59)

Defining the ratio's

240
CHAPTER 5: FLEXURE SYSTEMS

(5.60)

The condition for zero slope at the left support is

1
a = - - - ----=- (5.61)
8 - 24P+I2P2

For example, choosing the value a= LP/2, j3 = 0.5, from which we find a= -1,
corresponding to s = -L1 or, in words, the actuating force is at a position above
the flexures at a distance equal to their length. A p lot of the normal distortion
along the platform for such a position of the actuator is shown in figure 5.8.

1.5

,_._
!
0.5

s:l
0
·.z:: 0
0
Q) 0.2
Q
Q)
Q ·0.5

-1

-1.5

x /L

Figure 5.8: Distortion of the upper face of a simpler linear spring platfonn, s
= -L1 ,E = 70GPa, LP = 40mm, L1 = 8 mm, tp = 3 mm, If = 0.5 mm,
d = 10 mm, a = 20 mm, F =34 N.

The condition for zero slope at the other end of the beam can be readily
shown to be

(5.62)

As a further example, choosing a value of a= 3LP/4, j3 = 0.75, from which

equation (5.62) gives a value for a of 4/11 (contrast with -4/13 for zero slope at

241
FLEXURES

the other end). Figure 5.9 shows the subsequent deflection given by equation
(5.56) using these values for zero slope. Because, in this case, the point of
application of the drive is inside the flexure and produces a smaller bending
moment than the previous case, such a -design is likely to be more desirable.

0.4 0.6 0.8

-1

ia -2
0
·a
~ -3
csQ
-4

-5

-6

x /L

Figure 5.9: Deflection of the upper platform for the values s = 2.91 mm, a. =
4/11 and 13 = 3/4

5.2.2 Effect of axial strains in the flexure supports

In the previous analysis, the equal and opposite reaction forces, R, & Rli along the
axis of the flexures due to the moment produced by the drive have been ignored.
This will result in a net tilt of the complete platform given by the difference
between the axial extensions of each flexure divided by the length, LP, of the
platform separating them. Again, assuming small deflections are, the angular tilt
(anti-clockwise positive) is

(5.63)

This reduces to zero when the drive is positioned midway between the
platform and base, a conclusion in agreement with the more complete analytical
results of Plainevaux, 1956. It should also be noted that, in this equation, the
slope is independent of the cross sectional area of the platform while its length

242
CHAPrER 5: FLEXURE SYSTEMS

reduces value in inverse proportion. For the parameters used to generate figure
5.9, the slope predicted by equation (5.63) for a displacement of 100 ,.un is -17
j.Lrads. Current design goals for precision flexures are in the region of 1 1-lrad of
angular error and this is often considered a significant effect.
5.2.3 Combined effects
For small deflections, the principle of superposition can be applied to refine the
analysis of slope deflection of the platform support. It should be noted that this
ignores secondary effects, such as, the influence of the platform slope on the
values of bending moments from equation (5.52) which relies on a zero slope
boundary condition in its derivation. Summing equation (5.58) and (5.63), for
small deflections, the total slope of the platform is given by

dy= [ -
- t; (k - 2) -61- x +(k --2)- -x
1 (
2
-k(x - a]+-kLP +-
2
ka - ka -LP)]
- -o
dx LP IP LP 2 3 2LP 6 L}
2
(k ---
=[ - 2) -6t1 ( x +
(k-2)x
- -- .
-k(x - a]+-+ ka 2- ka -LP)]t;
kLP - - -- o
LP t! Lp 2 3 2Lp 6 L}
(5.64)

s
"""' -2

§
-3
•

•l:J -4
~
~ -5
~ -6
-7 • • -Parametric data
-8
• • • • • • FEM results

-9 ~--------------------------------~

x/L
Figure 5.10: Theoretical prediction for the full equations and a comparison
with results from finite element analysis. Parameters are given in the caption of
figure 5.8 except for the modified geometric values discussed in the text
From this rather lengthy expression, determination of the geometric
conditions for zero slope at either end of the platform support are not as
straightforward. However, a number of important features are still readily
apparent. The slope is independent of the material from which the monolithic it

243
FLEXURES

is comprised, is proportional to the flexure displacement, and inversely

proportional to the square of the flexure length. All of these parameters do not
influence the conditions for zero slope.
Figure 5.10 shows the theoretical prediction of the platform distortion with
results from finite element analysis for comparison. Although there is a relatively
large discrepancy between displacements, the theoretical model appears to be a
reasonable predictor of the end slopes of the platform.

5.3 Planar mechanisms

A general approach to the design of planar mechanisms is presented here.
For a planar mechanism, it has already been shown, section 2.9, that the
mobility is given by

M 3 = 3(n-1)- :tc, =3(n- 1) - :t(3- .t;)

±it
1• 1 i:l
(5.65)
=3(n-j-1)+
1•1

where n is the number of links, j is the number of joints and f is the number of
freedoms of the ith joint and c represents the corresponding constraints (3 - /).
For a mechanism made up entirely of single degree of freedom joints such as
the notch hinge, this reduces to Grubler' s equation

M = 3(n - 1) - 2j (5.66)

Clearly, based on this analysis, the four bar link (j = 4, n = 4) represents a

single degree of freedom mechanism. It is important to maintain some caution
when applying this formula to real systems. The planar assumption reduces the
problem to motion in a two dimensional space. For a more complete analysis, the
mobility should be computed from the three dimensional formula.
j
M6 =6(n- j - 1) +Lit (5.67)
1•1

In three dimensions, analysis of the four bar link yields M = -2 indicating

overconstraint. As a consequence, it is to be expected that the performance of
such a mechanism will rely upon precision of manufacture. For most
applications, over-constraint will be accommodated by the off-axis compliance of
the flexures and such errors are often insignificant. This surprising result
illuminates some of the pitfalls that, in some cases, might introduce significant
errors in a design. Consequently, the following section discusses the reasons for
the differences between planar and three dimensional mobility analysis.

244
CliAPTBR !>: FLEXURE SYSTEMS

To create a mechanism, it is necessary to connect a number of rigid links

using hinges. In the design process, there are a number of parameters that can be
selected. While mobility analysis is a direct measure of the feasibility of a
mechanism in terms of the freedoms it may possess, it is also necessary to
consider other parameters that can be varied to produce an optimal design.
Before discussing planar mobility analysis in more detail, some other important
factors that effect the function of a mechanism are
1. The depth of the mechanism. Increasing the depth results in an increased
stiffness both in and out the plane of motion. Consequently, this will increase
the load capacity of the flexure and reduce the influence of masses being
added to moving platforms. However, the linear increase in stiffness with
depth is matched by an equally proportionate increase in mass therefore
nulling any increase in natural frequency. An increased stiffness also requires
more work from the drive. As a rule, it is assumed that deeper flexures are
better. Some manufacturing issues are addressed in chapter 8.
2. The loads to be supported and the length of translation will dominate the size
of the mechanism. Consequently, planar flexures can range in size from
micro-mechanical systems to large structural supports weighing many tons,
see for example Thorpe, 1953.
3. Manufacture. Using computer-controlled machine tools, wire electro-
discharge machines and micro-electronic manufacturing facilities, planar
shapes of almost arbitrary complexity can now be produced over an
enormous range of dimensions. ·
4. As shown in the previous section, forces will distribute throughout the
mechanism resulting in deformation of all component parts. An alternative
view of a flexure mechanism is that of a series of connected flexible elements
of varying compliance. Such a continuum approach to flexure design can be
modeled using the method of finite elements. However, except where stated,
subsequent discussions will assume that links are infinitely rigid and
freedoms of connecting hinges provide combinations of pure rotation and
translation.
5.3.1 Discussion of the simplifying assumptions for reduction of mobility analysis to
planar mechanisms
Reduction of mobility analysis to planar mechanisms should be approached with
caution. Consider again the simple four-bar link. Implicit to this two-dimensional
mechanism is the assumption that the compliant axes of all joints are parallel.
Assuming such manufacturing perfection is possible, it is no longer necessary to
consider the effects of out-of-plane constraints. Alternatively, it may be assumed
that the perfection of manufactu.r e infers additional freedoms of the joints in
view of the fact that some theoretical constraints will not be 'tested' in practice.
With this caveat, consider first the conceptualized planar mechanism shown in
figure 5.11. This indicates the essential features of a model for assessing

245
FLEXURES

mechanism mobility in which individual joints are capable of providing one to

three freedoms.

Freedom/,"'

n,

Figure 5.11: A conceptualized. general planar mechanism consisting of 5 links and

6 joints
Considering the case of a mechanism in which all connections utilize single
freedom joints. Available joints providing a single freedom are flexure hinges,
slideways (including screws) or a more complex mechanism that, of itself, forms
a single freedom joint when used to connect two links. Using equation (5.66) it
can be shown that the minimum number of links to form a mechanism from
single degree of freedom joints is 2 with 1 joint (it can also be seen from (5.67)
that this is a single freedom mechanism in three dimensions). An example might
be a rigid pendulum hanging from a hinge. The next available lowest number of
links is 4 with 4 joints which is, of course, the four bar link. In general, it is
relatively simple to show that the relationship between the number of links and
joints for mobility 1 is given by

j =3n/2 - 2 n =2,4 :. oo (5.68)

Figure 5.12 shows a possible mechanism that co.mprises 6 links and 7 joints.
Increasing the number of links so that n = 8 and j = 10 results in the somewhat
modified form of the Peaucellier mechanism shown in figure 5.13. Detailed
kinematic analysis reveals that, by selection of suitable length of link and
distribution of hinge, such a mechanism can be produce perfect rectilinear

246
CHAPrER 5: FLEXURE SYSTEMS

Link
1
/ joints

Figure 5.12: A mechanism of mobility M = 1 comprising 6 links and 7 joints

motion of the point P in the figure, see for example Bevans, 1948. Another
mechanism of interest is that comprising 12 links and 16 joints which is
commonly referred to as the double compound rectilinear spring, see figure 5.14.
This has the particular advantage of a high degree of symmetry about its center
and also along the axis of motion.

M = 3(8-10-1)+10=1

Figure 5.13: A modified Peaucellier linkage, mobility M= 1

247
FLEXURES

Figure 5.14: Schematic representation of the double compound rectilinear

spring, M = 3(12-16-1)+ 16 = 1.

Table 5.3.1.1: Number links and single freedom joints to construct a planar
mechanism of mobili9' 1
Links Joints
2 1
4 4
6 7
8 10
10 13

For mobility 2, a minimum of 3 links and 2 joints are required. An example of

this is the chaotic double pendulum. Another combination might be 5 links and 5
joints or the more interesting 7 links and 7 joints. An example of the latter is the
compound linear spring shown in figure 5.15.
Using the full equation (5.65) for planar mechanisms, the above analysis can
be extended to include 2 and 3 freedom hinges such as those outlined in chapter
4. Considering only planar mechanisms, a two axis joint must comprise either
two orthogonal translations (not necessarily rectilinear) or a rotation plus one
translation. Obviously a three-axis hinge contains all three and is considered free
in the plane (and therefore non-existent!). The utility of the three-freedom hinge
is in its ability to withstand loads normal to the plane of the mechanism. To
reduce the mobility of a mechanism, the objective of most flexures, it is apparent
that the fewer links the better. For single mobility, a possible combination is 2

248
CHAPI'ER 5: FLEXURE SYSTEMS

Figure 5.15: A compowtd linear spring of mobility 2 providing two orthogonal

linear translations i.e. an x-y stage.

Figure 5.16: A planar mechanism of mobility 1 comprising 2 links and 2 two-

freedom joints (M= 3(2-2-1)+4)

links and 2 two-freedom joints, figure 5.16. A similar single degree of freedom
mechanism can be produced using a single degree of freedom hinge connecting 2
joints. However, there is an essential difference between these two. That is, using
a single degree of freedom rotational hinge, the only remaining freedom is a
rotation whereas two 2-freedom joints provide the possibility of either a rotation
or translation. Of course adding together combinations of single freedom
rotational hinges and links can produce linear mechanisms but this is at the
expense of added materials, complexity, and production cost. In an attempt to
produce a more symmetric mechanism, figure 5.17 shows a mechanism with 2
links and four joints. Using 2 freedom joints this is slightly over constrained
while three freedom joints result in a mobility of 3 and therefore the links can
adopt arbitrary relative positions within the plane. Accepting that this requires
precise manufacture, a possible pseudo single degree of freedom mechanism is
shown in figure 5.18. This utilizes the two-beam, two-axis hinges discussed in
section 4.4.2.

249
FLEXURES

M = 3(2 - 4 - 1) + Lf
2
= -1 (/ joints)
= 3 (/3 joints)

Figure 5.17: A mechanism having 2links and 4 joints

Some three axis hinges, or joints, are shown in figure 5.19. Literature on three
axis flexures is scarce and detailed analysis awaits further research.

Figure 5.18: A translation mechanism using two-beam, two-axis hinges

As evidenced by the few mechanisms shown in previous figures, there are an

endless array of mechanisms that can be produced using combinations of links
and joints. The interested reader will find some stimulating designs presented in
papers listed in the bibliography of interesting flexures at the end of this chapter.

250
CHAPTER 5: FLEXURE SYSTEMS

2JY_____.(\..__'1
b)

Link 2
c)

d)

g)

Link2

Link2

Figure 5.19: Some three degree of freedom, planar joints, a) series notch type, b) series half
notches, c) mobility model with three links and three single freedom joints i.e. M = 3(3-1 )+3 = 3,
d) symmetric array of two-beam, two-axis hinges (note that the legs connecting to the links are not
necessary for this joint to provide the requisite three freedoms), e) isometric diagram of three half
notch joint produced by slotting, f) hinge formed from a cylindrical band, g) combined single
beam, two axis hinge plus linear flexure

5.4 Dynamics of Ideal planar flexures (some common mechanisms)

In the following sections some general methods for assessing the performance of
planar and three-dimensional flexures will be introduced. Such generality
necessarily results in more complex and abstract mathematical formulisms. As a
consequence it is likely that the underlying principles may be lost. Fortunately, in
many applications, flexures contain symmetries that considerably simplify
analysis. This section presents a dynamic model for some of the more common
geometries used for planar mechanisms. The simple four-bar link for
parallelogram motion of the upper platform assessed in section 5.1 will not be
further assessed in this section. However, in practice, flexure hinges will have
finite stiffness values in the so-called constraint directions. Therefore, a complete
planar analysis of even the simple four-bar link should consider these
compliances. Necessarily, each joint must then be modeled as a provider of three
degrees of freedom leading to a mechanism of high mobility. In this case

251
FLEXURES

generalized dynamic models become necessary and these are discussed in

section 5.5 where the apparently simple four-bar link is reassessed.
5.4.1 The compound rectilinear spring
One of the major
concerns with the simple
linear spring is the
arcuate motion of the

u
upper platform relative
to the base. One
configuration that may
be used to overcome this
A limitation is achieved by
attaching two such
hinges together to
Base produce the geometry
shown in figure 5.20.
Figure 5.20: Mathematical model for the compound rectilinear Assuming all hinges are
spring of equal compliance and
all support legs are of
the same length with parallel axes, it is relatively easy to see that, if platform A is
the only element of the mechanism subject to a static, externally applied force it
will undergo a purely rectilinear motion. Platform B can also be shown to
displace exactly half as far as A when both displacements are measured relative
to the base. However, when considering free vibrations of such a structure, the
rectilinear motions of platforms A and B must be considered to be independent
and are given the coordinates q 1 and q2 respectively (i.e. this is a mechanism of
mobility= 3(7-1)+2(8) = 2). To derive expressions for kinetic and potential energy
in the flexure, although not necessary, the following assumptions have been
made for compactness of the solution
1. All motions are small
2. The support legs (C and D) are prismatic with the central axis passing
through the centers of the hinges at each end.
3. As a consequence of 2, the center of mass for each support leg lies on
the axis of the leg at a height midway between the two notch hinges.
4. Second moments of mass have been measured about center of mass
5. For the time being, it will be assumed that the rotational stiffness ( Nm
rad-1) for the notches are equal on support legs C and D, although the
stiffness of the notches on C need not necessarily be equal to those on
support legs D.

252
CHAPTER 5: FLExURE SYSTEMS

Based on the above considerations, it is relatively straightforward to derive

expressions for the kinetic, T, and potential, V, energy of the flexure

(5.69)

(5.70)

Considering, for example, the coordinate corresponding to displacement of

the platform A, the various elements of Lagrange's equation are

Proceeding in like manner for the second coordinate, from Lagrange's

equatiqn, the equations governing motion of this flexure are given by

(5.71)

This rather cumbersome equation can be expressed in the simpler form

a,,ij, +a,lij2 +c.. q, +c,lql = F;

(5.72)
a2•ii• + anii1 + c2,q, + cnq2 =F2

253
FLEXURES

where the a's and c's are constant and it is noted that a 12 = a 21 and C12 = c21 •
Equation (5.72) can be further simplified by writing it in the matrix form

or (5.73)
[M]{<i} +[KHq}= {F}
For the homogeneous (or steady state) solution, equation (5.73) can be
rearranged and expressed in the form of an eigen equation

[M]{<i}+ [K){q} = {o}

(5.74)
[1]{<i}+ [Ml '[K Hq} ={o}
Following the techniques outlined in chapter 3, equations (5.73) and (5.74)
can be used to derive the two natural frequencies (eigenvalues) and mode shapes
(eigenvectors) and the four steady-state frequency response functions. Extension
of this analysis for systems containing damping involves little additional
difficulty although the aforementioned parameters must be expressed in
complex form. To demonstrate this, a solution is readily obtained assuming
displacements of each coordinate of the form

{q} =Re{{<l>}e" 1
•) (5.75)

To produce the eigen-equation of the system, it is necessary to substitute

equation (5.75) into (5.74) after which

QA)- tv 2 (11){<1>} =
aA)- A(l ]){<1>} ={0} (5.76)

Solutions to equation (5. 76) for the two eigenvalues ( A1 2 =tv122 ) and the four
eigen vectors {<l>) can be readily obtained using matrix solution algorithms.
Alternatively, equation (5.76) can be written in expanded form

{c11 -a11 tv 2 }:1>1 +{c12 -a12m2 }1> 2 =0

2 2
(5.77)
{c2, - 021liJ }:1>, + {c22 - OnliJ Jt>2 = 0

From which the ratio of the two eigenvectors is readily obtained

(5.78)

254
CHAPrER·S: FLEXURE SYSTEMS

However, it is necessary to know the values of the natural frequencies before

this can be computed. These can be readily solved by substituting (5.78) into the
second of (5.77) to yield

(5.79)

The last of equation (5.79) is a quadratic in a> 2(=A.) for which the two
solutions can be readily computed.
The steady-state response for an arbitrary, harmonically varying input force
is similarly obtained. In this case it is only necessary to assume that the response
is linearly related to the input. For example assuming an applied force at
coordinate 1, of the form

(5.80)

the response can be readily written as

q 1 =Re(H 11 (ia>)F;e"'" )
(5.81)
q2 =Re(H 21(ia>)F;e' .. )

From (5.73), it is relatively simple to obtain an equation of the form

(c11 - a11 a> 2 }H11 (ia>) + (c11 - a12a> 2 )H21 (ia>) =l

(5.82)
(ell - a21lV 1 )HII (ia>) + (c22 - a22a>2 )H2• (ia>) = 0
This can be expressed in abbreviated matrix form

(5.83)

Note also that e12 = e21 • Two of the solutions for the values of the H(a>)'s at
any frequency can be readily obtained using Cramer's rule for solving these
simple simultaneous equations. Similar methods can be used to determine the
remaining two frequency responses for a force applied at the second coordinate.
Analytic expressions for the eigenvalues can be obtained by setting the
determinant of the left-hand matrix in (5.83) to zero. Chapter 3 contains a more
complete discussion of these methods.

255
FLEXURES

5.4.2 The double compound linear spring (including the lever driven spring)
An advantage of the previous design is that, provided the support legs are all of
equal length and the notches of equal stiffness, then it is possible to achieve
perfect rectilinear motion. This is due to the fact that the arcuate motion of the
first simple spring is matched by an equal arcuate motion of the other in a way
that the mutual approaches of each moving platform to its respective base both
cancel with respect to the stationary frame. However, a penalty for this is the
introduction of a second independent coordinate. Figure 5.21 shows a more
symmetric design consisting of two compound springs attached at the moving
platform. Such a mechanism is called a double compound linear spring. As a
consequence of the two compound springs being joined together, it can be seen
that a deviation from rectilinear motion is no longer possible in this design. This
is confirmed by a mobility analysis for which

M = 3(n - 1) - 2j =3(12 -1)- 2(16) =1 (5.84)

Figure 5.21: The double compound, notch type rectilinear spring

256
CHAPTER 5: FLEXURE SYSTEMS

Consequently, one coordinate is sufficient to determine the kinetic and

potential energy of all elements of the mechanism which, in the absence of a
driver are given by

(5.85)

The derivation of the kinetic energy is easily seen by considering each of the
terms in the first of equation's (5.85). Consecutively, these terms correspond to;
linear motion of the primary platform, approximate linear motion of the two
secondary platforms (assuming all lengths of the support legs are equal the
secondary platforms will experience half of the velocity of the primary); linear
motion of the centriod of the four central support legs, the common angular
velocity of all the support legs and the linear motion of the center of mass of the
four outer support legs. Substitution of equation (5.85) into Lagrange' s equation
gives the equation governing motion

(M A
M8+SM.
+-
2
--+ -21•
2
2 )··
ql +-2
kq I -F.
L
- I
4
L
(5.86)

The natural frequency of this flexure immediately follows

2 4k
m =-~--------------~ (5.87)
n L2(M +-M 8 +-
SM.
- +-21.)
A 2 2 2 L

Modeling the support legs as thin rods, the second moment of mass about
the center of mass (i.e. at the center of the leg) is

I = McL2 (5.88)
c 12

After which equation (5.87) becomes

257
FLEXURES

(5.89)

This is the same as the result given in Smith and Chetwynd, 1991.

L
Figure 5.22: A simple lever drive applied to the moving platform of the
double compound rectilinear spring

The addition of a lever drive merely adds to both the potential and kinetic
energy terms of the Lagrangian. Considering the simple lever of figure 5.22,
there will be three additional compliance's contributing to the potential energy
plus the additional inertia of the lever, driver and 'wobble pin'. The lever pivot
and primary platform drive flexures have angular stiffness values denoted by
k 6 & kd respectively. Assuming that the lever is driven by a piezoelectric element
which is rigidly coupled to the lever (unlikely in practice, see chapter 7) this
particular drive can be considered as a linear spring having stiffness k P • From
Rayleigh's method (section 3.11.2.3), it would seem reasonable to lump one third
of the mass of the piezoelectric element, mP' to the lever at the point of
connection as shown in the figure. A wobble-pin is necessary to provide freedom
for the y-axis component of the motion of the lever. Assuming that the notch
flexures behave as perfect notches, the contribution of the platform notch to the
potential energy may reasonably be neglected for small motions. Based on these
assumptions, the additional kinetic, T0 , and potential, V0 , energy terms are

258
CHAPTER 5: FLEXURE SYSTEMS

(5.90)

where n is the lever ratio b/a.

Consequently, the fundamental mode natural frequency of the lever driven,
double compound linear spring becomes

4k kb+kd lkp
- + +- -
~2= ----~L~2--~b~2____~2~n~2_____
(5.91)
n M 8M I m
M +- 8 + - - < +~+-P +m
A 2 3
2 b2 3n ,.

Clearly, whether the fundamental frequency increases or reduces will

depend on the relative values of the lever ratio, stiffness and mass components of
the lever mechanism.
5.4.3 A coupled two-axis flexure
Consider the flexure mechanism of figure 5.23 overleaf that possesses 5 links and
5 joints. Grubler' s equation immediately gives

M =3(5 - l) - 2(5) =2 (5.92)

Having established that this is a two-degree of freedom system, the next step
is to decide on a suitable set of generalized coordinates. Recognizing that the two
independent freedoms of the central mass are a translation in the vertical
direction and a rotation about its mid-point which, in this case, coincides with its
center of mass, suggests a suitable origin for the generalized coordinates. As
shown in figure 5.23(b), these coordinates are designated x. for the translation
and B. for the rotation. Input drives are located at the two lever points and, for
small motions, the input translations are given the temporary coordinates
x, & x 2 • To derive the Lagrangian of this system in terms of the generalized
coordinates, it is necessary to determine the functional relationships between
these coordinates. For small motions the transformations are given by the
equations

259
FLEXURES

a)

b)

x,
Figure 5.23: A two-degree of freedom notch hinge based flexure mechanism, a) solid
model, b) parametric model

260
CHAPTER 5: FLEXURE SYSTEMS

(5.93)

The inverse transformation is

2XD -/(}D
XI = --....;;...__--=.,
2n1
(5.94)
l(}D + 2X0
x2 =
2n 2

Defining the displacement ratios and angular deflections (two further

temporary coordinates) of the levers arms

b.
n.=-
a.
n=!2_
2
a2
(5.95)
(}I=--
x.
al
B2 = x2
a2

the kinetic energy of this flexure is given by

Assuming the lever and platform have similar geometry (i.e. a 1 = a 2 = a,

n1 = n2 = n =II a, I 1 = I 2 =I, b1 = b2 =/),equation (5.96) reduces to

261
FLEXURES

(5.97)

Similarly, it can be shown that the potential energy is

(5.98)

Again, assuming similar geometry of the two levers, (5. 98) can be rearranged
to give

(5.99)

Substituting (5.94) into (5.99) gives the potential energy in terms of the
generalized coordinates

(5.100)

Equation's (5.97) and (5.100) can be written in the generalized forms

(5.101)

As usual the coefficients show reciprocity i.e. axo = afJK & c xo =cfJK.
Substituting equations (5.101) into Lagrange's equation, the equations governing
free motion of this system are given by

auxo + ax8(jo + cuxo + cx8() =0

(5.102)
axoxo + a898o + c89oo + c zoxo = o

262
CHAPTER 5: FLEXURE SYSTEMS

This can be more conveniently expressed in the familiar matrix form

(5.103)

In this simplified form, this can be rearranged in the form of an eigen

equation from which two natural frequencies can be determined. To determine
the frequency response due to forces applied at the drives, it is necessary to
transform these to equivalent forces in the generalized coordinates.

5.5 General model for dynamics of planar flexures

While many flexures can be adequately modeled as the ideal mechanisms
considered above, in some cases the effects of flexure compliance in off-axis
directions may be of some concern. In such a case, the flexure must be modeled
as an element that provides an additional freedom in each compliant axis. Since
most mechanisms consist of a number of rigid bodies connected by flexure
'hinges', the addition of degrees of freedom to each hinge results in a
considerable increase in the overall mobility. As a consequence, even the
simplest flexure mechanism must necessarily be considered a multi-degree-of-
freedom system.
Either as a consequence of planar manufacture or inherent geometric
symmetries, it is often reasonable to consider the important characteristics of a
flexure system to take place in a plane. As a consequence, it is possible to derive
much useful information from an analysis of the dynamics of such a planar
system. If the correct plane is considered (as is most likely), results of this
analysis will include the lowest mode frequencies therefore representing limiting
dynamic performance. While still relatively simple, other designs may require a
considerably more complex three-dimensional analysis. It so happens that the
techniques necessary for planar analysis are merely simplifications of the more
complete three-dimensional case. Because the steps involved in the former can be
relatively easily followed and visualized, the analysis of this section can be
considered as an introduction to the more abstract, generalized dynamics of
compliantly coupled, rigid bodies presented in the following section.
In view of the large·number of degrees of freedom for the general analysis of
planar mechanisms, it is important to introduce a rigorous definition of the
various coordinates of the flexure. To illustrate the general principles involved in
the determination of the static and dynamic equations governing motion of a
planar mechanism, consider the four-bar link modeled in figure 5.24. Each of the
four rigid links is connected by compliant elements. To ease analysis, a number
of assumptions have been made. For more advanced applications, the book of
Goldstein, 1980, provides a more complete continuum analysis of rigid bodies as
well as some of the transformations relevant to this and the following section.

263
FLEXURES

Figure 5.24: Lumped model of a four bar link connected by flexure springs, a) geometric
parameters and coordinates for a generalized four bar link (for clarity, auxiliary coordinates 21,
43 and 41 are omitted), b) mathematical model for an arbitrary link.j.

For each rigid body of the planar mechanism, three coordinates (two
translations plus one rotation) are necessary to completely specify its position. To
determine the kinetic energy in the moving body, it is also necessary to know

264
CHAPI'BR '5: FLEXURE SYSTEMS

both the geometry (more correctly, the mass distribution) of the rigid body and
the position of the origin of the local coordinate system. Before proceeding, it is
necessary that the meaning of the local coordinate system be dearly understood.
By the local coor9inate system it is implied that a fixed origin is attached to a link
in its initial state. At this time, the position of each element within the link is
known. The origin of this coordinate system is placed at an appropriate point (to
be discussed) and thereafter remains stationary with respect to the global
coordinate system. Motion of any element on the link is then measured as the
difference between the start point and its instantaneous position as monitored by
the local coordinate system. Being separate entities, each link can be identified by
a unique single valued number. In the example shown in figure 5.24, links are
numbered consecutively from 1 to 4, the first link representing the rigid body to
which all other links are referred.
Compliant elements connecting the links of the mechanism can also be
ascribed a unique number for identification purposes. However, in this case it is
necessary to identify which two links are being connected by the flexure element.
One further complication arises. This is due to the fact that hinge compliances
may be more easily determined in a direction that is not coincident with its
adjacent local coordinates. Before considering how to deal with this, it is
probably timely to define the coordinate systems and notation used to describe
the position of the elements within the flexure mechanism.
5.5.1 Coordinate systems
In this and all subsequent analysis it will be necessary to discriminate three
different coordinate systems. These are
1. The global coordinate system
2. The principal local coordinates of a link
3. Auxiliary local coordinates
The global coordinate system refers to the inertial frame to which the
'stationary' link is affixed. Invariably this will correspond to the earth or some
vehicle, travelling at constant velocity, to which the flex'-lre is attached.
Throughout this text, the stationary link, containing the global coordinates, is
ascribed the number 1. All motions of flexure elements will be referred to this
coordinate set.
For the purpose of deriving the total kinetic and potential energy, it is
convenient to define a local coordinate system for each of the rigid links within
the flexure. As already mentioned, choice of the origin of this coordinate system
is arbitrary. However, once made, it is important to use inertial properties
measured about this point. For planar motion, the inertia corresponding to
rotation about the z-axis perpendicular to the plane of operation of the
mechanism is simply computed from

265
FLEXURES

(5.104)

where the radius vector r is measured from the z-axis of the local coordinate
system in this case.
For a homogeneous material of constant mass density, equation (5.104) can
be expressed in the alternative form

(5.105)

In much of the following, the inertia properties are known and it is not
'necessary to compute the integral of equation (5.105). For the three dimensional
analysis there will be nine components of inertia relative to the local coordinate
system. Denoting the general coordinates by the symbol q, the elements of the
inertia matrix can be computed from the equation

I "= fp(r)~ 2 51J - q,qJtv (5.106)

v

where 5" represents the Kronecker delta of matrix and vector analysis.
Motion of the rigid link in terms of the local coordinates may be visualized
by initially marking an imaginary dot on the rigid body at the origin. Motion of
the body will produce a displacement of the dot relative to the local coordinate
system the origin of which remains stationary with respect to the global.
Translation and rotation of a point at the origin will subsequently define the
position of all points in the rigid body, thereby effecting a complete vector
description of the flexure system. Coordinate transformations represent the
linear motion of a this dot in each coordinate system. By definition, rotation
measured at any point on the rigid link will be the same.
To compute the total forces in the flexure system, it is necessary to determine
the position and relative motion of the individual flexure springs at the two
points of attachment. For this purpose, an auxiliary coordinate set is defined at
these points. In practice, the position of this point is fully defined by the principal
local coordinates and it is only necessary to determine the relevant
transformations to express all forces in terms of the principal coordinates.
5.5.2 Notation
Clearly, in view of the large number of variables involved in this analysis, it is
necessary to carefully choose a unique notation that enables a full geometric
description of a general planar mechanism.
For each link of the mechanism it is necessary to define its position in terms
of one rotation and two translation coordinates. For link j of a flexure, the
unique principal local coordinates are given by

266
CHAPTER'S: FLEXURE SYSTEMS

(5.107)

To conveniently separate the components of rotational and translational

kinetic energy, the origin of this coordinate system has been chosen to be at the
center of mass of the link. In practice this axis will often coincide with the
principal axes of the inertia ellipsoid. In such a case, the mass matrix (to be
defined) is diagonal which greatly simplifies analysis.
To compute natural frequencies, it is necessary to describe the geometry of
the flexure mechanism in its initial state. For this purpose, two parameters are
required, one providing the angle of the link relative to the points of connection
of each flexure element, the other the orientation of the flexure axis. The initial
angle (anti-clockwise positive) between the x-axis of the link, j, and a straight line
between the origins of the link coordinate system and the auxilliary coordinate
system of the flexure connecting to an adjacent link k is, see figure 5.27,

(5.108)

For each flexure, there will be three stiffness values. In practice, these are
likely to be known in the direction of the flexure axis, which is unlikely to be
coincident with that of the local coordinates. The three-valued stiffness vector for
the flexure connecting links j to kin the direction of its axis is denoted

(5.109)

To compute the potential energy stored in the spring it is more convenient to

express the stiffness in terms of the components in the direction of the local
coordinate system. In this case, the transformed stiffness is denoted

(5.110)

Using this notation, the angle between the flexure element connecting links j
and k and the x-axis of the local coordinate system is denoted

(5.111}

Finally, the auxiliary local coordinates of link j at the point of attachment to a

flexure connecting link k are marked primes. Hence, an auxiliary coordinate
system is given by

267
FLEXURES

(5.112)

In principle, equations (5.104) to (5.112) contain the complete set of geometric

parameters necessary to describe the geometry of the flexure system and
subsequently determine the total potential and kinetic energy for substitution
into Lagrange's equation.
Before it is possible to derive the equations of motion, it is necessary to
determine the forces in the system as a function of arbitrary displacements in all
degrees of freedom. Because each rigid body, except for the stationary link,
possesses three degrees of freedom the potential energy must reduce to a matrix
equation containing independent coordinates numbering

3(n - 1) (5.113)

where n is the number of links in the mechanism. To reduce the number of

variables to this value, the following transformations are necessary.
5.5.3 Transformations
I. Stiffness transformations
Consider the simple leaf type flexure of figure 5.25 for which the stiffness
equations relating isolated forces and subsequent displacements are

(5.114)

As a word of caution, for

combined loads, more precise
stiffness equations should be
computed from the
trigonometric functions
presented in chapter 3.
However, for relatively low
Figure 5.25: A simple leaf type flexure
loads it may tentatively be
assumed that the principle of superposition applies and the flexure can be
modeled in terms of ideal springs as shown in figure 5.26. Also shown in this
figure are the two coordinates indicating the flexure (denoted x', y' and fJ) and

268
CHAPTER 5: FLEXURE SYSTEMS

local coordinate axes (denoted x, y and B ). Considering the measurement of

point P in this figure, it can be readily shown that the relation between

__
displacements and rotations in the two coordinate sets are given by
y'
lix' ,
..-- p
X
o/'

x'

Figure 5.26: Idealized model of the simple leaf-type flexure

&' =&cos(¢)-o/sin{¢)
o/' = o/cos{¢)+ &sin(¢) (5.115)
liB' = liB

In matrix form

{!:H~~~ :~~~ ~l{!} (5.116)

{ox'} = [A 1 f {ox}
where

269
FLEXURES

cos(¢) sin(¢) 0]
[AJ= -sin(¢) cos(¢) 0 (5.117)
[ 0 0 1

To determine the generalized component of force in the direction of the local

coordinates, it is necessary to derive an expression for the force in the
generalized coordinate directions as a function of the corresponding
displacements. Because ;•¢="' ¢ it is relatively simple to show that the
generalized forces are given by

sin(1• ¢) 0 ][ cos(l• ¢)
sin(;•¢)
cos(1•¢) 0 - sin(;•¢) cos(;•¢)
0 k 81k 0 0

This rather cumbersome equation can be written more succinctly in the

matrix form

Jk{F} = -(A 1t k(k);k [A 1r·,k{{ox~k }-{ox~}}

(5.119)
=- [K.tk{{ox;k } - {ox~}}
All that remains is to express the forces as they apply to the center of mass of
the link (i.e. the forces applied to the origin of the principal local coordinates).
With reference to figure 5.27, it can be seen that the forces applied to the center of
mass of the link j due to forces applied by the flexure connecting to link k are
given by

(5.120)

Combining equations (5.119) and (5.120), the externally applied forces on

link j due to the flexure jk can be expressed in terms of the principal local
coordinates by

270
CHAPTERS: FLEXURE SYSTEMS

Figure 5.27: Planar forces applied to a rigid link

(5.121)

P'

Ox)
Figure 5.28: lllustration of the geometric relationship between the auxiliary and principal local
coordinates
Equation (5.120) still contains the auxiliary local coordinates. Consequently,
with reference to figure 5.28 it is necessary to translate these to the principle
coordinates using the transformations

271
FLEXURES

(5.122)

Consequently, (5.121) can be expressed in terms of the principal local

coordinates for the forces at link j due to a flexure connected to an adjacent link k
given by

{FJk }=- (Al)Jk fK .l k(A3tk {c:5'x};

(5.123)
+ [A3L [K. Lk [A3]Tk} {Ox.}k

Finally, for each link, the total applied force will be the sum of forces from all
flexures connecting adjacent links. If link j connects to l 1 adjacent links, k, the
total applied force is given by

(5.124)

Clearly, a link cannot connect to itself so that j :t: k. The inertial reaction to
the sum of all applied forces is

(5.125)

From D' Alembert's principle, and assuming the incremental displacements

are both small and occur about the initial state, the requirement for dynamic
equilibrium of the n links gives

(5.126)

Again, some of the terms can be collected to form a more compact matrix
equation of the form

272
CHAPI'BR 5: FLEXURE SYSTEMS

(5.127),

Note that the difference in the two A matrices is to be found in the order of
the subscripts and both are distinctly different matrices, the first involving terms
that contain geometric variables for the link j only while the second includes the
geometry of the adjacent links.
Equation (5.127) indicates the elements of a larger matrix equation of the
form

(5.128)

Again, this can be written in condensed form by the familiar second order,
linear, differential equation

(5.129)

In this equation, the inertia matrix is simple diagonal while the stiffness
matrix is symmetric.
Damping elements at the individual hinges can be included using the same
transformations as those for the flexures, after which the general equation of
motion is of the form

(5.130)

Having derived this linear form, the eigenvalues and subsequent frequency
response can be obtained using the methods outlined in section 3.9. In particular,
equations (5.129) and (5.130) can readily be reduced to an eigen equation of the
form

[A]{z} =.t{Z} (5.131)

where

(5.132)

273
FLEXURES

In view of the complexity of the above algebra, this method is probably best
illustrated by selecting the simplest of examples
5.5.4 Case study 1: The simple linear spring flexure
As an example of the implementation of the above matrix procedure, the simple
parallel spring utilizing leaf type hinges is revisited. Figure 5.29 shows the
geometry of the flexure with relevant dimensions included in the caption.
Comprising a simple four-bar link, even this simple mechanism possesses 9
degrees of freedom. Fortunately, because all of the flexure hinges are of identical
geometry (and orientation, a factor that will be made use of shortly}, the three
orthogonal stillness values for each of the hinges are identical. Consequently the
mechanical characteristics of this flexure can be given by the values in table
5.5.4.1
· 1. pr~rties of t h. e st·mple flexure
Table 5.5.4 1: Mecharuca
Parameter Value
m2 =m3 (kg) 0.012

ml (kg) 0.032

/2 =I. (kgm2) 9.0xlQ-7

/3 (kgm2) 6.667x1Q-6

k, (N m-1) 2.1xlOS

ky (Nm-1) 2.lx1()6

k9 (Nmrad-1) 4.375
Again, the symmetry of this mechanism considerably reduces the complexity
of the transformation matrices. In particular, because all hinges are of the same
orientation, the angle of the axis of the flexures relative to the local coordinates is
the same for each hinge and of value -rt/2. Consequently, the rotation matrices
are all equal i.e.

0 -1 0]
[
[A J = I
0
0
0
0
1

This leads to a common stiffness matrix in which the linear components of

the flexure stiffness are interchanged i.e.

274
CHAPI'ER 5: F LEXURE SYSTEMS

2 4

Figure 5.29: A simple linear spring (s = 0.05 m, I" = 0.035 m, I= 0.04 m, L = 0.005 m, depth=
0.01 m, h = 0.008 m, width = 0.005 m, t = 0.5 mm, E = 210 GPa, p =8000 kg m3). Note that,

l
from equation (4.6), r has been chosen as a point at a distance of L/3 from the base to a
corresponding point measured from the upper linear motion platform

ky 0 0
(K.)= 0 k, 0
[
0 0 k(J

Derivation of the linear transformation matrices is not as convenient. The

necessary geometric parameters for this transformation are the length of the line,
I connecting the origins of the auxiliary and principal local coordinates and the
angle, ~k, between the x-axis and this line connecting the origins of the two
coordinates (measured positive anti-clockwise). These are given in table 5.5.4.2
below
... .
Table 5 542 Geometrxc parameters £or th e ..mear flexure
Parameter Value Parameter Value (m)
(rad)
821 =841 -n/2 [ 21 = / 23 =141 = 143 0.0225
823 = 843 n/2 132 = 134 0.025
832 1t

834 0

Systematically inserting the angular values of table 5.5.4.2 into equation

(5.124) yields the components of force acting on the three links

275
 FLEXURES

These forces can be combined to give the 9 x 9 stiffness matrix corresponding

to equation 5.129

276
CHAPTER 5: FLEXURE SYSTEMS

Of less complexity is the mass matrix given by

m2 0 0 0 0 0 0 0 0
0 m2 0 0 0 0 0 0 0
0 0 12 0 0 0 0 0 0
0 0 0 ml 0 0 0 0 0
(M)= 0 0 0 0 m3 0 0 0 0
0 0 0 0 0 /3 0 0 0
0 0 0 0 0 0 m. 0 0
0 0 0 0 0 0 0 m. 0
0 0 0 0 0 0 0 0 /4

Inserting the values for this case study into the above matrices, an eigenvalue
and vector analysis produces 9 eigenvalues (the squares of the natural
frequencies) with 9 eigenvectors associated with each eigenvalue. The results
from this analysis are given in tables 5.5.4.3 and 5.5.4.4 below
Table 5.5.4.3: Natural frequencies based on square root of the eigenvalues for the
simple linear sprin\g m
· uruts
· o f:<;J__c1es~r second or Hertz
Mode Value (Hz)
li)l 74.88
li)2 2978
li)) 3237
m. 7752
aJs 7861
(i)6 11666
li)1 29775
m, 31582
(i)9 32908

From these tables, it is possible to visualize the respective modes of this

mechanism. In the above table, every third row corresponds to rotation of the
links while the first of every three rows represents horizontal displacement. Most
striking is the value of the lowest mode natural frequency, 75Hz, in comparison
to all other modes, which are in excess of 3 kHz. As would be expected this mode
corresponds to parallelogram motion of the flexure between the upper platform
and base. Assuming that the hinges are perfectly rigid in all but the rotational
degree of freedom, the fundamenta l mode natural frequency can be
approximated from the equation

277
FLEXURES

(i)" =( 2~.) ,_ __ 4/k_.IJ'---1-.-2 = 76.47 (Hz)

1u ml + 2 ,,.,1 I

Table 5.5.4.4: Eigenvectors cor responding to t he eigenvalue modes of t he p revious

t able
Model Mode2 Mode3 Mode4 ModeS Mode6 Mode7 ModeS Mode9
-0.0162 -0.7071 0.0722 0 -0.0001 0 0 0 0
0 0 0 0 0 0.0068 -0.7071 0.0990 0.0541
0.7066 0 0.7032 -0.7071 0.7071 -0.0016 0 0.0001 -0.0001
-0.0319 0 -0.0262 0 0.0009 0 0 0 0
0 0 0 0 0 0.0365 0 0 0.0009
0 0 0 0 0 0.9986 0 0.9902 0.9968
-0.0162 0.7071 0.0722 0 -0.0001 0 0 0 0
0 0 0 0 0 0.0363 0.7071 -0.0990 -0.0584
0.7066 0 0.7032 0.7071 0.7071 -0.0016 0 -0.0001 -0.0001
...
Artificially mcreasmg the linear stiffness terms of the hmges by a factor of
100 increases the fundamental mode frequency to a value 75.02 Hz indicating a
relatively close agreement between the simple and complete planar analysis.
The second mode can be visualized as equal and opposite horizontal (x-axis)
displacements of the two support legs. For this mode the natural frequency can
be approximated from the equation

To the expected precision of the above analyses, the predicted frequencies

from the two models are the same. The reader is left to ponder the remaining 7
modes.
5.5.5 Commen ts on general planar analysis
In the preceding analysis, it was found that the fundamental natural frequency
was of considerably lower value than the higher modes. Clearly, the dynamic
performance of the mechanism is predominantly governed by this lowest, or
gravest, mode. As a consequence, it is reasonable in many cases to ignore all but
the lowest mode and consider the hinges as 'ideal'. This reduces the degrees of
freedom of each joint by two (which now becomes a ' hinge'), in many cases
enabling a simple analytic solution for the lowest modes using Lagrange's
equation or Rayleigh's method.

278
CHAPTER 5: FLEXURE SYSTEMS

It would be erroneous to assume that all modes of interest can be assessed

from this simple planar analysis. As already shown in section 5.1.1, in many
designs, the flexure mechanism may possess relatively low natural frequencies in
directions perpendicular to the plane of interest. Fortunately, because of the
orthogonal geometry of most mechanisms (they are commonly manufactured
using machines with orthogonal slideways and spindles) it is not necessary to
resort to full three-dimensional models. In these cases, it is possible to view the
mechanism from different, orthogonal planes and assess the natural frequencies
using any of the preceding analyses. In practice, if there are natural modes at
significant frequencies (either near to the required bandwidth or to vibration
frequencies transmitted to the structure) these will invariably become excited
and are likely to couple to any directions of interest. Ignoring higher modes may
result in problems if there are excitation forces that coincide with one of the
frequencies. In this case, it is possible (and more likely as the modes become
increasingly large) that the high frequency excitation will couple vibrations to the
lower frequency modes. To see this, consider the ratio of two numbers computed
to a precision of, for example, 1%. As the ratio of the difference between the two
numbers becomes greater than one hundred the computed value will always
appear as an integer. Even for mechanisms of modest complexity, if the mode is
out of the plane of interest, there is almost always some unwanted coupling.
Considering an isolated rod, which must constitute the simplest flexure, separate
excitation and the interaction of perpendicular longitudinal and lateral modes
has confounded experimentalists for centuries. Both of these important points
were clear to Rayleigh who wrote
..in actual experiments with bars which are neither uniform in material nor
accurately cylindrical in figure it is often found impossible to excite
longitudinal or torsional vibrations tvithout the accompaniment of some
measure of lateral motion. In bars of ordinary dimensions the gravest lateral
motion is far graver than the gravest longitudinal or torsional motion, and
consequently it will generally happen that the principal tone of either of the
latter kinds agrees nwre or less perfectly in pitch with some overtone of the
former kind. Under such circumstances the regular modes of vibrations become
unstable, and a small irregularity may produce a great effect. The difficultt; of
exciting purely longitudinal vibrations in a bar is similar to that of getting a
string to vibrate in one plane.
In all of the previous analysis, the links were assumed to be perfectly rigid
bodies. Real links will have a finite compliance and for more precise estimates it
is recommended that this be appropriately apportioned to the flexure elements
connecting links. If, for example, a link of compliance c1 is connected by two
flexures of compliance's c2 &c3 , it is recommended that the compliance's of the
c; c;
flexures be modified to & according to

279
FLEXURES

(5.133)

In terms of stiffness values, the above becomes

k' - k2k,(k2 +k3)

2
- k,(k2 +kJ)+l
(5.134)
k' - k3kl (k2 + k3)
3
- k,(k2 +k3)+1

It can be readily shown that, setting the link compliance to zero (or stiffness
to infinity) corresponds to the previous rigid body model.

5.6 General dynamics of flexures

Construction of a generalized three dimensional model represents an extension
of the planar analysis of section 5.5. Having derived the appropriate
transformations, the procedure for solution of the matrices is the same. Such
transformations, however, are not straightforward and great care should be
exercised in their development. Inclusion of such an analysis would take us too
far afield in a text of this length and must await further exposition, possibly in a
second volume. The necessary transformations are discussed in considerable
depth in the texts of Arya, 1990, Haug, 1989, Goldstein, 1980, Pars, 1965 and
many other texts concerning kinematics of rigid bodies. However, to help the
interested reader, the necessary transformations are mentioned here for
completeness. The first transformation necessary is that of relating elements of
the original stiffness matrix in terms of the orientation of the local coordinate
system. Assuming that superposition applies for both linear and angular
components of stiffness, this is most easily performed through successive
rotations about three axes using Euler angles. The first rotation is about the z-axis
of magnitude ~ and the appropriate coordinate transformation corresponding to
that of equation (5.116) is

cos(¢) sin(¢) 0 0 0 0
-sin(¢) cos(¢) 0 0 0 0
0 0 1 0 0 0
[D]= 0
(5.135)
0 0 cos(¢) sin(¢) 0
0 0 0 -sin(¢) cos(¢) 0
0 0 0 0 0

280
CHAPrER 5: FLEXURE SYSTEMS

After this rotation, the new coordinate system comprises the original set with
a rotation in the xy plane. By definition of the x-c.onvention, the second rotation B
is then taken about the x-axis of this new coordinate set and is effected by the
matrix transformation

1 0 0 0 0 0
0 cos(B) sin(B) 0 0 0
0 - sin(B) cos(B) 0 0 0
(C)= (5.136)
0 0 0 0 0
0 0 0 0 cos(B) sin( B)
0 0 0 0 -sin(B) cos(B)

Finally the transformation from one coordinate system to the other is

completed by a third rotation, VI about the z-axis of this second intermediate
coordinate set produced from the matrix

COS(Ijl) sin(VI) 0 0 0 0
- sin(IJI) COS(IJI) 0 0 0 0
0 0 1 0 0 0
(B) = (5.137)
0 0 0 COS(IJI) sin(IJI) 0
0 0 0 - sin(IJI) COS(IJI) 0
0 0 0 0 0

The complete coordinate transformation is a linear combination of these

three rotations. Comparing with equations (5.117) and (5.119) the appropriate
transformation matrix becomes

[AJ= [nJcJo] (5.138)

Assuming that the principle of superposition applies in this case, the stiffness
matrix is chosen to be simple diagonal w ith the linear stiffness components in the
first three cells of the diagonal and the rotational components in the three
remaining.
Determination of the linear transformation matrices requires some additional
definition of terms. In three dimensions, the relevant length of the link can be
reduced to three components projected onto the three planes of the loca l
coordinate system. For example, the length of the projection of the link of true
length l onto the xy plane is denoted I, and can be readily computed from the
equation

(5.139a)

281
FLEXURES

Projected lengths on the other two planes are readily computed from
2 112
lx = /(1 - cos (a))
(5.139b)
I, =l{l - cos 2 (r)) 112

The angles a, fJ and rare the direction cosines of the line from the local to the
auxiliary local coordinate system given by

X
cosa =-
/
cos/]=!!... (5.140)
I
z
cosy =-
1

Values for x, y & z are derived from the vector connecting the principal and
auxiliary local coordinates with the principal local coordinate system as the
origin (i.e. I = xi+ yj + zk ). To determine the forces in a Cartesian frame, it is
necessary to derive a transformation matrix in terms of linear translations in the
direction of the principal local coordinates of the link plus rotations about each of
the three axes. Consequently, the displacement vector of an individual link must
be of the form

{ox}= {& ~ ~ o(}x ooy oo,t (5.141)

In terms of this coordinate set, the linear transformation matrix is of the form

0 0 0 - ly sin(o>J - I, sin(B, 0 )
0 l 0 - lx sin(Bxo ) 0 I, cos(B, 0 )
0 0 I lx cos(Bxo) IY cos(Byo ) 0
[AJr = (5.141)
0 0 0 I 0 0
0 0 0 0 1 0
0 0 0 0 0

Eliminating columns and rows 3, 4 & 5 reduces to the transpose of the

transformation matrix of equation (5.120). The initial angles of the projections of
the line connecting the local coordinates are given by

282
CHAPTER 5: FLExURE SYSTEMS

z
tanBxo =-
y
X
taneyo =- (5.142)
z
X
tan8, 0 =-
y

Solutions for the three dimensional mechanism proceed similarly to the

planar case discussed in the preceding section.

5. 7 Sources of interesting flexure mechanisms

A full discussion of the novel flexure designs that have been presented in the
literature over the last few decades would contain sufficient material for the
production of many volumes of literature. However, these include a large
number of novel implementations of lever mechanisms and methods for
combining links and joints to produce displacements in desired directions. Most
of these can be appreciated at a glance and therefore the reader will be amply
rewarded after perusing these references.
Alemanni M., Mana G., Pedrotti G., Strona P.P. and Zosi G., 1986, On the
construction of a zerodur translation device for x-ray interferometric
scanning, Metrologia, 22, 55-63. One of few examples utilizing elliptic hinges
in this case manufactured from a single monolithic block of the low thermal
expansion glass ceramic Zerodur™. X-ray interferometry represents one of
the more demanding applications for flexure mechanisms requiring
rectilinear motion accurate picometer levels with parasitic rotations below
tens of nano-radians. Translation distances of tens of millimeters are under
investigation.
Bamford R.M. and Glaser R.J., 1997, Induced-strain actuator with mechanical
amplification, NASA Tech. Briefs, Jan., 60. An entertaining page.
Becker P., Seyfried P. and Siegert H., 1982, The lattice parameter of highly pure
single silicon crystals, Z. Phys. B. -Condensed Matter, 48, 17-21. An interesting
variant on the double compound linear spring with a high mobility to enable
electromagnetic correction of parasitic errors.
Deslattes R.D., 1969, Optical and x-ray interferometry of a silicon lattice spacing,
Appl. Phys. Letts., 15, 386-388. A levered, notch type rectilinear sprihg of
extraordinary rectilinear characteristics.
Furukawa E., Muranaka Y., Takenouchi Y., Moriya T. and Andoh T., 1992,
Development of a torsional ground motion pickup, Int. ]. JSPE, 26(2), 152-
157. Some novel lever mechanisms and an alternative analytic approach to
that adopted in chapter 7 of this book.
Furukawa E., Mizuno M. and Terada K., 1991, A magnifying mechanism for use
on piezo-driven mechanisms, Bull. Japan Soc. Prec. Engg, 25(4), 315- 320

283
FLEXURES

Furukawa E., Mizuno M and Hojo T., 1994, A twin-type piezo-driven translation
mechanism, Bull. Japan Soc. Prec. Engg, 28(1), 70-75
Hart M., 1968, An Angstrom ruler, Brit. f. Appl. Phys. U. Phys. D), 1,1405-1408. The
first paper presenting the concept of x-ray interferometry for metrological
applications. The flexure used in this study was produced from a single
crystal of silicon by making two pairs of orthogonal cuts.
HowelJs M.R., 1995, Design strategies for monolithic adjustable radius mirrors,
Optical Engineering, 34(2), 410-417
Jones R.V., 1987, Instruments and experiences, Wiley & Sons, London. Containing a
compilation of papers by this author, this is an excellent introduction to
flexure applications with extensive discussion of practical implementation.
Also contains many references plus an attempt to provide a brief historic
perspective on the developments of this field
Kyusojin A. and Sagawa D., 1988, Development of linear and rotary movement
mechanism by using flexible strips, Bull. Japan Soc. Of Prec. Engg., 22(4), 309-
314. Some examples of flexures combined to produce hinges with defined
instantaneous centers.
Nashimura K, 1991, A spring guided micropositioner with linearized
subnanometer resolution, Rev. Sci. Instrum., 62(8), 2004-2007.
Ryu J.W., Gweon D-G. and Moon KS., 1997, Optimal design of a flexure hinge
based XYB wafer stage, Precision Engineering, 21(1), 18-28. An interesting
analysis of lost motion in levered flexure systems (this is discussed in chapter
7)
Sydenham P.H., 1981, Mechanical design of instruments. 5: Putting elasticity to
use (Part A), Measurement and Control, 14, 179-185, (part B), Ibid, 219-227, -
Elastic design of fine mechanism in instruments, f. Phys. E: Sci. lnstrum., 17,
922-930. These references contain many examples of flexure elements plus an
extensive list of references.
Teague E.C., Young RD., Scire F. and Gilsinn D., 1988, Para-flex stage for
microtopographic mapping, Rev. Sci. Instrum., 59(1), 67-73. Example of
application of a variant of the hinge shown in figure 5.13(b).
Xu W. and King T., Flexure hinges for piezo-actuator displacement amplifiers:
Flexibility, accuracy and stress considerations, Precision Engineering, 19(1), 4-
10.

References
Arya A.P., 1990, Introduction to Clnssical Mechanics, Allyn and Bacon, Boston, Ma.
A careful reader should be able to correct a small transcribing error in
equation (13.115) and a typographical mistake in (13.117) of this edition.
Bevan T., 1948, The theory of machines, Longmans, Greem and Co., Lond.
Goldstein H., 1980, Classical Mechanics, Addison-Wesley, Reading, Ma., chapters
4 & 5. Note that Goldstein uses the x-convention in the sequence of rotations

284
CHAPTER 5: FLEXURE SYSTEMS

of Euler angles. Because the rotations are not commutative, one must use
these transformations with great care.
Haug E.J., 1989, Computer aided Kinematics and dynamics of Mechnnical Systems:
Volume 1: Basic Methods, Allyn and Bacon, Boston, Ma.
Keown R.A., 1921, Mechnnism, McGraw-Hill Book Co., NY.
Mabie H.H. and Reinholtz C.F., 1987, Mechnnisms and Dynamics of Machinery,
John Wiley &Sons, NY.
Pars L.A., 1965, A Treatise on Analytical Dynamics, Heinmann, London, chpt. VII.
Note that Pars uses they-convention in the sequence of rotations of Euler
angles.
Plainevaux J.E., 1956, Etude de deformations d'une lame de suspension elastique,
Nuovo Cimento,. 4(5), 922-928 (in French).
Plainevaux J.E., 1956, Mouvement de tangage d'une suspension elementaire sur
lames elastiques, Nuovo Cimento, 4(5), 1133-1141 (in French).
Rayleigh J.W.S.,1894, The Theon; of Sound, vol. I, Dover Publications, section 149,
chapter VII .
Smith S.T. and Chetwynd D.G, 1990, Optimisation of a magnet/ coil force
actuator and its application to linear spring mechanisms, Proc. Inst. Mech.
Engrs., 204(C4), 243-253.
Smith S.T., Chetwynd D.G. and Harb S., 1994, A simple two-axis ultraprecision
actuator, Rev. Sd. Instrum., 65(4), 910-917.
Thorpe IT, A.G., 1953, Flexure pivots -Design formula and charts, Product
Engineering, Feb., 192-200.
Timoshenko S.P. and Goodier J.N., 1970, Theory of Elastidty, 3•d Ed., McGraw-
Hill, NY.

285
6 Hinges of rotational symmetry

6.0 Overview
This chapter discusses the design offlexure elements that are constructed from solids
of revolution to provide compliance in defined axes for the purposes of precise motion
control. Because of the ease of manufacture, such mechanisms are often formed from
cylindrical tubes and circular disks. Due to their omnipresence in many instrument
designs, a brief discussion of the bellows and coil spring as flexure elements are
included. Not surprisingly, notch and leaf type Mnges can also be integrated into
rotan; couplings and two such designs are presented towards the end of this chapter.

6.1 Introduction
Design of circularly symmetric compliant elements is not new. Established
markets already exist that provide various springs and elastic couplings that may
be readily used for applications where transmission of linear translations or
rotations are desired. This chapter draws from the wealth of information already
developed for such mechanisms. In many cases it is desired that a flexible
coupling be provided between two rigid bodies having rotational symmetry. In
fact, for most motor driven shaft assemblies, a flexible coupling is necessary to
accommodate the small misalignments inevitable in any assembly. In addition to
misalignments, the ends of the shafts may also require some relative motion.
Probably the most familiar couplings are the universal (or Hooke) joints and
constant velocity couplings used to transmit rotary motion from engines to the
driven wheels on automobiles. Although less familiar, there is a not-less-
common requirement for couplings to connect between many types of machines
such as electric motors to gearboxes and, subsequently, gearboxes to machinery.
Reflecting the enormous range of applications in which electric and other
rotary motors are used, a bewildering array of mechanisms are available to
provide the requisite freedoms and power transmission requirements, see for
example Shigley 1989, Neale et al., 1991. Fortunately, the emphasis of this chapter
is towards applications that best utilize the advantages of flexure based designs
i.e. couplings requiring smooth, friction and wear-free motion, minimal
hysteresis and well defined distortions. The latter of these requirements
immediately rules out the enormous variety of couplings linked by intermediate
compliant elements such as; rubber inserts (spider, Barwell couplings), thin
metallic strips (chain coupling) or loose fitting rigid connections (gear and chain
FLEXURES

couplings). Of interest within the context of flexure design, this chapter will
discuss the following coupling geometry's
I. Coil springs.
II. Disk couplings.
m. Intermediate disk and coil spring designs: The rotary hinge.
IV. Flexible membranes.
V. Notch and leaf-type hinges applied to rotational couplings.
VI. The bellows.

6.2 Coil springs

Because of its high compliance in all degrees of freedom, the coil spring, when
used to connect two rigid bodies, is not ideally suited as a flexure for precise
motion control. It is included in this book because it is simple to analyse and
identify the reasons why this is used as an energy storage device, a more
complete analysis can be found in the book of Wahl, 1963. In its simplest form, a
coil spring comprises a single wire formed into a helix. Resistance to an applied
load is as a consequence of this applied force being transmitted along the wire
from one end to the other. Because the load passes through the center of the
spring, this must be supported by a torsional reaction of equal magnitude
throughout the wire. There will, of course, be a vertical shear force in the wire
providing an additional distortion. However, this may be ignored for most
practical designs. For a coil spring comprising a thin circular wire of diameter d
the torque, T, due to an axial applied load, F, at any position along the wire is
given by

T = FR = GJB (6.1)
L

where R is the radius of the coil.

The expressions for the total angular twist, B, and length of spring, L, for
inclusion in the torsion equation on the right hand side of (6.1) are given by
L~ 2nRn
0
e~ - (6.2)
R
1U4
J =-
2

where n is the number of turns of the helix, R is the nominal radius of the spring
and ois the total deflection between the two ends of the spring
Substituting equations (6.2) into (6.1) and rearranging gives

288
CHAPTER.6: HINGES OF ROTATIONAL SYMMETRY

(6.3)

More generally, it is the stiffness, k, of the spring that is of interest. From (6.3)
the stiffness can immediately be written in terms of radii or diameters

(6.4)

where d and D correspond to the wire and helix diameters respectively.

Based on this simple model, the maximum shear stress occuring at the
surface of the wire can be obtained from the bending equation (2.100)

Tr 2FRr 2FR
r =-=--
1 nr 4
=-nr -3 (6.5)

As might be expected for such a common machine element, extensive studies

of coil springs have been presented over many years. Probably the most notable
compilation of this body of work is provided by Wahl, 1963. A more complete
solution for the compliance and maximum stress in a coil spring includes the
pitch angle, a, and poisson's ratio, v, of the wire. For a spring of circular wire, the
compliance can be calculated from, Young, 1989,

!.._= 4nR3 [~-~(!...)2 + 3+v {tanaY] (6.6)

F Gr 4 16 R 2(l+v)

The maximum shear stress in the circular wire is given by

(6.7)

Correspondingly for a coil spring of square wire (R > 3b)

o 1
2.79R 3n
F =-;; = Ga 4
(6.8)
4
<= ~~R[l + l~a +05{~)' +0.{~)']
To see why coils springs are used in such a broad range of applications, it is
informative to investigate various spring geometry's in terms of utilization of

289
FLEXURES

material. A straight rod subject to an axial applied load provides the simplest
spring element to analyze and represents the most efficient geometry. At the
limit of performance, the whole volume of the spring will be subject to the same
maximum stress.. Consequently, such a design may be considered to represent
100% material utilization as defined by the ratio of average stress in the volume
of the spring divided by the total maximum stress-volume product. Although
such a geometry might be reasonable for materials that can sustain high strains
(i.e. rubber and other elastomers), for metallic, ionic and covalent solids, such a
spring would be impracticably long if appreciable displacements were required.
Other simple spring geometry's include the coil spring, a simple cantilever
subject to a bending moment at the free end, a simple cantilever subject to a force
at the free end in a direction perpendicular to the neutral axis and a leaf type
flexure constrained to produce an 's' shaped deflection curve as discussed in
section 4.1. For the coil spring made from wire of circular section, the shear
stress in the plane of the wire section is a linear function of the radius from the
center of the wire and at maximum deformation is given by

(6.9)

By definition, the materials utilization, u, for this spring geometry is

u= JJL rd(vol) (6.10)

(vol)rmax

Material utilization can be readily evaluated for the other geometry's. As an

example, the simple cantilever beam with a bending moment applied at the end
will have a constant stress along the axis with a linear variation from maximum
compressive on one face to an equal, tensile stress on the other. Clearly, the
average stress and, therefore, materials utilization is 50% of maximum as
measured across the face and constant along the axis. For an applied force at the
free end of the beam there will be a linear variation of bending moment starting
at zero at the free end to a maximum value at the support. For these simple
geometry's the material utilization is tabulated overleaf. Clearly, the coil spring
represents one of the best geometry's if we desire to utilize as much material as
possible. They also provide relatively large displacements for a given strain,
occupy a small volume and are, almost, rotationally symmetric. Clearly, the coil
spring is an efficient and useful strain energy storage element.

290
CHAPTER6: HINGES OF ROTATIONAL SYMMETRY

Table 6.2.1: Materials utilization for different spring geometry's

Spring geometry Materials utilization

(%)
Rod subject to axial load 100
Coil spring wound from circular wire 67
Cantilever subject to moment at free 50
end
Cantilever subject to force at free end 25
Leaf type flexure undergoing 's' 25
shaped deformation

With a substantial knowledge of coil spring performance and materials

selection accumulated over many decades, it is somewhat of a disappointment to
realize that they are far from ideal for use as a flexure. The main reasons for this
are
• The compliance in all other degrees of freedom is very high. Consequently,
this does not provide 'guidance'. Equivalently stated, the coil spring often
provides too many freedoms.
• With compression of the spring the diameter of the helix increases.
• As the coil spring is freely compressed the two ends will rotate resulting in a
coupling of linear translation and axial rotation, or twisting, of the spring.

6.3 The disc coupling (freedoms Bx,B,,x )

Connecting two shafts via a thin disc can readily produce an inexpensive
coupling. In practice there are two common configurations for such a design, see
figure 6.1. Most common for applications involving high pressure sealed
couplings is the connection from the center of a disk to the outside rim, Wolff,
1951. For this coupling one p ipe is joined to a hub in the center of the disc, or
diaphragm. The second connects to the outer rim, figure 6.1(a), and will be
referred to as an inner to outer rim disk coupling. In the second, both shafts are
connected symmetrically about the circumference, figure 6.1(b). Henceforth, this
is called an outer rim disc coupling. Design equations for each of these will be
discussed in turn. Because this coupling is designed to provide one axial and two
angular freedoms while transmitting a torque it is easy to confuse the terms
describing the various components of stiffness. Considering this coupling as a
simple idealized joint connecting two rigid bodies (in this case connecting the
ends of two shafts), a simplified model of this mechanism is shown in figure 6.2.
For most applications, the disc coupling is required to transmit torque while
permitting 'free' translation along the axis and two rotations perpendicular.
Consequently, these are considered to be three degree of freedom couplings. In

291
FLEXURES

L
a)

X y

Cross section

b)
Isometric view

Top view
Side view

z z

Figure 6.1: Two types of disk coupling, a) the inner to outer edge coupling, b) the outer edge
disk coupling

the following two sections, equations for the stiffness provided by the coupling
between shafts in relevant coordinate directions are presented. For clarity, the
following terminology will be used to describe individual components of
stiffness

292
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

Shaft 1 Shaft 2 ... y

- _j

"\.__ Coupling

Figure 6.2: A schematic representation of a coupling (the disc coupling in this

case) modeled as a joint connecting two rigid shafts

1, I Axial stiffness', k6F1

y y
relates the relative motion of the shafts, 0, v

towards or away from each other as a consequence of an axial force

FY.
2. In most applications, disc couplings are used to transmit torque from
one shaft to another. The ability to faithfully transmit rotation can be
assessed from the 'torsional stiffness', this being the ratio of the
applied torque, MY, to the relative angular motion, ¢>' (i.e.
k1,My =MYI ¢Y) between the two shafts.

3. Correspondingly, for a large number of designs, misalignment of the

axes of two shafts is notoriously difficult to eliminate. Again the
compliance of this coupling to accommodate these misalignments is a
common concern. Hence the 'angular stiffness' (which, by symmetry,
is constant about the x - z plane), refers to the ratio of the applied
torque perpendicular to the shaft axis, Mx or M:, to the subsequent
angular rotations ¢x or ¢z respectively i.e. kii,M, =k1,M, =Mx I ¢x.
6.3.1 The inner to outer rim disk coupling
Of primary interest for this coupling is its ability to transmit angular rotation
about the axes of the shafts. This is the torsional stiffness of the coupling for
which an approximate value can be computed from, Cherukuri, 1997,

(6.11)

293
FLEXURES

where a and b are the inside and outside radii respectively, t is the disc thickness
and G is the modulus of rigidity, see section 2.4.3.
A torque of magnitude, M Y' will produce pure shear, the contours forming
concentric circles. At a radius r the shear stress is given by

M
-r =--y-
riJ 21Cir 2
(6.12)

From von Mises failure criterion, the maximum torque that can be
transmitted is given by

(6.13)

Care should be exercized here in that an upper case Yin the subscript of u is
used to denote the yield stress and not stress in they-direction.
Considering only radial bending of the disk, the axial stiffness of this
coupling is, Young, 1989,

k = FY = 2traD[C C
1 2 - C ] _, = 2traD C (6.14)
6, F, Oy b3 C4 3 b3 S

where

c, =±H*n'·2'{~J)J
C, = :b[(*)' -1+2ln(;)]
{[(*)' .,H~H*)' -~}
(6.15)

c, = ~

c. =~Hm
Stresses in the disc can be computed (i.e. u = 6M 12nrt 2 ) from the maximum
bending moment which occurs at the inner edge of the disc and is given by

294
CHAPTER'6: HINGES OF ROTATIONAL SYMMETRY

a)
100000

10000 •
cS 1000 ~

100 ..... ~

10
0 0.2 0.4 0.6 0.8
alb

b)
0.3
0.25
0.2

u 0.15
'0

0.1
0.05
0
0 0.2 0.4 0.6 0.8
alb

c)
1000

z
.L

~ 100
.~

~ ..........
10
0 0.2 0.4 0.6 0.8
alb
Figure 6.3: Geometry factors for determination of the deflection and moments for an inner to
outer djsc coupling subject to an axial load, a) the deflection constant C5 , b) the moment
constant C6 , c) the product C5 C6

(6.16)

295
FLEXURES

From which

(6.17)

The constants Cs and C6 and their product are plotted for ratios ajb ranging
from 0.1 to 0.9 in figures 6.3(a- c). From this figure, it can be seen that, for ratios
of ajb between 0.1 and 0.35, the product of these two factors is nearly constant at
a value of around 20. In this case, equation (6.17) can be rewritten as

10Et 2 t5y
a max ~ (1 - v2 )ab2 (6.18)

The two other

freedoms of this coupling
have identical stiffness
corresponding to angular
deflections perpendicular
to the coupling axis. A side z
view of this disc coupling
indicating modeling
parameters is shown in
figure 6.4. It can be shown
that the maximum stress
for a given angular Figure 6.4: Cross section of a clamped circular disc with a
deflection again occurs at moment applied to a central annulus of radius a
the inside edge and can be
obtained from

(6.19)

Correspondingly, the moment required to induce an angular deflection ¢can

be obtained from the equation

(6.20)

The stress and angular stiffness factors, a and p, are plotted in figure 6.5 for
the ratio ajb ranging from 0.1 to 0.8.

296
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

1000

...0 100
-·
-+-a.
13 •
/
•

. • ----
1:S
.:3 •
.... . .
~ 10
u
8
~ ~

0.1
0 0.2 0.4 0.6 0.8
alb
Figure 6.5: Geometric factors for prediction of maximwn bending stress and angular
deflections a of equation (6.19) p of equation (6.20)

6.3.2 The outer rim disc coupling

Analysis of the out rim disc coupling is less straightforward. This section assesses
the stiffness of the coupling in the three freedoms that correspond to; angular
misalignment that requires two rotational freedoms and axial motion requiring a
linear freedom in the axial direction. Because this type of coupling is used for the
transmission of torque, analysis of torsional stiffness is also presented. For an
assessment of the equations developed in this section see Woody and Smith,
1999.
An approximate expression for the bending stiffness of a disc type coupling
can be derived, provided that the following assumptions are made,
1. The width of the disc (i.e. difference between inside and outside radius) is
small in comparison to its radius so that straight beam theories can be
applied. In almost all cases the thickness of the 'beam' will also be
substantially less than its width.
2. Displacements due to shear induced by loads normal to the beam axis can
be ignored.
3. The distribution and magnitude of external loads is not affected by
displacements of the beam.
4. Superposition is applicable.
5. Twisting about the axis of the disc induces no bending. Twisting can be
visualized by cutting the circular disc and straightening it out so that it
becomes a straight beam. The twisting couple is applied about the
longitudinal axis of this beam. When bent back to a circular arc, the
twisting moment is always tangential to the axis.

297
FLEXURES

Under these conditions, the equation governing deflections of a circular ring

due to symmetrically applied bending, M8 , and twisting moments, To, is given
by,Levy,1962

y + ()y) = El(y.,
3
EI(o + ')= _ oM8 +A-T.8 (6.21)
R2 ofi oB R2 y oB

The primes indicate partial differentiation with respect to B, I is the second

moment of area about the neutral axis of bending and A, is the ratio of the
bending to torsion rigidity given by

A-= EI (6.22)
GJ

For a beam of rectangular section, this can be computed from equations

(6.57) and (6.61) of the following section. From the expressions for J, it can be
shown that the above ratio is equal to 1+ u for a disc of circular cross section
section (i.e. a circular ring, made from a wire of circular section, an unusual
design), (1 + u)/ 2 for a thin plate ( J = b£3/3) and up to approximately 1.66 for a
square cross-section made from a material having a Poisson's ratio of 0.3.
Consequently, values ranging from 2/3 to 5/ will be used in the following
analyses. Solutions to equation (6.21) must satisfy three boundary conditions
given by the symmetric condition (in this case) of zero slope and deflection at the
supports plus the requirement for prescribed slope or deflection at B = 90°.
Ultimately, for rotation of the axes of the disc coupling the slope at the fixed
points are constrained so that the surfaces of the disc, where clamped, always
forms a straight line perpendicular to the axis of the clamp. This rather complex
boundary condition can be obtained from the combination of individual
solutions for twisting and an applied normal force at B =90° as shown in figure
6.6.
Having obtained an expression for the displacement, the twist of the beam
fJ (rad) can be derived from

(6.24)

As a first step to the solution of this problem, consider the case of the circular
ring subject to a twisting moment, 2T, at the center (i.e. B = 90°). From this,
because the loads will be equally supported by the two quadrants, the
subsequent bending and twisting moments in one quadrant are

M 8 =Tcos(B)
(6.25)
To =Tsin(B)

298
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

y
z

2T

X
Figure 6.6: Parametric representation of a clamped semi-circular ring

Substituting (6.25) into (6.21) gives

E~ (y"' + y')= T(I +A.)sin(O)

R (6.26)
y"' + y' =bsin(B)

where

b =T(l + A.)R
2
(6.27)
EI

A general solution to this linear, differential equation can be obtained from

the complementary function, y<I, that provides a solution to the homogeneous
component of equation (6.26) plus a particular integral, Ypt, that necessarily
contains a non-coincident term in sin(B)

Y<t = A cos(B)+ Bsin(B) + C

(6.28)
Ypt =(D+EB)sin(B)

Differentiating the second of equations (6.28) gives

y~ =Dcos(B) + Esin(B)+ EBcos(B)

y;,=- Dsin(B) + 2Ecos(B)- EBsin(B) (6.29)
y: =- Dcos(B)- 3Esin(B)- EBcos(B)
Substituting (6.29) into (6.26) yields

299
FLEXURES

2Esin(B) = bsin(B)
(6.30)
b
E =-
2

Hence the complete solution can be obtained from the sum of the two terms
in (6.28) from which

y = Acos(B)+ Bsin(B) + C + !!.Bsin(B) (6.31)

2

The boundary conditions for this case are

(6.32)

From the first of these A= -C and equation (6.31) reduces to

y = C(l - cos(B))+ Bsin(B) +!:Bsin(B)

2
(6.33)
y' = C sin(B) + B cos(B) + .£.sin(B) + .£. Bcos(B)
2 2
The second of the three boundary conditions result in the equations

y'(O) =B = 0
b (6.34)
y'(7!/ 2) = c + - = 0
2

Substituting (6.34) into (6.33), the general solution for the displacement of the
disc is given by

2
Y = cos(B)+ Bsin(B) - 1 (6.35)
b

It can be shown that this corresponds to case 5 of the examples given in the
paper of Levy, 1962 and, henceforth, results will be based on equations generated
in this paper. Plotting the displacement y as a function of Bcorresponds to a view
of the displacement of the ring as observed from the origin of the coordinate
system, figure 6.7(a). To view the displacement from the side as a function of the
distance from the center (i.e. viewing the deflected beam as projected onto the x -
y plane) it is necessary to plot (6.35} as a function of xjR as shown in figure 6.7(b).
The slope of the disc due to bending as a function of the distance x from the
center (i.e. once again, as woul~ be observed from the projection of the neutral
axis onto the the x - y plane) is given by

300
CHAPTER'6: HINGES OF ROTATIONAL SYMMETRY

Figure 6.7: Dimensionless displacement of a semi circular ring due to an applied twisting
moment at its center, a) radial view from the origin, b) view of the neutral axis projected onto
the x - y plane

dy dy dB bB T(l+A.)B
-=-- =-=--'--~ (6.36)
dx dB dx 2R 2R

The twist of the beam is given by

f3r = TR [(cos(B) + Bsin(B))+ A.(Bsin(B)- cos( B))] (6.37)

2EI

Figure 6.8 shows plots of the function in parentheses for three values of the
parameter A.. Examination of this graph reveals one of the potential errors
associated with the assumptions of this model. At the fixed end a finite twist is
predicted, the magnitude of which is related to the deviation of A. from unity. In

301
FLEXURES

7
6
5
El:: 4
!-:;
~ 3
~ 2

B (deg.)

Figure 6.8: Twisting along the beam as a function of the angular position B for values
orA = 2/3, 1 & 4/3
practice, twisting moments will be, at least partially, constrained at these
supportsand is a potential source of error. Such a complication is ignored in the
proceeding analysis.
From Levy, 1962, the displacement, y, of the semicircular section due to a
normal force of magnitude 2F is

4El [2BsinB + (n- 2Xt- cos B)- 2sinB + 2BcosB ]

(6.38)
FR 3 y = - A.((6 -nXl-cosB)+ 6sinB -48 -2B(cosB +sin B))
This is plotted in figure 6.9 for a range of values for A.. Proceeding as for the
previous example, the slope of this curve in x is

4EI dy [2BcosB + (n- 2B)sinB J1 (6.39)

FR 2 dx = -A.((4 -n+2B)sin0+(4-2B)cosB -4) cosB

The twist due to this bending moment is given by

2
f3M = FR (1 +A.X2BsinB+(n-2)cosB-2sinB+2BcosB] (6.40)
4El

For the purpose of matching the boundary condition at the point of applied
load, it is necessary to determine the expressions for the slopes and
displacements at the center of the semi circular ring. The relevant equations are
tabulated below

302
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

3.5
3
2.5
...
-!!:
~
..,. 1.5
2

0.5
0
0 0.2 0.4 0.6 0.8
xJR
Figure 6.9: Displacement of a semicircular beam due to a nonnal force applied at
the center

Table 6.3.2.1: Equations of displacement, slope and twist in one quadrant of a circular
ring

Parameter E quation at tr /2 Equation

no.
y torsion TR 2 6.35
- ( I +A.){tr-2)= DrT
4EI
dy . T(I + A.)Rtr = C T 6.36
-torsiOn
dx 4EI I

Pr TR 6.37
- t r(1 +A.)= C2T
4El
y bending FR 3 6.38
- ((2n -4)+ A.(4n -12)) = D8 F
4El

dy FR 2 6.39
El (n- A.(4 -n)) =C3F
dx 4
bending
PM FR 2
- (1 + ..tXn- 2):= C4 F
6.40
4El

It is informative to compare equation (6.38) with the expression derived by

Timoshenko, 1941, using Castigliano's equation shown overleaf

303
FLEXURES

Figure 6.10: Side view of coupling indicating distortion due to applied bending
moment (for the purposes of fmite element modeling the bending moment can be
generated by applying equal and opposite forces, F, on faces A and B)

_2 E
_ I_Y.:.::.:Io.==='"':..:...2 = 0.57 +A-(0.142) 6.38
(2F)R 3
=0.47 + A-(0.037) Timoshenko

6.3.2.1 Angular stiffness

A constraint imposed by the disc coupling is that of common slopes at the
surfaces of the ring at the point of clamping. Also, the center lines of the cross
section must also pass through the pivot point, see figure 6.10. In this model the
displacement and angle of the surface at the clamp has been assigned the
symbols y and ¢r respectively. Clearly, these will comprise the sum of the effects
of both 'torsion' and 'bending' due to a normal force which, if known, can be
computed from the formulae given in table 6.3.2.1 above. Using the expressions
in this table, the parameters for the model of figure 6.10 yield the equations

y =Ys- Yr = DsF - DrT (6.41)

¢=~-~=~F+~F - ~T-~T=~F-~T

The boundary condition of common slopes coincident with the pivot

provides the condition

y
- =¢ (6.42)
R

304
CHAPTER6: HINGES OF ROTATIONAL SYMMETRY

"
1.2 ,--- - - - - - - - - - - - -

1.1 +-~""'
---------
1+----~-------------

~ 0.9 ~--~
--""~
~------
0.8~-----~~
----

0.7 +-- - - - - - - - - - - =-...-....

.::----
0.6 ...__ _ _ ____._ _ _ _ ___,__ _

0.6 1.1 1.6

A.
Figure 6.11: Stiffness factor JM(.:j.) for the bending of the disk
coupling
It is important to remember that, in this section, ¢ refers to rotation about
either the x or z axes. Substituting (6.41) into (6.42), it is possible to eliminate the
twisting couple applied to the ring from

(6.43)

Additionally, it is known that the applied bending moment is simply the

sum of all moments generated in the four quadrants of the disk Consequently,
the force can be expressed in terms of the applied moment in terms of the
relationship

M = 4FR+4T=4F( R+~) (6.44)

Substituting (6.44) into the second of (6.41) yields

M =k =16El[ (tr+4)+A.(8 - tr) J

¢ M; R (2tr 2
- 4tr- 4) + A.(8tr 2 -18tr -16)+ A.2 (6tr 2 - 141Z' - 12)

= 16:/ JM(.i)

(6.45)

The functional relation JM(A.) is shown plotted figure 6.11. For common
value of A. this is usually close to 1. In practice, it is the ratio of the bending to
torsional stillness that is of concern to the designer and this is discussed at the
end of section 6.3.2.3.

305
FLEXURES

6.3.2.2 Axial stiffness

The axial stiffness of the disc type flexure can be obtained from the boundary
condition

(6.46)

From this, the displacement, y, due to an applied force, F . (= 4F), is given by

(6.47)

An equation for predicting an important case in which the axial stiffness of a

coupling comprized from segments of included angle less than 90 degrees (i.e. (}
< 1r/2) is developed in section 6.4.

z t 6.3.2.3 Torsional stiffness

To determine the torsional
stiffness, it is necessary to
evaluate distortions due to
forces and moments acting in
the plane of the disk only.
Again, the disc coupling can
be modeled as comprising the
sum of four arcuate segments,
each of included angle If, fixed
at one end and subject to
combined forces and moments
at the other. Displacement of
the 'free' end of the segment
can be derived from the
superposition of the effects of
an axial and radial force plus
an applied moment. Both the
forces, F.,,FR, and applied
moment, M, result in in-plane
distortion of the segment,
Figure 6.12: Parametric model of a single segment of a figure 6.12. From symmetry, it
leaf spring subject to in-plane loads is assumed that the deflection
of the 'free' end is in the

306
CHAPTER 6: HINGE$ OF ROTATIONAL SYMMETRY

direction of the axis. The displacements, u,,uR, of the free end and subsequent
angular deflection, 88 , can be calculated using Castigliano's theorem from the
integral energy equation, Young, 1989

"JlFRR sin( B) + F, R(l - cos(B)) + M jMB

U= (6.48)
0 2EJ

where, in this case, I corresponds to the second moment of area about the neutral
axis which is normal to the plane of the disc (i.e. I = t(R0 - R1 ) 3 / 12 ).
Correspondingly, the displacements and angular deflections due to the
applied forces and moment are given by

u = oU 2
= !f_[M(l{l - 2sin(I{I))+FRR(l+cos (1{1) - 2cos(l{l))l
" oF" 2El .. +FtvR(3fJ1+cos(fJI)sin(l{l)-4sin(fJI))
uR = oU = Jf_[2M(l - cos(I{!))+ FRR(I{I- cos(l{l)sin(l{l))l
oFR 2El .. +Ftv R(l+cos 2 (fJI)-2cos(l{l))

8= oU = ~[FRR(l-cos(fJf))+MI{I+F,R(I{I - sin(fJI))]
oM EI (6.49)

Subjecting the segment to the condition of zero angular and radial deflection
at the free end yields

F =F sin(l{l)[2cos(l{l)-2+sin(l{l)'lf]
2
R " 2 sin (fJI) + 4(COS(fJI) -1)- COS(l{I)Sin(l{l)f/1 + f// 2
(6.50)
M = F R 3sin 2 (VI) + 4(cos(l{l) -1)- cos(l{l)sin(l{l)'lf + 1{1 2 - sin(l{l)f/1
" 2 sin 2 ('I')+ 4(cos(I{!) - 1)- cos(fJI) sin(fJI)f/1 + fJ1 2

Substituting equations (6.50) into (6.49), the axial deflection, " "'' of the beam
becomes
3 3
u = F,R [4sin(l{l)cos(l{l)+51{1-fJ1 -4(sin(l{l)+cos(fJ1)'11) - cos 2 (1{1)1{1]( . )
6 51
" 2EI 2(1 + cos 2 (VI))+ cos(l{l)sin(l{l)f/1 - 4cos(fJ1) - 1{1 2

For a coupling comprised from 4 segments, the applied axial toque, T, is

T =4FR. Defining the angular rotation, ¢Y =u., I R , about the central axis, the
angular stiffness, kr1, ,of the disk coupling is

k _
Tl, -
8£/(
R
2
2(l + cos (fJI))+cos(fJI)Sin(fjf)1f-4COS(fJI) - 1{1 2
4sin(fJ1)cos(fJI) +Sf//- fJ1 3 - 4(sin('l') + cos(l{l)l{l) - cos 2 (fJ1)fJI
)
6 52
( · )

307
FLEXURES

45 55 65 75 85
If/ (degrees)

Figure 6.13: Factor /(If/)= kr8 R/8EI for the torsional stiffness of a disc
coupling

The expression in parentheses of equation (6.52) for the torsional stiffness as

a function of the included angle of the segment is plotted in figure 6.13. In
practice, additional clamping, boundary, shear and other compliances c.an
significantly reduce this effective stiffness, Woody and Smith, 1999. For a disc
coupling consisting of four quadrants, equation (6.52) reduces to

(6.53)

Equation (6.53) applies to the torsional stiffness of the hinge of the following
section. In practice, it is unlikely that a disk coupling comprised from arcuate
segments of included angle 90° will be produced. However, viewed as a coupling
to provide specific constraints and freedoms, it is more relevant to consider the
ratio of stiffness in different coordinate directions. Particularly relevant to the
disk coupling is the ratio of torsional to angular stiffness. Dividing equations
(6.53) by (6.45) yields

kr;, (R0 -R,)2 (8-.7r 2 ) ll(Ro -RY

kM; ::= t 2 (201r-32-.7r 3 ) jM(J.,) ~ /2

In most designs, this ratio will range from 50 to 500 or more and increases as
the square of the ratio of the breadth to thickness of the disk. For couplings of
included angle less than 90° this is ratio not expected to be significantly different.
6.4 Rotationally symmetric leaf type hinge (axial stiffness of the disc
coupling revisited)
In many ways similar to the disk coupling of the previous section, this type of
flexure is also used to provide three freedoms, one translation and two rotations.

308
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

Figure 6.14: The rotationally symmetric hinge, a) isometric view, b) exaggerated axial
distortion, c) modeling parameters for deriving axial stiffness equations

Although, in principle, the same technique could be employed for the prediction
of axial stiffness of this flexure, satisfaction of the boundary conditions for the
sectors leads to considerable mathematical complexity. An alternative analysis is
presented in this section.
Figures 6.14(a & b) show both an isometric view of such a flexure and a
exaggeratedly distorted flexure after a linear distortion in the axial, or z-axis,
direction. In its simplest form, this flexure can be easily manufactured by making
four slits in a tube. First, two slits are cut into either side of the tube directly
opposite one another. This can be readily achieved by slitting the tube in a
milling machine using a slitting saw. With the tube vertical, the first cut would
make a horizontal slit perpendicular to its axis the depth of this being less than
half the outside diameter of the tube. A second identical slit is made on the other
side of the tube after which two small bridges remain that prevent the tube from
separating into two pieces. Following this, the tube is then rotated about its axis
by 90 degrees and the slitting saw moved down a distance corresponding to the

309
FLEXURES

thickness of the saw plus the desired flexure depth. Repeating the same two slits
at this depth will then produce the complete flexure. It is important that the
upper slit is made first so that the final machining forces are not supported by
the thin flexure (it is left to the reader to see why this is so). Other manufacturing
methods such as electro-discharge machining or fabrication do not impose a
sequential manufacturing requirement, see chapter 8. Once machined, this
flexure is simply a monolithic version of the disc coupling discussed above with
each segment necessarily having an included angle, If!, of less than 90 degrees.

6.4.1 Axial stiffness and maximum stress calculations for the rotationally symmetric
leaf type flexure system
To derive a suitable formula for the axial stiffness of a rotationally symmetric
hinge it is possible to consider the full flexure consisting of the superposition of
two cantilever arches subject to a force, Fy, at the ends that is normal to the plane
of curvature, see figure 6.14(c). Under these conditions the deflection at the end
of the cantilever can be determined by integrating the bending moment and
torsion equations and assuming that superposition holds, see Seely and Smith,
1932. Under these conditions, the moment Mx and torque Mz at a position
specified by the polar coordinate If/can be obtained from the equations

Mx = FyR,. sin If!

(6.55)
M t = FyR.. (1 - COS If/)

Using a techniques called the 'dummy load method', the end deflection 0y of
the arched beam can be determined from the equation
2
"'JFY R~ sin 1f1 "JFYR~(l-coslfl)l
o, = EI
dlfl +
GJ
dlfl
0 0
(6.56)
=FYR~
- -[("'
- sin(21f/)) + 2(l+v)I(31f1 .
- - 2 Stnlfl+ sin(21f/))]
--'-'-...:..
EI 2 4 J 2 4

where Rm is the mean radius of the beam and If! is the angle subtended by
cantilever.
From equation (2.104) of chapter 2, the polar second moment of area for a
beam of rectangular section is given by the equation

310
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

J
3
= (Ro- R, )d [I - 19; d f ~tanh(nn(Ro- R, ))]
3 tr (R -RJ,~J.l . n
0 2d

= 4I[I-I~ _s_ f ~tanh(nn(I-r))] (6.57)

fr (1 - r) ItA ), ) •• n 2£
=4lfAs,y}
where y and & are the dimensionless ratios given by

R
r=-· (6.58)
Ro

(6.59)

It can be readily shown that

o<r < I (6.60)

From (6.57), the ratio of bending to torsional rigidity (A. of the previous section) is

A. = 2(1 + v )I = (I + v) (6.61)
J 2/J(s,y)

Although these values have been arbitrarily chosen, it would be expected

that y and & would typically be within the range

0.75 <r <0.95

0.01 <& <0.1
Noting that

equation (6.56) can be rearranged to provide the linear stiffness of the curved
beam given by

4E(I - r)ds vr- sin(2vr)) + (1 + v) (3vr . +-..:......:.....;.

sin(2vr))J-'
2
k = [(
- - 2smvr
•V. 3(1+3y+ 3y 2 +r 3 ) 2 fAs ,y) 2 4
(6.62)

311
FLEXURES

Connecting two curved cantilever beams together at their ends will produce
a curved spring represented by a cord of total angle 21f/ with clamped ends
subject to a linear translation of the supports in a vertical, or y axis, direction
perpendicular beam axis. Clearly equation (6.62) corresponds to the y axis
stiffness of the rotational flexure subtending a total angle 0. The stiffness of this
new configuration will be equivalent to half of that for each cantilever.
Consequently, the stiffness of the complete flexure is given by the equation

ks " -_ ( 2E(I - r)de )

1
[( sin(21f/)) (1 + v) ( 31f/ 2 . sin(21f/))]-'
If/ + - - sm 8 + --'---'--=-
, 31+3r+3r 2 +r 3 2 /;(e,r) 2 4

= 2E(l - r)de [(~_sin(B))+ 2.1.(38 _ 2 sin(~) +sin(B))]-'

1

3(1+3r+3r 1 +r 3 ) 2 2 4 2 4
(6.63)

An alternative, and considerably simplified, approximate formula for the

stiffness of the flexure can be obtained by conceptually splitting it into radial
strips. Assuming that each strip will distort along its axis in the same way as a
simple leaf flexure undergoing an's' shaped distortion the elemental stiffness of
each strip can be obtained from equation (4.9) of chapter 4 i.e.

(6.64)

2.5 ----- ---------------

O··r::::=~~~=;:;i
0.75 0.8 0.85 0.9 0.95

r
Fig ure 6.15: Stiffness of a single arcuate element using both the full and simple
theoretical models of equations (6.63) and (6.66). E = 210 GPa, R0 = 1, v = 0.3
values of£ correspond to 0.04, 0.03, 0.02 and 0.01 with the larger values
corresponding to high stiffness

312
CHAPTER ·6: HINGES OF ROTATIONAL SYMMETRY

0.5
g 0
·Q 5
"' ·1
~ ·1.5
~ ·2
0

~ ·2.5
.tl ·3
"*- -3.5
-4

r
Figure 6.16: Percentage deviation between the simple and full
theoretical models for the plots shown in figure 6.15

Integrating (6.64) yields the total stiffness, k, of the flexure

7}Ji3 = 283
3 3
1 1)
R;
Ed R. dR Ed (
k61F1 = R,2 -
(6.65)

In terms of the dimensionless parameters E and y this can be expressed by the

equation

(6.66)

Equations (6.65) and (6.66) have been plotted as a function of the variable y of
the range 0.75 to 0.95 with values of E ranging from 0.01 to 0.1, figure 6.15 with
the percentage stiffness deviation between these two equations plotted in figure
6.16. Both graphs have been computed using an included angle of 8 = 75 degrees.
For symmetry, in minimum of four hinges per flexure is necessary and therefore
this is considered to represent an upper bound on the angles likely to be
encountered with this particular hinge design. It can be seen that the error for all
plotted values of y does not exceed 4%. For higher values of y the error tend to
deviate to greater than this limit with rapid deviations occuring earlier for the
higher values of E. Considering that the range of values considered in this
analysis represents the limits of design commonly used, these may be considered
upper limits.
6.4.2 Simplified equation for maximum stress
Using the simple model of equation (6.66), the maximum stress is found to occur
at each end of the flexure at the inside edge and can be determined from the
equation

313
FLEXURES

(6.67)

or

(6.68)

The range of validity of this equation has yet to be experimentally

determined. However, finite element methods provide a complimentary method
for comparison. The results of such a study are outlined in the following section.
6.4.3 Assessment of approximate equations using finite element models
Ultimately, equations must be verified experimentally. In the absence of such
information, finite elements represent a reasonable check on the validity of
mathematical models. In this analysis, finite element models were generated and
computed. In all analyses solid pyramidal elements were used with nodal points
midway along each line. Although these elements provide reliable results for
stiffness calculations, they are less suited for solutions for determination of stress.
As a consequence considerably more deviation between stresses predicted from
simple beam bending and finite element analysis should be expected. A simple
curved beam was used with one end fixed and the other constrained with a
freedom in the direction of traverse only. Force was applied as a point load at the
central node on the face of the moving end of the beam. As a result of this, there
was a stress concentration at this node. Consequently, the stress was determined
from the maximum value at the fixed end. No significant differences were found
in a comparison the results from an alternative model in which a traction force

15.00 -+-gamma "'0.75 1 - - - - - - - - -

'0' - gamma =0.85
.._,
~ 10.00

=
0
5.00
·~
·~
-8 0.00

I -5.00

-10.00
0 0.02 0.04 0.06 0.08 0.1
&

Figure 6.17: Stress deviation between simple formula and results from
finite element analysis (8= 45 degrees)

314
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

10.00 ~ gamma=0.75
,.... - gamma =0.85
~ 5.00 ~-3o...,.-------; _._gamma =0.95
~
d
0
'D 0.00
·s:"'u
-o -5.00
~
~ -10.00

-15.00
0.01 0.02 0.03 0.04 0.05
&

Figure 6.18: Stress deviation between simple formula and results from finite
element analysis (8 = 60 degrees)

was applied to the moving face. Deflections were measured at the moving face
with variations across the face due to strains being insignificant in comparison to
the macroscopic displacements of the elements.
The closeness of the finite element model to the stresses and deformations of
a real flexure is always difficult to assess. In practice, it is more general to
consider accuracy as the extent to which the model corresponds to an analysis
with an infinite number of elements or the 'exact' solution based on continuum
models incorporating the same boundary conditions. For all of the geometries
analyzed, the stress was concentrated predominantly in the inside region of the
hinge at the end suppports. As a consequence, finite element meshes were
automatically generated with the ends of the beams partitioned in these regions
to contain a fine mesh having at least four elements in the radial and thickness
directions while the central region of the hinge was reduced to fewer elements.
In general, for such meshes, results appeared to converge to within a few percent
for meshes having more than a thousand elements. For some geometries,

--
10.00 r -· --

- - ......
,....
'.._,
?!. 5.00
~
8 0.00 ~

'D /,
-5.00
·s:u"'
-10.00 / -+-gamma =0.75
-o
~ -15.00 • gamma=0.85

~ -20.00 - . -gamma =0.95

-25.00
0.01 0.02 0.03 0.04 0.05

Figure 6.19: Stress deviation between simple formula and results from finite
element analysis (8= 75 degrees)

315
FLEXURES

20.00
-+-g =0.75 (simp)
- g = 0.85 (simp)
15.00 +-- - - - -- - -- - -- ---1 --.-g =o.95 (simp)
-++-g "0.76 (MI)
"$. 10.00 -t---"-....-::__::'Ko.;;;;:::-- -- - - -----; - 9 .. 0.85 (full)
d
0 ~g = 0.95 (fuJI)
".cl
-~ 5.00

.,
41)
'"«::

t
0.00
0.12
-5.00

-10.00

-15.00

Figure 6.20: Stiffness deviation between simple and full analytic models and results from finite
element analysis (8 = 45 degrees)

pathalogical results appeared in which case the number of elements was

increased significantly. In all cases the deviations between model and finite
element analysis are derived from the equation

. d . ti. ( results from finite element analysis)

Frae t10na1 ev1a on = 1 (6.69)
values calculated from formulae

Percentage deviations are simply obtained by multiplying equation (6.69)

by 100. Based on the above equation, negative errors correspond to the values
based on finite elements being higher than those predicted from the theoretical
models. Assuming that the finite element models correpsond to the 'true' value
-+-g =0.75 (simp)
- g • 0.85 (simp)
18.00 - -·-- ······-- - ---- -·---------1--.-g =0.95 (simp)
"$. -++-g . 0.75 (full)
- g = 0.85 (full)
§ 13.00 +-;:::,-::-=,.,..---==--~-------l
•.cJ
-~ ,__~~;......:__j

-8

-2.00 ·-.
0.01 0.02 0.03 0.04 0.05 0.06

Figure 6.21: Stiffness deviation between simple and full analytic models and results from finite
element analysis (8= 60 degrees)

316
CHAPTBR.6: HINGES OF ROTATIONAL SYMMETRY

20.00
18.00
16.00
"$.
..,
d 14.00
0
12.00
·s:"' -+- g = 0.75 (simp)
10.00 -+----.:::.-......::----=~~-; -~~ = 0.85 (simp)
-8
~ 8.00 +-"...._: : : - - - - --=--........,=-------1 -+-g =0.95 (simp)
6.00 r-- - -----"'- -=:::::::::;;;-1 --*""11 =0.75 (full)
~ 4.00 +-------------~ ----~~ =0.85 (full)
-+-g =0.95 (full)
2.00
0.00 +--~----r---~-~~--~--~----r---~· s
0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Figure 6.22: Stiffness deviation between simple and fulJ analytic models and results from finite
element analysis (8 = 75 degrees)

then a negative error would correspond to an 'unsafe' design. However, the

'true' values for such a hinge system are not known. Consequently, the
deviations of this analysis represent the magnitude of the safety margin that
should be incorporated into such a design. Results of this analysis are plotted in
figures 6.17 to 6.22. Figures 6.17 - to 6.19 show the stress deviations between
those predicted &om the simple formula (6.68) and those computed using finite
element analysis for include hinge angles, e, of 45, 60 and 75 degrees
respectively. Figures 6.20 to 6.22 show the corresponding deviation between
stiffness predicted from simple and full theoretical models and finite element
results. It must also be emphasized that the equations should not be used for
geometries outside of those assessed here without further verification.
Approximate equations for the angular stiffness of this flexure can be
obtained &om the preceding section for the limited case of a individual segments
of the flexure comprising arcuate sectors with a total angle of 90 degrees.

6.5 The bellows as a flexure element

Few can have gone through life without encountering the bellows being used for
a flexible coupling. Sometimes called gators, examples include; the seal
connecting a manual shift gearbox to the gear-stick, the grease retaining seal
across constant velocity couplings in most vehicles and often being used by
blacksmiths in days of yore (and, as a consequence, becoming a regular backdrop
in movies set in medieval times). A little closer observation and these can be seen
on many pipelines to accommodate thermal and seismic motion and also in
instrumentation for transmission of rotation &om motor drives to micrometer
spindles, etc. In many cases, when used to accommodate motion of pipelines
under high pressure or carrying hazardous materials, their reliable function

317
FLEXURES

~ ~ ~
_j_ _ _
I
L_ _ _
-----

?~=t ~
_j_ _ _ ~
- -- - -
t
L_ _ _ i)

~~=+
_j_ _ _
cBF-~ ~LI
f) _ _ ~ i)

Figure 6.23: Ten different bellows configurations indicating mathematical

modeling parameters, a) triangular, b) parabolic or sinusoidal, c) elliptical, d) ' s'-
shaped, e) semi-circular, f) ' s'-shaped, g) single omega, h) crimped plates, I)
rectangular, j ) ' u'-shaped

becomes a safety issue. Consequently, there is extensive literature on the bellows

including the effects of pressure and fatigue life estimation, Hamada and Tsuda,
1997. Recent reviews by Wilson, 1984, and Becht, 1986, contain a wealth of
information plus extensive reviews of the literature on this subject. In practice,
the instrument designer is likely to choose a particular bellows from those
commercially available. In this section, only the important conclusions relevant
to flexures will be discussed. Consequently, this section should not be taken as a
reference for bellows design as much as a guide towards important
considerations in the choice of bellows for integration into a particular design.
Like the simple coil spring, the bellows has a relatively high compliance in
all freedoms except for torsion about its longitudinal or, z-axis. It is possible to
produce a high ratio of torsion to the other five stiffness values with a bellows
design and there are a large number of different bellows geometry's, each
possessing intrinsically different merits and limitations. Figure 6.23 taken from
Wilson, 1984, shows ten different cross sections commonly in use. The goal of
the bellows is to provide a compliant link between two rigid bodies and,

318
CHAPTER-6: HINGES OF ROTATIONAL SYMMETRY

therefore, it is only the relative displacements and rotations between each end
that is of importance. Consequently, this coupling can be modeled as being fixed
at one end (i.e. a cantilever) after which the displacement of the free end
represent the relative motion between the two rigid bodies that it connects. One
other form of distortion is that due to misalignment between the two shaft axes.
This will cause the bellows to undergo an 's' shaped distortion as shown in
figure 6.24.

-------[
Figure 6.24: 'S' shaped distortion of a bellows coupling due to parallel, mis-aligned shafts
Not surprisingly, because any particular bellows consists of repeated cycles
of an axial meander shape (sometimes called a convolution), all can be
mathematically modeled in terms of the same dimensionless groups. Therefore,
conclusions from one analysis are generally applicable across the range of
designs. The 'U' shaped bellows is most commonly used in instrumentationl.
Unfortunately, analysis of this geometry is complex and has, so far, only been
carried out using finite element methods, Hamada et al., 1976. Comparison
between these finite element results and the experimental data of Berliner and
Vikhman, 1976, deviated by a factor of almost two. Deviations of better than ±15
% were observed between these experimental values and those based on the
theoretical model of Haringx, 1952. It is the object of this section to provide the
reader with a 'feel' for the relatively compliance's of bellows type flexures.
Consequently, because of its relative simplicity, analysis of the rectangular
bellows will be chosen for illustration purposes. Although commonly used for
connecting sealed piping, it will be assumed throughout the following
discussions that the bellows are not pressurized (for those interested in this issue,
there exists substantial literature, some of which is referenced in the works cited
here). Most of the following analysis has been extracted from the review paper of
James Wilson, 1984. For the simple cantilever mode displacements, Donnell, 1932
showed that, by substituting an equivalent elastic modulus E' for E, simple beam
bending formulae could be applied. For a bellows of N convolutions, average
radius R_ , the length L and mean cross sectional area A are given by

1
These are commercially available in a variety of forms from Servometer® Corporation,
Cedar Grove, NJ 07009-1291. This company also produces an excellent design guide.

319
FLEXURES

L = 4Nb
(6.70)
A = 2nR_ t

Based on these equations, the axial stiffuess of the bellows is given by

K _ E' A _ E'21CR"'.,t
6,F,- L - 4Nb (6.71)

Similarly, the angular stiffness can be obtained from the equation

(6.72)

Consequently, for any given design it is only necessary to determine the

parameter E' that corresponds to a particular compliance. Fortunately, for
predicting the torsional stiffness of a bellows, reasonable approximations can be
derived directly from the torsion equation, see chapter 3. Some manufacturers
refer to this as 'wind up' and has the units of angular rotation per unit torque
(e.g. rad N m-1 or arc sec per inch oz, etc.).
6.5.1 Torsional stiffness of the rectangular bellows
The rectangular bellows may be considered to consist of pairs of cylinders of
length 2b connected by annular disks, or washers. Assuming that the connecting
disk has a stiffness considerably higher than that of the cylinders, the total
compliance of the bellows can be calculated for the sum of compliance's over all
cylindrical elements, N. Based on these assumptions the equation for the bellows
compliance is

1 1 1
--=-+- (6.73)
Ko,M. K, Ko

where K0/of, is the stiffness of the bellows while the subscripts i and o
represent the inner and outer cylinders respectively. Although simple to include,
the compliance of the connecting disk is ignored.
Treating the bellows as a thin shell of thickness t, a reasonable assumption,
equation 6.71 can be expressed in the form

(6.73)

Defining the dimensionless parameter

320
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

(6.75)

equation (6.73) can be written in the equivalent form

K
O,M,
= nGt [ R~ Ri ] _ nGtR:
2b R: + Ri - 2b 1+ 173
[_![__] (6.76)

Theoretically values of 17 can range from 0 to 1. In practice, this is rarely less

than 0.5 and typically greater than 0.6 for applications in which the torsion
stiffness is important. It is interesting to contrast this calculation with one in
which the mean radius R_ is used and the bellows considered as a simple thin
tube. Based on this assumption the torsion stiffness can be written immediately
as

K = GmR!.,t
(6.77)
O,M, 4b

The average radius is given by

R = Ro +R,
(6.78)
,.., 2

from which

R!., = i[R~ +3R: R, +3R~R12 +Ri ]

(6.79)
Rl
=_o [1 + 377 + 3772 + 'T/3]
8

Substituting equation (6.79) into (6.77) gives

(6.80)

The difference between these two equations can be seen from the ratio of the
stiffness values

(6.81)

Figure 6.25 shows this ratio as 17 is varied from 0.5 to 1. It can be seen that for
17 > 0.72 the difference is always less than 15%. Typically, a bellows will contain

321
FLEXURES

1
.g ~
e
0.95
o.9 .......-
~ 0.85
/
!§ 0.8 /
/
•£! 0.75 /
0.7
~ 0.85 /
/
- 0.6
] 0.55 /
0.5
0.5 0.6 0.7 0.8 0.9

Figure 6.25: Torsional stiffness ratio of equation (6.81)

five or more convolutions. However, it is not necessary to know the number of

convolutions to predict the torsional stiffness. This is apparent by assuming that
instead of the meander cross-section, all of the inner radii cylinders and all of the
outer ones are joined together. The torsional stiffness will remain unchanged
while the cross-section will resemble that of the single convolution already
analyzed. Consequently, it is possible to rewrite the formula for the single
convolution in terms of the total bellows length, L (= 4b), and this should be valid
for any number of convolutions (remembering, of course, that the shell thickness
of the bellows has been ignored). Based on this assumption, the stiffness of a
rectangular bellows based on the average radius is given by

(6.82)

Of practical interest is a comparison of the stiffness of a typical bellows over

the range of geometry commercially available. Figure 6.26 shows a comparison
between theoretical predictions based on the average radius of a rectangular
flexure, equation (6.82), and 16 different flexure geometry's for nickel based 'U'
shaped bellows commercially available from the Servometer® Corporation.
Throughout the range of available geometry the deviations between these two
values are always less than 10%. This iiiustrates a more generally accepted
assumption that, for most tJ;pes ofbellows, the torsional stiffness may reasonably be
calculated using the torsion equation for thin shells based on the average radius
of the bellows. It can be seen that there is a deviation of much less than ±10%
over the complete geometric range, even though the equation is based on the
assumption of a rectangular geometry!
6.5.2 Axial stiffness of the rectangular bellows
An expression for the axial stiffness of a rectangular bellows was first
demonstrated by Haringx, 1952. A later finite element analysis by Hamada et al.,

322
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

600

500
en
en
-+- Nm per rad
- Theo!y

~]
400

300
0
·~ 200
g
100

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
'U' shaped bellow nwnber

Figure 6.26: Comparison between predicted torsional stiffness of a 'U' shaped bellows, equation
(6.82), and values obtained from the data sheets for an 'FC' series bellows manufactured by
Servometer Inc.

1976, proved to be at variance with experimental results. However, based on its

relative success, Haringx' theory will be presented in this section. Before
presenting these results, it is worth mentioning that the relevant parameters for
the rectangular bellows can be expressed in terms of the dimensionless groups

(6.83)

where h =(R.. - R, )I 2 and represents the distance from the average to either the
inside or outside radius. Another important dimensionless group is given by

(6.84)

Consistent with these thin shell groups, the axial stiffness a single
convolution of the rectangular bellows can be determined from equation (6.71)
with E' given by

E' = z
3
(l-7J2 )
Et4A.(b l R.,...)
(6.85)
3R.,...(l +hl R,_) (l-v l(l-7] Y-{27Jln7JY]
2 2 2

323
FLEXURES

Unfortunately, this introduces a further two dimensionless groups x and 17

given by

(6.86)

It can be readily verified that the first of these is identical to the

dimensionless parameter, equation (6.75), used in the previous analysis. The
coefficients for the second of (6.86) can be obtained from

(6.87)

All but the those of an unusual mathematical inclination can derive any
insight from such a convoluted series of equations. It is possible to get a 'feel' for
the magnitude of this by calculating the ratio E'/E for a range of bellows sizes.
For the bellows geometry's of the previous analysis the values of this parameter
are shown in figure 6.27. It can be seen from this figure that the modified
modulus is less than 0.1 % of the value based on a solid tubular beam model.
This can be compared directly with the torsion equation (6.76) and interpreted as
a decoupling between the torsional stiffness along the axis and rotational
stiffness perpendicular. Alternatively, it can be said that torsion forces for a given
relative angular motion between the two ends of the bellows are three orders of

324
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

magnitude or more
0.00065
greater in the axial
0.0006 than the tangential
0.00055 direction for a
~
0.0005 typical bellows
w 0.00045 design.
0.0004 In the axial
0.00035 direction, a direct
0.0003 comparison
1 2 3 4 5 6 7 8 9 10 11121314 1516 between linear and
Bellows number torsional
compliances is
Figure 6.27: Elastic modulus scaling factor computed from equations difficult. However,
the reduction in
axial stiffness is again of the order of 1,000 or more than would be the case for a
solid tube.
6.5.3 'S' Shaped distortion of the bellows
As mentioned in the introduction to this section, misalignments of the shafts at
which the two ends of the bellows are connected will result in a 'S' shaped
distortion. Superposition can be used and this configuration can be considered as
the addition of two cantilevers of half the total length of the bellows. Assuming
the original displacement at the ends of a cantilevered bellows of length L, is
given by

8 =F L! (6.88)
z 3£'/

It is relatively straightforward to show that the stiffness of the bellows can be

obtained from

(R""" ) (Rave )
3 3
K l2E' I 12E'm 3E'1lt
F,S, =~ = (4N) 3 b =16N 3 b (6.89)

Typically, this is considerably higher than the axial stiffness of the flexure.
However, the small value of the modulus ratio in conjunction with the inverse
cube power law in N generally dominates this expression to provide a relatively
high compliance.

325
FLEXURES

a)
6.6 The notch and leaf type
hinge applied to couplings of
rotational symmetry
In some applications, it is
desired to connect two shafts
using a coupling of relatively
high torsional and axial
stiffness. Freedoms to
accommodate angular and
linear misalignments in a plane
normal to the shaft axes can be
provided by separating two
ends of a cylinder using notch
b) or leaf-type hinges. Typical
examples of this are shown in
figures 6.28(a and b). Each shaft
connects to either end of the
cylinder that constitutes the
coupling. Both couplings in this
figure are two freedom
mechanisms. The first of these
has been produced by drilling
two, closely spaced pairs of
holes through the cylinder in a
direction perpendicular to its
Figure 6.28: Two freedom cylindrical couplings, a) axis. Slots are then removed
notch hinge coupling with constraint in y, z, (}x and (}Y from the cylinder to leave notch
b) similar coupling utilizing leaf-type hinges hinges that provide, in this case,
an angular freedom about the z-
axis. Each hinge consists of two notch hinges either side of the cylinder with their
axes being collinear. Consequently, these two hinges are considered to behave as
a single hinge. For the coupling of figure 6.28(a), there are two such hinges each
providing a single angular freedom. Consequently, this can be modeled as three
rigid bodies series connected by single freedom joints. If one end of the cylinder
is considered to be the fixed link, it is clear from the figure that the other end
would be free to rotate about the z-axis as well as translating in x with the
intermediate cylinder acting as a rigid link. The second coupling of this figure
shows a mechanism of similar .function. However, this coupling utilizes leaf type
hinges instead of notches. Manufacture of such a coupling could proceed starting
from a solid cylinder and cutting wide slots perpendicular to its axis. In many
ways, this is a variant of the rotationally symmetric hinge discussed in section
6.4. In this case, however, the requisite compliance has been produced by letting
the circumferential leaf springs become thick enough to be considered rigid

326
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

while the material connecting the solid bodies has been thinned and lengthened
to form a leaf spring the axis of which is parallel to that of the cylinder. Equations
for the prediction of hinge compliances and stresses can be readily derived from
those for the notch and leaf type hinges presented in chapter 4.
Couplings of other freedoms are readily produced by adding flexure
elements or changing their orientation about the axis of the cylinder. For
example, a coupling providing angular freedoms in ()x and 81 can be produced
by arranging for the axes of the two flexure elements to be perpendicular to each
other so that the hinge axes are parallel to the x and z axes respectively. In fact,
such a coupling would result if a large hole were machined along the vertical
axis of the two axis hinge shown in the center of the photograph of figure 4.42.
The number of freedoms can be extended by adding hinges disposed at 90
degrees about and at intervals along the cylinder axis. Examples of such
mechanisms can be found in the patents of Linley, 1976, and Agius, 1994.

6.7 Case study: A metrological three..axis translator for constant force

profllometry
This case study has been extracted from the paper of Badami et al., 1996, and is
chosen because it illustrates a variety of flexure applications as well as a simple
drive coupling, the latter being a topic discussed in chapter 8.
Stylus and constant force transducers form the basis of a broad range of
instruments for metrological characterization of fine surface features, see for
example Thomas, 1982, Chetwynd and Smith, 1991, Sarid, 1994. Typically, stylus
based instruments consist of a long range traverse mechanism for translation of
the specimen relative to the probe, with the probe system mOUf\ted with its axis
of measurement perpendicular to the traverse direction. In practice, it is the Abbe
errors associated with the misalignment between the axis of the probe that
interrogates the surface and the measurement axis of the transducer used to
sense the displacement of the probe that dominate the achievable precision in
profile height measurement. This case study presents a force probe system
developed to measure fine optical surfaces with an accuracy of 1 nm and a
resolution of 0.1 nm for vertical surface profile deviations of up to 3 mm. The
heart of this profiler consists of a constant force probe similar to that developed
by Smith and Howard, 1994, combined with a laser interferometer. The force
probe consists of a simple cantilever flexure with a stylus attached at the free
end. Forces are monitored by measuring deflection of the cantilever using
capacitance gaging. Since the Abbe error is due to both the offset and the angle of
tilt, control of either parameter can be used to minimize this error. The precision
of alignment between the axis of the laser interferometer and the stylus probe is
limited by manufacturing considerations. By controlling the angle of tilt, it is
possible to minimize this error. To this end a novel, three-axis flexure
arrangement has been designed and evaluated for active control of two rotations

327
FLEXURES

Laser beams to remote

interferometer

- Top plate

PZT actuator PZT actuator

Capacitance
gage

Bottom
plate

Figure 6.29: Relative dispositions of capacitance gages, interferometer and piezo-electric (PZT)
actuators

over a range of 30 IJiad while allowing a displacement of approximately l.Sj.llll.

The tilt control mechanism may also be exploited to obtain an estimate of the
Abbe offset by monitoring the displacements as measured by the laser and the
probe while rocking the probe head through a known angle.
6.7.1 Principle of Operation
A schematic diagram indicating the relationship between the various parts of the
device is shown in figure 6.29. The force probe is mounted to the target mirror
assembly, which, in turn serves as a plane mirror for a remote interferometer.
The mirror assembly also functions as a target for capacitance gages. The PZT
actuators can be controlled individually in combination to provide tilt and
translation of the force probe relative to the top plate. Since tilt in two axes is
required, a minimum of three actuators and three capacitance gages are required.
The force probe assembly used to interrogate the surface is shown in figure
6.30. It consists of a stylus probe mounted on a glass cantilever. Electrodes
deposited on the back of the cantilever and the top of the base form the plates of
a capacitor. Deflections of the cantilever change the spacing between these
plates, changing the capacitance, thus enabling the deflections of the cantilever to
be monitored. This design is based on the work of Howard and Smith and is
capable of sub-nanometer resolution over a 1 j.llll range.
In addition to providing tilt cancellation, the probe head also serves as a
high-bandwidth actuator th~t is used to provide constant force stylus
measurement. This is achieved by servo controlling the extension of the probe

328
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

ase

-.......- -Capacitance
electrodes

'---Stylus

Figure 6.30: Schematic diagram of force probe

head to maintain constant cantilever deflection. The probe head in turn is to be

mounted to a low-bandwidth, long-range vertical slide that provides the travel
required by the large vertical profile deviations envisaged for its particular
application. For absolute measurements, the position and tilt' of the force probe
must be measured relative to a metrology frame by interferometric methods. For
this purpose, the rear surface of the probe is an optically flat reflector.

Adjustment

PZT actuator
Flexure
spring
Bottom plate

Force probe

Figure 6.31: Schematic diagram of flexure arrangement

329
FLEXURES

The force probe assembly is mounted to the probe head that consists of the
actuators, flexures and capacitance displacement sensors. The flexure consists of
two sets of three flexure elements (represented schematically in Figure 6.31 by
the springs). The lower set of flexures provides the compliance for the PZT
actuators to enable the required displacement and tilt. The upper set is used to
adjust the pr~load on the actuators during assembly and allow for slight
dimensional mismatch between the length of the flexure and that of the
actuators. The adjustment screws control the deflections of the more compliant
upper flexures that in turn control the pr~load on the actuators due to the
deflection of the lower flexures. Once the pre-load is set, the controlling stiffness
is that of the lower flexures. The deflections of the individual PZT actuators are
monitored by capacitance gages, which can be used in combination to measure
both displacement and tilt of the probe. Although the position of the force probe
is monitored by a remote interferometer, the capacitance gages provide a more
tightly coupled metrology loop that enhances the controllability of the entire
system. In addition, the present design allows for the operation of the device as
a stand-alone device to facilitate bench testing.
Figure 6.32 shows an exploded view of the complete assembly. The two sets
of rotationally symmetric flexures (section 6.4) are part of a monolithic structure
made out of aluminum, which also includes slots for the actuators. The flexures
are equi-spaced around the circumference of a circle giving the monolithic
flexure thr~fold symmetry. Three PZT actuator segments are mounted on the
top plate and pass through slots in the flexure. The actuators are made from a
single piece of ceramic and exploit the Poisson's ratio effect to produce the
desired fine motion. These actuators bear on the bottom plate via spherical
contact surfaces that allow for the angular tilt. Capacitance gages mounted in the
top plate monitor the displacement of each PZT with the bottom plate forming
the target for the capacitance gages. The force probe is mounted to the bottom of
the target mirror, which in turn is clamped to the lower surface of the bottom cap
using a semi-kinematic clamping arrangement (not shown). Changes in length of
the PZT actuators cause the bottom plate to tilt and translate relative to the top
plate, the latter of which has a large aperture to allow the laser beams of a remote
interferometer to pass through to the target mirror. While the aluminum
structure of the probe head is susceptible to thermal changes, the force probe
assembly and target mirror are constructed out of ZeroduirM and are directly
referenced to the metrology frame through the interferometer, thus minimizing
the thermal sensitivity of the main measurement loop. Monolithic construction
has been employed to minimize joint losses and hysteresis and to obtain a
compact, high stiffness structure.

330
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

Laser
aperture

PZT

Capacitance gage
target surface

Bottom plate

M;~
Force prob~ -:-
Stylus~

F igure 6.32: Exploded view of complete assembly

References
Aguis J.G., 1994, Constant velocity flexible coupling and connection with the
drive and driven members, USA patent US5299980.

331
FLEXURES

Badami V.G., Smith S.T. and Patterson S.R., 1996, A metrological three-axis
translator and its application for constant force profilometry, Proc. ASPE, 14,
391-395
Berliner Y.I. and Vikhman Y.L., 1976, The axial rigidity of of flexible elements of
compensator& with U- and n-shaped profiles of waves, Chemistry and
Petroleum Eng., 12, 31.
Brecht IV C., 1986, Predicting bellows response by numerical and theoretical
methods, Trans. ASME: ]. of Pressure Vessel Technology, 108,334-341
Cherukuri H., 1997, personal communication.
Chetwynd, D. G., and Smith, S. T., 1991, High precision surface profilometry:
From stylus to S1M, in From Instruments to Nanotechnology, ed. Gardner, J.
W., and Hingle, H. T., Gordon and Breach, London, 273-300.
Donnell L.H., 1932, The flexibility of corrugated pipes under longitudinal forces
and bending, ASME Trans. ]. Appl. Mech., APM-54-7, 69
Hamada M. et al., 1976, Design diagrams and formulae for U-shaped bellows,
Int. ]. of Pressure Vessel Technology, 4, 315
Hamada M. and Tsuda T., 1997, On design formulas of U-shaped bellows, Trans.
ASME: ]. of Pressure Vessel Technology, 119,127-131
Haringx J.A., 1952, Instability of bellows subjected to internal pressure, Phillips
Res. Rep., 7,189-196.
Howard, L. P., and Smith, S. T., 1994, A metrological constant force profiler, Rev.
Sci. Instrum., 65(4), 892-902.
Neale M., Needham P. and Horrell R., 1991, Couplings and Shaft Alignment,
Mechanical Engineering Publications, London
Juvinall R.C. and Marshek K.M., 1991, Fundamentals of Machine Component
Design, 2nd ed., John Wiley and Sons Inc., NY.
Levy R., 1962, Displacements of circular rings with normal loads, Proc. ASCE,].
Struct. Div., 88(ST1), 23-54. Equation (1) in this paper has been erroneously
typeset.
Linley P.M., 1976, Anti-backlash, self-aligning nut, USA patent US3977269.
Pilkey W.D., 1994, Formulas for Stress, Strain and Structural Matrices, John Wiley
and Sons, Inc, NY.
Sarid, D., 1994, Scanning force microscopy: with applications to electric, magnetic and
atomic forces, Oxford University Press, NY.
Seely F. B. and Smith J. 0., 1932, Advanced Mechanics ofMaterials, J. Wiley and
Sons, NY chpt. 15.
Shigley J.E., 1989, Mechanical Engineering Design, McGraw-Hill, NY.
Smith, S. T., and Howard, L. P., 1994, A precision, low force balance and its
application to atomic force probe characterization, Rev. Sci. Instrum., 65(4),
903-903
Timoshenko S.P. and Woinowski-Kreiger S., 1959, Theory of Plates and Shells,
McGraw-Hill Book Company, NY (NB Symbols for geometry factors of
equations (6.17) and (6.18) correspond to the form given on page 289)

332
CHAPTER 6: HINGES OF ROTATIONAL SYMMETRY

Timoshenko S.P., 1941, Strength of Materials: Pt. II Advanced Theory and Problems,
2nd ed., D. Van Nostrand, Inc., NY, chapter II.
Thomas, T. R., 1982, Rough Surfoce.s, Longman Press, London.
Wahl A.M., 1963, Mechanical Springs, 2nd ed., McGraw-Hill, NY.
Wilson J.P., 1984, Mechanics of bellows: a critical survey, Int. f. Mech. Sci.,
26(11/12), 593-605
Woody S.C. and Smith S.T., 1999, Design and assessment of disk couplings, to be
published.
Wolff P.H.W., 1951, The design of flexible disk misalignment couplings, Proc.
Jnst. Mech. Engrs., 169, 165-175
Young W.C., 1989, Roark's Formulas for Stress and Strain, McGraw-Hill Book Co.,
NY.

333
 7
7.0 Overview
Levers

For a broad range of flexure applications, a displacement actuator is necessan; to

provide the drive forces. For a number of practical reasons, the most common drives
are the feedscrew and piezoelectric actuators. Because these drives provide either long
motion with limited resolution or short ranges with continuous control, it is often
necessary to introduce some form of lever mechanism to amplifiJ or attenuate the
drive action. For example, displacements produced using micrometers often require
attenuation while piezoelectric drives may have insufficient range thereby requiring
motion amplification. Although the concept of a lever is simple, its practical
implementation involves a number of compromises. Two types of lever are identified,
these being the mechanical lever and soft-spring/stiff-spring attenuation. Following a
discussion on the types of lever used in flexures, this chapter assesses the
consequences and sotne issues. In particulnr, a method for analyzing the effects of
'lost motion', and influence on the dynamic performance are considered. A detailed
analysis of an attenuation mechanism is presented to illustrate some important
considerations in any flexure design, in tltis case, with emphasis on lost motion and
associated modeling errors. Manufacture of the flexure in the case study also
provides information relevant to discussions presented in chapter 8.
7.1 Introduction
Many flexure designs necessarily include some form of drive mechanism. Of all
type available, these can be split into two categories these being contact or non-
contact. Non-contact drives are electromagnetic, electrostatic or hydrostatic.
Contacting drives all consist of actuators designed to impose a force or strain in a
solid body that is some way connected between two points in the flexure.
Contacting drives can be motor driven micrometers, friction driven pushers,
piezoelectric, magneto-strictive, electro-strictive, hydraulic piston, pressurized
diaphragms, Poisson's ratio, shape memory, bimetallic and thermal expansion
(for a rather fun version of this latter drive see Snyder, 1993). Some of the
important attributes of these mechanisms are discussed in chapter 8 while the
interested reader is referred to Smith and Chetwynd, 1991 for a more detailed
discussion. Non-contact drives impose a force on the flexure without
significantly effecting stiffness or adding mass. Contact drives however can be
considered to add, often, significant stiffness at their point of application. In
FLEXURES

general, the added mass of the actuator can be lumped into the 'stationary' link.
For example, in the case of a micrometer the stationary link is the one to which
the micrometer is rigidly secured while the moving link is attached to the anvil
of the micrometer via some form of coupling. Consequently, the only additional
mass on the moving link is that due only to the rigidly attached component of
the coupling.
As mentioned in the summary, in many cases, either the range or resolution
of a particular drive may be inadequate for a given application. Under such
conditions, it is often necessary to introduce a lever mechanism either for the
amplification or attenuation of the motion from the actuator to the flexure.

7.2 Mechanical levers

In its simplest form, the mechanical lever consists of a rigid body connecting the
moving, or 'output', body to some reference, or 'base' link. Applying a
displacement in a specific direction on this rigid body results in relative motion
between the moving platform and base. Such a rigid lever mechanism is directly
analogous to the electrical transformer. An alternative lever (in this case
attenuation) effect can be achieved by imposing a compliant coupling between the
drive and flexure. This can be suitably modeled as actuating a stiff spring

a)
a b

b)
b

Figure 7.1: The simple, mechanical lever, a) pivot 2 between driver 1 and driven 3, b)
pivot outside of driver and driven

336
CHAPTER 7: LEVERS

through a soft one. This latter mechanism is directly analogous to the potential
divider commonly employed in electrical circuits. Clearly, the displacement at
the stiff spring will always be less than that of the actuator.
Before moving on, it should be noted that there are a large number of
alternative lever mechanisms. In particular, the feedscrew, which can be
considered a leveJ, probably represents one of the most common machine
elements. Other lever mechanisms such as friction drives and the wedge may be
advantageous for some applications, see for example Smith and Chetwynd, 1991.

7.2.1 Tbe rigid lever

In its simplest form, the lever can be modeled as a rigid rod rotating about a
pivot as shown in figure 7.1(a). Assuming that the.input is applied at 1, the lever
ratio, n, is defined as

output displacement
n=
input displacement
==-ab (7.1)

Alternatives to this mechanism can be obtained by interchanging the input

and output (in which case the lever ratio is inverted) or moving the pivot as
shown in figure 7.1(b). For this configuration the lever ratio becomes

a +b
n = - (7.2)
a

Fortunately,
there is no
restriction on the
shape of the lever
arm and in its
more general
form, the lever
can be
represented as
shown in figure
7.2(a). Lengths of
the lever arms are
readily obtained
based on the fact
that the vector of
rotation about the
Figure 7.2: Levers of different shape, a) the general lever, b) lever arm
pivot is always at
with orthogonal axes between driver and driven.
right angles to
the applied forces

337
FLEXURES

for both input and output. It is often convenient to arrange that the input and
output displacements are at right angles which, for small displacements, can be
achieved using the link of figure 7.2(b). For a perfect lever, and assuming small
displacements, the following may be deduced:

1. The work done at the input will be equal to that at the output
2. Assuming that the links are infinitely rigid, the angular rotation about
the pivot of both the input and output links are equal
3. In static equilibrium, the vector of all three forces forms a closed loop.
4. The lever mechanism is reversible. That is, transposing the driver and
driven inverts the lever ratio.
5. Motion between any two fixed points in a solid body must be in a
direction perpendicular to an axis passing through the two points. This
is theorem I of chapter 5.

From 7.2(a), it can be seen that, as the force vector is directed towards the
pivot point, the lever ratio will increase asymptotically. This can be exploited in
flexure mechanisms to achieve large amplifications from small displacements. In
many applications, because of their smooth, continuous action, piezo-electric
actuators are chosen. Because of their limited displacement range, a large variety
of amplifying mechanisms have been implemented, see for example Furukawa et
al., 1990, 1991, 1994, Xu and King, 1996. In the following, it will be assumed that
piezoelectric actuators requiring amplification are to be employed. Although this
helps to illustrate some important design details, such an assumption is not
necessary and driver and driven can be reversed in all of the levers to produce a
reduction mechanism.
Figure 7.3 shows three piezo-electrically driven amplifying elements. The
first of these figures represents what might be considered a conventional lever. In
this figure, however, some further design details have been added. The hinges
that constitute the coupling and pivot are included. Assuming that the applied
force at the output acts in the direction shown it is obvious that the coupling will
experience a compressive load while the pivot will be in tension. A notch hinge is
chosen because of its relatively high compressive strength and resistance to
buckling. For the pivot, a leaf type spring is considered adequate. It has already
been shown in chapter 4 that the tensile force will induce an increase in the
rotational stiffness of the flexure. From a simple bending moment analysis it can
be shown that the forces in the pivot and coupling are

b
F2 =- F3 = (n - 1)F3
a (7.3)
F; = F2 + F3 = nF3

338
CHAPTER 7: LEVERS

a) a b
,· - ·1
~--- t--
• - -----i•l

b)

Figure 7.3: Three, piezo-electrically driven

amplifying elements, a) the direct lever, b)
indirect, two-bar lever for linear
amplification, c) indirect lever for linear or
angular output amplification

The subscripts 1 and 2 refer to the coupling and pivot respectively while the
lever ratio n is given by

a+b
(7.4)
a

It is clear that the forces in the pivot increase in proportion to the lever ratio.
Additionally, there will be a corresponding increase in bending moment about
the lever arm. Bending of the lever arm will contribute to lost motion in the
mechanism.
A more complex two-arm lever arrangement is shown in figure 7.3(b). In
this, the lever ratio can be envisaged as that due to the rotation of an arm fixed to
the pivot due to the forces from another free arm connected at some angle. In the
particular example shown, it is obvious that the output displacement is due to

339
FLEXURES

the motion of the midpoint necessary as the spacing between opposite ends of
the two rods changes. An equivalent mechanism with a piezo-electric drive is
shown in figure 7.3(c). This last, and probably most common, mechanism can be
modeled as shown in figure 7.4. Again, point 1 represents the input, 2 the pivot
and point 3 is the output. In this particular model, the input displacement is
parallel to the link connecting the pivot to the output.

x3

'
pivot

a)

' ~~ ~----------------------~
)6 lo
I '
'
b) 3~3
1
X3

1
2 I '
1 XI '

Figure 7.4: Parameters for a mathematical model of the two-arm levers shown in figures 7.3(b
& c), a) mathematical model, b) displacement diagram

By far the simplest way to determine a relationship between the input and
output is to use a 'displacement' diagram. The interested reader will find an
alternative approach adopted by Furukawa and his coworkers. This is analogous
to velocity diagrams often used to determine the instantaneous velocity in
complex mechanisms. Obviously over short time intervals the instantaneous
velocities and displacements will be directly proportional. Consequently, it is
possible to construct a displacement diagram in the same way as one would a
velocity diagram. The first step in the construction of this diagram is to define a
'stationary' point that coincides with the instantaneous position of the pivot, 2.
The motion of the actuator, x 1, relative to the pivot can be immediately plotted.
To determine the motion of point 3, the general directions of relative motions can
be plotted. Firstly, it is known that the point 3 moves vertically with respect to
the pivot Consequently, a vertical line passing through point 2 is drawn on the
diagram. Finally, because only rigid links are being considered, the motion of
point 3 relative to point 1 is perpendicular to the link /1 • Consequently, the

340
CHAPTER 7: LEVERS

displacement line passing through point 1 can be drawn as shown. Clearly,

because point 3 occupies a unique position in space (i.e. it can only be in one
place) this must be located at the point where the two lines intersect. The length
of the line connecting points 1 and 3 represents the relative displacement due to
rotation of the link /1 • FTom this displacement diagram, provided that the small
displacement assumption is valid, a number of simple input-output relationships
may be readily derived.
Firstly, it is necessary to introduce some geometric relationships provided by
continuity considerations. These are

(7.5)

'·sine. =h (7.6)

Combining (7.5) and (7.6) to eliminate 01 gives

(7.7)

From the displacement diagram, the lever ratio represented by the

relationship between input and output displacement is given by

xJ
x1
=n=-1-=ctnO, =!,}-(h)2
tan01 h /1
(7.8)

Clearly, for small angles, or as h reduces, the lever ratio can take on very
large values. Alternatively, as 8 1 tends towardsn/2 the lever ratio tends towards
zero. The change in the ratio as a function of the initial angle can be simply
obtained by differentiating (7.8) to give

(7.9)

This may be considered a measure of the non-linearity of the ratio, and

therefore, of the lever. This shows that, as the lever ratio increases linearly with a
reduction in h, the non-linearity will increase as the inverse square.
It is interesting to note that this ratio is independent of the length of the link
connected to the pivot. However, under some circumstances it might be
desirable to provide a large rotation from a small displacement. Such an action
can also be provided due to the rotation at the pivot of the mechanism in figure

341
FLEXURES

7.4. Rotation of the other link is rarely of interest. In this instance, the lever ratio
(having dimensions of rad m·l) is given by

(7.10) 0

The sensitivity of this ratio as a function of the angle is given by

(7.11)

Extension of the
above analysis to the

__
1
general, two-arm lever
involves little
additional complexity.
In this case, the input
can be directed in an
arbitrary direction. The
modified model for the
generalized two-arm
lever is shown in figure
7.5. Based on the
displacement diagram,
the lever ratio is readily
obtained as
Figure 7.5: Parameters for a mathematical model of the
generalized, two-ann lever, a) mathematical model, b) x3 • B cosB2
displacement diagram - = Stn 2 +--
X1 tanB1

(7.12)

In terms of the output rotation, 83 , the ratio becomes

- I ( sm
fJ3= - . fJ +cos
-B-
2)
(7.13)
2
x1 /2 tanB1

For a particular class of two-arm levers, it is arranged that the actuator action
is co-linear with the axis of link /1 (i.e. 81 = 8 2 ). In this particular case, the lever
ratio becomes

342
CHAPTER 7: LEVERS

(7.14)
x1 sin01

Again, a measure of the linearity of this ratio can be obtained from the
derivative of (7.14) given by

(7.15)

It can be seen that, as the

- - angle reduces, non-linear
effects increase as the inverse
square and as a consequence
may significantly reduce the
available displacement range.
Figure 7.6 illustrates a two
bar lever that is equivalent to
that of figu.re 7.5 with
()
actuator different angles between the
links. The displacement
diagram is included from
which the lever ratio can be
b)
similarly determined as
above.

7.2.2 Soft spring- stiff spring

attenuation
Figure 7.6: Alternative arrangement of the two bar lever,
Assuming that all pivots are
a) equivalent geometric models, b) displacement diagram
infinitely rigid, all of the
above mechanisms can be reversed to operate as either a displacement
magnifying or attenuating mechanism. In some instances it may be necessary to
introduce a large attenuation. In particular, relatively precise and inexpensive
feedscrews can be obtained in the form of a micrometer spindles and precision
ground ball screws. The resolution of these drives will always be limited by their
size (determines manufacturing process), stiffness (pre-load forces introduce
hysteresis in the form of oblique frictional tractions, Mindlin and Deresiewicz,
1953, or rolling resistance, Johnson, 1985) and manufacturing tolerances
(continuously improving). A relatively simple way to increase the resolution of a
long-range drive mechanism is to use a reduction mechanism. Many of the
mechanisms outlined in the previous section are adequate for such a task.
However, for high precision applications, the inherent non-linearity of these
devices will have to be dealt with. An alternative is to attenuate the input using

343
FLEXURES

the soft spring - stiff spring concept illustrated in figure 7.7. Those readers
familiar with electrical circuits will recognize this as analogous to a potential
divider. Again, it is assumed that this mechanism is driven by a prescribed
displacement, x1, and this imposes a force, F, on an intermediate spring, k1, and
this same force is transmitted to the displacement flexure of stiffness k 2 • The
subsequent displacement of the flexure, x3 , is given by

(7.16)
x k
- 3 = --1
x, k, + k3

Use of this attenuation mechanism has the relative advantages that

1. It remains linear over relatively large displacement ranges

2. Extremely large attenuation ratios are possible.
3. A highly compliant flexure can also be used to provide precisely
controlled, small forces (see case study III).

The only drawback to such a lever is

1. The high compliance of the intermediate spring between the

drive and moving stage adds little to the dynamic performance
while the relative advantages of coupling to a stiff drive are also
lost.
A planar mechanism utilizing such a lever reduction is shown in figure 7.8.
Other implementations are left to the designer's imagination. An interesting
example can be found in the paper of Hart, 1968. In this paper, such a reduction
mechanism is used to drive a single crystal silicon flexure with precise control of
motion at sub-nanometer levels suing a simple micrometer. In this particular
case, a reduction ratio of around 10,000:1 is achieved. Included in this paper is a
novel method for obtaining fine angular adjustment using hanging chains to
provide a torque.

Figure 7.7: 4tput/output model for the soft

spring, hard spring attenuation mechanism

344
CHAPTER 7: LEVERS

Figure 7.8: A planar mechanism utilizing a soft-spring, hard-spring attenuation

mechanism

Torsion bars may provide a simple high compliance intermediate flexure,

particularly for mechanisms requiring fine angular positioning. In its simplest
form this consists of a long, thin rod attached to a rotary drive at one end and the
flexure at the other. Figure 7.9 shows a simple, high-resolution angular
adjustment mechanism. In this case, the stiff spring is achieved by using a
similar, but larger diameter (and/ or shorter length) rod. Stiffness values for each
of the rods, k, can be readily computed from the torsion equation (2.100) while
the lever ratio is easily shown to be

(7.17)

Sometimes referred to as a filar suspension, examples of these can be found

in the papers of Marsh, 1961, and case study II at the end of this chapter.

7.3 Lost motion

Piezoelectric or magnetoI electro-
strictive actuators are frequently used
to drive flexures requiring translations
ranging from tens of micrometers or
more. A compact form of actuator
consists of a stack of actuators
connected in series. Unfortunately, for
such a large range, the physical
dimensions and number of elements in
Figure 7.9: A simple angular adjustment the stack become unwieldy. For a more
mechanism compact device, considerable space can
often be saved using a lever to amplify
the motion. To appreciate the models of this mechanism, consider the simple

345
FLEXURES

notch type linear spring driven by a piezoelectric element as shown in figure

7.10. Three different configurations are shown. All three designs utilize one of
the support legs as a lever arm. In figure 7.10(a), leverage is achieved by pushing
the support leg near the stationary pivot. Analogously, it is possible to close a
door by pushing near to the hinge without having to move ones hand a large
distance (although considerably more force is required). For an infinitely rigid
coupling between the actuator and support leg and also assuming that the
notches act as perfect hinges, then the ratio, n, between displacement of the
actuator and upper platform is simply

b+a
n= - (7.18)
a

b)

Figure 7.10: A simple linear

spring driven by a
piezoelectric element via a
lever, a) direct leverage on
the support, b) alternative
hinge design, c) alternative
c) a lever design

Intuitively, it is apparent that there will be a large sideways, or lateral, force

on the flexure. As a consequence it might be expected that this might introduce

346
CHAPTER 7: LEVERS

unacceptable distortion due to the compliance in this direction given by equation

(4.109). A alternative design is shown in figure 7.10(b) in which the actuator will
induce forces in the longitudinal direction of the notch. The benefit of this
alternative design can be seen from ratio between lateral and longitudinal
stiffness values given in equations (4.109) and (4.111) given by

r7i)
k6 P.
-k 6,F,
- -
..
-
8/12 (1 - 2'/3
----;.-,--~
9n 2 (l - /1 2 ) (7.19)

,...,. Equation (7.19) is

~ 0.004

g
plotted in figure 7.11.
0.003 From this, it can be seen
0
that, over a typical range
"-+=! 0.002 of values for the
~
Cll
0.001 dimensionless ratio p,
~
the lateral stiffness is
~ 0
0 0.05 0.1 0.15 0.2 0.25 0.3
more than two orders of
magnitude lower than
that in the longitudinal
p direction. It will be
Figure 7.11 : The ratio oflateral to longitudinal stiffness for a shown shortly that
notch hinge
compliance of this pivot
is responsible for most of
the lost motion within a lever system.
Finally, figure 7.10c) shows an alternative lever mechanism in which the
actuator has been moved 'outside' of the flexure. In this case the lever ratio is
given as n = bja. Such a design is often favored when it is desired to use the space
'within' the flexure.
To assess the effects of coupling and pivot compliance on the lever ratio, it is
possible to model the single degree of freedom system as shown in figure 7.12.
In this, the flexure is modeled having the same stiffness, k 3, that it would have
were it driven directly at the moving platform while the pivot and coupling are
identified by the subscripts 2 and 1 respectively. For a perfect pivot and
coupling, the relationship between a force applied at the coupling to the
displacement at the moving platform can be obtained by equating the work done
when a force is applied at the moving platform to that when a force is applied at
the lever. From this, the following is readily obtained

(7.20)

347
FLEXURES

a)

m.

F2,x2
0
Base
"'· piVOt
kJ
F3 ,x3

Base

Figure 7.12: Mathematical model of a lever driven flexure

mechanism, a) pivot under compression for an expansion of the
actuator, b) expansion of the actuator induces tensile force

Clearly, the stiffness of the flexure as seen at the coupling varies in direct
proportion to the lever ratio. Analysis is not as straightforward when the
compliance at the coupling and pivot are included.
The objective of the lever is to provide some form of amplification or
attenuation between the force or displacement at the actuator and that produced
at the moving stage of the flexure. In terms of the model of figure 7.12 this
consists of the desired ratio x 3 I x0 • From Newton's 3rd law it is known that, for
the flexure to be in static equilibrium, the vector sum of forces must vanish i.e.

348
CHAPTER' ? : LEVERS

(7.21)

It can also be seen that the spring forces applied to the flexure are

F1 =k 1(x1 - x 0 )
F2 =k2x2 (7.22)
F3 =k3 x3

Ignoring the angular stiffness of the springs, taking moments about the pivot
gives

(7.23}

Finally, because the flexure is assumed to be rigid, it is possible to derive two

equal expressions for the angular deflection

() = x1 - x2 = x2 - x3 (7.24)
a b

An immediate outcome of equations (7.21-7.23) is that there will be a

displacement in the pivot of magnitude

F;(l + 1/ n)
xl = k2
(7.25)

Substitution of equations (7.21 - 7.23) into (7.24) gives

(7.26)

From the first of (7.22), the actuator displacement is given by

(7.27)

Finally, the relationship between output and input displacements is given by

349
FLEXURES

(7.28)

It is informative to study some of the characteristics of this expression. To

further simplify (7.27) the dimensionless stiffness ratios are replaced by the
algebraic symbols a and p to yield

x3 -n
(7.29)
x, = (n'a +n'p(r + ;;)' + r]
Where a represents the stiffness ratio between the flexure and coupling and
p the stiffness ratio between the flexure and pivot.
The fractional lost motion, f, is defined by

(7.30)

As either the flexure stiffness tends to zero or both the coupling and pivot
stiffness tend to infinity, the lever ratio tends towards the ideal value -n. For an
ideal pivot, p =:) 0, equation (7.29) reduces to

x3 - n
(7.31)
Xo = (n a + 1)
2

while the fractional lost motion becomes

(7.32)

Both of the above can be readily verified using a simplified static model.
For the case of a perfect coupling

350
CHAPTER 7: LEVERS

flO

&:1
0.8 f15 -
0
'll
flO
a
0

til
0.6 ~ -
..9
Ol f6 _
~ 0.4

1 0.2
f4
fl

0 f1 ~

0 0.002 0.004 0.006 0.008 0.01 0.012

a=p
Figure 7.13: Fractional lost motion as a function of the ratios ofstiffuess
between flexure and pivot. The stiffuess of the actuator coupling has been made
equal to that of the pivot. In the legend fl represents the fractional lost motion
for a flexure mechanism with a nominal lever n =2.

& (7.33)

J ~ n'~l+H
(•'~I+;;)' + 1)
In both cases it can be seen that the fractional lost motion can only be
reduced by minimizing the products n 2 a or n 2 f3. Assuming that the stillness of
the pivot and couplings are identical, figure 7.13. For example, if an ideal lever
ratio of 10 is assumed and 20 % lost motion is considered acceptable, the stiffness
ratios for both pivot and coupling must be below 0.0012.
Surprisingly, the effect of cascading levers requires little additional
complexity. At the outset, the stiffness of the flexure, k3 , was defined as the force
required to produce a displacement. For a lever system, this is in direct
proportion to the ideal lever ratio and independent of the stiffness of the pivot.
Consequently, to determine the lost motion of the first lever, it is adequate to
replace the stiffness of the new flexure to include that of the second lever.
However, the output then becomes the displacement of the second lever.

351
FLEXURES

Fortunately, this second lever can be modeled as a separate lever having an

ideally stiff coupling. So, if two levers are cascaded and the first and second have
ideal lever ratio's n1 and ~ the total fractional lost motion becomes

(7.34)

If the stiffness ratios are the same as the previous analysis and the new lever
ratios are split and equal to the square root of the ratio for a single lever (7.34)
becomes

(7.35)

Figure 7.14 shows the ratio of the fractional lost motion for a single lever of
ideal ratio's 10 and 100
1000 and two levers of ratio
..flO and 10. As the
100 stiffness ratios a and fJ
are reduced, it is clear
~
..:::: that a cascaded pair
10 nl O represents COJ1Siderable
improvement. However,
- -- - - - nlOO
this advantage
1
0 0.002 0.004 0.006 0.008 0.01 0.01 ~
diminishes as the ratio
increases.
a= J3

Figure 7.14: Ratio of fractional lost motion between a

single lever and two levers in series

7.4 Effects of levers on flexure dynamics (influence of the drive)

In previous discussions, the influence of the driver upon the dynamic
performance of the·flexure has been ignored. In practice, with devices such as the
feedscrew or piezo-electric ac~ators, the added stiffness of these drives can
dominate over that of the flexure. However, the necessity of an intermediate

352
CHAPTER 7: LEVERS

coupling often results in a considerable reduction of the drive stiffness as 'seen'

at the flexure. As mentioned in section 5.1.4, this drive stiffness will add to the
potential energy storage and therefore to the effective stiffness of the flexure.
This section considers the influence of an intermediate lever on the dynamic
response of the flexure.

a)

Figure 7.15: Mathematical

X~;--_m_ __,
models for levered simple
linear spring mechanism;
a) simple notch type linear
b)
spring directly driven at the
moving platform, b) simple
linear spring driven via a
lever mechanism (solid
view and mathematical
a
model

To compare the influence of the lever on the dynamic response of the

mechanism, it is necessary to determine the natural frequency of a mechanism
directly driven. Consider the simple, notch type linear spring show in figure
7.15(a) in which the drive is directly attached to the moving platform via a
coupling of effective stiffness kc . Assuming the hinges to act at a point, it is
relatively straightforward to show that the potential and kinetic energy of such a
system is given by

(7.36)

353
FLEXURES

The second moment of mass for each of the support legs corresponds to that
for rotation about the stationary pivots. Lagrange's equation readily provides the
equation governing motion for a prescribed input displacement, x1

(M + bl2/)·· + -r +
X3
(4k8,M, k c ) X3 = k <XI (7.37)

From the homogeneous solution, the natural frequency is

(1)2 =
n (7.38)

Clearly the coupling stiffness adds to the natural frequency of the flexure as
mentioned in section 5.1.4 (see also equation 5.46)). Turning our attention now to
levered mechanisms, of which three examples are shown in figure 7.10, the
kinetic and potential energy equations become, see figure 7.15(b)

2 2 (7.39)
I XJ I XJ
V = -2 kc X --
I n )
+-(4k
2 )-
e,M, b2
(

Lagrange's equation immediately gives

(M + -2/,b2 )··x +-kn. (X-n3- x )+ (4ke,M, )X-b23 -_ Q

J 1
(7.40)

Rearranging equation (7.40) and adding energy dissipation, the equation

governing motion can be written in the form

(7.41)

To determine the frequency response of this levered mechanism a prescribed

input displacement is assumed to be of the form

(7.42)

Substituting (7.42) into (7.41), the general frequency response can be readily
obtained as

354
CHAPTER 7: LEVERS

(7.43)

Although containing different terms, equation (7.43) is of the same form as

that of a single stage vibration isolator discussed in section 3.6. From this
equation, the undamped natural frequency of the lever driven mechanism is
given by

(7.44)

Comparison between equations (7.38) and (7.44) readily shows the influence
of the lever on the dynamic performance of the mechanism. For an amplifying
mechanism (i.e. n(= b / a)> l) there are two deleterious effects. Firstly, and
probably most importantly, the effective stiffness of the coupling is reduced by
the inverse square of the coupling. For ratio's greater than one, this will always
result in a reduction of the natural frequency of the flexure and this is
compounded by the increased inertia of the lever mechanism itself. Figure 7.16(a)
over leaf shows the magnitude frequency responses of a levered flexure system.
At high frequencies, the response of the system is limited by the inertial term
in equation (7.41). To assess the inertial effect on the drive, it is interesting to look
at the relationship between the drive displacement and that produced at the
input to the lever. In this case (7.41) can be rewritten in the form

(7.45)

From this equation, it can be seen that the effective mass of the flexure as
'seen' by the drive increases with the square of the lever ratio.
It can be readily shown that the dynamic 'lost motion' is given by

355
FLEXURES

(7.46)

a)

{\
25

20 I\
15 I \
10 1--
/ \
5 ~
0
0 2 3 4 5

b)

1.5

1
g
·c 0.5
a
0

t; 0
.9 4 5
l -0.5
·c0
l -1

-1.5

-2

%n
Figure 7.16: Dynamic response of a levered flexure, a) frequency response, b)
=
fractivnal lost motion; kc = 100 N m·', k fl•xure 0.1 Nm·1, M = 1 kg, b12 = 3
1
Nsm·', b3 =0.1 Nsm· ,n= 10

It is informative to consider the above equation in terms of three major

regions, these being low frequency, near resonance and high frequency. Each of
these regions corresponds to stiffness, damping and mass dominated regimes
respectively. For the first region, the 'static' lost motion can be obtained by
setting a> :::) 0 at which the above reduces to

356
CHAPTER 7: LEVERS

f = n2K, -kc
n 2K 1
(7.47)
n24koM
= ;t :t

n24k9,M, +blkc

Being identical to equation (7.32), it can be seen that this tends to zero with
large coupling or low flexure stiffness.
At high frequencies, the mass term will dominate causing the right term of
(7.46) to tend to zero at which the fractional lost motion goes to 1.
The intermediate frequency near to resonance, at which damping has the
dominant influence, is a little more complicated. Figure 7.16(b) shows the
fractional lost motion corresponding to the frequency response plot of figure
7.16(a). Clearly, the choice of damping, as always, is a compromise. Such
compromises are left to the reader.

7.5 Case study 1: A fine adjustment mechanism for motion attenuation

In the design of an instrument for the measurement of hardness indentations, it
was required that a small displacement sensor be positioned alongside the
indenter to measure its depth of penetration. Figure 7.17(a) shows a solid view of
the lever mechanism designed to enable fine adjustment of the sensor so that its
zero point could be positioned adjacent to the tip of the hardness indenter.
Figures 7.17 (b & c) are engineering drawings of a lever design indicating its
dimensions. In the undistorted condition, the lever consists of a rigid bar with
two leaf-type springs attached to its corners at one end. These springs comprise
the lever pivot while it is the larger linear flexure that provides the motion of the
sensor (mounted on the platform) relative to the base. A line diagram of the lever
mechanism in the distorted condition is shown in figure 7.18. In fact, the lever
mechanism in this design corresponds to the two-beam two degree of freedom
flexure already discussed in section 4.4.2. In this case, it is the axial rigidity of the
beams that is being exploited. One of the main design features is that the output
motion of the stage is perpendicular to the lever adjustment enabling manual
adjustment from, in this instrument, an accessible direction that does not
obstruct the sensor or mounting mechanisms.
To translate the platform relative to the base, a force is applied to the lever at
a height d as measured from the pivot point of the flexures, figure 7.18. In the
distorted condition the lever flexures will provide an equal and opposite axial
force to the base and moving platform. Idealistically, the lever flexure might be
modeled as two ' perfect' pivots. Based on this assumption, the lever ratio, n, can
be approximated from the equation

n=
dispalcement of platform
displacement of lever
=-ds (7.48)

357
FLEXURES

where the distanced based on the analysis of section 4.1.1 is measured from a
point 1/3 of the length of the cantilever from the base to the point of application
of the force and s is the separation of the neutral axes of the lever flexures.

a)

b)
---,
2 0.2
r 12 I [
I
I r-

t--•- t--•-
13.9 8
16.1
,....-
13.9

r0-1 f __[A

..-- I
12 -r
f
1 - - - - - - - - - 1 6- - -- - - - - - 1
figure continues over

358
CHAPTER 7: LEVERS

r
8
c)

~3r-~·-
5.2
8

L
Figure 7.17(cont'd): Lever mechanism for
fine adjustment of hardness penetration depth
2 1- 1-0.1
sensing probe, a) solid view, b) major
•
-
dimensions, c) exploded view of central
2 :.- o.8 portion showing dimensions oflever
j__ _c Ol

t
For this design, a total deflection of the platform of 40-50 J.liil was required.
To enhance the resolution of the screw adjuster, a lever ratio of 1:6 between input
and output was considered adequate. In the final design, this flexure was
manufactured using wire-electro-discharge machining, see chapter 8, from a 12
mm thick plate of bronze. Some of the relevant parameters for this design are
given in Table 7.5.1 below. Equation (7.48) immediately gives an approximate
lever ratio of 1:6.37. This is slightly larger than the desired value.

7.5.1 Finite element analysis of the lever flexure

The objective of this study is to assess the fidelity of this design. In this case the
two parameters of interest are
1. Stiffness of the upper platform
2. Rectilinearity of motion of the lower face of the moving platform
Stiffness implies both dynamic response and immunity to vibration. Usually,
the dynamic response is limited by the fundamental, or gravest, mode, natural
frequency. Invariably this fundamental mode is proportional to the square root
of the most compliant relative motion in the mechanism, this being the desired
freedom. Consequently, the fastest response corresponds to the highest flexure
stiffness that, in turn, is limited only by the maximum stress tolerated by the
material of construction (where possible, minimizing mass by removing material
where stresses are small and maximizing the ratio of elastic modulus to density
should also be considered).
Initially, a design based on maximum permissible stresses is analyzed as an
example of a flexure to provide a fast dynamic response. Based on this simple
analysis, it is shown that the design will contain significant lost motion and
undesirable distortions of the structural components (i.e. the 'rigid' links) of the
flexure.

359
FLEXURES

Base

:r I Moving platform

t---~----- -------- - ---- --------

Figure 7.18: Line diagram of the distorted lever mechanism
Validity of the assumption that it is feasible to use simple beam bending
equations can be evaluated from the dimensionless axial load parameter y of
section 4.1. In general, the simple beam bending equation is usually adequate if y
significantly less than one. This parameter depends on the axial force applied to
the beam, which, in this design, corresponds to the force required to displace the
translation spring. The stiffness of the translation spring is given by

24
k = El = 3.36x106 (Nm-1) (7.49)
1 £3

Based on this, the applied force for 40 J.Lin output deflection of is 134 N. Being
the axial load applied to the lever flexures, the dimensionless load parameter is

r =!...2'{ii
fF: =1.02 (7.50)

360
CHAPTER 7: LEVERS

b)

a)
d

Moving
platform

Figure 7.19: Mathematical model oflever mechanism a) complete model, b) reduced form to
be equivalent to that in figure 7.13

Having a value of greater than 1, this design is uncomfortably near to the

buckling load and may be unstable at the extreme extension of the platform.
Ignoring this potential error, the axial stiffness of each lever cantilever is

k EA
tJJtlal/tvcr =-~- =7.8 X 101 (N m-1) (7.51)

Although this is significantly higher than that of the linear translation spring,
there are other errors due to the compliance of the lever acting as a linear spring,
k unwltvcr, and the cantilever displacement of the lever itself, k 1,vcrcant • For these

deformation modes the stiffness values are respectively

24£/
k llntarltvcr =f =390000
(Nm-1) (7.52)
k lowrcant = 3EJ 3 =3.0 x l0 6

(d - 21 / 3)

361
FLEXURES

Based on these three stiffness values, it is possible to develop a model of the

lever as shown in figure 7.19. Rather surprisingly, it can be seen that compliance in
the lever can be added to that of the coupling.
In fact, distortion of the base also adds to the linear distortion of the lever,
this is ignored in the following analysis. These compliant elements are connected
in series for which the equivalent stiffness is given by

kl = - - - - -1- = 345112 (Nm·t) (7.53)

- - - +- -
klln•arl•••rr kIIWII'Ullll

For computation of lost motion, the above stiffness values can be used to
compute the ratios p = 4.13x10·2 and a = 9.33. Assuming an infinitely stiff
coupling between the drive and the lever, the fractional lost motion is given by
(7.29)

(7.54)

From this, it is expected that the lever ratio will change from 1:6.4 to 1:8.2
corresponding to an increased attenuation. It is also necessary to estimate the
stresses in the two flexures at the maximum deflection of 40 J.Ull· For the linear
flexure, this will occur at the root of the flexures and is approximated by (2.69)
and (4.78)

u = Mt = 6Etx4 = 200 (MPa) (7.55)

2
/2 2£

This is reasonably far from the yield stress for this particular material
although some care should be exercised to reduce the effects of stress
concentrations at the roots of the flexure by incorporating a fillet radius equal to
the thickness of the flexure. For such dimensions the stress concentration factor is
likely to add a further 10% or less to the stress computed from the bending
equation.
For the lever flexure, stresses will be comprised fron:t axial forces and
bending of the cantilever. Bending stresses may reasonably be computed by
considering each flexure to distort as the sum of a simple cantilever with a
bending moment applied at the free end plus linear spring mode deformation
due to the applied force. The maximum angular deflection of the lever, emu ' is
computed from

362
CHAPTER 7: LEVERS

(rad) (7.56)

Table 7.5.1: Material property and geometric parameters for the levered
linear sprmg
· mecharus ·m
MATERIAL PROPERTIES
E (GPa) 130
ar (MPa) 600
General
Depth b (mm) 12
Lever parameters
Thickness t (m.m) 0.1
Length l (mm) 2
0.9
Separations (m.m)
Force offset d (m.m) 6.4
Lever ratio n 1:6.37
=0.157
Translation spring parameters
Length L (mm_l 8
Thickness t (mm) 0.82

The axial stress turns out to be 112 MPa while the bending stress in the lever
is approximated by the equation

Mt Et
a = - = -B
12 21
=145 (MPa) (7.57)

Assuming the lost motion is almost entirely due to 's' shaped distortion of
the lever springs, the bending stress associated with this deformation mode can
be approximated from

3
a = Et
2
x 3 (!n - ~) = 725 (MPa) (7.58)
1 n

Because the axial stress may be assumed uniformly distributed, the worst
case stress is simply the sum of axial and bending stresses and has a value of 982
MPa which is above the yield stress of the material. Clearly the lever flexures are
likely to yield before buckling, a disaster either way. A number of design

363
FLEXURES

modifications are
a) 1.0&05 possible to reduce
~==~~----~--­ this stress. Because
1
-1.06-05 a-----~- "~:----;ro----
the axial stiffness of
-3.06-05 +--------~.......-""' -----
" """ IU

the lever flexures

.,
"';; -5.0&05 -f-----------~----
.0 "'\.
'\.
are adequate, one
~ -7.06-05 +---------~-- solution is to
~ -9.06-05 +-- - - - - - - -- - - "'--'1.---- reduce the
thickness of the
-1.16-04 +-- - -- - - - -- -"--T-
legs of the linear
·1.36-04 '\
translation flexure
distance from upper face [mm] to 0.4 mm for
which a stress of
400 MPa is
b) 3.5E-04
predicted.
3.0E-04
To visualize
l- 2.5E-04 the deflections and

I
2.0E-04 assess the fidelity
1.5E-04
of the above
"' 1.0E-04 calculations finite
l S.OE-05
element methods
have been used to
O.OE+OO analyze this
0 5 10 15 mechanism.
distance along lower face [mm] Additionally, these
flexures have been
c) manufactured and
measured using a
t~~~~:s~=:::=::==:;;=~
1.0E-05
-1.0E·05 vision-based,
~ -3.0E-05 f-------~------ coordinate
measuring machine
~ :~:~::~: i-- - - - - - - ---'1.-- - - to assess the
t:;::l -9.0E-05 +-- - -- - - - - ---'lo.----- accuracy of a
-8 -1.1E-04 of-----------~..--- commercial wire-
-1.3E·04 electro-discharge-
-1.5E·04 machine. This will
distance from upper face [mm] be discussed in the
following section.
Figure 7.20: Distortions of outer faces of the lever, a) distortion From the finite
of the left face, b) distortion of the lower edge showing tilt, c) element analysis it
distortion of the right face is possible to
predict the
distortions around the mechanism. To illustrate some of the previous discussions

364
CHAPTER 7: LEVERS

presented in this and the preceding chapters, distortions along three edges are
shown. These correspond to; horizontal distortion of the left hand side of the
vertical edge on the base, the parasitic displacement of the lower face and
distortion of the vertical edge of the outer face of the of the moving platform in
the direction perpendicular to the motion. These plots correspond to figures
7.20(a-c) respectively. The first and last of these figures show similar distortion
corresponding to a net rotation primarily caused by a net rotation of the base
support. Such a conclusion was made from a series of tests, not presented here,
in which tilt was measured while the thickness value of both the left and right
hand side supports were varied. This bending of the base support results in a
displacement at the lower face of approximately 0.1 J.Lm corresponding to around
4% of the output motion. Subtracting this tilt from these plots reveals distortion
of the members of a shape similar to those analyzed in section 5.2. A simple
calculation indicates a net rotation of around 8.7 J.Lrad for 2.5 J.1ID output motion.
Over 50 J.Lm this would correspond to 170 J.Lrad. This rotation induces an
additional displacement of the lever relative to the base. Consequently, this will
result in additional lost motion thereby increasing the predicted attenuation from
the previous value of 1:8.16. Figure 7.20(b) shows the parasitic motion of the
lower surface. This has been computed from the vertical displacement relative to
the point B that is used as the output displacement, which has a value of 2.5 J.1ID
in this analysis. Tut of this member is approximately 2 1/• times worse and at the
extreme end contributes around 12% of the total motion. In fact, for this
mechanism, the lever ratio between the displacement of the point B in figure
7.17(a) and the lever was found to be 1:9.54. Clearly, for this rather poor design
there are a number of significant factors influencing the mechanism behavior that
can be summarized as follows

1. Bending of the base will result in additional lost motion and unwanted
rotations
2. Distortions of the supports also result in lost motion and parasitic
rotations
3. Relatively large axial forces experienced by the lever flexures could
result in buckling
4. Large lateral forces induce significant linear spring mode distortions
that result in lost motion plus some aClditional stresses not accounted
for in the above models
5. These large forces produce combined loads in the flexures that are
likely to result in significant departure from behavior predicted from
simple beam theories.

All of the above problems are significantly influenced by the relatively high
stiffness of the linear spring. Reducing the thickness of the linear translation
flexure will correspondingly reduce the stresses and forces in the mechanism. To

365
FLEXURES

assess this, flexures of thickness 0.6 and 0.4 nun have been modeled using finite
elements. The lost motion predicted by these models is shown in figure 7.21. This
plots theoretical predictions of the lever ratio are computed from the
displacement of the midpoint of the lower edge of the moving platform to that at
the lever with results normalized to an ideal lever of ratio 1:6.37 and the results
from three different finite element models. Because the effect of finite stiffness of
the base support is not included in the theoretical model, an increased lost
motion, as observed with the finite element analysis, is expected. Additionally,
the effects of stress concentration are not included in the model. For the relatively
small lost-motion, when the thickness of the linear translation spring is 0.4 nun,
the theoretical and finite element models correspond reasonably well.

1.1
g
a
'::;2

t;; 0.9
,g

1 0.8

0.7
g~
0.6
0 0.2 0.<4 0.6 0.8
flexure thickness (mm)

Figure 7.21: Lost motion of the lever mechanism as a function

of the thickness of the linear translation flexures

It is also interesting to compare the theoretical stresses in the lever with

results from the finite element analysis for these three designs. Maximum
stresses occur at the ends of the lever and linear translation flexures. These have
been analyzed using finite elements and the results compared with theoretical
predictions. Because of the complexity of this mechanism and the finite mesh
size, FEM results are probably accurate to within 5%. Based on results for the
three different thickness of linear translation flexure, the error between the
calculated stress and that predicted by the FEM analysis are tabulated below.

Table 7.5.2: Comparison of the differences between theoretical stress

pred'tc ti ons and thoseprod uced from f tru'te e1ement ana!Y_sis
1 ·
t(mm) % difference linear spring % difference lever flexure
0.82 -10.00 -24.42
0.60 -8.10 -22.75
0.40 -2.50 4.90

366
CHAPTER 7: LEVERS

a)
R£SlA..TS: 2.. B.C. 1,STR£SS_2 ,LOAD SET 1
STRESS - VOH MISts HIH: 1.02£-Qe - : 6.<6£• 0<
DEfDRHATlOth i• 8.t. 1 , DlSPUICDtiHT_1,LOAD stl 1
DISPUIC[IIEHT - IIAG HIH : 0.00£+00 - : 2 .f2£-o2 INUJE OPT!!»< :ACTIA
rRAH£ or ltH : PART SHaL SU!fAC£ : TOP

b)

Figure 7.22: Alternative lever design in which horizontal motion of the lever is constrained, a)
Solid model showing distortion, b) typical finite element mesh used in lever analyses

At both locations it can be seen that the difference between the two values

367
FLEXURES

increases with the thickness of the linear translation flexure. In this case a
negative error corresponds to the calculated stress being higher than that
produced by the finite element analysis.
Finally, it is informative to analyze the parasitic rotation of the lower edge of
the moving platform as the linear flexures reduce in thickness. To illustrate this,
the parasitic tilt as the percent motion of point A contributed by the tilt in
comparison to the motion at point B will be considered. For the 0.82 mm thick
linear flexures, this corresponded to about 12%. This value reduced to 9% at a
thickness of 0.6 mm and 3% for 0.4 mm. For this last design, a parasitic tilt of
0.03*0.05/15 mm = 100 J.Lrad is predicted for a displacement of 50 J.Lm of the
moving platform. As for the lost motion and analysis of stresses, the trend with
thickness shows an improved performance for the linear flexure thickness of 0.4
mm and this is considered to represent a good design in terms of maximum
stress and minimum parasitic motion of the output platform.
Because the lost motion is primarily due to the linear mode translation of the
lever flexures, constraining this motion using a further leaf spring might result in
a more predictable and consistent performance. Such a design is shown in figure
7.22. The addition of the horizontal flexure results in half of a cartwheel hinge
producing a pivot at the lower left comer of the lever. Assuming an ideal pivot at
this point, a lever ratio of 0.9/4.4 = 0.205 would be expected. Based on finite
element analysis, figure 7.22(b), a measured value of 0.194 corresponds to within
5%. In this analysis, a maximum stress of 64.6 MPa occurred in the cartwheel
hinge for a 43 J!ID displacement of the moving platform. The reader is left to
ponder this design.

7.5.2 Manufacture of the lever using wire electro discharge machining

Both of the above designs have been manufactured with a wire electro-discharge
machine. Typical machining times were around 2 tf2 hours for a flexure of 10 mm
depth. Once produced, accuracy of manufacture was assessed using a vision
based coordinate measuring machine. Dimensions were obtained by measuring
the positions of a number of points along an edge and computing a best-fit line to
the data. Mean separation between these lines is then used as a measure of the
relative dimension. In general, from measurements on five nominally identical
flexures, it was found that the dimensions were consistently in error by an
amount that would occur if the machining radius of the wire were approximately
6-7 J.tm larger than that programmed into the path-generation algorithm.
However, it is not clear that this apparent error is that of the geometry of the
flexure or a systematic effect due to the edge detection method used to measure
this part. Either way, in principle, programming the cutting path for a larger wire
easily compensates th.is. Typically, the standard deviation of the measured
dimensions was of the order 2-3 J.Lm which was only a little in excess of the
measurement standard deviation of around 1.5 - 2 J!ID. In general, it may be

368
CHAPTER 7: LEVERS

concluded that the machine is capable of producing components with accuracy's

of better than a few to 5 f..Lm as is currently claimed by manufacturers of electro-
discharge machines.
There were complications when trying to produce the thin flexures. First
attempts resulted in the complete removal of these thin beams. It was speculated
that this could be a consequence of combined forces from the high-pressure
water jet used to both cool and flush the debris from the machining zone and
thermal damage during the 'roughing' cuts. Reducing the water pressure from a
few to less than 1 bar and leaving a larger amount of material for fine finishing
resulted in acceptable flexure devices.

7.6 Case study II: Optical levers, galvanometers and the filar suspension (a
long, happy marriage)
Towards the latter part of the 19th century, the quest to connect the fll'Ildamental
electrical units with mechanical force led such esteemed scientists as Helmholtz,
Joule, Kelvin, Maxwell and Weber into flexure design. Many of the instruments
developed for this purpose consisted of null balances in which two equal and
opposite forces are created. For example, currents between two adjacent wires
will induce a force that can be balanced by that of a known mass in a known
gravity field (which can be computed from equation 1.1). In many cases, because
of the relatively small forces it is necessary that the disturbed mechanism be
brought back to the null position will minimal residual forces. Requiring precise
force control of small motions, this is an ideal application for a well designed
flexure. Probably the most common design for such applications is to suspend
the moving element from a filar (long thin rod) suspension. Based on the
remarkably accurate results from t!tese initial experiments this mechanism has
continued to enjoy widespread use to the present day, particular examples being
the galvanometer, torsion magnetometer and optical lever (called the 'mirror
method' in Maxwell's treatise). A relatively low cost competitor to the filar
suspension is the jeweled pivot, which provides a more robust rotary bearing
with a surprising low torsion resistance. Figure 7.23 shows a typical moving coil
mechanism originally designed by William Thompson (later Lord Kelvin). In
this, the coil is suspended between two filar suspensions arranged so that the
common axis is coincident with the center of gravity of the coil. Surrounding the
coil are the poles of a magnet (either permanent or electro-magnetically
generated) arranged so that the magnetic flux cuts the wires of the coil
perpendicular to the current. Within this coil, but not touching it, is a soft iron
core to direct the magnetic flux. Passing a current down the filar suspension to
the wires results in a Lorentz force acting in equal an opposite directions either
side of the coil. Under the action of this pure torque, the coil will twist through
an angle (}. The torque can be computed from the torsion equation of chapter 2.

369
FLEXURES

Coil

./""~---- Soft iron

core (fixed)

suspension

Plan view
Cross-section
IF

~ ~
IF
Figure 7.23: Schematic diagram of a moving coil galvanometer in whlch the coil is
suspended, and pivots, by a filar suspension

A carefully wound coil and a magnetically soft iron core ensures the fidelity of
this instrument, for a full discussion see Maxwell, 1891.
Three parameters determine the operation of this actuator. These are the
angular twist due to torque applied by the coil, the current through the coil and
cross product of the magnetic field and current Instead of an electromagnet, the
magnetic field can be produced using a second coil and, in this configuration, the
direct force between the two conductors can be measured. Although systems

370
CHAPTER 7: LEVERS

involving two, three and four suspended coils were developed to null the error
due to 'terrestrial magnetism', the principle of operation is similar. If either the
two of the above are known, the third can be deduced. For the multiple coil
measurements, the current is passed through all of the coils and subsequent
deflections or forces measured. Two methods were commonly used. One is to
suspend the coils on long wires and measure the pendulum rotation. Based on a
knowledge of the mass of the coil and length of the suspension, the electro-
magnetic force can be deduced. Alternatively, to avoid complications associated
with the 'lost' forces in the filar suspension, the twisting or displacements of the
coil could be nulled with masses of 'known' value appropriately placed. The
former techniques employs the so-called pendulum lever, Howard and Fu, 1997,
the latter being referred to as a 'force balance'. In both cases, it is necessary to
have an accurate knowledge of the geometry of the instrument and mass of the
moving parts. Problems associated with a definition of mass have plagued
researchers to this day and the current standard relies on an artifact. For large
masses (1 g to 1 kg), uncertainties of better than 1 part in a million are possible.
Bellow this, influences due to environmental contamination and instrument
limitations can result in unacceptably large uncertainty, Howard and Fu, 1997.
Because the electrical units can now be derived and experimentally realized from
fundamental quantum phenomena, it is being proposed that the kilogram be
replaced by the electrical definition of the Watt thereby linking the artifact
kilogram, the meter and the second to the ohm and the volt. Recent experiments
are reported in the paper of Williams et al., 1998. To this day, a realizable
definition of mass still hangs on a wire (in this case SO strands).
In instruments such as the torsion magnetometer, the optical lever continues
to be the main method for the measurement of angular deflection of the coil, see
for example Hadfield, 1962. Typically, the coil of the galvanometer is suspended
from one wire while a long rod hangs from bellow with the 'free' end hanging
between two poles of an electromagnet. Specimens can then be placed on a
platform attached to this end of the rod and subject to known magnetic fields. If
the specimen has magnetic properties this will result in a torque on the rod and
subsequently on the coil. Deflections are measured using the optical lever and
fed back through an integrating controller to a current amplifier, which acts to
null the deflection. If the instrument is calibrated, subsequent measurement of
the current through the coil and knowledge of the applied field and orientation
of the specimen enables magnetic properties of the specimen to be measured.
Applying a current to the galvanometer will produce a known angular
deflection. A mirror placed on the moving element can then be used to direct a
beam of light. Such devices are often used in printing devices, 'light shows' and
may have a future as color displays (the electromagnet may be replaced by an
electro-static actuator). Whether or not this latter application transpires, the
marriage of the filar suspension to moving mirror looks like it will see a second
diamond anniversary.

371
FLEXURES

Another, more exacting, application of the torsion balance is that used for the
measurement of the Newtonian gravitational constant. Initially used by
Cavendish and Boys towards the end of the 2()th century, new torsion balances
are being proposed in an effort to reduce the uncertainty from the current value
of 0.1% to better than a few parts in 100,000, Quinn et al., 1997.

References
Furukawa E. and Mizuno M., 1990, Displacement amplification and reduction by
means of linkage, Bull. Japan Soc. Prec. Engg, 24(4), 285-290
Furukawa E., Mizuno M. and Terada K., 1991, A magnifying mechanism for use
on piezo-driven mechanisms, Bull. Japan Soc. Prec. Engg, 25(4), 315- 320
Furukawa E., Mizuno M. and Hojo T., 1994, A twin-type piezo-driven translation
mechanism, Bull. Japan Soc. Prec. Engg, 28(1), 70-75
Hadfield D., 1962, Pemument Magnets and Magnetism, J. Wiley and Sons, London.
Hart M., 1965, An Angstrom ruler, Brit.J. Appl. Phys. 0· Phys. D.), 1, 1405 -1408.
Howard L.P. and Fu J., 1997, Accurate force measurements for miniature
mechanical systems; a review of progress, Proc. SPIE, 3225, 2- 11.
Johnson K.L., 1985, Contact Mechanics, Cambridge University Press, chpt. 9.
Marsh D.M., 1961, Micro-tensile testing machine, /. Sci. Instrum., 38, 229 - 234
(also discussed in Smith and Chetwynd, 1991, chapter 6)
Maxwell J.C., 1891, A Treatise on Electricity and Magnetism, Vol. ll, 3rd Ed., Dover
Publications Inc., NY. The reader will be well rewarded for perusing part ill,
chapters Vll and VIll and part IV, chapter XV which includes a first hand
account of these early developments will close attention being paid to the
mechanical construction of a variety of coil suspension apparatus. The
optical lever was well known and of great utility to these workers.
Mindlin R.D. and Deresiewicz H., 1953, Elastic spheres in contact under varying
oblique forces, Trans. ASME, Series E, ]. Appl. Mech., 16, 327 - 344
Quinn T.J., Speake C.C. and Davis R.S., 1997, Novel torsion balance for the
measurement of the Newtonian gravitational constant, Metrologia, 34, 245-
249.
Smith S.T. and Chetwynd D.G., 1991, Foundations of Wtraprecision Mechanism
Design, Gordon and Breach, NY.
Snyder J.J., 1993, Accurate, inexpensive, thermal expansion microtranslator, Rev.
Sci. lnstrum., 64(5), 1351 -1354.
Williams E.R., Steiner R.L., Newell D.B. and Olsen P.T., 1998, Accurate
measurement of the Planck constant, Phys. Rev. Letts., 81(12), 2404-2407.
Xu W. and King T., 1997, Flexure hinges for piezo-actuator displacement
amplifiers: Flexibility, accuracy and stress considerations, Precision
Engineering, 19(1), 4 - 10

372
 8 Manufacturing and
assembly considerations
8.0 Overview
Having selected the appropriate nulterial and geometry for a flexure there remains
the task of building and assembling the mechanism. At this stage it is necessan; to
consider the methods of manufacture and fabrication. Often, conventional nulchining
methods will adequately produce the component to the required tolerances. In many
cases the flexure will consist ofa thin-walled section. Some care is necessan; to avoid
· imposing large forces during nulchining. Two approaches can be adopted. Either the
tool geometry and cutting path strategy can be chosen to direct the cutting forces
towards the bulk mllterial or temporary supports in the form of conforming nulterials
can be applied to the opposite face. For this latter purpose, a range of reusable waxes
and low-melting temperature alloys are discussed. However, for some designs,
conventional methods are either incapable of producing the complex shapes required
or mllY introduce unacceptable surface danulge in the form ofresidual stresses. Other
times, there mny exist more economic techniques for large-scale manufacture. In
these circumstances, it is necessan; to utilize less conventional methods of
manufacture and it is these that occupy the first part of this chapter.
Flexures may be wrought from a single monolith of material or fabricated from
individual components. While the former of these eliminates a lot of the problems
associated with assembly, it is often impractical due to the flexure geometn; or
because the mechanism must integrate with other components for which a monolithic
element would not be suitable. Other times, it might be desirable to change the
stiffness of the flexure by interchanging flexible elements that, for example, might
simply comprise flat plates. Consequently, the fabricated flexure is commonly found
in nulny instruments and machines. For reliable perfornulnce, it is necessary that the
flexure is designed for assembly and a few tips and tricks comprise the second part of
this chapter. By definition, elastic mechanisms do not provide energy dissipation.
Although in practice some losses will be intrinsic to all elastic mechanisms, it is often
necessary to add viscous or visco-elastic damping. Some considerations are also
covered in this section.
For some mllterials selected for these applications, it may be necessary to machine
the nulterinl in its annealed state after which it is then necessary to heat treat the
component. In some cases a post heat-treatment finishing process such as grinding
or polishing are required. For conventional steels, such processes are common and
procedures can be determined from any of the many handbooks on machining
available in any good library. Beryllium copper is often used because of its high yield
FLEXURES

strain. The machining and subsequent heat treatment of this material is briefly
described.

8.1 Manufacture
In most workshops, available machine tools usually comprise lathes, milling
machines, grinders and boring machines. For large flexures, these machine tools,
in combination with standard fastening techniques (including welding and
gluing), probably represent the optimal manufacturing method. Commonly,
either separate components are made and subsequently assembled or the flexure
may be directly machined into a solid to form a monolithic mechanism. For the
production of small, instrument flexures, a broad range of alternative processing
techniques may better achieve a particular design. Also, in some cases, it may be
necessary to produce the flexure from hard or brittle materials that are not
amenable to single point cutting or grinding processes. Under these
circumstances, it is necessary to utilize less conventional production methods.
Techniques commonly used for the manufacture of flexures include
• Conventional machining
• Electro~discharge machining

• Lithographic etching
• Electroplating (or electr~forming)
• Diamond grinding

These are discussed, in turn, below

8.1.1 Conventional machining

The objective of this brief section is to provide a few tips on the use of
conventional machine tools to produce leaf or notch-type flexure geometry's.
Detailed machining parameters for most materials can be found in available
literature and therefore are not covered here.
Consider a flexure mechanism produced from a single monolithic solid. By
definition, toward the end of the machining cycle, the monolith will consist of
two or more rigid bodies connected by relatively compliant elements. Clearly, it
will be necessary to ensure that these rigid bodies are suitably constrained on the
machine tool. With care, such constraints can often be supplied using standard
fixtures. Additionally, during manufacture, it will be necessary to machine the
leaf and notch hinges. This is problematic as the limited strength of the thin webs
of the hinge render it susceptible to deformation if large cutting forces are
applied in the wrong direction. If possible, machining using small, sharp, cutters
running at relatively high speeds will tend to minimize these undesirable forces.

374
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSIDERATIONS

Additionally, it may be possible that the flexure can be machined in a way that
ensures that the cutter is always removing material from a relatively stiff region
on the monolith. When machining, the general rule is to direct the cutting forces
towards stiff or s~pported solid material.
a) b)

Figure 8.1: Using the side face of an end mill to produce a thin walled section, a)
using a conventional cutter, b) use of a modified cutter to reduce surface damage
in the presence of chatter
During the cutting of thin walled structures using a milling machine, it is
important to keep in mind the machining forces at the point of cutting. Cutting
with an end mill will introduce both axial and radial forces. Figure 8.l(a)
illustrates the cutting of a thin walled section using a standard end mill. In this
figure the cutter is moving towards or away from the reader (i.e. it is moving in a
line that is perpendicular to the plane of the paper) and the lower face of the
workpiece is considered to be rigidly fixed to the machine tool frame.
Compliance with the principle of machining from solid or high stillness it is
obvious that the most desirable cutting planes are either on the upper face of the
solid block or, near to the bottom of the thin web. Clearly, the worst place to
attempt machining is at the surface of the thin web towards the top. Also
illustrated in this figure is an exaggerated distortion of the thin web when it is
vibrating. Predominantly, it is the periodic cutting forces introduced by the teeth
of the cutter the source of excitation. When the rotational speed of the tool
produces periodic forces in harmony with the natural frequency of the workpiece
then the forced vibration amplitudes increase dramatically. Often, after excitation
of the workpiece begins the interaction between the moving surface and the tool
will introduce additional forces. Because these forces will be at a similar (or
'sympathetic') frequency as the vibration, they will further amplify the distortion
leading to a form of self-excited vibration commonly called 'chatter'. Inevitably
this will diminish surface finish and geometry of the web. It might appear that
choosing a different spindle speed or irregularly spacing the cutting teeth around
the tool periphery could solve this. However, the natural frequency of the web

375
FLEXURES

will change as machining progresses. Although it might not be possible to

eliminate vibration of the web, the effect of this on the geometry of the final
component can be reduced by changing to the cutter geometry shown in figure
8.l(b), for more detailed discussions see Smith and Dvorak, 1997. In this case, the
flutes of the end mill extend a short axial length from the free end of the cutter
after which they are removed. Above these flutes, a clearance provides a space in
which vibrations of the web will have little effect. A drawback with this
approach is that the web must be produced by multiple step-and-repeat
machining cycles.
a)
An alternative
b)
method for producing thin
sections is represented in
figure 8.2. In this case, the
machining direction is the
same as for the previous
figure but the flexure is
being produced from the
Figure 8.2: Machining of a monolithic flexure using an end horizontal surfaces
mill, a) a slot is produced in a solid block. b) the block is machined by the end face
turned upside down and another slot produced. The flexure
hinge consists of the surfaces generated by the end face of the
of the cutter. While some
cutter during the two machining operations. vertical forces will be
produced during this
machining operation, these are relatively small compared to those imparted on
the rigid blocks by the side face of the cutter. In this illustration it is assumed that
the flexure is to be produced by first of all machining the upper slot after which
the block is turned upside down. The flexure will emerge after the second slot is
complete (this is when the cutter, which is moving away from the reader, is clear
of the workpiece). In practice, to reduce stress concentrations at the roots of this
flexure, a radius end mill should be used.
For flexures of higher slenderness it may be necessary to use smaller cutters
operating at higher speeds in an effort to reduce forces in the direction of the thin
web. Again, it may be possible to generate the thin web using end milling using
the side face of the cutter to machine primarily on the side face from solid
material. Such a scenario is shown in figure 8.3. The slot is machined in a similar
manner to the previous example. However, in this case, it is also necessary to
translate the cutter along the length of the flexure as shown. More detailed
discussions and examples giving all machining parameters for thin web
manufacture from an aluminum workpiece can be found in the papers of Tlusty
et al., 19%, and Smith and Dvorak, 1998. Under suitable machining conditions
manufacture of webs of thickness 0.5 mm and lengths of 50 mm or more (i.e.
slenderness ratio's of 100) can be manufactured to relatively high precisions.

376
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSIDERATIONS

Direction of tool path and

a) b)
1---.-l final position of cutter
I I

Figure 8.3: Machining a slender, monolithic flexure, a) a slot is produced

in a solid block, b) the block is turned upside down and the process
repeated
An important advantage of notch hinges is that they are easier to produce
using conventional machine tools. In principle, it is only necessary to drill the
holes for the webs with any precision. Because this can be performed on the solid
blank, provided the webs are of sufficient thickness, these webs will have
tolerances near to the precision of the machine tool used to produce them. After
the holes for the notches have been produced, removal of the remaining material
will have little influence on the performance of the mechanism. Often, in the
production of the webs, the drilling of the second hole will significantly distort
the web of the notch. This is particularly troublesome because the most distortion
will occur at the thinnest portion of the web that, in tum, imposes the most
significant influence on its performance. In many cases, a remedy to this problem
is to drill (and, preferably, ream) the first hole and insert a closely fitting dowel
pin. This will then provide the necessary support of the thin web while the
second hole is being drilled or reamed.
For non-circular webs, manufacture of fixtures to support the rear face
during final machining on the other side becomes time consuming and
expensive. To ensure stiff support, it is also necessary that the support closely
conforms to the surface and hence the fixture must be either precisely
manufactured or some fine adjustment must be incorporated. Alternatively, if a
pocket can be formed that incorporates one flexure surface, support can be
added by temporarily filling this with liquid filler that can be solidified and
removed after machining is complete. Choosing the appropriate liquid is still a
matter of debate. Contenders include

• Waxes
• Low temperature metals that do not contract upon cooling
• Two part epoxy adhesives
• Solders

377
FLEXURES

In principle, application of these temporary fillers is readily implemented.

While in the liquid state the filler material is poured into the plenum and
solidified after which it functions as a support on the opposite face of the web
during machining. After machining, the filler is removed, dissolved or
transformed back to its liquid form and drained from the flexure. With this latter
option, the filler can be reused.
Waxes possessing a very wide range of thermal and mechanical properties
can be readily obtained. Most commonly these are supplied for the purposes of
optical bonding, candle making or as reusable machine tool prototype material.
The so~called 'optical waxes' have been in use for centuries for the temporary
bonding of optical components during lapping, polishing etc. These can be
obtained with a large variety of hardness values and softening temperatures.
Primarily these are made from varying mixtures of beeswax and rosin sometimes
with additives such as turpentine or rosin oil. In general, the higher the
proportion of rosin the harder the wax. A soft optical wax can be readily
produced from a mix of equal parts refined bee's wax and optical rosin. A typical
medium hardness wax is produced from 16 parts rosin, 2 parts bee's wax, 1 part
Carnauba wax and a small addition of Venetian turpentine or rosin oil. A large
variety of waxes for different applications are commercially available. Generally,
softening temperatures tend to correlate with hardness of the wax. Typically the
softer waxes will soften at temperatures of around 40 to 60 oc while the harder
waxes tend to soften at temperatures in the region 70 to 140 oc. Usually,
increasing the temperature above the softening point will result in a reduction of
the viscosity. For example, a hard wax with a softening temperature of 80 oc
typically will not flow 'freely' to form a thin surface film until heated to 140 oc.
In the softened state these waxes have good wetting properties and will adhere
to almost any clean metallic or ceramic surface. Upon cooling this will provide a
relatively stiff bond or, in bulk, a strongly adhered solid surface. Although the
wax contracts upon cooling, this does not cause many problems. Because of its
low conductivity it is likely to cool slowly from the outer surfaces and, therefore,
shrinkage tends to result in a slight'coring' in the center. Most waxes will form a
strong bond that will resist moderate machining forces. A lot of the commercial
waxes will slowly dissolve in detergents and many coolants used in machining
processes. However, at room temperature, dissolution usually occurs slowly
with reduction in strengths of bonded components only becoming noticeable
after having been immersed in coolant for many days.
When machining is complete, the wax and any temporary backing materials
can be removed by re~heating the assembly above the softening point. When the
wax has been removed or scraped away, there will often be a residual film. These
films can be removed using a suitable solvent. Typically, warm acetone or
petroleum provide relatively rapid action while alcohol, which is most
commonly used, is rather slow. In all cases, solubility increases with
temperature. Information of other, more efficient, solvents can often be obtained

378
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSIDERATIONS

from the supplier. Excessive heating can result in the wax film becoming burnt or
fused to the surface. Invariably, this will be very difficult to remove.
Waxes provided for the purpose of prototyping in machine tools have been
used to provide support during the machining of thin flexures.
Some reusable, low-melting point metal alloys are supplied specifically to
provide a temporary supporting material for machining thin-walled structures.
Because these alloys can expand upon solidification, components held in the
frozen matrix will experience a clamping force, and so, they are also used for the
securing of awkwardly shaped components. For this reason they are also
referred to as 'fixturing alloys'. One such alloy has a melting point of around 70
oc and can therefore be removed by placing it in a container and putting it in a
hot bathl.
A disadvantage with both of the above techniques is that it is necessary to
heat up the workpiece. Often, to do so, it is necessary to remove it from the
machine tool, after which, precise relocation of the workpiece is problematic.
Heating the workpiece in the machine tool may cause unacceptable time delays
because it is often necessary to wait for the system to return to an acceptable
thermal equilibrium. Room temperature adhesives appear to provide a solution.
Current efforts to utilize solders as a filling material have not met with
success. There are three reasons for this; they shrink upon solidification, they do
not wet many of the materials used for flexures and their high conductivity
causes problems due to rapid solidification and, therefore, inadequate flow
during pouring.
8.1.2 Electro-discharge machining

Two types of electric discharge machines are commonly found in engineering

workshops. These are the plunge and wire electro discharge machines outlined
in the following sections

8.1.2.1 Plunge electro-discharge machining

Electro-discharge machining, as the name suggests, utilizes electric discharges to

erode local areas of a surface as a shaped electrode is brought into close
proximity to a grounded workpiece surface. Over the last four decades, this
technology has undergone significant development and such machines are now
considered to be capable of producing components to a relatively high precision.
The principle of operation is relatively simple. A shaped electrode is
produced using conventional machining, sintering or forming operations. The
work-piece is immersed in a dielectric fluid (mineral oil, kerosene or purified
water). A potential is then applied to the electrode and it is lowered towards the
workpiece surface until electric discharges are produced. A servo-control
maintains a constant separation to enable pulsed discharges at rates ranging
1
Supplied by MSC Industrial Supply Co., PA.

379
FLEXURES

from tens to hundreds of kiloHertz. Continuously plunging and flushing debris

from the work zone enables the surface to be eroded to any desired depth. After
machining the workpiece will become a replica of the electrode. To produce the
discharge, it is necessary to apply potentials ranging from tens to hundreds of
volts resulting in currents of 0.1 to 500 amps. At the highest removal rates, the
surface will melt and re-solidify leaVing a relatively thick damage layer with
poor surface finish. After rough machining, a lower power finishing process is
required to remove this re-solidified, or recast, layer (see below).
Electrodes are usually made from materials with a high melting temperature
and high latent heat of melting. Graphite is often used for shaped electrodes,
while tungsten wires are commonly used to produce smaU holes of high aspect
ratio. Alternatively, copper tools can be used in a process called 'no wear' electric
discharge machining, Kalpakjian, 1995.

8.1.2.2 Wire electrodis~barge machining

Although the material removal mechanism is similar to the above, the principle
of operation for this process is entirely different. In this case, the electrode is a
thin wire that is continuously fed through the work-piece. This is then
electrically energized and used to erode a thin line through the workpiece
material. Any wood worker will recognize this as an equivalent to jig sawing, or
grocers, the cheese cutter. Moving the wire in a controlled path through the
specimen results in the desired geometry. In practice, wires ranging in size from
0.025 to 1 mm are common. Presently, capital cost of these machines is high and
the continued costs of wire and filtering of the flushing fluid limits this process
to the production of precise, high value flexures. It is not clear that this will
always be so. Some of the advantages of wire electro-discharge machining for
flexures are
• Components of almost arbitrary complexity can be produced.
• Any reasonable conductor can be machined enabling choice from the
hardest materials
• For thin wires, the gaps between surfaces can be correspondingly
small, enabling the design of compact mechanisms.
• Being capable of machining to depths of 100-200 mm, geometry's of
high aspect ratio can be produced making possible designs with high
ratios of stiffness between the free and constrained directions.
• Highly automated machines capable of re-threading accidentally
broken wires can be left unattended, therefore reducing labor costs.
• Work-pieces can be stacked to produce multiple components with a
single machining cycle.

380
CHAPTER 8: MANuFACTURING AND ASSEMBLY CONSIDERATIONS

The economics of this process are not straightforward and the reader is left to
assess their own particular needs and applications. In passing, it is noted that the
erosion process requires the volumetric removal of material in contrast to
machining in which only new surfaces are produced through extensive plastic
deformation. However, unlike machining using conventional cutting tools, wire
EDM can be used for coring. Because the core remains in one piece, it can be
further used as stock for other, smaller, components. For removing large
volumes of material it is sometimes faster and requires less energy to use wire
electro-discharge machining. For comparative purposes, some typical machining
parameters are outlined below.
Cutting speeds through the work-piece typically range from less than one up
to a few millimeters per minute for thin wires of diameter 0.25 mm or less. In
general, the feed-rate will slow down with increasing thickness of the work-
piece. At the faster cutting speeds or higher removal rates, surface finish tends to
become rougher with R, values typically in the region of a few micrometers.
Applying 'finishing' cuts can produce better surfaces with Ra values of a small
fraction of a micrometer. However, this order of magnitude gain in surface finish
is obtained at the expense of discharge power. Consequently, machining times
are likely to increase with each finishing cut (sometimes the feed rate for the
finishing cut is one third of that for the 'fast burn'). For low surface damage and
surface finish values, it is not unusual to perform four finish cuts. Consequently,
the machining time may increase dramatically. Typically, the electrode wire
feeds through the specimen at a rate of 100 - 300 mm s-1.
Under controlled conditions, it is not unreasonable to produce components
holding tolerances within a few micrometers and with sub-micrometer surface
finish. Flexures measuring 0.1 mm in the thinnest region are routinely
manufactured. Although some surface alteration must occur, no significant
problems caused by this have been observed in flexures of these dimensions.
Current machines can produce components to accuracies of better than a few
micrometers, see for example the case study 1 of the previous chapter.
As an example, a semicircular notch hinge of radius 6.4 mm, P= 0.172 and a
depth of 15 mm could be produced with four finishing cuts in a total time of
approximately 30 minutes.

8.1.3 Lithographic etching

Manufacture of flexures using lithographic etching processes is a widely used

technique for the mass production of complex planar mechanisms. One does not
have to look far to find one. Magnetic read heads on almost all hard disk storage
devices are supported on thin stainless steel flexures, accelerometers commonly
used in automobiles as crash detectors contain flexures integrated into micro-
electronic sensing circuits, and thin beryllium copper flexures are commonly
used in fine scientific instruments. In fact, the former two devices are produced

381
FLEXURES

in quantities measured in tens of millions per year. The first uses a proprietary
etching process.

8.1.3.1 Flexures produced using microelectronic processing techniques (MEMS)

Production of flexures using microelectronic manufacturing techniques is well
known. Laboratory devices manufactured more for scientific interest than market
potential were being made in the late 1970's. It was not until these devices were
presented in a seminal paper by Petersen, 1982, that the scientific community
became alerted to the potential of micro-electro-mechanical devices (now often
referred to as MEMS). Since this time, the number of applications has grown
enormously although the main markets appear to be for displacement sensors,
accelerometers or resonators for mechanical frequency selection.
In fact a simple dimensional analysis of flexures indicates the advantageous
dynamic characteristics of miniature devices. Consider a simple cantilever beam
for which the natural frequency is given by (3.232)

(8.1)

If we now change all of the dimensions by a factor k, equation (8.1) above

becomes

(8.2)

For a reduction in the dimensions there is a corresponding linear increase in

the dynamic response of the system. Such a scaling is, at least intuitively, obvious
when observing relative heights that can be jumped from a standing start by
elephants, dogs and their fleas.
Manufacture of the micro-mechanical devices traditionally starts with a flat
silicon wafer. Patterned masks are transferred to the surface using photographic
techniques and thin layers are either grown or deposited. Removal of the mask
leaves a thin film on the surface that is a replica of the original pattern. Common
film deposits are silicon dioxide (often used as a spacing layer to be removed
during a later processing stage), polycrystalline silicon (structural components),
silicon nitride (insulator) and aluminum (conductor usually on the upper face).
After multi-layered structures have been deposited, selective etches can then be
used to remove underlying layers leaving thin films 'free'. Components
deposited on these layers can be left completely free to form rotors or more
complex shapes or may remain attached at certain points so that the substrate
becomes the fixed link of a planar flexure mechanism.

382
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSIDERATIONS

Currently, the thickness of the films and, therefore, these devices are limited
from a few to tens of micrometers. Consequently, devices with planar
dimensions considerably larger than this tend to be rather slender. Another
technique for producing components that have a higher aspect ratio (ratio of long
dimension to thickness) is called UGA, see Guckel et al., 1991. This utilizes thick
films that can be decomposed using high energy x-rays. Screening the thick film
with a mask and exposing it to high intensity x-rays can produce high aspect
ratio holes. These can then be filled with, for example, nickel, after which,
removing the surrounding film will leave correspondingly high aspect parts.
Fortunately, the analytic techniques outlined in the previous chapters are
equally applicable to these devices. It is clear that the mass production of flexures
using micro-scale lithographic processing will provide inexpensive, fast and
repeatable flexure systems with integrated sensing and computing capacity.
8.1.3.2 Lithographic etching of copper sheet

For many decades, workers in electrical industries have been producing complex
shapes from thin copper films. Printed circuit manufacture can be readily
adapted for the production of planar flexures in thin beryllium copper sheet.
Advantages of this process are
• Arbitrary planar shapes can be produced from computer generated
masks
• Beryllium copper can be processed in worked and hardened state as
supplied by the manufacturer2 thereby avoiding post process heat
treatments.
During the production of a flexure, the following steps represent a typical
manufacturing procedure
1. The thin sheet is first of all cleaned using a solvent such as acetone or
isopropyl alcohol.
2. A thin layer of photo-resist is deposited onto the surface preferably
spinning the sheet to produce a thin and relatively uniform deposit.
Note: Positive acting photo resists tend to be easier to process.
3. Baking to around 100° Celsius for approximately one minute then
hardens the photo resist.
4. An opaque mask is placed over the surface, which is subsequently
exposed to ultra-violet light. A 150 W light source requires an exposure
of around 10 seconds. U this is not available, exposure in strong sunlight
for ten to twenty minutes is often adequate.

2
Brush Wellman Inc., Ohio 44110 will supply detailed booklets relating the properties and heat treatment
of a range of beryllium copper alloys

383
 FLEXURES

5. The photo resist is then developed (sodium hydroxide developer)

leaving the original mask pattern on the surface and this is then baked
for around twenty minutes at 80 -100 Celsius.
Finally, the rear face of the sheet is painted with etch resist (often referred to as
'dope' by printed circuit manufacturers) and the sheet then etched in ferric
chloride. To accelerate etching, the ferric chloride solution is often heated to
around 40 - 60 Celsius. Etching continues until all unexposed copper has been
dissolved. Removal of the flexures as soon at this occurs ensures minimal
undercutting.
In general, if the flexures are removed as soon as etching is complete,
undercutting by the etchant into the pattern will only extend a distance roughly
equal to the thickness of the film. For large thin sheets, such an effect will be
insignificant. In practice, sheets of up to 200 J.IID thick can be readily etched at
minimal expense.

8.1.4 Electroplating (or electro-forming)

Bellows produced by electro-deposition are commercially available3. If the film

deposited is removed to become the component, this process is called electro-
forming. Typically, these are produced by forming a mandrel, depositing a
desired thickness of metal (nickel), trimming the metal and then dissolving out
the mandrel. All metals can be electro-deposited and, because precise control of
deposition is possible, films ranging from near atomic dimensions to tens of
micrometer thickness can be readily produced. Other metals commonly used for
electro-forming include copper, gold and silver.
For coating onto insulating materials such as plastics, the surfaces must be
pre-coated using electro-less processes. Nickel is amenable to a relatively
straightforward electro-less deposition.
The advantages of this process are
I. Complex three dimensional geometry's can be readily produced
II. Because the film can be made very thin, relatively large distortions can
be achieved. This is immediately apparent from the bending equation,
which, for a flat plate of thickness t bent to a radius R, there will be a
stress at the surface of magnitude

Et
(j = - (8.3)
2R

As the thickness t reduces the curvature (1/ R) of the film can increase.
ill. Seamless, non-porous geometry's can be produced.
3
Servometer Corporation. Cedar-Grove, NJ 07009. This company supplies infonnative booklets outlining
applications and detailed design calculations.

384
CHAPTER 8: MANuFACTURING AND ASSEMBLY CONSIDERATIONS

IV. Components of almost arbitrary size can be produced ranging from

bellows of less than 1 mm outside diameter to components·weighing
hundreds of kilograms
It should also be noted that thin film devices often exhibit excellent fatigue
resistance.

8.1.5 Diamond grinding

In some applications, it might be desired to use ceramics or other brittle materials
for the flexure. The advantages offered by these materials include
• Being brittle, they will exhibit. linear behavior only. In contrast, flexures
produced from metals and other ductile materials will only exhibit linear
behavior if the maximum stresses are below the elastic limit. In brittle
materials, the elastic limit and catastrophic failure occur at the same stress.
• Materials can have excellent combinations of properties. For example, the
glass ceramic Zerodur™ has a thermal expansion coefficient that can be a
factor of 200 or more lower than steels at room temperature; silicon can be
obtained in very pure, single crystal form, titanium boride has a very high
hardness and speed of sound (however, it is difficult to machine with any
cutter); diamond has superior mechanical and thermal properties to any other
material, etc. (for a more complete discussion see Smith and Chetwynd, 1992).
• Brittle materials tend to maintain dimensional stability over extended time
periods.
• Many ceramics and insulators provide excellent immunity to hostile
environments.
Many ceramics and non-metalli~ solids can be readily machined by grinding
with either diamond, cubic boron nitride or other hard grits impregnated in a
matrix. Because of the broad range of materials that can be processed using
diamond cutters, these shall be the focus of this section. However, in many
aspects, similar process parameters (i.e. speeds, feeds, etc.) apply to grinding
wheels comprising the alternative grit types.
Diamond wheels can be purchased in a variety of shapes and sizes.
Invariably, the diamond grits are attached to a metal wheel or to the ends of
spindles (often called 'pins'). These, in turn, mount into standard grinding
spindles that can be intrinsic to the machine or retrofit to milling machines and
lathes. Grit sizes ranges from very fine 400 grit (this number indicates the
number of grits on average that would form a straight line one inch long) up to
coarse sizes of around 50 (approximately 0.5 mm). For super fine finishing, 800
grit wheels are sometimes used. The graph of figure 8.4 provides a rough guide
to the expected surface finish as a function of grit size. Although it would appear
that the finer grits are desirable for all applications, the maximum depth of cut

385
FLEXURES

0.8

0.6
•
......
~ 0.4
.._,
r:l
0.2

0
50 100 150 200 250 300
Grit size (um)

Figure 8.4: Expected surface finish as a function ofthe grit size used for
grinding
reduces at nearly the same rate. As a consequence, the volumetric removal rates
drops off even more rapidly as the surface finish improves.
Typically the abrasive part of the wheel is produced using one of two
techniques. Possibly the least expensive method is to coat the metal shank with
diamond grits and then deposit over these a thin nickel plating for retention.
Although inexpensive, it is difficult to true such a wheel and the grits tend to
have a relatively low packing density. Consequently, these are usually used for
rough machining where surface finish and dimensional tolerances are not
important.
A better quality wheel can be produced by mixing the diamond grits in a
softer matrix (any matrix will be softer!) and bonding this to the wheel or pin.
Common matrix materials are either resin or metal (brass or bronze) based. The
latter matrix proving to be more difficult to produce and, therefore, more
expensive to purchase. However, both wheels have the advantage that they can
be trued for optimal cutting conditions. Because of their excellent durability, it is
often possible to dynamically balance the metal bonded wheels after which they
can be used for many hours without further attention. It is the longevity of the
metal bonded wheels that makes them desirable for precision machining
applications. In the author's experience, the quality of all types of wheel can be
very variable with wheels appearing to wear out almost immediately or some,
seemingly, last forever. As a consequence, many machinists can be very
possessive of their 'favorite' wheels.
When machining ceramics, there are some general rules that can help the
novice to get started. These are

386
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSIDERATIONS

• Keep the workpiece surface as cool as possible. Flush the grinding

process with a fast flow of coolant taking care to direct the flow
directly into the cutting region.
• Make. sure that there is plenty of opportunity for the debris to be
removed from the cutting interface. Clogging of ceramics at high speed
can lead to rapid heating and solidifying of aggregates, often with
catastrophic result. Unlike metals, which will tend to make a lot of
noise and glow before failing, ceramics give no warning signs before
shattering.
• Use safety guards.
• Do not machine steel, it will wear out the diamond wheels very
rapidly.
• For efficient machining rates, use surface speeds of the grinding wheel
of the order 5 to 15m s-1.
This last parameter is important for the machining of complex shapes using
small cutters similar to those used for end milling. Under these conditions, the
cutter consists of a small end mill often having a diameter of a few millimeters or
less. To achieve a requisite surface speed of 10m s-1, a cutter of diameter 4 mm
must rotate at an angular speed of 47,750 rpm. Fortunately, a.ir turbine and
electric motor driven spindles that can readily achieve such speeds are
commercially available. Although it may not have sufficient rigidity for precision
applications, a common dentist drill will typically be rotating at speeds higher
than 100,000 rpm.
Usually, combinations of slicing and end milling are sufficient to produce
most flexures. Provided that the debris can be removed from the cutting zone,
arbitrary depths of cut are possible. In general the feed rate will depend on the
size of the grits. Typically the depth of cut per pass of a grit through the work
piece should not exceed one hundredth of the grit size and for fast cutter speeds
and deep cuts this should be reduced by an order of magnitude. However, this is
only a guide and can vary considerably depending on the machining
configuration. For example, with small cutters rotating at speeds of 1000
revolutions per second, a grit size of 100 /liD corresponds to a feed rate of 0.1-1.0
mm s-1. As an example, single crystal silicon is readily machined with a 100 grit,
2 mm diameter wheel rotating at 50,000 rpm with a feed rate of 2 mm ffiin·l or 0.3
mm s-1 . Machining with slow feeds and large depths of cut is usually referred to
as creep-feed grinding. This has the advantage of utilizing more of the cutter
surface therefore increasing the life of a tool.
Holding a brittle workpiece in a machine tool is a problem. Most materials
are non-(f~rro)magnetic and, therefore, cannot be held directly using a magnetic
chuck. Being brittle, clamping between the hard jaws of a chuck can cause the
component to crack if there are any local stress concentrations. High stresses can

387
FLEXURES

be caused either by hard dust particles or if the faces of the vise are not
coincident with the work piece so that there is a damping force along an edge.
Both scenarios are likely. A common method for mounting the work piece in a
machine tool is to first stick them to the surface of a metal block using optical
waxes, see section 8.1.1. This metal block can then be mounted in the machine
tool in the usual manner. If the component is to be sliced through, it is usual to
place an intermediate glass or ceramic plate between the work piece and metal
mounting block. Float glass, being inexpensive and readily available in most
hardware stores, is often the material of choice for this.
Again, the economics of creep feed grinding are not straightforward.
Typically, it is feasible to finish the machining process with a single cut. When
manufacturing metal components, it is common to saw the blank from larger
stock, remove bulk material using roughing cutters and complete the finishing
process with a relatively slow, fine finishing (usually grinding) operation. Creep-
feed grinding, although requiring a slow feed rate, can combine roughing and
finishing in one process. Consequently, it is likely to become economic for the
cutting of large slices. Another consideration is that of mounting these
specimens. Melting and cooling of optical waxes is time consuming. This may
only be economic for components with large profit margins or in cases in which
multiple flexures can be produced from a single machining cycle.
Drilling represents a more difficult challenge. In principle, it should be
simple to use a diamond-coated pin as a 'drill'. The key problems appear to· be
the removal of debris from the hole and the fact that no matter how fast the
cutter is rotating, the surface speed is always zero at the bottom of the hole in the
center. Although it is feasible to drill using pin type cutters rotating at high
speeds, it is difficult to prevent the occasional clogging. Invariably this results in
the drill 'binding', often with catastrophic result. A more conservative approach
is to rotate the cutter slowly (the author has used 70 rpm with a 3 mm cutter for
'drilling' a 10 mm deep hole) and apply constant force to the cutter. To avoid
dogging, the cutter is continuously retracted and the hole flushed with coolant.
Such an operation corresponds to a pecking cycle in which the cutter is fed at
constant force instead of constant feed rate.

8.2 Assembly
Invariably, whether it be fabricated from many components or machined from a
single monolith, a flexure will eventually be integrated into more complex
instrumentation or machinery. Not only will it be necessary to clamp the flexure
into a larger frame, but actuators for translation and sensors to monitor position
must be added. Fortunately, most sensors consist of electromagnetic coupling
(capacitance, inductance, optical interferometry) which adds little mass or
stiffness to the flexure. Parallel plate capacitance electrodes can add significant
damping due to shearing or squeezing of the air in the gap between them. There

388
CHAPTER 8: MANuFACTURING AND ASSEMBLY CONSIDERATIONS

remain some issues associated with coupling of actuators to the flexure and
subsequently attaching the completed assembly to a rigid frame. Finally, for
optimal dynamic response, it is often necessary to add damping (or energy
dissipation) mechanisms. Consequently, this section addresses
Assembly of flexures
Coupling the actuator to the flexure
Flexure mounts
Adding damping to flexure systems (extracting energy)
8.2.1 Assembly of flexures
Consider the exploded view of a single degree of freedom parallelogram flexure
shown in figure 8.5. It is apparent that all components are relatively simple to
manufacture and assemble. Important features of this assembly are the clamping
blocks, leaf type flexures and the upper and lower platforms.
Upper

L
platfonn

Clamping
block

Section of clamped
flexure in distorted
state

Figure 8.5: Exploded view of a simple leaf type linear flexure assembly

389
FLEXURES

It can be seen that the clamping blocks have a recess to provide clamping
forces along two lines. There are two reasons for this. Firstly, the function of the
clamp is to provide a bending moment to the flexure. Resolving this into two,
equal and opposite forces acting at an equivalent distance either side of the bolt,
it is clear where the forces act in this case. Obviously, for a given bending
moment these forces reduce with increasing separation. There are obvious limits
to this separation. To ens'ure that the clamping forces applied to the leaf are a
minimum, the contacts will be lines at the extreme edge of the clamp. However,
this will result in large stresses. Consequently, it is necessary to extend this to a
rectangular area contact preferably having a minimum thickness much greater
than that of the leaf spring. Another consideration is that of stresses imposed by
the clamp screws. In practice, screws are usually tightened until near to the yield
stress. Most screws, being made from steel, have a yield stress, at least, near to
that for a typical leaf. Consequently, to reduce the stress applied to the leaf, it is
necessary that the clamping area of the contact is larger than the combined cross-
sectional area of the screws. This is less important if the hardness of the leaf
spring is considerably higher than that of the screws and materials of the clamp.
Typically, it is better to design the recess to be a little wider than the thread of the
clamping screw. Consideration of the two clamps shown in figure 8.6 should
reveal the difference between good and bad design. Always desirable are spring
washers that help to regulate the rate at which forces are applied with rotation of
the bolt.

Flexure
platform

Spring
washer and
washer

-·- -
Figure 8.6: A typical, leaf spring clamp, a) clamp element too thin, of incorrect geometry and
fastened with undersized bolt (dashed line shows exaggerated distortions), b) correct clamping
method
Another consideration is the maximum stresses in the leaf which coincide
with the edge of the clamp. Firstly, there will be a compressive stress due to the
clamps. Applying superposition, the total stress will be the sum of this plus the
stress due to bending. For a given displacement, 8, the radius of curvature due to
bending for the leaf spring shown above is

390
CHAPTER 8: MANUFACTURlNG AND ASSEMBLY CONSIDERATIONS

(8.4)

If the comers of the clamp surfaces were sharp at this point there would be a
local stress concentration of infinite magnitude. Clearly, something would yield.
Choosing clamp materials of lower hardness will cause the yielding of these
sharp comers and not of the flexures. Consequently, if the flexure is deflected to
its maximum position after assembly, there will be an automatic plastic
deformation of the clamp edge, after which, the flexure should behave in a
reasonably linear fashion. A more reliable design would apply a small radius at
this edge of the clamps and piatforms. Often, a machined de burr will suffice.

a) r b)
~b+sb~

Figure 8.7: Potential parasitic errors of a simple linear spring due to manufacture and assembly
tolerances
Manufacturing and assembly tolerances present another source of error. Two
sources are those due to variations between separation of the springs at platform
and base, sb, and those due to difference in length of the flexures, s,, S'ee figure
8.7. For a linear translation 0, the parasitic rotations due to these errors can be
approximated from the equations

(8.5)

(8.6)

Tolerances on assembly are tighter for equal length of the flexures than they
are for lengths of the platforms. To minimize the latter, both moving and

391
FLEXURES

stationary platforms should be machined as pairs. The former error is usually

introduced during assembly and can be minimized by placing a parallel
separator (i.e. gage blocks) between the platforms during assembly.
Often, a long, leaf type spring has insufficient torsion or buckling resistance.
Splitting the leaf springs and separating them as far as possible from the center of
the platform can enhance the torsion resistance. Buckling resistance can
sometimes be enhanced, at the expense of maximum displacement range, by
applying a stiffening element in the middle portion of the flexure. The bending
moment along the axis of the flexure is zero in the center and increases linearly to
a maximum value at each end. Consequently, applying constraints at the middle
portion does not have a large effect on bending stresses for a given deflection but
can dramatically increase its ability to withstand buckling loads. Applying
central constraint results in a linear flexure comprising four, leaf-type hinges.
This results in more precise location of hinge action and, therefore, assessment of
parallelogram platform motion.
F

L, l t- ,..-- t-

Lr ~S'fii.
ti emng
elements
/
..._
'-- f- t-

Base

Figure 8.8: A simple Linear spring in which four, leaf type hinges are produced by
clamping the central portion of two leaf springs

To assess the effect of applying stiffening to the central portion of the

flexures, consider the simple spring system shown in figure 8.8. For a simple leaf
type spring, the maximum stress for a platform displacement ois

(8.7)

For the same mechanism of four leaf type hinges this is

E&
cr.L =--
2L L"
(8.8)
I

392
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSlDERATIONS

7
!:l
b
!. 5
b

1
0 0.05 0.1 0.15 0.2
K
Figure 8.9: Graph of the ratio (j L•/(j 1-r as a function of the
dimensionless ratio K
Assuming that the length of the leaf hinges is small in comparison to the total
length between platforms, the ratio of stresses in the two flexures for a given
displacement is given by

6L1 L. 6(L1 Lr-L} )

=-zr-=
(j L,

(jL. L~ (8.9)
= 6K(l - K)

where K= L1 /Lr
Plotting this in figure 8.9 indicates that there will be little difference in the
stresses for the two designs if the hinges are around one sixth of the length
between two platforms (or with stiffeners of 2/3 the length). For shorter hinges,
the displacement range drops off rapidly.
8.2.2 Coupling the actuator to the flexure

Actuator couplings represent a classic 'Catch 22' design problem. On the one
hand, it is desired to rigidly connect the actuator to the flexure to avoid, as far as
possible, 'lost motion'. Denoting the equivalent stiffness values of the flexure in
the desired and undesired axes kd and k. and of the actuator and coupling
ka and kc respectively gives an approximate model for the lost motion (for a more
detailed discussion of this problem see chapters 5 and 7)

Xo kakckw
7. = kakt kw + kAk.ka + kcka + k. kJ
(8.10)

393
FLEXURES

where x, and x, are the actuator and flexure platform displacements

respectively, see figure 8.10.
-

Figure 8.10: Generalized 'static' model representing displacements within a single degree of
freedom actuator and flexure mechanism

For an ideal system, the ratio in (8.10) should be unity. Unless elements
having negative stiffness are added (i.e. elements in a buckled state, see chapter
4), it is only possible to approach this ideal as all of the actuator, coupling and
stiffness values in undesired freedoms are large and the flexure stiffness becomes
vanishingly small. Unfortunately, as the stiffness of the flexure reduces in the
desired freedoms, so too will its resistance to off-axis (or parasitic) forces.
Consequently, the flexure will be become increasingly susceptible to parasitic
errors in the drive mechanism. In the limit, the motion of the flexure will
correspond exactly to that at the drive. The only reason for using a flexure in the
first place is to provide a guided motion and so this latter option is not available.
A stiff coupling that provides a high compliance in the off-axis freedoms always
presents a design challenge.
Potentially, a near perfect coupling could be achieved using either an
electromagnetic or hydrostatic drive. Sensing and maintaining a specified gap
between driver and platform often produces the effect of an extremely stiff
actuator with high compliance in the off-axis freedoms. For relative large
devices, such drive couplings have been successfully used to provide precisely
controlled motion, Peirce et al., 1994, Chen et al., 1995. However, for small
flexures, such drives tend to contribute a significant cost while introducing
problems of heat dissipation. For most instrument and machine flexures, the
favored drive is either a direct motor driven screw or, more popularly, the
piezoelectric actuator. Curiously, for much smaller scales in which the flexures
are measured in micrometers, electrostatic drives which in principle dissipate
little to no heat, are commonly used. A variety of couplings between a rigid
actuator and moving flexure stage for intermediate sized mechanisms are shown
in figure 8.11. Generally, these consist of a mechanism that is required to be stiff
in the drive axis while freedom of motion to accommodate errors in the drive. As
might be expected, flexures can be readily designed for this purpose and three
variations are shown in the figure. A simple wobble-pin or sphere on flat is also
commonly used. Jeweled pivots for this purpose are commercially available. A
low friction equivalent to the sphere on a flat is provided by contacting the outer

394
CHAPTER 8: MANuFACTURING AND A SSEMBLY C ONSIDERATIONS

a) Notch or toroidal hinge

Flexure

Single contact or
wobble pin in
jeweled pivot otch or toroidal hinges

U ~ '- o.,.
d)

S- ~==I=;::==1 "~
wu. J
e)
Spring pre-load

Wobble pin in jeweled pivots

f)
Spring pre-load

Figure 8.11: A variety of couplings for connecting a rigid drive to a moving flexure
element

395
FLEXURES

races of two rolling element bearings with their axes at right angles, see Smith
and Chetwynd, 1992. It is important to note that some form of pre-load is
necessary if a contact based coupling is chosen.

8.2.3 Flexure mounts

Having produced a flexure mechanism and integrated it with drivers and other
auxiliary components, it is often necessary to secure this into some larger
instrument mechanism. Generally, this must be secured using bolts and clamps.
Again, good design principles will help to ensure minimal influence of the
fastening mechanism. Useful rules are
Use kinematic or pseudo-kinematic design of the mounting points.
Match symmetry between the flexure and the mount.
Choose mounting points to be symmetric about the centroid of the flexure.
Try to match thermal expansion effects and rate dependent thermal
responses.
Tighten clamps incrementally to ensure equal clamping forces at each
fastening point.
Introduce compliance at the clamp. This can be achieved by either locally
weakening one of the clamping members where it connects with the
bolt or introducing a spring washer (for the same purpose).

8.2.4 Adding damping to flexure systems (extracting energy)

For optimal response, it has already been suggested in chapter 3 that, for many
applications, it is desired that the critical damping ratio, q, lays in the range 0.4 -
0.8. Almost by definition, if a material has been chosen for its favorable
characteristics as a flexure, it is unlikely to provide such a large value of energy
dissipation due to work of distortion. In fact, energy dissipation can only occur
with an accompanied hysteresis in the stress-strain characteristic thereby directly
effecting performance. In this section, internal friction of materials will be briefly
discussed. This will be followed by a discussion of mechanisms for increasing the
energy dissipation in a flexure.
8.2.4.1 Internal friction in solids
All materials subject to cyclic loads will dissipate energy, Wert, 1986.
Unfortunately the physical mechanisms whereby this occurs are not easy to
determine or discriminate. For metals, a number of phenomena have been
identified. These are
Internal friction caused by thermal currents and atomic diffusion
The Snoek effect

396
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSIDERATIONS

1/Q
0.005

0.004

0.001

0~---------r----------r---------~--------~
1000 100 10 0.1
Frequency
Figure 8.12: Internal friction ofthennoelastic origin in vibrating reeds, the dark
squares represent the frequency at which maxima are predicted by the theoretical
model ofZener(reproduced from Wert, 1986)

Zener relaxation
Dislocation damping

Thermodynamic considerations led Carl Zener to the analysis of energy

losses due to temperature gradients developed across the face of a thin beam
subject to cyclic bending. Compression on one face leads to a temperature
increase and vice versa on the tensile side. Subsequent irreversible heat flow then
provides the mechanism for energy dissipation. Depending on the thickness of
the beam, the rate of heat transfer leads to energy loss that will be a function of
the frequency. In his original work Zener showed that the maximum energy loss
for a thin cantilever beam occurs at a frequency

[Hz] (8.11)

where c. is the specific heat per unit volume, K is the thermal conductivity, tis
the thickness of the cantilever and A is constant that depends on the cross-
section.

397
FLEXURES

Subsequent experimental results by Bennewitz and Rotger, 1936, validated

this mechanism, see figure 8.12. Extension of this analysis to include the effects of
heat transfer across crystal interfaces in polycrystalline materials also showed
considerable promise, see figure 8.13. For our purposes, it is informative to
consider the scale in figure 8.12. Being considerably less than unity, this can be
expressed using (3.90) as 2~ (it will be shown shortly that this can be directly
related to hysteresis as a material property). Being less than a few parts in 1,000
the damping ratio is far too low for most flexure applications. It may also be
reasonably assumed that for a limited range of frequencies the damping ratio
remains reasonably constant.
Two other loss mechanisms, due to either interstitial atoms or substitutional
alloys within a crystal, have been extensively studied over the years. The former
was first studied by Snoek for determination of macroscopic strains in the region
of interstitial carbons in bee iron directly relating to the martensite response.
Based on this analysis, it was determined that the internal friction could be
related to the relaxation time. Experimental studies over temperatures ranging
from near zero to around 700 K indicate a good correlation between relaxation
time and diffusion of carbon, with the diffusion coefficient varying by up to
twelve orders of magnitude over this range! Zener studied the effect of diffusion
in substitutional alloys. In particular, it was proposed that, in the unstrained
state, these alloys would form in local clusters. For example , zinc in 30 - 60 brass
will tend to form into pairs imposing a local tetragonal strain. Upon application
of stresses, these will move to a more favorable configuration by diffusion.
Again, this will result in a relaxation of the material that can, in turn, be used as a
measure of the diffusion coefficient.
Another dissipation mechanism, for which verifiable theoretical models have

1/Q 0.1

0.08 ··- ···-·· -

0.06 - ........ -

0.04
poly crystal
0.02

0 100 200 300 400 500

Temperature (C)

Figure 8.13: Grain boundary relaxation as a function of temperature for

single and poly-crystalline aluminum

398
CHAPTER 8: MANuFACTURING AND ASSEMBLY CONSIDERATIONS

yet to be realized, is that due to motion of dislocations. In general, significant

dissipation is often experienced only at high frequencies although a lesser effect
covering a broad range of frequencies is likely in metals. From the above, the
general rules for the avoidance of damping might be
• Use single crystals or poly-crystals of optimal size to minimize
thermal dissipation
• Avoid alloys with high diffusion rates or single elements
• Avoid interstitial additives
Unfortunately, most of the above are not pos'sible for the production of
flexures from common engineering metals. An alternative compromise is
• Reduce the mobility of dislocations, interstitial elements and alloy
compounds
This is a more realistic option and can often be readily achieved after
processing by a suitable heat treatment to increase hardness.

Figure 8.14: Stress-strain characteristic with hysteresis damping.

Arrows on the graph indicate progression of time.

Modeling of the energy dissipation due to internal friction turns out to be

relatively straightforward. Because this is due to hysteresis in the stress-strain
characteristic, it may be assumed that such a mechanism is independent of how
fast the stresses and strains are being applied. Within a lumped model, this can
be inserted in the form of a damping element that is independent of frequency
but provides a force that lags displacement by 90°. There are two ways to
introduce this effect into a mathematical model. Before proceeding, it is
informative to speculate the response of a spring/mass system for which
hysteretic energy losses are significant. If the spring mass system is deflected

399
FLEXURES

slowly and released, it is expected that the deflection will follow that of the stress
strain characteristic of the spring material, figure 8.14. Consequently, at low
frequencies of cyclic load, the displacement due to an applied force will not be
the same as that for a 'perfect' spring. At excitation frequencies near to
resonance, it would be expected that, for small dissipation, similar resonant
phenomena to that for viscous damping should be observed. To derive a
dissipation function that is independent of frequency, consider the models of
figure 8.15. In these, the hysteretic damping is represented respectively as

1. A damping factor that varies as the velocity of distortion divided by the

frequency
2. An additional complex term added to the spring.

For both systems the equation governing motion for a periodic excitation
force may be expressed in either form

or (8.12)
mi+(k +ih}x =Re{Foiot}

As usual, assuming a linear solution of the form

Fe'..
0

c)

Figure 8.15: Mathematical model of a single degree of freedom

spring mass systems with hysteretic damping, a) using a hysteretic
damping element, b) equivalent model using complex stiffness

(8.13)

The frequency response of this single degree of freedom system derived from
either of equations (8.12) is

400
CHAPTER 8: MANUFACTURlNO AND ASSEMBLY CONSIDERATIONS

H(iw) = II m
wn2 -w 2 +ih/m
llk (8.14)

Expressing the above 'normalized' to the displacement for a static load

applied to a perfect spring yields

(8.15)

The phase shift as a function of input frequency can be readily obtained as

(8.16)

The subsequent frequency response is plotted in figure 8.16. From equation

(8.15) it is apparent that the value for 1/Q, which occurs at the undamped natural
frequency, is simply k/h which is the ratio of real to complex stiffness. For solid
materials, this ratio will also be that of the real to imaginary elastic modulus and
is therefore a material property. Consequently, to assess this value it is common
to use a rod of the material as part of a torsional pendulum and quote the
material property in the form 1/Q as is the case in figures 8.12 and 8.13. Extensive
tables of complex elastic moduli have been developed over the years, references
to which can be found in Lazan, 1968, and Snowden, 1968.
Often, the compound elastic moduli in linear and shear loading are denoted
by

E" = E' +iE"

(8.17)
G. =G'+iG"

Also the phase shift at zero frequency is commonly quoted as the loss angle
that, for small values, is given by

401
FLEXURES

100

hlk =0.05

0.1

Q) / (j)n

0
-0.5 1.5 2

-1
., -1.5
~
p.. -2
-25
-3
-3.5

Figure 8.16: Frequency response for a spring mass system with hysteresis
damping, a) magnitude response, b) phase lag

¢ = tan -•IG"/ )
\ / G' =tan
-•(E"/ ) E•/
/ E' ~ / E'
(8.18)
~x
A more extensive discussion of this field of study plus some tabulated values
for materials having high loss angles can be found in Goodman, 1996.
Value of Q for resilient materials are difficult to measure. In practice,
measurements are hampered by the energy losses introduced by the specimen
mounts. Even for non-metal springs of optimal geometry, it is difficult to achieve
Q values of greater than 6,000, Schindel et al., 1997. With great care, and for some
low loss materials such as fused silica, Q values of greater than 107 have been
measured, Startin et al., 1998. However, in the experiments of this reference, it is
considered that the Q value was primarily limited by losses through the
mounting interfaces.

402
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSIDERATIONS

8.2.4.2 Adding damping to a flexure

In much of the above, it has been implicitly assumed that the damping
coefficients are inadequate. Simple analysis of systems for applications such as
vibration isolation, highly linear springs or optimal dynamic response reveals
that structural damping falls in the intermediate region of being either too high
or too low. For systems that are required to follow dynamically varying demands
on position it has already been stated that optimal values are likely to be in the
range ~ = 0.4 to 0.8. To achieve this value it is usually to provide some additional
energy dissipation. There are two techniques commonly used to achieve such a
goal. These being
Active dissipation
Passive mechanisms
Dissipation occurs if there is a force that is out of phase with the position and
acceleration. An obvious method is to derive a signal based on the velocity.

Active dissipation
To provide active dissipation, it is necessary to monitor the velocity of the
moving members and apply a force in direct proportion. Commonly, there are
two means of doing this
Exploiting the velocity directly to produce a feedback force.
Measuring the position and differentiating with respect to time prior to
feeding this back as a force to the moving member
The first of these is often referred to as velocity feedback or, for motor
controllers, tachometer control. Exploiting the Faraday induction is a common
method for both deriving a signal and creating a velocity dependent force. For
example, it might be possible to attach a magnet and coil to the moving and
stationary components of the flexure. Motion of the magnet will produce a
Lorentz force on the electric field in the coil and this can be used to produce a
current or voltage proportional to the velocity. In the former case, it is only
necessary to provide a resistor across the coil, which will then dissipate energy in
the form of i 2 R losses. In reality, the added mass of either the magnet or coil
attached to the moving platform of the flexure will outweigh the benefits of such
a simple, passive isolator. Leaving the coil open circuit, there will be an induced
voltage proportional to the products of magnet flux, length of coil and velocity.
This velocity signal may then be amplified and returned to a force actuator acting
directly on the platform. Liu et al., 1993, were able to used such a technique on a
flexure based stylus mechanism to adjust the critical damping ratio from an
intrinsic value of around 0.02 to values greater than one.
The second approach is to measure the position of the platform, differentiate
it with respect to time and then feed this back to a force actuator (usually

403
FLEXURES

electromagnetic or, less often, piezoelectric in origin). For a perfect system, this
will have the effect of improving the transient response of the flexure. Often, the
derivative is combined with an integrator and proportional feedback for
providing a fast response with zero steady state errors. Known as a PID
(proportional-integral-derivative) controller these individual terms effect the
general response, steady state errors and transient response respectively. A
disadvantage to this approach is the adverse effects of noise on the derivative
term. As frequency increases, so too does the derivative with respect to time.
Consequently, high frequency noise, albeit mechanical or electronic, will tend to
be amplified and fed back to the system, often resulting in reduced performance
in terms of both stability and precision. For a more detailed discussion of closed
loop control strategies see for example, Dorf, 1980, Richards, 1979.

Passive damping mechanisms

Passive damping mechanisms have the advantage that, in comparison to active
mechanisms, they are relatively 'noise free'. By passive mechanisms, it is implied
that some form of mechanical device is to be added, the physical nature of which
tends to dissipate energy due to relative motion between two links of a flexure.
Invariably, this will be achieved by applying forces (thereby inducing relative
velocities) through damping elements connecting these links.
Three techniques are commonly employed, these being
Addition of viscous dampers
Introduction·of interfaces
Coating with high loss materials

These are discussed in tum below.

Viscous dampers
Shear of any fluid or gas at low Reynolds number (defined as the ratio of inertial
to viscous force) will result in a force that is proportional to both velocity and
viscosity. For macroscopic structures involving large forces, it is possible to
utilize viscous damping by sealing a fluid in a hydraulic or pneumatic piston and
cylinder arrangement. Arranging for controlled leakage either through or around
the piston results in a force that is a d~finite function of the relative velocity
between piston rod and cylinder. For smaller instrument mechanisms, problems
due to frictional forces at the seals often results in unacceptable hysteresis. As a
consequence, it is more usual to utilize some form of 'free paddle' arrangement.
Three common types are
The paddle in a bucket
Shear film damping

404
CHAPTER 8: MANuFACTURING AND ASSEMBLY CONSIDERATIONS

Squeeze film damping

The principle of operation of paddles is relatively simple. A fluid is placed in
a container that is both stationary and upright. A paddle immersed in the fluid
and rigidly connected to the moving platform provides the damping forces. Such
mechanisms are common in torsion magnetometers, galvanometers and other
instrument mechanisms with masses relatively low in comparison with the
stiffness. Some care is necessary to arrange that the center of force of the
damping element does not introduce unwanted bending moments into the
mechanism.
Shear damping relies on the relative motion of two parallel, solid surfaces
with a fluid between them. For two parallel surfaces of area A and separation d,
the shear force as a function of the relative velocity v is given by

F = rJAv (8.19)
d

where 17 is the viscosity of the fluid between the two and has units of Ns m·2.
Examples of such mechanisms applied to flexures are given in the papers of
Chen et al., 1995, and Holmes, et al., 1997.
Interestingly, for relative small separations, it is possible that surface tension
forces, independent of orientation will retain the fluid. Consequently, for flexures
it is sometimes possible to place a surface near to the moving flexure and deposit
a small amount of fluid in the gap. Some mechanisms have been known to retain
damping fluid between two vertical, parallel surfaces for tens of years.
Sometimes, it might be arranged that the parallel surfaces change separation,
or, in other words, move in a direction towards or away from each other. Under
these circumstances, any intervening fluids will be physically squeezed out from
the gap. This will require a force that is a function of the ratio of the minimum
dimension of the plates to their separation i.e. the slenderness of the gap. In
general, it is found that the squeeze film force is related by

(8.20)

where dis the instantaneous separation of the plates.

For two parallel plates moving in a direction normal to their surfaces, n is 3
while for a sphere and flat this reduces to 2, Vidic et al., 1997. Alternatively,
squeeze film damping might be utilized for the damping of flexures undergoing
distortion. Under these conditions, it is necessary to solve Reynold equation to
determine the relationship between the squeeze film forces and separation
velocity, see for example Gross, 1962. Even when the intervening fluid is air,
shear and squeeze film effects can produce significant forces, particularly in
MEMS, Muller and Howe, 1990, and some small instrument mechanisms,

405
FLEXURES

Howard and Smith, 1992. Because of this near inverse cube dependence on the
nominal separation, this can dramatically increase the damping ratio in small
mechanisms, often to values considerably greater than unity. The effects of
squeeze film damping between a cantilever with a small mass attached to the
free end and a solid base parallel and near to one face has been assessed by Xu
and Smith, 1995. Significant squeeze films can also develop in larger structures if
the viscosity of the intervening fluid is high or surfaces conform closely. For
example, laying one large glass sheet on another can result in a squeeze film of
air that, under self-weight loads, is maintained for several seconds. Because the
lateral friction is very low, such an effect can be hazardous if the plate is
unconstrained.

8.2.5 Effects of manufacturing tolerance on the stiffness of flexure elements

Intrinsic to all manufacturing processes is the unpredictable deviation between
the desired and actual geometry of parts. Commonly, this is simply accounted
for in the tolerance. More often than not, tolerances represent worst case values
encompassing the complete scatter of differences over all possible geometry's for
a given process (i.e. this might be ±0.05 mm for a milling machine on all
dimensions for any component that might be produced in one operation on the
machine). Quite often, the tolerance arises because of hysteresis in a machine tool
drive and slideway causing 'dead' bands during reversal. If there are significant
Abbe errors between the line of action of the tool and the measuring scale used to
monitor slideway position, it is not possible to detect, and therefore compensate,
these errors, see for example Slocum, 1992. Consequently, this tends to remain
relatively constant and is independent of the size of components to be produced.
Other possible errors such as undetectable parasitic rotations of the slideway,
thermal expansion effects and tool wear tend to be relatively small in
comparison. It is possible to reduce hysteresis using software compensation.
However, because the direction of the backlash depends on the history of the
motion and instantaneous friction coefficients, it is not possible to accurately
predict, and therefore eliminate, this error (i.e. reduce it to a value considerably
lower than other error sources). In this section, the effects of manufacturing
tolerances on the stiffness of flexure elements are discussed.
Predominantly, flexure action is provided by deflection of a beam element or
a locally notched region in a solid. For small deflections and applied loads, it is
often possible to derive an approximate expression for the stiffness, k, of such an
element by

Eb"tv
koc - - (8.21)
L"'

To determine the influence of tolerances in the above, it is possible to use the

chain rule of differentiation to estimate the change in stiffness from

406
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSIDERATIONS

& = (!;)oE+(:: )&+( ~~ }r +(:~)oL

b"t• Eb•-•t• Eb 11 t•-• Eb"t•
=- -oE+u Ob+v 01-w--1- oL (8.22)
L.. L.. L"' L...
Eb"t• oE Eb"t• Ob Eb"t• 01 Eb"t• oL
=- --+u-- - + v - - - - w- --
L" E L" b L,. t L" L

The common multiplier in the last of equations (8.22) is simply the original
stiffness value for a perfect flexure. Consequently (8.22) can be rearranged to
produce the fractional uncertainty in stiffness

ok oE Ob 01 oL
- =- +u-+v--w- (8.23)
k E b t L

Each of the numerators in the fractions on the right hand side of this
equation represents the uncertainty of that particular parameter. To assess the
influence of manufacturing tolerance on the predicted stiffness of the flexure, it is
only necessary to determine the exponents u, v, w in (8.21). For example, the
stiffness of a leaf spring subject to uniaxial tension is given as Ebt/L, from which
it is readily apparent that u = v = w = 1. For a few flexure elements the
appropriate exponents are given in Table 8.2.5.1 below
Table 8.2.5.1: Exponents of equation (8.21) for a variety of flexure elements
undergoing small distortions (in the case of a notch hinge R replaces L)
Flexure Distortion Applied load u v w
type mode
Beam Axial Axial tension 1 1 1
Lateral Applied moment 1 3 2
Lateral Applied force 1 3 3
Lateral Uniformly 1 3 4
distributed
Angular Bending moment 1 3 1
Notch Angular Bending moment 1 5/2 1/2
Lateral Lateral force 1 5/2 5/2

Equation (8.23) indicates that the fractional variability of the stiffness is

directly proportional to the fractional uncertainty of the individual parameters
multiplied by the exponent in the original equation for predicting the stiffness. It
is worth spending a little time considering the implications of this in view of the
errors most likely to dominate during manufacturing.
Taking, for example, a leaf-type hinge element it is worth speculating the
relative contribution of the errors that are likely to be encountered. Surprisingly,
the first of these, elastic modulus, is difficult to know accurately. Usually, the

407
FLEXURES

error is of the order 1 % or more with higher accuracy often reqwrmg

considerable effort and expense. However, for material from the same batch the
value of the elastic modulus is likely to be constant to considerably better than
this. Similarly, the depth of the flexure is often dependent upon the parallelism
between the front and back surfaces of the material from which the flexure
elements are produced. Alternatively, this will depend upon the thickness of the
strips from which the hinge is fabricated. In most cases this dimension is
relatively large and is likely to be both accurate and constant. In view of the low
value of the exponent, this parameter is unlikely to be a source of significant
errors for commonly used manufacturing processes. In contrast, the last two
parameters can have relatively small dimensions and large exponents.
Consequently, if manufacturing tolerances are constant and independent of
dimensions, the ratio of tolerance to dimension can be relatively high. For
example, it is not uncommon for leaf or notch hinges to require thickness values
of the order of a few hundred micrometers. A manufacturing tolerance of 10 J.lm
on a thickness of 100 J.lm represents an uncertainty ( t5t It) of 10 %. In view of the
exponents given in table I, this can result in stiffness variations of around 25 - 30
%.
As is clear from equation (8.23), the variation in stiffness is simply the sum of
scaled variations for each parameter. The negative coefficient in w simply
indicates that stiffness will reduce with increasing L. Of more importance for
assessing the consistency of products is an estimation of the variance of the
stiffness. It might be speculated that this may be obtained by taking the expected
value (indicated by the quantities inside () ) of the square of (8.23) given by

((~)')=((~ +u~ +v~ -w~)')

=(( ~)}·'((~ )}v'(( ~)}w'(( ~J) (8.24)

Adopting standard statistical notation and denoting the standard deviation

by the symbol u (not to be confused with the stress!), equation (8.24) can be
expressed as

(8.25)

The subscripts in this equation refer to the ratios and therefore the variances,
u , are dimensionless. Choosing this notation, it is important to note that the
2

tolerances must be specified in terms of the standard deviation measured about

408
CHAPTER 8: MANuFACTURING AND ASSEMBLY CONSIDERATIONS

the mean. The symbol p represents the correlation coefficient between the
appropriate variables indicated by the subscripts.
In many instances, each individual parameter is likely to be independent of
the others. In this instance the correlation coefficients are zero and (8.25) reduces
to

(8.26)

One particularly troublesome consequence of variable stiffness hinges within

a flexure is the disturbance of the symmetry of the structure. This asymmetry of
forces within the mechanism can result in significant deviations in the reaction of
the flexure to applied forces. In many cases, although they will still exhibit a
linear response between displacement and applied force, this difference between
the direction of applied force and motion might introduce significant errors in
terms of the function of the mechanism. An analysis of these effects is presented
in the paper of Ryu and Gweon, 1998.
The variability of the stiffness may also introduce uncertainty in the
operation of the mechanism. In particular this may lead to considerable
variability in the work required to produce a given displacement. For many
flexure designs comprising many hinges a truly random variability of the hinge
stiffness will tend to average towards the design value. For example, for the
simple and compound rectilinear spring mechanisms discussed in section 5.4 for
which each hinge is subject to nominally similar strains, the mean stiffness of
hinges will tend towards the true value with the variance of the mean u i given
by

(8.27)

where j is the number of hinges and the variance of the mean hinge value
becoming normally distributed in accordance with the well known central
tendency of statistical distributions.
In some cases, it may not be reasonable to assume that there is no correlation
between geometric parameters of a hinge. As an example, for a notch hinge
produced by making two nearly 'touching' holes variations in the radius will
directly effect the thickness of the web. In this case there will be a unity
correlation coefficient connecting the two parameters. In reality, the radius of the
notches will tend to be relatively constant while the web thickness will be
effected by more significant variations in the center location of the holes plus the
undesirable effects of lobing of the holes common during drilling operations.

409
FLEXURES

8.2.6 Typical flexure drives

Some form of drive must be included in the design of any flexure mechanism.
Although outside of the scope of this book, drive techniques are intrinsically
linked to the flexure design process. Consequently, to indicate possible design
solutions, some of the more common drives and their relative merits and
limitations are listed below, see also Smith and Chetwynd, 1992,
Feedscrews. These provide an inexpensive, relatively stiff and precise
drive for long (potentially unlimited) range motions of larger scale
flexures. Backlash and non-linear stiffness for small displacements
can produce significant errors upon reversal of motiori, thereby
limiting performance.
Friction drives. A friction drive is achieved by gripping a prismatic rod
between two rollers. Rotation of the rollers is converted to a linear
translation of the prismatic bar. Such a drive can have stiffness
comparable to a feedscrew with significantly reduced backlash. A
drawback is the requirement for precise rotational control of the
motor drive. Transmission of rotational motion through contact
between the outer rims of smooth disks is also common.
Piezoelectric. These are commonly used for fine motion stages because;
they are stiff, can be directly voltage controlled, produce a fixed
displacement almost independent of size (and therefore are favorable
for small devices), have a relatively fast dynamic response, have a
low thermal expansion coefficient and can operate with a high
preload. Disadvantages include hysteresis, creep, sensitivity to
thermal variations and internal losses that generate heat. High (0-
1500 V) and low (150 V) voltage materials as well as electro-strictive
devices are commercially available. Longer motions are often
obtained by incorporating flexure based lever mechanisms into the
drive. Less commonly, translation range can be increased using
'bimorph' designs in which two piezoelectric beams are bonded to
produce a 'bi-metallic' mode translation when they are actuated in
equal and opposite directions.
Electromagnetic. When applied directly, electromagnetic actuators can
be categorized into four groups, these being

The solenoid and soft magnetic target

Coil in a field produced by a permanent magnet or 'voice coil'
Magnetostatic
Electrostatic

410
CHAPTER 8: MANUFACTURING AND ASSEMBLY CONSIDERATIONS

In general, the above might be considered to be ordered from largest to

smallest force generation mechanism and a very brief overview of the
relative merits and limitations of each is presented below.
Usually the solenoid is mounte'd to the base of the flexure and will
consist of a copper winding surrounding a soft magnetic material to
provide a low reluctance path for the magnetic field. The flux path is
then broken by an air gap with another magnetically permeable
'target' being on the other side of the gap and attached to the moving
platform. To estimate the force, F, on the target it is necessary to
compute the inductance, L, of the magnetic circuit. This can be
computed from the equation

L = N 1{1 =~ fJJB.Hdvol
I I 1101
N2

where p represents the permeability of the soft magnetic core/target

and the gap respectively, A the respective cross-sectional area across
the flux path, N the number of turns of the solenoid, B is the
magnetic flux density, H is the applied magnetic field, I is the
excitation current and I is the path length of the magnetic flux. When
the permeability of the core material is significantly greater than that
of air and for reasonably large air gaps, the above reduces to

Force on the target can be computed from the gradient of the

magnetic potential, U, given by

Although relatively large forces can be generated using such a

design, it is non-linear in the length of the gap and the permeability
will vary with magnetic field and temperature and exhibits hysteresis.
However, for systems under closed loop control, these effects may not
be particularly troublesome, Holmes et al., 1999:

411
FLEXURES

Potentially, a more reproducible force can be generated with a

voice coil design. Two designs are commonly used to drive flexure
mechanisms. Probably the most common electromagnetic flexure
drive exists in the loudspeaker. In this, An arrangement of permanent
magnets is positioned around a coil so that its windings are subject to
a uniform and constant magnetic flux of density B. Under these
circumstances, when energized with a current, I, the coil will
experience a force, F, given by

F oc BIZ

l is the length of the winding within the flux path and in this it has
been assumed that winding, flux and force are mutually
perpendicular. Of particular interest with such a design is the fact
that, at least from this simple analysis, the force is linearly
proportional to the coil current and is independent of the relative
position between the coil and permanent magnets.
Often, to reduce mass, the winding is attached to the moving
platform of the flexure. In some applications, this may produce an
undesirable heat source. One solution, might be to make a coil
consisting of two opposing windings and generate the coil magnetic
field by differential currents operating at constant power. However,
the extra winding will add mass. Alternatively, it is sometimes
desirable to attached the permanent magnet to the moving platform
and surround this by a winding that is rigidly attached to the base, the
moving magnet design. For a circular cylindrical winding
surrounding a permanent magnet, the maximum force on the magnet
is experience when it is positioned with poles along the coil axis and
these being equidistant about the ends of the coil, Smith and
Chetwynd, 1990. In this case, the force on the magnet in the direction
of the coil axis can be computed from

For suitably chosen coil and magnet geometry such an actuator can
again be designed to provide a force that is linear with current and, to
first order, is independent of relative coil/magnet position.
Performance characteristics of both of the 'voice coil' designs will
vary with temperature and magnitude of the applied field. However,
for ' hard' (usually rare earth) magnets subject to .relatively small fields
non-linearity and hysteresis may be negligible.

412
CHAPTER 8: MANuFACTURING AND ASSEMBLY CONSIDERATIONS

If the permanent magnets of the above design were replaced with a

'perfect' magnet, then the above voice coil designs would provide a
correspondingly perfect linear relationship between coil current and
force. A winding of fixed geometry through which constant current
flows can produce an effectively perfect magnet. Such 'perfect' designs
are currently being used to compare electrical to mechanical power,
Olsen et al., 1991 . .
Electrostatic drives utilize the force produced between two
electrodes separated by a small gap (usually air) with a potential
difference across them. Forces can be easily derived from the
derivative of the electrostatic potential computed from the product of
capacitance and potential difference divided by 2, see for example
Smith and Chetwynd, 1992. For mechanisms of typical engineering
dimensions, such forces are invariably small. However, at small scales
such as those typically used for MEMS devices significant forces may
be produced and, in many cases, this is the preferred drive mechanism.
Magneto-strictive. Manufactured from alloys based on terbium, iron and
dysprosium, this material is often called Terfenol. Suitably pre-
stressed, this material will undergo a strain when subject to an
applied magnetic field. Such actuators appear favorable for relatively
large, high-force drives.
Hydraulic. Direct drive mechanisms employing hydraulically driven
pistons can produce very large amounts of work in reasonably small
volumes. They differ from piezoelectric and electromagnetic systems
in that the pressurized fluids can be generated remotely. Hence
hydraulic actuators can produce these large amounts of work with
relatively low heat generation at the point of actuation. A current
drawback is the expense required for the generation and control of
high pressures and sealing of the fluids is always problematic.
Poisson's ratio or 'ballooning'. For solid chambers, the problem of
sealing is obviated. Application of a pressure change to a pressure
vessel will result in corresponding strains and subsequent motion of
points on the surface. Two methods for achieving fine motion over
small ranges have been employed. One consists of a simple cavity
that, upon application of internal pressure change, results in relative
motion of walls on opposite sides of the cavity. The other method
may be visualized by a device consisting of a cylindrical cavity with a
solid cylinder running through the center and a small clearance in
which the pressurized fluid can be fed. Application of pressure to this
fluid will impose a uniform stress on the central solid. By Poisson's
ratio this solid rod will extend in the axial direction. Both techniques

413
FLEXURES

can provide stiff actuators for short-range, fine motion control.

Similar costs of pressure generation and control exist as for the
aforementioned.
Pneumatic. Pneumatic (mainly air) driven actuators have the advantage
that reasonably high-pressure supplies are readily available in most
manufacturing environments in the form of a mains supply of in
pressurized canisters. Although the pressures are relatively low
compared with hydraulic systems, they can exhaust to the
environment without pollution or mess. The compressibility of gases
means that heat will be generated whenever irreversible work is done
on the system (it is always irreversible).
Thermal. Changing the temperature of a system element, often by
passing a current through it, w ill result in both thermal softening and
expansion with changing temperature. Using such an element to pre-
load a flexure so that it is in tension produces actuation, the effect of
temperature resulting in a change of the pre-load. Such designs are
relative slow for large-scale devices, are asymmetric in response (heat
can be 'driven in' to the material at a different rate than it can be
extracted) and, by definition, produce heat.

8.3 Machining and heat treatment of some common flexure materials

By definition, optimal materials for flexures should have a relatively high elastic
limit. For metals, this corresponds to the material being in its hardened state.
However, this often precludes machining of the material using single point
cutting (turning and milling) processes. This is unfortunate because these
machine tools represent
3000 some of the fastest and
cheapest methods for
2500
material removal. In many
l2000 cases, there are direct cost
~ benefits if the material can
~ 1500
be free machined in an
j
1000 annealed state and
subsequently hardened.
For steels, machining
0+-----------r---------~r----- technologies have existed
0 0.5 for well over a century and
Carbon(%) numerous texts adequately
cover the machining
Figure 8.17: The hardness of steel as a function of carbon characteristics of this
content material.

414
CHAPTER 8: MANuFACTURING AND ASSEMBLY CONSIDERATIONS

8.3.1 Steels

For the manufacture of precision flexures, it is desirable to use steels with the
maximum difference in hardness before and after heat treatment It is also
necessary that the hardness of the material in its softened state is sufficiently
hard so that it will machine relatively freely. In general this corresponds to steel
having a carbon content greater than 0.3 % by weight and, where possible, close
to the critical point at 0.83%, see figure 8.17. However, with such high carbon
content this material tends to be both difficult to machine and may contain large
residual stresses at the machined surfaces.

8.3.2 Beryllium copper

In its annealed state beryllium copper can be readily machined using either
carbide or high-speed steel tools. For flexure applications, it is usual to select a
high strength alloy with the addition of lead producing a 'free' machining alloy
without effecting its mechanical properties. Precipitation hardening of this
material is achieved by heating below a phase transition temperature of around
1100 F. for high strength applications it is important to use a material that has
undergone significant cold working. This material can be readily annealed by
heating above 1300 F, which results in a phase transition involving dissolution of
the beryllium into the copper to form an a phase. Rapid quenching effectively
freezes the material in its soft state. Heating in air will lead to discoloration from
a copper color to dark brown.
After machining, the component can be hardened by heating to a
temperature below the phase transition and leaving it at that temperature for a
specified time period. As an example, a typical age hardening might require
heating to a temperature of 700 F for 1 hour. Heating at too high temperatures or
for excessive time periods can result in over-aging of the material resulting in a
reduction in strength. For example, 800 F the optimal hardening time might
reduce from 1 hour at 700 F down to 10 minutes•. High strength alloys have good
fatigue resistance with 107 cycles endurance stresses of up to 600 MPa being
possible.
As an aside, it is noted that clean beryllium copper readily glues to ceramic
surfaces and can serve as a reasonably good conductor of electrical signals (high
conductivity alloys are available but these have less desirable properties for
flexure applications).

• A complete guide to berylliwn copper can be obtained from Brush Wellman Inc., 17876 St. Clair Ave.,
Cleveland, Ohio 4411 0.

415
FLEXURES

References
Bennewitz Von K. and Rotger H., 1936. Uber die innere reibung fester korper;
Absorptiosfrequenzen von metallen im akustischen gebiet, Phys. Z, 37, 578 -
57~588.

Chen Kuo-Shen, Montiero A., Trumper D.L., Smith S.T., and Williams M.E., 1995,
Spring dominated regime design of a high load capacity, electro-
magnetically driven X-Y-e Stage, Proc. ASPE, 12,199-202.
Dorf R.C., 1980, Modern Control Systems, Addison-Wesley, Ma.
Goodman L.E., 1996, Material damping and slip damping, in Shock and Vibration
Handbook, 4th edition, ed. Harris C.M., McGraw-Hill, NY, chapter 36.
Gross W.A., 1964, Gas Film Lubrication, J. Wiley and Sons, Inc., NY.
Guckel H., Skrobis K.J., Christenson T.J., Klein J., Hans S., Choi B., Lovell E.G.
and Chapman T.W., 1991, Fabrication and testing of the planar magnetic
micromotor, J. Micrmech. Microeng., 1, 135-138.
Holmes M., Trumper D. and Hocken R.J., 1999, Long range scanning of an
electro-magnetically suspended platform with shear damping, to be
published.
Howard L.P. and Smith S.T., 1992, Long range constant force profiling for
measurement of engineering surfaces, Rev. Sci. Instrum., 63(10), 4289-4295.
Kalpakjian S., 1995, Manufacturing Engineering and Technology, 3•d ed., Addison-
Wesley Publishing Co. Inc., Ma.
Lazan B.J., 1968, Damping of Materials and Members in Structural Mechanics,
Pergamon Press, London.
Liu X., Chetwynd D.G., Smith S.T. and W. Wang, 1993, Improvement of stylus
measurement fidelity by active damping control, Meas. Sci. Technol., 4(12),
1330-1340.
Muller R.S. and Howe R.T., 1990, Technologies for microdynamic devices,
Nanotechnologt;, 1, 8 - 12.
Olsen P.T., Williams E.R. and Elmquist R.E., 1991, Monitoring the mass standard
via the comparison of mechanical to electrical power, IEEE Trans. on Inst. and
Meas., 40(2}, 115-120.
Peirce S.B., Tran H., Wiedemann M. and DeBra D., 1994, Quiet hydraulics for
ultraprecision actuation, Proc. ASPE, 9, 20-25.
Petersen KE., 1982, Silicon as a mechanical material, Proc. IEEE, 70(5), 420-456.
Richards R.J., 1979, An Introduction to Dynamics and Control, Longman, London.

416
CHAPTER 8: MANuFACTURING AND ASSEMBLY CONSIDERATIONS

Ryu J.W. and Gweon D.G., 1998, Error analysis of a flexure hinge mechanism
induced by machining imperfection, Precision Engineering, 21(2/3), 83-89.
Schindel D.W., Hutchins D.A. and Smith S.T., 1997, A study of materials at high
temperature using miniaturized resonant tuning forks and non-contact
capacitance transducers, J. Acoust. Soc. Am.,102(3), 1296-1309.
Slocum A. H., 1992, Precision Machine Design, Prentice Hall, NJ.
Smith S. and Dvorak D., 1998, Tool path strategies for high-speed milling
aluminum workpieces with thin webs, Mechatronics, 8, 291-300.
Smith S.T. and Chetwynd D.G, 1990, Optimization of a magnet/ coil force
actuator and its application to linear spring mechanisms, Proc. Inst. Mech.
Engrs., 204(C4), 243-253.
Smith S.T. and Chetwynd D.G., 1992, Foundations of Ultraprecision Mechanism
Design, Gordon and Breach, London.
Snowden J.C., 1968, Vibration and Shock in Damped Mechanical Systems, J. Wiley
and Sons Inc., NY.
Startin W.J., Beilby M.A. and Saulson P.R., 1998, Mechanical quality factors of
fused silica resonators, Rev. Sci. Instrum., 69(10), 3681-3689.
Tlusty J., Smith S. and Winfough W.R., 1996, Techniques for the use of long
slender end mills in high-speed milling, Annals of tire CIRP, 45(1), 393-396.
Vidic M., Harb S. M. and Smith S.T., 1997, Observations of contact measurements
using a resonance based touch sensor, Precision Engineering, 22(1), 19-36.
Wert C.A., 1986, Internal friction in solids, f. Appl. Phys., 60(6), 1888 -1895.
Wilson F.W. and Cox R.W., 1965, Machining tire Space-Age Metals, The American
Society of Tool and Manufacturing Engineers, Michigan.
Xu Y. and Smith S.T., 1995, Squeeze film damping of a cantilever beam with a
concentrated mass on the free end, Precision Engineering, 17(2), 94-100.

417
Author index
Christenson T.J., 416r
A Cox R.W., 417r
Abbe E., 12
Agius J.G., 327, 331r D
Alemanni M., 283r Dale J.S., 219r
Andoh T., 283r Davis R.S., 372r
Arya A.P.,280, 284r Debra D., 416r

B
Den Hartog J.P., 144, 151r
Deresiewicz H., 343, 372r
Badami V.G., 219r, 327, 332r Deslattes R.D., 283r
Bamford R.M., 283r Dieter G.E., 19, 56r
Becht c., 318, 332r Donnell L.H., 319, 332r
Becker P., 283r Dorf R.C., 404, 416r
Beilby M.A., 417r Duarte R., 218r
Beltrami E., 36 Dvorak D., 417r
Bennewitz K., 398, 416r
Berliner Y.l., 319, 332r E
Bevan T., 247, 284r Eastman F.S., 158, 218r
Blevins R.D., 146, 151r Elmquist R.E., 416r
de Bono E., 7, 13r Ertas A., 6, 13r
Boyer HE., 39, 56r Estler W.T., 13r
Bosch J.A., 213, 218r Evans C.J., 10, 13r

F
Brentnall W.D., 19, 56r

c Feynman R.P., 65, 151r

Ford H., 31, 56r
Carlson H., 2, 13r
Cauchy A., 1 Frocht M.M., 181, 203, 218r
Chapman T.W., 413r Fu J., 4, 13r, 371, 372r
Chen K-S., 394, 405, 416r Furukawa E., 283-4r, 338, 372r

G
Cheru.k uri H., 293, 332r
Chetwynd D.G., ,5, 6, 14r, 56r, 173,
219r, 229, 237, 258, 285r, 327, 332r, Galileo, 1
335, 337, 372r, 385, 396, 410, 413, Geary P.J., 2, 13r
416-7r Gilsinn D., 284r
Choi B., 416r Glaser R.J., 283r
FLEXURES

K
Goldstein H., 66, 151r, 263,280, 284r
Goodier J.N., 32-33, 53, 56r, 181, 218r,
225, 285r Kalpakjian S., 380, 416r
Goodman L.E., 402, 416r Kelvin Lord, 1, 369
Gross W.A., 405, 416r Keown R.A., 233, 285r
Guckel H., 383 416r King T., 284r, 338, 372r
Gweon D-G., 284r, 409, 417r Kirchoff G.R., 136

H
Klein J., 413r
Kyusojin A., 284r
Hadfield D., 371, 372r
Hamada M., 318-9, 322 332r L
Hans S., 416r Lagrange J.L., 74
Harb S.M., 285r, 417r Lanczos C., 68, 152r
Haringx J.A., 193, 198, 218r, 319, 322, Lawn B.R., 21, 56r
332r Lazan B.J., 401, 416r
Harrison J., 1 Leighton R.B., 65, 151r
Hart M., 284r, 344, 372r Levy R., 298, 300, 302, 332r
Haug E.J., 280, 285r Ling C.B., 179, 181-2, 218r
Hazen R.M., 35, 56r Linley F.M., 327, 332r
Helmholtz H., 1 Liu X., 403, 416r
Henky H.,36 Longair M.S., 65, 151r
Hertzberg R.W., 38, 40, 56r Love A.E.H., 1, 14r
Hildebrand F.B., 140, 151r Lovell E.G., 416r
Hocken R.J., 13r, 219r, 416r
Hojo T., 284r, 372r M
Holmes M., 405, 411, 416r Mabie H.H., 233, 285r
Hooke R., 1, 15 Maltbaek J.C., 134-135, 152r
Horrell R., 287, 332r Mana G., 283r
Howard L.P., 4, 13r, 327, 332r, 371, Mandelbaum A., 1r
372r, 406, 416r Mariotte E., 1
Howe R.T., 405, 416r Marsh D.M., 345, 372r
Howells M.R., 193,198-9,203, 218r, Maxwell J.C., 1, 370, 372r
284r McGill R., 218r
Huber M.T., 36 Miller J.A., 219r
Hutchins D.A., 417r Mindlin R.D., 343,372r

J
Mischke C.R., 19, 56r
von Mises R., 36
Jackson D., 56r Mizuno M., 283-4r, 338, 372r
Johnson K.L., 343, 372r Mohr0.,29
Jones F.D., 8, 13r Montiero A., 416r
Jones J.C., 6, 13r Moon K.S., 284r
Jones R.V., 2, 13r, 176, 218r, ~84~ de Moivre A., 109
Joule J.P., 1

420
AliTHOR INDEX

Moriya T., 283r R<:>tger H ., 398, 416r

Muller R.S., 405, 416r Routh E.J., 103
Muralidhar A., 219r Ryu J.W., 284r, 409, 417r
Muranaka Y., 283r

N
s
Saint Venant B., 1
Nashimura K., 284r Sagawa D., 284r
Neale M., 287, 332r Sands M., 65, 151r
Needham P., 287, 332r Sarid D., 327, 332r
Newell D.B., 372r Saulson P.R., 417r
Newland D.E., 107, 152r Schindel D.W., 402, 417r

0
Scire F., 284r
Seely F. B., 310, 332r
Olsen P.T., 372r, 413, 416r Seyfried P., 283r

p Sherratt F., 38, 42, 56r

Shigley J.E., 19, 56r, 287, 332r
Paros J.M., 177, 180, 185, 190, 212, Shockley W.F., 146, 152r
218r Siegert H ., 283r
Pars L.A., 280, 285r Skrobis K.J., 413r
Patterson S.R., 332r Slocum A.H., 5, 14r, 406, 417r
Pedrotti G., 283r Smith J.O., 310, 332r
Peirce S.B., 394, 416r SmithS., 376, 417r
Pereira P.H., 213, 219r Smith S.T., 5, 6, 14r, 21, 56r, 173, 183,
Petersen R.E., 181, 219r 219r, 226, 229, 237, 258, 285r, 297,
Petersen K.E., 382, 416r 327, 332-3r, 335, 337, 372r, 385,
Phillips J., 55, 56r 396, 406, 410, 412, 416-7r
Pilkey W.O., 41, 56r, 182, 213, 219r, SnowdenJ.C.,401, 417r
306,332r Snyder J.J., 335, 372r
Plainevaux J.E., 172, 219r Sobel D., 1, 14r
Poisson S.D., 24 Spain I.L., 35, 56r
Popov E.P., 26, 56r Speake C.C., 372r
Startin W.J., 402, 417r
Q Steiner R.L., 372r
Strona P.P., 283r
Quinn T.J., 372-2r
Strutt J.W.S. (see Rayleigh)
R Sydenham P.H ., 2, 14r, 284r

Rayleigh Baron, 9, 14r, 74, 103, 121,

128, 130, 134, 152r,279,285r
T
Takenouchi Y., 283r
Reinholtz C. F., 233, 285r
Taylor B., 65
Richards R.J., 404, 416r
Teague E.C., 284r
Rivin E., 5, 14r
Terada K, 283r, 372r
Rostoker W., 19, 56r
Thomas T.R., 327, 333r

42 1
FLEXURES

Thompson W. (see Kelvin) Zahavi E., 38, 42, 56r

Thorpe A.G., 156, 159, 167, 219r Zosi G., 283r
Timoshenko S.P., 1, 14r, 32-33, 36, 53,
56r, 130, 132, 134, 152r, 225, 285r,
303, 332~3r
Tlusty J., 376, 417r
Torbilo V., 38, 42, 56r
Tran H., 416r
Trumper D., 416r
Tsuda T., 318, 332r

v
Vidic M., 405, 417r
Vikhman Y.L., 319,332r

w
Wahl A.M., 2, 14r, 2~9, 333r
Wang W., 416r
WeberW,1
Weidemann M., 416r
Weisbord L., 177, 180, 185, 190, 212,
218r
Wert C.A., 397, 417r
Whittrick W.H., 193, 219r
Williams E.R., 371, 372r, 416r
Williams M.E., 413r
Wilson J.F., 318, 333r
Wilson F.W., 417r
Winfough W.R., 417r
Woody S.C., 297, 333r

X
Xu Y., 219r, 406, 417r
Xu W., 284r, 338, 372r
y
Young D.H., 130,132, 134, 152r
Young R.D., 284r
Young W.C., 193, 198, 207, 219r, 289,
294,333r

z
422
Subject index
clamped-free, 126

A
clamped-free with a rigid mass at
the free end, 127
free-free or fixed-fixed, 126
Abbe error, see Errors Raleigh's method, 128
Adhesives, 377 's' shaped deflection
Assembly, 388 Raleigh's method, 133
adding stiffeners to leaf type Bellows, 317
flexures,392 axial stiffness of rectangular, 322
flexure clamps, 389 's' shaped distortion, 325
flexure mounting, 396 stiffness of, 319
tolerances, effect of, 391 torsional stiffness of rectangular, 320
types of, 318

B Beryllium copper
heat treatment, 415
Beams, bending, 43 et seq. machining, 415
bending equation, 43,45 Brittle materials, 20
boundary conditions, 49 failure, 21
deflection, 45 proof testing, 22
flexural rigidity, 45 Weibull modulus, 21

c
moment, shear force and rate of
loading relationships, 48
see also Leaf type flexures
sign convention, 46 Cartwheel hinge, 199,368
singularity functions, 49 center shift, 202
tabulated, 58 stiffness of, 201
Beams, vibrations of stresses in, 203
lateral, 129 Centros, see Virtual center
cantilever beam with rigid mass
attached at the free end, (Raleigh's Coil spring, 288
method), 134 see also Materials utilization
free-free or clamped-clamped, 131 stiffness, 289
hinged beam with a central mass, stresses, 289
(Raleigh's method), 135 Conventional machining, 374
hinged-hinged,130 end mill geometries, 375
longitudinal, 124 fixturing
FLEXURES

alloys, 379 Design, 6 et seq.

low temperature metals, 377 analysis, 7, 12
waxes, 377 assumptions, 12
notch hinges, 377 brainstorming, 8
thin walled section,375 cancellation, 10
Coordinates, compensation, 10
generalized, 72 concepbualization,7
transformations, 268 correction, 10
Cosine error, see Errors inversion, 7
Couplings, drive, 392 Ockham's razor, 8
see also Drives performance measures, 6
Critical damping ratio reduction, 7
definition, 81 sketching, see Sketching
half power points, 83 symmetry, see Symmetry
measurement of, 82 synthesis, 6
real and imagjnary, 85 Diamond grinding, 385
optimal values, 81 Disc coupling, 291
see also Quality factor inner to outer rim, 293
stiffnesses, 293
Cross strip pivot, 193 stresses, 296
center shift, 198 outer rim, 297
Euler buckling, 196 angular stiffness, 304
stresses, 197 assumptions,297
stiffness, 194 axial stiffness, 306
Cruciform hinge, 204 distortions, 303
stiffness, 205 see also Rotationally symmetric leaf
type hinge
torsional stiffness, 306
D Drives
coupling to flexure, 393
D' Alembert's principle, 68, 71 effect on natural frequency, '137
Damping electromagnetic, 410
active dissipation, 403 feedscrew, 410
adding to flexure systems, 396 friction, 410
hysteresis, 19, 400 hydraulic, 413
internal friction, 396 magneto-strictive, 413
modeling, 400 piezoelectric, 410
passive, 404 pneumatic, 414
see also Critical damping ratio Poisson's ratio, 413
shear, 404 thermal, 414
squeeze film, 405 types of coupling, 394
viscous, 404 Ductility, 17
Deflection of beams, see Beams, Dynamics
bending of general, 280

424
SUBJECT INDEX

transformations, 280 et seq. Fatigue, 37 et seq.

planar mechanisms, 251 damage,42
case study, 274 effect of mean stress, 42
compound rectilinear spring, 252 effect of notches, 40
comments on_ 278 endurance limit, 38
coordinate systems, 265 tabulated values, 57
coupled two-axis flexure, 259
Gerber parabola, 42
double compound rectilinear spring,
Goodman criteria, 42
256
effect of levered drive, 258 high cycle, 38
general model, 263 low cycle, 38
notation, 266 notch sensitivity, 41
transformations, 268 Palmgren-Miner equation, 43
see also Levers S/N curves, 37, 39
strength reduction factor, 40

E stress concentration factor, 40

Filar suspension, 369
Eigen analysis, 94, 106 et seq. Fixtures, 377, see also Conventional
conservative systems, 108 machining
interpretation, 113 Flexural rigidity
non-conservative systems, 112 beams,45
poles, 98, 107 plates, 136
values, 94, 107 Flexures
vectors, 108 advantages, 2
Elasticity, see Modulus of elasticity definition,x,2
Electro discharge machining disadvantages, 3
plunge,379 Leaf type mechanisms, 156, see also
see also Wire electro discharge Leaf type flexures
machining list of source literature, 283
Electroplating, 384 see also Assembly
Errors see Bellows
Abbe, 10 et seq. see Cartwheel hinge
definition, 12 see Cross strip pivot
cosine, 10 see Cruciform hinge
propagation, 407 see Disc coupling
separation, 10 see Four bar link
see Notch type flexures
see Two axis hinges
F Force loops, see Loops
Failure, see Fatigue Force probe, 329
ductile, 36 Four bar link, 222 et seq.
'fresca criteria, 36 adding stiffeners, 392
van Mises criteria, 37 leaf type, 222

425
FLEXURES

stiffness, 224 et seq. see Disc coupling

notch type, 226 see Mobility
optimal geometry, 237 see Two axis hinges
combined effects, 243 see Three axis hinges
effect of axial strains, 242
simple linear spring, 238
trapezoidal, 235 L
natural frequency, 237
virtual center, 231 Lagrange's equation, 71 et seq.
defined, 74
Frequency response, see Linear
steps in use of, 75
systems theory
Leaf type flexures
adding stiffeners, 392
G applied to couplings of rotational
symmetry, 326
Galvanometer, 369 cantilever beam, 154
Generalized coordinates, 60 axial compressive force, 158
Generalized dynamic analysis bending stiffness, 158
Generalized force, 64 combined axial and tangential loads,
159
Gravity acceleration, 4
combined axial and tangential loads
Grubler's equation, see Mobility plus a moment applied to the free
end of a simple cantilever, 167

H configurations of, 156

rigid body attached at free end, 156
single normal force, 154
Hamilton's principal, 65,66
's' shaped deflection, 168, 173
applied to spring mass system, 68
height deviation, 176
lateral vibration of a bar, 69
stiffness, 175
Hardness see also Four bar link
of steels, 414 see also Rotationally symmetric leaf
Heat treatment type hinge
see Beryllium copper Levers
see Steels cascading, see Lost motion
Hooke's Law, 15 case study, 357, see also Wire electro
Hysteresis, see Damping discharge machining
dynamic effect, 352

I perfect, 338
rigid, 337
modeling rigid lever compliance, 362
Instantaneous center, see Virtual center
see Lost motion
simple, 336
J soft spring-stiff spring, 343
attenuation, 344
Joints two arm,339

426
SUBJECf INDEX

analysis of, 340 see also Wire electro discharge

Link, see Mobility machining
Linear systems theory, 75 et seq. tolerances, effect of, 391
critical damping ratio, see Critical stiffness, 406
damping ratio Materials, 17 et seq.
definitions of, 78 brittle, see Brittle materials
eigen analysis, see Eigen analysis fatigue, see Fatigue
frequencyresponse,79 hysteresis, 19, 400
acceleration, 91 metals, 17
graphical representation, 98 non-metals, 20
Raleigh's approach, 103 resilience, 19
shortcuts, 100 toughness, 19
velocity, 91 Materials utilization, 290
general,87
Mean
multi-degree of freedom, 92 et seq.
steps in solution, 114 variance of, 409
poles, 98, 107 Measurement loops, see Loops
roots locus, 120 Metals, see Materials
spring mass damper system, 75 Micro electro mechanical systems
transmissibility, see Vibration (MEMS), 382, 413
isolation scaling, 382
zero's, 96, 103, 107 Mobility, 54
Lithographic etching, 381 Grubler's equation, 56,244
copper sheet, 383 joint, 55
Loops link, 55
definition, force, 9 planar, 55
definition, measurement, 9 three dimensional, 55
requirements for, 9 Modulus of elasticity
rules, 10 complex, 401
Lost motion, 345 definition, 23
cascade levers, 351 Modulus of rigidity, 34
generalized lever system, 347 et seq. Mohr's circle, 27 et seq.
steps in construction, 30
M three dimensional, 31
Multi-degree of freedom systems, see
Manufacture,373 Linear systems theory
see also Beryllium copper see also Generalized dynamic
see also Conventional machining analysis
see also Diamond grinding
see also Electroplating
see also Lithographic etching N
see also Steels Natural frequency

427
 damped, 86 Principal stress and strain, see Stress,
single degree of freedom, 81 see Strain
Notches
see also Fatigue
Notch hinge, 177
Q
applied to couplings of rotational Quality factor, 82, 402
symmetry,326 see also Critical damping ratio
case study, 217
circular, 180 et seq.
accuracy of estimates, 183 R
Paros Weisbord equations,
approximate, 190
Raleigh's dissipation function, 74
stiffness, 181 Raleigh's method, 60
stress concentration factor, 181 see also Beams, vibrations of
elliptic, 185 et seq. Resilience, 19
compliance in other axes, 188 resonance frequency, 86
stress concentration factor, 191
Rigidity modulus, see Modulus of
general, 177
rigidity
see also Conventional machining
Rotationally symmetric leaf type
Notch sensitivity, see Fatigue
hinge,308
assessment, 314
0 axial stiffness, 310
case study, 327
Optical lever, 369 stresses, 313

p s
Palmgren-Miner equation, see Fatigue Scaling
Planar mechanisms, 244 dynamic effect, 382
assumptions, 245 Sketching
dynamics, see Dynamics benefits,8
generalapproach,244 Solders, 379
mechanisms of mobility M = 1, 247 et
Spring, see Coil spring
seq.
Steels
Plates, vibrations of
hardness, 414
circular, 136
heat treatment, 415
governing equation, 136
clamped at the perimeter, 137 Stiffness
with a central mass, (Raleigh's as a measure of precision, 5
method), 140 effect of manufacturing tolerance,
rectangular, 145 406
simply supported, 145 Strain
other boundary conditions, 146 engineering, 16

428
SUBJECf INDEX

principal, 22 et seq. Transfer functions, see Linear systems

shear,32 theory
true, 16 Transmissibility, see Vibration
Strength reduction factor, 40 isolation
Stress Two axis hinges
biaxial, 24 comparative table, 211
elastic limit, 18 rod,211
endurance, 19 simple, 206
see also Fatigue stiffness, 207
engineering, 16 stress, 208
plane,26 toroidal, 211
principal, 22 et seq. case study, 215
non-principal, 26 et seq. stiffness, 212
see also Mohr's circle stress, 213
resilience, 19 stress concentration factor, 213
shear,32 two beam, 208
three dimensional, 31 stiffness, 209
stress, 209
triaxial, 25
true, 16 Three axis hinges, 251

u
yield, 18, 35
see also Failure
Stress concentration factor, 40
elliptic hinge, 191 Uncertainty, propagation of, 406 et
notch hinge, 181 seq.
toroidal hinge, 213 correlation, 408

v
Strings, vibrations of, see Vibrations
Structural loops, see Loops
Symmetry
benefits,8 Variance, 408
broken,9 of the mean, 409
Synthesis, see Design Variational operators
commutation, 64
Hamilton's principle, see Hamilton's
T principle
minima of a function, 64
Tin canning, 13
Vibrations
Torsion, 51 et seq.
beams, see Beams, vibrations of
of prismatic beam of circular cross
continuous systems, 120 et seq.
section, 52
Fourier analysis, 123
of prismatic beam of rectangular
plates, see Plates, vibrations of
cross section, 53
Raleigh's method, see Raleigh's
Toughness,19 method

429
FLExURFS

rod, see Beams, vibrations of

strings, 121
Vibration isolation
transmissibility, 88
choice of damping ratio, 81
Virtual center, 231
Kennedy's theorem, 233
rigid body motion, 232
see also Four bar link
Virtual work, 63
Voice coil drive, see Drives

w
Waxes, see Conventional machining
Wire electro discharge machining, 168,
380
lever, case study 368

y
Young's modulus, see Modulus of
elasticity

z
Zero stiffness, 165

430